Doctoraatsproefschrift nr. 1051 aan de faculteit Bio-ingenieurswetenschappen van de KU Leuven

RESTORATION AND SUSTAINABLE MANAGEMENT

OF FRANKINCENSE FORESTS IN ETHIOPIA: A BIO-ECONOMIC ANALYSIS

Mesfin Tilahun Gelaye

Dissertation presented in partial fulfilment of the requirements for the degree of Doctor of Bioscience Engineering

September 2012

Supervisors: Prof. E. Mathijs, KU Leuven Prof. B. Muys, KU Leuven Members of the Examination Committee: Prof. E. Smolders, KU Leuven, Chairman Prof. M. Maertens, KU Leuven Prof. M. Hermy, KU Leuven Prof. J. Deckers, KU Leuven Prof. R. Brouwer, Free University of Amsterdam

© 2012 Katholieke Universiteit Leuven, Groep Wetenschap & Technologie, Arenberg Doctoraatsschool, W. de Croylaan 6, 3001 Leuven, België Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of openbaar gemaakt worden door middel van druk, fotokopie, microfilm, elektronisch of op welke andere wijze ook zonder voorafgaandelijke schriftelijke toestemming van de uitgever.

All rights reserved. No part of the publication may be reproduced in any form by print, photoprint, microfilm, electronic

or any other means without written permission from the publisher.

ISBN: 978-90-8826-257-9

D/2012/11.109/43

Dedicated

to

My beloved wife

Bizumesh Hailu,

Our lovely children

Abel, Betelihem, Ana & Milka,

&

My parents

iv

Preface

“And when they had come into the house, they saw the young Child with Mary His mother, and

fell down and worshiped Him. And when they had opened their treasures, they presented gifts

to Him: gold, frankincense, and myrrh.” MATTHEW 2, 11.

Thanks GOD for this moment of my life and all my ways through.

This dissertation would not have been in this shape without the supervision, scientific guidance,

and encouragement of my promoters Prof. Erik Mathijs and Prof. Bart Muys. I am also very

grateful to Prof. Seppe Deckers, Prof. Liesbet Vranken, Prof. Miet Maertens, Prof. Martin

Hermy, Dr. Raf Aerts, Dr. Bruno Verbist, and Dr. Wim Aertsen for their invaluable scientific

comments. I am very fortunate to learn from their diverse scientific expertise. I would like to

extend my word of gratitude to the members of the examination committee: Prof. Erik Smolders,

Prof. Martin Hermy, Prof. Miet Maertens, Prof. Seppe Deckers, and Prof. Roy Brouwer for

proofreading the manuscript and providing their valuable comments.

I would like to thank the whole group of colleagues in the division of Bioeconomics as well as

the Forest Ecology and Management Research group at KU Leuven. My gratitude goes to

Abebe Ejigu, Kidanemariam Gebregziabeher, Jorge Alberto Cusiscaniqui Giles, Jeremy Valck,

Monica Schuster, Isabel Lambrecht, Ellen Verhofstadt, Basil Mugonola, Alice Nakiyemba, Pieter

Vlaeminck, Pieter Van Turnhout, Anneleen Kenis, Dr. Koen Dillen, Dr. Wouter Achten, Kitessa

Hundera, Aklilu Nigussie, Kidane Gidey and all other members. I would like to extend my

sincere gratitude to Rik Deliever and Kris Vandezande who have supported and trained me a lot

while I was doing the soil chemical analysis. I would also like to thank my colleague

Gebreyohannes for his advice and support in the course of analyzing my soil samples.

I gratefully acknowledge the financial support of the Interfaculty Council for Development

Cooperation (IRO) Scholarship Program of the KU Leuven, the International Foundation for

Science (IFS) and the VLIR-Mekelle University Institutional University Cooperation (VLIR-MU-

IUC) project.

I would also like to thank the local authorities of the study area for their support in facilitating the

survey. During the course of the field research, many individuals were involved and I would like

to extend my gratitude to all the VLIR- MU-IUC office administrative staff members and car

drivers without whom the data collection would not have been possible. My special thanks go to

ii

Nahusenay Teamer, Mulugeta Hagos, Selamawit Girmay, Eleni Girmay, Samison Gerbemeskel,

Kahsu Kiros, Haile G/Giorgis, and Ato Berhe Hadush.

I got a lot of encouragement from my friends and colleagues with whom I have been working at

Mekelle University. I would like to thank you all. My special thanks go to Solomon Geleta, Dr.

Bedru Babulo, Dr. Fredu Nega, Dr. Zaid Negash, Dr. Sintayehu Fisseha, Dr. Assefa Abegaz, Dr.

Woldegebriel Abreham, Prof. Mitiku Haile, Dr. Kindeya Gebrehiwot, Dr. Kassa Amare, Dr. Hans

Bauer, Dr. Gebrehawaria Gebreegziabeher, Abebe Damitew, Tekalign Simeneh, Mulugeta

Sibhatleab, Agazi Hailay, Alem Araya, Muluwork Kidanemariam, Assefa Werede, Abreha

Tesfay, Fisseha Abadi, Aregawi Gebremichael, Hailemichael Tesfay, Teklay Tesfay, Haftom

Bayray and colleagues of Mekelle University.

Of course, during the course of these four years as well as from the beginning of my

undergraduate study, my parents and my parents-in-law supported and encouraged me with

their love and by taking care of not only me but also my family during the course of my studies.

My special thanks go to my father Tilahun Gelaye, my mother Sosina Bezabih, my father in-law

Hailu Gashe, my mother in-law Kefene Debela, my uncle Goshu Bezabih and his family, my

aunt Amarech Bezabih and my uncle Mesfin Workineh, my sister Almaz and her family, my

brothers Sahelu Tilahun and his family, Birhanu Tilahun, Ketema Tilahun and his family, Dereje

Mekonnen, Yonas Tefera, Merid Tefera, Dereje Hailu, Getachew Hailu, Shimeles Hailu,

Abreham Mesfin, Endalkachew Hailu, and my sisters Senait Mekonnen, Tigist Hailu, Zinash

Tefera, Wosene Tefera, Rahel Tefera, and Misrak Mesfin.

I would like to take this opportunity to extend my deepest gratitude to my late grandmother

Alemush Moges, my late aunt Fanosie Bezabih and my late uncle Tefera Tesema who were

always supporting me and saw the seed of the importance of education in my childhood. I

always remember their wise advices, love, and kindness. I Pray to God to Rest Their Souls in

Peace!

Finally, my special thanks go to my beloved wife, Bizunesh Hailu, our lovely son Abel Mesfin

and lovely daughters Bethelihem Mesfin, Ana Mesfin, and Milka Mesfin. My wife and our kids

are really my strength. Specially, Buzye has been encouraging me besides the entire burden on

her in taking care of our kids. I am blessed in you and I really love you all forever!

Mesfin Tilahun

Leuven, September 2012.

iii

Summary Boswellia papyrifera (Del.) Hochst is a multipurpose deciduous tree species with high economic,

cultural and environmental values. Frankincense from this tree species is a traded commodity

used in the pharmaceutical, food, cosmetic and chemical industries, for clerical services in

different religions, and as a fragrance during coffee ceremonies in Ethiopia. However, the

resource has been declining due to unsustainable management, which includes shifting to crop

cultivation, free grazing and indiscriminate cutting of leaves for livestock feeding, and

overtapping for frankincense. This study aims at: a) assessing the effects of leaf lopping for

fodder, tapping for frankincense, and free grazing on the biophysical state of Boswellia

papyrifera forests of Ethiopia, b) assessing the stocks of biomass and soil organic carbon in

Boswellia papyrifera forests, c) evaluating the trade-offs between conservation, production

forestry, and shifting to crop cultivation, d) assessing rural households’ demand for conserving

the forest, and e) identifying the role of frankincense forests on rural livelihood and poverty

reduction.

The dissertation is a multidisciplinary piece of work based mainly on primary data from plot level

experiments and a household survey conducted in five villages with Boswellia papyrifera forests

in Tigray, northern Ethiopia. For analysing these biophysical and socio-economic data, a

number of statistical and econometric models are applied, which include: multilevel linear mixed

model, the standard allometric model, the model of environmental cost benefit analysis, the

double-bound dichotomous contingent valuation method, and standard impact assessment

parametric and non-parametric econometric models.

We found that leaf cutting caused significant declines in frankincense yield, and production of

inflorescence and fruits. Tapping showed a significant positive impact on frankincense yield, but

there was no evidence of a significant difference between tapped and untapped trees in terms

of flowering and fruiting. Some environmental variables like altitude, soil depth and nutrient

content also significantly affect tree productivity. Interestingly frankincense yield, flower and fruit

production significantly differ in relation to the bark colour that would be used as an indicator of

tree fitness.

The allometric model AGB = 0.061(DBH) 2.353 predicts the above ground biomass carbon in

Boswellia papyrifera forest with an average bias of less than 2%. The stored carbon in Boswellia

papyrifera forests was about 44 Mg ha-1 of which nearly 78% was accounted by soil organic

carbon and fenced plots had more concentrations of carbon and nutrients in the soil than

unfenced plots. Given the economic and ecological importance of this species and the limited

data on biomass and carbon stock especially in dry forests of Africa, the findings are useful for

iv

biomass and carbon stock estimation, and for assessing the implications of land use change on

carbon emissions. Moreover, it is also important for validating existing generalized allometric

models, which mostly lack evidence from Africa’s dry forest.

The cost benefit analysis assesses the economic benefits and costs of six options of

frankincense forest management that range from conservation in the form of exclosure to the

business as usual scenario that involves free grazing, leaf lopping and intensive frankincense

tapping. The Net Present Values of almost all the forestry options are negative if the benefits

from carbon and nutrient storage services are excluded indicating that direct benefits from

frankincense forests are less than the benefits from the competing land use. Moreover, as

shifting cultivation is the competing land use in the study area, estimates of the opportunity

costs of reducing emissions from deforestation and forest degradation (REDD) specific to the

forest were also provided. Accordingly, pure conservation of the forest could result in emission

reduction of about 142 tons of CO2 per hectare at an opportunity cost of about 33 USD per ton

of CO2 emission reduction. About 80% of this opportunity cost is incurred by rural people in the

form of forgone net benefits from not converting the forest to cropland through shifting

cultivation.

The contingent valuation study aims at assessing local evidence on whether any conservation

intervention will be welcomed by rural people in the Boswellia forest areas. Accordingly, the

study assessed the rural households’ willingness to pay and willingness to contribute labour for

Boswellia papyrifera forest conservation. We found that next to the bid level, willingness to pay

is influenced most by income and education, and willingness to contribute labour is significantly

affected by size of family labour and gender of the household head. A household is willing to

pay at least about 5 USD per year or contribute almost one week of free labour per year, which

amounts to close to 7 USD valued at per capita daily income of the households. The potential

local demand for conservation of Boswellia papyrifera forest could be mobilized effectively with

complementary policy interventions aimed at sustainable use and poverty reduction.

Recently there is a considerable debate on the role of non-timber forest products on poverty

alleviation. The study on the impact of frankincense membership on rural income and poverty

contributes to the empirical literature on the role of organized access to a traded non-timber

forest product on rural livelihood and poverty reduction. In the past, rural households in northern

Ethiopia had no access to frankincense production and trading. With recent developments in the

region, rural households are getting access to the harvesting of frankincense through organizing

cooperatives. We analysed the income and poverty effects of rural frankincense cooperative

firms and the results indicate that both membership as a binary variable and amount of

v

households’ investment on shares in the cooperative firms have significant positive welfare

impacts in terms of increasing household income and reducing rural poverty.

The negative impact of leaf looping on the tree’s capacity to produce frankincense, fruits and

flowers as well as the negative net present values of all the forestry alternative management

options indicate that the forest is in a very high risk of continuous degradation and perpetuation

of what is called “a tragedy of the commons”. Thus, conservation and sustainable management

practices should be in place to increase the resources’ competitiveness and hence avoid its

degradation. This could be possible through interventions that can make developments in

alternative livestock feed productions, through creating favorable conditions for the development

of businesses that can add value to the frankincense, which is now exported as a raw material,

and with the introduction of the other ecosystem service benefits of the resource like

development of bee keeping and ecotourism. Moreover, the participation of rural communities in

the conservation and sharing of benefits from the resource are very crucial in future

conservation and management interventions as it has been confirmed from the results of the

contingent valuation on rural households demand for conservation as well as the results from

the impact evaluation. However, further research is required on assessing the resource base at

a wider spatial scale, on whether bark color of the tree is a genetic or phenotypic phenomenon

affecting tree productivity, and on the problems associated with legal and institutional

frameworks in the use and management of Boswellia papyrifera forests in Ethiopia.

vi

vii

Samenvatting

Boswellia papyrifera (Del) Hochst is een multifunctionele loofboomsoort met een hoge

economische, culturele en ecologische waarde. Wierook van deze boomsoort wordt als

grondstof verhandeld en gebruikt in de farmaceutische, voedings-, cosmetische- en chemische

industrie, in verschillende religies en als geurstof in koffieceremonies in Ethiopië. Echter, deze

hulpbron neemt af door onduurzaam beheer, zoals de teelt van gewassen, vrij grazen van

dieren en willekeurig afsnijden van de bladeren voor veevoeder en overtapping voor wierook.

Deze studie heeft als doel: a) de effecten te beoordelen van het afsnijden van de bladeren voor

voedergewassen, het tappen voor wierook en het vrij grazen van dieren op de biofysische

toestand van de Boswellia papyrifera bossen van Ethiopië, b) de voorraden aan biomassa en

bodem-organische koolstof in Boswellia papyrifera bossen te beoordelen, c) de trade-offs

tussen de conservering, productieve bosbouw en de verschuiving naar de teelt van gewassen

te evalueren, d) de vraag naar bosbehoud van huishoudens op het platteland in te schatten, en

e) de rol die wierookbossen kunnen hebben op het levensonderhoud van plattelandsbewoners

en armoedebestrijding te identificeren.

Het proefschrift is een multidisciplinair werk dat voornamelijk gebaseerd is op primaire

gegevens van experimenten op perceelsniveau en een enquête bij gezinnen uitgevoerd in vijf

dorpen met Boswellia papyrifera bossen in Tigray, Noord-Ethiopië. Voor de analyse van deze

biofysische en socio-economische gegevens worden een aantal statistische en econometrische

modellen toegepast, waaronder: multilevel lineaire gemengde modellen, het standaard

allometrisch model, de ecologische kosten-batenanalyse, de double-bound dichotome

contingent valuation methode, en standaard impactevaluatie met behulp van parametrische en

niet-parametrische econometrische modellen.

We vonden dat het afsnijden van bladeren aanzienlijke dalingen in de wierookopbrengst, de

bloemen en het fruit veroorzaakt. Tappen toonde een significante positieve invloed op de

wierookopbrengst, maar er was geen bewijs van een significant verschil tussen de getapte en

niet-getapte bomen op het vlak van bloei en vruchtvorming. Sommige omgevingsvariabelen als

hoogte en de diepte en nutriënteninhoud van de bodem hebben ook een significante invloed op

de boomproductiviteit. Interessant is dat wierookopbrengst en de productie van bloemen en fruit

beduidend verschillen naargelang de kleur van de schors die kan worden gebruikt als een

indicator van de gezondheid van de boom.

Het allometrische model AGB = 0.061(DBH)2.353 voorspelt de bovengrondse biomassa koolstof

in Boswellia papyrifera bos met een gemiddelde afwijking van minder dan 2%. De opgeslagen

koolstof in Boswellia papyrifera bossen was ongeveer 44 Mg ha-1 waarvan bijna 78% werd aan

viii

organische koolstof in de bodem en omheinde percelen kenden een hogere concentratie van

koolstof en voedingsstoffen in de bodem dan niet omheinde percelen. Gezien het economisch

en ecologisch belang van deze soort en de beperkte gegevens over biomassa en

koolstofvoorraden vooral in de droge bossen van Afrika, zijn deze bevindingen nuttig voor de

schatting van biomassa en koolstofvoorraden, en voor de beoordeling van de gevolgen van

veranderingen in landgebruik op de uitstoot van koolstof. Bovendien is het ook van belang voor

het valideren van bestaande gegeneraliseerde allometrische modellen, die meestal

onvoldoende gegevens bevatten uit de droge bossen van Afrika.

De kosten-batenanalyse evalueert de economische kosten en baten van zes beheeropties in

wierookbossen gaande van conservering in de vorm van ‘exclosure’ tot de business-as-usual-

scenario, waarbij dieren vrij grazen, bladsnoei en intensief tappen van wierook. De netto huidige

waarde van bijna alle de opties zijn negatief als de voordelen van de opslag van koolstof en

voedingsstoffen niet worden meegerekend, wat aangeeft dat de directe voordelen van

wierookbossen kleiner zijn dan de voordelen van het concurrerende landgebruik. Bovendien,

omdat zwerflandbouw het concurrerende landgebruik in het studiegebied is, werden ook

schattingen van de opportuniteitskosten van de vermindering van emissies door ontbossing en

de degradatie van bossen (REDD) specifiek voor bossen gemaakt. Bijgevolg kunnen de pure

instandhouding van het bos kan leiden tot een emissiereductie van ongeveer 142 ton CO2 per

hectare aan een opportuniteitskost van ongeveer 33 USD per ton CO2-emissiereductie.

Ongeveer 80% van deze opportuniteitskost wordt gemaakt door mensen op het platteland in de

vorm van gederfde netto voordelen van het niet omzetten van het bos tot bouwland door

zwerflandbouw.

Het contingent valuation onderzoek is gericht op het gebruiken van lokale gegevens om te

beoordelen of instandhoudingsinterventies verwelkomd zullen worden door mensen op het

platteland in de Boswellia bosgebieden. Daarom evalueerde de studie de bereidheid om te

betalen en de bereidheid om arbeid bij te dragen door plattelandsbewoners voor het behoud

Boswellia papyrifera bossen. We vonden dat naast het biedniveau, de bereidheid om te betalen

het meest beïnvloed wordt door inkomen en opleiding, en de bereidheid om arbeid bij te dragen

sterk beïnvloed wordt door de grootte van het gezin en het geslacht van het gezinshoofd. Een

huishouden is bereid om op zijn minst ongeveer 5 USD per jaar te betalen of bijna een week

arbeid per jaar gratis te leveren, wat overeenkomst met bijna 7 dollar gewaardeerd aan het per

capita dagelijks gezinsinkomen. De potentiële lokale vraag naar behoud van Boswellia

papyrifera bossen kan doeltreffender worden gemobiliseerd aan de hand van aanvullende

beleidsinterventies gericht op duurzaam bosgebruik en op armoedebestrijding.

ix

Recent is er een aanzienlijk debat over de rol van niet-hout bosproducten op

armoedebestrijding. De studie over de impact van lidmaatschap van wierookcoöperaties op

inkomen en armoede draagt bij tot de empirische literatuur over de rol van georganiseerde

toegang tot verhandelde niet-hout bosproducten op de toestand van boeren en

armoedebestrijding. In het verleden hadden landelijke gezinnen in het noorden van Ethiopië

geen toegang tot wierookproductie en handel. Met de recente ontwikkelingen in de regio krijgen

gezinnen op het platteland toegang tot het oogsten van wierook door het opzetten van

coöperaties. We analyseerden het inkomen en de armoede-effecten van rurale

wierookcoöperaties en de resultaten geven aan dat zowel lidmaatschap (gemeten als een

binaire variabele) en de hoeveelheid die gezinnen investeren in aandelen van de coöperatieve

bedrijven belangrijke positieve welvaartseffecten hebben in termen van toenemend

gezinsinkomen en het verminderen van rurale armoede.

Zowel de negatieve impact van bladsnoei op de capaciteit van de boom om wierook, fruit en

bloemen te produceren, als de negatieve netto huidige waarde van alle alternatieve opties van

bosbeheer geven aan dat de bossen zich in een zeer hoog risico van voortdurende degradatie

en van bestendiging van wat de ‘tragedy of the commons’ wordt genoemd. Daarom moeten de

conservering en duurzame beheerspraktijken worden geïmplementeerd om het

concurrentievermogen van de hulpbron te vergroten en daarmee verdere degradatie

voorkomen. Dit zou mogelijk zijn alternatieve veevoederproductie te stimuleren, door middel van

het creëren van gunstige voorwaarden voor de ontwikkeling van bedrijven die waarde kunnen

toevoegen aan de wierook, die nu wordt geëxporteerd als grondstof, en met de introductie van

andere ecosysteemdienstbaten, zoals de ontwikkeling van de bijenteelt en ecotoerisme.

Bovendien is de deelname van landelijke gemeenschappen, zowel in het behoud als in de

verdeling van de baten van de hulpbron, zeer belangrijk voor toekomstige instandhouding- en

beheersinterventies, zoals bevestigd werd door de resultaten van de contingent valuation met

betrekking tot de vraag van rurale gezinnen naar bosbehoud, alsmede de resultaten van de

impactevaluatie. Er is echter verder onderzoek nodig naar de beoordeling van de hulpbronnen

op een bredere ruimtelijke schaal, naar de vraag of de kleur van de schors van de boom een

genetische of fenotypisch verschijnsel is dat boomproductiviteit beïnvloedt, en ook naar de

problemen die verband houden met de juridische en institutionele kaders in het gebruik en

beheer van de Boswellia papyrifera bossen in Ethiopië.

x

xi

Table of Contents

Preface................... ................................................................................................................ i

Summary................ ............................................................................................................. iii

Samenvatting ..................................................................................................................... vii

List of Figures .................................................................................................................... xv

List of Tables ................................................................................................................... xvii

List of Acronyms/Abbreviations ...................................................................................... xix

CHAPTER 1. General introduction ..................................................................................... 1

1.1. Background of the study .......................................................................................... 1

1.1.1. Frankincense, natural gums and resins export ................................................. 2

1.1.2. Boswellia papyrifera forest degradation ............................................................ 5

1.2. Research questions ................................................................................................. 7

1.3. Aim and objectives of the study ................................................................................ 8

1.4. Study area description ........................................................................................... 10

1.5. Research methodologies ....................................................................................... 12

1.6. Outline of the thesis ............................................................................................... 13

CHAPTER 2. Effects of leaf lopping for fodder and resin tapping from Boswellia papyrifera trees on frankincense yield, flowering and fruiting ................. 17

2.1. Introduction ............................................................................................................... 17

2.2. Materials and methods ........................................................................................... 19

2.2.1. Experimental plot design and soil data collection ............................................ 19

2.2.2. Sample tree selection, leaf lopping and tapping experiments ......................... 21

2.2.3. Data analysis .................................................................................................. 24

2.3. Results ................................................................................................................... 25

2.3.1. Characteristics of Boswellia papyrifera populations ........................................ 25

2.3.2. Environmental and dendrometric characteristics of sample trees ................... 26

2.3.3. Variables affecting frankincense yield ............................................................. 31

2.3.4. Variables affecting inflorescence and fruit production ..................................... 33

2.4. Discussion ............................................................................................................. 35

2.5. Conclusions ........................................................................................................... 38

CHAPTER 3. Biomass and soil organic carbon stocks in Boswellia papyrifera (Del.) Hochst forests .............................................................................................. 39

3.1. Introduction ............................................................................................................ 39

3.2. Materials and methods ........................................................................................... 41

3.2.1. Tree biomass carbon data collection and analysis .......................................... 41

3.2.2. Herbaceous biomass carbon .......................................................................... 43

3.2.3. Soil carbon and nutrients ................................................................................ 44

3.3. Results ................................................................................................................... 45

xii

3.3.1. Description of biomass input data of harvested trees ...................................... 45

3.3.2. Stand structure ............................................................................................... 46

3.3.3. Soil carbon and nutrient concentrations .......................................................... 46

3.3.4. Allometric relationships ................................................................................... 47

3.3.5. Carbon stocks ................................................................................................ 49

3.3.6. Effect of fencing on stocks of soil organic carbon and nutrients ...................... 50

3.3.7. Comparison of allometric model of Boswellia papyrifera with mixed species models .................................................................................................................... 52

3.4. Discussion ............................................................................................................. 54

3.4.1. Inclusion of dendrometric variables on goodness of fit of the allometric model ...................................................................................................................... 54

3.4.2. Carbon stock in Boswellia forests and implications to land use change .......... 55

3.5. Conclusions ........................................................................................................... 56

CHAPTER 4. Valuation of ecosystem services: a cost benefit analysis of forest management options and REDD+ opportunity costs specific to frankincense forests .................................................................................... 57

4.1. Introduction ............................................................................................................ 57

4.2. Cost benefit analysis (CBA): conceptual framework ............................................... 58

4.3. Materials and methods ........................................................................................... 59

4.3.1. Selecting forest management options ............................................................. 59

4.3.2. Determining physical quantities of the ecosystem services ............................. 60

4.3.3. Determining opportunity cost: crops from shifting cultivation ........................... 62

4.3.4. Valuation of benefits and costs ....................................................................... 63

4.3.5. Data analysis and decision criteria ................................................................. 64

4.4. Results ................................................................................................................... 67

4.4.1. Estimated quantities of Boswellia papyrifera forest ecosystem services ......... 67

4.4.2. Base case NPV of forest management options ............................................... 72

4.4.3. Opportunity cost of REDD+ specific to Bosswellia papyrifera forest ................ 74

4.4.4. Distributional effects ....................................................................................... 75

4.4.5. Multicriteria analysis ....................................................................................... 78

4.4.6. Sensitivity analysis ......................................................................................... 80

4.5. Discussion and conclusions ................................................................................... 84

CHAPTER 5. Rural households’ demand for frankincense forest conservation: a contingent valuation analysis ..................................................................... 87

5.1. Introduction ............................................................................................................ 87

5.2. Value to be estimated and the contingent valuation method .................................. 89

5.3. Materials and methods ........................................................................................... 89

5.3.1. Survey design and data collection .................................................................. 89

5.3.2. Model specification for measuring WTP and WTCL ........................................ 95

5.3.3. Data calibration .............................................................................................. 98

xiii

5.4. Results ................................................................................................................... 99

5.4.1. Households’ knowledge and attitude .............................................................. 99

5.4.2. Parameter estimates of WTP and WTCL ........................................................ 99

5.4.3. Robustness tests .......................................................................................... 103

5.5. Discussion ........................................................................................................... 104

5.6. Conclusions ......................................................................................................... 107

CHAPTER 6. Impact of membership in rural frankincense cooperatives on rural income and poverty ................................................................................... 109

6.1. Introduction .......................................................................................................... 109

6.2. Materials and methods ......................................................................................... 111

6.2.1. The data ....................................................................................................... 111

6.2.2. Frankincense cooperative membership, rural income and poverty: descriptive analysis ................................................................................................................. 112

6.2.3. Econometric models ..................................................................................... 115

6.3. Results ................................................................................................................. 118

6.3.1. Determinants of membership in frankincense cooperative firms ................... 118

6.3.2. Income effects of membership in frankincense cooperative firms ................. 120

6.4. Discussion and conclusions ................................................................................. 122

CHAPTER 7. Conclusions and recommendations ........................................................ 125

7.1. General conclusions ............................................................................................ 125

7.2. Recommendations ............................................................................................... 129

7.3. Limitations and implications for further research .................................................. 131

References............. .......................................................................................................... 133

Appendices...... ................................................................................................................ 147

xiv

xv

List of Figures

Figure 1.1: Frankincense by quality grades based on purity, size and color of the resin

granules .................................................................................................................... 3

Figure 1.2: Aggregated volume and value of exports of frankincense, natural gums and resins

from Ethiopia over the period 1979-2007 ................................................................... 4

Figure 1.3: Shifting cultivation, free grazing, and leaf lopping for fodder, and tapping for

frankincense in Central and Western Boswellia forests of Tigray ............................... 7

Figure 1.4: Location of the study area ....................................................................................... 12

Figure 1.5: Schematic presentation of the thesis outline and conceptual framework ................. 15

Figure 2.1: Frankincense tapping season (October to beginning of June) and phenological

periods for Boswellia papyrifera trees in Western and Central Tigray, Ethiopia ....... 23

Figure 2.2: Scatter plots of 420 trees within 140 plots sampled from five Boswellia papyrifera

forest populations in northern Ethiopia .................................................................... 29

Figure 2.3: Frankincense yield (a and b), inflorescence (c and d) and fruits (e and f) per

inflorescence as a function of diameter at breast height (DBH) ............................... 37

Figure 3.1: Ground floor herbaceous and grass biomass in freely grazed (left picture) and

fenced stands of Boswellia papyrifera forest in the K.Humera site of Western

Tigray ...................................................................................................................... 44

Figure 3.2: Diameter distribution of Boswellia papyrifera trees in the study area ....................... 46

Figure 4.1: Present values of opportunity costs ($ per tCO2) of maintaining Boswellia papyrifera

forest under different forest management options for a period of 30 years at real

discount rate of 5.99% ............................................................................................. 75

Figure 4.2: Supply chain of frankincense in Ethiopia ................................................................. 76

Figure 4.3: Distribution of present values of benefits and costs (%) of maintaining a hectare of

Boswellia forest under six management options to actors in the frankincense supply

chain ....................................................................................................................... 78

Figure 4.4: Sensitivity of NPV of conservation of Boswellia forest to (±40%) changes in

parameters .............................................................................................................. 81

Figure 5.1: Boswellia papyrifera stands in different state of degradation in Western and Central

Tigray ...................................................................................................................... 91

Figure 5.2: Distribution of willingness to pay in cash and willingness to contribute labour data

from the pre-test survey ........................................................................................... 94

Figure 5.3: Face-to-face contingent valuation survey (March 2010) .......................................... 94

xvi

xvii

List of Tables

Table 2.1: Description of the five Boswellia papyrifera populations and size and distribution of

sample plots ............................................................................................................. 20

Table 2.2: Mean values (SE) for environmental, dendrometeric, and response (flowers, fruits

and frankincense yield) variables of five Bosweliia papyrifera populations in northern

Ethiopia and the significance of differences between the populations ...................... 26

Table 2.3: Mean values (SE) for environmental, dendrometric, and response (yield, flowers and

fruits) variables in ten groups (Leaf lopping X tapping as factors) of 140 plots within

five Bosweliia papyrifera populations in northern Ethiopia (per group N=14) ............ 30

Table 2.4:Type III tests of fixed effects of three factors ( leaf lopping and tapping), plot level

environmental variables, and dendrometric variables (condensed in two PCA

dimensions) on two seasons frankincense yield (log10 transformed of yield per tree)

in 140 plots within 5 Boswellia papyrifera forest populations in northern Ethiopia ..... 32

Table 2.5:Type III tests of fixed effects of two factors ( leaf lopping and tapping), environmental

and dendrometric variables on two seasons log transformed number of inflorescence

and fruits per inflorescence of Boswellia papyrifera trees in northern Ethiopia .......... 34

Table 3.1: Mean (SE) of dendrometric, wood density and dry biomass and Mann-Whitney Test

statistics for the two-independent samples of trees from Abergelle and Kafta Humera

sites used for allometric modelling ............................................................................ 45

Table 3.2: Mean (SE) of soil properties and Mann-Whitney Test statistics for the two-

independent samples from the study sites by soil depth ........................................... 47

Table 3.3: Predictive performance of allometric equations for dry biomass of Boswellia

papyrifera tree components (leaf biomass (LB), branch biomass (BB), stem biomass

(SB), and aboveground biomass (AGB)) as a function of tree variables (diameter

(DBH), crown area (CA) height (H), wood density (ρ), product of DBH and H (DBHH),

product of DBH and CA (DBHCA), and product of DBH and ρ (DBHρ)) ................... 48

Table 3.4: Mean values (SE) and Mann-Whitney Test statistics for the two-independent samples

of carbon stock in soil, trees and herbaceous biomass in Boswelllia papyrifera forest

ecosystems of Avergelle and K.Humera sites ........................................................... 50

Table 3.5: Mean (SE) and Mann-Whitney Test statistics for comparison of soil organic carbon,

total nitrogen and available phosphorous stocks between fenced and unfenced

Boswellia stands (n1=n2=70) ................................................................................... 51

Table 3.6: Comparison of observed (real) aboveground biomass (AGB) data of the 30

destructively sampled Boswellia papyrifera trees with the predicted values of AGB for

the same trees using the model develope in this study and models published for

mixed species of tropical forests............................................................................... 53

xviii

Table 4.1: Mean (SE) values of Boswellia forest ecosystem services by management

options ..................................................................................................................... 68

Table 4.2: Market prices of outputs and inputs for frankincense production and trading ........... 70

Table 4.3: Mean areas of cultivated plots by sample farm households, quantities of outputs and

inputs and prices of crop outputs and input costs ..................................................... 71

Table 4.4: NPV of Boswellia papyrifera forest management options ......................................... 73

Table 4.5: Scores of multi-criteria analysis for Boswellia papyrifera forest management

options ..................................................................................................................... 79

Table 4.6: Sensitivity of NPVs to changes in prices of cost and benefit items and real discount

rate ........................................................................................................................... 82

Table 5.1: Bid design and number randomly assigned sample households ................................... 94

Table 5.2: Description and summary statistics of socioeconomic characteristics of sample

households of the study area (N = 473) ............................................................................. 98

Table 5.3: WTP Models: Single bound discrete choice probit (Model-I and Model-II), bivariate

probit (Model-III) and a double-bound interval data probit (Model-IV) models of

rural households’ willingness to pay in cash for Boswellia papyrifera forest

conservation ......................................................................................................................... 100

Table 5.4: WTCL models: Single bound discrete choice probit (Model-V and Model-VI), bivariate

probit (Model-VII) and a double-bound interval data probit (Model-VIII) models of rural

households’ willingness to contribute labour for Boswellia papyrifera forest

conservation. ........................................................................................................................ 101

Table 5.5: Models for checking robustness of the WTP models..................................................... 104

Table 6.1: Summary statistics and comparison of household characteristics .......................... 113

Table 6.2: Determinants of membership in the frankincense rural cooperative firms ............... 119

Table 6.3: Effects of membership and log transformed investment in frankincense cooperative

firm on log transformed rural income ...................................................................... 121

Table 6.4: Balancing properties of covariates in frankincense cooperative members and non-

member groups for kernel matching on propensity scores ...................................... 122

Table 6.5: Simulation-based sensitivity analysis for propensity score matching ...................... 122

xix

List of Acronyms/Abbreviations AGB Above Ground Biomass

AIC Akaike Information Criterion

ANOVA Analysis of Variance

ATE Average Treatment Effect

BB Branches Biomass

BD Bulk Density

CA Crown Area

CBA Cost Benefit Analysis

cdf Cumulative Density Function

CDM Clean Development Mechanism

CONS Conservation

COP Conference of The Parties

CSA Central Statistical Authority

CV Contingent Valuation

DBDC Double Bounded Dichotomous Choice

DBH Diameter at Breast Height

EC Electric Conductivity

ECBA Environmental Cost Benefit Analysis

ETB Ethiopian Birr (National Currency of Ethiopia)

FAO Food and Agricultural Organization of the United Nations

FGRAZ Free Grazing

GDP Gross Domestic Product

GHG Greenhouse Gas

GPS Global Positioning System

H Tree Height

Ha Hectare

HBC Herbaceous Biomass Carbon

HICES Household Income, Consumption and Expenditure Survey

IETA International Emissions Trading Association

IPCC Intergovernmental Panel on Climate Change

IV Instrumental Variable

Kg Kilo gram

LB Leaf Biomass

lCER Longterm Certified Emission Reduction

LNT Leaf lopping with no tapping

LTAF12 Leaf lopping with tapping at 12 spots after fruiting ends

xx

LTAF6 Leaf lopping with tapping at 6 spots after fruiting ends

LTFS12 Leaf lopping with tapping full season at 12 spots

LTFS6 Leaf lopping with tapping full season at 6 spots

MEA Millennium Ecosystem Assessment

Mg Mega gram (1 Mg = 1 ton = 1000 kilo gram)

MRV Monitoring Reporting and Verification

NLNT No leaf lopping with no tapping

NLTAF12 No leaf lopping with tapping at 12 spots after fruiting ends

NLTAF6 No leaf lopping with tapping at 6 spots after fruiting ends

NLTFS12 No leaf lopping with tapping full season at 12 spots

NLTFS6 No leaf lopping with tapping full season at 6 spots

NPV Net Present Value

OC Opportunity Cost

OECD Organization for Economic Cooperation and Development

OLS Ordinary Least Squares

P Phosphorous

PCA Principal Component Analysis

PV Present Value

REDD Reducing Emissions from Deforestation and Forest Degradation

RRMSE Relative Root Mean Square Error

SB Stem Biomass

SBDC Single Bounded Dichotomous Choice

SOC Soil Organic Carbon

tCER Temporary Certified Emission Reduction

TLU Tropical Livestock Unit

TN Total Nitrogen

TRAFFIC The Wildlife Trade and Monitoring Network

UNFCCC United Nations Framework Convention on Climate Change

UN-REDD United Nations Collaborative Programme on Reducing Emissions from

Deforestation and Forest Degradation in Developing Countries

USD United States Dollar

WTCL Willingness to Contribute Labour

WTP Willingness to Pay

CHAPTER 1

General introduction

1.1. Background of the study

Dry lands constitute about 41% of the earth’s surface and are home to more than one-third of

the world population (Mortimore, 2009). Africa’s dry land forests cover about 43% of the

continent and provide a number of ecosystem services to more than 235 million people (FAO,

2010a). Every day millions of people across the dry forests and woodlands countries of sub-

Saharan Africa collect non-timber forest products for their subsistence (Shackleton and Gumbo,

2010). These provisioning services of dry land forests have long formed a vital component of

people’s day-to-day livelihood needs, providing energy, food, medicines and raw materials for

building, crafts, tools and farm implements (Campbell and Luckert, 2002). Beside the

subsistence use, non-timber forest products serve as a source of cash income through trade in

local markets. Moreover, some non-timber forest products traded at the international market

contribute to the foreign exchange earnings of the exporting countries. According to Walter

(2001), there is a distinct geographical variation in the importance of non-timber forest product

categories across sub-Saharan Africa. For example, bush meat is important in most countries

across all the dry forests and woodland areas of the region, edible plants are more common in

the Sudanian and Zambezian woodlands, and honey and bees wax is widely common in the

Kalahari, Zambezian, and Somali-Masai phytoregions, and also to some extent in a few

countries in the Sudanian phytoregion. Tree exudates (frankincense (olibanum), gum Arabic,

myrrh, tannins, opopanax, and gum karaya) are important non-timber forest products in the

sudanian phytoregion and are among the most traded non-timber forest products in the

international market. The region is endowed with Boswellia Accacia and Commiphora species

that are known to produce frankincense gum arabic and myrrh, respectively.

Boswellia is among the 17 genera described in the family Burseraceae (Heywood et al., 1993).

The geographical distribution of the genus extends from Ivory Cost to the Horn of Africa and

southwards to the northern Madagascar. It is also found in the Middle East and India. More than

75% of the species are endemic to the dry lowland areas of north-eastern tropical Africa (White,

1983; Vollesen, 1989; Azene et al., 1993; Farah, 1994; FAO, 1995; Tadesse et al., 2007).

Boswellia papyrifera (Del.) Hochst is one of the 20 species in the genus. It is a deciduous tree

reaching a height of up to 13 m with thick branches tipped with clusters of leaves. The outer part

of its bark is smooth, pale yellow-brown where as a cut in the bark is red-brown and a fragrant

milky gum come out of the ducts in the bark. It has leaf and floral buds protruding on the apices

along the branches. The tree has compound leaves that contain 9-20 pinnate, veined, leaflets

supported by petioles. The flowers are white-pink on a red flower stalk of up to 35 cm long. The

2

fruit looks red capsules of about 2 cm long with three sides consisting of three hard seeds with

apical horn (Weiss, 1987; Vollesen, 1989; Bein et al., 1996; Ogbazghi, 2001).

Boswellia papyrifera (Del.) Hochst is widely distributed in Ethiopia, Eritrea, Sudan, Nigeria,

Central African Republic, Cameron, Chad, and northeast Uganda (Vollesen, 1989). In Ethiopia,

Boswellia papyrifera forms relatively pure stands and is found in the arid lowland areas of the

country, namely in Tigray, Gondar, Gojam and Shewa providences. It occurs as dominant on

steep rocky slopes, lava flows or sandy valleys, within 810-1800 m.a.s.l altitudes, temperature

of 20-25 ºC, and annual rainfall less than 900mm (Vollesen, 1998; Azene et al., 1993; Fitchl and

Admasu, 1994; Tadesse et al., 2007; Mengistu et al., 2012). There is no up-to-date national

level accurate and comprehensive information on land area covered by oleo-gum resin trees.

However, existing data indicate that the country has an estimated 2.9 million ha of land covered

by Oleo-gum resin bearing tree species that belong to the genus Boswellia, Acacia and

Commiphora with a natural gum production potential of over 300,000 metric tons (Girmay,

2000). Of the total land covered with oleo-gum resin bearing tree species in the country, nearly

33% (940,000 ha) is found in Tigray and consists of genera like Boswellia, Sterculia,

Commiphora and Acacia. Boswellia papyrifera is the main gum resin producing tree species in

the country covering more than 1.5 million hectares of land (Girmay, 2001; Tadesse et al.,

2007).

The tree has high ecological importance for combating desertification for it is capable of growing

on shallow or very stony soils in arid areas of the tropics (Stiles, 1988). The frankincense from

this tree is used as input in the pharmaceutical industries (Michie and Cooper, 1991; Schillaci et

al., 2008), in the food, perfume and cosmetics industries (Tucker, 1986), and as a traditional

medicine (FAO, 1995). It is widely used for rituals in different religions (FAO, 1995) and as a

fragrance during coffee ceremonies in Ethiopia. The leaves of the tree have a high nutritive

value for livestock feed (Melaku et al., 2010). Moreover, it is an income source to rural people

mostly as wage income from seasonal employment in the collection of the product, and the

country gets foreign currency from export.

1.1.1. Frankincense, natural gums and resins export

Frankincense, natural gums and resins from dry land tree species have been in use since 1700

B.C. at the latest (Howes, 1950). Ethiopia and Sudan are the major exporters of these natural

products to the world market. Ethiopia exports these products almost as a raw material and the

only processing activities taking place before exporting the product is to manually grade the

dried raw resin collected from the forest into different quality grades based on purity, size and

colour of the resin granules. Urban poor women employed in the trading firms carry out the

grading process. According to the Ethiopian Natural Gum Marketing Enterprise and other

3

frankincense trading firms in the country, there are seven grades for frankincense of which the

first six are for export market and only the last grade, which contains mostly bark of the tree and

small amounts of powder like granules of the resin (Figure 1.1).

Figure 1.1: Frankincense by quality grades based on purity, size and color of the resin granules. Grezo is

the dried resin as collected from the forest; G1 to G5 refer to Grades 1 to 5. Grades 1 and 4 have sub

categories based on the color of the frankincense; G4-S stands for Grade 4 Special. Grade 5 contains

mostly bark of the tree with small amount of fine resin granules

For about three decades (1979 to 2007) in the past the export volume as well as the foreign

exchange earnings from these natural products exhibited cyclical fluctuations (Figure 1.2).

Although rigorous time series econometric analysis is required to figure out the factors that have

been contributing to such fluctuations, the descriptive analysis in Figure 1.2 below suggests that

the major shocks (both policy change and political instability) that the country has experienced

over these period could be mentioned as the main factors. Prior to 1991 Ethiopia’s production

and export of frankincense, natural gums and resins was under the monopoly of a single state

company. During this period, the average annual volume of export was only 764 tones with an

average value of 1.5 million USD. Except for the years 1982 and 1983, the export volume in

each year over the period prior to 1991 was less than 1000 tons. During this period, the country

was under a military government and was in a civil war, which was mainly taking place in the

northern part of the country that is the major source of the product. The relatively higher

volumes of export in 1982 and 1983 may be due to exporting of the product in stock that had

been harvested in previous years. In actual practice, a harvest of a particular year at least

requires a considerable amount of time for drying it and grading the product for export.

The lowest export volume (56 tons) was in 1992, and this was the year after the fall of the

military government. The transitional government’s main priority during the initial years was

creating peace and stability in the war torn economy. Following the change from central

4

planning to open market economic policy, few private companies entered the frankincense,

natural gums and resin production and export sector and started to compete with the state

company. The other policy change relevant to the export sector is the adoption of market based

foreign exchange policy that started with the devaluation of the country’s currency from the fixed

rate of ETB 2.07 per USD that stayed for 17 years during previous government to ETB 5 per

USD in 1991. As a result, the export volume and the foreign currency earnings from the sector

increased from 56 tons (0.075 million USD) in 1992 to 2085 tons (2.975 million USD) in 1997.

After 1997, both the volume of export and revenue was increasing except for the years 1998

and 1999, during which the country was again in war with Eritrea and the major producing

region of frankincense was a war zone during these years. However, after 1999 the export

showed almost a continuous rise and reached the maximum of 4534 tons (6.55 million USD) in

2007. The volume of export in 2007 alone was equivalent to 49.43% of the total exports made in

12 years during the previous government.

Figure 1.2: Aggregated volume and value of exports of frankincense, natural gums and resins from

Ethiopia over the period 1979-2007. Sources: Central Statistical Authority (CSA) of Ethiopia and

Ethiopian Customs Authority

Besides the entry of private firms in the production and export business of frankincense and

natural gums and resins, developments in creating rural frankincense cooperatives in the

northern Ethiopia since 2002 have contributed in increasing the supply of particularly

frankincense to the export market. According to data from Tigray Bureau of Agriculture and

Natural Resources, there have been 89 registered firms engaged in frankincense production

and trading in the region. Of these firms, three companies are exporters, and the other 86 firms

are engaged in either only production or both production and trading in the domestic market. In

terms of type of business organization, 36 firms are partnerships, 25 are rural cooperatives, 18

5

are private limited companies, 3 are share companies, and one firm is a state owned company.

All the rural cooperative firms are engaged only in the production (harvesting) activity and sell

their harvests to the exporting and/or processing companies.

1.1.2. Boswellia papyrifera forest degradation

The General Assebmly of the United Nations adopted in December 2007 the most widely,

intergovernmentaly agreed definition of sustainable forest management. According to this

definition, sustainable forest management aims to mainatain and enance the economic, soicial

and environmental value of all types of forests, for the benefit of present and future generations

(UN, 2008). The definition comprises seven elements, which include: a) extent of forest

resource: expresses the desire to have significant forest cover and stocking to support the

social, economic, and environmental dimensions of forestry; and ambitions to reduce

deforestation, restore and rehabilitate degraded forest landscapes; b) forest biodiversity:

conservation and management of biological diverity at ecosystem, species, and genetic level to

ensure the diversity of life is maintained, and provide opportunities to develop new products, for

example medicines, in the future; c) forest health and vitality: forests need to be managed so

that risks and impacts of unwanted disturbances are minimized due to for example wildfires,

pests, diseases, invesive species etc.; d) productive functions of forest resources: the

ambition to maintain a high and valuable supply of wood and non-wood forest products, while at

the same time ensuring management options for future generations; e) protective functions

which refers to the ecological functions (regulation of soil, hydrological and aquatic systems) of

forests and trees outside forests; f) socio-economic functions of forest resources: adresses

the role of forests to the ooverall economy, their contributions to protect areas with high cultural,

spritual or recrational values; and g) legal, policy and institutional framewerk required for

achieving the above six elements.

Based on the strict sense of the above defintion of sustainable forest management, there is a

strong evidence for the lack of sustainable frankincens forest management in Ethiopia. Despite

the economic, social, ecological, and cultural importance of Boswellia papyrifera, there has

been a continuous degradation of the resource in the country and the region of Eastern Africa at

large. Studies reported widespread decline in the distribution and abundance of the species as

well as lack of regeneration and sapling recruitment (Ogabazghi 2001; Gebrehiwot 2003;

Groenendijk et al., 2012) as serious problems of natural regeneration. For example, in Tigray

region more than 177,000 ha of Boswellia forest is reported to be destroyed in a period of about

20 years between the late 1970s and late 1990s (Gebrehiwot et al., 2002). Moreover, although

seeds germinate during the wet season, it is extremely rare to find saplings and even small

trees with less than 5 cm in diameter in most Boswellia forest areas in Ethiopia. Nigussie et al.,

(2008) reported almost a 100% seedling mortality in dry seasons in central Tigray and more

6

than 76% of the trees in western Tigray were greater than 30 cm diameter at breast height

(DBH) (Gebrehiwot et al., 2002). In Eriterea Ogbazghi et al. (2006) also reported that Boswellia

papyrifera stands were predominantly composed of larger trees. Degradation of Boswellia

papyrifera has long been recognized as ecological concern and the species has been listed by

TRAFFIC as one of the endangered species that need priority in conservation (Marshall 1998).

A number of anthropogenic factors have been contributing to the degradation of the resource

(Figure 1.3). First, human population growth and the associated increasing demand for

agricultural land is causing an extensive shifting cultivation and continuous deforestation of the

resource. In the northern Tigay and north west of the Amahra regional states, where this

resource is found, human population is increasing due to both natural growth and migration of

people from the dry highlands as well as through government resettlement programs (Eshete,

2011). Secondly, overtapping of the remaining population for frankincense to meet the

increasing demand for the product both for domestic and international markets is a major threat.

Rijkers et al. (2006) suggested that Boswellia papyrifera trees need 4 years of rest in order for

trees to recover and heal their wounds. However, trees are intensively tapped for continuous

years without getting rest for healing their wounds. The current practice of tapping is too

intensive and damaging and it is negatively affecting survival, growth and reproduction of trees

(Ogbazghi, 2001; Rijkers et al., 2006; Eshete, 2011). Thirdly, the forest lands are serving as

grazing areas and heavy and unmanaged free grazing by livestock has been practiced for long

and is increasing with human population for the fact that mixed crop livestock farming is the

major economics activity of rural people in the country. Studies indicated that free grazing is one

of the factors for the degradatation of the resource through its negative effect on regeneration

and recruitment of seedling of Boswellia papyrifera (Gebrehiwot et al., 2002; Moges and Kindu,

2006; Adam and Tayeb, 2008; Nigussie et al., 2008). In addition, browsed tree species

constitute a vital component in livestock productivity by serving as an important source of animal

feeds (Bamualin et al., 19980; Aregawi et al., 2008). In addition to its use as a source of

frankincense, Boswellia papyrifera is among the browsed species in the dry forests of eastern

Africa (Melaku et al., 2010). Thus, there is high pressure on the species in terms of leave

lopping for livestock feed. Herdsmen facilitate the accessibility of the leaves to their livestock by

using sticks or stones, by shaking the tree and/or its branch and lopping the leaves (Aregawi,

2008). It is sometimes common to observe Boswellia papyrifera trees with barely few leaves in

the middle of the wet season in which the tree is supposed to have a full crown cover (Figure

1.3).

7

Figure 1.3: Shifting cultivation, free grazing, and leaf lopping for fodder, and tapping for frankincense in

Central and Western Boswellia forests of Tigray (Pictures taken in March 2009 and August 2009)

However, as far as our knowlege is concerened, there is no scientific study that looked into the

effect of leaf lopping on frankincene yield, flower and fruit production capacity of Boswellia

papyriera (Del.) Hochst. The handful of previous as well as recent studies on the species

ovelooked the effect of leaf lopping and their focuses were rather much on the effects of tapping

intensities on yield (Rijkers et al., 2006; Tilahun et al., 2011; Eshete et al., 2012a), flowers and

fruits (Rijkers et al., 2006), seed quality, germination and storage behaviours (Eshete et al.,

2012b), and phothosyntheic carbon gain (Mengistu et al., 2012). In addition, there is also no

scientific eveidence on the amount of carbon stock in the biomass and soils of Boswellia

papyrifera forests, which is important for understaning both the forest’s capacity to sequester

CO2 from the atmosphere and the emission of this greenhouse gass due to the deforestation

and degradation of the forest. Moreover, sustainable management and conservation of the

resource requires designing an optimal forest management option for the species that can take

into account the trade-offs between the competing uses of the resource responsible for its

degradation. More specifically, understanding the bio-economic effects of leaf lopping for

fodder, tapping, and free grazing on frankincense yield, flower and fruit production is important

for designing conservation and sustainable frankincense production options. Thus this research

tries to fill these research gaps through adressing the following research questions.

1.2. Research questions

This research explores the effects of leaf lopping for fodder, tapping for frankincense, and free

grazing on the biophysical traits as well as the economic value of Boswellia papyrifera forests in

Ethiopia; the trade-offs between conservation, production forestry, and shifting cultivation; and

the role of the frankincense forests on rural livelihood and poverty reduction. Specifically, the

study addresses the following research questions:

1. How do leaf lopping, tapping as well as environmental and tree level dendrometric

variables affect the productivity (in terms of frankincense yield) and reproductive

capacity (in terms of flowering and fruit production) of Boswellia papyrifera trees?

2. How much biomass and soil organic carbon is stored in Boswellia papyrifera forests?

8

3. How much are the opportunity costs (in terms of net benefits from shifting cultivation) of

keeping Boswellia papyrifera forest under management options that range from pure

conservation to the business as usual practice that involves intensive frankincense

production, leaf lopping, and free grazing?

4. Are rural households residing in Boswellia papyrifera forest areas willing to contribute

for a conservation intervention?

5. What is the contribution of income from frankincense on poverty reduction in rural areas

with frankincense cooperative firms?

1.3. Aim and objectives of the study

The general aim of this research is to contribute to the very limited literature on the economic

and ecological values of dry land forest ecosystems and their role in the livelihood of rural

people. The study seeks to assess the bio-economic impacts of experiment-based options of

frankincense forest management, and the trade-offs in terms of the economic values associated

with the current practice of converting the forest to agriculture through shifting cultivation; and

identify the implications of the resource’s degradation on loss of biodiversity and greenhouse

gas emission. The study will enhance both ecological and economic rationality among the

decision makers and stakeholders in the urgent need for curbing the current state of

degradation and high risk of extinction of the species through restoration and/or conservation

and sustainable management practices. The study explores the effects of leaf lopping for

fodder, tapping for frankincense, and free grazing on the biophysical traits as well as the

economic value of Boswellia papyrifera forests in Ethiopia, the trade-offs between conservation,

production forestry, and shifting cultivation, and the role of the frankincense forests on rural

livelihood and poverty reduction. By answering the above research questions in section 1.3, the

corresponding objectives that this study aimed to achieve are:

1. To assess the effect of leaf lopping and tapping as well as environmental and tree level

dendrometric variables on the productivity (in terms of frankincense yield) and

reproductive capacity (in terms of flowering and fruit production) of Boswellia papyrifera.

2. To assess the stocks of biomass and soil organic carbon and nutrients in Boswellia

papyrifera forests and analyse the implication of deforestation through shifting

cultivation on these indirect ecosystem services.

3. To value the provisioning and regulating ecosystem service of Boswellia payrifera forest

under alternative management options that range from pure conservation to the

business as usual unsustainable use of the resource and compare these values with

the net benefits from the alternative land use, which is shifting cultivation.

9

4. Assess rural households’ demand for conservation of Boswellia papyrifera forest and

identify the determinants of their willingness to contribute for the conservation of the

forest.

5. To assess the role of Boswellia papyrifera forest on rural livelihood and poverty

reduction.

The specific objectives of the study are:

To assess the effects of leaf lopping on frankincense yield, flowering, and fruit

production in Boswellia papyrifera trees.

To compare the effects of tapping throughout the tapping season versus tapping after

the end of the fruiting period on frankincense yield, flower and fruit production in

Boswellia papyrifera trees.

To assess the effects of site level environmental factors on frankincense yield, flowering,

and fruit production in Boswellia papyrifera trees.

To assess the effects of tree level dendrometric characteristics on frankincense yield,

flowering and fruit production in Boswellia papyrifera trees.

To develop an allometric model specific to Boswellia papyrifera tree species.

To compare the predictive performance of the allometric model of Boswellia papyrifera

tree with the predictive performance of mixed species allometric models available in the

literature.

To estimate the biomass and soil organic carbon stock as well as soil nutrient stocks in

Boswellia papyrifera forests.

To assess the effect of fencing Boswellia paprifera stands for protection from free

grazing on concentrations and stocks of soil organic carbon and nutrients.

To make an economic analysis of different frankincense forest management options that

represent pure conservation, exclosures, and the business as usual practice (which

involves intensive frankincense tapping, leaf lopping and free grazing).

To assess the opportunity cost of reducing emissions of carbon dioxide from degradation

and deforestation of frankincense forests for the alternative forest management options.

To evaluate the distributional effects of alternative options of frankincense forest

management on the stakeholders involved in the supply chain of the frankincense

market.

To evaluate alternative frankincense forest management options in terms of their

contribution for achieving the objectives of United Nations Collaborative Program on

Reducing Emissions from Deforestation and Forest Degradation in developing countries

(UN-REDD Program). The UN-REDD+ program objectives include reducing emissions,

10

biodiversity conservation, poverty reduction, and sustainable forest management and

enhancement of forest carbon stocks.

To identify the factors that affect households’ willingness to make contributions either in

cash or free labour for the conservation of Boswellia papyrifera forest.

To estimate the rural households’ willingness to pay and willingness to contribute labour

for Boswellia papyrifera forest conservation.

To assess the level of asymmetry in the willingness to pay and willingness to contribute

responses of rural households for the conservation of Boswellia papyrifera forest.

To assess factors affecting rural households’ decision to involve as member in rural

frankincense cooperative.

To assess the effect of membership in rural frankincense cooperative firms on rural

income and rural poverty.

To assess the effect of non-timber forest products other than frankincense on rural

household income and poverty.

1.4. Study area description

This study was conducted in Tigray Regional State located in northern Ethiopia between 12° to

15° N and 36° 30′ to 40° 30′ E. Agriculture is the dominant economic activity in the region and

according to the Bureau of Plan and Finance of the Tigray Regional State, the sector

contributes about 40% of the regional gross domestic product (GDP). It consists of crop

production, livestock rearing and crop-livestock mixed farming. Natural gums and resins,

sesame and hides and skins are the major export items from the region (BoPF, 2010). Studies

indicate that about 85% of the population in the region depend on subsistence agriculture for

their livelihood, mainly by practicing mixed crop-livestock farming system (Tesfay, 2006). The

crop production is predominantly rain-fed and practiced on small and fragmented farmlands

using low input traditional technology. Moreover, land degradation and recurrent droughts have

been causing declining and highly variable land productivity (Holden et al., 2003; Yesuf et al.,

2008). Livestock rearing is both a means of income and source of wealth in rural areas of the

region. According to data from the Central Statistical Authority, the grazing livestock population

in Tigray, which includes cattle, sheep and goats, and equines summed into tropical livestock

units (TLU) increased from about 2.47 million to 3.41 million in five years time over the period

2005/06 to 2010/11.

Although reports indicate that poverty is generally declining in Ethiopia over the last one

decade, the incidence is still very high. According to the recent interim report of the Ministry of

Finance and Economic Development on the result of the 2010/11 national Household Income,

Consumption and Expenditure Survey, the proportion of people living in poverty was 29.6%

11

(with 30.4% in rural areas and 25.7% in urban areas). The poverty level for Tigray was higher by

2.2 percentage points than the national average.

In Tigray, central and western zones are known with Boswellia papyrifera forest areas.

According to Gebrehiwot et al (2002), the total Boswellia forest area in the two zones was

332,562 ha of which 98.5% was in six districts (Kafta Humera, Tahtay Adiabo, Welkayt,

Tselemiti, Tsegede, and Asegede Tsimbila) located in western Tigray whereas the remaining

small proportion was in central Tigray. Although much of the resource is found in western

Tigray, all previous experimental studies on the species from the region were carried out in

central Tigray. This might be for cost reasons because the latter is in close proximity to the

regional capital. Therefore, based on this fact, availability of frankincense cooperative firms, and

still taking somehow into account the practical cost reasons for undertaking the field research,

we selected four districts for the study. The three districts are from western Tigray (Kafta

Humera, Tahtay Adiabo, and Welkayit) and one district called Abergelle was selected from

central Tigray. Figure 1.4 shows the location of the study sites where household survey and plot

level experiments were conducted. According to the 2007 national population census, the four

districts have a total population of 416,366, which accounts 9.65% of the total population of the

region (CSA, 2008). The rural population in the four districts account 86.87% of the total

population whereas the remaining is an urban population. Similar to the regional economy,

agriculture is the main source of livelihood in these districts. Sorghum, sesame, teff, and maize

are the major crops grown in the districts. Moreover, the districts are also the main sources of

frankincense, which is the major export item from the region. According to the recent

classification of the vegetation resources of Ethiopia, these sites belong to the dry lowland with

Combretum-Terminallia and Acacia-Commiphora woodlands (Friss et al., 2010) in which gum

and resin bearing trees are the dominant species.

The site in Abergelle district in Central Tigray covers two villages or locally called ‘tabias’ with an

altitude range of 1530-1680 m above sea level. The second site with an altitude of 760-997 m

above sea level is in western Tigray and consists of three villages that belong to three

neighbouring administrative districts, namely Kafta Humera, Tahytay Adiyabo, and Welkayit

(see Villages of Household Survey in Figure 1.4). According to data based on the FAO’s (2005)

local climate estimator software, the rainfall at both sites is characterized by a unimodal

distribution with total annual rainfall of 818 mm and 1029 mm at sites in central and western

Tigray respectively. The rainy season is from June to September in both sites and lasts for 90

day in the first site and for 95 days in the second site. Diurnal minimum and maximum

temperature per month are 16 °C and 28.6 °C for the site at Abergelle and 16.70 °C and 36.10

°C for the site in western Tigray. Boswellia papyrifera is the dominant and relatively biggest tree

species in both sites.

12

We selected three Boswellia papyrifera populations at Abergelle and two populations at the site

in western Tigray for plot level experiments. The five populations differ in microclimate and soil

conditions (see Locations of Boswellia payrifera Populations in Figure 1.4). The distance

between populations within a site ranges from 2.5 to 7 km. We took GPS readings in March

2009 to determine the area of each Boswellia forest populations. Accordingly, the areas of the

five Boswellia populations of the study sites range from 8.23 to 85.39 ha with a total area of

272.05 ha.

Figure 1.4: Location of the study area

1.5. Research methodologies

This thesis is a multidisciplinary research work based on data from plot level experiment,

household survey, interviews with key informants, and secondary sources. Details on the

methods of data collection are given in the materials and methods sections of each of the main

chapters. For analysing these biophysical and socio-economic data, a number of statistical and

econometric models are applied. These include multilevel linear mixed model, the standard

allometric model, the model of environmental cost benefit analysis, the double-bound

dichotomous contingent valuation method, and standard impact assessment parametric and

non-parametric econometric models.

13

The multilevel linear mixed model approach was applied to evaluate the effect of leaf lopping,

tapping for frankincense, site level environmental variables, and dendrometric characteristics on

frankincense yield and production of flowers and fruits by Boswellia papyrifera trees. The

standard plant allometric model (Niklas, 2004) was used to develop an allometric model specific

to the species and apply it for estimating dry biomass and carbon stock in Boswellia forests.

The model of environmental cost benefit analysis (ECBA) is applied for assessing the trade-offs

among managing the forest under the six experiment based forest management options derived

from the results this study and that of shifting cultivation as the competing land use. Based on

recent developments on carbon accounting in the framework of the United Nations

Collaborative Program on Reducing Emissions from Deforestation and Forest Degradation in

developing countries (UN-REDD Program), the opportunity costs of REDD+ specific to

Boswellia paprifera forest were estimated. The method of multi-criteria analysis is used to rank

six forest management options according to objectively measurable criteria derived from the

results of the study that are in line with the UN-REDD+ program objectives.

In order to analysis rural households’ willingness to pay in cash and contribute free labour for

Boswellia papyrifera forest conservation a Double-Bound Dichotomous Choice model of

contingent valuation was applied (Hanneman, 1984; Hanemann et al., 1991; Cameron and

Quiggin, 1994). The probit, bivariate probit, and interval-data probit discrete choice models were

used to analyse the determinants of rural households willingness to pay (WTP) in cash as well

as willingness to contribute labour (WTCL) for Boswellia papyrifera forest conservation.

For assessing the welfare impact of frankincense cooperative membership, a number of

econometric models were applied. These include the method of ordinary least squares (OLS),

instrumental variable estimation techniques (IV-estimation), and Propensity Score Matching

(Rosenbaum and Rubin, 1983).

1.6. Outline of the thesis

This dissertation is a multidisciplinary piece of work that integrates biophysical aspects with

socio-economic dimensions of Boswellia papyrifera forest management using a bio-economic

analysis. The first chapter introduces the background and rational of the study, the objectives,

the study area, and outline of the dissertation. The next five chapters, which are the main

bodies of the dissertation, can be classified into two interlinked parts. The first group (Chapters

2 and 3) is devoted to the biophysical aspects where as the other three chapters (Chapter 4, 5,

and 6) are about the economic aspects of Boswellia forest management.

14

Figure 1.5 below shows the schematic presentation of the main chapters of the thesis and their

interrelationship as well as the guiding conceptual framework. Natural resources provide goods

and services that are important to human welfare. The level of goods and services that

Boswellia papyrifera forests provide depends on the natural environment through its supply of

CO2, H2O and solar energy that are important inputs for the process of photosynthesis. It also

depends on the type of forest management practices (conservation, exclosure, free grazing and

leaf lopping) in place. Following the work of Pearce (1993), a number of authors (Kengen, 1997;

Bishop, 1999; Cambell and Luckert, 2002) proposed a framework for valuing forests the goods

and services that Boswellia papyrifera forests provide could be classified into direct, indirect,

option and non-use values. The sum of these values is termed as the total economic value of

the resource. According to the Millennium Ecosystem Assessment, ecosystem services are the

benefits people get from ecosystems and these benefits are classified into four, which are

provisioning, supporting, regulating, and cultural services (MEA, 2003). Direct use values are

the goods and services that directly accrue to the consumers of the forest products.

Frankincense, grass and leafs used as animal feed, and wood for fuel and making of farm

implements are examples of the direct use values from Boswellia papyrifera forests. According

to the Millennium Ecosystem Assessment, these use values can be classified as the

provisioning services (MEA, 2003).

Indirect use values refer to some special functions of the forest ecosystem that accrue indirectly

to both users and non-users. These values include the supporting, regulating, and cultural

ecosystem services. The supporting services that Boswellia papyrifera forests provide include

provisioning of habitat for wildlife, honey bees frequently visit Boswellia papyrifera flowers

during the dry season, production of oxygen as a product of photosynthesis, and contributions

of the forest to soil formation and retention, and nutrient cycling, and water cycling. Boswellia

papyrifera forests store carbon, sequester carbon dioxide from the atmosphere, and control

erosion, which all can be considered among the main regulating services of the forest. The

forest also provides cultural service in that its frankincense is used for clerical services in

different religions, used as fragrance in Ethiopia’s cultural coffee ceremony, and has

educational value for health related research and developments.

Chapter 2 is about how forest management affects some of the provisioning and supporting

services of Boswellia papyrifera forests in biophysical terms. More specifically, the chapter

addresses research question 1 of the dissertation that inquires how frankincense yield,

flowering and fruit production are affected by leaf lopping and intensity of tapping as well as tree

level characteristics and site factors. The third chapter is about the regulating services of the

forest and aimed at answering research question 2. It discusses about the allometric model for

15

Boswellia paprifera tree species, the biomass and carbon stock at tree and per ha levels as well

as the soil carbon and nutrient stock per ha of Boswellia forests.

Figure 1.5: Schematic presentation of the thesis outline and conceptual framework (Adapted from Pearce

(1993); MEA (2003))

The economic implications of the results of chapter 2 and 3 are analysed in Chapter 4. This

chapter is related to research question 3 of the thesis. It analyses the economic and ecological

values of some of the provisioning, supporting and regulating services of the forest under

different forest management options that include conservation, exclosure, and free grazing

forest management scenarios and compares them with the alternative land use, which is

production of agricultural crops through shifting cultivation.

Once different forest management options are evaluated in economic terms, implementation of

the best management option requires the full participation and willingness of rural communities.

Natural Resource (Boswellia papyrifera forest)

DIRECT USE VALUES

Regulating services Carbon stocks in biomass and soils (avoided emissions), carbon sequestration, and erosion control

Provisioning services Frankincense (for medicine, perfume, chemical industries), grass and leaf fodder, fuel wood

INDIRECT USE VALUES

OPTION VALUES

NON-USE VALUES

Environment (CO2, H2O, Solar Energy)

Supporting services Habitat for wildlife, flowers for bees, oxygen, soil formation, nutrient cycling, water cycling wers, Fruits

Preservation Value (Rural Households’ Demand for Conservation)

Cultural services (Frankincense use for clerical services in different religions, coffee ceremonies), educational values

Forest management (Conservation, Exclosure, Free grazing and leaf lopping)

CHAPTER 3

CHAPTER 4

CHAPTER 2

WELFARE EFFECTS ON RURAL HOUSEHOLDS

CHAPTER 5

LAND CONVERSION VALUE (Net benefits from shifting cultivation)

CHAPTER 6

16

Conservation is one of the scenarios evaluated in Chapter 4 and it implies the other two

components of the total economic value, which are namely the option and non-use values.

Weisbrod (1964) first introduced the idea of option value, which can be viewed as an insurance

premium that one would be willing to pay to ensure the supply of the direct and indirect uses of

a given resource later in time. Non-use values refer to those elements of value that are

unrelated to current, future or potential uses (Krutilla, 1967) of the forest. It is composed of

existence, bequest and altruistic values of the resource. The existence value, as first explained

by Krutilla (1967), is the willingness to pay for the knowledge that a natural resource is

preserved without the intention of personal present or future use whereas bequest value refers

to the willingness to pay for the utility derived from conserving the resource for improving

welfare of future generations. The altruistic value is value based on the utility derived from

knowing that somebody else benefits. Preservation value includes option, existence, and

bequest values (Walsh et al., 1984). Therefore, a further analysis on whether there is demand

for conservation of Boswellia papyrifera among the rural households is the theme of chapter 5

that addresses research question 4 of the study.

Sustainable forest management and conservation is important in improving human welfare. On

the other hand, the welfare impacts of natural resource use, for example income from

frankincense as a provisioning service, in rural areas could be used as an incentive for

mobilizing rural households in conservation activities. In this regard, chapter 6 provides

evidence on the impact of membership in frankincense cooperative firms on rural households’

income and poverty reduction. Concluding remarks and policy recommendations are provided in

the last chapter.

CHAPTER 2

Effects of leaf lopping for fodder and resin tapping from Boswellia

papyrifera trees on frankincense yield, flowering and fruiting

2.1. Introduction

Boswelllia papyrifera is a deciduous resin bearing species of the tropical dryland forests.

Frankincense from this tree species is among the most widely traded non-timber forest products

in the world markets for its use as input in food, cosmetics, and pharmaceutical industries as

well as for ritual purposes. However, the resource base is under continuous degradation due to

a number of anthropogenic factors. Since the 1991 policy change that allows private sector

involvement in Ethiopia, the export of natural gums and resins of the country has been

increasing due to the entry of new private firms in to the market that previously was fully under

the monopoly of a state owned enterprise. In response to meeting the increasing demand, trees

have been tapped excessively both in terms of the number of tapping spots and frequency of

tapping. According to data from the Central Statistical Authority of Ethiopia (Figure 1.2), the

annual average export volume of natural gums and resins prior to 1991 was about 750 tons. It

was increasing afterwards and reached 3000 tons a year over the period 2003-2007. Recent

studies indicate that annual frankincense yield per tree increases with tapping intensity (Tilahun

et al., 2011, Eshete, 2011) and tapping frequency but in the long run yields may decrease in

view of damage to the tree that eventually may lead to the decline of the tree population

(Eshete, 2011).

Resins from trees are considered as a plant defence for and to seal off injured barks, to prevent

desiccation, protect against insect and fungal attacks, and injuries from decay (Hart 1989;

Philips and Croteau, 1999; Rijkers et al., 2006). Rijkers et al. (2006) found that tapping for

frankincense has negative effect on tree vitality and reproduction and strongly suggest that

there is a trade-off between resin production and production of tree reproductive organs in the

tree’s allocation of carbohydrates. In a very recent study, Eshete et al. (2012b) compared seed

germination success in tapped and untapped Boswellia trees and reported significantly higher

viability for seeds from untapped trees. Tilahun et al. (2011) also reported a relatively lower yield

per tree in the first four harvests from December to February, during which the tree produces

flowers, fruits and seeds, compared to a relatively larger yield in the remaining tapping

operations made over one harvesting season. This suggests competition in the allocation of

carbohydrates for growth process and resin synthesis during the first period. Moreover, a recent

study exhibits that tapping reduces non-structural carbohydrate concentration and the total non-

structural carbohydrate pool size (concentration X biomass) in the plant system of Boswellia

papyrifera trees (Mengistu, 2011).

18

Browse tree species constitute a vital component in livestock productivity in arid and semi arid

tropical Africa by serving as an important source of animal feeds (Bamualin et al., 19980;

Aregawi et al., 2008). Boswellia papyrifera is among the browse tree species in the dry forests

of Africa. Melaku et al. (2010) reported the nutritive values, which include chemical composition,

in sacco dry matter and degradability as well as in vitro dry matter digestibility, of foliages of 12

browsed species and ranked the leaves of Boswellia papyrifera as superior in its nutritive values

as animal feed. Tropical deciduous tree species absorb carbon when they have a full green

canopy and store it in the form of non-structural carbohydrates for future use and for uses

during the leafless period (Newell et al., 2002; Bansal and Germino, 2009; Mengistu, 2011). The

stored carbohydrates will be allocated to resource capture and growth, for storage purposes,

reproduction, and defence (Rijkers et al., 2006; Mengistu, 2011). However, the erratic and

unreliable rainfall in arid and semi-arid tropical regions limits biomass yield in the growing

season of herbaceous species and hence exacerbating the poor supply and quality of livestock

feed in the region leading to high demand and overexploitation of the natural vegetation of

browse species. The leaf period of Boswellia payrifera is during the wet season. However, due

to the shortage of livestock feed, the leaves are highly exposed to browsing by livestock and

cannot reach full canopy cover (Figure 1.3, the 3rd picture). Moreover, herdsmen facilitate the

accessibility of the leaves to their livestock by using sticks or stones, by shaking the tree and/or

its branch and lopping the branches with leaves (Aregawi, 2008). This will limit the plant’s

capacity to undertake photosynthesis and hence store carbohydrates for its future requirements

for reproductive purposes and synthesis of frankincense, which both follow immediately at the

beginning and/or after the tree sheds its leaves.

As far as our knowledge is concerned, no studies reported the effects of leaf lopping from

Boswellia papyrifera trees on frankincense, flowers and fruit production. Existing literature on

the species rather focus on the effect of tapping on frankincense yield, flowers and fruit

production, seed germination, and photosynthesis carbon gain (Rijkers et al., 2006; Tilahun et

al., 2011; Eshete et al., 2012a; Eshete et al., 2012b; Mengistu et al., 2012). Therefore, it is our

hypothesis that leaf lopping is an important and underestimated factor decreasing the fitness of

the Boswellia resource and its ability to produces frankincense, flowers, and fruits. Moreover,

Rijkers et al. (2006) argued that tapping causes competition for carbohydrates between

frankincense production, and fruit and seed setting. Their argument is in line with the fact that

frankincense harvesting by tappers traditionally take place during the dry season in which the

tree also produces flowers and fruits. In addition, their tapping experiments were based on the

conventional full season tapping. Taking this knowledge in to account, we further hypothesised

that tapping throughout the dry season (conventional tapping) has a significant negative effect

on the fitness of the tree and its capacity to produce flowers and fruits whereas tapping after the

19

end of the fruiting period has no significant negative effect on flowers and fruits production by

the tree.

Frankincense from Boswellia papyrifera is one of the few oldest commodities that have been

traded in the international market. However, the resource base is under inexorable degradation

and so far, there is no scientific method of sustainable harvesting system developed for the

species. Sustainability and optimization of the different ecosystem services that the species

provides, mainly frankincense production, is still the major challenge. In addition to the effect of

tapping, understanding the effects of leaf lopping for fodder on both frankincense yield and

reproduction parameters of the species is crucial in designing an appropriate sustainable

harvesting system specific to the species. It is also important to understand the effects of

environmental and tree level traits on the productivity of the tree. Moreover, such studies help to

make further cost benefit analysis of leaf lopping from the tree. Thus, we evaluate what factors

determine frankincense, inflorescence, and fruit production by Boswellia papyrifera. Specifically,

we deal with the following questions. a) Does leaf lopping decrease frankincense, inflorescence,

and fruits production by the tree? b) What is the effect of phenotypic based tapping (tapping at

different tapping spots after the end of the fruiting period versus tapping at comparable tapping

spots throughout the season) on frankincense, inflorescence, and fruit production?, and c) how

do environmental and dendrometric variables affect production of frankincense, inflorescence

and fruit in Boswellia papyrifera trees?

2.2. Materials and methods

2.2.1. Experimental plot design and soil data collection

To assess the effects of leaf lopping and tapping on the production of frankincense, flowers and

fruits by Boswellia papyrifera trees plot level experiments were conducted for two seasons from

August 2009 to June 2011 in the five Boswellia papyrifera populations of the study area (Figure

1.4). The region of the study area is characterised by erratic and unreliable rainfall owing to the

semi arid environment. Depending on the objectives of our study and the experiments we

carried out, mainly for the leaf looping experiment, the period of investigation was sufficient in

that there were relatively good rains during the rainy seasons in both years and we observed

complete phenology of the tree (leaf, flowering, and fruiting) in both periods. In the five

Boswellia populations (Figure 1.4 see Locations of Boswellia payrifera Populations), we

established 14 permanent plots each with an area of 1-2 ha that sum up to 18.7 ha. Seven of

them were fenced with barbed wire to protect from free grazing and we employed guards to look

after all the permanent plots. In each population, the distance between a fenced and an

unfenced permanent plots ranges from 5 to 10 meters. Table 2.1 below summarizes some

characteristics of the five Boswellia populations and distribution of the permanent plots.

20

Table 2.1: Description of the five Boswellia papyrifera populations and size and distribution of sample

plots

Population Sample Plots

Name Area in ha Attitude in masl (Min-max) Unfenced Fenced Total Area in ha

K. Humera I 85.39 841-989 1 1 2 2.16

K. Humera II 85.12 760-997 1 1 2 2.16

Abergelle I 24.17 1610-1680 2 2 4* 6.28

Abergelle II 8.23 1568-1656 1 1 2* 4.00

Abergelle III 69.14 1530-1631 2 2 4* 4.16

Total 272.05 7 7 14 18.76

Note: * two of the plots (one fenced and one unfenced in each case) were established in August 2007 by Abeje et al.

(2011) whereas all the other plots were established in August 2009.

We divided each permanent plot into 25-50 quadrants of each 20mx20m size and we selected

10 quadrants as subplots for inventory and soil data collection as follows: we numbered all the

quadrants and selected one random quadrant as a subplot after which the selection of the other

subplots happened by leaving a quadrant as a gap between subplots. For each subplot, we

measured environmental variables (altitude and slope, stoniness, soil depth, and soil chemical

properties). In each subplot, we lay three transect lines at 5m, 10m and 15m from the corner of

the length along the width of the plot. First, we measured altitude at the corner points of the

subplot and its centre using GPS and the average of the five readings was taken as the altitude

for the subplot. Slopes in percent were measured at three points along the transect using a

Clinometer and the average of the tree readings was taken as the slope for a subplot. Second,

to measure stoniness we lay a meter tape on the ground along each of the three transects and

recorded each of the distances covered by soil, gravel, cobble, stone, and boulder. We sum the

distances covered by gravel, cobble, stone, and boulder and divided the result by 20m to get

the proportion of stone cover per transect. The average value of the three transect was then

multiplied by 100 to get the percent of stoniness for the subplot. Third, along each transect we

measured soil depth using metal pine at 5 points (at 0, 5m, 10m, 15m, and 20m). The

measurement was done first by hammering the metal pine down to the soil until stone blocks it

and then measuring the part of the metal pine not in the soil and finally deducting this length

from the total length of the pine to get the soil depth. Depending on the soil depth at the

sampling points, soil samples at depths of 0 to 0.2m, 0.2 to 0.4m, and 0.4-0.6m were collected

from three of the five points of each transect (at 5m, 10m, and 15m). The samples collected

from the 9 points of the three transects were mixed thoroughly to form one composite sample

per soil depth per subplot. For the whole study area, we collected 175 composite soil samples,

which include 140 composite soil samples for the depth 0 to 0.2m, 31 samples for depth of 0.2

to 0.4m and only 4 samples for the depth of 0.4 to 0.6m. Soil samples were air-dried, sieved

using a 2mm sieve and then ground before analysis. We used the Walkley and Black method

(Walkley and Black, 1934) for determining soil organic carbon, the Kjeldahl method (Bremner

21

and Mulvaney, 1982) for determining total soil nitrogen, and the Olsen and Sommers (1982)

method for determining available phosphorous. We also measured the electric conductivity and

pH-water of each composite sample.

2.2.2. Sample tree selection, leaf lopping and tapping experiments

We measured the diameter of all trees in the subplots at 1.3m height and determined the

frequency distribution of the diameter for each population separately. Based on the distribution,

Boswellia papyrifera trees were stratified into small, medium, and large diameter classes, in

which the number of trees in each class is one-third of the total number of trees in a population

(see Appendix 2A). Accordingly, three diameter classes were determined for each population

and used for selecting representative samples from each population. Based on the diameter

classes specific for each population, three sample trees (one from each diameter class) were

randomly selected per subplot. In few subplots where we encountered no trees in the range of a

specific diameter class, we took trees from other diameter classes. Accordingly, we selected

420 trees from the five populations for the whole experiment. Tree height, stem length, crown

diameter of each sample tree were measured and bark colour (yellow or orange) of each tree

was recorded. Total tree height was measured using clinometer whereas stem height (which is

the height of the stem from the ground to the first branch of the tree that forms the crown) is

measured using meter tape. Crown diameter was taken as the average of measurements taken

from two directions (first is along the direction of the longest branch of a tree and the second

measurement was along the direction perpendicular to the first measurement) using a meter

tape. The bark colour of each tree was classified as either yellowish or orange by visual

observation of the colour of the bark of the tree.

We randomly allocated the subplots of each permanent plot to leaf lopping and tapping factor

combinations. The leaf lopping has two levels (no leaf lopping and leaf lopping) and the tapping

factor has five levels (no tapping, tapping at 6 spots for full season (TFS6s), tapping at 12 spots

for full season (TFS12s), tapping at 6 spots after the end of fruiting (TAF6s), and tapping at 12

spots after the end of fruiting period (TAF12s)). We thus have a 2 by 5 design with 14

replications of subplots (with a total of 42 sample trees) for each combination of the levels of

leaf lopping and tapping factors. These combinations are:

1. No leaf lopping with no tapping (NLNT).

2. Leaf lopping with no tapping (LNT).

3. No leaf lopping with tapping full season at 6 spots (NLTFS6).

4. Leaf lopping with tapping full season at 6 spots (LTFS6).

5. No leaf lopping with tapping full season at 12 spots (NLTFS12).

6. Leaf lopping with tapping full season at 12 spots (LTFS12).

7. No leaf lopping with tapping at 6 spots after fruiting ends (NLTAF6).

22

8. Leaf lopping with tapping at 6 spots after fruiting ends (LTAF6).

9. No leaf lopping with tapping at 12 spots after fruiting ends (NLTAF12).

10. Leaf lopping with tapping at 12 spots after fruiting ends (LTAF12).

Number of flowers per tree and number of fruits per inflorescence were the response variables

to the above ten treatments whereas for assessing the effect of leaf lopping and tapping on

frankincense yield as a response variable, a 2X4 design with similar replications as above can

be established by dropping the level “no tapping” from the tapping treatment. All the sample

trees in each of the leaf lopping-tapping treatment combinations were marked to identify them

during the experimental period.

The extent of leaf lopping from a particular tree may vary on its height, crown area, and the

slope of the specific location of the tree, which all determine the accessibility of leaves for

looping. Therefore, it was not easy to distinguish different classes of leaf lopping intensities that

can hold for all sample trees. Therefore, for this practical reason we distinguished only two

classes, which are absence and presence of leaf lopping. All accessible leaves were looped

manually from the sample trees selected for the treatment with presence of leaf lopping while

the leaves of the samples for the other treatment level were not lopped. It was done by climbing

on the trees and lopping the leaves by hand. The first season leaf lopping was carried out in

August 2009. However, the leaf period started earlier in June and we had no control over what

happened prior to the start of our leaf lopping experiment. Therefore, by the time we started the

experiment, leaves had already been lopped from some trees. For the fact that this will cause a

bias on the results of the experiment, we repeated the experiment in the next season in July

and August 2010. To avoid the possible bias that could arise due to cutting of leaves by local

people, we hired permanent guards to look after our experiments over the experimental period

from July 2009 until June 2011.

Tapping and frankincense collection: Frankincense tapping from Boswellia papyrifer trees

involves making a number of very small incisions (tapping spots) into the bark layer at the start

of the season and widening them over the course of the next tapping rounds to draw out the

white liquid resin from the resin canals of the wood. It is carried out by shaving the bark with the

help of a scalpel-like tool locally called ‘mengaff.’ The collection of the solidified and hardened

resin called the frankincense starts from the second round of tapping. After collecting the

hardened resin from the previous round of tapping, the tappers do widening of the tapping spot

by shaving into the bark so that new liquid resins will flow out and dry for the next collection. The

tool used for this purpose has small, flattened blade and round wooden handle with a prominent

edge (see panel (h) in Figure 2.1).

23

Figure 2.1: Frankincense tapping season (October to beginning of June) and phenological periods for

Boswellia papyrifera trees in Western and Central Tigray, Ethiopia. Leaf (a) =starting period, (b) = full leaf

period and (c) = period of leaf shading. Flowering (d) and fruiting (e). Panel (g) shows a tapper while

making the first incisions (tapping spots) at the start of the tapping season in October using a tool locally

called ‘Mengaff’ (Panel (h)). The overlay of panels (c) to (f) on panel (g) indicates that trees undertake

shading of their leaves, flowering, and fruiting and are at the same time subject to tapping

The tapping experiment was conducted for two production seasons. The first season was from

October 2009-June 2010 and the second was from October 2010-June 2011. The five levels of

the tapping experiment were no tapping, tapping at 6 spots for full season, tapping at 12 spots

for full season, tapping at 6 spots after the end of fruiting, and tapping at 12 spots after the end

of the fruiting period. The five levels of tapping were chosen considering two factors, which are

tapping intensity and period of tapping. In the case of tapping intensity measured in terms of

number of tapping spots, we selected 0 tapping spots representing no tapping, tapping at 6

spots representing normal tapping according to the traditional local tapping practice, and

tapping at 12 spots representing heavy tapping. Rijkers et al. (2006) suggested that trees

subject to the conventional full season tapping are required to allocate their stored carbohydrate

for resin production and this negatively affects their capacity to produce flowers and fruits. In

order to test this hypothesis, we selected two taping periods one representing full season

tapping (October to June) which is the conventional tapping period and the other tapping only

after the end of the fruiting period of the tree (March to June). The full season tapping was from

October to beginning of June (Figure 2.1). In this period, sample trees were tapped for 15

rounds and frankincense collection was carried out for 14 rounds. The first tapping was done at

the start of a season and collection of the resin started from the second round of tapping

24

onwards. The tapping after fruiting ends, was from March to beginning of June and it is

determined based on the phenology of the tree (Figure 2.1). Before the start of tapping in this

period, trees have already finished their flowering and fruiting functions. Over this period, trees

were tapped for 7 rounds and resins were collected for 6 rounds. Frankincense yield per tree

was determined by weighing the collected resin, which is locally termed as ‘grezo’ directly after

collection using a digital weight balance with a precision of 0.01g. The annual yield per tree was

then taken as the sum of the 14 collections for the trees tapped for the full season, and the sum

of the 6 collections for trees tapped after the fruiting ends.

Inflorescence and fruit data collection: Following Rijkers et al. (2006), the number of

inflorescences of each sample tree was counted during the flowering and fruiting periods for two

seasons (January-March 2010 and January-March, 2011). For each sample tree fruits were

counted for five inflorescences, of which four were from selected arbitralily from the periphery

and one from the centre of the crown. The counting of inflorecences and fruits was carried out

twice per year with an interval of one month during the flowering and fruiting period and we took

the larger value for the analysis. Description of all the variables measured in the experiment are

given in Appenix 2B.

2.2.3. Data analysis

In order to summarize the variation in terms of the dendrometric and environmental variables

between the different treatment groups, first the variables were reduced to principal components

using Principal component analysis (PCA). The dendrometric variables were reduced to two

PCAs (PCA1_Dend and PCA2_Dend, which together describe the major (71.13%) biotic

variation among trees). Each PCA explains 37.04 and 34.09% of the total variance,

respectively. The environmental variables were also condensed into two PCA’s explaining

71.74% of the variance of which the PCA1_ENV explains 37.04% and PCA2_ENV explains

34.71% the total variance. Second, scatter plots were used to observe whether there exist

differences in terms of environmental as well as dendrometric variables between sample trees

in among the different treatment groups. The differences were tested using Kruskal-Wallis H

test Moreover, difference in each of the environmental as well as dendrometric variables

between populations of sample trees as well as sample trees in each combination of the leaf

lopping and tapping factors were analyzed with multivariate ANOVA.

A multilevel linear mixed model approach (Singer, 1998) was applied to estimate the effects of

the factor variables (leaf lopping and tapping), dendrometric variables, number of trees per plot

and environmental variables on tree level frankincense yield, number of inflorescence, and

number of fruits. All the response variables (frankincense yield, number of inflorescence, and

number of fruits per inflorescence) were log10 transformed to meet assumptions of normality.

25

The dendrometric variables (condensed in two principal components) were the level 1 variables.

All the other variables (number of trees and environmental variable (altitude, slope, soil depth,

stoniness, and soil variables, that include pH-H2O, electric conductivity, concentrations of

carbon, phosphorus, and nitrogen)) were measured at the subplot level were level 2 variables

(Singer, 1998). Leaf lopping and tapping were used as fixed-effect factors and the other

variables (levels 1 and 2) were used as covariates in the model. The level 1 variable

(PCA1_dend and PCA2_dend) were included in the random effect variables, together with the

intercept. The variances of the random effect variable were both unity and they were

uncorrelated. Therefore, a variance-covariance structure of scale identity was used. Tree was

treated as the subject-grouping variable. The two-way interaction of the factor variables (leaf

lopping and tapping) was included. The Akaike information criterion (AIC) was used to compare

a set of reduced models to the full model and select the model with the lowest AIC as the best

model. We used SPSS 16.0 (SPSS Inc., Chicago, IL) for the statistical analysis.

2.3. Results

2.3.1. Characteristics of Boswellia papyrifera populations

The five Boswellia populations were distinct in terms of all assessed environmental

characteristics (Multivariate ANOVA: Wilks’ lambda = 0.00, F36, 477.7 = 180.70, P < 0.001) as well

as in terms of number of trees per plot and dendrometric variables (Multivariate ANOVA: Wilks’

lambda = 0.149, F24, 454.7 = 13.15, p < 0.001) (Table 2.2). The three populations in the Abergelle

site are on a higher altitude with shallower soils and greater stoniness than the two populations

in the Kafta Humera site. The very shallow soil depth in the three populations of the first site

with relatively better concentrations of soil nutrients (carbon, phosphorous, and nitrogen), higher

electric conductivity and pH-H2O than the two populations in the Kafta Humera site is in line with

the presence of limestone and schist in the Abergelle site. This indicates that Abergelle sites are

relatively better suggesting a higher productivity potential than the Kafta Humera sites. Yet the

dendrometric differences tend to show, at least partly the opposite. Boswellia populations in the

Kafta Humera site were relatively with larger total tree height, stem heights, and have wider

crowns than tree populations in the Abergelle site. This could be due to differences in past

management history of the stands in the two sites. The stands in Abergelle sites have been

used for frankincense production for relatively longer periods compared to the stands in Kafta

Humera, in which human settlement and intensive production of frankincense is relatively a

recent phenomenon to the area.

Majority (85.7%) of Boswellia trees in the Kafta Humera populations had yellowish bark whereas

the bark colour for the majority (81.3%) of Boswellia trees of the Abergelle populations was

orange. Trees in the five populations were significantly different in their dimension (DBH) only at

5% level and there was no statistically significant difference in terms of the number of trees per

26

plot (Table 2.2). Moreover, as indicated in Table 2.2, there were also statistically significant

differences between the five populations in terms of the mean values of frankincense yield,

number of inflorescences, and number of fruits per inflorescence of both survey years.

Table 2.2: Mean values (SE) for environmental, dendrometeric, and response (flowers, fruits and

frankincense yield) variables of five Bosweliia papyrifera populations in northern Ethiopia and the

significance of differences between the populations

Variables

Population F4, 135 pa

Abergelle-I Abergelle-II Abergelle-III K.Humera-I K.Humera-II

Environmental

Altitude (m a.s.l.) 1648(2.73) 1630(1.62) 1580(0.81) 983(1.45) 804(1.66) 33464 ***

Slope (%) 8.11(0.60) 22.13(3.15) 13.02(0.72) 11.75(1.36) 19.03(1.31) 17.36 ***

Soil depth (m) 0.15(0.01) 0.08(0.00) 0.07(0.00) 0.30(0.02) 0.24(0.01) 91.72 ***

Stoniness (%) 69.76(1.56) 78.78(2.20) 76.47(1.17) 45.50(2.14) 39.77(3.43) 73.21 ***

pH-H2O 7.67(0.01) 7.41(0.04) 7.83(0.02) 5.90(0.07) 6.51(0.13) 242.99 ***

EC (µS/cm) 149(2.03) 223(26.23) 123(3.28) 40.75(3.76) 77.75(18.79) 31.93 ***

Carbon (%) 6.86(0.43) 9.48(0.66) 3.65(0.35) 1.58(0.15) 2.80(0.19) 47.61 ***

Phosphorus(mg/kg) 17.61(0.81) 23.60(1.35) 13.86(0.66) 7.97(1.53) 9.20(0.60) 34.18 ***

Nitrogen (%) 0.37(0.01) 0.43(0.43) 0.18(0.01) 0.08(0.00) 0.11(0.00) 166.45 ***

Dendrometric

DBH (cm) 20.93(0.37) 19.22(0.55) 21.55(0.51) 20.70(0.93) 23.53(0.64) 5.56 ***

Height (m) 4.80(0.08) 5.04(0.13) 5.30(0.12) 8.22(0.52) 9.12(0.39) 66.44 ***

Stem height 2.08(0.06) 2.31(0.13) 2.39(0.07) 2.71(0.13) 2.81(0.11) 10.72 ***

Number of trees 9.18(0.66) 7.80(0.81) 7.73(0.48) 9.30(0.81) 5.75(0.33) 1.98 0.10

Bark color (% of yellow)

17.00(0.04) 22.00(0.06) 18.00(0.04) 92.00(0.04) 83.00(0.07) 56.58

***

Crown diameter (m) 4.54(0.13) 4.31(0.18) 3.82(0.18) 4.94(0.28) 4.90(0.18) 6.60 ***

Flowers and Fruits

Inflorescence/tree 1 18.18(1.92) 19.42(2.48) 3.49(0.89) 22.73(3.35) 19.36(2.67) 15.89 ***

Inflorescence/tree2 42.82(3.60) 38.93(6.28) 14.52(2.64) 60.87(8.69) 38.30(3.71) 13.91 ***

Fruit/inflorescence1 12.80(1.44) 15.56(1.62) 3.58(0.82) 9.99(2.48) 10.58(1.79) 10.94 ***

Fruit/inflorescence2 41.57(2.38) 42.07(2.92) 17.37(1.70) 26.86(3.05) 20.66(1.87) 26.02 ***

Yield F4,107

Frankincense1

(g/tree) 387.57 (48.27)

333.23 (63.83)

189.42 (31.43)

503.55 (114.48)

360.81 (74.48)

3.77 0.007

Frankincense2

(g/tree) 325.26 (39.53)

356.81 (62.60)

273.09 (46.29)

481.64 (110.92)

302.41 (64.92)

1.54 0.197

Note: 1 year 2010, 2 year 2011; *** p < 0.001; a

A Multivariate ANOVA; for all environmental variables: Wilks’ lambda

= 0.00, F36, 477.7 = 180.70, p < 0.001; for number of trees and dendrometric variables: Wilks’ lambda = 0.149, F24, 454.7

= 13.15, p < 0.001.

2.3.2. Environmental and dendrometric characteristics of sample trees

Before analysing the effects of leaf lopping and tapping treatments on frankincense yield,

flowering and fruiting, the variations in terms of environmental variables, dendrometric variables

as well as number of trees per plot among the different treatment groups have to be tested. The

27

scatter plots in Figure 2.2 (a and c) indicate that the trees in the leaf lopping experiment (not

lopped and lopped) were not separated when ordination was based on both PCAs of

environmental variables as well as PCAs of dendrometric variables. The two PCAs of the

environmental variables explain 71.74% of the total variance. PCA1_ENV explains 37.04% of

the total variance of all the environmental variables whereas PCA2_ENV explains the remaining

34.71%. Among all the environmental variables, the value of PCA1_ENV is relatively very much

affected by the values of soil pH-water, altitude, soil depth, and stoniness whereas the value of

the PCA2_ENV was relatively affected much by soil nutrients, electric conductivity of the soil,

altitude, and slope. In the case of the PCAs of the dendrometric variables, PCA1_Dend and

PCA2_Dend together account 71.13% of biotic variation among trees. PCA1_Dend explains

37.04 and much of the value is affected by DBH and crown diameter whereas PCA2_Dend

accounts the rest 34.09% of the total variance. Much of the value of PCA2_Dend is affected by

tree height and bark colour. Accordingly, the samples in the two levels of the leaf lopping

experiment did not reflect differences in terms of site potential (Figure 2.2a), dendrometric

variable (Figure 2.2c) as well as level of competition among trees (Table 2.3).

In terms of tapping treatment as well, trees in each of the corresponding levels (not tapped,

TFS6s, TFS12s, TAF6s, and TAF12s) did not form separated group when ordination was based

on environmental (Figure 2.2b) as well as dendrometric variables (Figure 2.2d). A further test

using the Kruskal-Wallis H test proves the descriptions in figure 2.2. First in the leaf lopping

treatment groups, there was no statistically significant difference in PCA1_Dend between the

two leaf lopping treatments (H(1) = 1.478; p = 0.224) with a mean ranks of 217.70 for leaves not

lopped group and mean rank of 203.30 for the leaves lopped group. In the case of PCA2_Dend

the corresponding values were (H(1) = 0.536; p = 0.464) with mean ranks of 215.84 for trees in

the leaves not lopped group and 206.16 for the leaves lopped group indicating no statistically

significant difference in denedrometric variables.

In terms of the environmental variables that explain the site characteristics, there was no

statistically significant difference in PCA1_ENV between the two leaf lopping treatments (H(1) =

0.023; p = 0.879) with a mean ranks of 211.40 for leaves not lopped group and mean rank of

209.60 for the leaves lopped group. In the case of PCA2_ENV the corresponding values were

(H(1) = 0.126; p = 0.723) with mean ranks of 212.60 for trees in the leaves not lopped group

and 208.40 for the leaves lopped group indicating no statistically significant difference in

environmental variables.

Similar statistical test result was also observed for the case of the tapping treatment groups. We

found no statistically significant difference in PCA1_Dend between the five tapping treatments

(H (4) = 5.109; p = 0.276) with a mean rank of 224.58 for not tapped, 224.55 for TFS6s, 208.43

28

for TFS12s, 188.62 for TAF6s, and 206.32 for TAF12s. In the case of PCA2_Dend the

corresponding values were (H(4) = 0.510; p = 0.973) with a mean rank of 211.18 for not tapped,

218.15 for TFS6s, 206.67 for TFS12s, 206.58 for TAF6s, and 209.92 for TAF12s indicating no

statistically significant difference in denedrometric variables. In the case of the environmental

variables too, there was no statistically significant difference in PCA1_ENV between the five

tapping treatments (H (4) = 1.020; p = 0.907) with a mean rank of 206.16 for not tapped, 216.71

for TFS6s, 217.36 for TFS12s, 201.88 for TAF6s, and 210.39 for TAF12s. In the case of

PCA2_ENV the corresponding values were (H(1) = 1.007; p = 0.909) with a mean rank of

211.30 for not tapped, 205.68 for TFS6s, 203.54 for TFS12s, 211.30 for TAF6s, and 220.68 for

TAF12s indicating no statistically significant difference in environmental variables among the

different tapping treatment groups.

Therefore based on that above analysis and a results of multivariate ANOVA in Table 2.3, the

sample trees in the different treatment groups did not show statistically significant differences in

terms of site variables (Multivariate ANOVA: Wilks’ lambda = 0.701, F81, 797.4 = 0.54, p = 0.999)

and number of trees and dendrometric characteristics (Multivariate ANOVA: Wilks’ lambda =

0.716, F54, 642 = 0.81, p = 0.837) (Table 2.3).

29

Figure 2.2: Scatter plots of 420 trees within 140 plots sampled from five Boswellia papyrifera forest

populations in northern Ethiopia. (a-b) principal component analysis of environmental variables and (c-d)

principal component analysis of dendrometric variables. Sample trees are labeled according to levels of

the leaf lopping and tapping factors

30

Table 2.3: Mean values (SE) for environmental, dendrometric, and response (yield, flowers and fruits) variables in ten groups (Leaf lopping X tapping as factors) of 140 plots

within five Bosweliia papyrifera populations in northern Ethiopia (per group N=14)

Variables NLNT LNT NLTFS6 LTFS6 NLT12 LTFS12 NLTAF6 LTAF6 NLTAF12 LTAF12 Fdf1,df2 Pa

Environmental F9.130

Altitude (m) 1409(91) 1411 (91) 1411(92) 1412(92) 1411(92) 1410(92) 1411 (92) 1411 (92) 1409(93) 1413 (91) 0.00 1.00

Slope (%) 15.57(3.52) 12.89(2.51) 13.69(2.27) 13.93(2.50) 14.48(2.20) 12.79(2.04) 15.40(2.26) 12.62(1.32) 12.43(2.10) 12.16(1.58) 0.29 0.98

Soil depth (m) 0.15(0.02) 0.16(0.03) 0.15(0.03) 0.16(0.03) 0.16(0.03) 0.15(0.03) 0.15(0.02) 0.15(0.02) 0.14(0.02) 0.16(0.03) 0.07 0.99

Stoniness (%) 67.86(4.44) 62.99(6.41) 67.03(4.91) 63.36(4.97) 66.30(5.33) 59.14(5.15) 68.75(3.94) 64.04(4.66) 65.37(4.53) 67.31(3.89) 0.35 0.96

pH-H2O 7.14(0.28) 7.29(0.16) 7.26(0.20) 7.27(0.19) 7.26(0.19) 7.33(0.21) 7.21(0.21) 7.23(0.21) 7.33(0.18) 7.28(0.20) 0.07 0.99

EC (µS/cm) 146(23.87) 125(16.08) 116(13.89) 122(14.88) 116(15.44) 116(11.94) 149(44.05) 124(19.95) 129(15.93) 123(14.99) 0.31 0.97

Carbon (%) 4.05(0.72) 4.77(0.83) 4.78(0.92) 4.52(0.84) 5.02(1.00) 5.06(0.93) 4.99(0.95) 5.14(0.82) 6.37(1.21) 5.12(0.98) 0.41 0.93

Phosphorus (mg/kg) 16.20(2.24) 15.21(1.55) 13.32(1.58) 14.90(2.03) 16.20(1.94) 14.33(1.65) 14.07(2.29) 12.85(1.46) 15.83(2.29) 15.24(1.93) 0.37 0.95

Nitrogen (%) 0.23(0.04) 0.24(0.04) 0.22(0.03) 0.25(0.04) 0.26(0.05) 0.25(0.04) 0.25(0.04) 0.24(0.03) 0.25(0.04) 0.25(0.03) 0.09 0.99

Denderometric

DBH (cm) 21.82(0.96) 21.33(0.86) 22.17(0.86) 21.95(0.76) 20.98(0.65) 19.76(0.68) 21.02(0.84) 20.70(1.12) 21.63(0.74) 20.65(1.02) 0.72 0.69

Height (m) 5.78(0.44) 6.41(0.78) 6.33(0.53) 6.17(0.66) 6.21(0.48) 6.16(0.49) 6.13(0.61) 6.31(0.70) 5.74(0.47) 5.60(0.30) 0.24 0.99

Stem H (m) 2.41(0.16) 2.42(0.17) 2.31(0.10) 2.39(0.14) 2.34(0.15) 2.54(0.14) 2.26(0.17) 2.51(0.14) 2.34(0.12) 2.43(0.16) 0.35 0.96

Crown diameter (m) 0.26(0.24) 4.41(0.30) 4.24(0.26) 4.78(0.32) 4.54(0.23) 4.40(0.33) 4.86(0.31) 4.51(0.35) 4.28(0.24) 3.80(0.21) 1.13 0.35

Number of trees 8.79(2.00) 6.64(0.63) 9.79(2.33) 9.36(1.69) 6.93(0.85) 8.79(1.06) 7.14(1.61) 11.50(2.31) 6.79(1.26) 6.86(1.80) 0.99 0.45

Bark color (% Yellow) 0.48(0.11) 0.43(0.10) 0.43(0.10) 0.43(0.11) 0.29(0.10) 0.31(0.10) 0.38(0.12) 0.38(0.12) 0.38(0.10) 0.33(0.11) 0.31 0.97

Flowers and Fruits

Inflorescence/tree 1 15.02(3.11) 19.33(4.55) 18.95(4.16) 13.21(3.95) 14.76(3.53) 8.08(2.10) 14.81(3.38) 15.07(2.88) 17.83(3.85) 12.74(3.27) 0.89 0.54

Inflorescence/tree2 40.17(8.23) 36.03(9.95) 43.48(8.10) 29.45(6.62) 37.55(8.53) 22.98(5.05) 40.55(7.71) 34.02(7.01) 41.14(8.13) 35.76(6.72) 0.63 0.77

Fruit/inflorescence1 8.97(1.98) 15.67(3.41) 12.67(2.62) 6.82(1.93) 7.90(1.83) 5.95(1.22) 11.47(3.11) 9.60(2.27) 11.70(1.76) 8.73(1.91) 1.65 0.11

Fruit/inflorescence2 28.05(3.69) 29.83(5.73) 31.79(4.05) 30.92(2.92) 30.15(4.70) 21.69(2.78) 31.10(4.91) 28.05(5.08) 36.59(5.16) 28.22(4.66) 0.71 0.70

Yield F7.104

Frankincense1 (g/tree) 476(69.46) 458(52.52) 729(84.92) 557(76.87) 96(13.72) 99(12.96) 143(15.53) 130(19.42) 22.71 ***

Frankincense2 (g/tree) 469(80.16) 402(47.49) 732(83.75) 527(62.24) 103(12.54) 102(10.36) 186(22.00) 152(16.35) 21.13 ***

Note: 1 year 2010,

2year 2011;*** P < 0.001;

a Multivariate ANOVA; for all environmental variables: Wilks’ lambda = 0.701, F81, 797.4 = 0.54, p= 0.999; for number of trees and dendrometric variables: Wilks’

lambda = 0.716, F54, 642 = 0.81, p= 0.837.

31

2.3.3. Variables affecting frankincense yield

The two factor variables (leaf lopping and tapping) and the PCAs of the dendrometric variables

(PCA_Dend_1 and PCA_Dend_2) as well as some of the environmental variable (altitude,

stoniness, and soil pH-H2O) have significant effects on frankincense yield of both production

seasons (Table 2.4). We found no statistically significant interaction effects of the factor

variables. At tree level, the main effect of leaf lopping on frankincense yield was negative and

significant for the 2009/10 production season (F1, 184 = 3.14, p = 0.08) as well as the 2010/11

production season (F1, 182 = 3.83, p = 0.05). The main effect of tapping on yields of both seasons

was positive and significant (F3, 187 = 79.72, p < 0.001; F3, 184 = 70.56, p < 0.001).

Among the environmental variables, the main effects of altitude and nitrogen concentration in

the soil on yield of 2009/10 were both positive and significant (F1, 171 = 7.52, p= 0.01; F1, 193 =

10.24, p < 0.001) but only altitude had significant effect on the yield of 2010/11 (F1, 177 = 19.46, p

< 0.001). The other environmental variables with significant but negative main effects on yields

of both production seasons were percent of stoniness (F1, 191 = 8.50, p < 0.001; F1, 189 = 6.28, p =

0.01) and the soil pH-H2O (F1, 154 = 14.31, p < 0.001; F1, 158 = 12.89, p < 0.001).

Both the PCAs of the dendrometric variables, which were considered as random in the model,

had positive and significant effects on yield of the last production season (F1, 20 = 37.80, p <

0.001; F1, 245 = 3.22, p = 0.07). Nevertheless, in the case of yield of 2010/11 it is only

PCA1_Dend with significant main effect (F1, 223 = 23.45, p < 0.001). The number of trees per plot

had no significant effect on yields of 2009/10 (F1, 173 = 2.03, p = 0.16) and 2010/11(F1, 172 = 0.28,

p = 0.60).

32

Table 2.4:Type III tests of fixed effects of three factors (leaf lopping and tapping), plot level environmental

variables, and dendrometric variables (condensed in two PCA dimensions) on two seasons frankincense

yield (log10 transformed of yield per tree) in 140 plots within 5 Boswellia papyrifera forest populations in

northern Ethiopia

Source df Season

2009/10 2010/11

df(error) F P df(error) F P

Linear mixed model

AIC

312.99 241.65

Intercept 1 144.90 36.44 *** 145.55 50.28 ***

Leaf lopping 1 166.39 3.81 0.05 165.22 5.43 0.02

Tapping 3 173.20 70.63 *** 171.43 62.40 ***

Leaf lopping X Tapping 3 172.78 1.02 0.39 171.58 0.99 0.40

Altitude (masl) 1 159.96 4.30 0.04 169.82 18.69 ***

Slope (%) 1 188.03 0.01 0.94 187.21 0.74 0.39

Soil depth (m) 1 151.94 0.35 0.55 150.92 0.94 0.33

Stoniness (%) 1 167.05 3.94 0.05 169.07 8.04 0.01

pHwater 1 136.53 6.05 0.02 141.57 12.09 ***

EC (µS/cm) 1 211.32 0.00 0.95 208.23 0.00 0.97

Carbon (%) 1 183.59 0.84 0.36 180.59 1.35 0.25

Phosphorus (mg/kg) 1 187.85 0.09 0.76 183.60 0.29 0.59

Nitrogen (%) 1 188.38 5.95 0.02 185.04 1.41 0.24

Number of trees 1 173.04 2.03 0.16 171.75 0.28 0.60

PCA_dend_1 1 219.77 24.60 *** 218.01 36.96 ***

PCA_dend_2 1 251.54 0.00 0.95 250.78 2.37 0.12

Restricted linear mixed model

AIC

265.46

192.27

Intercept 1 180.16 129.67 *** 180.64 113.88 ***

Leaf lopping 1 184.40 3.14 0.08 181.63 3.83 0.05

Tapping 3 186.87 79.72 *** 183.67 70.56 ***

Altitude (masl) 1 171.41 7.52 0.01 177.02 19.46 ***

Stoniness (%) 1 190.70 8.50 *** 188.45 6.28 0.01

pH-water 1 153.48 14.31 *** 157.68 12.89 ***

Nitrogen (%) 1 193.30 10.24 *** 189.77 0.18 0.68

PCA_Dend_1 1 223.25 23.45 *** 220.01 37.80 ***

PCA_Dend_2 1 248.11 0.02 0.88 244.74 3.22 0.07

Note: *** p < 0.001

33

2.3.4. Variables affecting inflorescence and fruit production

Leaf lopping and tapping did not show significant effects on inflorescence and fruit production in

the 2010 phenological period, but their interacting effect on fruits for the same season was

negative and significant (F4, 199 = 3.10, p = 0.02) (Table 2.5). However, in the 2011

phenological period it was only leaf lopping that exhibited negative significant effects on both

inflorescence (F1, 225 = 6.25, p = 0.01) and fruit production (F1, 222 = 6.44, p = 0.01). The effects

of tapping and its interaction with leaf lopping on both response variables were insignificant.

In the case of the environmental variables, the nitrogen content in the soil had significant

positive effect on both flowering (F1, 269 = 18.23, p < 0.001; F1, 256 = 29.45, p < 0.001) and fruiting

(F1, 249 = 60.97, p < 0.001; F1, 259 = 61.05, p < 0.001) of the two seasons. Soil depth had also

similar effect on both flowering (F1, 246 = 29.46, p < 0.001; F1, 249 = 37.68, p < 0.001) and fruiting

(F1, 219 = 16.43, p < 0.001; F1, 245 = 14.63, p < 0.001) of both seasons. The main effect of percent

of soil carbon on the number of inflorescence was positive and significant (F1, 257 = 2.91, p =

0.09; F1, 245 = 5.39, p = 0.02) whereas its effect on the number of fruits was insignificant for both

seasons. The main effects of both PCAs of the dendrometric variables on flowering of both

seasons were positive and significant. Their effects on fruiting in the two seasons were also

positive and statistically significant except for that of PCA2_dend. The statistics corresponding

to each season for the effects of PCA1_Dend on inflorescence are (F1, 228 = 53.00, p < 0.001; F1,

241 = 101.40, p < 0.001) and on fruits are (F1, 207= 55.10, p < 0.001; F1, 237 = 41.76, p < 0.001).

The statistics for effects of PCA2_Dend on inflorescence are (F1, 218 = 17.47, p < 0.001; F1, 253 =

12.38, p < 0.001), and on fruits are (F1, 205 = 9.95, p < 0.001; F1, 247 = 0.38, p = 0.54).

In this study, we found that the conventional dendrometric variables (DBH, H, stem height, and

crown diameter) of the tree have a positive correlation with the frankincense and fruit

productivity. In addition, we observed that tree bark colour significantly affects both frankincense

yield (ANOVA: F2, 333 = 3.68, p = 0.026) and the flower (F2, 417 = 18.22, p < 0.001) and fruit (F2, 417

= 3.29, p =0.038) production of the tree.

34

Table 2.5:Type III tests of fixed effects of two factors (leaf lopping and tapping), environmental and

dendrometric variables on two seasons log transformed number of inflorescence and fruits per

inflorescence of Boswellia papyrifera trees in northern Ethiopia

Source df Inflorescence Fruits per inflorescence

2010 2011 2010 2011

dferror F P dferror F P dferror F P dferror F P

Linear mixed model

AIC 655 653 666 647

Intercept 1 173 5.10 0.03 178 12.68 *** 150 0.68 0.41 177 6.13 0.01

Leaf lopping 1 212 2.47 0.12 204 8.46 *** 181 1.10 0.30 204 8.38 ***

Tapping 4 219 0.80 0.53 211 0.87 0.48 189 1.79 0.13 212 1.15 0.33

Leaf lopping X

Tapping

4 218 2.00 0.10 210 0.45 0.77 187 3.09 0.02 210 0.58 0.67

Altitude (masl) 1 183 0.27 0.60 198 0.37 0.54 163 0.23 0.63 195 2.05 0.15

Slope (%) 1 240 1.69 0.19 230 0.07 0.79 209 0.14 0.70 231 0.03 0.87

Soil depth(m) 1 179 2.96 0.09 187 2.18 0.14 156 3.06 0.08 185 1.26 0.26

Stoniness (%) 1 184 0.01 0.92 193 0.86 0.35 162 0.00 0.96 191 0.60 0.44

pH-H2O 1 157 1.70 0.19 169 4.05 0.05 137 0.25 0.62 166 2.32 0.13

EC (µS/cm) 1 249 0.40 0.53 224 2.16 0.14 211 0.44 0.51 226 0.46 0.50

Carbon (%) 1 237 4.58 0.03 224 4.55 0.03 204 0.10 0.75 225 1.76 0.19

Phosphorus

(mg/kg)

1 266 0.40 0.53 235 0.26 0.61 226 0.11 0.74 239 1.53 0.22

Nitrogen (%) 1 235 16.04 *** 223 25.15 *** 203 14.98 *** 224 22.35 ***

No. of trees 1 210 0.82 0.36 210 0.00 0.99 183 0.00 1.00 209 0.57 0.45

PCA_dend_1 1 245 50.07 *** 254 91.26 *** 225 48.06 *** 253 43.37 ***

PCA_dend_2 1 280 11.39 *** 313 7.83 0.01 275 4.56 0.03 309 1.08 0.30

Restricted linear

mixed model

AIC 592 586 594 582

Intercept 1 250 8.70 *** 238 65.07 *** 218 0.00 0.95 236 71.49 ***

Leaf lopping 1 231 1.92 0.17 224 6.25 0.01 201 1.19 0.28 222 6.44 0.01

Tapping 4 229 0.83 0.50 224 0.97 0.42 200 1.81 0.13 222 0.69 0.60

Leaf lopping X

Tapping

199 3.10 0.02 221 0.49 0.74

Soil depth (m) 1 246 29.46 *** 249 37.68 *** 218 16.43 *** 245 14.63 ***

Carbon (%) 1 257 2.91 0.09 245 5.39 0.02

Nitrogen (%) 1 269 18.23 *** 256 29.45 *** 249 60.97 *** 259 61.05 ***

PCA_dend_1 1 228 53.00 *** 241 101.4 *** 207 55.10 *** 237 41.76 ***

PCA_dend_2 1 218 17.47 *** 253 12.38 *** 205 9.95 *** 247 0.38 0.54

Note: *** p < 0.001

35

2.4. Discussion

The study indicated that in both production seasons, trees subjected to leaf lopping produced

on average less frankincense (Figure 2.3a) and fewer flowers and fruits (Figures 2.3 c and e)

compared to trees from which leaves were not lopped, and the effects on yield and flowering

were consistent across all diameter distribution and the effects on fruits as well for the small and

middle diameter class. For example for the production season of 2010/11, among trees of less

than 20 cm in diameter, the mean value for frankincense yield was 312.06(SE = 35.18) for the

trees in the leaves not lopped group whereas it was 234.95(SE = 23.95) grams for trees in the

leaves looped group. The corresponding average values for the number of inflorescence were

22.87(SE = 2.26) and 16.67(SE = 1.98), and the values for fruits per inflorescence were

32.10(SE = 2.54) and 23.65(SE = 2.14) for the not leaves lopped and leaves lopped groups

respectively. For trees with diameter between 20 cm and less than 25cm, the mean value for

frankincense yield was 335.55(SE = 44.67) for the trees in the leaves not lopped group whereas

it was 324.17(SE = 40.37) grams for trees in the leaves looped group. The corresponding

average values for the number of inflorescence were 39.37(SE = 4.29) and 35.80(SE = 4.55),

and the values for fruits per inflorescence were 31.24(SE = 2.40) and 28.67(SE = 2.29) for the

not leaves lopped and leaves lopped groups respectively. For trees of 25cm diameter and

above, the mean frankincense yield for the leaves not lopped group was 528.30 (SE=72.34)

grams whereas it was 376.84(SE=54.15) grams for the leaves lopped group. The corresponding

average values for the number of inflorescence were 68.70(SE = 6.94) and 54.28(SE = 7.21),

and the values for fruits per inflorescence 31.03(SE = 2.00) and 234.26(SE = 2.97) for the not

leaves lopped and leaves lopped groups respectively. Unlike the small and medium diameter

classes, the number of fruits per inflorescence for leaves-lopped trees in the large diameter

class was smaller than the number for the not lopped group.

This study suggests that leaf lopping reduces the capacity of Boswellia papyrifera trees to

synthesise resin as well as produce sexual reproduction organs. Tilahun et al. (2011) reported

that trees in exclosures produced more frankincense than trees in free grazing and browsing

stands that are subject to damages of branches and twigs. Latt et al. (2000) reported that

frequent pruning in two agroforestry species reduces reserve of non-structural carbohydrates

and suggested a negative impact on biomass production. In a study on the reproductive biology

of Boswellia serrata, Sunnichan et al. (2005) suggested that the low fruiting level during the

leafless period was due to insufficient resource reserves in the stem. A recent study by

Mengistu (2011) indicated the presence of seasonal variation in the size of non-structural

carbohydrates in Boswellia papyrifera trees with very low values during the dry season

compared to the level at the end of the wet season indicating exhaustion and re-fill. Moreover,

he also found a significant positive correlation between crown leaf area and the annual crown

carbon assimilation rate as well as between crown carbon assimilation rate and photosynthetic

36

rate for the same species. Therefore, any disruptions, like browsing, frequent pruning and/or

leaf lopping during the active wet season could lead to less capacity to absorb carbon during

the wet season and hence reduce the reserve of total non-structural carbohydrate for future

use. Large trees (DBH > 25cm) (Figure 2.3e) exhibited that despite leaf lopping, they produced

on average larger number of fruits per inflorescence than trees of the same diameter range

whose leaves were not lopped. This indicates that leaf lopping affects fruit production in trees of

only the lower and middle diameter class. This could be due to larger reserves of carbohydrates

in the large diameter class trees that could help them to overcome the effect and produce either

more or equal number of fruits per inflorescence compared their counterparts.

Our study indicates that yield per tree increases significantly with number of tapping spots.

Regardless of their size, trees tapped at 12 tapping spots had more frankincense yield than

trees tapped at 6 spots. Moreover, trees tapped after the fruiting ends yield less frankincense

than tree subject to full season tapping (Figure 2.3b, Table 2.3). Similar results were reported in

recent studies in Ethiopia that looked into frankincense yield modelling (Tilahun et al., 2011) and

production (Eshete, 2011) using tree dendrometric variables and tapping intensity as

explanatory variables. In our study, we found no statistically significant difference between

untapped and tapped trees in terms of flowers and fruits they produced (Figure 2.3 d and f,

Table 2.3). Trees tapped after the fruiting period also on average had slightly larger number of

flowers and fruits than trees subjected to the full season tapping regime. Moreover, tapped trees

larger than 22 cm in diameter produced more flowers and fruits than untapped trees. However,

in all cases the differences were statistically insignificant (Figures 2.3 d and f). This is quite

strange and different from what was reported by Rijkers et al. (2006). A very recent study by

Eshete et al. (2012b) as well as Rijkers et al. (2006) also reported lower germination success of

seeds from tapped stands compared to seeds from untapped stands.

The lack of significant difference in flower and fruit production between tapped and untapped

trees in our study could be associated with the fact that the species requires low carbohydrates

for these processes or is not related to tapping the same year. Moreover, availability of sufficient

stored carbohydrates and new supplies from photosynthesis could enable the trees to

undertake both resin synthesis and growth processes at least for a given period. In the work of

Mengistu (2011) on carbohydrate allocation patterns of Boswellia papyrifera trees, it has been

shown that the annual net primary productivity for fruit production accounted for less than 0.5%

of the annual gross primary productivity of the tree. The study also indicated that the annual

gross primary productivity of the tree was more than sufficient to cover the annual carbon sinks

that the work takes into account, one of which was the sink in the form of resin production.

37

Figure 2.3: Frankincense yield (a and b), inflorescence (c and d) and fruits (e and f) per inflorescence as

a function of diameter at breast height (DBH).The regression lines show the linear fit of the standardized

Y values as a function of DBH. Panels (a, c, and e) show effect of leaf lopping and panels (b, d, and f)

show effect of tapping

Previous studies on the effect of tapping on yield as well as growth process variables of the

species were based on analysis of variance. They did not explicitly take into account the effect

of specific environmental factors. The linear mixed model we applied has allowed us to take into

account some specific environmental variables in our analysis. Thus, we found that at least for

the altitude ranges of the study sites, trees at higher altitude performed significantly better in

terms of yield than trees located at lower altitudes. However, in terms of production of flowers

and fruits, altitudinal difference has no significant effect. Although Boswellia papyrifera trees

38

generally grow on degraded and stony sites, it is revealed that extreme level of degradation that

can be expressed in terms of very high percentage of stoniness, low soil nutrients for example

nitrogen, and very shallow soil depth has significant negative impact on frankincense, flowers

and fruit productivity of the species.

In this study, we found that bark colour significantly related to the productivity of frankincense

trees. We found that compared to trees with orange bark and that of trees with intermediate

bark colour (mix of yellow and orange); trees with yellowish bark were the most productive in

terms of frankincense, inflorescence and fruits per inflorescence. Trees with orange bark colour

were the least productive. The bark could considered as a fitness indicator where orange bark

shows drying or less vigour trees with less chlorophyll which implies less photosynthetic ability

of the bark.

2.5. Conclusions

We found that lopping leaves from Boswellia papyrifera for livestock feed has a negative effect

on the fitness of the tree and its capacity to produce frankincense. In order to increase both the

productivity in frankincense yield as well as reduce the decline in flowers and fruits production

and hence enhance natural regeneration, Boswellia trees need protection from leaf lopping by

humans as well as from livestock browsing. This could be possible through strong community

based exclosure management. However, because mixed crop-livestock farming is the main

source of livelihood in Boswellia forest areas, protection of Boswellia trees from leaf lopping can

only be effective if alternative livestock feed are available. Therefore, other development

interventions and research that can solve the shortage of livestock feed in the area are needed

to reduce the pressure on the Boswellia forests. As a deciduous tree species, research on how

to harvest the litter fall and store it for livestock feed in periods of shortages could be considered

as alternative for reducing the impact of the current practice of leaf looping on the tree’s

productivity. Finally, future research is required to verify whether the difference in bark colour of

Boswellia trees is due to genetic factor or is an indication of vitality. It is also important to take

into account the bark colour in undertaking seed germination and propagation studies. The

results of such studies are important for the sustainable management of Boswellia papyrifera

forests.

CHAPTER 3

Biomass and soil organic carbon stocks in Boswellia papyrifera (Del.)

Hochst forests

3.1. Introduction

Global forest ecosystems store 650 Gt of carbon of which 44% is in biomass, 11% in dead wood

and litter, and the remaining 45% in soils (FAO, 2010b). However, deforestation and

degradation in the tropics accounts for 17% of global anthropogenic greenhouse gas (GHG)

emissions (IPCC, 2007), destroying vital carbon sinks that are sequestering carbon dioxide from

the atmosphere that threatens future climate stabilization (Stephens et al., 2007). The

importance of including emission reductions from tropical deforestation in the global climate

change policy has been growing in recent years. Gullison et al. (2007) estimated that reducing

tropical deforestation rates by 50% by 2050 and controlling them for another 50 years would

avoid a direct flux of 50 Gt of carbon underlying the crucial role of a climate policy initiative to

reduce emissions caused by tropical deforestation. The Conference of the Parties (COP) 13

decisions taken at the 2007 COP to the United Nations Framework Convention on Climate

Change (UNFCCC) in Bali have considered the possibility for reducing emissions from

deforestation and forest degradation in developing countries (REDD) to become part of a post-

2012 global climate regime (Bond et al., 2009). REDD was one of the few themes where

observers agreed that progress was made during the Copenhagen and Cancun negotiations.

The implementation of REDD plans requires feasible and scientific approaches (Gibbs et al.,

2007) which can enhance the monitoring, reporting and verification (MVR) of carbon emissions

from deforestation and forest degradation in developing countries. Methodological guidance

from UNFCCC calls on nations to apply the most recent Intergovernmental Panel on Climate

Change (IPCC) guidelines as a basis for estimating forest related GHG emissions that require

the use of a combination of remote sensing together with ground-based forest carbon inventory.

Ground-based forest inventory measurements and direct estimation of aboveground biomass

through destructive harvesting could greatly reduce uncertainty and improve the estimation of

forest carbon stocks (Gibbs et al., 2007). Allometric equation is the most accurate method to

relate the dry biomass data collected from destructive sampling to easily measurable tree

variables. These equations are also often used to validate other methods, such as the

estimation of carbon stock using non-destructive in-situ measurements and remote sensing

(Wang et al., 2003). Allometric equations can be developed for single species (Fuwape et al.,

2001; Ong et al., 2004; Cole and Ewel, 2006) or mixtures of species (Brown et al., 1989; Chave

et al., 2005) of tropical forests. However, mainly dicotyledonous tree species differ in allometry,

wood density, and architecture, all of which can affect the relationship between the

40

measurements taken during forest inventories and the biomass of individual trees (Chave et al.,

2003). Mixed species equations provide useful estimates for large-scale global and regional

comparisons, but they are of limited use for application to specific species on specific sites

(Cole and Ewel, 2006). This pooled-species approach is a reasonable method in cases where

the database to which it is to be applied includes a very large number of species, which is often

the case in tropical forests. In cases where species-specific information is available, single

species equations can provide more accurate estimates of biomass than mixed species

equations (Litton and Kauffman, 2008). The pan-tropics generalized allometric equations of

Chave et al. (2005) are developed based on an exceptionally large dataset of 2410 destructively

sampled trees. It can be used for estimating forest carbon stocks across a wide range of forest

types. However, none of the trees Chave et al. (2005) used in these equations was from an

African forest (Gibbs et al., 2007). Accurate data on forest and tree biomass from Africa are very

limited and much of our knowledge on biomass carbon pool and carbon sequestration of

tropical forests is based on studies from South and Central America (Hertel et al., 2009).

Tropical dry forests are the most extensive land cover type in the tropics and are the

predominant forest types in Africa, but there are very few studies on biomass of these forests

and its specific tree species (Urqiza-Haas et al., 2007). Therefore, research from Africa’s dry

land forests with ground-based data sets are highly important to develop site specific allometric

relationships as well as to validate existing mixed species allometric models.

The present spotlight of REDD is on countries with high forest cover and/or high deforestation

rates and fails to notice the dry land forests which are also an important carbon pool (Verbist

and Muys, 2010). There is a growing demand from developing countries with vast areas of dry

land forest to get support from the UN-REDD Programme. Sudan, Central African Republic and

Nigeria are countries with Boswellia papyrifera dry land forest and are in the current list of 29

partner countries of the UN-REDD programme. Boswellia papyrifera (Del.) Hochst is among the

economically important tree species of the dry land forests of these parts of Africa. However, an

allometric equation developed and validated for Boswellia papyrifera tree and its tree

components (leaves, branches, and stem) as well as the level of soil organic carbon stock in

this forest type has not yet been published. Such a study is important for accounting the

biomass and carbon stock in dry land forests of Africa and in the valuation of the ecosystem

services of this dry land forest in the current and future UN-REDD partner countries endowed

with this important tree species. Moreover, the allometric model for the species can also be

used for validating existing mixed species allometric models, which mostly lack field-based data

from tropical dry forests of Africa. In addition, developing allometric equations for each tree

component is important for a number of reasons. First, as a deciduous tree species, Boswellia

papyrifera produces and shade leaves every season. Therefore, developing allometric model for

the leaf could help for estimating the annual carbon dioxide that can be sequestered by the tree

41

and the forest. Second, developing allometric equations for each tree component is important

because different tree components biologically differ in their structure and size of their biomass,

which may not have the same functional relationship with tree dendrometric variables (like

diameter, height, crown area) that will be used as predictor variables in the models.

Thus, the objectives of this study are: a) to develop and validate allometric equations for the

species and each of its tree component (leaves, branches, bole, total above ground, and roots;

b) to assess the soil organic carbon, tree and herbaceous biomass carbon stocks in Boswellia

papyrifera forest ecosystem, and c) to identify the effect of fencing Boswellia paprifera stands

for protection from free grazing on concentrations and stocks of soil organic carbon and

nutrients.

3.2. Materials and methods

3.2.1. Tree biomass carbon data collection and analysis

From the 14 permanent plots of Boswellia stands of the study sites (Figure 1.4 of chapter 1), 30

sample trees representing the diameter distribution of the trees in the study sties were selected

for measuring biomass in a destructive way. Almost equal number of sample trees from each

diameter class in each site were randomly selected, which in total were 13 from the site in K.

Humera and 17 from the site in Abergelle for the destructive biomass measurement (see the

distribution of number of sample trees selected for destructive sampling in last column of the

table in Appendix 3A). The number of samples from the K.Humera site was less because of

relatively extreme hot temperature that made the laborious work of the destructive sampling

more difficult. The selection of the sample trees was made not from the subplots but from the

quadrants that were the gaps between the subplots (see section 2.2.1 of chapter 2 on plot

design).

The following measurements were made before felling the selected sample trees: diameter at

1.3m height (DBH), total tree height (H), stem length (which is the length from the ground to the

first branch that forms the crown), and average crown diameter. For crown diameter the

average of two measurements (using a meter tape) was taken, first is along the direction of the

longest branch of a tree and the second measurement was along the direction perpendicular to

the first measurement. After felling, each sample tree was divided into leaves, branches, and

stem. Roots were excavated from 4 trees of the 30 randomly selected trees. Out of the four

trees on which roots were excavated, one is from the small diameter class and is 12.1cm in

diameter, one from the medium and 18cm in diameter, and the other two trees were from the

large diameter classes were with diameters 25.4 and 29.6cm. Fresh weight of each tree

component was measured in the field using a spring weighing scale of 25 kg capacity with an

accuracy of ± 10 %. The fresh weight of samples of each component was measured using a

42

table balance with a capacity of 3500 g and accuracy of 0.01g. Samples were then stored in

sealed plastic bags and transported to Mekelle University, where they were oven dried at a

temperature of 105 °C until a constant weight was achieved (Ketterings et al., 2001; Basuki et

al., 2009). To determine the wood density (ρ), the volume of stem slices from the lower, middle,

and upper parts of the stems including bark (Basuki et al., 2009) were measured with the water

replacement method (Basuki et al., 2009). Wood density was calculated as oven dry weight

divided by volume.

For each destructively sampled tree, the total dry weights of leaves, branches and boles were

calculated by multiplying the tree level fresh weight of each component by the corresponding

dry to fresh weight ratios of the subsamples from the respective tree component. Total tree level

dry weight of aboveground biomass was then estimated by summing the dry weights of the

leaves, branches and stem. Dry root biomass was estimated for each sample tree by multiplying

the aboveground dry biomass by the average root to shoot ratio of the four sample trees for

which root excavation was carried out. Details on variable types, units of measurements, and

sample sizes for all variables measured from the destructively sampled trees are given in

Appendix 3B. Descriptive statistics and the Mann-Whitney non-parametric test for two

independent samples were used to compare the destructively sampled trees from the Abergelle

and Kafta Humera sites in terms of dendrometric characteristics and sizes of dry biomass.

The following standard form of the allometric equation was applied: Yi = a(DBH)b where Yi is dry

biomass of tree component i, DBH is the diameter at breast height, and a and b are statistical

parameters (Niklas, 2004). The scaling parameters ‘a’ and ‘b’ were determined after log-

transformation of the data using the ordinary least squares. The bias correction factor of

Snowdon (Bi et al., 2001) was used. This factor is the ratio of the average of the untransformed

biomass data used to fit the model to the average of fitted values from the logarithmic

regression model, after back-transformation from logarithms.

Following Sampaio et al. (2010), different allometric regression equations (linear, quadratic,

cubic, logarithmic, inverse, growth, S-function, exponential, and logistic) were also applied. In

these models, the oven-dry weight of each biomass component was considered as dependent

variable, and DBH, H, and wood density (ρ), and crown area (CA) were taken as covariates

individually or as a product of two variables used as a single variable.

The fit of each model to the data was evaluated using the adjusted coefficient of determination

(Adj. R2), the relative root mean square error (RRMSE), and the Akaike information criterion

(AIC). A 10-fold cross-validation (Stone, 1974) was performed to assess each model’s

predictive performance. This was done first by sorting the data of the 30 trees from the

43

destructive sampling by diameter and then stratifying it into three diameter groups each

containing 10 samples. Second sampling without replacement was applied to select one sample

at random from each diameter group to form a subset of the data. In such a way, the data was

divided into 10 subsets of equal size. For each allometric model, 10 regressions were made,

each time leaving aside one of the subsets and using that subset to calculate the prediction

accuracy (Aertsen et al., 2010). To measure the predictive performance adjusted R2, RRMSE,

and AIC were calculated using the predicted values for cross validation. We used SPSS 16.0

(SPSS Inc., Chicago, IL) for all the statistical analysis.

The data of this study was also fitted to a number of mixed species models in the literature that

are developed for dry tropical trees and to models developed for all tropical forest types. These

models are: a) the pan-tropic allometric models of Chave et al. (2005) for a dry tropical and all

tropical forest types, b) Dawkins’(1961) single model based on eight tropical species from

Trinidad, Puerto Rico and Honduras, c) the mixed species model of Ogawa et al. (1965)

developed for a dry monsoon, a mixed monsoon, a savanna forest, and a tropical rain forest; d)

the models for tropical dry forest species by Brown et al (1989) and Brown (1997), e) the model

of Brandies et al. (2006) for six Puerto Rican subtropical dry forest species, and f) Navar’s

(2009) model for tropical dry forest species of eastern Sinaloa of Mexico. We further calculated

the deviation of each prediction from the mean value of the observed aboveground dry biomass

of the destructively sampled trees and the statistical difference was tested using a t-test.

Based on the inventory data, the total amount of carbon sequestered in the biomass of the

Boswellia papyrifera trees in each subplot was determined as follows. First, the oven-dry

biomass of tree components of each tree in the subplot was predicted from its diameter using

the corresponding allometric models (Table 3.3). Root oven-dry biomass was then predicted

from the predicted aboveground biomass using an average root-to-shoot ratio (Table 3.1). The

amount of carbon sequestered in each tree component was calculated as 50% of the predicted

dry biomass (Houghton et al., 2001; Martin and Thomas, 2011). The estimates of the amounts

of carbon sequestered in each tree of each subplot were then summed to give an estimate of

the total tree biomass of each subplot. The subplot level average for each permanent plot was

then calculated and scaled up to per ha values by multiplying it with 25 (the ratio of one ha to

the size of subplot). Finally, the averages per ha were calculated for each site and for the study

area by taking the corresponding averages of per ha values of the permanent plots.

3.2.2. Herbaceous biomass carbon

To determine the aboveground herbaceous biomass, which includes grasses and herbs on the

floor of the Boswellia stands, harvesting and measurement of the aboveground grass and herbs

was carried out from five 1mx1m randomly selected quadrants in each of the 20mx20m sub-

44

plots of the fenced plots (Figure 3.1). The harvesting and measurement was done from 350

quadrants after one year since the fencing. The fresh weight (F) of sub-samples of grasses and

herbs was measured in the field and they were taken to lab for determining the oven dry weight.

The dry herbaceous biomass (DHB) of the samples was calculated through multiplying the fresh

weight by the dry to fresh weight ratio of the subsample herbaceous biomass. The herbaceous

biomass carbon (HBC) in (Mg C ha-1) was finally calculated using:

HBC = ((0.5 X DHB) (kg m-2) X 10-3) X (10000 m2 ha-1 – tree basal area (m2 ha-1))

Figure 3.1: Ground floor herbaceous and grass biomass in freely grazed (left picture) and fenced stands

of Boswellia papyrifera forest in the K.Humera site of Western Tigray. (Pictures taken in October 2010)

3.2.3. Soil carbon and nutrients

Soil samples were collected and analysed as described in subsection 2.2.1 of the materials and

methods section of chapter 2. In addition, from each permanent plot, one subplot was randomly

selected and soil core sample was taken from the selected subplot for bulk density

determination. The bulk density samples were dried to a constant weight at 105°C for a

minimum of 48 hours and then weighed. Descriptive statistics and the Mann-Whitney non-

parametric test for two independent samples were used to compare the Abergelle and Kafta

Humera sites in terms of soil characteristics. The following equations were used to calculate the

stocks (in Mg ha-1) of soil organic carbon, nitrogen and phosphorus:

SOC= (% C X 10-2) X BD (Mg m-3) X depth (m) X 10,000 m2 ha-1

TN= (% TN X 10-2) X BD (Mg m-3) X depth (m) X 10,000 m2 ha-1

Olsen-P= (P in mg/kg X 10-6) X BD(Mg m-3) X depth (m) X 10,000 m2 ha-1

where SOC is soil organic carbon, TN is total nitrogen, Olsen-P is available phosphorus, BD is

bulk density, and depth is soil depth.

45

3.3. Results

3.3.1. Description of biomass input data of harvested trees

The mean values of the input variables of the harvested trees used to determine structure-

biomass relationships and develop the allometric models for estimating stock of tree biomass

carbon are given in Table 3.1 below.

Table 3.1: Mean (SE) of dendrometric, wood density and dry biomass and Mann-Whitney Test statistics

for the two-independent samples of trees from Abergelle and Kafta Humera sites used for allometric

modelling

Variables Mean (SE) Mann-Whitney UTest

Pooled

(n=30)

Abergelle

(n=17)

K.Humera

(n=13)

Mean Rank Z-

value

2-tailed

Asymp.

Sign.

Abergelle

(n=17)

K.Humera

(n=13)

Diameter (DBH) (cm) 19.84

(1.07)

20.51

(1.42)

18.96

(1.67)

16.47 14.23 -0.69 0.49

Height (H) (m) 6.62

(0.35)

5.39

(0.14)

8.23

(0.50)

23.92 23.92 -4.60 ***

Crown depth (m) 4.10

(0.33)

3.04

(0.20)

5.49

(0.51)

9.47 23.38 -4.30 ***

Crown diameter (m) 4.23

(0.28)

4.31

(0.35)

4.11

(0.45)

14.69 14.69 -0.44 0.66

Crown area (CA) (m2) 15.42

(1.82)

15.81

(2.43)

14.92

(2.84)

16.06 14.77 -0.40 0.69

Wood density (ρ) (g/cm3) 0.58

(0.01)

0.61

(0.02)

0.54

(0.02)

10.62 10.62 -2.66 0.01

Dry leaf biomass (LB) kg) 3.57

(0.50)

2.73

(0.52)

4.66

(0.85)

12.94 18.85 -1.82 0.07

Dry branch biomass (BB)

(kg)

46.80

(7.16)

51.11

(9.84)

41.16

(10.62)

13.85 13.85 -0.90 0.37

Dry stem biomass (SB) (kg) 28.34

(2.74)

32.49

(4.09)

22.91

(2.89)

17.71 12.62 -1.57 0.12

Dry above ground biomass

(AGB) (kg)

78.70

(9.79)

86.32

(13.79)

68.73

(13.72)

13.92 13.92 -0.86 0.39

Root weight / shoot weight =

root to shoot ratio*

0.12

(0.03)

Note: *** p < 0.001; *sample size =4

The diameter of harvested trees ranges from 11 to 30.5 cm. Tree height varied from 4.57 to 13

m, and crown diameter varied from 1.55 to 6.60 m. Crown area varied from 1.65 to 33.43 m2,

whereas wood density ranges from 0.44 to 0.77 g/cm3. In the case of dry biomass

measurements, the variation was highest in leaf biomass (which varied from 0.29 to 9.56 kg per

tree, almost a 33-fold range), followed by branches (4.46 to 136.38 kg per tree, almost a 31-fold

46

range), and lowest in stem biomass (which varied from 9.59 to 67.44 kg per tree, a 7-fold

range). The Mann-Whitney Test statistics indicated that except for the values in tree height,

crown depth, wood density, and dry leaf biomass the samples from Abergelle and Kafta Humera

sites did not show statistically significant differences in terms of the other variables. We found

that wood density and total tree height have significant correlations with population (n = 30; r =

0.386; p = 0.035 for wood density Vs population and n =30; r = -0.607; p < 0.001 for height Vs

population) whereas the correlation of DBH with population is not significant. Trees from the

Abegelle populations have relatively higher wood density but shorter than trees from the

K.Humera populations.

3.3.2. Stand structure

The diameter distribution of Boswellia papyrifera trees in the study area resembles a normal

distribution (Figure 3.2). However, in both sites we found no trees below 8 cm DBH indicating

lack of recruitment. The number of trees for the pooled sample was 202.32 ha-1 with standard

error of 7.52 and mean basal area of 6.59 m2 ha-1.

Figure 3.2: Diameter distribution of Boswellia papyrifera trees in the study area

3.3.3. Soil carbon and nutrient concentrations

The soil depth in both sites was very shallow and the maximum soil depth we found was

0.542m. The soil was deeper in the K.Humera site than the Abergelle site and the difference

was statistically significant. There was no statistically significant difference between the sites in

terms of slope and bulk density. Compared to K.Humera site, soils in the Abergelle site have

higher values of pH-water, electric conductivity and concentrations of soil organic carbon, total

nitrogen and available phosphorus. The differences in all these soil properties were also

statistically significant at 1% level of error probability (Table 3.2).

47

Table 3.2: Mean (SE) of soil properties and Mann-Whitney Test statistics for the two-independent

samples from the study sites by soil depth (n1= 100 and n2=40 for depth 0 to 0.2m; n1=4 and n2=31 for

depth 0.2 to 0.4m; and n1=0 and n2=4 for depth 0.4 to 0.6m; n1 refers sample size for Abergelle site and

n2 for K.Humera)

Variables Mean (SE) Mann-Whitney UTest

Pooled

Abergelle

K.Humera

Mean Rank Z-

value

2-tailed

Asymp.

Sign.

Abergelle K.Humera

Slope (%) 13.60(0.71) 12.88(0.89) 15.39(1.10) 64.66 85.10 -2.70 0.007

Soil depth (m) 0.15(0.01) 0.11(0.01) 0.27(0.01) 51.72 117.46 -8.67 ***

Bulk density (g cm-3

) 1.29(0.03) 1.30(0.03) 1.27(0.04) 69.80 72.25 -0.32 0.746

pH-Water

0 to 0.2m 7.26(0.06) 7.68(0.02) 6.21(0.09) 90.29 21.02 -9.13 ***

0.2 to 0.4m 6.20(0.10) 7.79(0.04) 5.99(0.03) 33.50 16.00 -3.40 0.001

0.4 to 0.6m 5.88(0.44) 5.88(0.44) 0.00 2.50

Electric conductivity

(µS/cm)

0 to 0.2m 126.5(6.51) 153.5(6.50) 59.25(9.91) 88.79 24.78 -8.44 ***

0.2 to 0.4m 48.11(5.65) 130.8(7.22) 37.42(2.63) 33.50 16.00 -3.30 0.001

0.4 to 0.6m 27.00(1.23) 27.00(1.23) 0.00 2.50

Organic carbon (%)

0 to 0.2m 4.98(0.298) 6.10(0.34) 2.19(0.15) 86.00 31.75 -7.15 ***

0.2 to 0.4m 3.01(0.59) 4.81(2.50) 2.78(0.59) 26.75 16.87 -1.86 0.063

0.4 to 0.6m 0.41(0.19) 0.41(0.19) 0.00 2.50

Total nitrogen (%)

0 to 0.2m 0.24(0.01) 0.30(0.01) 0.09(0.004) 89.68 22.55 -8.85 ***

0.2 to 0.4m 0.07(0.01) 0.24(0.04) 0.05(0.002) 33.50 16.00 -3.30 0.001

0.4 to 0.6m 0.03(0.004) 0.03(0.004) 0.00 2.50

Available phosphorous

(mg/kg)

0 to 0.2m 14.82(0.60) 17.31(0.61) 8.58(0.82) 87.51 27.98 -7.85 ***

0.2 to 0.4m 5.77(0.49) 12.46(2.41) 4.90(0.13) 33.50 16.00 -3.30 0.001

0.4 to 0.6m 4.34(0.31) 4.34 (0.31) 0.00 2.50

Note: *** p < 0.001

3.3.4. Allometric relationships

In Table 3.3 a distinction is made between the goodness of fit values used for model calibration

and values for the 10-fold cross validation. For each of the goodness of fit measures, the values

for the calibration set are better than the values for the validation set in all the models. In other

words, for each model, the calibration set has higher value of Adj. R2, and lower values of

RRMSE and AIC compared to the corresponding values for the validation set. For selection of

the best model for each biomass component, selections based on the values of both calibration

and validation sets provided same result. The best model for each biomass component is

48

highlighted in bold. A model is selected as the best fit if it resulted in lowest values of RRMSE,

AIC, and highest value Adj. R2. In addition to these measures of predictive performance, the

ecological interpretability and user-friendliness (simplicity) of a model is used as a decision tool

(Aertsen et al., 2010).

Table 3.3: Predictive performance of allometric equations for dry biomass of Boswellia papyrifera tree

components (leaf biomass (LB), branch biomass (BB), stem biomass (SB), and aboveground biomass

(AGB)) as a function of tree variables (diameter (DBH), crown area (CA) height (H), wood density (ρ),

product of DBH and H (DBHH), product of DBH and CA (DBHCA), and product of DBH and ρ (DBHρ))

Variable (X) Model *BCF Adj. R2 RRMSE AIC

*Cal. *Valid Cal. Valid. Cal. Valid.

LB

DBH lnLB = ln(0.045) + 1.369(lnX) 1.30 0.18 0.05 0.90 0.95 -3.50 1.62

CA lnLB = ln(0.656) + 0.549(lnX) 1.28 0.22 0.10 0.87 0.92 -4.39 1.51

DBHH lnLB = -149.011X-1

+ 2.262 1.24 0.34 0.23 0.80 0.84 -6.78 1.27

DBHρ lnLB = -9.805X-1

+ 1.886 1.34 0.11 -0.01 0.93 1.00 -2.40 1.79

DBHCA lnLB = ln(0.261) + 0.42(lnX) 1.28 0.23 0.10 0.87 0.92 -4.46 1.53

BB

DBH lnBB = ln(4.42E-03) + 3.016(lnX) 1.02 0.92 0.91 0.08 0.08 -38.11 -1.88

CA lnBB = ln(2.947) + 0.965(lnX) 1.14 0.71 0.69 0.15 0.15 -17.98 -0.02

DBHH lnBB = -208.92X-1

+ 5.326 1.16 0.65 0.62 0.16 0.17 -15.00 0.32

DBHρ lnBB = ln(0.093) + 2.441(lnX) 1.07 0.81 0.79 0.12 0.12 -24.06 -0.73

DBHCA lnBB = ln(0.455) + 0.785(lnX) 1.10 0.82 0.80 0.12 0.17 -25.21 0.35

SB

DBH lnSB = -29.303X-1

+ 4.819 1.03 0.81 0.79 0.08 0.08 -40.97 -2.19

CA lnSB = ln(6.841) + 0.52(lnX) 1.06 0.62 0.57 0.11 0.11 -30.44 -1.18

DBHH lnSB = -107.014X-1

+ 4.162 1.07 0.50 0.45 0.12 0.12 -26.46 -0.80

DBHρ lnSB = ln(0.782) + 1.445(lnX) 1.02 0.85 0.84 0.07 0.07 -44.85 -2.63

DBHCA lnSB = ln(2.496) + 0.424(lnX) 1.04 0.72 0.68 0.09 0.09 -34.80 -1.59

AGB

DBH lnAGB = ln(0.061) + 2.353(lnX) 1.01 0.95 0.95 0.04 0.04 -53.33 -3.52

CA lnAGB = ln(9.785) + 0.749(lnX) 1.08 0.72 0.70 0.09 0.09 -26.56 -0.83

DBHH lnAGB = -161.374X-1

+ 5.566 1.10 0.65 0.63 0.11 0.11 -23.23 -0.51

DBHρ lnAGB = ln(0.586) + 1.951(lnX) 1.03 0.88 0.86 0.06 0.09 -38.55 -0.97

DBHCA lnAGB = ln(2.286) + 0.61(lnX) 1.05 0.84 0.82 0.07 0.10 -34.63 -0.59

*BCF is bias correction factor; cal. refers to calibration, and valid. refers to validation.

The allometric equations generally fit the data well, and in all the models except for leaf

biomass, more than 81% of the variation in dry biomass was explained by diameter at breast

height (Table 3.3). Taking each independent variable separately, the best fit for stem, branch

and AGB was with DBH, followed by crown area whereas for leaf biomass the best fit was with

crown area followed by DBH. Tree height and wood density as separate covariates resulted in

the worst fit with mostly insignificant coefficients. Multiplying height, wood density, and crown

49

area with DBH, resulted in better fits only for the case of stem and leaf biomass. In the case of

stem biomass models, the product of DBH and wood density provided the best fit whereas in

the case of the leaf biomass models, the product of DBH and tree height provided the best fit.

However, all the coefficients of determination for the leaf biomass models were less than 0.5

implying a relatively low fit compared to the models for other biomass components.

In this study, no correlations between height and DBH (r = 0.182; p=0.335) and between height

and aboveground dry biomass (r = 0.156; p = 0.411) were observed for the destructively

sampled trees (n=30). Using height as separate independent variable and as a combined single

variable like DBH x H have resulted in a decrease in the fitness of the model. Similar results

were observed for branch and stem biomass models. However, leaf biomass is positively

correlated with tree height (r = 0.433; p=0.017) and including height in leaf biomass model has

resulted in a relatively better improvement in the fitness of the model compared to a model with

only DBH as independent variable.

3.3.5. Carbon stocks

The total amount of ecosystem carbon stock in Boswellia papyrifera forests, which includes soil

carbon, tree biomass carbon, and herbaceous biomass carbon, was 43.48 Mg ha-1. Soil organic

carbon stock accounted 77.95% of the total ecosystem carbon (Table 3.4). Much of the soil

organic carbon was stored in the top soil constituting 67.17% of the total carbon stock of

Boswellia forest ecosystem. The total biomass and carbon stock in Boswellia papyrifera trees

were in the order 19.69 ± 2.99 and 9.85± 1.49 Mg ha-1respectively. The biomass carbon in

Boswellia trees accounted for 22.25% of the total ecosystem carbon. The largest tree biomass

allocation (ratio of biomass in tree component (organ) to total tree biomass) and carbon stock

was accounted by branches (54%) followed by stem (32%), and roots (10.4%) and with leaves

accounting only 3.6%.

The mean ecosystem carbon stock as well as soil carbon stock of Boswellia papyrifera forests

in K.Humera was higher than the corresponding figures in Abergelle site, and the differences

were statistically significant. Similar significant difference was observed when looking at the

carbon stock at the three soil depth classes. However, in terms of the stock of tree biomass

carbon, there was no statistically significant difference between the two sites.

50

Table 3.4: Mean values (SE) and Mann-Whitney Test statistics for the two-independent samples of

carbon stock in soil, trees and herbaceous biomass in Boswelllia papyrifera forest ecosystems of

Avergelle and K.Humera sites

Variables Mean(SE) Mann-Whitney UTest Pooled

Mean Rank Z-

value

2-tailed

Asymp.

Sign.

Abergelle K.Humera Abergelle K.Humera Mean (SE) %TCS

Soil organic carbon

stock (Mg ha-1)

depth: 0-60cm 25.00(2.79) 55.21(6.52) 59.59 97.78 -5.03 *** 33.63(2.95) 75.95

0-20cm 24.79(2.74) 42.13(3.78) 60.81 94.72 -4.47 *** 29.74(2.32) 67.16

20-40cm 0.21(0.17) 12.84(3.37) 55.51 107.98 -9.09 *** 3.82(1.08) 8.63

40-60cm 0.00(0.00) 0.24(0.16) 68.5 75.50 -3.20 0.001 0.07(0.05) 0.16

Tree Biomass carbon

stock (Mg ha-1)

Total 10.19(0.92) 8.99(1.33) 72.3 66.00 -0.83 0.406 9.85(0.76) 22.24

Leaf 0.36(0.03) 0.34(0.03) 70.1 71.50 -0.19 0.854 0.36(0.02) 0.81

Branches 5.52(0.53) 4.78(0.76) 72.3 66.00 -0.83 0.406 5.31 (0.44) 11.99

Stem 3.23(0.27) 2.93(0.39) 71.89 67.02 -0.64 0.521 3.15(0.22) 7.11

AGB 9.14(0.82) 8.07(1.19) 72.3 66.00 -0.83 0.406 8.83(0.68) 19.94

Roots 1.05(0.10) 0.92(0.14) 72.3 66.00 -0.83 0.406 1.015(0.08) 2.29

HBC (Mg ha-1) 0.64(0.03) 1.21(0.09) 55.01 109.22 -7.15 *** 0.80 (0.05) 1.81

Total carbon

stock (TCS)

35.83(3.05) 65.13(6.69) 60.61 95.22 -4.56 *** 44.28(3.09) 100.00

Note: *** p < 0.001

3.3.6. Effect of fencing on stocks of soil organic carbon and nutrients

The result of this study indicates that on a per ha basis soils in Boswellia papyrifera stand

fenced for a period of two to about four years have more stocks of organic carbon, nitrogen, and

phosphorus than soils in the unfenced stands and the differences are statistically significant

(Table 3.5). As indicated in Table 2.1 of chapter 2, out of the seven fenced permanent plots, 4

were fenced for two years and the other 3 were fenced for about four years.

This difference can be explained by the fact that fencing protects livestock grazing and hence

decomposition of the grass biomass and litter fall contributes to the enhancement of soil organic

carbon and nutrient contents and reduction of soil erosion. A further comparison of the physical

and chemical properties of soils in the fenced and unfenced plots suggests similar reasoning.

These variables are the variables used in the calculation of the stocks of carbon and nitrogen

and phosphorous. The fenced and unfenced plots have no statistically significant difference at

5% level in terms all the chemical properties (concentrations of carbon, nitrogen and

51

phosphorous). However, soils in the fenced plots had significantly higher bulk density and lower

percentage of stoniness than the unfenced plots (Table 3.5). The soil depth in fenced plots was

also higher than the depth in the unfenced plots, but the difference was not significant. These

variables are both influenced by free grazing and browsing.

Table 3.5: Mean (SE) and Mann-Whitney Test statistics for comparison of soil organic carbon, total

nitrogen and available phosphorous stocks between fenced and unfenced Boswellia stands (n1=n2=70)

Mean(SE) Mann-Whitney UTest

Fenced

Unfenced

Mean Rank Z-value 2-tailed

Asymp.Sign. Fenced Unfenced

Soil organic carbon stock (Mg ha

-1)

depth: 0-60cm 42.85(5.06) 24.41(2.66) 62.31 78.69 -2.39 0.017

0-20cm 35.596(3.72) 23.89(2.62) 62.90 78.10 -2.22 0.027

20-40cm 7.12(2.09) 0.52(0.17) 63.70 77.30 -2.61 0.009

40-60cm 0.14(0.090) 0.00(0.00) 68.50 72.50 -2.02 0.043

Total nitrogen stock (Mg ha-1

)

depth: 0-60cm 1.84(0.16)

1.18(0.13) 60.70 80.30 -2.86 0.004

0-20cm 1.70(0.14) 1.15(0.13) 60.86 80.14 -2.81 0.005

20-40cm 0.13(0.029) 0.032(0.010) 63.93 77.07 -2.52 0.012

40-60cm 0.008(0.0054) 0.000(0.000) 68.50 72.50 -2.02 0.043

Available phosphorus stock (Mg ha

-1)

depth: 0-60cm 0.0124(0.0014) 0.0085(0.0010) 62.71 78.29 -2.27 0.023

0-20cm 0.0107(0.0011) 0.0083(0.0010) 63.19 77.81 -2.13 0.033

20-40cm 0.0016(0.0004) 0.0003(0.0001) 64.00 77.00 -2.49 0.013

40-60cm 0.0002(0.0001) 0.0000(0.0000) 68.50 72.50 -2.02 0.043

Input data for calculating stocks

Carbon (%): 0–20cm 4,74(0.40) 5,22(0.41) 73.64 67.36 -0.92 0.359

20 -40cm 4.15(0.80) 0.82(0.04) 18.78 16.50 -0.64 0.521

40 -60cm 0.41(0.19) 2.50

Total Nitrogen (%): 0-20cm 0,24(0.02) 0,24(0.02) 70.86 70.14 -0.11 0.915

20 -40cm 0.08(0.02) 0.05(0.004) 16.04 21.75 -1.61 0.109

40 -60cm 0.03(0.004) 2.50

Phosphorus (mg/kg): 0 -20cm 14,19(0.85) 15,44(0.83) 74.76 66.24 -1.24 0.214

20 -40cm 6.37(0.70) 4.60(0.33) 18.22 17.58 -0.18 0.858

40 -60cm 4.34(0.31) 2.50

Bulk density (g/cm3) 1.32(0.02) 1.25(0.04) 63.08 77.92 -2.18 0.030

Soil depth (m) 0,17(0.01) 0,13(0.01) 65.42 75.58 -1.48 0.132

Stoniness (%) 62.46(2.17) 67.97(2.06) 77.16 63.84 -1.94 0.052

52

3.3.7. Comparison of allometric model of Boswellia papyrifera with mixed

species models

As far as our knowledge is concerned, there is no study of allometric equation specific to the

species Boswellia papyrifera in the literature to compare with our results. However, our dataset

allows fitting to a number of mixed species model that require DBH, H, and wood specific

density as input variables. Moreover, most allometric models in the literature are developed

based on data from tropical Asia and Latin America and the input from Africa is almost

negligible. Therefore, fitting our data from Africa will also give insight on whether the existing

mixed species models are valid to at least the dry land forests of Africa. Table 3.6 below

provides the mean above ground biomass predictions for the destructively sampled Boswellia

papyrifera trees based on the model developed in this study and other 11 mixed species models

from the literature. All the predictions are also compared against the real (observed) dry above

ground biomass of the 30 sample trees.

Except for two of the pan-tropic models of Chave et al. (2005) and the model of Ogawa et al.

(1965), all other models applied to our data overestimated AGB with average deviations from

29% to 171% (Table 3.6). Moreover, the mean of the observed data is significantly different (at

5% level) from each of the mean values predicted using these models in the literature. Among

the pan-tropical models of Chave et al. (2005), the models that the authors suggested as the

best models for dry tropical trees are the models in serial numbers 5 and 12 of Table 3.6. The

first model requires input data on diameter, wood specific density and tree height and the latter

is applicable if data on total tree height are not available. Fitting our data to these models

provided an average over prediction of 35.73% and 170.62% respectively and the mean of

observed data is significantly different from the mean of the predictions using each of these

models. Rather, two other models from Chave et al. (2005), indicated in serial number 2 and 3

in Table 3.6, and which are developed for all tropical forest types (dry, moist , or wet) fit to our

data relatively well and provided the lowest average deviation compared to all the other models.

We found no statistically significant difference between the mean of the observed and the

predicted mean proposed by our model for Boswellia papyrifera and the predicted mean values

using these two models. Our model fits best to the observed data and its prediction deviates

only by an average of less than 2%. Moreover, it is a simple model for it requires only the

measurement of the tree diameter as a predictor variable.

53

Table 3.6: Comparison of observed (real) aboveground biomass (AGB) data of the 30 destructively sampled Boswellia papyrifera trees with the predicted values of

AGB for the same trees using the model develope in this study and models published for mixed species of tropical forests

Model Mean 95% confidence

interval limits

Deviation in % Paired t-test

(Observed - Predicted)

Lower Upper Average Min Max t-value P

Model from this study

1. AGB = 0.061(DBH)2.353

78.23 59.05 97.41 1.91 -21.27 36.79 0.203 0.840

Models for mixed species tropical forests from literature

2. Chave et al. (2005): AGB = exp(-2.922 + 0.99ln(ρ(DBH)2H)) 83.48 62.25 104.70 10.79 -36.18 81.08 -0.996 0.328

3. Chave et al. (2005): AGB = exp(-2.994 + ln(ρ(DBH)2H)) 83.84 62.31 105.37 10.80 -36.02 81.41 -1.057 0.299

4. Ogawa et al. (1965): AGB = 0.043(ρ(DBH)2H)

0.95 82.96 62.24 103.68 13.83 -36.49 86.87 -0.720 0.477

5. Chave et al. (2005): AGB = exp(-2.843 + ln(ρ(DBH)2H)) 97.51 72.47 122.54 28.87 -25.60 110.98 -3.278 0.003

6. Chave et al. (2005): AGB = exp(-2.187 + 0.916ln(ρ(DBH)2H)) 99.07 75.70 122.44 35.73 -23.27 118.68 -3.937 0.001

7. Dawkins (1961): AGB = 0.0694(ρ(DBH)2H) 116.17 86.34 146.00 53.53 -11.36 151.36 -5.050 ***

8. Navar (2009) : AGB = 0.081(DBH)2.413

125.34 93.91 156.77 61.49 24.51 116.59 -7.414 ***

9. Brandeis et al. (2006): AGB = exp(-1.94371 + 0.84134ln(DBH)) 113.96 88.78 139.13 63.77 -11.22 159.09 -5.189 ***

10. Brown et al. (1989): AGB = 34.4703 – 8.0671DBH + 0.6589(DBH)2 155.65 113.53 197.77 91.17 31.92 160.73 -6.800 ***

11. Brown (1997): AGB = exp(-1.996 + 2.32ln(DBH)) 157.17 119.11 195.23 106.00 58.68 176.57 -8.3619 ***

12. Chave et al. (2005):

AGB=ρ(exp[-0.677+1.784ln(DBH)+ 0.207(ln(DBH))2 – 0.0281(ln(DBH))

3])

208.06 156.32 259.80 170.62 95.93 311.70 -8.1327 ***

Note: *** p < 0.001.

54

3.4. Discussion

3.4.1. Inclusion of dendrometric variables on goodness of fit of the allometric

model

Biomass allocation to the different tree organs differ among different biomes. A recent meta

analysis on this subject found that in tropical forests plants have the highest biomass allocation

(82%) in stem and the lowest (2%) in leaves (Poorter, 2011). Our study also indicated that much

(86%) of the biomass of the tree is accounted by branches and stem. The study indicates that

DBH explains most of the variation in aboveground biomass of Boswellia papyrifera trees. A

number of studies confirm that DBH is the best indicator of tree biomass (Ter-Mikaelian and

Korzukhin, 1997; Basuki et al., 2009). Precise height measurements are difficult to make and

are rarely available in large surveys for they are costly in terms of both time and resources.

Therefore, many authors suggest not to include height in allometric equations for biomass

estimation (Chave et al., 2005; Smith and Whelan, 2006; Noguerira et al., 2008) and even in

some studies where it is included it resulted only in a small improvement in the fit of the model

(Sampaio et al., 2010). Measuring crown area is even more difficult than measuring height, and

both variables are thus rarely used in allometric equations and their contribution to explain

variation tends to be small (Samba et al., 2001; Sampaio and Silva, 2005; Sampaio et al.,

2010). In this study, we found no statistically significant correlation between crown depth and

DBH (n = 30; r = 0.289; p=0.121) but a very significant correlation between crown area and

DBH (n = 30; r = 0.812; p < 0.001). However, the inclusion of these variables improves the fit of

the model only in the case of leaf biomass. Wood density is even more difficult to determine and

its contribution to allometric equations of single species is small because it varies little among

individuals of the same species (Sampaio et al., 2010). In our case, the range between the

maximum and minimum value of wood density for the species is less than twofold (0.44 g/cm3 to

0.77 g/cm3). The correlation between wood density and DBH is positive but not statistically

significant (n = 30; r = 0.186; p = 0.325) waheras wood density and tree height have significant

negative correlation (n = 30; r = -0.418; p = 0.022). We found that the inclusion of wood density

as a joint variable with DBH has resulted in a small improvement of the fit of the model, but only

in the case of stem biomass. However, in case of developing pooled-species allometric

equations, average wood density may be an important variable (Chave et al., 2005; Fehrmann

and Kleinn, 2006). This variable has been usually related to the scaling coefficient (a) of the

power equation (Ketterings et al., 2001; Pilli et al., 2006; Navar, 2009). However, a recent study

of Stegen et al., (2009) claims no evidence of paternal relationship between forest biomass and

wood density in tropical forests.

For stem, branches and above ground biomass as a whole, the power equations resulted in the

best fit to the data. Even in the case of leaf biomass for which we found generally lower values

of goodness of fit measures, a power functional form with DBHCA as a single independent

55

variable is the second best compared to the linear model with the product of DBH and H as a

single variable that relatively fits best. Most previous allometric studies for specific and mixed

species formulate the same conclusion (Chave et al., 2005; Pilli et al., 2006; Navar, 2009) and a

large body of theory on plant growth, structure and scaling supports the power equation (Niklas,

2004; Fehermann and Kleinn, 2006; Navar, 2009) for modelling dry biomass. Camel browses

the leaves of Boswellia payrifera trees and rural households lop and feed their livestock. This

impact of browsing by camels and lopping by humans might be a source of bias on the

measured leaf biomass of the destructively sampled trees and could be the reason for lowest fit

of the leaf biomass with tree variables (Table 3.3). Overall, leaf biomass is small when

compared to woody biomass (Table 3.4).

3.4.2. Carbon stock in Boswellia forests and implications to land use change

This study showed that fencing of Boswellia stands enhances the stock of soil organic carbon

and nutrients due to its effect of protecting litter falls and grass biomass from livestock grazing.

The soil organic carbon in the plots fenced for a period of two to three years was higher by

18.44 Mg ha-1 than the value in the unfenced plots, which are free for livestock grazing.

Mekuria et al. (2011) reported similar result in their comparison of exclosures of different ages

with communal grazing land in the same study region. They reported that a five-year-old

exclosure had 31.94 Mg ha-1 more soil organic carbon stock than a communal grazing land of

the same size. In terms of soil organic carbon and nutrient concentrations, fenced plots

exhibited higher concentrations particularly in the lower soil depth (20 to 40 cm) but the

differences were not statistically significant. There is an indication that fencing has positive

effect in enhancing soil organic carbon and soil nutrients and significant effects could be found if

longer periods of exclosure managements are practiced. According to the Booker Tropical Soil

Manual (Landon, 1991), the concentrations of both soil organic carbon and total nitrogen in the

study sites (Table 3.2) can be rated as medium but both values are very close to the lower

margins of this rating.

In the case of biomass carbon, the values for aboveground and total tree biomass calculated for

Boswellia papyrifera are in the range of aboveground tree biomass in African savannas, which

varies from 5.7 to 33.0 Mg ha-1 (House and Hall, 2001). A recent study on the valuation of land

restoration in Tigray estimated aboveground biomass and carbon stock of 17.8 Mg ha-1 and 8.9

Mg ha-1 for a 20 years old exclosure (Mekuria et al., 2011), which is very close to our result.

Using the recently updated database on the geographical distribution of woody biomass carbon

stocks in tropical forests and the Biome average method, Gibbs and Brown (2007) estimated

mean biomass carbon stock of tropical dry forests of sub-Saharan Africa at 17 Mg ha-1.

According to the IPCC Tier 1 global carbon biomass data, the biomass carbon stock in Africa’s

56

dry tropical forest for the year 2000 was 72 Mg ha-1 for broadleaf forests and 36 Mg ha-1 for

forest-cropland mosaic (Aaron and Gibbs, 2008).

3.5. Conclusions

The model AGB = 0.061(DBH) 2.353 is the most suitable allometric equation for Boswellia

papyrifera dry forests. The 95% confidence interval for predicted aboveground biomass using

this model overlaps with the confidence interval of the real biomass, indicating that the model is

sufficiently reliable for carbon accounting. This allometric equation can be used for estimating

carbon stocks in dry land forests, especially for tropical dry forests that are dominated by

Boswellia papyrifera and related species.

Tree diameter at 1.3m height is the single explanatory variable used in this model and it is easy

to measure and generally available in standard forest inventories. The inclusion of height, wood

density, crown depth, and crown area as a predictor either together with DBH or as a product of

DBH and each of these variables does not improve the goodness of fit of the model for AGB.

Comparison of the real aboveground biomass of the destructively sampled trees with the

predicted result from the allometric equation developed in this study for the species and

predictions based on other allometric equations in the literature developed for mixed species

tropical forests indicated that the model from the current study provided the best prediction.

While the average over prediction using the model of this study is only 1.91% than the real

biomass measured for the sample Boswellia papyrifera trees, the predictions using the

previously developed models for mixed species range from 6.53 to 164.37% higher than the

real biomass of the destructively sampled trees. This implies that site and species specific

equations provide less biased estimates than existing mixed species models used for predicting

carbon stock of tropical dry forests.

The study indicated that about 76 % of the carbon stock in Boswellia forests is accounted by

soil organic carbon with tree biomass accounting about 22%. Policies that aim to reduce

emissions from deforestation, therefore, need to focus on management interventions that can

enhance both soil carbon and at the same time avoid the degradation of the forest stand. From

the fencing experiment, the study indicated that fencing enhances soil organic carbon.

However, large area fencing is very costly but effective management in the form of exclosure

could bring similar result as evidenced by few studies in the highland parts of the study region.

Moreover, such management could assist natural regeneration and recruitment of Boswellia

papyrifera trees.

CHAPTER 4

Valuation of ecosystem services: a cost benefit analysis of forest

management options and REDD+ opportunity costs specific to

frankincense forests

4.1. Introduction

Forest ecosystems and the services they provide make significant direct and indirect

contributions to the global economy and human welfare. The benefits can be grouped into direct

and indirect use values, option values and non-use values and the sum of these values is

known as total economic value (Pearce and Tuner 1990; Barbier 1993; Kolstad 2000; Cambell

and Luckert, 2002). Valuation of forest ecosystem services has been recognized as an

important tool that can aid decision makers to evaluate trade-offs between alternative forest

management regimes and courses of social actions that change the use of forest ecosystems

and the services they provide (MEA, 2003). Although dry land forests provide a variety of

ecosystem goods and services there is inadequate documentation and evaluation of the

benefits they provide. This is mainly the case in Africa where these forests account for more

than 43% of the land and provide a number of ecosystem services in the day-to-day life of more

than 235 million people in the region (FAO, 2010a). Valuation of dry land forests could help with

land use planning, macroeconomic policy analysis, and for assessing the potential of REDD for

climate change mitigation.

Most forest valuation studies in developing countries mainly focus on the direct use value in

terms of wood and non-wood forest products (e.g. fuel wood, natural gums and resins,

medicinal plants, edible wild fruits, insect foods, honey, bush meat etc…) and their contribution

to rural income and livelihood (Reddy and Chakravarly, 1999; Cavendish 2000; Godoy et al.,

2002; Mamo et al. 2007; Tilahun et al. 2007). Some undertake comparison of alternative land

use changes (Tilahun et. al. 2007; Gunawardena and Rowan 2005; Hein and Gatzweiler 2006).

With the growing awareness of the non-marketed environmental benefits of forests and

developments in valuation methods for non-market benefits, there are studies that take into

account the non-market benefits of forest ecosystems in their valuation works (Ramirez et al.,

2002; Wise and Cacho, 2005, Gunawardena and Rowan, 2005; Hein and Gatzweiler, 2006;

Reichhuber and Requate, 2006; Balana et al., 2012).

Continuous land degradation and deforestation and the associated loss of biodiversity and

economically important tree species of the dry land forests in northern Ethiopia is one of the

serious problems aggravated by the extreme poverty in the region. The natural dry land forests

are highly fragmented and the risk of extinction of the rich biodiversity they hold is increasing

58

(Muys et al., 2006). Boswellia papyrifera is one of such species that is declining due to

overtapping for frankincense, cutting of leaves for fodder, free grazing by livestock that hinders

natural regeneration of the species, and conversion of the forest to agricultural land for crops

cultivation. Groenendijk et al. (2012) reported high adult mortality and lack of sapling recruitment

that is illustrated in the presence of clear gaps in the population structure of both tapped and

untapped Boswellia populations. Unless certain conservation measures are taken, the same

study projected a 90% decline in the size of tapped and untapped populations within half a

century and a 50% decline in frankincense production within fifteen years.

However, pure conservation of Boswellia forests excludes the forest from the conventional

competing uses that certainly entails an opportunity cost. Thus, to guarantee the optimal social,

environmental and economic use of this forest resource the present and future flows of benefits

and costs from conserving the resource need to be identified, valued and evaluated against

alternative uses. We identified six management options that range from conservation of the

forest with no extraction of provisioning services to the business as usual practice that involves

free grazing, leaf lopping for fodder, and heavy tapping of frankincense. Based on experiments

we identified the benefits of each forest management options. Next, a cost benefit analysis

(CBA) is applied to evaluate the net economic benefits from the different ecosystem services of

Boswellia forest under these management options and assess the opportunity cost in terms of

forgone benefits from not shifting to crops cultivation, which is the competing land use in the

study area.

4.2. Cost benefit analysis (CBA): conceptual framework

In most CBA of natural resources, valuing all the benefits and costs is a challenging task for the

fact that forests as a natural resource provide lots of goods and services for some of which

market prices do not exist. In the case where market prices exist for traded goods and services,

they may not reflect the true scarcity of the resource because of market imperfections. The

existence of monopolistic elements in the input and output markets, excessive state

interventions designed to control market failures, difference between border and domestic

prices of inputs and outputs due to tariffs and other trade policies are examples of market

imperfections that lead to distorted prices. CBA tries to overcome these shortcomings by finding

alternative prices called shadow prices. Shadow prices capture the real scarcity and hence

economic value of resources (Gittinger, 1982). For those goods and services for which market

prices do not exist, indirect valuation techniques like replacement cost methods, productivity

change methods, damage cost avoidance, and opportunity cost methods can be applied.

However, these values may not reflect the real economic value of the resource.

59

In CBA there is a need to discount the future flows of benefits and costs at a given discount rate

to get the present value. The discount rate should reflect the social time preference and the

relevant time horizon (project duration) should also be based on the perspective of the

stakeholders affected. Although the choice of the discount rate for CBA of environmental

projects remains controversial, there is a universal agreement among economist that a real

interest rate, which is the nominal interest rate adjusted for inflation, has to be used (Perman et

al., 2003). After making the CBA for alternative projects, the choice of the best resource

allocation option that maximizes social welfare has to be selected. The theoretical foundation of

CBA is welfare economics. In a society where markets are perfectly competitive market forces

will lead to a Pareto efficient (optimal) condition of resource allocation. According to the Pareto

optimal decision rule, a policy change and/or the introduction of new project is socially desirable

if everyone or at least some in the society are made better off without making anybody worse

off. However, in real world decision making a policy change and/or the introduction of a project

makes some better off and worse off others. The same principle could be applied in CBA of

forest management and utilization options. In such a case, there is a need for comparing the

increased gains by some with the amount of losses by the others. Thus, Pareto optimal criterion

is incomplete to assess all the possible welfare changes in a society due to changes in resource

allocation decisions (Johansson 1991). CBA utilizes an alternative decision rule with somewhat

less conceptual appeal, but much greater feasibility, than the Pareto efficiency rule. This rule is

known as the Kaldor-Hicks criterion. According to this criterion, a policy and/or project should be

adopted if and only if those who will gain could fully compensate those who will lose and still be

better off. This criterion provides the basis for the potential Pareto efficiency rule, or, more

commonly known as the net benefits criterion that states to choose only projects with positive

net benefits. If a project with positive net benefit is to be selected, there will be a possibility of

compensating the losers and the policy and/or project could potentially be Pareto improving

(Boardmann et al., 2001).

Finally, in order to incorporate risk and uncertainty due to changes in prices of inputs and

outputs, quantities of output levels, and the interest rate, a sensitivity analysis of the discounted

net benefit to changes in these variables has to be worked out to make the analysis complete

(Zerbe and Bellas, 2006). Moreover, further decision criteria like multi criteria analysis tools can

be used to incorporate project effects for which monetary values are either inexistence or

impossible (Perman et al., 2003).

4.3. Materials and methods

4.3.1. Selecting forest management options

In order to select the possible forest management options for the valuation study, we first

conducted an experiment on the effect of leaf lopping and tapping on frankincense yield,

60

flowering and fruiting (Chapter 2). Based on the result from the analysis of the leaf cutting and

tapping experiments in chapter 2, it is found that leaf lopping has significant negative effect on

frankincense yield, flowering and fruiting. However, there is no evidence of significant difference

between tapped and untapped stands in terms of both flowering and fruiting. Therefore, the first

six of the ten combinations of the 2 by 5 experiment (see section 2.2.2 of Chapter 2) are

selected for the economic valuation. These management options are: a) pure conservation

option in the form of exclosure that does not allow both extraction of frankincense, leaf lopping,

and free grazing (CONS); b) exclosure with normal tapping (tapping at 6 spots per tree) but

prohibits both leaf lopping and free grazing (EXCLOSUR1); c) exclosure with heavy tapping

(tapping at 12 spots per tree) but no leaf lopping and no free grazing (EXCLOSUR2); d) no

tapping with leaf lopping for fodder and free grazing (FGRAZ1); e) normal tapping with leaf

lopping for fodder and free grazing (FGRAZ2); f) heavy tapping with leaf lopping for fodder and

free grazing (FGRAZING3).

4.3.2. Determining physical quantities of the ecosystem services

The ecosystem services of Boswellia papyrifera forest considered in this analysis are

provisioning (frankincense, grass and leaf as fodder), regulating (avoided emission from

standing biomass carbon and soil organic carbon stocks, and carbon sequestration in soils),

and supporting services (avoided loss of soil nitrogen and available phosphorous, and

sequestration of nitrogen and phosphorous in soils). Our analysis does not take into account a

number of other services of the forest. As such, our valuation of the benefits is a lower boundary

of all the benefits. For example, in case of frankincense we take only the value as a raw

material. However, frankincense is used as input in pharmaceutical, cosmetic, food, and

chemical industries in the importing countries. It has also cultural and religious values in that it is

used as a fragrance for clerical services in the Catholic and Orthodox Christian churches and

other religions. In Ethiopia, it is a tradition to use frankincense during coffee ceremonies.

Moreover, the flowering period of Boswellia paprifera is during the dry season and serves as

important bee forage and the forest serves as habitat for wildlife.

Frankincense: To determine the frankincense yield, a tapping experiment was conducted for

two production seasons on 42 sample trees per each forest management with tapping options.

The first season was from October 2009-June 2010 and the second was from October 2010-

June 2011. The tapping was from October to June in which sample trees were tapped for 15

rounds and frankincense collection was carried out for 14 rounds. Frankincense yield per tree

was determined by weighing the collected resin, which is locally termed as “grezo”, directly after

collection using a digital weight balance with a precision of 0.01g. After collection it was air-dried

and graded into seven grades (Figure 1.1) based on purity, color, and size at the Ethiopian

Natural Gum Marketing Enterprise. The yield per ha for each quality grade frankincense was

61

determined by multiplying the tree level mean value by the mean number of trees per ha for the

study site. Accordingly, we determined yields per ha for each forest management options. The

outputs for the management options (CONS and FGRAZ1) are zero because both management

scenarios prohibit tapping throughout the project duration.

Wood, leaf, and grass biomass: The methods used for determining the wood, leaf and grass

biomass in the fenced and unfenced plots are discussed in section 3.2.1 and 3.2.2 of chapter 3.

For the fact that conversion of the forest to cropland in the study area is carried out through

shifting cultivation and the cleared wood biomass is used for fuel wood, we assumed 100% of

the biomass carbon as emission reduction benefit in the case of maintaining the forest.

Therefore, the same mean biomass carbon stock is considered for all the forest management

option. This is because first, the Mann-Whitney Test in table 3.4 of chapter 3 indicated that the

differences in biomass carbon stock in standing trees of fenced and unfenced experimental

plots were not statistically significant (z = -0.83, p = 0.406). Secondly, the differences are not

due to our experiments of leaf lopping and tapping as well as fencing.

Soil organic carbon, total nitrogen, and available phosphorous: The data collection and

results of the lab analysis for soil organic carbon, total nitrogen and available phosphorous are

discussed in sections 2.1.1 of chapter 2 and 3.2.3 of chapter 3. According to the result of Mann-

Whitney Test in table 3.5 of chapter 3, there are statistically significant differences between the

fenced and unfenced experimental plots in terms of stocks of soil organic carbon (z =-2.39, p =

0.017), total nitrogen (z = -2.86, p=0.004), and available phosphorus (z = -2.27, p = 0.023).

Thus, for the forest management options CONS, EXCLOSUR1, and EXCLOSUR2 we take the

mean values of carbon, nitrogen and phosphorous stocks in fenced experimental plots whereas

for the other three forest management options that allow free grazing and leaf lopping for

fodder, we considered the mean values of the unfenced experimental plots. The differences in

stocks of this regulating ecosystem service between the fenced and unfenced experimental

plots are considered as the amount of sequestrations in the soil due to excluding free grazing

and lopping of leaves, which are the elements of the forest management options CONS,

EXCLOSUR1, and EXCLOSUR2. For the other three management option, zero sequestration of

carbon, nitrogen and phosphorous are considered. This is to avoid double counting and for the

fact that the grass and leaf biomass are considered as benefits in these management

scenarios. Because if grass and leaf biomass are removed through free grazing, there will be

less or no organic matter left for to be decomposed and increase the level of soil organic

carbon.

In order to take into account the benefit of emission reduction due to maintaining the forest, we

assumed 100% of the standing biomass carbon and 25% of the soil organic carbon in the forest

62

as emission reduction benefits. This is consistent with literature on loss of soil carbon after

conversion of forest to herbaceous croplands (Guo and Gifford, 2002; Murty et al., 2002;

Houghton and Goodale, 2004; West et al., 2010). For converting these benefits to temporary

certified emission reduction units (tCER) which are measured in terms of tCO2, we multiplied the

carbon stock by the molar conversion factor of 3.67. To determine the avoided loss in soil

nitrogen due to not converting the forest to cropland, we applied 15% following the meta-

analyses of Murty et al. (2002).

4.3.3. Determining opportunity cost: crops from shifting cultivation

A farm household survey that integrates contingent valuation survey design with structured

questionnaire related to socio-economic variables was conducted in the study area in March

2010.

The household survey integrates contingent valuation with socio-economic survey designs to

generate data for this chapter and chapters 5 and 6. The survey was structured in two sections

in which the part of the contingent valuation questionnaire was followed by the questionnaire for

collecting the socio-economic data (Appendix 4A). The local people were not benefiting much

from the resource because of lack of the skill of tapping frankincense. Mostly migrant workers

from the highlands of Tigray have been practicing the tapping of frankincense. However, in the

last few years, some of the local people have started to generate income from frankincense, as

they became member of local frankincense cooperatives. Therefore, the five villages for the

survey were selected based on the location of the plot level experimental study sites with the

objective of integrating biophysical aspect of the resource management with social and

economic dimensions. Three of the sample villages were the villages where we established plot

level experimental study and the other two villages from the neighbouring districts were selected

based on proximity to the experimental plots and availability of frankincense rural cooperative

firms. The survey was conducted in March 2010 on 520 sample households of which 120 were

frankincense cooperative members and 400 were non-members. Further details on the

sampling procedure are given in Chapter 5.

Relevant to this chapter, the survey was designed to collect data that can be used to determine

the net benefits per ha from crop production in the study area. In the survey, respondents were

asked about the type of crop they cultivated in each of their farm plots, the size of each plot, the

input quantities used, and the crop output harvested in the crop year 2008/09 (Appendix 4A).

We determined the weighted average opportunity cost of the land from crop production using

the proportion of area of land cultivated by each crop category as a weight. To determine the

dry mass of crop residue we multiplied the crop output by the residue to crop ratio from scientific

literature (Keftasa, 1988; Kim and Dale, 2004). The ratios are 1.3 for sorghum, 1.2 for sesame,

1.0 for maize, and 3.4 for teff, and 1.2 for barely. The value 1.2 is used for calculating the crop

63

residue for the other single crop category in this study. For the other categories, we applied the

ratio for sorghum.

4.3.4. Valuation of benefits and costs

In order to understand the supply chain from harvest to export of frankincense and facilitate the

valuation exercise, interviews were conducted in March 2009 with three companies engaged in

the business. However, data for the base year (2009/10) on the harvesting cost of grezo, export

prices for each grades of frankincense, and the costs of grading and trading each quality grade

was obtained from Guna Trading Private Limited Share Company, which is one of the three

interviewed companies.

Market prices of oxen rent for farming activities, wage for daily labour hired for farming activities,

prices of other agricultural inputs used (pesticides, herbicides, fertilizer, and seeds), prices of

crop outputs, and prices of fuel wood and grass were obtained through the household survey.

The opportunity cost of family labour used in farming activities was assumed zero due to the

following reasons. First, the main economic activity of the farm households in the study area is

mixed farming and there are very limited off-farm employment opportunities. Second,

determining the opportunity cost of family labour for each farm household requires detail

information about the off-farm employment opportunities that each member of the family in the

working age group actually missed due to working on the family farm. The land use fee paid by

farmers was determined based on the ratio of government revenue from rural land use fee (239

million ETB) to total cultivated cropland (8.77 million ha) for the fiscal year 2008/09 (CSA, 2011).

Maintaining the land as a forest entails opportunity cost in terms of wood consumption for fuel

whereas converting the land to agriculture implies that the wood biomass will be used for fuel.

Therefore, to value the dry wood biomass we used the local market price of fuel wood. The

price of grass is used to value grass, leaf biomasses, and crop residues. Labour cost of guards

to protect the forest from leaf lopping, free grazing and other human interference was based on

the wage paid for guards for exclosure management in Tigray (Mekuria et al., 2011).

The conference of the Parties (COP9) of the Kyoto Protocol decided on the accounting rules for

non-permanent Certified Emission Reductions (CER in tCO2) of reforestation and afforestation

projects and provided a framework for establishing markets for carbon credits. CER are

certificates of green house gas emission reductions obtained from project activities in

developing countries and include permanent reduction through emission avoidance as well as

non-permanent reduction by forestry projects (Olschewski and Benitez, 2005). The CER units,

which have to be verified by independent entity, are expected to be traded like any other

commodity (Olschewski and Benitez, 2010). Long-term and temporary credits are the two

possible ways to account for non-permanent emission reductions in forests during the first

64

commitment period from 2008 to 2012 (UNFCCC, 2003). A temporary credit (tCER) is a CER

issued for an afforestation or reforestation project activity under the clean development

mechanism (CDM) and expires at the end of the commitment period following the one during

which it was issued but can be reissued several times during the project as long as the forest

exists (Olschwski and Benitez, 2010). Contrary, a long-term credit (lCER) expires at the end of

the crediting period of the overall project (Olschewski and Benitez, 2005), and cannot be

reissue (Neef and Henders, 2007; Olschewski and Benitez, 2010). Both long term and

temporary credits are measured in tons of carbon dioxide. For determining benefits from

emission reduction, we transformed price per tCO2 of permanent CER into price for tCER in

accordance with Olschewski et al. (2005):

,

where: p5 is the price of tCER, p∞ is price of permanent credit and d is the discount rate.

Assuming tCER expiring time of 5 years, and average price of $ 25 per permanent credit and a

discount rate for Annex I countries of 3 per cent, results in a price of $ 3.43 per tCER. Mekuria

et al. (2010) used the same price for valuing the emission reduction benefits of exclosures in

Ethiopia.

Following the decision of The Conference of the Parties (COP13) on considering REDD

(reducing emissions from deforestation and forest degradation in developing countries) as part

of a post-2012 global climate regime; efforts are taking place on developing the concept and

creating crediting mechanism. At the COP15, the idea of REDD+ came into picture with the plus

referring additional objectives like biodiversity conservation, poverty reduction, sustainable

forest management and enhancement of forest carbon stocks. The UNFCCC and several

national and state governments are currently considering the development of REDD+ crediting

mechanism that would reward REDD+ efforts in tropical countries with issuance of

emission/sequestration credits that could be traded in carbon markets (IETA, 2012). REDD+

entail opportunity costs to the forest country in the form of mainly forgone economic benefits of

alternative land use and to some extent social and cultural costs that are not easily measured in

economic terms (White et al., 2011). In this study, we calculated the economic opportunity costs

(per tCO2) forgone from shifting cultivation as alternative land use.

4.3.5. Data analysis and decision criteria

We used the net present value (NPV) decision criterion to evaluate the economic profitability of

the alternative forest management options. NPV sums up the discounted annual flows of net

benefits, which in turn is the difference of discounted benefits and discounted costs, over the life

of the project. The NPV of a project is the amount by which it increases net worth in present

value terms. Therefore, the decision rule is to accept a project with non-negative NPV and reject

otherwise (Perman et al., 2003):

65

where:

NPV is Net Present Value (US$ ha-1)

Bt is Benefit at time t (US$ ha-1)

Ct is Cost at time t ($ ha-1)

r is real discount rate

t is time in years (t = 0, 1, 2, …T)

j is the forest management option (j = 1, 2, 3…6)

The real discount rate is calculated from the nominal interest rate, i, and the expected inflation

rate π (Fisher, 1930) as:

.

Current consumer price and/or general price indices are often used as an estimate of future

inflation. However, these indices reflect the general development of all prices, which might

either over estimate or underestimate the future price development of the specific project

outputs. Therefore, it would be better to use inflation forecasts if available. Therefore, based on

EconomyWatch.com’s Econ Stats database, Ethiopia’s inflation forecasts for the years 2012-

2016 are available. Since the base year for our evaluation is 2010, in addition to the five year

forecasts we used the 2011 inflation rate and calculated a geometric mean inflation rate of

9.914% for the six years data (the inflation rates are 12.924 for 2011, 11.199% for 2012, and

9% for each of the years 2013-2016). According to the Central Bank of Ethiopia, the maximum

lending rate for the base year is 16.50%. Thus using these rates and equation 6, we calculated

a real interest rate of 5.99% and used for discounting.

The project duration over which the economic analysis has to be carried out is another

important parameter that has to be chosen. For lack of data on the productive life of Boswellia

paprifera tree, we determined the project duration based on the common tapping schedule of

five years, which is practiced in the study area and a four year resting period between tapping

cycles recommended to heal the wound (Rijkers et al., 2006). As shifting cultivation is the

alternative land use, we also considered three years of crop cultivation with three years fallow

based on the data from the household survey. According to the farm household respondents,

fallowing is practiced for a period of 1 to 6 years and about 98% of the respondents fallow their

land for a period of 1 to 3 years. Moreover, about 77% of the farm households in the survey

cultivate their land for a period of 1 to 3 years before and after fallow. Therefore, we used 3

years of cultivation followed by another 3 years of fallow in accounting the costs and benefits of

crop production as an alternative land use. Accordingly, a period of 30 years was taken as the

project duration which comprises 3 full cycles (1 cycle = 5 years tapping + 4 years resting) of

66

frankincense production with one incomplete cycle of 3 years harvesting. In the case of crop

production the 30 years project duration implies 5 full production cycles (1 cycle = 3 years

cultivation + 3 years fallow). The selected project duration is also in accordance with the project

duration given by official carbon accounting rules (UNFCCC, 2003) and related studies

(Olschewski and Benitez, 2005; Mekuria et al., 2010). Opportunity cost of reducing CO2

emission from deforestation and forest degradation (REDD+ opportunity cost) was calculated

for each forest management option through dividing the corresponding present value of the

opportunity cost by the total CO2 emission reductions.

After the CBA, the six forest management options were assessed using a multi-criteria analysis

based on five categories of criteria, which are opportunity cost per tCO2 emission reduction,

carbon enhancement, nutrient storage, biodiversity conservation, and income distribution to the

poor. For the multi-criteria analysis, we followed the steps described in Perman et al. (2003).

The first step involves that the best outcome of a criterion was set equal to 1 and each of the

other outcomes of the same criterion is converted to dimensionless form by expressing the

criterion outcome for each option as a ratio to the best outcome which is set equal to 1. In the

case where the best outcome is a lower value, like the case of cost in which the best has to be

the minimum one, then the reciprocal of the ratio of each outcome to the value of the best

outcome was taken. For example let if we assume three forest management options, say A, B,

and C, which result in the CO2 emission reduction at a cost of 500, 1000, and 2000 per unit

area, the management with minimum cost is A and it is the best option and has to be set at 1 =

(500/500). Then for the other options the reciprocal of the ratio of each outcome to the value of

the best outcome can be determined as (1000/500)-1 = 0.5 for B and (2000/500)-1 = 0.25 for C.

The next step is calculating the scores by multiplying the results in step 1 by weights to each

criterion that reflect the preferences of the decision makers. For the fact that these criteria were

not presented to decision makers for weighting, we assumed first equal weights to all the criteria

and then applied different weights to see the effect on the result. The final step involves

aggregating in step 2 across all criteria to reach at the sum of scores used for ranking.

Finally, sensitivity analysis was performed taking into account three areas of changes, which

include changes in quantities of cost and benefit items, changes in prices of cost and benefit

item, and changes in the real discount rate, and investigated the sensitivity of the base case

NPVs to these changes. We selected 15% and 40% changes in prices based on historical data

on price changes over two selected years (2002/03 and 2009/10). We used the same values for

changes in output quantities and the real discount rate. We selected year 2002/3 because we

have data on all prices relevant to this study, particularly for frankincense out and input in all the

supply chain. The year 2009/10 was selected just because it is the base year for the current

analysis. Accordingly, we found average annual increase of 13.45% for inputs of frankincense

67

processing and trading, an increase in 23.29% in prices of agricultural inputs, a rise of 29.38%

in prices of agricultural outputs, and a rise of 37.56% in prices of grezo and graded

frankincense.

4.4. Results

4.4.1. Estimated quantities of Boswellia papyrifera forest ecosystem services

Table 4.1 summarizes the physical quantities of the ecosystem services analyzed in this

valuation. The provisioning services from Boswellia forest under the six management options

considered in the analysis are frankincense and biomasses of grass and leaf. In the case of

frankincense, out of the other four management options that allow tapping at different tapping

intensities, EXCLOSUR2 has the highest yield per tree in terms of both ungraded frankincense

(locally called grezo) and all of the quality grades except G1B. In terms of this grade, the forest

management option EXCLOSUR1 is with the highest yield. Using the mean number of trees per

ha for the study site, which was 202.32 with standard error of 7.52, the calculated grezo

frankincense yield (Kg ha-1) are 96.30 for EXCLOSUR1, 92.66 for FGRAZ1, 147.49 for

EXCLOSUR2, and 112.69 for FGRAZ2 forest management options. In the case of grass and

leaf biomasses, the values for the first three management options were taken zero with the

assumptions that these management options (CONS, EXLOSURE1 and EXCLOSURE2)

prohibit free grazing and leaf lopping. For each of the other three options (FGRA1, FGRAZ2,

and FGRAZ3), the sum of grass and leaf biomass was 2.31 Mg-1 ha-1 yr-1.

The regulating services from the forest considered in the analysis include the stock of biomass

carbon in Boswellia trees, which includes the biomass in all parts of the tree except for the leaf

biomass, the sock of soil carbon, and the annual amount of carbon sequestered in the soil. The

carbon stocks in biomass and its equivalent tCER units used for the valuation indicate that all

the management options have equal value of tCER units (34.80 Mg CO2 ha-1) from avoided

biomass carbon emission that can be realized from maintaining the frankincense forest from

conversion to other land use (shifting cultivation). In terms of soil carbon stock and the

equivalent tCER unit from avoided soil carbon emission due to maintaining the forest from

conversion to agricultural land differs among the first group of the management options (CONS,

EXCLOSURE1, and EXCLOSURE2) and the other group that includes FGRAZ1 FGRAZ2, and

FGRAZ3. In the case of carbon sequestration in soils due to decomposition of litter fall and

grass biomass, the first group of the management options (CONS, EXCLOSURE1, and

EXCLOSURE2) have equal values whereas the values for the other group (FGRAZ1, FGRAZ2

and FGRAZ3) are all zero for the fact that the litter fall (leaf biomass) and grass were

considered in the part of the provisioning services of the latter group. Thus, the sum of carbon

emission reduction and carbon sequestration benefits per ha was 141.70 tCER units of CO2 for

68

each of the management options in the first group whereas the corresponding value for each of

the other group was 57.18 tCER units of CO2.

Table 4.1: Mean (SE) values of Boswellia forest ecosystem services by management options

Ecosystem services Management

CONS EXCLOSUR1 EXCLOSURE2 FGRAZ1 FGRAZ2 FGRAZ3

Frankincense (Kg tree-1

yr-1. n*=42)

Grezo 476(69.46) 729(84.92) 458(52.52) 557(76.87)

G1A 40.95(9.73) 49.00(8.49) 25.42(4.76) 48.42(10.01)

G1B 56.84(13.97) 41.34(4.58) 42.34(8.69) 41.12(6.21)

G2 16.42(2.66) 21.11(3.43) 9.36(1.61) 22.46(5.41)

G3 10.00(1.58) 17.89(2.49) 9.72(1.66) 17.21(3.10)

G4s 139(20.50) 221(30.75) 129(16.22) 163(22.44)

G4s 152(33.24) 263(36.42) 167(27.49) 177(33.26)

G5 59.83(8.60) 115(18.65) 74.68(14.83) 88.09(12.95)

Grass (Mg ha-1 yr

-1. n=70) 1.60(0.28) 1.60(0.28) 1.60(0.28)

Leaf (Mg ha-1 yr

-1. n=140) 0.71(0.04) 0.71(0.04) 0.71(0.04)

Biomass carbon (Mg ha-1. n=140) 9.49

(0.74)

9.49

(0.74)

9.49

(0.74)

9.49

(0.74)

9.49

(0.74)

9.49

(0.74)

tCER from avoided biomass carbon

emission (Mg CO2 ha-1)

34.80 34.80 34.80 34.80 34.80 34.80

Soil carbon stock

(Mg ha-1. n=70)

42.85

(5.06)

42.85

(5.06)

42.85

(5.06)

24.41

(2.66)

24.41

(2.66)

24.41

(2.66)

tCER from avoided soil carbon

emission(Mg CO2 ha-1)

39.28 39.28 39.28 22.38 22.38 22.38

Sequestered carbon in soil

(Mg ha-1 yr

-1. n=70)

18.44

(5.72)

18.44

(5.72)

18.44

(5.72)

tCER from soil carbon

sequestration(Mg CO2 ha-1)

67.62 67.62 67.62

Total Nitrogen

(Mg ha-1. n=70)

1.84

(0.16)

1.84

(0.16)

1.84

(0.16)

1.18

(0.13)

1.18

(0.13)

1.18

(0.13)

Nitrogen loss avoided

(Mg ha-1. n=70)

0.28 0.28 0.28 0.18 0.18 0.18

Sequestered nitrogen in soils

(Mg ha-1 yr

-1. n=70)

0.65

(0.21)

0.65

(0.21)

0.65

(0.21)

Available phosphorous

(Mg ha-1. n=70)

0.012

(0.001)

0.012

(0.001)

0.012

(0.001)

0.008

(0.001)

0.008

(0.001)

0.008

(0.001)

Phosphorous loss avoided

(Mg ha-1. n=70)

0.0019 0.0019 0.0019 0.0013 0.0013 0.0013

Sequestered phosphorous in soils

(Mg ha-1 yr

-1. n=70)

0.004

(0.002)

0.004

(0.002)

0.004

(0.002)

Note: n* refers number of sample trees, n refers number of sample subplots, Grezo is the dried resin as collected from the forest;

G1 to G5 refer to Grades 1 to 5. Grades 1 and 4 have sub-categories based on the color of the frankincense; G4-S stands for Grade

4 Special. Grade 5 contains mostly bark of the tree and with small amount of fine resin granules.

The only regulating services of Boswellia papyrifera forests considered in the analysis are the

benefits in terms of avoiding loss of soil nutrients (total nitrogen and available phosphorous) and

69

benefits in terms of increase in the nutrient contents due to sequestration in soils that can be

realized through keeping the forest from conversion to crop land. The sum of avoided nitrogen

loss and sequestered nitrogen in soils of Boswellia papyrifera forests under the first group of

management was about 490 Kg ha-1 whereas it is only about 180 Kg ha-1 for the case each of

the other group. In terms of available phosphorous, the corresponding benefits were about 4 kg

ha-1 and 1.3Kg ha-1 respectively.

The data in Tables 4.2 and 4.3 on input and output prices were used for the valuation of the

corresponding ecosystem services. Table 4.2 summarizes the output and input prices for

frankincense production and trading. The data is based on information from Guna Trading

Company. Table 4.3 below summarizes the number of plots cultivated with different crop

categories by the survey households and the corresponding output quantity and price as well as

input quantities and input costs used in the valuation. All prices and costs are given in USD after

converting the values in local currency to USD at the annual average exchange rate (1USD =

ETB 13.3778) for the fiscal year 2009/10. According to the data from the household survey

(Appendix 4A), the 520 sample households had 1021 cultivated plots with a total area of 1193

ha of which sesame cultivation accounted about 22% of the cultivated area, sorghum accounted

21.4% and the two crops cultivated in a single plot accounted another 25.7% of the cultivated

areas. Therefore, the two crops in total accounted 69.08% of the cultivated croplands. The other

crops include teff covering 8%, maize with 4.22%, two crops other than sorghum and sesame

accounting 8.52%, and three or more crops per plot accounting 7.90% of the total cultivated

area of cropland. Accordingly, three crops cultivation per plot was the most productive followed

by sorghum and sesame cultivation on single plot and then other two crop types other than

sorghum-sesame combination cultivated on single plot. The crops cultivated on single plot were

maize, sorghum, sesame, teff, and other crops that include barely, haricot bean, chickpea, and

horse bean.

70

Table 4.2: Market prices of outputs and inputs for frankincense production and trading

Output Price(USD/100Kg) Input Price(USD/100Kg)

Frankincense Grezo frankincense harvesting 51.80

G1A 397.67 Grading and packaging (G1A to G5)

G1B 248.55 G1A 14.20

G2 298.26 G1B 13.08

G3 268.43 G2 12.33

G4s 198.84 G3 13.83

G4 178.95 G4s 11.21

G5 112.13 G4 9.34

*Grass 2.45 G5 7.48

Fuel wood 2.92 Transport, port and transiting cost 15.17

tCER in tCO2e 3.43 Overhead expenses 4.19

Nitrogen 47.84 Payment for guards (USD ha-1

yr-1

) 31.40

Phosphorous 56.81

Note:* the price of grass was used to calculate the value of leaf fodder as well as crop residues. Data for

the frankincense output and input prices are based on Guna Trading Private Limited Share Company.

Grezo is the dried resin as collected from the forest; G1 to G5 refer to Grades 1 to 5. Grades 1 and 4 have sub-

categories based on the color of the frankincense; G4-S stands for Grade 4 Special. Grade 5 contains mostly bark of

the tree and with small amount of fine resin granules.

71

Table 4.3: Mean areas of cultivated plots by sample farm households, quantities of outputs and inputs and prices of crop outputs and input costs

Crop types cultivated on a single plot

Cultivated plots by sample households

Output Input (**cost is in USD/ha)

N Area in ha

kg/ha Crop Price USD/100kg

*Crop residue

Seed Herbicides and pesticides

Hired labour Oxen rent Land use fee

Mean (SE)

Mean (SE)

Mean (SE)

Kg/ha Kg/ha Mean (SE)

Cost

Litter/ha Mean (SE)

cost No./ha Mean (SE)

Cost

No./ha Mean (SE)

Cost Cost

Sorghum 274 0.93 (0.04)

487.39 (30.71)

35.28 (1.30)

633.61 34.09 (1.86)

17.84 0.87 (0.08)

4.14 0.00 0.00 7.26 (4.45)

47.29 2.04

Sesame 186 1.41 (0.09)

463.74 (38.84)

112.13 (1.96)

556.49 30.24 (1.21)

35.33 0.82 (0.07)

3.90 9.01 (1.32)

63.97 12.00 (5.79)

78.18 2.04

Teff 126 0.76 (0.04)

234.56 (15.68)

59.80 (1.18)

797.50 36.95 (2.28)

22.84 0.61 (0.06)

2.90 0.00 0.00 0.00 0.00 2.04

Maize 75 0.67 (0.05)

497.60 (52.99)

33.64 (0.77)

497.60 36.00 (2.96)

12.11 0.45 (0.08)

2.18 0.00 0.00 0.00 0.00 2.04

Other single crop

42 0.65 (0.08)

243.25 (26.65)

45.97 (3.49)

291.90 22.95 (4.57)

10.96 1.12 (0.12)

5.28 0.00 0.00 0.00 0.00 2.04

Sorghum and sesame

169 1.81 (0.09)

603.55 (34.72)

62.08 (1.21)

784.62 28.91 (1.57)

18.67 1.96 (0.07)

9.03 13.04 (2.69)

92.63 5.56 (4.39)

36.22 2.04

Other two crops

103 0.99 (0.06)

511.58 (48.52)

50.65 (2.96)

665.06 22.99 (2.22)

11.64 0.97 (0.24)

4.72 0.00 0.00 4.00 (0.21)

26.06 2.04

Three and more crops

46 2.05 (0.16)

717.96 (97.71)

73.28 (9.31)

933.35 48.38 (2.24)

36.83 1.16 (0.13)

5.49 10.79 (2.47)

76.60 4.00 (0.34)

26.06 2.04

***Wood biomass

140 1695 2.45

*The values were derived by multiplying the crop output per ha by the residue to crop ratio. according to (Keftasa, 1988; Kim and Dale, 2004)(residue to crop ratios are according to

(Keftasa, 1988; Kim and Dale, 2004) ; 1.3 for sorghum, 1.2 for sesame, 1.0 for maize, and 3.4 for teff, and 1.2 for barely (Keftasa, 1988; Kim and Dale, 2004) and the value for barley

is used for calculating the crop residue for the other single crop category in this study and for the other categories we applied the ratio for sorghum. ** refers cost calculated as a

product of price and quantity and hence unit prices can be derived as a ratio of the cost to mean value of an input quantity; *** refers the above ground wood biomass excluding leaf

that can be used as fuel wood in conversion of the forest to crop land.

72

4.4.2. Base case NPV of forest management options

The present values of benefits, direct costs, opportunity costs, and the NPV that take into

account different opportunity costs of land are presented in Table 4.4. The weighted average

NPVs for all forest management options except FGRAZ3 are negative. Moreover, in the case of

high opportunity costs of forestland (Sesame, sesame with sorghum, other two crops per plot,

and cultivation of three crops per plot) the NPV of Boswellia papyrifera forest under any of the

management options is negative. These NPVs of maintaining the Boswellia paprifera forest for

30 years under the exclosure and free grazing system vary from -97 to -4490 $ ha-1 depending

on the type of crop taken into account as opportunity cost. Among the management options with

negative NPVs, the CONS option under exclosure system costs the highest whereas

EXCLOSUR2 costs the lowest. If relatively less opportunity costs of land are considered (Maize,

Teff, and Sorghum cultivation), maintaining the forest provides positive NPV only under the

FGRAZ1 and FGRAZ2 options. These positive NPVs vary from $105 ha-1 with maize production

as opportunity cost to $ 659 ha-1 with forgone Sorghum cultivation as opportunity cost of

forestland.

For each forest management option, comparing the PV of the total benefits as well as each the

PV of benefits from each ecosystem with the PV of costs give some indication of the relative

importance of each benefit item in the overall weighted average NPV. First, in the case of the

CONS option, the PV of total benefits, which is mainly in form revenue from tCERs and value of

avoided loss and sequestered total nitrogen in soils, covers only 41.97% of the PV of total

weighted average costs. There are no direct benefits in the form of either revenue from

frankincense, grass, or leaf fodder. Therefore, if the indirect benefits are deducted, the negative

NPVs will increase in absolute value and be exactly equal to the PV of the total cost. Second, in

the case of EXCLOSUR1, the PV of total benefits can cover 84.96% of the PV of total weighted

average cost. Under this management option the PV of benefits from frankincense alone, which

is the only direct benefit, can compensate 41.53% of the PV of total weighted average costs

whereas the total indirect benefits account 43.43%. Third, in the case of EXCLOSURE2, the PV

of total benefits can compensate 88.22% of the PV of total weighted average costs and the

direct benefit from frankincense alone accounts 51.04% whereas indirect benefits account

37.18%.

73

Table 4.4: NPV of Boswellia papyrifera forest management options

Ecosystem services CONS EXCLOSUR1 EXCLOSURE2 FGRAZ1 FGRAZ2 FGRAZ3

PV of Benefits

CO2 emission reduction 1104.74 1104.74 1104.74 577.52 577.52 577.52

N-Replacement cost benefit 861.95 861.95 861.95 63.31 63.31 63.31

P-Replacement cost benefit 1.35 1.35 1.35 0.52 0.52 0.52

Frankincense benefits 0.00 1881.92 2701.79 0.00 1532.46 2080.01

Grass biomass 0.00 0.00 0.00 573.79 573.79 573.79

Leaf biomass for fodder 0.00 0.00 0.00 255.57 255.57 255.57

Total 1968.04 3849.96 4669.83 1470.71 3003.17 3550.72

PV of direct costs

Frankincense harvesting

and trading

0.00 1419.96 2181.82 0.00 1218.06 1595.18

Payment for guards 458.45 458.45 458.45 0.00 0.00 0.00

PV of Opportunity costs

Grass fodder 573.79 573.79 573.79 0.00 0.00 0.00

Leaf fodder 255.57 255.57 255.57 0.00 0.00 0.00

Frankincense 1578.21 0.00 0.00 1578.21 0.00 0.00

Sorghum 807.38 807.38 807.38 807.38 807.38 807.38

Sesame 2719.36 2719.36 2719.36 2719.36 2719.36 2719.36

Teff 867.26 867.26 867.26 867.26 867.26 867.26

Maize 1190.70 1190.70 1190.70 1190.70 1190.70 1190.70

Other single crop 635.53 635.53 635.53 635.53 635.53 635.53

Sorghum and sesame 1755.30 1755.30 1755.30 1755.30 1755.30 1755.30

Other two crops 1658.42 1658.42 1658.42 1658.42 1658.42 1658.42

Three and more crops 3102.38 3102.38 3102.38 3102.38 3102.38 3102.38

Weighted average crop 1333,97 1333,97 1333,97 1333,97 1333,97 1333,97

Fuel wood 489.62 489.62 489.62 489.62 489.62 489.62

Total 4689,61 4531,36 5293,22 3401,8 3041,65 3418,77

NPV at opportunity cost of

not converting land to

cultivation of:

Sorghum -2195 -155 -97 -1404 488 659

Sesame -4107 -2067 -2009 -3316 -1424 -1253

Teff -2255 -215 -157 -1464 428 599

Maize -2578 -538 -480 -1788 105 275

Other single crop -2023 17 75 -1233 660 830

Sorghum and sesame -3143 -1103 -1045 -2352 -460 -289

Other two crops -3046 -1006 -948 -2255 -363 -192

Three and more crops -4490 -2450 -2392 -3699 -1807 -1636

Weighted average -2722 -681 -623 -1931 -38 132

The PV of total benefits from FGRAZ1 can cover only 43.23% of the PV of total weighted

average costs with the direct benefits from grass and fodder accounting 24.38%. Of the total

benefits, the largest share is from grass followed by revenue from tCER, whereas lowest share

74

is by the benefits in the form of avoided loss and sequestered phosphorous. The other two

management options, FGRAZ2 and FGRAZ3, have both negative and positive values of NPV

depending on crop type considered in determining the opportunity cost, and the weighted

average NPV is positive for FGRAZ3. For the FGRAZ2, the PV of total benefits can compensate

98.73% of the PV of total weighted average costs with frankincense benefit alone accounting

50.38% whereas indirect benefits account only 21.09%. In the case of FGRAZ3, the PV of total

benefits covers 103.86% of the PV of total weighted average costs. For sorghum, teff, maize,

and other single crop cultivation options as opportunity cost of the land, the NPVs are positive

for both FGRAZ1 and FGRAZ2. However, if the indirect benefits are going to be deducted from

the total benefits, all the NPVs of these last two management option will turn to negative values

except for the cases of single crop cultivation and sorghum in the case of FGRAZ3.

4.4.3. Opportunity cost of REDD+ specific to Bosswellia papyrifera forest

Compared to the competing land use, which is shifting cultivation, maintaining Boswellia

papyrifera forest as exclosures can result a total of 141.70 tCO2 emission reductions whereas it

can only result in 57.18 tCO2 emission reductions under the free grazing and leaf lopping

management option. In the case of exclosures, the sum of avoided soil carbon loss and soil

carbon sequestration account 75.44% of the total emission reduction the remaining share is

accounted by avoided emission from biomass. In the case of free grazing and leaf loping

management option, because there is no addition in the form of soil carbon sequestration, the

only emission reduction sources are from avoided biomass carbon emission (60.86% of the

total) and loss of soil carbon (39.14%) that could arise if the forest is converted to agriculture.

In the REDD+ mechanism, opportunity costs are expressed in terms of $ per tCO2 emission

reductions and/or CO2 sequestration. Therefore, we calculated the PVs of each opportunity cost

items in $ per tCO2 as a ratio of the PVs and 141.7 tCO2 for the case of all forest management

options under exclosure and as a ratio of the PVs and 57.18 tCO2 for the forest management

options under the free grazing and leaf lopping scenario.

The PVs of opportunity costs per tCO2 emission reduction for Boswellia papyrifera forest are

presented in Figure 4.1. The PV of total REDD+ opportunity cost specific to Boswellia papyrifera

forest ranges from $21.61 to 66.64 per tCO2. This range of opportunity cost could maintain a

hectare of the forest for 30 years and reduce emissions of 57.18 to 141.70 tCO2. Because

exclosures allow more emission reductions than the scenario of free grazing and leaf lopping,

the opportunity costs per tCO2 in exclosure management is lower. EXCLOSUR1 and

EXCLOSUR2 options have the same and lowest opportunity cost (21.61 $ per tCO2) and

followed by the CONS (32.75 $ per tCO2) whereas the FGRAZ1 has the highest opportunity

cost (66.64 $ per tCO2). The important point here, is if REDD+ financing is possible, which

75

opportunity cost rates will best maximize the REDD+ objectives? The next section of multi-

criteria analysis of our results answers this question.

Figure 4.1: Present values of opportunity costs ($ per tCO2) of maintaining Boswellia papyrifera forest

under different forest management options for a period of 30 years at real discount rate of 5.99%

4.4.4. Distributional effects

Looking at the actors involved in the frankincense market channel and classifying them into

actors from rural and urban areas helps to understand how the benefits and costs of

frankincense forest are distributed among these groups and to derive policy implications on

income distribution effects of each management options.

Figure 4.2 below shows the actors in the frankincense market channel. The channel starts in the

rural areas where the forest is located and harvesting of the product is taking place. The

processing of the harvest for domestic and export markets takes place in urban areas.

13.89

27.17

32.96

25.72

52.52

62.10

-32.75

-21.61 -21.61

-66.64

-39.03 -39.03

-80

-60

-40

-20

0

20

40

60

80

CONS EXCLOSUR1 EXCLOSUR2 FGRAZ1 FGRAZ2 FGRAZ3

PV

of

be

ne

fits

an

d o

po

rtu

nit

y c

os

ts (

$ p

er

tCO

2)

Forest Managemnt Options

tCER Benefits

Nutrients Benefits

Frankincense Benefits

Grass Benefits

Leaf fodder Benefits

Forestry Benefits

Leaf fodder OC

Fuel wood OC

Grass fodder OC

Frankincense OC

Forestry OC

Sorghum OC

Teff OC

Maize OC

Sesame OC

≥ 3 crops OC

Other single crop OC

Sorghum + Sesame OC

Other two crops OC

Agriculture OC

Total OC

Benefits

Opprunity costs

76

Figure 4.2: Supply chain of frankincense in Ethiopia

According to the information from the interviews with frankincense trading firms, the actors in

the supply chain of frankincense can be classified into five major groups based on their duties

and level of involvement in the frankincense harvesting, processing, and trading. These groups

are tappers, local middlemen (called locally as ‘squadra’) who organize the tappers, local

farmers’ cooperative firms, companies engaged only in the domestic trade, and companies

involved in exporting. Frankincense forests from which the product is harvested are under the

concession of the last three groups and they pay royalty to the local government. The tappers

are mostly young and landless people from the highlands of Tigray. They work on these forests

as labour suppliers and get wage in return depending on the amount collected. They enter

contract with the companies mostly through the middlemen to work during the harvesting

season and supply a minimum quantity of frankincense at a predetermined price (wage per unit

of harvested frankincense). In addition to this wage, the local middlemen will cover a specified

amount of food and health expenses of the tappers over the contract period. The middlemen

perform this task as agents to the company that administers the forest under its concession. In

return, the company pays a pre-determined price per unit of frankincense harvested as a

commission to the middlemen plus the food and health expenses that they paid to the tappers.

Local farmers’ cooperative firms directly employ and pay wage to the tappers for harvesting

frankincense from the forest under their concession. They sell the grezo frankincense to either

exporting companies or companies engaged in processing the product for sale to the exporting

companies. The companies engaged in export also undertake grading of the grezo into seven

quality grades for export (see Figure 1.1). Poor women who work as seasonal wage employees

Harvesting

(Rural)

Processing

(Urban)

Export

(Urban)

• Tappers

(Young poor men)

• Middle men ('Squdra ')

• Rural Farmers'

Cooperative firms

• Transporters

• Loading & unloading

(Young poor men)

Urban Poor

• Grading

(Urban poor women)

• Loading& unloading

(urban poor men)

Urban High Income

• Domestic trading

business firms

• Exporters

• Transporters

• Business firms

• Foreign stakeholders

(Port service)

77

in the processing and exporting companies perform grading of the grezo into the different

quality grade. Urban poor men also get seasonal labour work of unloading grezo frankincense

coming from the forest for grading and loading the graded frankincense in transporting it to the

export market. Transporters involve in the chain as service providers in transporting the grezo

from the forest to the central processing place and then to the port of Djibouti for export.

In addition to the benefits from frankincense to the above actors in the rural areas, benefits from

grass and leaf as fodder and forgone opportunity costs from crop production are attributed to

the general rural farm households. We further classified the urban group into urban1

representing both the urban poor women working in the grading task and poor men working the

loading and unloading task for the trading companies in the urban areas, and Urban2

representing the business firms. Based on the data on the cost structure of frankincense trading

that we got from one of the business firms, handling cost at the port is one component and

accounts for 4.64% of the total processing and trading cost. Because Ethiopia has no port, we

considered this as foreign stakeholders in analyzing the distributional effects of the forest

management potions.

Accordingly, figure 4.3 shows the distribution of the present values of benefits and costs of

keeping a hectare of the forest under the six forest management option to the actors in the

frankincense supply chain. In the case of the distribution of present value of benefits, the PV of

the total benefit under the CONS scenario is all a positive externality in the form of emission

reduction, and enhancement and reductions of nutrient loss. These positive externalities are

indirect benefits and attributed to none of the groups (Figure 4.3). However, in order to achieve

this level of positive externalities, the rural actors bear much of the PV of the weighted average

total cost. The PV of the emission reduction benefit is $ 7.80 per tCO2 and the PV of soil nutrient

enhancing and loss reduction benefit is 6.09 $ per ton. Assuming REDD+ as the possible

financing mechanism for these benefits, it is important to know the opportunity costs to each

groups of the society per tCO2 emission reductions. Accordingly, of the $32.75 opportunity cost

per tCO2 emission reduction under the CONS option the rural actors bear $28.94 per tCO2

emission reduction whereas all the other actors bear the remaining small amount.

78

Figure 4.3: Distribution of present values of benefits and costs (%) of maintaining a hectare of Boswellia

forest under six management options to actors in the frankincense supply chain

From the EXCLOSURE1 management onwards, the share of the indirect benefits to total

benefits decreases whereas the share of direct benefits increases. In addition, much of the PV

of the direct benefits from the resource is attributed to the rural actors. However, the rural actors

incur much of the opportunity cost of keeping the forest (not shifting to crop cultivation) under

each of these management options as well. Except for the FGRAZ1, in each of the other four

cases 100% of the opportunity cost is incurred by the rural actors the opportunity costs account

50 to 60% of the PV of the total cost. The other actors bear opportunity costs in case of the

CONS and FRGRAZ1 and their share in total account 20.38% and 14.5% of the total

opportunity costs respectively.

4.4.5. Multicriteria analysis

Table 4.5 shows the results of a multi-criteria analysis. Twelve criteria derived from the results

of this and previous studies on frankincense forest were used for the analysis. The criteria can

be grouped into five categories, which are consistent with the objectives of the REDD+ program.

These criteria are opportunity cost per tCO2 emission reduction, quantities of emission

reductions and enhanced carbon stocks in soils, nutrient loss reduction and enhancement,

effects related to sustainability and biodiversity conservation, and effects related to income

distribution and poverty reduction.

-79.72 -58.55 -50.12

-85.52

-59.95 -53.33

0

33.64 39.81

56.39 39.81

63.67

-100

-75

-50

-25

0

25

50

75

100

CONS EXCLOSUR1 EXCLOSUR2 FGRAZ1 FGRAZ2 FGRAZ3

Rural actors OC Direct costs Urban2 OC

Foreign stakeholder OC Urban1 OC Urban1 Benefit

Foreign stakeholder Beneefit Urban2 Benefit Rural actors Benefit

Indirect Benefits

79

Table 4.5: Scores of multi-criteria analysis for Boswellia papyrifera forest management options

Criteria CONS EXCLOSUR1 EXCLOSURE2 FGRAZ1 FGRAZ2 FGRAZ3

I. Average OC of REDD+ in $

per tCO2

32.74

(0.13)

21.61

(0.20)

21.61

(0.20)

66.64

(0.06)

39.03

(0.11)

39.03

(0.11)

II. Emissions

Emission reduction in tCO2 ha-1

74.08

(0.10)

74.08

(0.10)

74.08

(0.10)

57.18

(0.08)

57.18

(0.08)

57.18

(0.08)

Enhanced carbon soil stock in

tCO2e ha-1

67.62

(0.10)

67.62

(0.10)

67.62

(0.10)

0.00

(0.00)

0.00

(0.00)

0.00

(0.00)

III. Nutrients

Nitrogen loss reduction in

(Mg ha-1

)

0.28

(0.05)

0.28

(0.05)

0.28

(0.05)

0.18

(0.03)

0.18

(0.03)

0.18

(0.03)

Nitrogen enhancement in

(Mg ha-1

)

0.65

(0.05)

0.65

(0.05)

0.65

(0.05)

0.00

(0.00)

0.00

(0.00)

0.00

(0.00)

Phosphorous loss reduction in

(Mg ha-1

)

0.00

(0.05)

0.00

(0.05)

0.00

(0.05)

0.00

(0.03)

0.00

(0.03)

0.00

(0.03)

Phosphorous enhancement in

(Mg ha-1

)

0.00

(0.05)

0.00

(0.05)

0.00

(0.05)

0.00

(0.00)

0.00

(0.00)

0.00

(0.00)

IV. Biodiversity conservation

Effect on flowering (mean

log10(#of flowers)

1.34

(0.07)

1.34

(0.07)

1.34

(0.07)

1.21

(0.06)

1.21

(0.06)

1.21

(0.06)

Effect on fruiting (mean

log10(#of fruits per flower)

1.37

(0.07)

1.37

(0.07)

1.37

(0.07)

1.27

(0.06)

1.27

(0.06)

1.27

(0.06)

Risk of population decline

(Groenendijk et al ., 2012)

1.00

(0.07)

2.00

(0.03)

3.00

(0.02)

6.00

(0.01)

5.00

(0.01)

4.00

(0.02)

V. Income effects

Direct benefits in $ per tCO2

emission reduction

0.00

(0.00)

13.28

(0.06)

19.07

(0.08)

14.50

(0.04)

41.30

(0.08)

50.88

(0.10)

Income distribution effect (% of

REDD+ revenues based on

opportunity costs to the poor)

0.80

(0.10)

0.67

(0.08)

0.53

(0.07)

0.72

(0.05)

0.55

(0.07)

0.52

(0.05)

Sum of scores in ( ) 0.832 0.904 0.912 0.430 0.545 0.546

Rank 3rd

2nd

1st

6th

5th

4th

Sum of scores* 0.896 0.870 0.854 0.429 0.516 0.510

Rank 1st

2nd

3rd

6th

4th

5th

*Different weights for the five groups of criteria (0.1; 0.3; 0.1; 0.3; and 0.2 for criteria I to V respectively).

The multicriteria analysis ranks EXCLOSUR2 as first and CONS as third. In general the

exclosure options are better than the free grazing and leaf lopping management options for

achieving REDD+ objectives in the context of Boswellia papyrifera forest. If more weights are

considered to biodiversity conservation, emission reduction and poverty reductions, the CONS

option ranks first followed by the EXCLOSUR1. The free grazing and leaf loping management

options are in general inferior to the exclosure options in meeting the criteria and FGRAZ1

scenario ranked the least of all.

80

4.4.6. Sensitivity analysis

Although most of the forest management options analyzed do not pass the NPV criteria, it is

worth to make a sensitivity analysis. It helps to take into account uncertainties in future flows of

benefits and costs that could arise due to changes in the discount rate, input and output prices

and quantities and hence their likely effect on the NPV as well as the opportunity cost of REDD+

specific to the species. Specifically, it helps to identify the most important parameters that affect

NPV, which of the changes in these factors result in a positive NPV compared to the base case

negative NPV, and assess how the opportunity cost per tCO2 emission reduction changes due

to these changes.

Table 4.6 summarizes the impact of changes of all the parameters on NPV estimates based on

the weighted mean opportunity cost of land. Changes in the opportunity cost of land, total cost

of frankincense processing and trading, frankincense prices, and quantities of crop output have

strong influences on the NPVs of most of the forest management options. For example, other

things remain the same; a 40% decrease in the opportunity cost of land causes a 62.20% rise in

NPV of the CONS option whereas an increase in the total cost of frankincense processing and

trading by 40% results in a 41.29% rise in the NPV of the CONS. However, none of these

proportionally higher changes in NPV relative to the 40% change in the three parameters result

in a non-negative NPVs of this management option (Figure 4.4). Moreover, the NPVs for the

CONS management option remain negative irrespective of changes in each of the parameters

at both rates of change. Similar result is observed for the case of FGRAZ1 (Table 4.6).

81

Figure 4.4: Sensitivity of NPV of conservation of Boswellia forest to (±40%) changes in parameters

In the case of the other four forest management options, a given percentage change in most of

the parameters causes proportionally higher percentage change in NPVs. Changes in ten of the

fifteen parameters cause a relatively larger change in the NPVs of the two exclosure forest

management options. These parameters are total cost of frankincense processing and trading,

prices of all benefits, opportunity cost, frankincense prices, crop output, prices of crops, price of

tCER, prices of fertilizer, price of grass, and the real discount rate. For example, a 40% rise in

the prices of all benefits results in a 172.39% rise in the NPV of EXCLOSUR1 and a 241.09%

rise in the NPV of EXCLOSUR2. Moreover, a 40% change in each of five of the above ten

parameters results in non-negative NPVs in both management options. These parameters are

total cost of frankincense processing and trading, prices of all benefits, opportunity cost,

frankincense prices, and prices of crops. In addition, a 40% decline in the real discount rate also

results in a non-negative NPV in the case of EXCLOSUR2.

-2722

-1641

-3642 -3737

-1029

-3173

-4000

-3500

-3000

-2500

-2000

-1500

-1000

-500

0

Base

Fra

nkin

cense p

rices

Fra

nkin

cense h

arv

estin

g c

ost

Fra

nkin

cense y

ield

TC

fra

nkin

cense tra

din

g

Prices o

f cro

ps

Cro

p o

utp

ut

Price o

f fe

rtili

zer

Gra

ss p

rice

Gra

ss b

iom

ass

Leaf bio

mass

Price o

f tC

ER

Fu

el w

ood p

rice

Prices o

f all

outp

uts

Opport

unity c

ost

Dis

count ra

te

CONS

NPV at 40% decrease in parameters

-2722

-3802

-1598

-4414

-5000

-4500

-4000

-3500

-3000

-2500

-2000

-1500

-1000

-500

0

Base

Fra

nkin

cense p

rices

Fra

nkin

cense h

arv

estin

g c

ost

Fra

nkin

cense y

ield

TC

fra

nkin

cense tra

din

g

Prices o

f cro

ps

Cro

p o

utp

ut

Price o

f fe

rtili

zer

Gra

ss p

rice

Gra

ss b

iom

ass

Leaf bio

mass

Price o

f tC

ER

Fu

el w

ood p

rice

Prices o

f all

outp

uts

Opport

unity c

ost

Dis

count ra

te

CONS

NPV at 40% increase in parameters

82

Table 4.6: Sensitivity of NPVs to changes in prices of cost and benefit items and real discount rate

CONS EXCLOSUR1 EXCLOSURE2 FGRAZ1 FGRAZ2 FGRAZ3

Base NPV -2722 -681 -623 -1931 -38 132

Parameter

Change (%)

+15

-15

+40

- 40

+15

-15

+40

- 40

+15

-15

+40

- 40

+15

-15

+40

- 40

+15

-15

+40

- 40

+15

-15

+40

- 40

Frankincense

prices

-3127 -3802 -399 71 -218 457 -2336 -3012 191 575 444 964

-2316 -1641 -964 -1434 -1029 -1704 -1526 -850 -268 -651 -180 -700

Frankincense

harvesting cost

-2695 -2650 -715 -770 -675 -761 -1904 -1860 -67 -115 95 32

-2748 -2793 -648 -593 -572 -485 -1958 -2002 -10 38 169 232

Frankincense

yield

-2958 -3353 -612 -497 -545 -415 -2168 -2562 9 87 205 326

-2485 -2090 -751 -866 -701 -831 -1694 -1300 -86 -164 59 -62

Total cost of

frankincense

processing &

trading

-2237 -1598 -1293 -2094 -1564 -2805 -1446 -808 -565 -1255 -556 -1459

-3003 -3642 -331 471 -76 1165 -2213 -2852 263 953 527 1429

Prices of crops

-3002 -3469 -962 -1429 -904 -1371 -2212 -2679 -319 -786 -148 -616

-2441 -1974 -401 67 -343 125 -1651 -1183 242 710 413 880

Crop output

-3038 -3566 -998 -1526 -940 -1468 -2248 -2775 -355 -883 -185 -712

-2405 -1877 -365 -296 -307 -238 -1614 -1087 278 806 449 976

Price

of fertilizer

-2573 -2325 -533 -285 -475 -227 -1902 -1854 -10 38 161 209

-2870 -3118 -830 -1078 -772 -1020 -1960 -2008 -67 -115 103 55

Grass price

-2859 -3087 -819 -1047 -761 -989 -1819 -1633 73 259 244 430

-2584 -2356 -544 -316 -486 -258 -2043 -2229 -150 -336 20 -166

Grass biomass

-2808 -2951 -767 -911 -709 -853 -1845 -1702 48 191 218 362

-2635 -2492 -595 -452 -537 -394 -2017 -2161 -124 -268 46 -97

Leaf biomass

-2760 -2824 -720 -784 -662 -726 -1893 -1829 0 64 170 234

-2683 -2619 -643 -579 -585 -521 -1969 -2033 -77 -141 94 30

Price of tCER

-2556 -2280 -516 -239 -458 -181 -1844 -1700 48 193 219 363

-2887 -3737 -847 -1697 -789 -1639 -2018 -2162 -125 -269 45 -99

Fuel

wood price

-2795 -2917 -755 -877 -697 -819 -2004 -2127 -112 -234 59 -64

-2648 -2526 -608 -486 -550 -428 -1858 -1735 35 157 205 328

Prices of

all outputs

-2969 -3381 -241 493 -60 879 -2128 -2458 399 1129 652 1518

-2474 -2062 -1122 -1856 -1187 -2125 -1734 -1405 -476 -1206 -388 -1254

Opportunity

cost

-3356 -4414 -1079 -1742 -1021 -1685 -2441 -3292 -312 -768 -142 -597

-2087 -1029 -283 380 -225 438 -1421 -570 235 691 406 861

Discount rate

-2595 -2422 -804 -973 -805 -1055 -1845 -1728 -174 -363 -44 -288

-2869 -3173 -539 -247 -413 20 -2033 -2245 118 438 336 752

In the case of FGRAZ2 and FGRAZ3, a given percentage change in any of the parameters

causes a proportionally higher percentage change in NPVs. Changes in total cost of

frankincense processing and trading, prices of all benefits, and quantity of crop output are the

83

three most important parameters with very large influences. For example, a 15% fall in the cost

of frankincense processing and trading causes a 792.11% rise in the NPV of FGRAZ2 and a

299.24% rise in the NPV of FGRAZ3. Moreover, for all parameters except frankincense

harvesting cost and price of fertilizer, a 15% increase in a parameter that correlates positively

with NPV or vise versa causes non-negative NPV for the case of FGRAZ2. At a 40% similar

change, all parameters result in non- negative NPVs. For the FGRAZ3, which is the only

management option with positive NPV at the base case and represents the business as usual

case, the value remain positive for a 40% rise or fall in frankincense harvesting cost, price of

fertilizer, and quantity of leaf biomass. However, for all of the other parameters, a 40% rise in a

parameter with which NPV has negative correlation and vice versa results in a negative NPV.

A given percentage change in the discount rate causes a lower proportional change in the

NPVs of the CONS and FGRAZ1 options whereas it causes a higher proportional change in the

case of NPVs of the other forest management options. For the lower case, for example, a 40%

reduction in the discount rate causes only a 16.57% rise in the NPV of the CONS and 16.26%

rise in the NPV of the FGRAZ1. Contrary, the same percentage fall in the discount rate results

in a 103.21% rise in the NPV of EXCLOSUR2 and a 1252.60% rise in the NPV of FGRAZ2.

In the case of the three exclosure and FGRAZ1 management options, the present value of

REDD+ opportunity cost per tCO2 emission reduction is less sensitive to changes in any of the

parameters. A given percentage change in a parameter causes a less percentage change in the

PV of REDD+ opportunity costs of these forest management options. However, changes in

prices of all benefits, quantity of crop output, prices of frankincense, the real discount rate, and

total cost of frankincense processing and trading are the most important parameters in

influencing the present values of REDD+ opportunity costs for the CONS and FGRAZ1

management options. For example, a 40% increase in prices of all benefits (frankincense, grass

and leaf biomass, fertilizer, and tCER) at the same time, results in a rise of 31.37% in the PV of

the REDD+ opportunity cost per tCO2 emission reduction under the CONS option and a 29.49%

rise in the case of the FGRAZ1 management option. In the case of the other two exclosure

management options, the PV of REDD+ opportunity cost is relatively more sensitive to changes

in quantity of crop output, prices of crops, and the real discount rate. For example, a 40% rise in

prices of crop output causes a 36.48% increase in the PV of REDD+ opportunity cost of both

EXCLOSUR1 and EXCLOSUR2.

For FGRAZ2 and FGRAZ3 options, changes in quantities of crop outputs and prices of crops

have a proportionally higher influence on the PV of REDD+ opportunity cost. For each of these

two management options, while a 40% rise in crop output causes a 50.06% raises in the PVs of

REDD+ opportunity costs, a 40% rise in crop prices result in a 44.81%. A change in the

84

discount rate is the third influencing factor but causes a relatively lower proportional change in

the PV of the REDD+ opportunity costs of these two management options.

4.5. Discussion and conclusions

The objective of this chapter was to evaluate the economic implications of experiment based

Boswellia forest management options and compare them with the competing land use, which is

shifting to crop cultivations. The standard model of cost benefit analysis is used to analysis data

mainly from plot level experiments and household surveys. The results revealed that the

weighted average economic NPV of Boswellia papyrifera forest is positive only for the free

grazing and leaf lopping with heavy frankincense tapping forest management option, which

resembles the business as usual case. The present value of benefits from the indirect

ecosystem services (benefits from carbon and soil nutrient storage) are $ 1968 ha-1 for

exclosures and $ 641 ha-1 for free grazing Boswellia forest management option. These benefits

are positive externalities from keeping the forest. Although it may be possible to trade carbon

storage benefits under the REDD+ mechanism, the benefit in the form of nutrient storage, which

can serves as a buffer for ensuring the provision of nutrients in the future, can hardly be traded

on markets. If these indirect ecosystem service benefits are deducted, none of the forest

management options evaluated in this study are economically profitable compared to the

competing land use option, which is shifting cultivation. Mekuria et al., (2011) also reported that

benefits in the form grass biomass and carbon revenues from tCER units could not compensate

the opportunity cost of exclosures from wheat, teff or barely productions. In a recent study on

economic analysis of exclosures, Balana et al. (2012) reported negative NPV for a conversion

of productive cropland to exclosure. However, an earlier study by Tilahun et al., (2007) on the

same species reported positive NPVs for both exclosure and open for free grazing sites and

indicated that the values are higher than the NPV of crop production. Our result is different

because there have been significant inflationary changes occurred in the country since 2003,

which was the base year for the analysis in the study of Tilahun et al. (2007), and these

changes have impact on the relative values of the benefit and cost items used in our analysis.

The relative changes in forestry output and input prices, which are relevant to this study are not

proportional to the relative changes in output and input prices of crop production. On the other

hand, the present study covers a wider study site with a relatively more productive land, and

hence, the crop productivity per ha applied for the present study is on average 62.37% higher

than the average crop output in the study of Tilahun et al. (2007). There is also a significant

difference in the quantity of grass biomass used between the two studies, which is due to both

site difference and difference in the methods used for quantifying the grass biomass. The

quantity of grass biomass per ha in the current study is 78% lower than the estimated quantity

by Tilahun et al. (2007). The other source of the difference is that the NPVs in this study are

based on economic rather than financial analysis. Therefore, in the case of exclosures grass

85

and leaf fodder benefits are considered as opportunity costs whereas the value of grass in the

previous study was considered as benefit.

Both the base case and sensitivity analyses results clearly indicate that unless certain

conservation interventions are in place, the current trend of deforestation of the resource and

hence emission of CO2 both from biomass burning and soil organic carbon loss will continue.

Moreover, in the long term, the risk of extinction of the species is likely to increase and a

number of studies on the species have been alarming this fact (Gebrehiwot et al., 2002;

Gebrehiwot et al., 2003; Rijkers et al., 2006; Eshete et al., 2012b, Groenendijk et al., 2102).

Although frankincense is a low carbon forest, from the points of view of its widely recognized

economic, medicinal, cultural, and religious importance as well as biodiversity conservation, it

requires consideration in the REDD+ framework so that it would be possible to conserve it

through compensating local people for their opportunity costs. Accordingly, our study indicated

that the REDD+ opportunity cost per tCO2 is lower for the exclosure management than the free

grazing with leaf lopping option. The present value of the opportunity cost for the conservation

scenario is $ 32.75 per tCO2 and decreases to 21.61 per tCO2 if frankincense tapping with either

normal or heavy tapping options is allowed. With this level of opportunity cost it is possible to

ether purely conserve the forest for a period of 30 years or manage as exclosures with

frankincense extraction for the same period. However, in the case of the free grazing with leaf

lapping management it is $ 66.64 per tCO2 for the no tapping scenario and deceases to 39.03 $

per tCO2 for frankincense extracting options. Our estimates for the exclosure forest

management scenarios are lower than both the 2008 price for carbon market of the EU

Emission Trading Scheme, which were running about 35 to 40 $ per tCO2 and the latest

PointCarbon (2011) estimate of global carbon price of $ 35 per tCO2 for 2020. However, our

estimates are larger when compared to REDD+ opportunity cost estimates based on regional

studies, the Stern Review, and Global Economic Models. Based on a review of 29 regional

empirical studies, Boucher (2008) found an average REDD+ opportunity 2.51 per tCO2. A

conversion of the area based Grieg-Gran’s estimate for the Stern (2006) and Eliasch (2008)

Reviews to per-ton costs provides a range of $2.67 to $8.28 per tCO2 (Boucher, 2008).

Estimates from based on global economic models range from $6.77 to $17.86 with an average

of $11.26 per tCO2 (Kindermann et al., 2008). The reason for relatively high opportunity cost

estimates in our case could be explained by the fact that, frankincense forest is a low carbon

forest but relatively high commercial value. Frankincense and other natural gum and resin

exports are one of the sources of foreign currency to the country. According to data from the

Central Statistical Authority of Ethiopia, the country exported 14978 tons of oleo-gum resins

(mainly Frankincense, Myrrh, and Gum Arabic) over the period 2003-2007 and earned US$

21.53 Million. Moreover, the domestic and global price rise in agricultural crops also contributes

to the rise in the opportunity cost of land.

86

The result of the multi criteria analysis indicates that exclosure management options are

superior to free grazing. Moreover, the conservation option is the best option followed by the

exclosure with a normal tapping option for achieving REDD+ objectives, which are reducing

emissions and enhancing carbon stock, biodiversity conservation, and poverty reduction, and

sustainable management of forests. The current practice of leaf lopping from the tree for fodder,

free grazing, and heavy tapping of frankincense are unsustainable and for both future

frankincense production as well as the species itself. Thus, both a conservation measure as well

as management of the resource as exclosures with tapping of not more than 6 spots per tree

has to be practiced for sustainable use of the resource. Moreover, tapped trees have to get

sufficient resting period as recommended by previous studies.

CHAPTER 5

Rural households’ demand for frankincense forest conservation: a contingent valuation analysis

Submitted to Journal

(Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H. & Mathijs, E.)

5.1. Introduction

The export sector in Ethiopia is dominated by primary commodities with a share of about 90% of

the total export earnings of the country and the sector accounts for a small share (13.6%) of

GDP (World Bank, 2010). Exploiting local comparative advantages may lead to increases in

primary product exports but also in more environmental degradation (Rock, 1996; Muradian and

Martinez-Alier, 2001). In case of forest resources, a number of studies argued that trade in non-

timber forest products can promote conservation and sustainable management of tropical

forests while improving rural livelihoods (Belcher and Schreckenberg, 2007). However, it often

involves overuse and extreme degradation of forest resources (Arnold and Pérez, 2001) mainly

in the case of open access forests. Literature on international trade based on the comparative

advantage of developing countries from their natural resource endowments also gives insight on

how trade in renewable resources from open access resources could lead to overexploitation of

the resource and to a loss of social welfare (Azqueta and Sotelsek, 1999).

Forest resources in Ethiopia are state properties and this is true for Boswellia papyrifera forests

too. However, there is no management for sustainable extraction of resins from the forest.

There is also a very limited technical expertise as well as skilled labour in the forestry sector in

the country at large. Therefore, we can reasonably consider the resource as ‘open access’

despite it is legally state property and see how profit maximizing private firms behave. Given the

fact that the international market for primary products is nearly competitive, private firms could

maximize their profit by minimizing the private (financial) costs of extracting the resource.

Moreover, the critical shortage of foreign exchange and balance of payments problems also

force the government to adopt a number of incentive mechanisms to promote the export sector

and hence the commercial exploitation of the resource. Prior to 1991 the state-owned Ethiopian

Natural Gums and Resins Marketing Enterprise was the only company in the country exporting

oleo-gum resins to the international market. According to CSA data, the average annual export

of the company over the period 1978-1991 was about 750 tons. Following the post-1991 market

based economic policy that allowed the participation of private companies into the business, the

country’s export volume of oleo-gum resins has increased to 2000 tons per year in the period

1996-2003, and reached to about 3000 tons per year over the period 2003-2007. The country

88

exported 4,533 tons in 2007 alone to 27 different countries around the world with China, United

Arab Emirates, Germany, France and Greece importing 64% of the total.

Although the opening of the market generated more revenue to the country, it is hardly

improving the poverty situation of particularly the rural households living around these forests.

On the supply side, the low cost of tapping the resource implies low income to rural households

who are mostly in the beginning of the market channel and receive low wages for the labour

they supply for tapping the resource. They will not have an incentive to maintain the resource as

a forestland and will go for shifting cultivation to meet immediate subsistence needs. In effect,

the deforestation and degradation of the resource leads to the emission of carbon dioxide into

the atmosphere as well as desertification. Moreover, oleo-gum resins are important inputs in

pharmaceutical industries. Degradation of the resource also implies a negative effect on the

services it will provide to research and development in the pharmaceuticals and other sectors,

mainly in the importing countries. The international market price for these primary products does

not take into account these negative externalities. The public good nature of the resource has

contributed to a continuous degradation and has led to this ‘tragedy of the commons’ and a

number of studies reported lack of regeneration and continuous degradation of the resource in

East Africa (Ogbazghi et al., 2006; Negussie et al., 2008). TRAFFIC (the wildlife trade

monitoring program of WWF) has reported Boswellia papyrifera as endangered species that

needs priority in conservation (Marshall, 1998). Therefore, local, national and international

solutions are required for conservation and sustainable use of the resource.

The general objective of this study is to assess rural households demand for conserving

Boswellia papyrifera as a potential local solution and draw some policy measures required at

national and international levels for the conservation and sustainable management of the

resource. Thus, we hypothesize that despite poverty, people in rural areas are aware of the

importance of conserving the resource. They are typically income poor but labour abundant.

Therefore, they might not be able to pay cash for resource conservation, but they might be more

willing to contribute labour. Hence, we assessed the demand for conservation in terms of their

willingness to contribute cash and labour for Boswellia papyrifera forest conservation. In

addition, this paper identifies the factors affecting the willingness to contribute and compares

the asymmetry in WTP and WTCL using the market wage, the minimum wage and the per

capita daily income for converting WTCL into monetary values. Our study is the first to apply

both cash and labour for eliciting rural households’ demand for Boswellia papyrifera forest

conservation.

89

5.2. Value to be estimated and the contingent valuation method

A review of the environmental economics literature suggests that the total economic value of a

resource is the sum of its use, option, existence and bequest values (Campbell and Luckert,

2002). Frankincense of Boswellia papyrifera is an input in the pharmaceutical, chemical, food

and cosmetic industries. The leaves of the tree have nutritive value as livestock feed (Melaku et

al., 2010). These direct uses can be valued using market-based methods. In addition, the

resource has non-market benefits. With uncertainty about future demand, there may be an

interest to keep the resource as an option for future use. Given the current trend of degradation,

people are concerned about the possibility of extinction and one would get utility from

contributing for its conservation and improving the welfare of future generations.

Contingent valuation (CV) methods have received increasing attention as a means to estimate

option and existence values. Money is the conventional unit of account of value used in CV

questions. In Ethiopia where 39% of the population lives below a per capita income of US$ 1.25

a day (World Bank, 2010), household incomes are often inadequate to meet the basic needs

and asking only WTP from their income may not fully capture households’ valuation of

environmental amenities and hence their demand for such services. Considering this fact, there

are few studies in developing countries (Swallow and Woudyalew, 1994; Echessah et al., 1997;

Kamuanga et al., 2001; Hung et al., 2007) that applied both labour and cash for eliciting

willingness to pay for tsetse fly control and prevention of forest fires.

Many CV studies rely on a single bound dichotomous choice (SBDC) approach that asks

respondents whether they would accept a randomly assigned predetermined single bid amount.

However, this method can be statistically very inefficient and requires a large sample to attain a

given level of precision (Cameron and Quiggin, 1994). A double-bounded dichotomous choice

(DBDC) approach in which the respondent is asked a follow-up question if s/he would pay a

higher or lower bid depending on the response to the initial bid (Hanemann et al., 1991) is often

used to improve the efficiency of the SBDC model, and we applied this to our study.

5.3. Materials and methods

5.3.1. Survey design and data collection

To assess households’ demand for Boswellia papyrifera forest conservation, a CV survey on

520 sample households was conducted in five villages located in the frankincense producing

districts of central and western Tigray in Northern Ethiopia (Details on selection of the villages is

given in section 4.3.3 of chapter 4). In each of the five rural communities of the study area

(Figure 1.4 in chapter 1), a group of key informants of 6-8 people comprising local leaders,

agricultural development agents, members of frankincense cooperative firms, and residents

90

were organized. We obtained a list of resident household heads from the local administration.

Based on the list we first classified households based on membership in local frankincense

cooperative firm. The group further set a wealth-ranking criterion (based on livestock units

owned) for three groups (rich, medium, and poor) and determined the total number of residents

in each of the wealth groups. Using this information, the proportion of samples from each group

was then determined and the sampling process was carried afterwards. Based on the sampling

frame, we did the sample selection for the members and non-members using the following

procedure. According to the list, each time a random number was drawn and the name of the

household corresponding to the selected number on the list was read, the key informants

decide to which wealth class the selected household belongs. The procedure was continued

until the required number of sample households was selected. Accordingly, 120 cooperative

member and 400 non-member sample households were selected for the study.

According to CV experts, it is important to provide respondents with adequate and accurate

information to make them fully aware of the hypothetical market situation and arrive at correct

WTP measures. Therefore, a survey with four major parts was designed for this study

(Appendix 4A, the first three parts are section 2.1 to 2.3 and the fourth part include the sections

III-V of the questionnaire). The first part describes the uses of Boswellia papyrifer and

respondents were told about the different uses of the tree and the forest. They were asked a

follow up question to evaluate what they knew before with what they were told during the

interview and their level of understanding of these uses. In the description of the state of

degradation, respondents were informed about the results of previous studies on the problems

of continuous degradation of the resource due to shifting cultivation, overtapping for

frankincense, and lack of regeneration and the fact that the species is on the verge of extinction.

This verbal description of the state of degradation of the resource was accompanied by

illustrations using three photographs of Boswellia stands. The photographs were taken in March

2009 and August 2009 from the same study area. These pictures show the trend in the

degradation of the resource. They were presented on an A4 size photo paper to make a clear

visual contrast (Figure 5.1).

Following the verbal and visual description of the state of degradation of the resource,

respondents were asked follow-up questions. The questions are about whether they noticed the

pictures well or not; in which state they prefer to transfer the resource in their village to their

grand children; and to what extent they are concerned about the rate of deforestation and risk of

extinction of the species.

91

Figure 5.1: Boswellia papyrifera stands in different state of degradation in Western and Central Tigray

Respondents were told about the need for conservation of Boswellia forest, prevent it from

extinction, and keep the resource for future use by existing and future generation. Three goals

of the conservation program were clearly explained to the respondents as: Goal 1: Keeping the

Boswellia forest as option of meeting future demand for the different products and services from

the resource; Goal 2: To realize the existence of the resource and curb the possibility of

extinction; Goal 3: Improve the welfare of the future generation of the society. For each of the

conservation goals, respondents were asked to rank their interest on a scale from 1 to 10 where

1 is “not at all interested and 10 is “very much interested”.

The second part deals with the method of provision, the payment mechanism, the decision rule

and time frame of the payment. Respondents were informed about the requirement of their

financial and labour contributions for achieving the goals of the conservation program. They

were told that payment of cash contributions will be collected as in the form of annual tax by the

local government and the money will be used purely to the implementation of the conservation

project. The annual voluntary labour contribution will be in terms of patrolling the conservation

site for controlling free grazing, cutting of trees, shifting cultivation practices, and tapping of

frankincense. They were informed that enough volunteers would have to participate in the

annual tax or labour contributions for the conservation program and achieve its goals. After one

year of participation if in case support is going to be insufficient, the government will guarantee

the refund of their contributions in cash and the equivalent cash value of their labour

92

contributions. Following this, respondents were asked about whether the decision rules would

increase or decrease their interest to join the program.

The third part is the contingent valuation questions. We used the double bounded discrete

choice elicitation (Hanemann et. al., 1991) method. In this elicitation method, a respondent is

asked whether he/she is willing to pay a pre determined and randomly assigned value of the

non-marketed good. If the answer to the first question is positive, the respondent will be asked

whether he/she will be willing to pay a pre determined higher value. If the answer to the first

question is “no” the respondent will be asked whether he/she is willingness to pay pre

determined lower value.

Before the willingness to pay question, respondents were reminded of their household income

and expenditure. In the case of the willingness to contribute labour they are reminded of the

number of household labour physically capable of contributing labour for the planned

conservation, the total labour time required to accomplish farming and other household activities

as well as community works. The wording for initial valuation questions for cash and labour

contributions read as: a) would you (your household) contribute for the Boswellia papyrifera

forest conservation program if it costs you X ETB per year? b) would you (your household)

contribute for the Boswellia papyrifera forest conservation program if it requires you L full labour

days per year as a guard for patrolling the conservation site given that 1 full labour day = 8

working hours?

The wordings of the follow up questions were the same except for the X and L values. In both

questions for a respondent with positive response to either the initial or the follow up bids, on a

scale from “1” for very uncertain to “10” for very certain we asked how certain he/she is that

he/she would join the program and pay the value he/she voted yes. For a respondent that

respond “no” for both the initial and follow up bids, we asked a question that describes his no

response. At the end of this section a respondent was asked which contribution mechanism

he/she prefers, cash or labour or indifferent.

The last section of the survey inquires the respondent about the household’s demographic

characteristics (family size, age, gender and education), their wealth (size of land holding,

livestock assets, agricultural and non-agricultural household durables, and house value) and

income (income from crop production, livestock products, livestock sales, off-farm income and

forest environmental income), and others socioeconomic characteristics.

Survey pre-test and bid design: A pre-test survey was carried out in January 2010 on 50

randomly selected samples households in the study area. The pre-test survey had dual

93

purpose. The first was to design bid levels for the final survey based on open-ended contingent

valuation responses and the other purpose was test the survey questionnaire. In discrete choice

contingent valuation, the choice of monetary amounts that are used as bids is an important

issue. Welfare estimates could be sensitive to the specification and size of bid amounts (Cooper

and Loomis, 1992). There are a number of methods proposed by different authors (Boyle et. al.,

1988; Cooper, 1993; Alberini, 1995) for designing bids for discrete choice contingent valuation.

We used the Boyle et al. (1988) method for two practical reasons. First, compared to the other

methods it is not complex and does not require much computational effort. Secondly, according

to the study by Elnagheeb and Jordan (1995) in which they compared Boyle et al. ‘s method

with Coopers’ and an ad-hoc approach of their own using a Monte-Carlo models of willingness

to pay, they found that the Boyle at al.’s method generally performed better, particularly if

variation in willingness to pay estimates was small. They also found with increased variability

and sample size, willingness to pay estimates became more similar for each of the three

methods.

The Boyle et al.’s method of bid designing uses WTP data from pre-test survey collected

through open-ended questions. The method applies the system of complementary random

numbers (Ehrenfeld and Ben-Tuvia, 1962; Kleijnen, 1975; Hillier and Lieberman, 1980) as a

sampling procedure. The objective of the method is to find a preliminary estimate of the

distribution of WTP values and minimize the potential for the cumulative density function to have

fat tail. Therefore, following the Boyle et al.’s method:

1. The expected value of the WTP/WTCL were calculated by integrating the area under the

curve 1 - F(X) and 1 - F(L) where F(X) and F(L) are the cumulative density functions

(c.d.f) of the willingness to pay and willingness to contribute labour values from the pre-

test survey ( Figure 5.2) ;

2. For the total sample size, N = 520, N/2 = 260 random numbers were generated as

probabilities (pi’s) from a uniform distribution on the interval [0, 1]; and derived another

260 calculated probability points (qi’s) where qi = 1 - pi. Therefore, for each pi we have got

corresponding value of qi.

3. To generate monetary and labour day offers that will be assigned randomly to

respondents of the final survey, the probability points were converted into monetary

values by multiplying them with the expected value from the empirical c.d.f. of the pre-test

data and rounded the product to even monetary and labour day amounts.

94

Figure 5.2: Distribution of willingness to pay in cash and willingness to contribute labour data from the

pre-test survey

Following the recommendations of Hennenmann and Kenninen (1998), Alberini (1995) and

Kanninen (1995) on the optimal bids design, we have selected five initial bid levels from the

generated monetary and labour offers and these selected initial bid levels are clustered near the

median of the distribution. The follow up bids for a yes and no responses for the initial bids were

taken by doubling and halving the initial bids respectively (Table 5.1).

Table 5.1: Bid design and number randomly assigned sample households

Bid :[WTP in ETB per year]; [WTCL in labour days per year]

[Initial, upper, lower]

Probabilities Sample

size

A: [19, 38, 10]; [2, 4, 1] 0.15 103

B:[47, 94, 24]; [5, 10, 3] 0.35 104

C:[68, 136, 34]; [7, 14, 4] 0.50 106

D:[84, 168, 42]; [9, 18, 5] 0.65 104

E:[113, 226, 57]; [12, 24, 6] 0.85 103

After incorporating the findings of the pre-test survey, determining the bid levels and random

assignment of bids to sample households, we trained interviewers and conducted the final face-

to-face survey in March 2010 (Figure 5.3).

Figure 5.3: Face-to-face contingent valuation survey (March 2010)

0

0.2

0.4

0.6

0.8

1

0 40 80 120 160

(1

-F(X

); (

1-F

(L)

WTP ('0 Birr ) and WTCL (days) per household per year

WTP WTCL

95

5.3.2. Model specification for measuring WTP and WTCL

The analysis of respondents’ choice is based on the random utility model (McFadden, 1974;

Hanemann, 1984) in which utility (Ui) arising from a “Yes” or “No” response to a contingent

valuation question is comprised of a deterministic (Vi), observable component and a random (i),

unobservable component:

)1(VU iii

If the household answers ‘Yes’ it receives the public good (BPF conservation program) and its

income is reduced by the amount of the bid. The index i reflects the alternatives Yes (i=1) and

No (i=0). The dichotomous choice elicitation model relies on the assumption that households

maximize their utility function. In this case, utility is assumed to arise from income (mj), the

presence of a BPF conservation program and other socio-economic characteristics. The jth

household will accept the randomly assigned initial bid only under the following condition

(Hanemann, 1984):

)2()B;S;m(v)B;S;tm(v j00jjj0j11j1j1j1j1

where, v1j is the indirect utility in a state of BPF conservation B1 and v0j is the indirect utility in

the status quo Bo. B1 is higher than Bo indicating higher non-market benefits from the forest in a

state of conservation than the status quo. The variable mj is income if the bid, tj, is in cash

whereas mj is leisure if the bid is in terms of labour. Sj captures other socio-economic variables.

0j and 1j are the identically, independently distributed random variables with zero means.

The jth household will say ‘Yes’ to the valuation question if the condition in equation (2) is

satisfied and the probability of a yes response will be:

)3()v(F

))B;S;m(v)B;S;tm(vPr(

))B;S;m(v)B;S;tm(vPr()Yes(Pr

j1j00jjj01jjjj1

j00jjj0j11jjjj1j

with =0j-1j, v=v1j-v0j, and F(v) as the cumulative distribution function of . If the indirect

utility, (vj) is assumed to be linear and depends upon income and socio-economic

characteristics when the bid is in cash, or alternatively, on leisure and socio-economic

characteristics when the bid is labour, then vj can be defined with and without the BPF

conservation program and the utility difference can be expressed as follows:

)4(t)(v

)m(v

)tm(v

j01

j0j0

jj1j1

96

where the vector Sj with the socio-economic variables is suppressed. The probability of a Yes

response becomes:

)5()t(F)1iPr( j

where =1-0. If a normal distribution for , the difference in the error terms, is assumed, then

the probability to say Yes becomes

)6()t()1iPr( j

where is the normal cumulative distribution function, is the parameter estimate of the bid

amount and is either the estimated constant (if no other explanatory variables are included in

the equation) or the “grand” constant which is computed as the sum of the estimated constant

plus the product of the other explanatory variables times their mean (Holmes et al., 2004).

Welfare estimates in the form of compensating surplus can be derived and when the bid results

in a single change in a policy option, the welfare estimate reduces to (Christie et al., 2004):

)7()vv(1

CV 01t

where t is the estimate of the marginal utility of money if the bid is expressed in cash or the

marginal utility of labour if the bid is expressed in labour contributions. Following Hanneman

(1984) who advocated the use of median WTP as a measure of economic welfare, the median

WTP was computed from the parameter estimates in equation 5 as WTPmedian=/.

While most empirical studies that used SBDC data assumed a logistic distribution of the

difference in error term and use the logit model for its simplicity to compute, this study assumed

normality of the difference in the error terms since aims to model each household’s two discrete

responses jointly using a bivariate normal probability density function. This function allows

testing for a non-zero correlation between initial and follow-up responses, whereas the standard

logistic distribution does not (Cameron and Quiggin, 1994). The hypotheses that the follow-up

response is independent of the initial response )0:H,0:H( Ia

I0 ) can be tested using the

likelihood ratio test. A restricted version of the bivariate probit model (i.e., if initial and follow-up

responses are assumed to be motivated by same latent WTP value, observed differences are

due to randomness in the WTP distribution, and the correlation coefficient 1 leads to a

double-bound interval data probit model (Hanemann et al., 1991). The hypotheses that the two

stochastic valuation functions of the bivariate probit model have identical distributions of error

terms )1:H,1:H( a0 so that they are agreeable to the DB probit analysis can also be

tested using the likelihood ratio test.

97

Empirical specification: The binary choice model in equation (6) can be estimated using a

probit model with the following specification:

)8()rSmt()Yes(Pr jjjj0j

where tj is the bid amount, mj is the household income for a bid in cash or leisure for a bid in

labour, Sj is a vector of variables measuring household socio-economic characteristics, rj is a

vector of regional dummy variables. The , , and are vectors of parameters related to

respectively the bid, household income or leisure, socio-economic characteristics and regional

dummies. We used STATA 11.0 for the statistical analysis.

The dependent variable in equation (8) is a binary variable taking the values of 0 or 1 depending

on the response to the randomly assigned predetermined bid levels. LogINC, AGE, EDU,

GENDER, and LABOUR are the variables designed to capture the sample households’ socio-

economic background (Table 5.2). Income refers to annual income from crop and livestock

production plus off-farm income after deducting costs. Age, education and gender refer to the

household head, while labour refers to the amount of labour available in the household. Most of

these variables have shown significant predictability in other related studies (Swallow and

Woudyalew, 1994; Echessah et al., 1997; Kamuanga et al., 2001; Gürlük, 2006; Hung et al.,

2007).

We expected a positive impact of LogINC on the likelihood to say yes to the contingent

valuation question. This is because economic theory tells us that conservation is like a normal

good, for which peoples’ demand increase with income. Unlike logINC, however, the study has

no ex ante expectations in terms of whether AGE and GENDER have a positive or negative

impact on the likelihood to accept the bid levels. We anticipated that higher education would

imply higher awareness and concern for degradation of natural resources, which should result

in higher WTP and WTCL. We anticipate that a large amount of family labour in a household

would imply a higher level of WTCL, because they are more likely to have relatively more labour

in excess of farming and other activities compared to households with a small amount of family

labour.

The dummy variable RESID is expected to have a positive impact on the likelihood to accept the

bid levels in both the WTP and WTCL. We anticipate that those who have been living longer in

the area would be more concerned with the existing state of degradation than the newcomers,

who are in fact contributing to the degradation of the resource in the course of the resettlement.

The variable USER is expected to have a positive impact on the likelihood to say yes for the

valuation questions of WTP and WTCL. Because the RESID variable was highly correlated with

the age and education level of the household head, the interaction term EduRESID and

AgeRESID are included in the model to capture the pure EDU and pure RESID as well as the

98

pure AGE and pure RESID effects on the response variable. We include dummies for each

villages to capture village fixed effects like differences in natural resource endowments and off-

farm employment opportunities that could affect household income and hence WTP and WTCL.

Table 5.2: Description and summary statistics of socioeconomic characteristics of sample households of

the study area (N = 473)

Variables Description Mean(SD)

BID: WTP Randomly assigned amount to each household in ETB per year 65.89(32.03)

: WTCL Randomly assigned amount to each household in days per year 6.97(3.41)

logINC Log transformed household income 9.44(0.92)

LABOUR Number of household members 15-64 years old 2.70(1.30)

AGE Age of the household head 40.49(12.46)

EDU Education of the household head: 1=Illiterate, 2=Read and write but no

formal schooling, 3= Elementary, 4= Junior High school, 5= High school

2.12(1.10)

GENDER Gender of the household head: 1=Male; 0=Female 0.91(0.29)

RESID 1= The household head has been living in the area since before the

resettlement year (2002); 0 = otherwise

0.72(0.45)

USER 1= if the household has been using the resource as either fodder, tapper ,

or member of frankincense cooperative firms; 0 otherwise

0.54(0.49)

EduRESID Interaction term for EDU and RESID 1.43(1.27)

AgeRESID Interaction term for AGE and RESID 30.43(21.97)

VIL1 Village dummy: 1= Adi Aser, 0=otherwise 0.20(0.40)

VIL2 1= Adigoshu 0=otherwise 0.35(0.48)

VIL3 1= Moguu, 0=otherwise 0.14(0.35)

VIL4 1= Jijike, 0=otherwise 0.16(0.37)

VIL5 1= Siye, 0=otherwise 0.15(0.36)

DEPEND Number of household members <15 and >64 years old 2.59(1.60)

LAND Land size in hectares 2.32(1.71)

DisAWRoad Walking distance to all weather roads in hours 1.89(3.93)

5.3.3. Data calibration

Before using the data for statistical analysis, we first calibrated it by dropping protest responses.

In the CV literature, reasons other than financial constraints and the good having no value to the

household are protest responses (Labao et al., 2008). In the WTP bidding, 35 households

(6.7%) answered no/no. Among these, 16 (3.1%) replied that they have some interest in the

conservation but would not pay anything to join the program. These could be free riders and

classified as protest responses. In case of WTCL bidding, we found 51 ‘no/no’ (9.8 %) of which

38 (7.3 %) were protest responses. From the total protest responses of both biddings, 7

households answered ‘no/no’ to both biddings, 9 households only to the WTP bidding, and 31

households only to the WTCL bidding. Therefore, 47 households respond ‘no/no’ at least to one

of the biddings. Therefore, in order to see the effect of each covariate on WTP and WTCL using

99

the same sample, we considered them as protest responses. After the protest screening, we

were left with 473 households for statistical analysis. We further calibrated the data using the

asymmetric uncertainty model (Champ et al., 1997) to correct for potential hypothetical bias.

Households who replied yes/yes, yes/no and no/yes to initial and follow-up bids were asked

how certain they were in making the payment. We used a ten-point scale with 1 as ‘Very

uncertain’ and 10 as ‘Very certain’. We classified all yes/yes, yes/no and no/yes responses as

no/no responses if the household chooses certainty scores 1 to 9. Accordingly, we classified

responses of 181 households in the WTP and 158 households in the WTCL as ‘no/no’

responses.

5.4. Results

5.4.1. Households’ knowledge and attitude

The survey showed that a majority of the households (68.7%) had previous knowledge about all

or most of the uses. A little more than a quarter (28.5%) reported that they had previous

knowledge about few of the uses but were able to learn more from the interview and understood

all uses very well. Almost all (99.2%) reported that they would prefer to transfer the Boswellia

papyrifera forest areas in their village to their grandchildren in good state. All households

reported that they are concerned about the deforestation and the risk of extinction of which the

majority (66.4%) are very concerned. To elicit households’ attitudes on the conservation goals,

they were asked to state their interest on a scale from 1 to 10, with 10 as ‘Very interested’ and 1

as ‘Not at all interested’. Nearly one-third (65.0%) attributed an option value for the resource as

they were very interested in keeping it as an option of meeting future demands. Almost the

same proportion (66.0%) recognized an existence value for the resource, as they were very

interested in realizing its existence and curbing the possibility of its extinction. The majority

(72.7%) observed a bequest value for the resource, as they were very interested in conserving it

for improving the welfare of future generations. No household selected scale 1 in any question

implying that all households attributed option, existence and bequest values for Boswellia

papyrifera forest with different degrees of interest. A majority (72.1%) were very much interested

in the overall conservation policy.

5.4.2. Parameter estimates of WTP and WTCL

To estimate equation (8), we used four econometric models for each of the WTP and WTCL

responses (Tables 5.3 and 5.4). Therefore, two of the models (Models I and V) were single-

bounded dichotomous choice (SBDC1) models for the initial response on the WTP and WTCL

question. Two (Models II and VI) were SBDC2 models for the follow-up response, two (Models

III and VII) were bivariate probit models, and the other two (Models IV and VIII) were double-

bounded interval data probit models (DB).

100

The BID term is statistically significant at 1% for all the models and has the expected negative

sign consistent with the theory of demand. This indicates that, in the case of both WTP and

WTCL, the higher the BID price for the Boswellia paprifera forest conservation, the less likely

households would be willing to pay and contribute labour respectively. The variable LogINC has

the expected positive sign and is significant for all the WTP models. This implies that the

probability of accepting a higher bid increases with income. In other words, conservation is like

a normal good to the rural households for which their demand increases with an increase in

income. In the case of the WTCL models, the coefficient of LABOUR is positive as expected

indicating that the larger the number of family members in the productive age group, the higher

the probability of accepting a higher bid. Soil and water conservation works have been common

practices in Tigray for which rural households make free labour contributions every year.

Table 5.3: WTP Models: Single bound discrete choice probit (Model-I and Model-II), bivariate probit

(Model-III) and a double-bound interval data probit (Model-IV) models of rural households’ willingness to

pay in cash for Boswellia papyrifera forest conservation

Variables Model-I Model-II Model-III Model-IV

BID -0.022(0.002)a -0.007(0.001)

a -0.021(0.002)

a -0.195(0.055)

a

LogINC 0.842(0.109)a 0.710(0.098)

a 0.828(0.105)

a 5.832(1.587)

a

AGE 0.010(0.015) 0.007(0.014) 0.007(0.014) 0.048(0.109)

EDU 0.266(0.138)c 0.088(0.119) 0.202(0.126) 1.874(1.088)

c

GENDER 0.440(0.255)c 0.167(0.234) 0.359(0.238) 1.156(1.755)

RESID 1.226(0.780) 1.055(0.707) 0.944(0.736) 6.890(5.680)

USER -0.008(0.148) 0.008(0.136) 0.004(0.145) -0.032(1.065)

AgeRESID -0.018(0.016) -0.014(0.015) -0.013(0.016) -0.100(0.120)

EduRESID -0.319(0.159)b -0.177(0.140) -0.254(0.149)

c -2.173(1.253)

c

VIL1 0.281(0.264) -0.027(0.245) 0.323(0.259) 2.090(1.933)

VIL2 0.443(0.257)c 0.478(0.240)

b 0.442(0.257)

c 3.006(1.957)

VIL3 0.349(0.280) 0.330(0.261) 0.454(0.278) 3.300(2.147)

VIL4 0.104(0.249) -0.102(0.235) 0.146(0.253) -0.029(1.786)

Constant -7.817(1.162)a -6.716(1.028)

a -7.470(1.102)

a -50.081(14.453)

a

LogL -212.192 -252.080 -356.800 -266.908

Pseudo R2 0.347 0.229

0.938(0.020)a

Median WTP* 5.73 9.70 5.68 4.82

Mean WTP* 5.71(0.15) 10.06(0.41) 5.62(0.15) 4.86(0.12)

95% CI* 5.4 to 6.01 9.26 to 10.85 5.32 to 5.92 4.63 to 5.10

*Values are in USD (1USD = ETB 13.3778), values in ( ) are standard error, a= significant at p<1%, b= at p< 5%, c=

at p< 10%.; VIL5 is dropped due to collinearity.

GENDER is positive for all the models and significant at 5% for model VI and at 10% for models

I, V and VII. This indicates that the likelihood to accept higher bids mainly in terms of labour is

higher for male-headed households than female-headed households. AGE has a negative sign

101

for all models of WTCL, but only significant at 10% for model VII. This implies that the older the

household head the lower the probability of accepting higher bids in terms of labour. The

variable RESID has no significant impact on WTCL, but the interaction term AgeRESID has a

positive sign and is significantly different from zero in all WTCL models. Chi2-test results (0.58,

p=0.447 for model V; 0.12, p=0.724 for model VI; 0.41, p=0.521 for model VII; and 0.50,

p=0.478 for model VIII) revealed that the sum of the coefficients of AGE and AgeRESID is not

significantly different from zero. This implies that older households who immigrated into the

community as settlers after 2002 are less likely to accept a higher bid than older household

heads who have been already living in the community before 2002. This might be because the

older households who have been living in the areas for a long time have also been observing

the degradation of the resource relatively for a long time than the older settlers who have come

to the areas. This might have created a difference in their understanding of the problem and

hence difference in their WTCL.

Table 5.4: WTCL models: Single bound discrete choice probit (Model-V and Model-VI), bivariate probit

(Model-VII) and a double-bound interval data probit (Model-VIII) models of rural households’ willingness

to contribute labour for Boswellia papyrifera forest conservation.

Variables Model-V Model-VI Model-VII Model-VIII

BID -0.180(0.021)a -0.072(0.010)

a -0.175(0.020)

a -2.194(0.640)

a

LABOUR 0.210(0.059)a 0.185(0.055)

a 0.200(0.058)

a 1.948(0.748)

a

AGE -0.021(0.014) -0.021(0.013) -0.023(0.013)c -0.217(0.141)

EDU 0.115(0.109) 0.083(0.108) 0.099(0.108) 0.642(1.036)

GENDER 0.395(0.227)c 0.509(0.221)

b 0.388(0.220)

c 2.050(2.250)

RESID -0.677(0.696) -0.533(0.672) -0.843(0.672) -7.833(6.757)

USER -0.164(0.135) -0.103(0.129) -0.155(0.133) -1.313(1.322)

AgeRESID 0.026(0.015)c 0.023(0.014)

c 0.028(0.014)

b 0.264(0.152)

c

EduRESID 0.021(0.132) -0.038(0.129) 0.053(0.131) 0.415(1.245)

VIL1 -0.079(0.230) -0.184(0.217) -0.107(0.226) -0.717(2.192)

VIL2 0.041(0.233) 0.227(0.221) 0.033(0.232) 1.033(2.245)

VIL3 0.075(0.251) 0.058(0.236) 0.061(0.247) 1.197(2.416)

VIL4 0.082(0.236) 0.248(0.225) 0.122(0.237) 0.230(2.262)

Constant 0.992(0.688) 0.453(0.648) 1.148(0.667) 13.360(7.162)c

LogL -259.421 -284.195 -428.594 -312.608

Pseudo R2 0.182 0.128

0.919(0.022)a

Median WTCL 8.56 14.64 8.72 7.02

Mean WTCL 8.78 14.59 8.87 7.17

95%CI 8.59 to 8.97 14.14 to 15.04 8.69 to 9.06 7.04 to 7.29

Values in ( ) are standard errors, a= significant at p<1%, b= significant at p< 5%, c=significant at p< 10%.; VIL5 is

dropped due to collinearity.

102

EDU has the expected positive sign for all models, but is only significant at 10% for models I

and IV. This implies that the higher the level of education of the household head, the higher the

probability of accepting higher bids in cash. The interaction term EduRESID has a negative sign

in all WTP models and is significantly different from zero in models I, III and IV. Chi2 -test results

(0.41, p=0.523 for model I; 0.40, p=0.527 for model III; and 0.25, p=0.615 for model IV) revealed

that the sum of the coefficients of EDU and EduRESID is not significantly different from zero.

This implies that higher educated households who immigrated into the community as settlers

after 2002 are more likely to accept a higher bid, while this does not hold for higher educated

household heads who were already living in the community before the resettlement in 2002.

This might be because severe environmental degradation at the origin is one of the main push

factors for the migration of the settlers and they have had experience with the negative

consequences of environmental degradation on their livelihood. Therefore, higher educated

households who migrated into the community would be more willing to contribute for the

conservation compared to earlier residents with a higher level of education for they have been

living in a relatively better environment, and at least were not displaced due to environmental

degradation. USER is negative for all the models except models II and III and insignificant in all

the models indicating no difference in the probability of accepting higher bids between user and

non-user households.

In both the WTP and WTCL models, the coefficient on the bid term in the follow-up response

model (model II and VI) is much lower in absolute value than the coefficient of the bid term in

the initial response model (model I and V). As a result, the spread parameter, which is the

inverse of the negative of the coefficient on the bid term, of the follow-up response probit model

is higher (by 228.59% in the WTP models and 152.04% higher in the WTCL models) than the

spread term of the initial response probit model. This indicates that if we fit distinct valuation

functions to each response, the follow-up response is much ‘noisier’ than the initial. This may be

due to the presence of some strategic responses in the follow-up. The likelihood ratio test result

(Chi2 = 79.78; p= 4.19E-19 for model I Vs II) for the WTP models confirms this fact and shows

that the restriction that the follow-up response (model II) is independent of the initial response

(model I) is not valid and hence is rejected. Similar results were found for the WTCL models

(Chi2 = 49.55; p = 1.94E-12 for model V versus VI). The statistical significance of the correlation

coefficients for models V and VII also confirm these outcomes.

The hypothesis 1:0 H cannot be rejected both in the case of the WTP models (Chi2 = -179.78;

p = 1 for models III Vs IV) and the WTCL models (Chi2 = -231.97; p = 1 for models VII versus

VIII) suggesting the DB models would lead to more efficient estimates of WTP and WTCL than

the bivariate models. Nevertheless, the results suggest that the SBDC models I and II have high

103

predictive power and statistical reliability. The percentage of correct prediction in these models

is reasonably high.

In all model of WTP and models VI and VII of WTCL, the median values are within the 95%

confidence interval of the mean implying no statistically significant difference between mean

and median values. In the case of both median and mean values of WTP and WTCL, the SB

models give higher estimates than the DB models and the differences are statistically

significant.

5.4.3. Robustness tests

In order to check the robustness of our results, we did some robustness tests. We treated

household income as endogenous anticipating unobservable factors affecting a respondent’s

discrete choice response for the WTP questions may also affect household income and applied

IV-Probit (Probit model with endogenous repressors) to solve the problem of endogeneity using

DEPEND, LABOUR, LAND and DisAWRoad as instrumental variables. These variables do not

directly affect WTP, but indirectly through their effect on income. The correlation of each of

these variables with the error term of equation 9 is statistically insignificant (r = -0.0060, p <

0.897 for DEPEND; r = -0.0095, p< 0.837 for LABOUR; r = 0.0037, p<0.936 for LAND; and r =

0.0144, p<0.754 for distance to all weather roads). LAND, LABOUR, and DEPEND are

important inputs in agricultural production and hence affect income positively. DisAWRoad

captures access to lower prices in the agricultural input markets and higher prices for the output

markets, which have both a positive effect on household income. All of these variables have

statistically significant correlations (all at p< 0.001) with log transformed household income (r

=0.1576 for DEPEND; r = 0.1474 for LABOUR; r= 0.3523 for LAND; and r = -0.3658 for

DisAWRoad).

In the IV Probit model we found no evidence of endogeneity of logINC (/athrho = 0.076; chi2 =

0.696). All the coefficient estimates of the model are almost similar, in both magnitude and

significance, to those in Model I. To see potential multicollinearity problems due to the inclusion

of logINC together with variables capturing the household characteristics, we ran a probit model

in which income is excluded (SBDC1a). This resulted in more number of significant variables

compared to model I. Therefore, in order to investigate causality and disentangle the effect of

household characteristics on the one hand and income on the other, we implement the

following two-stage procedure. In the first stage, we did a regression of logINC on some

household characteristics. Next, we saved the residuals under the variable logINCcorr. This

variable is by construction orthogonal to the household characteristics. In a second stage, we

reran the SBDC1b of the form displayed in Table 5.5, but we replaced logINC with logINCcorr.

The coefficient estimates of income in model I of Table 5.3 and that of logINCcorr in the

104

SBDC1b in Table 5.5 are almost the same. Moreover, the confidence intervals for each

corresponding coefficient estimates of model SBDC1b model I overlap each other implying no

statistically significant differences. This is also evidence that WTP does not merely depend on

household characteristics, which are also affecting income and attitude towards conservation of

the Boswellia paprifera forest, but that the WTP for BFP conservation is strongly dependent on

the financial means of the household as such.

Table 5.5: Models for checking robustness of the WTP models

Variable IV-Probit SBDC1a SBDC1b

BID -0.022(0.002)a -0.019(0.002)

a -0.022(0.002)

a

LogINC 0.749(0.267)a 0.844(0.115)

a

AGE 0.011(0.015) 0.013(0.014) 0.017(0.015)

EDU 0.288(0.148)c 0.355(0.125)

a 0.455(0.139)

a

GENDER 0.492(0.286)c 0.752(0.234)

a 0.907(0.253)

a

RESID 1.282(0.791) 1.312(0.731)c 1.727(0.781)

b

USER 0.010(0.155) 0.128(0.136) 0.164(0.146)

AgeRESID -0.019(0.017) -0.019(0.016) -0.025(0.017)

EduRESID -0.336(0.165)c -0.367(0.146)

b -0.469(0.160)

a

VIL1 0.386(0.374) 1.089(0.231)a 1.248(0.247)

a

VIL2 0.521(0.323) 1.054(0.233)a 1.158(0.248)

a

VIL3 0.437(0.357) 1.022(0.255)a 1.150(0.270)

a

VIL4 0.118(0.251) 0.265(0.236) 0.235(0.247)

Constant -7.127(2.150)a -1.231(0.713)

c -1.589(0.764)

b

LogL -718.447 -248.525 -216.592

Pseudo R2 0.236 0.334

Values in ( ) are standard errors, a= significant at p<1%, b=significant at p< 5%,

c= significant at p< 10%.; VIL5 is dropped due to collinearity.

5.5. Discussion

The aim of this research was to assess rural households’ demand for Boswellia papyrifera

forest conservation and identify the factors determining their willingness to pay and willingness

to contribute labour. For this, we designed a contingent market situation in which the benefits

include option, existence and bequest values of the resource. The results give insight about

these values of the resource for the rural population and their demand for conserving the

resource. Almost all respondents attributed option, existence and bequest values for the

resource and the majority (72%) were very much interested in conserving Boswellia payrifera

forest in their villages.

The study shows that the face-to-face contingent valuation survey worked well as revealed in

terms of high survey response rate (100%) and low protest rates (3.1% for WTP and 7.3% for

105

the WTCL responses). The protest rates are considerably lower than the protest rates for most

contingent valuation surveys in developed countries. Meyerhoff and Liebe (2010) did a meta-

analysis of 254 samples from stated preference surveys of environmental valuation, and

reported mean a protest rate of 14.19% for 15 studies for which forest resources are the subject

of the valuation, and mean of 17.97% protest rate for 108 studies with dichotomous choice

format. Hung et al. (2007) also found a very low (3%) protest rate for forest fire prevention in

Vietnam compared to related studies in developed countries.

All the probit regression analyses revealed that the probability of a ‘Yes’ response to the

willingness to pay and willingness to contribute labour biddings vary with a number of

covariates in a reasonable and expected manner, thereby offering some support for the

construct validity of our contingent valuation applications. We found that the variables bid, log-

transformed income, education, and the interaction of education and residence length as the

most important factors affecting willingness to pay. The variables bid, labour, gender, and the

interaction term of age and residence length are the most important factors affecting willingness

to contribute labour. The results of the statistical tests for independence of initial and follow-up

responses as well as on the correlation of error terms of the bivariate probit models indicated

that the double-bounded model gives more efficient estimates than the single-bounded model

and this result is consistent with the findings of Hanemann et al. (1991).

The monetary value of lower bound mean willingness to contribute labour (7.17 days per

household per year) at the market wage rate was 23.34 USD with 99% confidence interval of

22.21 to 24.46 USD which implies a very large asymmetry with the lower bound mean

willingness to pay in cash (4.86 USD per household per year). We found similar result for the

upper bound mean willingness to contribute labour. Results from previous studies (Swallow and

Woudyalew, 1994; Echessah et al., 1997; Kamuanga et al., 2001; Hung et al., 2007) that used

labour and cash contributions for eliciting willingness to pay for natural resource management

interventions with benefits of public good character indicated that respondents were willing to

contribute more in labour than in money. Specifically Echessah et al. (1997) found higher mean

willingness to pay in labour converted at the market wage rate for casual labour than the mean

willingness to pay in cash for tsetse fly control in Ethiopia. Eom and Larson (2006) suggest that

higher willingness to pay in terms of non-monetary payment vehicles than the monetary

payment could be linked to low valuation of time and hypothetical bias. Another reason could be

that liquidity constraints could force respondents to pay more in terms of labour than cash,

which is a reflection of market imperfections restricting substitution among different resource

endowments (Vondolia, 2011). Hung et al. (2007) argue that higher willingness to pay in labour

may be due to prior experience of such payments.

106

It will be correct to value labour contributions at the market and/or minimum wage rates in case

the government intends to implement such a program for the fact that these rates are what it

would have paid. However, converting labour contributions at the market wage and/or minimum

wage may bias our conclusion for the following reasons at least in the context of households

characterized by subsistence farming. First, the market for daily labour is imperfect with few

employers than the supply in which rural households will not have perfect employment

opportunity. Therefore, in making decisions on the labour amount they want to contribute it is

less likely that they compare it with the market wage as opportunity cost of their decision.

Secondly, there is no minimum wage rule in some developing countries as it is in the developed

world. If it does exist, rural households may not know how much this minimum wage is and use

it to calculate the opportunity cost their choices on the amount of labour they would to contribute

for the contingent market. Third, in the contingent valuation survey with money as payment

vehicle respondents are reminded to take into account their income and expenditure before the

valuation question. Similarly, for labour as payment vehicle they need to be reminded of the

total labour time they have and the amount they require to undertake their household activities

(farming and community works, etc…). This information is the basis for them for comparing the

opportunity cost of their choice. Labour time in the off-farm labour market might be part of this

for some respondents but in general, rural households allocate much of their labour on their

subsistence farming activities than off-farm labour markets. For those who participate, they

mostly do this after the crop season and if they could not get employment in this period, the

opportunity cost of labour will then be zero. Therefore, it makes more sense to use the per

capita daily income of the respondent himself instead of the market or minimum wage rates as

the opportunity cost of labour and test the convergent validity of willingness to pay and

willingness to contribute labour values. Furthermore, if we think of a CV survey in a developed

world with money and labour time contributions as payment vehicles is to take place,

respondents would most probably choose to pay more in cash than in labour. In doing so a

rational respondent compares the opportunity cost of his/her time valued not at a flat market

wage rate but at his own personal earning capacity per unit of time, which will either be large or

smaller than the market wage rate.

Therefore, we did a comparison of willingness to pay in cash with willingness to contribute

labour converted at the per capita daily income, the minimum and market wage rates. For the

minimum wage, Ethiopia has not yet a law for minimum wage. However, public employers and

private companies set their minimum wages. We took the public employers minimum wage

(320 ETB per month) for our calculation. The monetary value of the lower bound mean

willingness to contribute labour valued at the per capita daily income was 6.64 USD and this

value is much closer to the lower bound mean willingness pay in cash. Moreover, there is an

overlap in the 99% confidence intervals of the two values (4.97 to 8.31 versus 4.55 to 5.18).

107

The monetary value of the lower bound mean willingness to contribute labour at the minimum

daily wage rate is 5.72 USD however; the 99% confident interval for this value is (5.58 to 5.85)

and does not overlap with the corresponding confidence interval of the mean willingness to pay

in cash.

5.6. Conclusions

This study used labour and cash as payment vehicles to measure the non-market benefits of

environmental amenities to income constrained rural households in Ethiopia. Our results

suggest that labour time contribution valued at the per capita daily income of the respondent is

more appropriate to test the convergence validity of willingness to contribute labour and

willingness to pay in contingent valuation studies in subsistence economies. From the point of

view of the conservation and sustainable management of Boswellia papryifera forest, the study

indicated that despite Ethiopia is a low-income country, people are clearly aware of the

importance of conservation of the resource and that they are indeed interested to contribute for

the conservation. However, mobilizing this local demand for conservation is only part of the

solution to the problem, which requires improving their livelihood.

Therefore, improving the livelihood of the rural communities through increasing the benefits from

the resource on the one hand and managing the resource on a sustainable basis on the other is

the challenge that calls for national and global level policy interventions. At the national level,

increasing the capacity of rural farmers’ cooperatives by providing the necessary training and

credit facilities so that they can directly engage themselves in the export market could increase

their share of the benefits from the resource. It is also important to build a local and national

level skilled work force in forestry and related fields that is now a critical problem in the country.

The other possible area of policy intervention could be the development of forest-certification

that promotes the sustainable use and management of the resource so that the international

market for oleo-gum resins could provide price premiums for resins from sustainably managed

forests.

Frankincense forests in particular and oleo-gum resins dry land forests in general could also be

worth investing if included in global REDD+ mechanisms for both poverty reduction, biodiversity

conservation and for reducing emissions from deforestation and forest degradation. However, a

sound justification for such policy interventions may require more research on values of all the

ecosystem services from the resource (carbon stock and sequestration, opportunity cost of the

forestland, and direct use values from oleo-gum resin production). This research only addresses

the option, existence and bequest values of the resource to the local people.

108

CHAPTER 6

Impact of membership in rural frankincense cooperatives on rural

income and poverty

6.1. Introduction

There is a growing understanding on the role of forest resources as income source to rural

people in developing countries. In their review of 51 case studies from 17 developing countries

in Africa, Asia, and Latin America Vedeld et al. (2007) indicated that forest environmental

income accounts for 22% of the total income of the sampled population. In relative terms, the

poorest of the rural poor depends very much on forest environmental income. From his poverty-

environment relation study in Zimbabwe Cavendish (2000) indicated that the lowest income

quintile earn a little more than 40% of their income from forest resources and reported 29% for

the highest income quintile. Mamo et al. (2007) reported a 59% of forest environmental income

to the lowest income quintile and 30% to the relatively wealthiest group of rural households in a

study conducted in the mid highland part of Ethiopia.

The inclusion and analysis of forest environmental income in a household survey are very

important in analyzing the extent of poverty, its causes and find out policy measures that could

alleviate poverty (Sjaastad et al., 2005). Studies argue that rural poverty surveys that do not

take into account the forest environmental income will underestimate rural income and welfare

(Cavendish, 2000; Vedeld et al., 2007; Babulo et al., 2009). In addition, it will overestimate the

poverty level and exaggerate the inequality among the survey units (Reddy and Chakravarty,

1999). Moreover, including environmental income in poverty surveys will also provide

information on the effect of any policy measures (conservation and management) on the rural

households whose livelihoods depend on the resource (Cavendish 2000; Mamo et al., 2007).

Most of the literature on the role of forests on rural livelihood associate non-timber forest

products with lower incomes, subsistence strategies and as income diversification to reduce risk

(Cavendish, 2000; Angelsen and Wunder, 2003; Belcher and Kusters, 2004; Vedeld et al.,

2007). However, economically important tree species with high value export non-timber forest

products would have a beneficial impact on rural income and poverty reduction given that rural

households are involved actively in the supply chain of such relatively commercialized non-

timber forest products. Frankincense from Boswellia papyrifera is one such high value non-

timber forest product that Ethiopia exports. According to data from Guna Trading Share

Company, the free on board (FOB) price for first grade (G1A) frankincense in 2010 was about

$4000 per ton and the price for the lowest grade G5, which is only sold at the local market, was

110

$1120 per ton. The wage for the ‘grezo” (ungraded frankincense) tapping was $ 518 per ton

(Table 4.2 in chapter 4).

Land and forest resource in Ethiopia are state properties. According to the Forest Development,

Conservation and Utilization Proclamation No. 542/2007 of the Federal Democratic Republic of

Ethiopia, state forests that could be developed on concession shall be given to community,

associations or investors on concession so that they conserve and utilize them in accordance

with government directives. In the case of Boswellia payrifera forest, both private and state

companies take forestland on concessionary basis and pay royalties to the government. These

companies create seasonal employment for the tapping, collection, and grading frankincense.

At the country level, natural gum, resin, and frankincense tapping and collection activities create

seasonal employment opportunities for 20,000 to 30,000 people (Shackleton and Gumbo, 2010,

Girma, 1998), who are mainly poor young migrant workers from the highland areas. Resident

rural households in the low land Boswellia forest areas have not been benefiting from such

seasonal employment opportunities due to a number of constraints. Lemenih et al. (2007)

identified a number of constraints that hinder resident farm households that not to involve in

frankincense production in north western Ethiopia. These constraints include cultural influence

that considers the job as inferior and culturally unacceptable, unattractiveness of the wage rate

relative to income from agriculture, property tenure and government policy that prohibit

individual household level ownership of Boswellia forestland as well as incense production for

sale, and lack of the skill of tapping frankincense.

Therefore, resident rural households have not been benefiting for long either form the seasonal

employment or from the rent that is generated from marketing the product. Reducing poverty

levels of resident rural households in Boswellia woodland areas of the country requires among

other things providing people’s access to productive resources for income generation. Rural

cooperatives have been cited as a goal in rural participation for rural development processes

(Aref and Sarjit, 2009). They are voluntary business associations formed by people of limited

means through contribution of share capital that forms the basis of sharing out the profits that

accrue from the business (Wanyama et al., 2008). Rural cooperatives also provide opportunities

for organizing particular local economic interests and/or for protecting common pool natural

resources (Simmons and Birchall, 2008). Moreover, they have significantly contributed to the

mobilization and distribution of financial capital by creating employment and income generating

opportunities for both their members and non-members alike, given that membership is open to

all persons without class, ethnic, or professional biases (Wanyama et al., 2008; Aref, 2011).

Based on the values, principles, and essential nature of rural cooperatives and their historical

positive achievement in poverty alleviation in developed countries, Birchall (2004) argued that

111

despite the presence of failures in cooperatives in developing countries, rural cooperatives have

a great potential role for poverty reduction in developing countries.

Over the last few years, there are efforts, particularly in northern Ethiopia, to organize rural

households in the form of cooperatives and engage them as new actors in the frankincense

supply chain (see section 4.4.4 and figure 4.2 in chapter 4). These cooperatives get forestland

from the local government on concession and pay royalty to the government. They undertake

the collection of frankincense by directly employing tappers. They also sell their harvest to

exporting companies on through auctioning or a direct sale based on market price. Rural

household members of the cooperatives get income in the form of dividend from the profit that

their cooperatives make. Thus, this study aimed to assess the impact of membership in

frankincense rural cooperative firms on rural household’s income and incidence of poverty and

contribute to the limited literature on the beneficial impact of non-timber forest products on rural

income and poverty reduction.

6.2. Materials and methods

6.2.1. The data

To study the welfare impact of rural frankincense cooperative firms, which were nonexistent

prior to 2002, a household survey was conducted in Boswellia papyrifera forest areas of

northern Ethiopia household survey (Appendix 4A, sections III-V of the survey) in March 2010

as discussed in section 4.4.3 of chapter 1 and section 5.3.1 of chapter 5. Accordingly, 120

cooperative members and 400 non-member sample households were selected for the study.

Before using the data for the econometric analysis, samples with missing values in either the

covariates or response variable are considered as non-responses and the sample size was

reduced to 511, of which 120 are members and the rest are in the control group. If data were

missing for an observation on either the dependent or independent variables, then the

observation could not be used for empirical analysis. There are two approaches in the literature

on dealing with missing values (Little, 1992; Little and Rubin, 2002; Horton and Kleinman, 2007;

Dardanoni et al., 2011). One is to use imputed values constructed by the researcher or provided

by the data producing agency and the other is to use the observations with complete data using

the methods of list wise (case wise) and/or pair wise deletion of missing values. However, there

is a trade-off between these approaches and imputing data increases precision but may bias

the results (Dardanoni et al., 2011). If missing data were not behaviour related, the only problem

of using only the observations with complete data would be that the reduced sample size would

make the standard estimator less precise, while no bias would be induced (Hsiao, 1979; Liu et

al., 2009).

112

The structured household survey (Appendix 4A, sections III-V of the survey) data include

demographic characteristics, size of land holding, access to irrigation, non-agricultural

consumer durables, non-land agricultural assets, agricultural production from owned, rented in

and rented out land, livestock product income, off-farm labour income, dividend from

frankincense cooperative membership, forest environmental income (income from other non-

timber forest products), and recall data on asset holdings prior to the introduction of

cooperatives in the study area. In addition to the structured questionnaire, data on areas of

Boswellia forestland in each village and number of cooperative firms at district level were

collected from respective local government offices.

6.2.2. Frankincense cooperative membership, rural income and poverty:

descriptive analysis

Frankincense cooperative membership: Rural frankincense cooperative firms were

established in the study area in 2002. They undertake harvesting of frankincense from the forest

under their concession. According to interviews with leaders of the cooperative firms in the

study villages and data from the survey, membership requires contributions in the form of

buying of shares. Profits of the cooperatives are distributed to members as dividends

proportional to their investment in shares. A member has to purchase a minimum of 1 share

with a value of ETB 50 and we found that the maximum share value per member household in

our samples is ETB 3000, which is equivalent to 60 shares. Similar to the other companies

involved in the sector, rural cooperative firms also employ tappers on wage basis for the

collection of the product from the forest under their concession.

Table 6.1 compares the characteristics of cooperative member (treatment group) and non-

member (control group) households. Although households in the treatment group are older,

have more family members in the productive age group as well as larger dependency ratio than

the control group households, and have more access to irrigation, the differences are not

statistically significant. In terms of education, proportion of female headed households as well

as distance to all weather roads, and access to irrigation the means for the control group are

larger than the treatment group but the differences are not significant. However, the two groups

have a statistically significant difference in terms of their asset holdings prior to the introduction

of frankincense cooperative firms. Households in the treatment group have higher initial per-

adult equivalent landholdings, tropical livestock units, non-land assets, and consumer durables

than the control group households. In addition, the proportion of settler households among the

treatment group is less than the proportion in the control group and the difference is statistically

significant. Therefore, this comparison suggests that membership in frankincense cooperative

firms is not biased in terms of demographic characteristics, but is biased towards the better-off

113

households in terms of larger initial per-adult equivalent asset holdings and to those of original

residents than the settlers.

Table 6.1: Summary statistics and comparison of household characteristics

All ( N=511)

Cooperative members (n=120)

Cooperative non-members (n=391)

t-value

Mean SE Mean SE Mean SE

Income in (1000 ETB) by source

Total Income 14.45 0.60 19.49 1.51 12.90 0.62 4.73a

Dividend 0.37 0.08 1.59 0.30 0.00 0.00 9.56a

Forest environmental incomed 1.97 0.08 2.26 0.18 1.88 0.09 1.96

c

Off-farm incomee 2.43 0.25 3.37 0.72 2.14 0.24 2.09

b

Livestock product incomef 1.79 0.16 2.37 0.26 1.61 0.20 1.97

b

Crop income 7.89 0.44 9.89 1.03 7.27 0.47 2.57b

Poverty rate 0.33 0.02 0.16 0.03 0.39 0.02 4.77a

Human and social capital

Labour 2.68 0.06 2.83 0.13 2.63 0.06 1.41

Dependency ratiog 1.06 0.03 1.11 0.07 1.05 0.04 0.77

Female headed households 0.11 0.01 0.09 0.03 0.11 0.02 -0.57

Household head with at least elementary education

0.39 0.02 0.38 0.04 0.39 0.02 -0.06

Age of the household head 41.38 0.57 42.83 1.11 40.93 0.66 1.41

Household head is settlerh 0.32 0.02 0.24 0.04 0.34 0.02 -2.03

a

Walking distance to all weather roads in hours

1.88 0.17 1.52 0.42 2.00 0.19 -1.15

Initial per-adult equivalent assetsi,j

Landholdings (ha) 0.82 0.04 1.06 0.10 0.75 0.05 3.13a

Tropical livestock unitsk 3.35 0.20 4.53 0.57 2.99 0.19 3.31

a

Non-land assets (1000 ETB)l 0.27 0.07 0.71 0.31 0.14 0.02 3.26

a

Consumer durables (1000 ETB)m 0.29 0.02 0.46 0.06 0.24 0.02 4.09

a

Irrigationn 0.10 0.01 0.13 0.03 0.09 0.01 1.40

Statistical significance: a = p < 1%; b= p<5%; c=p<10%; dIncome from non-timber forest products which include fuel wood for home

consumption and for sale, food (fruits, cactus, honey), traditional medicinal plants, wild washing soda, materials for making

household equipments (fibers used to make sweeping material, baskets, woven mats, woven hats); eIncludes income as local

masonry, carpentry, hair dressing, pottery, blacksmithing, petty trades, food-for work and cash for work community conservation

works; fLivestock product income includes value of honey, milk, butter, skins and hides, eggs, and income from renting out donkey,

mule, camel, and oxen; g

Dependency ratio is calculated as the ratio of number of dependent household members to number of

households in the working age group (age between 15 and 64 years); h Household head is a settler is a dummy variable equals to 1

if the household head came to the village as a settler following the government resettlement program held in the region, and 0 if the

household was born in the village and/or has been living there prior to the resettlement; iInitial assets are for the year 2002;

jexpressed per adult equivalent terms, using the modified OECD adult equivalence scales: a weight of 1 is assigned for the

household head, 0.5 for all other adults and 0.3 for children under the age of 17; k Conversion factors for tropical livestock unit: 1

for camel, 0.8 for horse, 0.7 for cow/oxen, 0.5 for donkey, 0.1 for sheep/goat, and 0.01 for chicken; lNon-land assets include all

agricultural household durables (farm implements) like set of traditional plowing implements, hammer, sickle, spade, saw, shovel,

motor pump, treadle pump, oxen cart; mConsumer durables include radio, tape recorder, kerosene stove, buckets, household

utensils, bed, chairs, tables, jewelries, wrist watch.nDummy variable and has value 1 if the household has irrigated land and 0

otherwise.

114

Household Income: Total household income and income from different sources (crop

production, off-farm income, forest environmental income, livestock product income, and

dividends from frankincense cooperative membership) is calculated for the 12 months prior to

the survey (March 2010). The main source of income to rural households in the study area is

rain-fed agricultural crop production, mainly sorghum, sesame, maize, and teff with very few

horticultural products like banana, onion and other vegetables in the case of households with

access to irrigation water along the Tekeze river. Income from crop production accounts for

54.60% of the mean household income of the total sample households. Off-farm income from

labour and self-employment in off-farm activities is the second largest source accounting for

16.82% of the average income of the total sample households. Most conventional rural

household surveys do not account the value of forest environmental income, which includes fuel

wood, edible wild fruits, traditional medicinal plants, and other non-timber forest products. Our

survey indicates that these non-timber forest products valued at the local market price account

for 13.63% of the average household income of the total sample households. Together with

income as dividend from frankincense cooperative membership, non-timber forest product

income accounts for 16.19% of the household income for the pooled sample. Mixed crop-

livestock production is the dominant economic activity in most rural areas of Ethiopia. The value

of livestock products accounts for 12.39% of the average household income. Income as

dividend from frankincense cooperative membership accounts for 2.56% of the average income

of total sample households and for 8.16% of the average income for members of cooperative

firms.

Poverty rates: According to the 1995/96 Ethiopia’s Household Income, Consumption and

Expenditure Survey (HICES), the national poverty line was ETB 1075.03 per adult per year and

annual per adult- equivalent income below ETB 806.27 was the measure for extreme poverty

(Woldehanna, 2004). We calculated ETB 2772.84 per adult for poverty and ETB 2079.62 for

extreme poverty for the study area by updating the national poverty lines to the period of our

survey (2008/09) using consumer price indices (CSA, 2011). The incidence of poverty in the

study area is 33.46% and extreme poverty at 19.57%. Poverty is much lower among

cooperative member households and is only 15.83% whereas it is 38.88% among the control

group households. While the absolute poverty was 23.27% among the control group, it was only

7.50% among the treated group households. As often done in conventional income and poverty

studies, if only conventional income sources are considered and income from non-timber forest

products are not accounted for, the poverty rates would be much higher. Without the forest

environmental income, the poverty rate would be 43.83% for the study area, 26.67% for

frankincense cooperative members, and 49.11% for the control group. Therefore, forest

environmental income reduces the poverty line for the study area by 10.37 percentage points.

Together with income derived from frankincense cooperative membership, forest income in total

115

reduces the poverty rate by 11.55 percentage points. These figures suggest that membership in

frankincense cooperative firms is associated with higher rural incomes and lower poverty rates.

However, it is important to undertake a rigorous econometric analysis to verify such a relation.

6.2.3. Econometric models

The above descriptive analysis indicates that frankincense cooperative member households

have a statistically significant higher income than the non-member households. However, in

order to verify that this difference is due to membership in the cooperatives, further econometric

analysis is required for solving the basic problem of impact evaluation, which is the lack of a

simultaneous observation of outcomes of an individual sample from receiving or not receiving

the treatment. Therefore, comparing the outcomes of the treatment and control groups could be

subject to selection bias due to observables (Rosenbaum and Rubin, 1983) and unobservables

(Heckman, 1997).

We applied econometric models to estimate the effect of membership in frankincense

cooperative firms on rural household income. The econometric analyses are based on two

different types of models in which the first type has a binary treatment variable (Model A) and

the second type has a continuous treatment variable (Model B). Therefore, membership in the

cooperative firms as a treatment variable, Ti, is expressed as either a binary variable measuring

whether or not a household has a share in frankincense cooperative firm in the village, or as a

continuous variable measuring the value of investment in shares in the cooperative firms that

the household has invested to be a member. The outcome variable income, Yi, is also

determined by other relevant covariates represented by the vector Xi, including household

productive asset holdings and demographic characteristics that may affect productivity and

profits:

)1(uXTY iiii

Estimating the effect of Ti on Yi requires Ti to be uncorrelated with the error term. However, Ti

can be arbitrarily correlated with the error term or unobserved heterogeneity. It is likely that the

participation of rural households in frankincense cooperative firms is non random. To address

the potential bias, we used parametric and non-parametric econometric models. First, the

problem can be treated as an endogeneity problem where the partial effect of the treatment

variable Ti depends only on observed exogenous variables. Accordingly, we first applied simple

OLS estimation to see the effect of membership in frankincense cooperative, measured either

as a binary variable or as a continuous variable, on household income. Next, instrumental

variable estimation (IV) approaches are applied to solve the problem of endogeneity of both the

116

binary and continuous membership variables. Third, we can consider the unobserved

heterogeneity as a sample selection problem and estimate the average treatment effect of the

binary membership variable using the propensity score matching method, which is a non-

parametric approach. We used STATA 11 for the statistical analysis.

OLS estimation: Equation 1 is estimated using ordinary least squares (OLS) using a large set

of covariates Xi, with the anticipation that these control variables can correct for selection bias

(Maertens et al., 2011). The model is estimated with binary variable Ti (Model 1A in Table 6.3)

to the selection-on-observables method of Rubin (1974) and Heckman and Robb (1985), and

with the continuous variable Ti (Model 2A), which corresponds to a general multiple regression

model. The covariates included in the vector Xi are: household labour, which is the number of

household members in the productive age group, and its squared term, the number of

dependent household members, a dummy variable for gender of the household head (1 for

female and 0 for male), a dummy variable for household head with at least elementary

education (1 for a household head with at least elementary education, and 0 otherwise), a

dummy variable for household heads who came to the village as settlers from other places (1

for a settler and 0 for original resident), age of the household head, walking distance to all

weather road (in hours), a dummy for initial access to irrigation water (1 for household with

irrigation access, and 0 otherwise), initial per adult equivalent land holding size (in ha), value of

initial per adult equivalent non-land assets (in 1000s of ETB), and value of initial per adult

equivalent household consumer durables (in 1000s of ETB). The initial land and the asset

holdings refer to the situation in 2002 before the introduction of rural frankincense cooperative

firms in the study area. It is based on recall data and included in the model to control for

potential endogeneity problems related to these asset holdings of the sample households.

Instrumental variable estimation: In order to deal with possible bias due to unobserved

heterogeneity, the effect of Ti on Yi can be estimated applying the IV method using the following

model in which the treatment variable is estimated in the first-stage equation (3) and the

estimated Ti used as an instrumented covariate in the second-stage structural equation (2):

)2(uXTY iiii

With

)3(ZXT iii

where the variable Zi is an instrument for Ti in the first-stage equation and Xi is the same vector

of covariates as in the OLS model in equation 1. The initial per adult equivalent livestock

holdings in tropical livestock units, the area of Boswellia forest in the study villages, and the

number of frankincense cooperative firms in the district per 1000 households are used. We

select these variables based on the intuition that, ceteris paribus, the probability of membership

117

and the amount to invest in a rural frankincense cooperative firm depend on the lagged

household wealth, the area of Boswellia paprifera forest available, and the number of

cooperative firms at the district level. We anticipated that these variables have no correlation

with the error term in structural equation 2. Models 1B for the dummy treatment variable and 2B

for the continuous treatment variable are based on the two-stage least square (2SLS)

instrumental variable estimation and we applied the ivreg2 command (Baum et al., 2010) in the

STATA software. The application allows for testing endogeneity of the treatment variable and

testing the validity of the instrumental variables used to address the endogeneity problem.

In the 2SLS IV estimation of Ti as a continuous variable, the variable is censored at zero, as the

majority of the sample households (76.52%) are among the control group. Therefore, following

Maertens et al. (2011) the first stage equation is estimated using a Tobit model in which T i* is a

latent variable and used as an instrument in the structural equation 2 to (Model 2C):

)4()]T,0max[(Tw ithZXT *iiii

*i

Propensity score matching: In order to solve the problem of selection bias the propensity

score matching technique described first by Rosenbaum and Rubin (1983) is used. This method

involves the estimation of the average treatment effect as a weighted average of the outcome

difference between treated and control groups that are matched based on their similarity in

observable characteristics (Wooldrige 2002; Abadie et al. 2004; Becker and Ichino 2005). The

average treatment effect (ATE) (Model 1C) is estimated as the expected value of the difference

between the income with treatment Y1 and the income without treatment Y0 using the following

formula:

)5()YY(EATE 01

The procedure involves first specifying and estimating the propensity score using a probit

model. We estimated the propensity scores as a conditional probability of membership in

frankincense cooperative firms P (Ti=1| Si) using a probit model. Propensity score matching

techniques are sensitive to the choice of a proper set of covariates (Imbens, 2004; Becker and

Ichino, 2005) and requires some testing on whether to use unrestricted or restricted probit

model for estimating the propensity scores (Maertens et al., 2011). Therefore, we test the

unrestricted probit model with the same set of covariates Xi and Zi as in the first stage IV

equation, against a restricted probit model that includes only those variables with a significant

effect (at < 0.1 level). A log-likelihood ratio test indicates that the restricted model is better and

the propensity score is estimated using variables with significant effect (at the 0.05 level) in the

restricted model. The kernel matching method is applied and only observation in the common

support region where the propensity scores of the treated unit is not higher than the maximum

118

or less than the minimum propensity score of the control units (Becker and Ichino, 2005), are

used for calculating the ATE.

The balancing and conditional independence assumptions have to be satisfied for the

propensity score matching to provide a consistent estimate of the ATE. The first assumption

refers to the balancing in the covariate distribution between treated and control group

observations (Dehejia and Wahba, 2002; Imbens, 2004). The balancing properties are taken

into account by testing for equality of means between treated and matched controls (Table 6.4).

The result of the t-test indicates that 6 out of the 7 covariates in the unmatched cases were

statistically significant indicating the presence of a strong bias amounting an average of 22.06%

(with standard deviation of 14.48%), and matching strongly reduces this bias to 4.99% (with

standard error of 4.03%). Moreover, the statistically insignificant difference between matched

treated and controls in terms of all the covariates refers to the fulfilment of the balancing

properties (Table 6.4).

The second assumption of conditional independence requires that if all variables that affect the

receipt of treatment and the potential outcomes are observed, then conditional on these

variables the potential outcomes with and without treatment are randomly independent of the

receipt of treatment (Imbens, 2004). However, this assumption is not inherently testable

because the data are uninformative about the distribution of neither the untreated outcome for

the treated nor the treated outcome for the control group (Imbens, 2004). However, the

robustness of matching estimators to failure of the conditional independence assumptions can

be tested using a simulation method proposed by Ichino et al. (2006). The method is based on

Rosenbaum and Rubin (1983) and it assumes that the conditional independence assumption is

not satisfied, but would be satisfied if an additional binary variable could be observed.

Accordingly, the method simulates this binary confounder in the data that is used as an

additional matching factor. Thus, following Ichino et al. (2006), we use a neutral confounder and

a confounder calibrated to mimic observable binary covariates in the model. The results in Table

6.5 indicate that the estimate based on the neutral binary confounder differs by less than 5%

from the baseline matching estimator whereas the estimate with the binary confounder

calibrated to mimic the covariate that ‘a household is a settler’ differs by 10.12% from the

baseline propensity score matching estimator. These results confirm that the results of ATE are

robust.

6.3. Results

6.3.1. Determinants of membership in frankincense cooperative firms

Table 6.2 shows the determinants of membership in frankincense cooperative firms. The

unrestricted and restricted probit models estimate the probability of membership in the

119

cooperatives and the tobit model estimates the log-transformed value of shares that households

made as investment in the frankincense cooperative firms. The relatively high statistics for the

McFadden Pseudo R2 and percents of correct predictions in both the unrestricted and restricted

models implies that membership in frankincense cooperative firms in the study area is well

explained by household and village level characteristics.

Table 6.2: Determinants of membership in the frankincense rural cooperative firms

Variables Full model Restricted model Tobit

Labour 0.032(0.211) 0.333(1.357)

Labour2 0.005(0.028) 0.020(0.181)

Female headed household -0.030(0.236) -0.442(1.508)

Education 0.167(0.153) 0.868(0.962)

Age of the household head 0.008(0.007) 0.044(0.045)

Irrigation -0.261(0.236) -1.251(1.452)

Dependent 0.088(0.045)c 0.098(0.043)

b 0.497(0.282)

c

Household head is settler -0.789(0.167)a -0.768(0.162)

a -4.698(1.074)

a

Distance to all weather road -0.003(0.017) -0.025(0.104)

Initial Land 0.123(0.072)c 0.136(0.069)

b 0.882(0.453)

c

Initial Consumer durable 2.73E-04 (1.17E-04)

b

2.48E-04 (1.16E-04)

b

0.002(0.001)b

Initial non land assets 1.83E-04 (9.57E-05)

c

1.69E-04 (9.31E-05)

c

4.42E-04 (2.05E-04)

b

Initial TLU 0.023(0.016)d 0.024(0.015)

d 0.139(0.092)

d

Boswellia forest area in the village 0.001(1.05E-04))a 0.001(1.02E-04)

a 0.003(0.001)

a

Number of cooperatives per1000HH in

the district

-6.507(1.204)a -6.313(1.167)

a -32.694(7.674)

a

Constant -2.517(0.467)a -2.001(0.207)

a -15.890(3.173)

a

LogL -223.80 -226.86 -541.13

McFadden Pseudo R2 0.197 0.186 0.090

LR 2 109.45

a 103.31

a 106.44

a

% of Correct predictions 80.23 80.23

% of Sensitivity 35.00 35.83

% of Specificity 94.12 93.86

Likelihood ratio test of

restricted versus full model

LR 2 6.14

Prob > LR 2 0.524

Statistical significance: a = p < 1%; b= p<5%; c=p<10%; d=p<15%.

The results of the both probit and tobit models show that the covariates that whether a

household head is settler to the village, the area of Boswellia forest in the village in which a

household lives, and the number of frankincense cooperative firms in the district per 1000

households are the most significant (at p < 1%) determinants of membership in the frankincense

cooperative firms. The negative coefficient for the dummy variable indicating whether the

household head is a settler indicates that membership was biased against the settler

120

households who came to the villages as part of the resettlement programme of the government.

The positive coefficient on the covariate on forest area and the negative coefficient on the other

covariate are consistent with our expectation. The larger the area of Boswellia forests in a

village the higher the probability of a household to be member of a frankincense cooperative

firm. Contrary, the larger the number of frankincense cooperative firms at district level, the

higher the competition for the forestland and hence the lower the probability of forming a

cooperative and become a member. Households with more dependent family members and

households that are relatively wealthier in terms of per adult equivalent initial land holding sizes,

consumer durables, other asset holdings, and total livestock units had also a significantly higher

probability in becoming a member as well as investing in frankincense cooperative firms. The

results of the probit and tobit models on the determinants of membership in cooperative firms

are consistent with the descriptive analysis in Table 6.1. The results in both models show that

participation in frankincense cooperative firms is not biased towards most of the demographic

characteristics of households. Rather, it is biased towards the relatively wealthier households

and households who are original residents.

6.3.2. Income effects of membership in frankincense cooperative firms

The results of the econometric models applied for the estimation of the effect of membership

and log transformed value of investment in frankincense cooperative on log transformed

household income are presented in Table 6.3. The results of the OLS estimation (Model 1A),

the IV-2SLS (Model 1B), and the propensity score matching (Model 1C) with the binary

membership variable all indicate that frankincense cooperative member households have

statistically significant higher income than non-member households. Model 2B indicates that

membership as a binary variable is endogenous (Wu-Hausman F-statistics is significant at 1%)

and hence the OLS estimates (Model 1A) are biased.

Our instrumental variables satisfy all three identification restrictions. The Saran statistics for

over-identification restriction is insignificant (p=0.546 for Model 1B and p= 0.139 for Model 2B)

indicating that the instrumental variables used in the model are valid instruments and are

uncorrelated with the error term of the structural equation 2 and that they are correctly excluded

from the estimated equations. The Anderson Canon. Corr. Lagrange Multiplier statistic for the

under-identification test is also significant (at p < 1%) indicating that the model is correctly

identified. In addition the Cragg-Donald Wald F-statistic for both Model1B and Model 2B is

larger than the critical value of the Stock-Yogo (2005) weak identification test, which equals

22.30 for 10% maximal IV size for 3 instruments and 1 endogenous variable case, indicating the

rejection of the null hypothesis that the instruments are weak (at p < 0.05 level).

121

Table 6.3: Effects of membership and log transformed investment in frankincense cooperative firm on log

transformed rural income

Variables Model 1A Model1B Model 1C Model 2A Model 2B Model 2C

OLS IV-Estimation PSM OLS IV-Estimation IV-Estimationd

Membership (dummy)

0.375 (0.076)

a

1.044 (0.210)

a

0.316 (0.098)

a

Ln(Investment) 0.060 (0.012)

a

0.157 (0.033)

a

0.043 (0.008)

a

Labour 0.315 (0.097)

a

0.327 (0.103)

a

0.304 (0.097)

a

0.298 (0.101)

a

0.308 (0.097)

a

Labour2 -0.027 (0.013)

b

-0.031 (0.014)

b

-0.026 (0.013)

c

-0.027 (0.014)

c

-0.029 (0.013)

b

Female headed household

-0.319 (0.106)

a

-0.302 (0.112)

a

-0.312 (0.106)

a

-0.284 (0.111)

b

-0.299 (0.106)

a

Education 0.166 (0.069)

b

0.162 (0.074)

b

0.169 (0.070)

b

0.171 (0.073)

b

0.160 (0.069)

b

Age of the household head

-0.013 (0.003)

a

-0.012 (0.003)

a

-0.012 (0.003)

a

-0.012 (0.003)

a

-0.012 (0.003)

a

Irrigation 0.538 (0.110)

a

0.536 (0.116)

a

0.542 (0.110)

a

0.545 (0.115)

a

0.541 (0.110)

a

Dependent 0.029 (0.020)

0.022 (0.022)

0.030 (0.020)

0.024 (0.021)

0.023 (0.020)

Household head is settler

0.515 (0.069)

a

0.560 (0.075)

a

0.505 (0.069)

a

0.530 (0.073)

a

0.506 (0.069)

a

Distance to all weather road

-0.016 (0.008)

b

-0.013 (0.008)

-0.017 (0.008)

b

-0.016 (0.008)

c

-0.013 (0.008)

c

Initial Land 0.160 (0.035)

a

0.134 (0.038)

a

0.157 (0.035)

a

0.129 (0.037)

a

0.130 (0.035)

a

Initial Consumer durable

2.56E-04 (6.06E-05)

a

1.72E-04 (6.87E-05)

b

2,55E-04 (6,06E-05)

a

1.78E-04 (6.79E-05)

a

1.90E-04 (6.36E-05)

a

Initial non land assets

-1.28E-05 (1.97E-05)

-3.18E-05 (2.16E-05)

-1,21E-05 (1,97E-05)

-2.82E-05 (2.12E-05)

-1.94E-05 (1.98E-05)

Constant 8.565 (0.190)

a

8.446 (0.204)

a

8.575 (0.190)

a

8.482 (0.201)

a

8.952 (0.199)

a

R2 0.348 0.247 0.330 0.266 0.334

F-test 20.42a 17.98

a 20.34

a 18.16

a 20.67

a

Root MSE 0.701 0.744 0.702 0.734 0.700

Test for endogeneity of treatment variable

Wu-Hausman F-test statistic 13.685a 11.384

a

Tests for validity of the instruments

Weak identification test Cragg-Donnaled Wald F-statistic

28.435 29.679

Under identification test Anderson Canon. Corr. LM statistic

75.116a 77.902

a

Over identification test Sargan Statistic

1.215 3.943

Statistical significance: a = p < 1%; b= p<5%; c=p<10%; dThe IV-Estimation is based on a Tobit model in the first stage.

122

Table 6.4: Balancing properties of covariates in frankincense cooperative members and non-member

groups for kernel matching on propensity scores

Variable

Sample

Mean %Bias

%Reduction in bias

t-test

Treated Control t-Value P-value

Dependent Unmatched 2.69 2.48 13.40 1.28 0.20

Matched 2.61 2.60 0.40 97.30 0.03 0.98

Household head is settler

Unmatched 0.24 0.34 -21.80 -2.03 0.04

Matched 0.26 0.27 -2.80 87.30 -0.21 0.83

Distance to all weather road

Unmatched 1.06 0.75 31.20 3.13 ***

Matched 1.00 0.89 10.50 66.20 0.79 0.43

Initial Land

Unmatched 0.46 0.24 39.40 4.09 ***

Matched 0.39 0.42 -4.40 88.70 -0.29 0.78

Initial consumer durables (in 1000ETB)

Unmatched 10.73 7.47 50.60 4.63 ***

Matched 10.69 10.94 -3.80 92.40 -0.30 0.77

Boswellia forest area

1000ha in the village

Number of

Unmatched 0.86 0.61 45.20 4.19 ***

Matched 0.86 0.88 -4.10 91.00 -0.31 0.76

cooperatives per 1000 households in the district

Unmatched 0.86 0.61 45.20 4.19 ***

Matched 0.86 0.87 -1.70 96.20 -0.13 0.90

Note: *** p < 0.001

Since the dependent variable is in natural logarithmic form, the coefficients indicate the elasticity

of income to changes in the covariates. From the results of Model 1C and Model 1A, we can

derive that frankincense cooperative member households on average have incomes that are

31.60% to 37.5% higher than non-member households. In absolute terms, this means member

households have incomes in between ETB 4076 to ETB 4838 higher than the average income

of households in the control group.

Table 6.5: Simulation-based sensitivity analysis for propensity score matching

Estimated treatment

effect

Outcome

effecta

Selection

effectb

Kernel matching

Baseline propensity score matching estimator 0.316

Matching estimator with simulated binary confounder

Neutral confounder 0.303 1.537 3.212

Confounder calibrated to mimic a household head is settler 0.348 5.936 0.617

aThe outcome effect measures the estimated effect of the simulated binary confounder on the outcome variable log

transformed household income; bThe selection effect measures the estimated effect of the simulated binary

confounder on the selection into treatment.

6.4. Discussion and conclusions

Ethiopia’s export of frankincense, natural gums and resins to the world market has been

increasing since 1993 following the policy change that allows entry of the private sector in the

harvesting, processing and trading of these non-timber forest products. Although much of the

123

harvesting and all the processing and trading activities are still operated by the state owned

company and the private business firms involved in the sector, in Tigray rural cooperative firms,

which are owned by rural households, have entered in the harvesting of frankincense since

2002. The ability of non-timber forest products for achieving the Millennium Development Goal

of poverty alleviation has been the subject of considerable debate (Shckleton and Gumbo,

2010). This chapter has analysed the impact of membership in rural frankincense cooperative

firms on income and poverty reduction using household survey data. The study shows that

frankincense cooperative firms have contributed an important benefit to rural households. The

results of the different econometric models indicate robust, significant and large positive effects

on income and poverty reduction. Income from frankincense cooperative membership resulting

in a 5 percentage point reductions of the poverty rate among the cooperative members and a

1.19 percentage point reduction in the poverty rate for the total sample of the study area. The

16% poverty rate among the frankincense member households is much lower than the rural

poverty rate for the country in the year 2005, which was 38.9% (World Bank, 2011) as well as

the 29.6% rate, which was announced recently as an interim report on the Household Income

and Consumption Expenditure Survey held in 2010/11 in Ethiopia.

This study also found that entry to a rural cooperative firm is biased towards the better off

households in terms of asset holdings like land, non-land assets and household consumer

durables, and livestock holdings. Moreover, the cooperative member households have higher

total income as well as income from other non-timber forest products. The sum of income from

frankincense membership and other non-timber forest products accounts for 16% of the total

income of cooperative members but only 14.57% of the total income of non-members. This

indicates that in relative terms the better-off households depend more on non-timber forest

products than the relatively poor households. Ambrose (2003) has also reported that non-timber

forest products contribute significantly to middle income households than low income

households in the humid regions of Cameroon. However, most previous works in Africa suggest

that the poorest households are the most dependent on non-timber forest products for their

livelihood and income (Cavendish, 2000; Neumann and Hirsch, 2000; Fisher, 2004; Shackleton

and Schackleton, 2006; Babulo et al., 2009; Shckleton and Gumbo, 2010). This case study

assessed the effect of frankincense cooperative membership on poverty based on the income

effect. Therefore, as a case study one has to note the limitations that it can only tell us the

effects at a point in time. An important question remains regarding the sustainability of the rural

frankincense cooperative firms and their effects on rural income and poverty reduction in the

future, which mainly will depend on their capacity to manage the forest resource on sustainable

basis. Frankincense forests throughout the country are poorly managed and there is lack of

both institutional and technical capacity at all levels, which is an important challenge that has to

be addressed in order to sustain the role of the resource base on rural poverty reduction.

124

CHAPTER 7

Conclusions and recommendations

7.1. General conclusions

This dissertation is a multidisciplinary piece of work with five main research questions. In order

to answer these research questions and achieve the aims and objectives of the research

(Chapter 1 section 1.3), primary data collected through plot level biophysical experiments and a

household level socioeconomic survey were collected from the study sites as the main data

sources and various analytical tools were applied. A multilevel linear mixed model was used to

investigate the effects of leaf lopping, tree level characteristics, and environmental factors on

frankincense yield, flower and fruit production and hence identify empirical management options

for the species. The standard plant allometric modelling that shows the relationship between

size and weight of dry biomass was applied to develop an allometric model specific to Boswellia

papyrifera trees and hence determine the dry biomass and carbon stock in Boswellia forests. In

addition, soil chemical analysis was undertaken using standard lab procedures and the data

were used to determine the stocks of soil organic carbon and nutrients in Boswellia forests of

the study area. Based on the results of chapters 2 and 3, the economic implications of

conservation and other five different Boswellia forest management options were evaluated

using the standard model of cost benefit analysis (CBA) in chapter 4. A double-bound

dichotomous choice model of contingent valuation was applied to assess the demand of rural

households for conservation of Boswellia papyrifera forest. Finally, in the last chapter the

welfare impact of rural households’ access to the resource through membership in rural

frankincense cooperative firms was analysed using the method of ordinary least squares,

instrumental variable estimation and propensity score matching techniques.

The focuses of this study were the evaluation of the economic, ecological, environmental

impacts of experiment based Boswellia papyrifera forest management options, and the trade-

offs between these management options and the conversion of the forest to crop cultivation. It

also deals with the assessment of rural households’ willingness to contribute for the resources

conservation and the role of the resource in rural poverty reduction. Specifically, the study

answered the five research questions and derived key findings that are important inputs in

designing policies towards the restoration and sustainable management of frankincense forests

in Ethiopia. The summaries of the findings related to each research questions are given in the

next paragraphs.

How does leaf lopping for fodder affect frankincense yield, flower and fruit production in

Boswellia paprifera trees?

126

The empirical results in Chapter 2 indicate that the current practice of leaf lopping for livestock

feed has a detrimental effect on frankincense, flowers, and fruits production capacity of

Boswellia trees. Leaf lopping has resulted in a statistically significant negative effect on

frankincense, flower and fruit production capacity of the tree. Moreover, the negative effects of

leaf lopping on the three response variables are consistent across all diameter classes of the

tree. Unlike leaf lopping, tapping has a statistically significant positive effect on frankincense

yield. However, the main effects of tapping as well as the interaction effects of tapping and leaf

lopping on the capacity of Boswellia trees to produce flowers and fruits were negative but

statistically insignificant. In addition to the leaf lopping and tapping factors, environmental

variables (soil depth, total nitrogen and carbon concentrations in the soil) as well dendrometric

variables (tree diameter, height, stem height, crown diameter) and the bark colour of the tree

have statistically significant effects on inflorescence and fruits production capacity of the tree.

The environmental variables (altitude, percent of stoniness, pH-water and total nitrogen

concentration in the soil) as well as dendrometric variables and the bark colour of the tree have

statistically significant effect on frankincense yield of Boswellia papyrifera trees.

How much biomass and soil organic carbon is stored in Boswellia papyrifera forests?

The analysis of biomass and soil organic carbon stock in Boswellia papyrifera forests in Chapter

3 shows that among the different allometric relationships identified for the above ground

biomass as well as the biomasses for the different tree components (leaf, branches, and stem),

the standard power function Yi = a(DBH)b with diameter at breast height (DBH) as a single

explanatory variable provided the best fit. In addition, both the validation test using the 10-fold

cross validation and comparison of the predictions of above ground biomass of the tree using

different mixed species allometric models from the literature with the prediction using the model

from this study confirmed that the model (AGB = 0.061(DBH) 2.353) is superior for predicting the

above ground biomass of Boswellia papyrifera forests. The study also indicated that much of

the carbon stock in Boswellia papyrifera forest was found in soils than the tree biomass. Soils

organic carbon accounted about two-third of the total ecosystem carbon pool of Boswellia

papyrifera forest. Moreover, the carbon and soil nutrient content in fenced plots was higher and

statistically significant than the carbon and soil nutrient content in unfenced plots that were free

for livestock grazing.

How much are the opportunity costs (in terms of net benefits from shifting cultivation) of keeping

Boswellia forest under management options that range from pure conservation to the business

as usual practice that involves intensive frankincense production, leaf lopping, and free grazing?

127

The environmental cost benefit analysis of the six experimental Boswellia papyrifera forest

management options (namely, conservation, exclosure with normal tapping of frankincense,

exclosure with heavy tapping of frankincense, free grazing with no leaf lopping and no tapping,

free grazing with leaf lopping and normal tapping, and free grazing with leaf lopping and heavy

tapping) resulted in negative NPVs except for the last management option. In other words,

keeping the forest in any form of these first five management options is economically less

profitable than shifting the forest to crop cultivation. Moreover, all management options have

negative NPVs if the values of the indirect ecosystem services, which are carbon emission

reduction and sequestration benefits as well as the values of avoided loss of nitrogen and

phosphorous nutrients in soils and benefits of sequestered nutrients, are deducted from the total

benefits. The rural people incur much of the opportunity cost of keeping the forest in any of the

management options assessed. Except for the conservation and free grazing with no leaf

lopping and no tapping management options, in all the other four management options 100% of

the opportunity cost of not converting the forest to crop land is incurred by rural households.

This clearly explains the lack of incentives among rural households not to stop practicing shifting

cultivation in Boswellia forest areas. Shifting cultivation practices are highly visible in Boswellia

papyrifera forest areas of particularly western Tigray and the Meterma district of the south-

western part of the Amhara regional state.

Conservation of Boswellia papyrifera forest as well as management as exclosures could result

in emission reductions of about 142 tCO2 per hectare with equivalent opportunity costs of about

33 USD and 22 USD per tCO2 emission reduction for the conservation and exclosure

management options respectively. Moreover, compared to the three management options that

allow free grazing, the conservation and exclosure options of managing the forest could lead to

better achievements of the UN-REDD+ objective, namely reducing emissions from deforestation

and enhancing carbon stock in forests, conserving biodiversity, reducing poverty, and realizing

the sustainable management of forests. The selection of the best management option among

the conservation and the two exclosure management options, however, depends on the

distribution of the weights that decision makers will attribute to each objective used as criterion

in the multi-criteria analysis. If relatively, more weights are to be given to emission reduction,

biodiversity conservation, and poverty reduction criteria and less weight to the cost of emission

reduction and soil nutrient enhancement, the conservation option could lead to the best

achievement of the REDD+ objective.

Are rural households residing in Boswellia forest areas willing to contribute for a conservation

intervention?

128

The contingent valuation study indicates that despite the perceived poverty in the country, rural

people in the study area attribute option value to the resource and are very concerned about the

risk of extinction of the species. They are willing to contribute either in terms of cash or in terms

of free labour for the conservation of Boswellia papyrifera forests. Besides the bid level, while

household income is the most important factor affecting rural households’ willingness to pay in

cash for Boswellia papyrifera forest conservation, it is the number of household members in the

working age group that affects most their willingness to contribute labour for conservation. Rural

households are willing to pay at least about 5 USD per year on average in cash or contribute

about 7 days of free labour as guards for the conservation of Boswellia papyrifera forests. The

difference between the willingness to pay in cash and the monetary value of the lower bound

willingness to contribute labour valued at the average per capita daily income of the rural

households was not statistically significant indicating the presence of convergent validity of the

willingness to pay in cash and willingness to contribute labour values.

What is the contribution of income from frankincense on poverty reduction in rural areas with

frankincense cooperative firms?

The results of the analysis on the impact of frankincense cooperative membership on income

and poverty in the rural areas of the study site indicates that organized access to the resource

has a significant positive effect on rural income and poverty reduction. However, access to the

resource in terms of membership to the cooperative firms was biased towards the relatively

better off households in terms of livestock and asset holdings. Moreover, the share of the sum

of income from frankincense cooperative membership and other non-timber forest products to

total household income was higher for the cooperative members than the non members

indicating that the better off households depend more on the forest resource than the relatively

poor households.

In general, the key findings of this study were: a) leaf lopping from Boswellia papyrifera tree has

a significant negative effect on frankincense, inflorescence, and fruit production capacity of the

tree, b) the allometric model AGB = 0.061(DBH) 2.353 is the best equation for predicting the

above ground biomass carbon in Boswellia papyrifera forest, c) soil organic carbon accounts

about two-third of the carbon pool in Boswellia papyrifera forests; and fencing, which is a proxy

for excluding livestock from free grazing, enhances the carbon and soil nutrient stock in soils, d)

pure conservation as well as management of Boswellia papyrifera forests as exclosures

generate negative Net Present Values, e) pure conservation and/or management as exclosure

could result in emission reduction of about 142 tCO2 per hectare of Boswellia papyrifera forest

at an opportunity cost of about 33 USD per tCO2 emission reduction in the case of conservation;

and of this cost about 80% is incurred by rural people in the form of forgone net benefits from

129

not converting the forest to crop land through shifting cultivation, f) rural households are willing

to contribute either in cash or free labour for the conservation of Boswellia papyrifera forests;

and their willingness to pay in cash depends most on household income whereas willingness to

contribute labour is affected most by the size of family labour in a household, g) membership as

well as investment in shares in rural frankincense cooperatives have a significant impact on

increasing rural household income and reducing rural poverty.

7.2. Recommendations

The results of this study have important implications for designing policies towards the

conservation, restoration, and sustainable management of Boswellia papyrifera forests.

Therefore, based on the results of the study, the following recommendations or policy

implications are drawn:

1. The experimental study on the effect of leaf lopping and tapping on the productive

capacity of Boswellia papyrifera trees shows that leaf lopping has negative impacts on

frankincense yield as well as flower and fruit productions. Thus, from the point of view of

sustainability of both frankincense production and the tree species, development

interventions that could solve the problem of livestock feed, such as introducing

cultivable improved livestock feed, has to be carried out in the Boswellia papyrifera

forest areas so that it is possible to avoid the practice of leaf lopping from the tree.

Moreover, it is also important to create awareness among the rural communities on the

negative impact of the practise of leaf lopping on the forest resource.

2. The result of the biomass and carbon stock analysis in chapter 3 shows that much of

the carbon stock in Boswellia papyrifera forests is accounted for by soil organic carbon

and fenced stands have much more stored carbon in the soil than the unfenced stands.

Taking fencing as a proxy for strict exclosure management, this study implies that well

organized and efficient exclosure management can enhance the soil organic carbon as

well as soil nutrient content in Boswellia papyrifera forests. Moreover, exclosures can

also assist natural regeneration of Boswellia papyrifera through protecting seedlings

from damages that can be caused because of free grazing. Thus, policies that aim to

reduce deforestation and related emissions need to focus on management interventions

that can enhance the soil carbon and thereby protect the standing trees from

deforestation. Although experiences of management of exclosures as a means for

restoration/rehabilitation of degraded forests is not new to Tigray, it has been limited

only to the highlands and it needs to be applied in the western lowlands where almost

all of the remaining pockets of the region’s forest including Boswellia papyrifera forests

are found.

130

3. The cost-benefit analysis of the six Boswellia forest management options indicated that

except for the management that involves free grazing with leaf lopping and heavy

frankincense tapping, which resembles the business as usual scenario, all the other

options have negative Net Present Values. If the benefits from the indirect ecosystem

services are deducted, the net benefits from keeping the forest in any of the

management options are less than the net benefits from shifting to crop cultivation.

Besides, rural people incur almost all of this opportunity cost. Therefore, the current

trend of deforestation through shifting cultivation will continue unless certain

interventions like conservation of the resource in the framework of UN-REDD+

programme is in place to save the species from the existing high risk of extinction.

Establishing bee keeping sites in Boswellia papyrifera forests for rural households and

providing extension services in bee keeping could increase the tangible benefits from the

forest and increase its competitiveness against the competing land use – shifting

cultivation. Other possible means for increasing the competitiveness of the forest is to

exploit the ecotourism potential and generate income for the rural communities.

Moreover, for making the forest more competitive than the practice of shifting to crop

cultivation, promoting investments on industries that can add value on the raw

frankincense, and introducing forest-certification of oleo-gum resin products for

promoting sustainable use and management of the resource has to be considered.

4. The results of the contingent valuation survey suggest that rural households have an

interest in conserving Boswellia forests through contributions in terms of either cash or

free labour terms, which could help in reducing the costs of implementing conservation

measures. Therefore, the consent of these stakeholders has to be taken into account

for the effectiveness of any future conservation and management interventions with

regard to this particular resource.

5. Finally, the analysis of rural households’ access to frankincense forests through

membership in rural frankincense cooperative firms shows significant positive welfare

impacts in terms of increasing rural household income and reducing poverty. However,

it has also been found that entry to frankincense cooperative firms was biased against

the relatively asset-poor household groups. Therefore, the issue of equity in the use of

the natural resource has to be taken into account in the effort of increasing rural

households’ access to the use of the forest. Moreover, providing technical trainings to

rural frankincense cooperative firms on the management of the forest is crucial for

sustaining the resource and its impact on rural poverty reduction.

131

7.3. Limitations and implications for further research

The biophysical section of this study (Chapters 2 and 3) is based on plot level experiments. For

up scaling the results of this study, scientific data on the total Boswellia forest area in Tigay as

well as other parts of the country are not yet available. Therefore, future research is required on

the assessment of the resource base at a wider spatial scale using geographic information

systems with ground based plot level inventory data.

In this study, we identified that Boswellia papyrifera trees with yellowish bark colour are

relatively more productive in terms of frankincense, flowers, and fruits production than trees with

orange and intermediate (mix of yellow and orange) bark colours. However, whether such

colour difference is due to genetic variation, age, or vitality has to be researched.

The study indicated that leaf lopping has significant negative impacts on frankincense yield,

flowers and fruit production in Boswellia trees. However, livestock rearing is the major livelihood

system in Boswellia forest areas of the country and such an effect will continue to happen.

Therefore, further research is required on the carrying capacity of the forest for sustainable

livestock production system. Specifically it is important to investigate the capacity of the forest

for production of livestock feed in terms of harvesting litter fall from the tree during the shading

period, and grass harvesting through establishing exclosures. Moreover, such information could

help in introducing the zero grazing system, which involves cut-and-carry grasses from the

forest for livestock feed, to rural households in Boswellia papyrifera forest areas.

Our cost-benefit analysis does not take into account some of the services that frankincense

forests provide. For example, the study only considers the value of raw frankincense. However,

frankincense is used as input in the pharmaceutical, cosmetic, food, and chemical industries. It

is also used as a fragrance for clerical services in different religions, and it is commonly used for

traditional coffee ceremonies in Ethiopia. Moreover, the flowers of frankincense are important

forage for bee keeping. The forest also serves as habitat for wildlife and could have benefits for

ecotourism. Therefore, these benefits from the resource have to be taken into account in future

valuation studies. Moreover, since the forest provides global ecosystem services both in terms

of frankincense as a traded commodity with medicinal value and as store of biodiversity, studies

on importing countries’ demand for conservation of the forest through contingent valuation

survey may provide higher values of the resource than the value presented in this study.

Although our study indicated that membership in rural frankincense, cooperatives had a positive

impact on rural household income and poverty reduction, sustainability of such effects will

depend on the sustainability of the resource base. In other words, the legal, technical and

institutional capacity of the cooperatives in managing the forest has to be studied. Furthermore,

132

as far as our knowledge is concerned there are no studies dealing with the legal and

institutional frameworks for the use and management of Boswellia papyrifera forests in

particular and oleo-gum resin bearing tree species in general. Such studies are important to

identify the constraints and in designing and developing the appropriate legal and institutional

capacities for the conservation and sustainable management of the resource.

References

Aaron, R., & Gibbs, H.K., 2008. New IPCC Tier-1 Global Biomass Carbon Map for the Year 2000. Carbon

Dioxide Information Analysis Center [http://cdiac.ornl.gov], Oak Ridge National Laboratory, Oak

Ridge, Tennessee.

Abadie, A.D., Drukker, J.L., & Imbens, G.W., 2004. Implementing matching estimators for average

treatment effects in Stata. The Stata Journal 4, 290-311.

Adam, A.A., & Tayb, A.M.E., 2008. A comparative study of natural regeneration of B. papyrifera and other

tree species in Jebel Marra Darfur; Sudan. Research Journal of Agriculture and Biological Sciences

4, 94-102.

Aertsen, W., Kint, V., Van Orshoven, J., Özkan, K., & Muys, B., 2010. Comparison and ranking of

different modelling techniques for prediction of site index in Mediterranean mountain forests.

Ecological Modeling 221, 1119-1130.

Alberini, A., 1995. Optimal designs for discrete choice contingent valuation surveys: Single-Bound,

Double-Bound, and Bivariate Models. Journal of Environmental Economics and Management 28,

287-306.

Ambrose, O.B., 2003. The contribution of NTFPs to the livelihood of the “forest poor”: Evidence from the

tropical forest zone of south-west Cameroon. International Forestry Review 4, 231-234.

Angelsen, A., & Wunder, S., 2003. Exploring the forest-poverty link: key concepts, issues and research

implications. CIFOR Occasional Paper No. 40.

Aref, A., 2011. Rural Cooperatives for Poverty Alleviation in Iran. Life Science Journal 8, 38-41.

Aref, F., & Sarjit, S.G., 2009. Rural tourism development through rural cooperatives. Nature and Sciences

7, 68-73.

Aregawi, T., Melaku, S., & Nigatu, L., 2008. Management and utilization of browse species as livestock

feed in semi-arid district of North Ethiopia. Livestock Research for Rural Development 20, Article

#86. Accessed on February 15, 2012: http://www.lrrd.org/lrrd20/6/areg20086.htm

Arnold, J.E.M., & Pérez, M.R., 2001. Can non-timber forest products match tropical forest conservation

and development objectives? Ecological Economics 39, 437-447.

Azene, B., Birnie, A., & Tengnas, B., 1993. Useful trees and shrubs for Ethiopia. Identification,

propagation and management for agricultural and pastoral communities. Regional Soil

Conservation Unit, Swedish International Development Authority (SIDA), Nairobi.

Azqueta, D., & Sotelsek, D., 1999. Comparative advantages and the exploitation of environmental

resources. CEPAL REVIEW 68, 115-136.

Babulo, B., Muys, B., Nega, F., Tollens, E., Nyssen, J., Deckers, J., & Mathijs, E., 2009. The economic

contribution of forest resource use to rural livelihoods in Tigray, Northern Ethiopia. Forest Policy

and Economics 11, 109-117.

Balana, B.B., Myus, B., Haregeweyn, N., Descheemaeker, K., Deckers, J., Poesen, J., Nyssen, J., &

Mathijs, E., 2012. Cost-benefit analysis of soil and water conservation measures: The case of

exclosures in northern Ethiopia. Forest Policy and Economics 15, 27-36.

Bamualin, A., Jones, R. J., & Murray, R .M., 1980. Nutritive value of tropical browse legumes in the dry

season. Proceedings of Australian Society of Animal Production 13, 229-232.

134

Bansal, S., & Germino, M.J., 2009. Temporal variation of nonstructural carbohydrates in montane

conifers: similarities and differences among developmental stages, species and environmental

conditions. Tree Physiology 29, 559-568.

Barbier, E.B., 1993. Sustainable Use of Wetlands, Valuing Tropical Wetland Benefits: Economic

Methodologies and Applications. The Geographical Journal 159, 22-32.

Basuki, T.M., van Laake, P.E., Skidmore, A.K., & Hussin, Y.A., 2009. Allometric equations for estimating

the aboveground biomass in tropical lowland Dipterocarp forests. Forest Ecology and Management

257, 1684–1694.

Baum, C.F., Schaffer, M.E., & Stillman, S., 2010. ivreg2: Stata module for extended instrumental

variables/2SLS, GMM and AC/HAC, LIML and k-class regression.

http://ideas.repec.org/c/boc/bocode/s425401.html.

Becker, S., & Ichino, A., 2005. Estimation of average treatment effects based on propensity scores. The

Stata Journal 2, 358-377.

Bein, E., Habte, B., Jaber, A., Birnie, A. & Tengas, B., 1996. Useful trees and shrubs in Eritrea:

identification, propagation and management for agricultural and pastoral communities. Technical

Hand Book No. 12. Nairobi Regional Soil Conservation Unit, Nairobi.

Belcher, B., & Kusters, K., 2004. Non-timber forest product commercialization: Development and

commercialization lessons. In: Kusters, K., & Belcher, B., (eds). Forest Products, Livlihoods and

Conservation. Case Studies of Non-Timber Forest Product Systems 1 Asia, CIFOR, Jakarta.

Belcher, B., & Schreckenberg, K., 2007. Commercialization of non-timber forest products: a reality check.

Development Policy Review 25, 355–377.

Bi, H., Birk, E., Turner, J., Lambert, M., & Jurskis, V., 2001. Converting stem volume to biomass with

additivity, bias correction, and confidence bands for two Australian tree species. New Zealand

Journal of Forestry Science, 31, 298–319.

Birchall, J., 2004. Co-operatives and the Millennium Development Goals. International Labour

Organization (ILO), Geneva.

Bishop, T.H., 1999. Valuing Forests: A review of methods and applications in developing countries. IIED,

London.

Boardman, A.E., Greenberg, D.H., Vining, A.R., & Weimer, D.L., 2001. Cost-benefit analysis: Concepts

and practice (2nd

ed.). Prentice Hall, Inc., Upper Saddle River, New Jersey.

Bond, I., Grieg-Gran, M., Wertz-Kanounnikoff, S., Hazlewood, P., Wunder, S., & Angelsen, A., 2009.

Incentives to sustain forest ecosystem services: A review and lessons for REDD. Natural Resource

Issues No. 16. International Institute for Environment and Development, London, UK, with CIFOR,

Bogor, Indonesia, and World Resources Institute, Washington D.C., USA.

Boucher, D., 2008. Out of the Woods: A Realistic Role for Tropical Forests in Curbing Global Warming.

Union of Concerned Scientists. UCS Publications, Cambridge, UK.

Boyle, K.J., Welsh, M.P., & Bishop, R.C., 1988. Validation of empirical measures of welfare change:

Comment. Land Economics 64, 94-98.

Brandeis, T.J., Delaney, M., Parresol, B.R., & Royer, L., 2006. Development of equations for predicting

Puerto Rican subtropical dry forest biomass and volume. Forest Ecology and Management 233,

133–142.

135

Bremner, J.M., & Mulvaney, C.S., 1982. Nitrogen-total. In: AL Page, et al. (eds) Methods of soil analysis:

Part 2. Chemical and microbiological properties. ASA Monograph Number 9: 595–624.

Brown, S., 1997. Estimating biomass and biomass change of tropical forests: a primer UN FAO Forestry

Paper 134, Rome.

Brown, S., Gillespie, A.J.R., & Lugo, A.E., 1989. Biomass estimation methods for tropical forests with

applications to forest inventory data. Forest Science 35, 881–902.

Bureau of Plan and Finance (BoPF), 2010. Tigray Regenal State Five Years (2010/11-2014/15) Growth &

Transformation Plan. Tigray Bureau of Plan and Finance, Mekelle.

Cambell, B.M., & Luckert, M.K., 2002. Uncovering the Hidden Harvest: Valuation methods for Woodlands

and Forest Resources. Earthscan Publications, London.

Cameron, T.A., & Quiggin, J., 1994. Estimation using contingent valuation data from a “dichotomous

choice with follow-up” questionnaire. Journal of Environmental Economics and Management 27,

218-234.

Cavendish, W., 2000. Empirical Regularities in the Poverty-Environment Relationship of Rural

Households: Evidence from Zimbabwe. World Development 28, 1979-2003.

Central Statistical Authority of Ethiopia (CSA), 2008. Statistical abstract 2008. Central Statistical Authority

of the Federal Democratic Republic of Ethiopia, Addis Ababa, Ethiopia.

Central Statistical Authority of Ethiopia (CSA), 2011. Statistical abstract 2011. Central Statistical Authority

of the Federal Democratic Republic of Ethiopia, Addis Ababa, Ethiopia.

Champ, P.A., Bishop, R.C., Brown, T.C., & McCollum, D.W., 1997. Using donation mechanisms to value

nonuse benefits from public goods. Journal of Environmental Economics and Management 33,

151-162.

Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., Fromard, F.,

Higuchi, N., Kira, T., Lescure, J.P., Nelson, B.W., Ogawa, H., Puig, H., Riѐra, B., & Yamakura, T.,

2005. Tree allometry and improved estimation of carbon stocks and balance in tropical forests.

Oecologia 145, 87–99.

Chave, J., Condit, R., Lao, S., Caspersen, J.P., Foster, R.B., & Hubbell, S.P., 2003. Spatial and temporal

variation in biomass of a tropical forest: results from a large census plot in Panama. Journal of

Ecology 91, 240–252.

Christie, M., Warren, J., Hanley, N., Murphy, K., & Wright, R., 2004. Developing measures for valuing

changes in biodiversity: final report, Report to DEFRA, London.

Cole, T.G. & Ewel, J.J., 2006. Allometric equations for four valuable tropical tree species. Forest Ecology

Management 229, 351–360.

Cooper, J. & Loomis, J., 1992. Sensitivity of willingness to pay estimates to bid design in dichotomous

choice contingent valuation models. Land Economics 68, 211-224.

Cooper, J.C., 1993. Optimal bid selection for dichotomous choice contingent valuation surveys. Journal of

Environmental Economics and Management 24, 25-40.

Dardanoni, V., Modicaa, S. & Peracchi, F., 2011. Regression with imputed covariates: A generalized

missing-indicator approach. Journal of Econometrics 162, 362-368.

Dawkins, H.C., 1961. Estimating total volume of some Caribbean trees. Caribbean Forester 22, 62-63. In:

Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., Fromard,

F., Higuchi, N., Kira, T., Lescure, J.P., Nelson, B.W., Ogawa, H., Puig, H., Riѐra, B., & Yamakura,

136

T., 2005. Tree allometry and improved estimation of carbon stocks and balance in tropical forests.

Oecologia 145, 87–99.

Dehejia, R.H. & Wahba, S., 2002. Propensity score-matching methods for nonexperimental causal

studies. The Review of Economics and Statistics 84, 151-161.

Echessah, P.N., Swallow, B.M., Kamara, D.W., & Curry, J.J., 1997. Willingness to contribute labour and

money to tsetse control: Application of contingent valuation in Busia District, Kenya. World

Development 25, 239-253.

EconomyWatch.com. Inflation forecast for Ethiopia. Available at:

http://www.economywatch.com/economic-statistics/country/Ethiopia/

Ehrenfeld, S., & Ben-Tuvia, S., 1962. The efficiency of statistical simulation procedures. Technometrics 4,

257-275.

Eliasch, J. 2008. Climate change: Financing global forests (the Eliasch Review). Office of Climate

Change, Government of the United Kingdom. London, UK.

Elnagheeb, A.H., & Jordan, J.L., 1995. Comparing three approaches that generate bids for the

referendum contingent valuation method. Journal of Environmental Economics and Management

29, 92-104.

Eom, Y.S., & Larson, D.M., 2006. Valuing housework time from willingness to spend time and money for

environmental quality. Review of Economics of the Household 4: 205-227.

Eshete, A., 2011. The frankincense tree of Ethiopia: ecology, productivity and population dynamics. PhD

thesis, Wageningen University, Wageningen, The Netherlands.

Eshete, A., Streck, F.J., & Bongers, F., 2012a. Frankincense production is determined by tree size and

tapping frequency and intensity. Forest Ecology and Management 274, 136-142.

Eshete, A., Teketay, D., Lemenih, M., & Bongers, F., 2012b. Effect of resin tapping and tree size on

purity, germination and storage behavior of Boswellia papyrifera (Del.) Hochst. Seeds from Metema

District, northwestern Ethiopia. Forest Ecology and Management 269, 31-36.

FAO, 1995. Non-wood forest products 1: Flavors and fragrances of plant origin. Food and Agricultural

Organization of the United Nations (FAO), Rome.

FAO, 2005. New_LocClim CD-ROM. Available at:

http://www.fao.org/nr/climpag/pub/en3_051002_en.asp.

FAO, 2010a. Guidelines on sustainable forest management in drylands of sub-saharan Africa. Arid Zone

Forests and Forestry Working Paper No. 1. Rome.

FAO, 2010b. Global Forest Resources Assessment 2010: Main Report. FAO Forestry Paper 163. Food

and Agricultural Organization of the United Nations, Rome.

Farah, Y., 1994. The milk of the Boswellia forests: frankincense production among the pastoral Somali.

Uppsala University, Uppsala.

Fehrmann, L., & Kleinn, C., 2006. General considerations about the use of allometric equations for

biomass estimation on the example of Norway spruce in central Europe. Forest Ecology and

Management 236, 412-421.

Fichtl, R. & Admasu, A., 1994. Honey Bee Flora of Ethiopia. DED/Margraf Verlag, Germany.

Fisher, I., 1930. The Theory of Interest, as determined by Impatience to Spend Income and Opportunity

to Invest it. Macmillan, New York.

137

Fisher, M., 2004. Household welfare and forest dependence in Southern Malawi. Environment and

Development Economics 9, 135-154.

Friis, B. I., Demissew, S., & Breugel, P., 2010. Atlas of the potential vegetation of Ethiopia. The Royal

Danish Academy of Science Letters, ISBN 978-87-7304-347-9.

Fuwape, J.A., Onyekwelu, J.C., & Adekunle, V.A.J., 2001. Biomass equations and estimation for Gmelina

arborea and Nauclea diderrichii stands in Akure forest reserve. Biomass and Bioenergy 21, 401–

405.

Gebrehiwot, K., 2003. Ecology and management of Boswellia papyrifera (Del.) Hochst. Dry forests in

Tigray, Northern Ethiopia. PhD Thesis. George August University of Göttingen. Cuvillier Verlag,

Göttingen, Germany.

Gebrehiwot, K., Muys B., Haile, M., & Mitloehner, R., 2003. Introducing Boswellia papyrifera (Del.) Hochst

and its non-timber forest product, frankincense. International Forestry Review 5, 348-353.

Gebrehiwot, K., Muys, B., Haile, M., & Mitloehner, R. 2002. Boswellia papyrifera (Del.) Hochst: a tropical

key species in northern Ethiopia, Proceedings of the Conference on International Agricultural

Research for Development, Deutscher Tropentag 2002, Witzenhausen, October 9-11, 2002,

Germany.

Gibbs, H.K., Brown, S., Niles, J.O., & Foley, J.A., 2007. Monitoring and estimating tropical forest carbon

stocks: making REDD a reality. Environmental Research Letters 2, 1-13.

Girma, D., 1998. Non-wood forest products in Ethiopia. EC-FAO Partnership Program (1998–2000).

December 1998, Addis Ababa, Ethiopia.

Girmay, F., 2000. The status of gum Arabic and resins in Ethiopia. Report of the Meeting of the Network

for Natural Gum and Resins in Africa (NGARA) 29th -30

th May, Nairobi, Kenya. In: Tadesse, W.,

Desalegn, G., and Alia, R., (2007). Natural gum and resin bearing species of Ethiopia and their

potential applications. Investigación agrarian: Sistemas y Recursos Forestales 16, 211-221.

Gittinger, J.P., 1982. Economic Analysis of Agricultural Projects. John Hopkins University Press,

Baltimore.

Godoy, R., Overman, H., Demmer, J., Apaza, L., Byron, E., Huanca, T., Leonard, W., Perez, E., Reyes-

Garcia, V., Vadez, V., Wilkie, D., Cubas, A., McSweeney, K., & Brokaw, N., 2002. Local financial

benefits of rain forests: comparative evidence from Amerindian societies in Bolivia and Honduras.

Ecological Economics 40, 397-409.

Goodale, C.L., Heath, L.S., Houghton, R.A., Jenkins, J.C., Kohlmaier, G.H., Kurz, W., Liu, S., Nabuurs,

G.J., Nilsson, S., Shvidenko, A.Z., Apps, M.J., Birdsey, R.A., & Field, C.B., 2002. Forest carbon

sinks in the Northern Hemisphere. Ecological Applications 12, 891–899.

Grieg-Gran, M., 2006/updated 2008. The cost of avoiding deforestation. Report prepared for the Stern

Review/Eliasch Review of the economics of climate change. International Institute for Environment

and Development. London, UK.

Groenendijk, P., Eshete, A., Sterck, F.J., Zuidema, P.A., & Bongers, F., 2012. Limitations to sustainable

frankincense production: blocked regeneration, high adult mortality and declining populations.

Journal of Applied Ecology 49, 164-173.

Gürlük, S., 2006. The estimation of ecosystem services’ value in the region of Misi Rural Development

Project: Results from a contingent valuation survey. Forest Policy and Economics 9: 209-218.

138

Gullison, R.E., Frumhoff, P.C., Canadell, J.G., Field, C.B., Nepstad, D.C., Hayhoe, K., Avissar, R.,

Curran, L.M., Friedlingstein, P., Jones, C.D., & Nobre, C., 2007. Tropical Forests and Climate

Policy. Science 316, 985-986.

Gunawardena, M., & Rowan, J.S., 2005. Economic Valuation of a Mangrove Ecosystem Threatened by

Shrimp Aquaculture in Sri Lanka. Environmental Management 36, 535-550.

Guo, L.B. & Gifford, R.M., 2002. Soil carbon stocks and land use change: a meta analysis. Global

Change Biology 8, 345-360.

Hanemann, W.M., 1984. Welfare evaluations in a contingent valuation experiments with discrete

responses. American Journal of Agricultural Economics 66, 332-341.

Hanemann, W.M., & Kanninen, B., 1998. The statistical analysis of discrete-response contingent

valuation data. In: Bateman, I.J. & Willis, K.G. (eds.). Contingent valuation of environmental

preferences: Assessing theory and practice the USA, Europe and Developing Countries. Oxford

University Press, Oxford.

Hanemann, W.M., Loomis, J., & Kanninen, B., 1991. Statistical efficiency of double bounded dichotomous

choice contingent valuation. American Journal of Agricultural Economics 73, 1255-1263.

Hart, J.H., 1989. The role of wood exudates and extractives in protecting wood from decay. Natural

Products of Woody Plant 2, 861-880. Springer-Verlag, Berlin, Germany.

Heckman, J.J. & Robb, R., 1985. Alternative methods for evaluating the impact of interventions: An

overview. Journal of Econometrics 30, 239-267.

Heckman, J.J., 1997. Sample selection bias as a specification error. Econometrica 47, 153-161.

Hein, L., & Gatzweiler, F., 2006. The economic value of coffee (Coffea arabica) genetic resources.

Ecological Economics 60, 176-185.

Hertel, D., Moser, G., Culmsee, H., Erasmi, S., Horna, V., Schuldt, B., & Leuschner, C.H., 2009. Below-

and above-ground biomass and net primary production in a paleotropical natural forest (Sulawesi,

Indonesia) as compared to neotropical forests. Forest Ecology and Management 258, 1904–1912.

Heywood, V.H., Moor, D.M., Richardson, I.B.K., & Stearn, W.T., 1993. The flowering plants of the world.

2nd

edition. BT Batsford Ltd. London.

Hillier, F. S., & Liebermann, G.J., 1980. Introduction to operations research. Holden-Day Inc., San

Francisco.

Holden, S., Barrett, C.B., & Hagos, F., 2006. Food-for-work for poverty reduction and the promotion of

sustainable land use: can it work? Environment and Development Economics 11, 15-38.

Holmes, T.P., Bergstrom, J.C., Huszar, E., Kask, S.B., & Orr IIIF., 2004. Contingent valuation, net

marginal benefits, and the scale of riparian ecosystem restoration. Ecological Economics 49, 19-

30.

Horton, N.J., & Kleinman, K.P., 2007. Much ado about nothing: a comparison of missing data methods

and software to fit incomplete data regression models. The American Statistician 61, 79–90.

Houghton, R.A., & Goodale, C.L., 2004. Effect of Land-Use Change on the Carbon Balance of Terrestrial

Ecosystems. Ecosystems and Land Use Change, Geographical Monograph Series 153, American

Geophysical Union, Washington, DC.

Houghton, R.A., Lawrence, K.T., Hackler, J.L., & Brown, S., 2001. The spatial distribution of forest

biomass in the Brazilian Amazon: a comparison of estimates. Global Change Biology 7, 731-746.

139

House, J.I., & Hall, D.O., 2001. Productivity of tropical savannas and grasslands. In: Dias, A.T.C., Mattos,

E.A., Vieira, S.A., Azeredo, J.V., & Scarano, F.R., 2006. Aboveground biomass stock of native

woodland on a Brazilian sandy coastal plain: Estimates based on the dominant tree species. Forest

Ecology and Management 226, 364-367.

Howes, F.N., 1950. Age-old resin of the Mediterranean region and their uses. Economic Botany 4, 1307-

1316.

Hsiao, C., 1979. Linear regression using both temporarily aggregated and temporarily disaggregated

data. Journal of Econometrics 10, 243-252.

Hung, L.T., Loomis, J.B., & Thinh V.T., 2007. Comparing money and labour payment in contingent

valuation: The case of forest fire prevention in Vietnamese context. International Development 19,

173-185.

Ichino, A., Mealli, F. & Nannicini, T., 2006. From temporary help jobs to permanent employment: What

can we learn from matching estimators and their sensitivity? IZA Discussion paper No. 2149.

Institute for the Study of Labour, Bonn.

Imbens, G., 2004. Nonparametric estimation of average treatment effects under exogeneity: a review.

The review of Economics and Statistics 86, 4-29.

International Emissions Trading Association (IETA), 2012. IETA Information Note on Design Issues for

REDD+ Mechanism No.2: The Nested Approach. Available at:

http://www.ieta.org/assets/LUWG/ieta%20redd%20information%20note%203-%20nested%20approach.pdf

IPCC, 2007. Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth

Assessment Report of the Intergovernmental Panel on Climate Change [B. Metz, O.R. Davidson,

P.R. Bosch, R. Dave, L.A. Meyer (eds)], Cambridge University Press, Cambridge, United Kingdom

and New York, NY, USA.

Johansson, P.O., 1991: An introduction to modern welfare economics. Cambridge University Press,

Cambridge.

Kamuanga, M., Swallow, B.M., Sigue´, H., & Bauer, B., 2001. Evaluating contingent and actual

contributions to a local public good: Tsetse control in the Yale agro-pastoral zone, Burkina Faso.

Ecological Economics 39, 115-130.

Kanninen, B.J., 1995. Bias in discrete response contingent valuation. Journal of Environmental

Economics and Management 28, 114-25.

Keftasa, D., 1988. Role of crop residues as livestock feed in Ethiopian highlands. In: Pasture Network for

Eastern and Southern Africa (PANESA), 1988. African forage plant genetic resources, evaluation

of forage germplasm and extensive livestock production systems. Proceedings of the Third

Workshop held at the International Conference Centre, Arusha, Tanzania, 27-30 April 1987. ILCA,

Addis Ababa.

Kengen, S., 1997. Forest Valuation for Decision-making: Lessons of Experiences and Proposals for

Implementation. FAO, Rome.

Ketterings, Q.M., Coe, R., van Noordwijk, M., Ambagau, Y., & Palm, C.A., 2001. Reducing uncertainty in

the use of allometric biomass equations for predicting aboveground tree biomass in mixed

secondary forests. Forest Ecology and Management 146, 199–209.

Kim, S. & Dale, B.E., 2004. Global potential bioethanol production from wasted crops and crop residues.

Biomass and Bioenergy 26, 361-375.

140

Kindermann, G., Obersteiner, M., Sohngen, B. Sathaye, J., Andrasko, K., Rametsteiner, E.,

Schlamadinger, B., Wunder, S., & Beach, R., 2008. Global cost estimates of reducing carbon

emissions through avoided deforestation. Proceedings of the National Academy of Sciences 105,

10302-10307.

Kleijen, J.P.C., 1975. Statistical techniques in simulation. Marcel Dekker, New York.

Kolstad, C.D., 2000. Environmental Economics. Oxford University Press, New York, Oxford.

Krutilla, J.V., 1967. Conservation reconsidered. American Economic Review 47, 777-786.

Labao, R., Francisco, H., Harder, D., & Santos, F.I., 2008. Do colored photographs affect willingness to

pay responses for endangered species conservation? Environmental and Resource Economics 40,

251-264.

Landon, J.R., 1991. Booker Tropical Soil Manual: A handbook for soil survey and agricultural land

evaluation in the tropics and subtropics. Longman Scientific & Technical, Longman Group Ltd., UK.

ISBN: 0-582-00557-4.

Latt, C.R., Nair, O.K.R., & Kang, B.T., 2000. Interactions among cutting frequency, reserve

carbohydrates, and post cutting biomass production in Gliricidia sepium and Leuceana

leucocephala. Agroforestry Systems 50, 27-46.

Lemenih, M., Feleke, S., & Taesse, W., 2007. Constraints to smallholders production of frankincense in

Metema district, North-western Ethiopia. Journal of Arid Environments 71, 393-403.

Little, R.J.A. & Rubin, D., 2002. Statistical Analysis with Missing Data, 2nd

ed. Wiley.

Little, R.J.A., 1992. Regression with missing X’s: a review. Journal of the American Statistical Association

87, 1227-1237.

Litton, C.M., & Kauffman, J.B., 2008. Allometric models for predicting aboveground biomass in two

widespread woody plants in Hawaii. Biotropica 40, 313-320.

Liu, E., Hsiao, C., Matsumoto, T., & Chou, S., 2009. Maternal full-time employment and overweight

children: Parametric, semi-parametric, and non-parametric assessment. Journal of Econometrics

152, 61-69.

Maertens, M., Colen, L., & Swinnen, J.F.M., 2011. Globalization and poverty in Senegal: a worst case

scenario? European Review of Agricultural Economics 38, 31-54.

Mamo, G., Sjaastad, E., & Vedeld, P., 2007. Economic dependence on forest resources: a case from

Dendi district, Ethiopia, Forest Policy and Economics 9, 916-927.

Marshall, N.T., 1998. Searching for a Cure: Conservation of Medicinal Wildlife Resources in East and

Southern Africa. TRAFFIC International, Cambridge, United Kingdom.

Martin, A.R. & Thomas, S.C., 2011. A Reassessment of Carbon Content in Tropical Trees. PLoS ONE 6,

e23533. doi:10.1371/journal.pone.0023533.

McFadden., D., 1974. Conditional logit analysis of qualitative choice behaviour. In: Zarembka, P. (Ed.),

Frontiers in Econometrics. Academic Press, New York.

Mekuria, W., Veldkamp, E., Tilahun, M., & Olschewski, R., 2011. Economic Valuation of Land

Restoration: The Case of Exclosures Established on Communal Grazing Lands in Tigray, Ethiopia.

Land Degradation and Development 22, 334-344.

Melaku, S., Aregawi, T., & Nigatu, T., 2010. Chemical composition, in vitro dry matter digestibility and in

sacco degradability of selected browse species used as animal feeds under semi-arid conditions in

Northern Ethiopia. Agroforestry Systems 80: 173-184.

141

Mengistu, T., 2011. Physiological ecology of the frankincense tree. PhD thesis, Wageningen University,

Wageningen, The Netherlands, ISBN: 978-90-8585-927-7.

Mengistu, T., Sterck, F.J., Anten, N.P.R., & Bongers, F., 2012. Frankincense tapping reduced

photosynthetic carbon gain in Boswellia papyrifera (Burseraceae) trees. Forest Ecology and

Management 278, 1-8.

Meyerhoff, J. & Liebe, U., 2010. Determinants of protest responses in environmental valuation: A meta-

study. Ecological Economics 70, 366-374.

Michie, C.A. & Cooper, E., 1991. Frankincense and myrrh as remedies in children. Journal of the Royal

Society of Medicine 84, 602-605.

Millennium Ecosystem Assessment (MEA), 2003. Ecosystems and human well-being: a framework for

assessment. Millennium Ecosystem Series. Island Press. Wahington DC, USA.

Moges, Y., & Kindu, M., 2006. Effects of fencing and ground cultivation on natural regeneration of

Boswellia papyrifera in Metema Wereda, Ethiopia. Journal of The Drylands 1, 45-51.

Mortimore, M., Anderson, S., Cotula, L., Davies,J., Faccer, K., Hesse, C., Morton, J., Nyangena, W.,

Skinner, J., & Wolfangel, C., 2009. Dryland Opportunities: A new paradigm for people, ecosystems

and development, IUCN, Gland, Switzerland; IIED, London, UK and UNDP/DDC, Nairobi, Kenya.

Muradian, R., & Martinez-Alier J., 2001. Trade and the environment: from a ‘Southern’ perspective.

Ecological Economics 36, 281-297.

Murty, D., Kirschbaum, M.U.F., McMurtrie, R. E., & McGilvary, H., 2002. Does conversion of forest to

agricultural land change soil carbon and nitrogen? a review of the literature. Global Change Biology

8, 105-123.

Muys, B., Gebrehiwot, K., & Bruneel, S., 2006. The ecology and silviculture of dryland forest rehabilitation

in Ethiopia. Journal of The Drylands 1, 1-2.

Navar, J., 2009. Allometric equations for tree species and carbon stocks for forests of northwestern

Mexico. Forest Ecology and Management 257, 427-434.

Neef, T., & Henders, S., 2007. Guidebook to Markets and Commercialisation of Forestry CDM projects.

Tropical Agricultural Research and Higher Education Center (CATIE). In: Olschewski, R. & Benitez,

P.C., 2010. Optimizing joint production of timber and carbon sequestration of afforestation projects.

Journal of Forest Economics 16, 1-10.

Negussie, A., Aerts, R., Gebrehiwot, G., & Muys, B., 2008. Seedling mortality causes recruitment

limitation of Boswellia papyrifera in northern Ethiopia. Journal of Arid Environments 72: 378-383.

Neumann, R.P. & Hirsch, E., 2000. Commercialization of Non-timber Forest Products: Review and

Analysis of Research, CIFOR, Bogor.

Newell, E., Mulkey, S., & Wright, S., 2002. Seasonal patterns of carbohydrate storage in four tropical tree

species. Oecologia 131, 333-342.

Niklas, K.J., 2004. Plant allometry: is there a grand unifying theory? Biological Reviews 79, 871–889.

Nogueira, E.M., Fearnside, P.M., Nelson, B.W., Barbosa, R.I., & Keizer, E.W.H., 2008. Estimates of forest

biomass in the Brazilian Amazon: new allometric equations and adjustments to biomass from

wood-volume inventories. Forest Ecology and Management 256, 1853–1867.

Ogawa, H., Yoda, K., Ogino, K., & Kira, T., 1965. Comparative ecological studies on three main types of

forest vegetation in Thailand II Plant biomass. Nature and Life in Southeast Asia 4, 49-80. In:

Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., Fromard,

142

F., Higuchi, N., Kira, T., Lescure, J.P., Nelson, B.W., Ogawa, H., Puig, H., Riѐra, B., & Yamakura,

T., 2005. Tree allometry and improved estimation of carbon stocks and balance in tropical forests.

Oecologia 145, 87–99.

Ogbazghi, W., 2001. The distribution and regeneration of Boswellia papyrifera (del.) Hochst in Eritrea.

Tropical Resource Management Papers, No. 35. PhD. Desertation. Wageningen University.

Ogbazghi, W., Rijkers, T., Wessel, M., & Bongers, F., 2006. The distribution of frankincense tree

Boswellia papyrifera in Eritrea: the role of environment and land use. Biogeography 33, 524-535.

Olschewski, R., & Benitez, P.C., 2010. Optimizing joint production of timber and carbon sequestration of

afforestation projects. Journal of Forest Economics 16, 1-10.

Olschewski, R., & Benitez, P.C., 2005. Secondary forests as temporary carbon sinks? The economic

impact of accounting methods on reforestation projects in the tropics. Ecological Economics 55,

380-394.

Olsen, S.R., & Sommers, L.E., 1982. Phosphorus. In A.L. Page, et al. (eds).Methods of soil analysis: part

2. Chemical and microbiological properties. ASA Monograph 9, 403-430.

Ong, J.E., Gong, W.K., & Wong, C.H., 2004. Allometry and partitioning of the mangrove Rhizophora

apiculata. Forest Ecology and Management 188, 395-408.

Pearce, D., 1993. Economic Values and the Natural World. Earthscan, London.

Pearce, D.W., & Turner, R.K., 1990. Economics of natural resources and the environment. BPCC

Wheatsons Ltd., Exeter, U.K.

Perman, R., Ma, Y., McGilvary, J., & Common, M., 2003. Natural Resource and Environmental

Economics. Pearson Education Ltd., London.

Phillips, M.A., & Croteau, R.B., 1999. Resin-based defenses in conifers. Trends in Plant Sciences 4, 184-

190.

Pilli, R., Anfodillo, T., & Carrer, M., 2006. Towards a functional and simplified allometry for estimating

forest biomass. Forest Ecology and Management 237, 583-593.

PointCarbon, 2011. Carbon 2011. Thomson Reuters Point Carbon. Available at:

http://www.pointcarbon.com

Poorter, H., Niklas, K.J., Reich, P.B., Oleksyn, J., Poot, P., & Mommer, L., 2011. Biomass allocation to

leaves, stems and roots: meta-analyses of interspecific variation and environmental control. New

Phytologist, Tansley Review: www.newphytologist.com

Ramirez, O.A., Carpio, C.E., Ortiz, R., & Finnegan, N.B., 2002. Economic Value of the Carbon Sink

Services of Tropical Secondary Forests and Its Management Implications. Environmental and

Resource Economics 21, 23-46.

Reddy, S.R.C., & Chakravarty, S.P., 1999. Forest dependence and income distribution in a subsistence

economy: evidence from India, World Development 27, 1141-1149.

Reichhuber, A., & Requate, T., 2006. Alternative Use Systems for the Remaining Cloud Forest in Ethiopia

and the Role of Arabica Coffee - A Cost-Benefit Analysis. Economics Working Paper No. 2007-07.

Christian-Albreschts – Universtiate Kiel. Department of Economics, Berlin.

Rijkers, T., Ogbazghi, W., Wessel, M., & Bongers, F., 2006. The effect of tapping for frankincense on

sexual reproduction in Boswellia papyrifera. Journal of Applied Ecology 43, 1188-1195.

Rock, M., 1996. Pollution intensity of GDP and trade policy: can the World Bank be wrong? World

Development 24, 471-479.

143

Rosenbaum, P.R., & Rubin, D.B., 1983. The central role of propensity score in observational studies for

causal effects. Biometrika 70, 41-55.

Rubin, D., 1974. Estimating causal effects of treatments in randomized and non-randomized studies.

Journal of Educational Psychology 66, 688-701.

Samba, S.A.N., Camire, C., & Margolis, H.A., 2001. Allometry and rainfall interception of Cordyla pinnata

in a semi-arid agroforestry parkland, Senegal. Forest Ecology and Management 154, 277-288.

Sampaio, E., Gasson, P., Baracta, A., Culter, D., Pareyn, F., & Lima, K.C., 2010. Tree biomass

estimation in regenerating areas of tropical dry vegetation in northeast Brazil. Forest Ecology and

Management 259, 1135-1140.

Sampaio, E.V.S.B., & Silva, G.C., 2005. Biomass equations for Brazilian semiarid caatinga plants. Acta

Botanica Brasilica 19, 935-943.

Schillaci, D., Arizza, V., Dayton, T., Camarda, L., & Di Dtefano, V., 2008. In vitro anti-biofilm activity of

Boswellia spp. oleogum resin essential oils. Letters in Applied Microbiology 47, 433-438.

Shackleton, C.M., & Shackleton, S.E., 2006. Household wealth, status and natural resource use in the

Kat River Valley, South Africa. Ecological Economics 57, 306-317.

Shackleton, S., & Gumbo, D., 2010. Contribution of Non-wood Forest Products to Livelihoods and

Poverty Alleviation. In: Chidumayo, E.N., & Gumbo, D.J., (Eds.), 2010. The Dry Forests and

Woodlands of Africa: Managing for Products and Services. Earthscan Ltd., UK.

Simmons, R., & Birchall, J., 2008. The role of co-operatives in poverty reduction: Network perspectives.

The Journal of Socio-Economics, 37, 2131–2140.

Singer, J., 1998. Using SAS proc mixed to fit multilevel models, hierarchical models, and individual

growth curves. Journal of Educational and Behavior Statistics 24, 323–355.

Sjaastad, E., Angelsen, A., Vedeld, P., & Bojö, J., 2005. What is environmental income? Ecological

Economics 55, 37-46.

Smith, T.J., & Whelan, K.R.T., 2006. Development of allometric relations for three mangrove species in

South Florida for use in the greater everglades ecosystem restoration. Wetlands Ecology and

Management 14, 409-419.

Stegen, J.C., Swenson, N.G., Valencia, R., Enquist, B.J., & Thompson, J., 2009. Above-ground forest

biomass is not consistently related to wood density in tropical forests. Global Ecology and

Biogeography 18, 617-625.

Stephens, B.B, Gurney, K.R., Tans, P.P., Sweeney, C., Peters, W.,Bruhwiler, L., Ciais, P., Ramonet, M.,

Bousquet, P., Nakazawa, T., Aoki, S., Machida, T., Inoue, G., Vinnichenko, N., Lloyd, J., Jordan,

A., Heimann, M., Shibistova, O., Langenfelds, R.L., Steele, L.P., Francey, R.J., & Denning S.,

2007. Weak Northern and strong tropical land carbon uptake from vertical profiles of atmospheric

CO2. Science 316, 1732-1735.

Stern, N., 2006. The economics of climate change (the Stern Review). Cambridge University Press.

Cambridge, UK.

Stiles, D., 1988. Arid land plants for economic development and desertification control. Desertification

control Bulletin 17, 18-21.

Stock, J.H., & Yogo, M., 2005. Testing for Weak Instruments in Linear IV Regression. Working paper

version: NBER Technical Working Paper 284. Available at: http://www.nber.org/papers/T0284.

144

Stone, M., 1974. Cross-validatory choice and assessment of statistical predictions. Royal Statistical

Society Series B (Methodological) 36, 111-147.

Sunnichan, V.G., Mohan, R.H.Y., & Shivanna, K.R., 2005. Reproductive biology of Boswellia serrata, the

source of salai guggul, an important gum-resin. Botanical Journal of the Linnean Society 147, 73-

82.

Swallow, B.M., & Woundyalew, M., 1994. Evaluating willingness to contribute to a local public good:

Application of contingent valuation to tsetse control in Ethiopia. Ecological Economics 11, 153-161.

Tadesse, W., Desalegn, G., & Alia, R., 2007. Natural gum and resin bearing species of Ethiopia and their

potential applications. Investigación Agraria: Sistemas y Recursos Forestales 16, 211-221.

Ter-Mikaelian, M.T. & Korzukhin, M.D., 1997. Biomass equations for sixty-five North American tree

species. Forest Ecology and Management 97, 1-24.

Tesfay, G., 2006. Agriculture, resource management and institutions: a socio-economic analysis of

households in Tigray, Ethiopia. PhD Thesis. Wageningen University, The Netherlands.

Tilahun, M., Muys, B., Mathijs, E., Kleinn, C., Olschewski, R., & Gebrehiwot, K., 2011. Frankincense yield

assessment and modeling in closed and grazed Boswellia papyrifera woodlands of Tigray,

Northern Ethiopia. Journal of Arid Environments 75, 695-702.

Tilahun, M., R., Olschewski, C. Kleinn, & Gebrehiwot, K., 2007. Economic analysis of closing degraded

Boswellia papyrifera dry forest from human interventions - A study from Tigray, Northern Ethiopia.

Forest Policy and Economics 9, 996-1005.

Tucker, O.A., 1986. Frankincense and myrrh. Economic Botany 40, 425-433.

UNFCCC, 2003. Modalities and procedures for afforestation and reforestation project activities under the

clean development mechanism in the first commitment period of the Kyoto Protocol. Decion-/CP.9.

Available at: http://www.unfccc.int.

United Nations (UN), 2008. Resolution adopted by the General Assembly 62/98: Non-legally binding

instruments on all types of forests. Accessed on July 20, 2012 from: http://daccess-dds-

ny.un.org/doc/UNDOC/GEN/N07/469/65/PDF/N0746965.pdf?OpenElement.

Urquiza-Haas, T., Dolman, P.M., & Peres, C.A., 2007. Regional scale variation in forest structure and

biomass in the Yucatan Peninsula, Mexico: effects of forest disturbance. Forest Ecology and

Management 247, 80-90.

Vedeld, P., Angelsen, A., Bojo, J., Sjaastad, E., & Berg, G.K., 2007. Forest environmental incomes and

the rural poor. Forest Policy and Economics 9, 869-879.

Verbist, B., & Muys, B., 2010. Dryland areas, forgotten by REDD? KLIMOS Policy brief 2, KLIMOS,

Leuven.

Vollesen, K., 1989. Burseraceae. In: Hedberg, I. and Edwards, S., (editors): Flora of Ethiopia, Volume 3.

Addis Ababa University.

Vondolia, G.K., 2011. What do respondents bring into contingent valuation? A comparison of monetary

and labour payment vehicles. Working Papers in Economics No. 508, Göteborg University.

Walkley, A., & Black, I.A., 1934. An examination of the Degtjareff method for determining organic carbon

in soils: Effect of variations in digestion conditions and of inorganic soil constituents. Soil Science

63, 251-263.

Walsh, R G., Loomis, J. B., & Gillman, R. A., 1984. Valuing Option, Existence, and Bequest Demand for

Wilderness. Land Economics 60, 14-29.

145

Walter, S., 2001. Non-wood forest products in Africa: a regional and national review. FAO, Rome.

Wang, H., Hall, C.A.S., Scatena, F.N., Fetcher, N., & Wu, W., 2003. Modeling the Spatial and temporal

variability in climate and primary productivity across the Luquillo mountains, Puerto Rico. Forest

Ecology and Management 179, 69-94.

Wanyama, F. O., Develtere, P. & Pollet, I., 2008. Encountering the Evidence: Cooperatives and Poverty

Reduction in Africa. Journal of Cooperative studies 41, 16-27.

Weisbrod, B.A., 1964. Collective-Consuption Services of Individualized-Consuption Goods. The Quarterly

Journal of Economics 78, 471-477.

Weiss, E., 1987. Guide to plants tolerant of arid and semi-arid conditions: nomenclature and potential

uses.Republic of Kenya, Ministry of Livestock Development, Range Management, Division, and

German Agricultural Team, Nairobi.

West, P.C., Gibbs, H.K., Monfreda, C., Wagner, J., Barford, C.C., Carpenter, S.R., & Foley, J.A., 2010.

Trading carbon for food: Global comparison of carbon stocks VS. crop yields on agricultural land.

PNAS 107, 19645-19648.

White, D., Minang, P., Pagiola, S., & Swallow, B., 2011. Estimating the opportunity costs of REDD+: A

training manual. The World Bank, Washington DC, USA.

White, F., 1983. The Vegetation of Africa: a descriptive memoir to accompany the

UNESCO/AETFAT/UNSO vegetation map of Africa. UNESCO, Natural Resources Research 20, 1-

356.

Wise, R., & Cacho, O., 2005. A bioeconomic analysis of carbon sequestration in farm forestry: a

simulation study of Gliricidia sepium. Agroforestry systems 64, 237-250.

Woldehanna, T., 2004. The experience of measuring and monitoring poverty in Ethiopia. Paper

Presented at the Inaugural Meeting of the Poverty Analysis and Data Initiative (PADI). May 6-8,

Mombassa, Kenya.

Wooldridge, J. M., 2002. Econometric Analysis of Cross Section and Panel Data. The MIT Press,

Cambridge.

World Bank, 2011. World Development Indicators. The World Bank, ISBN: 978-0-8213-8709-2.

World Bank. 2010. World Development Indicators. The World Bank, ISBN: 978-0-8213-8232-56.

Yesuf, M., Di Falco, S., Deressa, T., Ringler, C., & Kohlin, G., 2008. The Impact of Climate Change and

Adaptation on Food Production in Low-Income Countries: Evidence from the Nile Basin, Ethiopia.

IFPRI, Washington DC.

Zerbe R.O.Jr., & Bellas, A.S., 2006. A Primer for Benefit-Cost Analysis. Edward Elgar Publishing Ltd.,

Cheltenham, U.K.

146

Appendices

Appendix 2A: Distribution of sample and total number of Boswellia papyrifera trees in three diameter classes of each of the five

Boswellia papyrifera populations

Population DBH-Class DBH in cm

(Min –Max)

Number of trees in all the

subplots of a population

Number of

sample trees

Abergelle I Small [9.6, 17.6) 122 25

Medium [17.6, 21.9) 123 44

Large [21.9, 32] 122 51

Subtotal 367 120

Abergelle II Small [9.6, 16.1) 73 17

Medium [16.1, 20.3) 72 22

Large [20.3, 40.5] 72 21

Subtotal 217 60

Abergelle III Small [11.0, 17.7) 82 32

Medium [17.7, 21.1) 83 41

Large [21.1, 35.5] 83 47

Subtotal 248 120

K.Humera I Small [9.5, 16.0) 62 13

Medium [16.0, 21.1) 62 20

Large [21.1, 33.9] 62 27

Total 186 60

K.Humera II Small [8.0, 19.6) 39 16

Medium [19.6, 23.6) 38 19

Large [23.6, 34.8] 38 25

Subtotal 115 60

Total 1133 420

148

Appendix 2B: Units of measurment and sample sizes for variables measured in the experimental plots.

Varibles Variable type Unit Sample size

Dependent variables

Frankincese yield per tree Ratio Gram 336

Number of flowers per tree Ratio Count 420

Number of fruts per inflorecense Ratio Count 420

Independent variables

Diameter (DBH) Ratio Centimeter 420

Height Ratio Meter 420

Stem height Ratio Meter 420

Crown diameter 1 Ratio Meter 420

Crown diameter 2 Ratio Meter 420

Average crown diameter Ratio Meter 420

Bark color Binary (1= Yellowish,

0 = Orange)

420

Altitude Ratio Meter above sea

level

140

Slope Ratio % 140

Soil depth Ratio Meter 140

Stoniness Ratio % 140

pH-water Ratio 175

Electric conductivity Ratio

µS/cm

(microSiemens

per centi meter)

175

Organic carbon Ratio % 175

Total Nitrogen Ratio % 175

Available phosphorous Ratio miligram/kilogram 175

Bulk density Ratio gram/cubic

centimeter

14

149

Appendix 3A: Distribution and number of sample trees selected for biomass measurement through destructive sampling by

population and diameter class.

Population DBH-Class DBH in cm

(Min –Max)

Number of

trees in all

subplots of a

population

Number of sample

trees

Number destructively

sampled trees for allometric

modelling

Abergelle I Small [9.6, 17.6) 122 25 3

Medium [17.6, 21.9) 123 44 1

Large [21.9, 32] 122 51 5*

Subtotal 367 120 9

Abergelle II Small [9.6, 16.1) 73 17 0

Medium [16.1, 20.3) 72 22 0

Large [20.3, 40.5] 72 21 0

Subtotal 217 60 0

Abergelle III Small [11.0, 17.7) 82 32 3*

Medium [17.7, 21.1) 83 41 2*

Large [21.1, 35.5] 83 47 3*

Subtotal 248 120 8

K.Humera I Small [9.5, 16.0) 62 13 2

Medium [16.0, 21.1) 62 20 2

Large [21.1, 33.9] 62 27 2

Subtotal 186 60 6

K.Humera II Small [8.0, 19.6) 39 16 4

Medium [19.6, 23.6) 38 19 2

Large [23.6, 34.8] 38 25 1

Subtotal 115 60 7

Total 1133 420 30

*represents the diameter classes from which the four sample trees on which root excavation was carried out.

150

Appendix 3B: Units of measurement and sample sizes for variables measured for allometric modelling using destructively sampled

Boswellia trees

Varibles Variable type Unit Sample size

Sub-samples form sample trees

Fresh leaf subsample weight Ratio Gram 30

Fresh branches subsample weight Ratio Gram 30

Fresh stem subsample discs weight Ratio Gram 30

Fresh roots subsample weights Ratio Gram 4

Oven dry leaf subsample weight Ratio Gram 30

Oven dry branches subsample weight Ratio Gram 30

Oven dry stem subsample discs weight Ratio Gram 30

Oven dry roots subsample weights Ratio Gram 4

Tree level variables

Diameter Ratio Centimeter 30

Height Ratio Meter 30

Crown diameter Ratio Meter 30

Crown area Ratio Square meter 30

Fresh leaf weight Ratio Kilogram 30

Fresh branches weight Ratio Kilogram 30

Fresh stem biomass weight Ratio Kilogram 30

Fresh root biomass weigh Ratio Kilogram 4

Root to shoot ratio Ratio 4

Volume of oven dry stem subsample discs Ratio Cubic centimeter 30

Wood density Ratio gram/cubic

centimeter

30

Dry leaf biomass Ratio Kilogram 30

Dry branch biomass Ratio Kilogram 30

Dry stem biomass Ratio Kilogram 30

Dry above ground biomass Ratio Kilogram 30

Dry root biomass Ratio Kilogram 30

151

APPENDIX 4A: Questionnaire for contingent valuation and socioeconomic survey of rural households in Boswellia forest

areas of northern Ethiopia

I. Household’s background on the uses of Boswellia papyrifera forest (Fodder and Frankincense)

1. Are you or any member in the household engaged in frankincense tapping? (Please circle ONE response)

0. No

1. Yes

2. Are you or any member in the household participating as shareholder in the local cooperative that is engaged in frankincense

production and trading? (Please circle ONE response)

0. No Skip to question No. 9.

1. Yes

3. In which year did you join the local cooperative as a shareholder? __________________________

4. How many shares do you own as member of the local cooperative? ___________________ shares

5. How much is the value of one share in Birr? _______________Birr/share

6. Have you ever received dividend from your cooperative in the last 7 years?

0. No

1. Yes

7. How many times did you receive dividend from your cooperative since you become a shareholder?

______________________

8. In which year (s) did you receive and how much was the amount each year?

Year in E.C 1995/96 1996/97 1997/98 1998/99 1999/00 2000/01 2001/02

Dividend received.

0 = No , 1 = Yes

Dividend received in

Birr

9. Have you ever used the leaf of Boswellia papyrifera tree as a fodder for your cattle and/or goats and sheep? (Please circle

ONE response)

Location of the Household

Household Code: __________________________________________

Name of the Household Head: _______________________________

Sex of the Household head: Male =1 Female = 0

Household/member engaged in frankincense tapping:

Yes = 1 No = 0

Household/member is share holder in a cooperative engaged in

frankincense production and trade:

Yes = 1 No = 0

Zone: Central = 1 Western = 2

Wereda:

Abergelle = 1 Kafta Humera = 2

Welkait = 3 4= Tahtay Adiabo

Tabia: _______________________

Kushet: ______________________

Data Collection

Enumerator’s Name: __________________________

Enumeration date: ___________________________

Time interview started: ______:_______

Time interview completed: ______:_____

Data Entry

(To be filled during data entry)

Name:____________________________

Date: _____________________________

Questionnaire code in Bold:

Bid A: WTP = (19; 38; 10) and WTCL= (2; 4; 1)

Bid B: WTP= (47; 94; 24) and WTCL= (5; 10; 3)

Bid C: WTP= (68; 136; 34) and WTCL= (7; 14; 4)

Bid D: WTP= (84; 168; 42) and WTCL= (9; 18; 5)

Bid E: WTP= (113; 226; 57) and WTCL= (12; 24; 6)

152

0. No skip to section II

1. Yes

10. How often are you using leaf of Boswellia papyrifera tree as fodder? (Please circle ONE response)

1. Every year

2. Occasionally when there is serious draught and shortage of fodder

3. Rarely

II. Households’ Willingness to Pay (WTP) for Boswellia papyrifera forest conservation

2.1. Description of the Boswellia papyrifera forest

2.1.1. Uses of Boswella papyrifera

No. Parts (product) of Boswellia

papyrifera tree

Uses

1 Frankincense Church, fragrance in coffee ceremony, medicine, perfume, food flavouring,

source of foreign currency through export, off-farm income source

2 Stem Construction of house, farm implements, fuel wood, fencing material

3 Branches Fuel wood, fencing material

4 Leaves Fodder for livestock

5 Flowers Nectar and pollen for bee keeping

6 Roots Protect soil erosion

The tree and the forest Shade, combat desertification, home for wildlife, regulate climate

11. How do you evaluate your previous knowledge about the uses of Boswellia papyrifera forest compared to the information you

have now in the above table? (Please circle one response).

0. I knew all of the uses listed in the table above and nothing is new for me.

1. I knew most of the uses listed in the table above and only few are new to me.

2. I knew few of the uses listed in the table. I learned more now and understand all the uses very well.

3. I knew few of the uses listed in the table, but now I learned more but could not understand all the uses very well.

4. I did not know any of the uses listed in the table, but now I learned and understood all of the uses very well.

5. I did not know any of the uses listed in the table, but now I learned but still could not understand all uses very well.

12. To what level is your household benefiting from Boswellia papyrifera forest in terms of the above uses from the different parts

of the tree in your area?

(Please, circle one response)

1. Not benefiting at all

2. Benefiting very little

3. Benefiting partially

4. Fully benefiting

2.1.1. Degradation of the resource

Despite the social, cultural, economic and environmental benefits that Boswellia forests have been providing, the species is under

threat. It is under high risks of extinction due to continuous and prolonged deforestation for agricultural land, over exploitation for its

frankincense and free grazing that hinders natural regeneration of the species.

In the 1970s, Tigray had around 500, 000 hectares of land covered by Boswellia papyrifera forest (Wilson, 1977). These decreased

to 330,000 hectares (BoFED, 2000) according to an estimate made 10 years ago. With the continuous trend of deforestation in the

region, the current population will be less than the 330,000 hectares.

Current studies indicate that there is problem of natural regeneration of the tree. It is usual not to see small trees, saplings and

seedlings of Boswellia trees in remaining stands of Boswellia forest in the region. The following pictures taken in 2009 from

153

Abergelle and Kafta Humera Boswellia forest areas are few examples of the current degradation status of the resource in Tigray

(Please show the separate big size Photo of pictures below to the respondent for 3 minutes and ask him/her questions no 13 to 15).

13. Have you noticed the three pictures?

0. No

1. Yes

14. In which of the above sates would you prefer to transfer

the Boswellia papyrifera forest areas of your village to

your grandchildren for their future use?

1. In state A

2. In state B

3. In state C

15. To what extent are you concerned about the rate of

deforestation of Boswellia papyrifera forest and the risk

of extinction of this resource in Tigray? (Please circle

one response)

1. Not at all concerned

2. It concerns me some how

3. I am concerned

4. I am very much concerned

2.1.3. Need for conservation/preservation

In order to prevent the Boswellia papyrifera forest from extinction and keep it for the future use by existing and future generation,

there is a need to conserve the forest. This conservation policy has the following goals:

Goal 1: Keeping the Boswellia forest as option of meeting future demand for the different products and services from the

resource.

Goal 2: To realize that existence of the resource and curb the possibility of extinction

Goal 3: Improve the welfare of future generation of the society

16. How interested are you in the goal of keeping option of using the resource in future (Goal 1)? On a scale from ‘1’ to ‘10’,

where ‘1’ is ‘Not At All Interested’ and ‘10’ is ‘Very interested’, how interested are you? (Please circle ONE response).

Not At All

Interested

Very

Interested

1 2 3 4 5 6 7 8 9 10

17. How interested are you in the goal of realizing the existence of the Boswellia papyrifera (Goal 2), provided that you do not

have the intention of using the resource directly in the future? On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Not At All Interested’

and ‘10’ is ‘Very interested’, how interested are you? (Please circle ONE response)

Not At All

Interested

Very

Interested

1 2 3 4 5 6 7 8 9 10

154

18. How interested are you in the goal of improving the welfare of the future generation (welfare of your grand children) of the

conservation policy on Boswellia papyrifera forest (Goal 3)? On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Not At All Interested’ and

‘10’ is ‘Very interested’, how interested are you? (Please circle ONE response)

Not At All

Interested

Very

Interested

1 2 3 4 5 6 7 8 9 10

19. How interested are you in the overall planned conservation policy? On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Not At All

Interested’ and ‘10’ is ‘Very interested’, how interested are you? (Please circle ONE response)

Not At All

Interested

Very

Interested

1 2 3 4 5 6 7 8 9 10

2.2. Method of provision, Payment Vehicle, Decision Rule and Time Frame of Payment

This conservation policy requires both financial and human resources. Let us consider that there is a plan to conserve Boswellia

papyrifera forest area in your village that aims to meet the above 3 goals and you are required to make voluntary contributions in

either money or daily labour terms. The payment of contribution in the form of money will be made every year to the local

government as a tax payment for Boswellia papyrifera forest conservation. The voluntary labour contribution to be made also every

year is in terms of patrolling the conservation site for controlling illegal grazing, cutting of trees, and tapping of frankincense in the

conservation site.

Enough volunteers would have to participate to pay for the conservation program and achieve its goals. However, if after one year

participation were insufficient to fund the conservation program, the local government would cancel the program and guarantee to

refund all the money collected from volunteers as well as the equivalent amount of money for the labour contributions made by

volunteers.

20. Does the fact that a minimum level of volunteers’ participation is required for the conservation program to operate make the

program of less interest to you, more interest, ore does it not affect your interest? (Please circle ONE response)

1. Decreases interest

2. Increases interest

3. Does not affect interest

4. Do not know

21. Does the fact that the local government would refund all the money it collects and the equivalent in terms of money for your

labour contributions, if support is insufficient, make the program of less interest to you, more interest, ore does it not affect

your interest? (Please circle ONE response)

1. Decreases interest

2. Increases interest

3. Does not affect interest

4. Do not know

We would like you to compare this Boswellai papyrifera conservation program with other environmental conservation activities you

might support, like Soil and water conservation works.

155

22. On a scale from ‘1’ to ‘10’, how would you rank the Boswellia papyrifera forest conservation program compared with other

programs, like soil and water conservation? A ‘1’ indicates you view the Boswellia papyrifera conservation program ‘Much

Less Favourably’ and ‘10’ ‘Much More Favourably’. (Please circle ONE response).

Much Less

Favourably

Much More

Favourably

1 2 3 4 5 6 7 8 9 10

2.3. The Contingent Valuation Question Format (Double-bounded Dichotomous-choice)

Given your household’s income and other expenses, we would like you to think about whether or not you would be interested in

joining the Boswellia papyrifera forest conservation program.

23. Would you (your household) contribute for the Boswellia papyrifera conservation program if it costs you Birr 19 per year?

(Please circle ONE response)

0. No If the response is ‘No’, skip to Question 27.

1. Yes

24. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are of

that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would join

the program and pay the extra Birr 19 a year if the conservation work were actually implemented? (Please circle ONE

response).

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

25. Would you (your household) contribute for the Boswellia papyrifera conservation program if it costs you Birr 38 per year?

(Please circle ONE response)

0. No If the response is’No’, skip to Question 27.

1. Yes

26. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are

of that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would

join the program and pay the extra Birr 38 a year if the conservation work were actually implemented? (Please circle ONE

response)

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

27. Would you (your household) contribute for the Boswellia papyrifera conservation program if it costs you Birr 10 per year?

(Please circle ONE response)

0. No If the response is ‘No’, skip to Question 29

1. Yes

28. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are of

that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would join

the program and pay the extra Birr 10 a year if the conservation work were actually implemented? (Please circle ONE

response).

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

156

29. So you think that you would not join the Boswellia papyrifera forest conservation program at a cost of Birr 10 per year.

Which of the following best describes your ‘No’ response to question 27? (Please circle ONE response)

1. I have no interest in the Boswellia papyrifera forest conservation program.

2. I have some interest in the Boswellia papyrifera forest conservation program, but I would not pay anything to join

the program.

3. I would join the Boswellia papyrifera forest conservation program if it costs me some money, but less than Birr 10

per year.

If the contribution to the Boswellia papyrifera forest conservation program is in the form of voluntary labour contribution, you will be

expected to patrol the conservation site for controlling illegal grazing, cutting of trees, and tapping of frankincense in the

conservation site and other works required to meet the goals of the program. So, given number of labour force (household

members aged from 15 years and above) who are physically capable of contributing labour for patrolling at the conservation site,

the labour time you require to accomplish your farming and other household activities as well as community works, we would like

you to think about whether or not you would be interested in joining the Boswellia papyrifera forest conservation program.

30. Would you (your household) contribute for the Boswellia papyrifera conservation program if it requires you 2 full labour

days per year as a guard for patrolling the site given that 1 labour day = 8 hours? (Please circle ONE response)

0. No If the response is ‘No’, skip to Question 34

1. Yes

31. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are

of that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would

join the program and contribute 2 labour days a year if the conservation work were actually implemented? (Please circle

ONE response)

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

32. Would you (your household) contribute for the Boswellia papyrifera conservation program if it requires you 4 full labour

days per year as a guard for patrolling the site given that 1 labour day = 8 hours? (Please circle ONE response)

0. No If the response is ‘No’, skip to Question 34

1. Yes

33. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are of

that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would join

the program and contribute 4 labour days a year if the conservation work were actually implemented? (Please circle ONE

response)

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

34. Would you (your household) contribute for the Boswellia papyrifera conservation program if it requires you 1 full labour

days per year as a guard for patrolling the site given that 1 labour day = 8 hours? (Please circle ONE response)

0. No If the response is ‘No’, skip to Question 36

1. Yes

157

35. So you think you would join the Boswellia papyrifera forest conservation program. We would like to know how sure you are of

that. On a scale from ‘1’ to ‘10’, where ‘1’ is ‘Very uncertain’ and ‘10’ is Very Certain’, how certain are you that you would join

the program and contribute of 1 labour day a year if the conservation work were actually implemented? (Please circle ONE

response)

Very

Uncertain

Very

Certain

1 2 3 4 5 6 7 8 9 10

36. So you think that you would not join the Boswellia papyrifera forest conservation program if it requires you 1 full Labour Day

as a guard per year. Which of the following best describes your ‘No’ response to question 34? (Please circle ONE

response)

1. I have no interest in the Boswellia papyrifera forest conservation program.

2. I have some interest in the Boswellia papyrifera forest conservation program, but I would not contribute any

labour days to join the program.

3. I would join the Boswellia papyrifera forest conservation program if it requires me some labour contribution, but

less than 1 labour day as a guard per year

37. Which type of contribution mechanism would you prefer as the easiest and more feasible to you?

1. Contributing in the form of money

2. Contributing in the form of labour as guard

3. Both are equal for me to contribute

Questions related to socioeconomic and demographic characteristics of respondent households

III. Basic Household Characteristics

No.

1. List all the names of household (HH) members starting with the HH head. The HH member includes any one that lives in the HH since the last three months and eats together and generally shares income and expenses of the HH.

2. What is the Relation of the HH member to HH head? See codes below the table

3. What is the sex of the HH member? Male =1 Female =0

4. What is the age of the HH member in Years?

5. What is the level of education of this HH member?

Illiterate = 0

Can read & Write but no formal schooling =1

Number of Years of formal schooling completed

Is the HH member attending school? Yes =1 No=0

1

2

3

Codes for Question 2: 0= Self (Head) 1=Wife/Husband 2=Son/Daughter 3=Father/Mother 4=Father/Mother In-Law

5=Grand Father/ Mother 6=Son/Daughter In-Law 7=Grand Son/ Daughter 8=Nephew/Niece 9=Aunt/Uncle

10= Brother/Sister 11= Other relatives 12= Non relative

6. Since when have you been living in this tabia? Since __________________________

7. Have you come to this tabia as a settler during the resettlement program implemented by the government?

0. No, I was born here or living here before the resettlement

1. No, I came here after the resettlement

2. Yes

158

8. How long does it take to make a one-way travel on foot from your house to the nearest all weather road to get transport

service?__________ hours

9. How long does it take to make a one-way travel on foot from your house to the nearest primary school? __________ hours.

10. How long does it take to make a one-way travel on foot from your house to the nearest secondary school? __________

hours.

11. How long does it take to make a one way travel on foot from your house to the nearest Vocational School (TVET)?

__________ hours.

VI. Household Asset (Wealth)

4.1. Livestock /Animal Possession and income from Livestock sales.

Liv

esto

ck

Code

Livestock

Type

Livestock Wealth Income from Sales of Livestock (Part

of section V)

1.

How

many of

these

livestock/

animals

do you

own at

present?

Number

2.

If sold

today,

what

would be

the value

of these

livestock/

animals?

Total Birr

3.

How many of

these

livestock/anim

als did you

own 7 years

ago (in the

year 1995

E.C)?

Number

4.

If sold 7 years

ago (in 1995

E.C), what had

been the value

of these

livestock/anim

als?

Total Birr

5.

How many of these

livestock/animals were

sold in the last 12

months?

Number

6.

What was the

total value in

Birr received

from the

sales made in

the last 12

months?

Total Birr

1 Beef Cattle

2 Dairy Cattle

(Cows)

3 Oxen

4 Bull

5 Heifer

6 Calf

7 Horse

8 Mule

9 Donkey

10 Camel

11 Goats

12 Sheep

13 Chicken

14 Traditional

Bee Hives

15 Modern Bee

Hives

159

4.2. Income from Livestock Products (Part of section V but integrated here for convenience of respondents)

C

O

D

E

Livestock

Product

Type

Unit

1= litre

2=

kilogram

3= Unit

4= Day

5= other

7.

How much

of these

livestock/ani

mals

products

were

produced in

the last 12

months?

8.

What

would be

the total

value in

Birr if sold

in the

same

period?

Total Birr

9.

How much

of these

livestock/a

nimals

products

were sold

in the last

12

months?

10.

What was

the total

value in Birr

received

from the

sales made

in the last 12

months?

Total Birr

11.

How much

of these

livestock/ani

mal products

are in stock?

12.

What will be

the total value

of the amount

in stock if sold

at present?

Total Birr

1 Honey

2 Milk

3 Butter

4 Skins

5 Hides

6 Eggs

7 Renting

out

donkey

8 Renting

out horse

9 Renting

out camel

10 Renting

out Oxen

11 Renting

out Mule

12

13

14

15

160

4.3. Agricultural household durables

Please tell us about the agricultural assets owned by all household members.

Asset

code

Agricultural asset type 13.

Do you

have the

agricultural

asset?

1 = Yes

0 = No

14.

If yes for Q13,

How many of

these assets do

the HH owns at

present?

If partial

ownership, put

fraction owned.

Number

15.

If sold today,

what would be

the value of

the owned

part of the

asset?

Total Birr

16.

How many of

these assets did

the HH own 7

Years ago (in

1995 E.C)? If

partial

ownership, put

fraction owned.

Number

17.

If sold 7 Years

ago (in 1995

E.C), what had

been the value

of the owned

part of the

asset?

Total Birr

1 Animal pulled Ploughing

set (Maresha)

2 Axe (mesar, mestrebia,

gejemo)

3 Hammer(Mertelo,

Melakino)

4 Sickle (meatsid)

5 Spade

Saw (Megaz)

6 Shovel

7 Sprinkler

8 Chisel

9 Motor pump

10 Treadle pump

11 Ox Cart

12 Cart pulled by donkey

13 Cart pulled by person

14 Gezemo

14 Careta (for holding

loads on donkey)

15 Other __________

161

4.4. Non-Agricultural Household durables/Assets

Can you please tell us about the non-agricultural assets owned by all household members?

Non-

Agri

asset

code

Non-Agricultural asset type 18.

How many of these

assets do the HH

owns at present?

If partial ownership,

put fraction owned.

Number

19.

If sold today,

what would be

the value of the

owned part of

the asset?

Total Birr

20.

How many of

these assets

did the HH

own 7 Years

ago (in 1995

E.C)? If partial

ownership, put

fraction

owned.

Number

21.

If sold 7 Years

ago (in 1995

E.C), what had

been the value

of the owned

part of the

asset?

Total Birr

1 Radio

2 Tape Recorder

3 TV

4 Kerosene stove (Gas midija)

5 Kerosene lamp (Masho)

6 Kerosene lamp (Fanos)

7 Kerosene lamp (Lamba)

8 Bucket (Baldi)

9 Jerikan

10 Utensils and other kitchen

equipments (plates, iron pot,

kubaya, wancha)

11 Bed

12 Mattress

13 Stool

14 Chair

15 Table

16 Sewing machine

17 Bicycle

18 Blacksmith equipments

19 Weaving equipments

20 Building equipments (Mason

tools)

21 Jewelleries

22 Wrist watch

23 Weight balance (Mizan)

24 Flour Mill (Weficho)

25 Barrel (Bermel)

26 Plastic pot (Tela metimekia

plastic bermel)

27 Lamp (Ye eji Batery)

28 Other, Specify

162

4.5. Land Holding

Please tell me the different plots of agricultural land that are owned and operated, owned and rented out, owned and fallow land,

and rented in and operated by the household in the last 12 months and 7 years ago (in 1995 E.C).

Number of

plots in the

last 12

months

Total hectares

of all the plots

in the last 12

months 1=

Irrig

ate

d

0=

Rain

fed

Number of plots

7 years ago (in

1995 E.C)

Total hectares of

all plots 7 years

ago (in 1995 E.C)

1=

Irrig

ate

d

0=

Rain

fed

22. How many plots and

hectares of land do you

own? Owned land

means land in which

HH has long term

usufruct-rights

A1

A2

B1

B2

23. How many of these

plots and how many

hectares are/were

rented out to other

farmers?

A3

_

A4

_

B3

_

B4

_

24. How many of these

plots and how many

hectares are/were

fallowed?

A5 - A6 - B5 - B6 -

25. How many additional

plots and hectares of

land did you rent in?

A7 + A8 + B7 + B8 +

26. Total plots and

hectares operated by

the HH

A9 = A10 = B9 = B10 =

Check

that

A9=

A1-A3-

A5+ A7

Check

that

A10= A2-

A4-A6+A8

Check that

B9= B1-B3-

B5+B7

Check that

B8=B2-B4-

B6+B8

27. If the household has a fallowed land, after how many years of continuous cultivation did the land fallowed. _________ years

28. For how many years will the land be fallowed until it is cultivated again? __________________ years

29. What are the three most important reasons for fallowing the land?

1.__________________________________________________________________________________________

2.__________________________________________________________________________________________

3.__________________________________________________________________________________________

163

V. Household Income

5.1. Agricultural Income/ output

5.1.1. Crop production

Please tell me about the plots of agricultural land, wood lots, or pasture land owned and operated by the household in the last 12

months.

P

L

O

T

C

O

D

E

1.

Ple

ase te

ll m

e about

each plo

t of

land

belo

ngin

g

to

the

HH

?

Ple

ase

describ

e o

r giv

e m

e t

he n

am

e o

f each

plo

t.

2.

What is

the a

rea o

f th

e p

lot?

Unit C

ode: H

ecta

re =

1 T

sim

di =

2

3.

What kin

d o

f la

nd is t

his

?

Code:

Irr

igate

d =

1 R

ain

fed =

0

4.

What crop have you grown on

this plot during the last 12

months?

See Crop Code below the

table

If two crops are grown on the

same plot put the crop with

larger proportion of the plot as

Main Crop and smaller

proportion as second crop.

5.

How much kilograms of crop did you harvest from

this plot during the last 12 months? What would

be the total value of the harvest from the plot in

Birr if it would be sold during the same period?

Name of the

plot

Are

a

Unit C

ode

Ma

in C

rop

2ndC

rop

3rd C

rop

4th C

rop

Main Crop 2nd

Crop 3rd Crop 4

th

Crop K

g

Birr

Kg

Birr

Kg

Birr

Kg

Birr

1

2

3

4

5

6

7

Crop Code:

1=White teff

2= Red & Mixed teff

3= Barley(gebis)

4= Wheat

5= Maize

6= Sorghum

7= Zengada

8= Oats (Aja)

9= Horse beans(Bakela)

10= Linseed(Telba)

11= Sesame(selit)

12= Lentils (misir)

13= Cheack peas (shimbira)

14= Cow peas (Ater)

15= Dekoko

16= Guaya (Vetch)

17= Sinar

18= Field peas

19= Haricot beans (Boloke)

20= Fenugreek (Abish)

21= Rice

22= Nueg

23= Pop corn /Embaba

24= Dagussa

25= Onion (key shinkurt)

26= Garlic (Nech shinkurt)

27= Cabbage

28= Spinach (qosta)

29= Selata (Lettuce)

30= Tikil gomen

31= Carrot

32= Orange

33= Lemon

34= Avocado

35= Ginger

36= Banana

37=Karia (Green paper)

38= Berbere (Red paper)

39= Ird

40= Cactu(Beles)

41= Tomato (Comidoro)

42= Potato

43= Gesho

44= Sugarcane

45= Mango

46= Papaya

164

6. Have you rented in land in the last 12 months from other farmers and operated on the rented in land? (Please circle ONE

response)

0. No Skip to Question 15 on the next page

1. Yes

If yes, please tell me about each plot of land you rented in and operated in the last 12 months.

P

L

O

T

C

O

D

E

7.

Ple

ase t

ell

me

about

each p

lot

of

land y

ou r

ente

d i

n a

nd o

pe

rate

d i

n t

he l

ast

12

mo

nth

s?

8.

What

is the a

rea o

f th

e p

lot?

Un

it C

od

e:

Hecta

re =

1 T

sim

di =

2

9.

What kin

d o

f la

nd is t

his

? C

od

e:

Irrig

ate

d =

1 R

ain

fed =

0

10.

What

type o

f contr

act re

gula

tes t

he land y

ou r

ent in

?

Co

de

Fix

ed r

ent=

1 S

hare

cro

p =

2 O

ther

= 3

11.

If fixed r

ent, w

hat

is the a

mo

unt

of birr

paid

for

the p

lot per

year?

12.

If S

hare

Cro

p,

what %

of th

e tota

l harv

est

pa

id t

o t

he

lan

d o

wn

er

per

harv

est?

13.

What crop

have you

grown on this

plot during the

last 12

months?

See Crop

Code page

163

If two crops

are grown on

the same plot

put the crop

with larger

proportion of

the plot as

Main Crop

and smaller

proportion as

second crop.

14.

How much kilograms of crop did you

harvest as your share from this plot

during the last 12 months? What would

be the total value of your share from the

plot if it would be sold during the same

period? Please record only the share

to this household.

Name

of the

plot

Are

a

Unit C

ode

Ma

in c

rop

2nd C

rop

3rd C

rop

4th

Cro

p

Main

Crop

2nd

Crop

3rd

Crop

4th

Crop

Kg

Valu

e B

irr

Kg

Valu

e B

irr

Kg

Valu

e B

irr

Kg

Valu

e B

irr

1

2

3

4

5

6

165

15. Have you rented out your land to other farmers in the last 12 months?

0. No Skip to section 5.1.2 on the next page

1. Yes

If yes, please tell me about each plot of land you rented out to other farmers in the last 12 months.

P

L

O

T

C

O

D

E

16.

Ple

ase t

ell

me

about

each p

lot

of

land y

ou

rente

d out

to oth

er

farm

ers

in

the la

st

12

mo

nth

s?

17.

What

is the a

rea o

f th

e p

lot re

nte

d o

ut?

Unit C

ode: H

ecta

re =

1 T

sim

di =

2

18.

What kin

d o

f la

nd is t

his

? C

od

e:

Irrig

ate

d =

1 R

ain

fed =

0

What

type o

f contr

act re

gula

tes t

he la

nd y

ou r

ent

out?

Fix

ed r

ent=

1 S

hare

cro

p =

2 O

ther

= 3

20.

If

fixed

ren

t, w

hat

is t

he a

mo

unt

of

birr

you r

eceiv

e f

or

the p

lot

per

year?

21.

If

Sh

are

Cro

p,

what

% o

f th

e t

ota

l harv

est

do y

ou r

eceiv

e a

s o

wner

per

harv

est?

22.

If share crop, What type of

crop have you received as a

share from this plot during

the last 12 months?

See Crop Code on page

163

If two crops are grown on

the same plot put the crop

with larger proportion of the

plot as Main Crop and

smaller proportion as

second crop.

23.

If share crop, How much kilograms

of the crop did you receive as your

share from this plot during the last

12 months? What would be the total

value of your share from the plot if

sold at present? Please record only

the share to this household.

Nam

e o

f th

e p

lot

Are

a

Unit C

ode

Ma

in C

rop

2nd C

rop

3rd C

rop

4th

Cro

p

Main

Crop

2nd

Crop

3rd

Crop

4th

Crop

Kg

Valu

e B

irr

Kg

Valu

e B

irr

Kg

Valu

e B

irr

Kg

Valu

e B

irr

1

2

3

4

5

6

166

5.1.2. Agricultural input

I would like to ask you about the agricultural inputs that the household purchased and used for the last 12 months’ agricultural

production?

C

od

e

Input Type 24.

How much of the agricultural

input did you use (both

purchased and from home) in

the last 12 months?

25.

How much birr did you spend in

total to purchase the input in the

last 12 months?

Unit Quantity Birr

1 Commercial Fertilizer DAP Kg

Commercial fertilizer UREA Kg

2 Improved Seed for Wheat Kg

Improved seed for Barley Kg

Improved seed for Sorghum Kg

Improved seed for Teff Kg

Improved seed for Maize Kg

Improved seed for sesame Kg

Other crops : specify____________

3 Local Seed for Wheat Kg

Local seed for Barley Kg

Local seed for Sorghum Kg

Local seed for Teff Kg

Local seed for Maize Kg

Local seed for sesame Kg

Other crops: Specify_____________

4 Herbicides Litre

5 Pesticides Litre

6 Rented in Oxen Number

7 Hired labour Number

8 Veterinary service #of animals treated

5.1.2. Sales of crop

I would like to ask you about the type and amount of crop consumed, currently in stock, and sold by any of the household member

in the last 12 months from the last 12 months production.

Crop

C

O

D

E

See

from

page

163

Unit

26.

How much of

the total crop

harvest of the

last 12

months has

been

consumed so

far?

27.

How much is

the Value of

the

consumed

amount at

an average

price in Birr?

28.

How much

of the crop

harvested in

the last 12

months is

currently in

Stock?

29.

How much

will be the

Value of the

crop in stock

if sold at

current

market price

in Birr?

30.

Did you sell any

of the crop types

from the harvest

of the last 12

months during the

same period?

Yes = 1

No =0

31.

If yes for Q28, How

much did you sell?

How much revenue did

you get from the sales

made in the last 12

months?

Amount

Revenue

Birr

Kg

Kg

Kg

167

5.2. Forest Environmental Income

I would like to ask you all the products your household obtained freely from your environment (common pool open forest areas or

woodlands or specific tree) in your village, around your homestead, near/around/in your farm, or in your garden. For each forest

environmental product your household harvested during the last 12 months, please answer the following questions.

Pro

duct C

ode

Unit Code:

1= Shekim (head-load) per man;

2= Shekim (back-load ) per women ;

3=Chinet (load per donkey);

4= Chinet (load pe camel);

5= Esir (bundles);

6= Piece(number);

7= Kilograms;

8=Bags;

9=Yard;

10= meter;

11= kareta,

12. other, specify ___________

Forest Environmental Product

1.

Have you

ever used

the forest

environment

al product?

2.

How much did

you harvest in

a week and if

the product

were to be

obtained

through

purchase, how

much would

you pay as a

value?

3.

How much did

you harvest in

a month and if

the product

were to be

obtained

through

purchase, how

much would

you pay as a

value?

4.

How much did

you harvest in a

year and if the

product were to

be obtained

through

purchase, how

much would you

pay as a value?

1=

Ye

s

;

0=

No

Unit

Code

Am

ount

Valu

e I

n B

irr

Am

ount

Valu

e I

n B

irr

Am

ount

Valu

e I

n B

irr

1 Fuel wood for home consumption

2 Fuel wood for sale

3 Timber

4 Construction woods for house (walls,

roofs, poles)

5 Construction wood for shade near house

(walls, roofs, poles)

6 Construction wood for shade in farms

(walls, roofs, poles)

7 Construction wood for livestock kraal

8 Fencing materials (for both farm plot and

homestead fencing)

9 Farm implements (farm tool handles,

mofer, kenber, hoes)

10 Crop storage ( gotera)

11 Household furniture (stool, plates, spoons,

mortars, drums, decorative)

12 Wild food item (Fruits, vegetables, honey,

roots, mushrooms, bush meat)

13 Medicinal Plants (Tradition medicine)

14 Cactus (Beles)

15 Tooth cleaning brush (twigs)

16 Wild washing soda (like endod)

168

Continuation of the previous page

Pro

duct C

ode

Unit Code:

1= Shekim (head-load) per man;

2= Shekim (back-load ) per women ;

3=Chinet (load per donkey);

4= Chinet (load per camel);

5= Esir (bundles);

6= Piece(number);

7= Kilograms;

8=Bags;

9=Yard;

10= meter;

11= kareta,

12. other, specify ___________

Forest Environmental Product

1.

Have you

ever used

the forest

environment

al product?

2.

How much did

you harvest in

a week and if

the product

were to be

obtained

through

purchase, how

much would

you pay as a

value?

3.

How much did

you harvest in

a month and if

the product

were to be

obtained

through

purchase, how

much would

you pay as a

value?

4.

How much did

you harvest in a

year and if the

product were to

be obtained

through

purchase, how

much would you

pay as a value?

1=

Ye

s

;

0=

No

Unit

Code

Am

ount

Valu

e I

n B

irr

Am

ount

Valu

e I

n B

irr

Am

ount

Valu

e I

n B

irr

17 Leafs and grass for livestock fodder

18 Thatching grass

19 Fiber/rope

20 Sweeping material (mekoster)

21 Input for basket and mesob making

22 Woven hat (barneta)

23 Woven mats (sele’n)

24 Frankincense (Etan)

25 Other gums and resins

26 Aye

27 Mekeo

28 Geba

29 Habhab (Dima)

30 Akat

31 Muleo

32 Other, specify

169

5.3. Off-farm income

Off-

farm

activity

Code

Off- farm income source type 1.

Number of

household

members

engaged in

the off-farm

activity

2.

What was the total net income in Birr earned from the

off-farm activity made in the last 12 months by all

participating members?

Total income

in the 1st

Season

(January-April)

in Birr

Total income in

the 2nd

Season

(May-August) in

Birr

Total income in

the 3rd Season

(Sept-December)

in Birr

1 Frankincense tapping self employed

2 Frankincense tapping as daily labourer

for firms

3 Frankincense tapping as member of

cooperatives

4 Income as a squadra leader of

frankincense tappers

5 Masonry

6 Carpentry

7 Hair dressing (qonanit)

8 Handicraft, including pottery,

Blacksmithing

9 Trade in livestock

10 Selling local drinks and food (Tela,

Arequi, Teji, Kolo, Injera, Shai )(Net

income = Revenue - Cost)

11 Other Petty trading (Net income =

Revenue –Cost)

12 Food for work

13 Cash for work

14 Safety Net program

15 Other transfers and remittances

16 Other works as daily labourer

17 Flour mills (Wefcho)

18 Other, specify

Curriculum Vitae

Mesfin Tilahun Gelaye was born on March 20, 1976 in Arsi zone, Oromia Region of Ethiopia. In 1996,

he joined Mekelle University to study the undergraduate degree program in Economics. He graduated the

four-year program with honours and received his Bachelor of Arts degree in July 2000. He joined the

department of Economics of the same university as full-time teaching faculty in July 2000. He served at

first for two years at the rank of graduate assistant and went to Germany for his second degree. He

received his M.Sc. degree in Forest Ecology and Forest Sciences with specialization in Tropical and

International Forestry in September 2004 from George-August University of Goettingen. After his M.Sc.

study, he returned back to his teaching post at the department and worked for another four years at

different levels. For a total of six years, he taught a number of courses in the undergraduate degree

program in Economics, which include Natural Resource and Environmental Economics, Statistics for

Economists, Calculus for Economists, Linear Algebra for Economists, International Economics,

Development Planning and Project Analysis, Forest Economics and Product Marketing, and Population

Economics. Besides teaching, he was actively involved in research, consultancy and community services

of the university. From November 2005 to September 2008, he also served at higher university

administration positions. He was director of the office of the Micro Finance Program for about two years

and later served as director of the Institute of Continuing and Distance Education of Mekelle University.

In October 2008, he received admission to the KU Leuven as a doctoral student with a scholarship from

the KU Leuven Doctoral Scholarship Program in the framework of the Inter-Faculty Council for

Development Cooperation (IRO). He joined the Bioeconomics division and the division of Forest, Nature

and Landscape Research in the department of Earth and Environmental Science to do his PhD research

on the topic “Restoration and sustainable management of frankincense forests in Ethiopia: a bio-

economic analysis.” Over the course of his study at KU Leuven, he received a research grant award

from the International Foundation for Science (IFS) and funding and logistic support from The VLIR-MU-

IUC research project. During his stay at Mekelle University for doing his PhD field research, he has

supervised one M.Sc. student of his home institution as main supervisor.

Mesfin has published five research papers, also actively participated, and presented scientific papers in a

number of international workshops and conferences, among which are:

The 2nd

International DAAD Workshop on Forests in Climate Change Research and Policy and

the Forest Day 5 of the COP17 Durban Conference in South Africa,

The 4th Ecosystem Services Partnership Conference 2011 in Wageningen, and

The XIIIth Congress of the European Association of Agricultural Economists (EAAE 2011) in

Zurich.

He received the best oral presentation award from the European Alliance on Agricultural Knowledge for

Development (AGRINATURA) for his presentation at the Tropentag-2011 in Bonn and the best poster

award at the Forest2011 Conference on Conservation and management of forests for sustainable

development: where science meets policy held in Leuven, Belgium, 23-24 November 2011.

Publications in peer reviewed journals

Tilahun, M., Muys, B., Mathijs, E., Kleinn, C., Olschewski, R. & Gebrehiwot, K., 2011. Frankincense yield

assessment and modeling in closed and grazed Boswellia papyrifera woodlands of Tigray, Northern

Ethiopia. Journal of Arid Environments 75, 695-702.

Mekuria, W., Veldkamp, E., Tilahun, M. & Olschewski, R., 2011. Economic Valuation of Land

Restoration: The Case of Exclosures Established on Communal Grazing Lands in Tigray, Ethioppia. Land

Degrad. Develop. 22, 334-344.

Tilahun, M., Olschewski, R., Kleinn, C., & Gebrehiwot, K., 2007. Economic analysis of closing degraded

Bosewellia papyrifera dry forest from human intervention - A study from Tigray, Northern Ethiopia. Forest

Policy and Economics 9, 996–1005.

Papers published in conference proceedings

Tilahun, M., Muys, B., Mathijs, E., & Verbist, B., 2012. Methods for valuing dry land forests in the context

of REDD+ and the role of multidisciplinary Sciences. In: Fehermann, L, Kleinn, C., (Eds.), 2012. Forests

in Climate Change Research and Policy: The Role of Forest Management and Conservation in a

Complex International Setting. Proceedings of the 2nd

International DAAD Workshop. Pietermaritzburg

and Durban, South Africa, 1-7 December 2011(pp. 193-200). Goettingen: Curvillier Verlag.

Mohammed, A., Tilahun, M., Muys, B., Mathijs, E., & Gidey, A., 2012. The effect of income from non-

timber forest products on rural poverty and income inequality: a case study from the dry woodland area of

Tigray. In: Fehermann, L, Kleinn, C., (Eds.), 2012. Forests in Climate Change Research and Policy: The

Role of Forest Management and Conservation in a Complex International Setting. Proceedings of the 2nd

International DAAD Workshop. Pietermaritzburg and Durban, South Africa, 1-7 December 2011(pp. 185-

192). Goettingen: Curvillier Verlag.

Paper submitted

Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H. & Mathijs,

E., 2012. Rural households’ demand for frankincense forest conservation in Tigray, Ethiopia: a contingent

valuation analysis. (Submitted to Journal of Land Degradation and Development).

Papers (posters) presented at International Scientific Conferences

Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H. & Mathijs,

E., 2011. Contingent valuation analysis of rural households' willingness to pay for frankincense forest

conservation. Oral presentation & full paper contribution. EAAE 2011 Congress: Change and Uncertainty,

Challenges for Agriculture, Food and Natural Resources. ETH Zurich-Switzerland, August 30 to

September 2, 2011. EconPapers No. 116085. http://purl.umn.edu/116085

Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H. & Mathijs,

E.,2011. Valuing dry land forest ecosystem services: a case of rural households’ willingness to pay and

contribute labour for frankincense forest conservation in Ethiopia. Oral presentation. 4th ESP Conference

2011: Ecosystem Services: Integrating Science with Practice. Wageningen, The Netherlands, October 4-

7, 2011.

Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H., &

Mathijs, E., 2011. Contingent Valuation Analysis of Rural Households’ Willingness to Pay and Contribute

Labour for Frankincense Forest conservation. Oral presentation. Tropentag 2011: Development on the

margin: Livelihood strategies, Bonn, Germany, October 5-7, 2011.

Tilahun M., Gidey, A., Gebregziabeher, K., Woldu, T., Daan, O., Deckers, J., Gebrehiwot, K., Bauer, H.,

& Mathijs, E., 2011. Rural households’ demand for irrigation water in Tigray: a contingent valuation

analysis. Poster. International Congress Water 2011: Integrated water resources management in tropical

and subtropical drylands: Mekelle, Ethiopia, 19-26 September 2011.

Tilahun, M., Vranken, L., Muys, B., Deckers, J., Gebreegziabher, K., Gebrehiwot, K., Bauer, H., &

Mathijs, E., 2011. Restoration and Sustainable Management of Frankincense Forests: A bio-economic

analysis. Poster. Forest2011: Conservation and management of forests for sustainable development:

where science meets policy. International Conference. Leuven, Belgium, 23-24 November 2011.

Tilahun, M., Mathijs, E., Muys, B., & Verbist, B., 2011. Methods for valuing dry land forests in the context

of REDD+ and the role of multidisciplinary sciences. Oral presentation. The 2nd

International DAAD

Workshop: Forests in climate change research and policy: the role of forest management and

conservation in a complex international setting. Organized by George-August University of Goettingen

and Stellenbosch University. Pietermaritzbeurg and Durban, South Africa, December 1-7, 2011.