Sustainable Forest Management Support to the Forest ... · Sustainable Forest Management Support to the Forest Stewardship Council Service delivery report Mondi Syktyvkar trials 26

Sustainable Forest Management

Support to the Forest Stewardship Council

Service delivery report

Mondi Syktyvkar trials

26 May 2011

Eyre Consulting – UK

Denis Alder Consulting – UK

sarmap – Switzerland

Metria – Sweden

Forest Stewardship Council – Germany


Support to Forest Stewardship Council

Mondi service delivery report

Page 2

This document is property of the consortium and cannot be distributed/duplicated without authorisation

Reference Documents

Ref Document

RD-3 Aon, Eyre C., sarmap, Ambiental, Earth Observation in Support of Corporate Social Responsi-

bility – Forestry Sector, Action Plan for increasing the use of Earth Observation across the

Sustainable Forest Sector, Final Report, ESA/ESRIN, 2008.

RD-4 Eyre C. and sarmap, Timber Volume Estimation in Forestry Plantations: Forest Modelling

from Earth Observation data, DUE Innovator II, ESA/ESRIN, 2009.

SFM-D1.3 Eyre C., DA Consulting, sarmap, Metria, FSC, Sustainable Forest Management - Service Trials

Specifications - Earth Observation component, Deliverable 1.3, Report, ESA/ESRIN, 2008.




Metria – Sweden





Page 3


Acronyms

ALOS Advanced Land Observation Satellite

BA Basal Area

DBH Diameter at Breast Height

DEM Digital Elevation Model

DSM Digital Surface Model

EO Earth Observation

FBD Fine Beam Dual (HH/HV polarization)

FBS Fine Beam Single (HH or VV polarization)

HR High Resolution (10 to 30 meter spatial resolution)

InSAR Interferometric Synthetic Aperture Radar

NIR Near InfraRed

SAR Synthetic Aperture Radar

SPOT Système Probatoire d’Observation de la Terre

SRTM Shuttle Radar Topographic Mission

SWIR Short Wave InfraRed

TH Tree Height

TPH Tree Per Hectare

TV Timber Volume

UTM Universal Transfer Mercator

VHR Very High Resolution (better than 10 meter spatial resolution)

WGS World Geodetic Datum




Metria – Sweden





Page 4


CONTENTS

Executive Summary 5

Introduction 7

1. Stakeholders 8

1.1 FMT provider 8 1.2 Service provider 8 1.3 Trial partner 8 1.4 Stakeholder Group 9

2. MondiSYK concession and the service trial areas 11

2.1 MondiSYK concession and trial area 11 2.2 MondiSYK field data collection 12 2.3 Data and service cost indications 13

3. Service portfolio 15

3.1 – Product list 15 3.2 Product description, validation and evaluation 16 Product 0 - Digital Surface Model 18 Product 1 - Forest area 22 Product 5 - Clear-cut, logging, and re-growth 26 Product 10 - Road mapping 31 Product 12 - Settlements 34 Product 13 - Land cover in non forest areas 37 Product 14 - Biophysical parameters 40

4. Service Portfolio Evaluation 43

4.1 EO Technology Utility 43 4.2 Summary of Portfolio Utility 44 4.3 Potential for cost saving in generating in situ or conventional data 45 4.4 Overall Value for money 45 4.5 Technology Risk and Reward 46 Annex 1 – Product Evaluation Criteria – Issues to be considered 47

Annex 2 - ALOS PALSAR-1 failure 48

Annex 3 Forest Succession Table – Komi 53

Annex 4 – Additional validation for Forest Area (Product 1) and Land Cover (Product 13) following the confusion matrix procedure 55

Annex 5 - Clarification on the sensitivity of the L Band data to saturation when estimating the timber volume above certain values. 60




Metria – Sweden





Page 5


Executive Summary

The project with Mondi Syktyvkar (MondiSYK) has been valuable and useful in that it has allowed the project

partner to test and demonstrate the resilience and utility of Earth Observation (EO) in a mega-forest envi-

ronment as an area wide mapping process for forest area and forest area changes while using the same EO

data to assess the potential to infer estimations of timber volume, a basic requirement for forest manage-

ment.

A customized service, aiming at the provision of transparent, comprehensive, and cost effective information,

has been designed and implemented for the provision of target products based on high to very high resolu-

tion (15m to 3m) spaceborne Synthetic Aperture Radar (SAR) and Optical data. In essence, the service con-

sists of two main components:

– Generation of thematic products (maps and change maps);

– Inference of timber volume.

In a series of trials in Komi Forest, Russia, the following products have been demonstrated and tested over

a range of operational conditions, namely:

– Product 0 - Digital Surface Model (DSM)

– Product 1 - Forest area – at landscape level;

– Product 2 - Forest type – Identification of the range of forest dynamic ecosystems;

– Product 3 - Burnt Areas – The extent of burnt areas within the concessions; – Product 5 - Clear cut and re-growth – The distribution of and area of clear-cut within prescribed

time frames; – Product 9 - Road mapping – comparison of EO derived forest road system and GPS derived road

maps. – Product 11 - Settlements – To locate and map settlements within the forest; – Product 13 - Monitoring of protected forest areas – using shape files to identify typical examples

of bio systems, to assess the potential for remote sensing to monitor status of HCVF - intact forest; HVCF

- FSC standard definition; Riparian wetlands; Boggy wetlands; – Product 14 - Biophysical parameters – An indication or estimation of timber volume to support in

situ operational data used in forest management and planning;

Estimations of the costs for each of the products are presented for both Palsar and SPOT sensors. The

package of mapping products 1-5 (excluding logging and regrowth) would sum up to €3.20/km2 for Palsar

and €1.46/km2 for SPOT. If Palsar data is used, timber volume could be included for a total €3.48/km2. The

cost of monitoring products (including Product 5 with logging and regrowth) would sum up to €4.7/km2 for

Palsar and €4.17/km2.

The validation of the results show the DSM resulted in an average error of 3% and is better than anything

currently available. The average error on forest area mapping, using Mondi’s validation process, was 13.5%

while sarmap, using a confusion matrix validated the maps at 74%, and if rivers were excluded, overall ac-

curacy rose to 92% (see Annex 4). None of the sensors were able to discriminate species within the com-

plex boreal forest of Komi. The temporal accuracy (clear cuts per year) of Palsar derived maps and Mondi-

SYK data was 100%. The spatial accuracy (i.e. detecting area clear felled) using Palsar was 10.2% less

than the measurement using ground plot data. However, using CSM data, accuracy rose to 95%-99%.




Metria – Sweden





Page 6


Road mapping validation was carried out on 116km of road of which 109km was confirmed as road, and the

rest, rivers or power lines. MondiSYK validated that they could discriminate larger settlements but not

smaller ones using EO. On mapping land cover on non forest areas, the average error was estimated by

MondiSYK to be 6.5% and monitoring changes in protected areas is possible but identifying areas that

should be protected is not possible as it required species detection. Combining DSM with Palsar could help

in identifying areas that could be considered for protection, but it would require field data to support a deci-

sion to classify forest for protection. Finally, on the estimation of biophysical parameters of the forest using

Palsar, where the forest was segmented across three classes (chosen by MondiSYK), accuracy on low vege-

tation (0-50 m3/ha) was 20%, 50-150m3/ha was 62% and for areas with more than 150m3/ha, accuracy

was assessed at 87.5% (see Annex 5 for discussion on saturation effect). Effectively, complex forest struc-

tures with mixed species and the damp forest conditions resulted in limited changes in radar signal and sig-

nal saturation which restricted the value and utility of this product. At a landscape level, MondiSYK consid-

ered it to be of value.

The evaluation of each product by MondiSYK team is provided. The overall assessment of products by

MondiSYK confirms that clearcut mapping (Product 5) and road mapping (Product 10) could be used to rein-

force existing data; the DSM (Product 0) could be used to replace existing data sources but cannot in Russia

for legal reasons; subject to some improvements, forest mapping (Product 1), mapping non forest areas

(Product 13) and estimation of biophysical characteristics (Product 14) could replace existing sources of

data; and finally, EO derived settlement mapping was considered unsuitable to replace existing sources of

data.

Generally, MondiSYK assessment is that the overall portfolio of demonstration products would be potentially

good value for money. Peter Gardiner, Manager Mondi Natural Resources, comments “Currently the portfolio of products could be used as valuable tools to support/enhance existing products in use by Mondi, certifica-tion bodies and other stakeholders but with refinements and the systematic introduction of latest technolo-gies, the potential of EO products to provide supporting and stand-alone solutions for the monitoring and management of the boreal forest is substantial. Provided that the technology and procedures are intro-duced with the approval of relevant state authorities, the introduction of high resolution EO technology and open source systems would be timeous and economic interventions to improve the current poor state of Russian maps”




Metria – Sweden





Page 7


Introduction

This project, funded by the European Space Agency (ESA) under the Value Adding Element program, aims to develop, demonstrate, validate and evaluate a range of Earth Observation (EO) products to support the better FSC certification. In this report, results are provided of the selected portfolio of space borne EO prod-ucts on Mondi FSC certified concessions in Komi Republic in Russia.

The project goal is to demonstrate that the latest advances in this technology can deliver:

• improvements in the quality of mapping and monitoring in forest management and certification;

• make certification data and processes more accessible and transparent to stakeholders;

• add rigor and discipline to the certification process by standardizing the range and quality of geo-spatial information;

• reduce the costs of certification by improving the efficiency of certification bodies in the audit proc-ess.

The report includes an initial section which provides background information on the Trial Partner’s for-estry operations and general overview of the other stakeholders that have reviewed or provided feedback on the value of demonstrated service portfolio. This includes the Certification Body and where possible, key NGOs.

The Product Trials are summarised, along with EO data sources used in the mapping process at the be-ginning of the section on the Demonstration Trials. Each Product Trial is then detailed, along with the Vali-dation which has been completed independently by the Trial Partner.

The report concludes with an overall evaluation of the service portfolio, including a risk analysis and cots benefit evaluation.

A key objective of this project is to make the maps and geospatial data generated more accessible and uti-lisable by FSC stakeholders and FSC itself. To support this, the project includes the development of an Open Source GIS software, the Forest Management Toolkit, which enables visualisation of the maps and the identification, location and extraction of forest and non-forest areas and area changes over time. This toolkit, based on the Open Source GIS software Q-GIS and can be downloaded from the FMT website, which is at http://www.eofmt.com.

This report has been prepared and contents approved by Eyre Consulting, Mondi and Sappi.




Metria – Sweden





Page 8


1. Stakeholders

1.1 FMT provider

Denis Alder Consulting - Denis Alder is a consultant in forest biometrics and information systems for forest management with some 36 years working experience. After working in research in forest growth and yield modelling at Oxford University 1974-79, the greater part of this work has involved designing systems, de-veloping software and analysing data for forest inventories, growth and yield studies, and the monitoring of forest management. His work generally has focussed on improving the sustainability of forest management through better control systems for yield allocation and monitoring, and as such has often been part of the process of bringing forest management up to the level necessary for certification to Forestry Stewardship Council standards.

1.2 Service provider

sarmap - An internationally recognised developer of remote sensing software and products. Extensive ex-perience in airborne and spaceborne SAR & optical data processing and fusion. Services in the domain of land applications, in particular Digital Elevation Model, land motions, agriculture, forestry and hydrology.

1.3 Trial partner

Mondi Syktyvkar (MondiSYK) is a wholly owned subsidiary of Mondi Group an international paper and pack-aging Group, with production operations across 31 countries. The key operations are located in central Europe, Russia and South Africa, servicing markets across the globe.

Mondi is fully integrated across the paper and packaging process, from the growing of wood and the manu-facture of pulp and paper (including recovered paper), to the conversion of packaging papers into corru-gated packaging and industrial bags. The Group is principally involved in the manufacture of uncoated fine paper (UFP), packaging paper and converted packaging products as well as speciality products.

Group revenue in 2010 was €6.2bn and cash generated by operations was €778m.

In 2010, Mondi employed an average of 29,000 people and, in addition, provided employment for an esti-mated average of 21,000 contractors during the year. (Mondi Syktyvkar employs about 8500 people.) Mondi is a dual listed company (DLC), with listings on the Johannesburg and London stock exchanges.

The Group has consistently been recognised for its sustainability performance through its inclusion in the FTSE4Good UK, in the European and Global Index Series since 2008, and in the JSE’s Socially Responsible Investment (SRI) Index from 2007 to date.

The Russian operations are situated in the Komi Republic (area 416,000 km2 and population 951,000 of which 250,000 live in Syktyvkar). Mondi has recently invested €550 m in the expansion of the Syktyvkar mill to lift pulp production to 1m tons and in the process consume in excess of 4m m³ of wood.




Metria – Sweden





Page 9


Fig 1 – Mondi Syktyvkar Mill with extension works

MondiSYK’s leased area in the boreal forests is 2.1m ha with an annual allowable cut of 5.1 million m3. Cur-rently, all 2.1 million ha of the MondiSYK leased forest areas are FSC certified and Mondi has supported in excess of 1m ha of group certification involving the state forests. MondiSYK is also upgrading its reforesta-tion and silvicultural capacity, having opened a mechanised nursery in 2009.

The project participants at Mondi are:

– Peter Gardiner, Manager – Natural Resources based in UK – Denis Popov, Planning Manager – Mondi Syktyvkar – Vladimir Ermakov, Forest Data Manager – Mondi Syktyvkar

1.4 Stakeholder Group

NEPCon, a partner company of Rainforest Alliance (FSC Certification Body)

NEPCon is a Danish registered non profit organisation with a mission to encourage sustainable development of natural resources with 8 wholly owned daughter companies in Russia and Europe. NEPCon undertakes certification against the FSC standard in collaboration with their FSC-accredited partner, the Rainforest Alli-ance. By agreement, NEPCon is managing the Rainforest Alliance's SmartWood forestry program in Europe, Russia, Turkey and Israel.

Rainforest Alliance is an international non-profit conservation organization and the world's leading FSC certi-fier of forestlands. Since its formation 20 years ago, Rainforest Alliance has certified some 40m ha in 58 countries and more than 50% of all FSC certified forests are certified by SmartWood.

Key contacts are:

– Peter Feldberg, CEO – NEPCon Estonia – Nicolay Tochilov, Director – NEPCon Russia – Sergey Pautov, Certifier– NEPCon Russia

Silver Taiga Foundation and Komi Model Forest (The key NGO)

"Silver Taiga" (http://www.silvertaiga.ru/) was founded in 2002 as a Komi Regional Non-Profit Foundation by the World Wide Fund for Nature (WWF). The founding of the Silver Taiga Foundation was supported fi-nancially by the Swiss Development Corporation (http://www.deza.admin.ch/) of the Swiss Government. The Foundation is the implementing organisation for the Komi Model Forest Project and other projects ori-enting on sustainable regional development, based on natural resources.




Metria – Sweden





Page 10


The Komi Model Forest (KMF) is the boreal forest conservation and management model forest in Priluzje Forest Unit (Lesnichestvo) of Komi Republic. Funding has also been provided by Swiss Agency for Develop-ment and Cooperation (SDC) and the implementing organisations are WWF (until 2002) and Silver Taiga (ongoing). It includes a certified area 800,000 ha of which a significant proportion is logged by commercial companies. MondiSYK is one of the largest concessionaires with 60,000 ha.

Silver Taiga is the responsible agency for the Komi Model Forest Project area, which was established to de-velop forest management practices that reflect the special ecological and environmental situation in Komi namely:

• Industrial logging only started 70 years ago and secondary forests are still very young. Sustainable forest management during this period was very limited and there is great scope for improvement. Most regeneration is natural or almost natural with very little replanting to date.

• Different succession stages and forest dynamics are very different to what we might experience in other areas (e.g. EU or Poland)

• The forests are very large scale and forests species dynamics varied

• The remoteness of the Komi region and the very limited roads away from logging areas, creates unique issues

• Fire (natural and human induced) is critically important in the forest succession dynamics

Key contacts are:

• Przemyslaw Majewski –Director – Silver Taiga

• Yuri Pautov – Deputy Director – Silver Taiga

• Lena Popova – Ecologist

FSC National Initiative

Contact has been made with, the Director of FSC, Russian NI. If time permits, they will be asked for com-ments and evaluation of the products demonstrated

– Andrei Ptichnikov, Director – Russian FSC National Initiative




Metria – Sweden





Page 11


2. MondiSYK concession and the service trial areas

2.1 MondiSYK concession and trial area

Komi is a frontier area and very remote with a poor road system. There are only 13 km of roads per 1000 km2 of forest which is approximately 70 times less roads than Sweden. The industry is emerging from So-viet era when there was little or very limited forest resource management. The climate and therefore forest operating conditions are more like Canada than Europe, and while forest extraction operations are based on latest Scandinavian techniques, there is generally poor forest management with limited monitoring systems in place. The majority of felling is done during winter when the ground is frozen, impacts are minimised and access to forests over frozen soil and rivers is optimum. In MondiSYK’s case, some of the harvested timber is floated down the river to the mill.

The Komi Boreal forests are very complex and dynamic, with fire playing a vital part in the natural (non-anthropogenic) dynamics modifying their appearance, composition, structure and functional organization. Natural dynamics can be damaged by human activities. There are 5 recognised natural forest dynamic types and which prevails is determined by the site quality. These natural forest dynamic types1 are:

• Deciduous-to-Spruce fire dynamics

• Pine-to-Spruce fire dynamics

• Pine fire dynamics

• Spruce non-fire dynamics

• Dynamics on extremely poor soils

Fig 2 – Mondi Syktyvkar concession leased areas and trial area

1 For a description of the forest dynamics, see the Table in Annex 3, which has been developed from a re-port by Ministry of Natural Resources of the Russian Federation, Forestry Agency for the Komi Republic, Sil-ver Taiga Foundation, Priluzje State Forestry Unit entitled “Recommendations on logging operations with conservation of ecological values in intact (pristine) forests in the Komi Republic”. July 2005




Metria – Sweden





Page 12


Management of logging operations in forests areas has been carried out in accordance to the Rules of Final Felling Operations in the Plain Forests of the European Part of the Russian Federation which do not ade-quately reflect the type and stage of natural forest dynamics. The new “Recommendations on logging op-erations with conservation of ecological values in intact (pristine) forests in the Komi Republic”, developed in Komi Model Forest Project, imitate natural forest dynamics during logging operations and forestry manage-ment, thereby attempting to:

• Mimic the impact of natural disturbances (i.e. forest fires) on the forest ecosystems;

• Accelerate the succession of natural tree species changes during forest re development;

• Maintain existing forest ecological values in both space and time.

Historically, as forest resources were exploited new logging operations moved into generally more inaccessi-ble and remote areas which then required the construction of new roads. The depletion of forest resources near existing roads infrastructure has forced a stark choice - to continue opening up new areas with all the associated infrastructure and transport costs or to initiate better forest management of exploited areas. The more open minded and advanced companies are realising that routine clearfell of new, pristine areas rather than management and selective logging is not necessarily the best approach and this historical extensive process of “cut and run” is now being replaced by more intensive FSC standards of forest and resource management including clear cut logging.

The present logging recommendations reflect the standard of the Forest Stewardship Council:

• Principle 6. Environmental impact;

• Principle 9. Maintenance of high conservation value forests.

As a result, protection of pristine HCVF areas and re-establishment and improved management of logged areas are being taken seriously, with MondiSYK a prime mover in this direction within Komi.

Historically, there was a strong Russian State Forestry organisation, responsible for all forest data collection and management (State forest inventory enterprise). However over the last decade, this has slowly been deteriorating in terms of timing of inventory periods and adequacy of data during post communist era, as no inventories in Komi have been done by the State Forest Service for the last 15 years. Where it has been done, quality and reliability have declined due to luck of finance. There is no alternative data source, as no private providers have emerged. There are new state organizations dealing with natural resources, but they are not developing inventory data.

Mapping systems were updated with GPS and geographic references were added, but it was not topographi-cally or geographically that accurate. Available topographic maps are old and in 1:100,000 to 1:200,000 and MondiSYK logging planning has been based on State Forest data sources 1997 – 2007. Geodesic maps are also available for some areas and purposes but substantial errors are frequently found and as these er-rors are not systematic, the value of these maps is limited. A grid of numbered quartals (2km x 4 km), which are marked in field, provide a link between old and any new forest mapping data.

Higher resolution maps are available but not for commercial purposes. If an organisation is prepared to share all data as open source, then these more detailed maps are more available. It will therefore be im-portant that the project operates in an open source way, to ensure that we can access the best available maps.

2.2 MondiSYK field data collection

MondiSYK have introduced a data collection and inventory management process, designed to upgrade the quality of data currently available on which to base both forest management activities as well as Forest Cer-tification. The above schematic shows the data collection and management process, and the use of EO is a critical part of this development.




Metria – Sweden





Page 13


Fig 3 – Organisational structure of forestry management inventory systems Mondi Syktyvkar

2.3 Data and service cost indications

MondiSYK’s estimated costs2 of detailed inventory is about €4 per ha.

MondiSYK has been provided the following summary indicative service costs by sarmap, based on the fol-lowing assumptions:

• Costs are based on whole "scene" utilisation and no account is taken of limited areas within a scene being utilised by a client. This can only be estimated for specific cases.

• Where required, all sample plot data has to be provided, in correct format and structure, by client.

• Validation is client responsibility and cost

• All mapping products are for maps at a specific time. Assuming the date of acquisition and sensor is the same, different products can be produced from the same scene(s). In such circumstances, as the same data is utilised for different mapping products, data /processing costs are not additional for prod-uct combinations. Service costs will be additional for each different product.

• Change monitoring would result in two classified images at different points in time, plus a map of changes between the dates of acquiring.

ALOS PALSAR (costs €/km2)

• All products, including Timber Volume are produced from Palsar (15m resolution).

Resolution of higher than 15m would require a different sensor and be additional. This could not apply to TV as PALSAR data is obligatory.

2 These numbers are provided as there is little else to go on – but accuracy is uncertain.




Metria – Sweden





Page 14


Table 1 - Indicative costs for services using ALOS PALSAR

SPOT 4 AND SPOT 5 service cost indication cost per square km

Optical data can be used for the generation of thematic products. Costs ranges are provided above for SPOT-4 and -5 on the same basis as above.

Table 2 - Indicative costs for services using SPOT-4/-5. Data cost is according to South Africa market price.




Metria – Sweden





Page 15


3. Service portfolio

3.1 – Product list

An overview of products demonstrated with MondiSYK is listed in the table below. Refer to D1.3 for a de-tailed description of the products.

Product Trial status

0. Digital Elevation Model Completed

1. Forest area Completed

2. Forest type requested – technically not possible

3. Burnt areas, severity, re-growth no burnt areas available – not completed

4. Status of fire break

5. Clear cut, logging, re-growth Completed

6. Health status

7. Homogeneity including gap identification

8. Detection of errors in forest stand databases

9. Status of young (3-5 y) stands

10. Road mapping Completed

11. Water bodies

12. Settlements Completed

13. Land cover in non forest areas Completed

14. Biophysical parameters Completed

Table 3 - List of all demonstration products and MondiSYK selection

MondiSYK selection includes:

– Forest area – at landscape level. – Forest type – Identification of the range of forest dynamic ecosystems. – Burnt Areas – The extent of burnt areas within the concessions. – Clear cut and re-growth – The distribution of and area of clear-cut within prescribed time frames. – Road mapping – comparison of EO derived forest road system and MondiSYK GPS derived road

maps. – Settlements – To locate and map settlements within the forest. – Monitoring of non forest areas – using shape files to identify typical examples of bio systems, to as-

sess the potential for remote sensing to monitor status of HCVF - intact forest; HVCF - FSC stan-dard definition; Riparian wetlands; Boggy wetlands

– Biophysical parameters – To provide an indication or estimation of timber volume to support in situ operational data used in forest management and planning.

Note that:

– Products are delivered by FTP; – Visualisation and further analysis of the products are performed using the FMT.




Metria – Sweden





Page 16


3.2 Data

The following data has been used:

– ALOS PALSAR-1 (each date includes 3 images for a total of 30 images)

– 12 Aug 2007 FBD (HH/HV), 15 meter resolution, coverage 70 by 70 km

– 27 Sept 2007 FBD (HH/HV), 15 meter resolution, coverage 70 by 70 km

– 29 Jun 2008 FBD (HH/HV), 15 meter resolution, coverage 70 by 70 km

– 14 Aug 2008 FBD (HH/HV), 15 meter resolution, coverage 70 by 70 km

– 02 Oct 2009 FBD (HH/HV), 15 meter resolution, coverage 70 by 70 km

– 09 Feb 2007 FBS (HH), 8 meter resolution, coverage 70 by 70 km

– 28 Dec 2007 FBS (HH), 8 meter resolution, coverage 70 by 70 km

– 30 Dec 2008 FBS (HH), 8 meter resolution, coverage 70 by 70 km

– 14 Feb 2009 FBS (HH), 8 meter resolution, coverage 70 by 70 km

– 04 Apr 2010 FBS (HH), 8 meter resolution, coverage 70 by 70 km

– Cosmo-SkyMed-1/-2/-3/-4

– 24 Sept 2010 StripMap HH, 3 meter resolution, coverage 40 by 40 km

– 25 Sep 2010 StripMap HH, 3 meter resolution, coverage 40 by 40 km

– ERS-1/-2

– 28 Sept 1995, 25 meter resolution, coverage 100 by 100 km

– 29 Sept 1995, 25 meter resolution, coverage 100 by 100 km




Metria – Sweden





Page 17


The colour composite below shows the extent of the ALOS PALSAR image (approximately 210 by 70 km) and of the Cosmo-SkyMed-1/-2 scene (approximately 40 by 40 km)

Fig 4 - The colour composite of Palsar, CSM and Spot data used in product generation




Metria – Sweden





Page 18


3.3 Detailed review of Products delivered, Product validation and evaluation

Product 0 - Digital Surface Model

Description Topographic height referred to ellipsoid or geoid.

Location Komi

Data sources - ERS-1/-2 data (25m) acquired on 28th and 29th Sept 1995

Product - Colour composite (8bit) in GeoTif format of ERS-1/-2 DEM data

- Digital Elevation Model (float data) referred to ellipsoid or geoid in GeoTif format

- Cartographic and geodetic system: UTM-39 North, WGS-84

Fig 5 - GTOPO-30 - 1km grid spacing (left) and ERS-1/-2 DEM - 25m grid spacing (right). Dark values do correspond to low altitudes, bright values to high altitudes.




Metria – Sweden





Page 19


Validation

The validation was based on 58 samples taken from absolute height topography maps and compared to a reference Digital Elevation Model. The comparison resulted in an average error of 3%. Detail level is higher than 1:200 000 contours lines, is the most common scale available in Russia and the currently available scale available to MondiSYK. However, it results in legal restrictions: due to Russian Legislation DEM has to be divided by parts 250 square km with blank strips between squares from security point of view. Therefore DEM cannot be used for detail road construction and forest cutting planning.

# DSM DEM Absolute dev Relative dev

1 150,4 149 -1,4 0,990691

2 151,8 152 0,2 1,001317 3 143,6 144 0,4 1,002785

4 175,9 177 1,1 1,006253 5 127,2 128 0,8 1,006289

6 150,4 152 1,6 1,010638 7 164,1 166 1,9 1,011578

8 170,9 173 2,1 1,012287 9 166,7 169 2,3 1,013797

10 154,6 157 2,4 1,015523 11 178,2 181 2,8 1,015712

12 176,2 179 2,8 1,015891

13 125,9 128 2,1 1,016679 14 167,2 170 2,8 1,016746 15 150,4 153 2,6 1,017287

16 146,4 149 2,6 1,017759

17 172,7 176 3,3 1,019108

18 127,5 130 2,5 1,019607

19 125,4 128 2,6 1,020733 20 120,4 123 2,6 1,021594

21 172,2 176 3,8 1,022067

22 134 137 3 1,022388

23 169,8 174 4,2 1,024734 24 153,2 157 3,8 1,024804

25 154,1 158 3,9 1,025308 26 177,3 182 4,7 1,026508 27 173,4 178 4,6 1,026528 28 171,4 176 4,6 1,026837

29 126,6 130 3,4 1,026856 30 172,2 177 4,8 1,027874

31 145,8 150 4,2 1,028806 32 166,8 172 5,2 1,031175

33 170,5 176 5,5 1,032258

34 146,8 152 5,2 1,035422

35 173,8 180 6,2 1,035673 36 171,8 178 6,2 1,036088

37 140,9 146 5,1 1,036195 38 174,6 181 6,4 1,036655 39 176,5 183 6,5 1,036827

40 181,3 188 6,7 1,036955

41 135,9 141 5,1 1,037527 42 167,6 174 6,4 1,038186 43 142,5 148 5,5 1,038596

44 166,4 173 6,6 1,039663 45 138,2 144 5,8 1,041968 46 134,1 140 5,9 1,043997 47 161,6 169 7,4 1,045792

48 175,9 184 8,1 1,046048 49 144,2 151 6,8 1,047156 50 162,2 170 7,8 1,048088 51 176,4 185 8,6 1,048752

52 171,4 180 8,6 1,050175 53 155,8 164 8,2 1,052631 54 169,1 178 8,9 1,052631 55 119,2 126 6,8 1,057046

56 129,6 138 8,4 1,064814 57 165,2 176 10,8 1,065375 58 138,3 149 10,7 1,077368

Average 156,1 160,8 4,7 1,03

Table 4 – Validation table




Metria – Sweden





Page 20


It is worth mentioning that the provided DSM indicates, in general, the height above the canopy. However, as there is some penetration of the radar signal into the canopy, where there is low vegetation, the derived height will more closely resemble a Digital Elevation Model. On the other hand, where there is high vegeta-tion, it is expected that the DSM will be higher than the reference DEM. An error of 3% is very acceptable for applications discussed below.

Fig 6 – DSM, UTM 39 North, with landscape shp 1:25,000 (5 m)

Fig 7 – DSM, UTM 39 North, with landscape shp 1:200,000 (20 m)




Metria – Sweden





Page 21


Evaluation

The interferometric DSM is significantly better than anything currently available to MondiSYK, and the fact that it is not a DEM has potential operational advantages for MondiSYK. However, its use in the Russian Re-public might have to be specified quite carefully, to comply with national security laws. The product would have to be generated in 250 km2 blocks with an unclassified or masked 1 km strip between each classified block providing a series of 16 by 16km maps from each 100 by 100 km scene. This is simply a matter of set up, and should not be regarded as a problem. It was agreed that a DSM with a scale of 1:75,000 would be a practical solution for MondiSYK as a mapping/monitoring tool. Potential applications of such a product are:

A. MondiSYK and forestry companies Used in conjunction with other Products, it is could be used for monitoring FSC compliance in following areas: - improved ecosystem classification for monitoring - defining watershed and water protection zones based on landscape approach - initial general planning of logging areas, - preliminary planning of the main roads layout within future logging areas. It would not be sufficiently high resolution for detailed logging road layout design, where a DEM of 2m geographic and 5m elevation is required.

MondiSYK also identify potential added value in combining the DSM with other EO classification layers to provide or confirm other forest/land cover characteristics, whereby small DEM ~ DSM deviations could indicate non forest area and high deviations would indicate forest areas.

B. Certification bodies Such a DSM could help MondiSYK provide better identification and detection of river valleys/basins to the CBs on water protection of bogs and riparian protective forests.

Overall the performance of the product portfolio in terms of each value criteria Acquisition frequency Poor Fair Good Excellent

Spatial coverage Poor Fair Good Excellent

Product spatial accuracy Poor Fair Good Excellent

Product dependency Poor Fair Good Excellent

Independence Poor Fair Good Excellent

Product timeliness Poor Fair Good Excellent

Product accessibility and Poor Fair Good Excellent transferability




Metria – Sweden





Page 22


Product 1 - Forest area

Description Geo-referenced information on location, extension, and distribution of forest area.

Location Komi

Data sources - ALOS PALSAR-1 FBD data (15m) acquired on 2nd Oct 2009

- SRTM Digital Elevation Model version 4

Product - Colour composite in GeoTif format of ALOS PALSAR-1 FBD data

- Forest area cover in GeoTif format and/or shape format


Fig 8 - ALOS PALSAR-1 FBD colour composite (left) and corresponding product (right).




Metria – Sweden





Page 23


Fig 9 - ALOS PALSAR-1 FBD colour composite (top) and corresponding product in raster (middle) and shape format (bottom).




Metria – Sweden





Page 24


Validation (see also the additional accuracy assessment based on a confusion matrix at Annex 4 to this re-port and Annex 3 for overview of the dynamic and complexity of Taiga forest environment).

Classification has to be checked in the field in summer, in order to make proper naming for classes. So far the following grouping and renaming has been adopted.

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

1. Mature old forest, intact, along rivers

2. Middle-aged secondary forest

3. Young forest, shrubs

4. Settlements

5. Bare soil / settlements

6. Wetlands

7. Wet lowlands

8. Rivers

9. Lakes & rivers

10. Grass/bushes (reforestation or abandoned fields near villages) 11. Bare soil, agricultural land, fresh logging sites

Table 5 – PALSAR classes (left) and renaming according internal classification (right).

For the validation, 148 circular plots with a total amount of trees more than 20 have been considered. The total area per plots is 500 sqm. The results are included in the table below. The average error is 13.5%.

Class Number of Plots

High vegetation - forest 38

Medium vegetation - forest 90

Rough soil 12

Short vegetation/very rough soil 8

Total 148

Table 6 – Validation table

Evaluation

A. MondiSYK and forestry companies This product can be used for:

- identifying and monitoring forest and non forest areas - comparing total area of clear-cuts for further comparing to a registered area of logging sites - monitoring clear fell sites and retained seed trees - vegetation classification (species identification) is not possible with Palsar alone, but this is not criti-

cal as this can be achieved with World View 2 and SPOT, though it is usually only possible to get one cloud free acquisition a year. There is a cost issue for World View 2 when used over very large areas.

Because MondiSYK would need this forests mapping in association with forest species distribution, this product would have to be used in association with additional data (ground and EO).

Overall the performance of the product in terms of each value criteria Acquisition frequency Poor Fair Good Excellent



Product biophysical accuracy Poor Fair Good Excellent




Metria – Sweden





Page 25






MondiSYK added “Product Dependence” because it is operationally very important that area mapping is combined to species separation. Therefore this product would have to be used in association with alterna-tive sources of remote sensing data and /or ground data. The key advantage of this product is its ability to operate in cloudy conditions.




Metria – Sweden





Page 26


Product 5 - Clear-cut, logging, and re-growth

Description Geo-referenced information on location and extension of the clear cuts and re-growth.

Location Komi

Data sources - ALOS PALSAR-1 FBD data (15m) acquired on 27th Sept 2007, 14th Aug 2008, and 2nd Oct 2009

- Cosmo-SkyMed-1/-2 StripMap InSAR data (3m) acquired on 24th and 25th Sept 2010


Product - Colour composites in GeoTif format of multi-temporal ALOS PALSAR-1 FBD and Cosmo-SkyMed-1/-2 InSAR data

- Clear-cut and re-grow in GeoTif and shape format


Fig 10 - Multi-temporal ALOS PALSAR-1 FBD colour composite (left) and corresponding product (right).




Metria – Sweden





Page 27


Fig 11 - Cosmo-SkyMed-1/-2 StripMap InSAR colour composite (top) and corresponding product (bottom).




Metria – Sweden





Page 28


Validation

Product based on Palsar Data

Fig 12 - MondiSYK plots (in red) and Clear-cut and re-grow derived from PALSAR in shape format.

Top image - Landsat TM-5 August 2010, Middle – Landsat TM-5 August 2007, Bottom – World-View-2 CIR_GIST August 2010.




Metria – Sweden





Page 29


Plot_OID Total plot Area PALSAR Area Deviation (%)

-2147482902 Итог 18,041574 15,56349 13,73541466

-2147482685 Итог 24,774199 23,50254 5,132993402

-2147482682 Итог 23,150261 21,03585 9,133430504

-2147482679 Итог 15,364433 13,60733 11,43614607

-2147482677 Итог 12,624699 12,5487 0,602026234

-2147481733 Итог 33,177881 31,08546 6,306680647

-2147481718 Итог 18,017871 14,36502 20,27348847

-2147481647 Итог 25,989114 23,85227

8,222080984

-2147481459 Итог 45,074659 38,36152 14,89337057

-2147481458 Итог 25,423217 22,58898 11,14824296

-2147481150 Итог 33,88233 29,98172 11,51223957

Subtotal '0' 51,104481 51,10448

Total 326,624719 297,5973

Average 10,21782855

Table 6 – Validation data of areas clearcut polygons derived from MondiSYK database and corresponding areas derived from PALSAR data. Unit of area is ha.

In terms of:

• Temporal accuracy of detection of clear cut areas using this product is 100 % when comparing annual clear cut areas against the MondiSYK database or Landsat TM-5 confirm clear cuts were done in summer of 2009.

• The spatial accuracy of the clear cut area derived from the PALSAR product is on average, 10,2 % less than total plots area using ground measurement (only polygons inside plots was taken). When comparing the Product using Cosmo-SkyMed, MondiSYK is sure that the accuracy will be between 95% and 99%, but this cannot be calculated, as MondiSYK do not have any clear cut plots within the area of Cosmo-SkyMed-1/-2 StripMap InSAR data (3m) acquired on 24th and 25th Sept 2010.

Product based on CSM data

No validation could be provided for the Cosmo-SkyMed product, because it lay outside the MondiSYK con-cession. Therefore MondiSYK do not have sufficient data from the CSM acquisition area to validate this prod-uct. We plan to perform this in the summer when field data can be collected.

Evaluation

A. MondiSYK and forestry companies

Each year, MondiSYK’s has to make detailed plans for some 1,000 different logging areas. Adequate track-ing of the activity in these logging sites and associated protected areas is critical to FSC compliance and cer-tification. Errors insufficient data is a basis for CARs. Therefore MondiSYK need to confirm the extent to which the planned annual felling programme has been met – both in overall area and well as the exact loca-tion of clear cut areas. This can be achieved with the Palsar product. For the monitoring the quality of log-ging activities, especially the retention of seed trees and potential impact on protected areas, the higher resolution CSM product could be used. The great benefit of a programme based on radar sensors, would be the significant reduction in cost from using World View 2 ($35/km2) as well as increased frequency and flexi-bility of planned acquisitions as this area has cloud cover on 80% of days.

The potential benefits of delivered products:




Metria – Sweden





Page 30


- Can be used for actualizing logging sites in the ForestIS system (Web-GIS system for registering de-lineated logging sites). It would simplify the removal of those sites which were planned but have not been logged because of weather or other restrictions.

- Can be used for monitoring of set-aside areas – retention trees, biotopes, HCVFs. Composite of re-tention trees with HVCF will give possibility monitor all year MondiSYK plots.

- Can be used for checking borders violation of delineated logging sites.

- Using a combination of radar sensors increases frequency and flexibility of and EO based monitoring programme.

Fig 13 – Clearcut areas showing environmental protection zones (image generated by Lantmateriet, METRIA in association with Holmen Skog).

Overall the performance of the Product based on Palsar providing temporal information Acquisition frequency Poor Fair Good Excellent


Product spatial accuracy (CSM) Poor Fair Good Excellent

Product spatial accuracy (PALSAR) Poor Fair Good Excellent

Product temporal accuracy (PALSAR) Poor Fair Good Excellent








Metria – Sweden





Page 31


Product 10 - Road mapping

Description Detection and extraction of roads in forest areas. This product serves in primis to the detec-tion of on-going illegal forest logging activities.

Location Komi

Data sources - ALOS PALSAR FBS data (8m) acquired on 09 Feb 2007, 28 Dec 2007, 30 Dec 2008, 14 Feb 2009, 04 April 2010 data.


Product - Averaged ALOS PALSAR FBS data in GeoTif format

- Road map in shape format






Metria – Sweden





Page 32


Validation

Fig 15 – Analysis of PALSAR derived road network (red).

The total length of product roads inside MondiSYK leased area is 116.8 km. The estimation of errors is 7.8 km, of which 1.3 km are in reality electricity lines and 6.5km rivers and valleys.

Evaluation

A. MondiSYK and forestry companies

Potential applications and benefits of EO products are:

- Mapping of roads are required for monitoring existing road networks and for layout and detail de-sign of roads within new logging sites. Road inventory is one of the biggest problems in planning process. Road mapping can provide MondiSYK with actual data about roads for strategic road con-struction planning.

- Monitoring MondiSYK’s existing concession roads is done using GPS mounted in logging trucks. The data is collected and then analysed and entered into the GIS systems at the end of the year. Roads are updated annually and ideally, this should be done in the key down time which is May and No-vember.




Metria – Sweden





Page 33


- Road network planning by MondiSYK also need two levels of spatial detail in planning new roads. At a lower level, layout of main roads could be done using EO data, followed by very detailed planning of logging sites when bogs and water courses need to be clearly identified and protected.

- Using Radar allows MondiSYK to optimise the timing of planning and monitoring activities.

- Combining different products (e.g. DSM, Forest and non Forest products) could support both moni-toring and planning process for roads.

- Using EO for road monitoring and mapping in new leased areas

Overall the performance of the Product 10 in terms of each value criteria Acquisition frequency Poor Fair Good Excellent










Metria – Sweden





Page 34


Product 12 - Settlements

Description Geo-referenced information on location and distribution of settlements.

Location Komi


- Cosmo-SkyMed-1/-2 StripMap InSAR data (3m) acquired on 24th and 25th Sept 2010


Product - Colour composites in GeoTif format of ALOS PALSAR-1 FBD and Cosmo-SkyMed-1/-2 StripMap InSAR data

- Settlements including other main land cover classes in GeoTif format and/or shape format






Metria – Sweden





Page 35


Validation

Settlements information at MondiSYK is only available in form of polygons, where towns/villages or the en-semble of houses are delineated by boundaries, hence generalized in cartographic terms; e.g. the resulting polygon does not necessary include all infrastructures. An example is illustrated in Figure 16 below.

Fig 17 - Google Earth image (top) and corresponding Palsar classification including MondiSYK provided shape of town boundaries (bottom).

The only way to compare the raster classification with these generalized polygons is therefore based on a visual inspection, resulting obviously in an under-classification of this class.




Metria – Sweden





Page 36


MondiSYK conclusion is that the PALSAR based product is suitable to identify medium-large settlements, but small ones are hardly mapped.

Evaluation

MondiSYK, forest companies and Certification Bodies

This product could be used to

- detect abandoned settlements for creating HCVF 5.

- detect abandoned industrial areas for requesting re-cultivation











Metria – Sweden





Page 37


Product 13 - Land cover in non forest areas

Description Thematic map of the land coverage in the surrounding forest area. In general, following ma-jor land cover classes (which depend on the sensor) are considered: bare soil, settlements, cropped areas, water bodies, low vegetation (grassland), average vegetation, rangeland, rocks. On request and whenever appropriate, specific classes are included.

Location Komi



Product - Colour composites of ALOS PALSAR-1 FBD in GeoTif format

- Land cover in GeoTif and/or shape format






Metria – Sweden





Page 38


Validation (see also the additional accuracy assessment based on a confusion matrix at Annex 4 to this re-port, which should be used in addition to the MondiSYK validation work below. Also, see Annex 3 for over-view of the dynamic and complexity of Taiga forest environment).

Validation was done based on forest inventory data. Shape files of forest inventory consist of segments (polygons) with associated forest description. Swamps are indicated separately in this shape. 31 sample plots have been taken for the validation. They are centroid of all swamps from forest inventory polygons (only in the MondiSYK leased area). The average error is 6.5%.

Class Number of sample

plots

Medium vegetation - forest 2

Rough soil 9

Short vegetation mixed with water or low dense settlement 2

Smooth bare soil/wet bare soil 18

Total 31

Table 7 – Validation table – environmental monitoring of land cover in non forest areas

Fig 19 – PALSAR derived land cover in shape format (red) and Landsat TM-5 in true colours.

Evaluation

Rivers and wetland are easily recognised on the Product. Clear cuts and re-growth can be detected (Refer to product 1 and 5). An advantage of this SAR derived product is that wet / dry forests can be identified, which is not possible using optical sensors.

MondiSYK, forest companies and Certification Bodies

This product can be used to determine:

- the area and location of wet forest (boggy and riparian forests)

- changes in the area ratio and spatial relationship between clear cuts and wet forest




Metria – Sweden





Page 39


- the location of future areas suitable for winter/non winter harvest.











Metria – Sweden





Page 40


Product 14 - Biophysical parameters

Description Estimation of biophysical parameter such as tree height, diameter at breast height, basal area, tree per hectare, and timber volume.

Location Komi

Data sources - ALOS PALSAR-1 FBD data (15m) acquired 2nd Oct 2009


- Geo-referenced homogeneous forest samples covering all ages of the bio-physical parame-ters to be estimated

Product - Timber volume in GeoTif format


The major problem experienced in doing this trial has been access to georeferenced ground data with which to calibrate the radar backscatter.

Change segments:

Fig 20 - Timber volume product (left) and detail (right: ALOS PALSAR-1 FBD colour composite and product).




Metria – Sweden





Page 41


Validation (also refer to Annex 5 on the sensitivity of L Band data to saturation when estimating timber vol-umes above certain values. Also, see Annex 3 for overview of the dynamic and complexity of Taiga forest environment).

Fig 21 - WorldView-2 CIR_GIST August 2010 (left) and WorldView-2 CIR_GIST August 2010 with Timber volume in GeoTif format, transparency 70% (right)

Fig 22 – Detail of selected area in Fig 19, with Timber Volume sample data (m3/ha)




Metria – Sweden





Page 42


Validation was done by 68 samples based (see Fig 21 above) of forest inventory data (cancroids of forest segments less than 30 ha). The results are summarised in the table below.

vol_class quantity errors %

0 10 3 30,00

0-50 5 4 80,00

50-150 21 8 38,10

>150 32 4 12,50

Total 68 19 27,94

Table 8 – Validation of biophysical classes (Vol units – m3/ha)

Volume per segment should be defined for more objective validation.

Evaluation

A. MondiSYK and Forestry companies

The three key commercial forest groups are:

- Lower productive forest which have a biomass below 50m3/ha

- Medium or pre-economic forest with biomass between 50-150m3/ha

- High vegetation or economic forest with biomass above 150m3/ha.

This product can be used:

- as a rough check for major errors of a forest inventory data (ground data base). In other words, it can be a kind of a reliability indicator.

- when choosing logging sites for cutting and defining general direction of new roads.

- Update errors – e.g. some clear-cut areas have not been reflected in the ground database or other deforested areas were not checked during forest inventory.

- Inventory errors – i.e. some field data was input and processed in a wrong way, when non-forest area could suddenly appear as a forest.




Product biophysical accuracy Poor Fair Good Excellent








Metria – Sweden





Page 43


4. Service Portfolio Evaluation

4.1 EO Technology Utility

The evaluation of demonstrated products, based technology and product issues is as follows:

CRITERIA Issues for consideration MondiSYK comments

Technical barri-ers

• Do you see any technical barriers in using EO data?

• Are there technical skills required which are not typically available within your operations?

• If so, could you usefully outsource this service?

• Any questions about “technical efficiency”?

• There have been some compatibility problems using products in ArcView 3.2 and ArcGIS 9.3, where coordinates did not fit.

• Under Russian Law, it might be necessary to customise the presentation of DSM as discussed

• Intermediate level in ArcGIS is needed. Currently there are 2 specialists in-volved in this process.

• There are about 4-5 companies selling GIS and remote sensing services

System Conti-nuity

(see Annex 2 concerning

Palsar failure)

• Are you satisfied that the service continuity can be assured? Are you satis-fied that data derived from optical and radar sensors can be substi-tuted?

• At high latitudes, there are at least 4 Palsar acquisitions per year. In addi-tion, Palsar has built in redundancy, so risk of non availability is low

• Cloud cover is an issue for 80% of days. Being able to get acquisitions throughout the year would be very valuable.

• Taking into consideration frequent and vast forest fires in Russia in is valu-able still to be able to use EO

Data transfer-ability

• Could the use of the EO data be of value to others in the FSC chain? If you provided CBs with data derived from this prod-uct, do you believe that this could justify lower certification costs?

• There are three key groups to whom spatial mapping data and derived area/volume data need to be transferred: o MondiSYK management, contractors, suppliers o Komi Forest Committee o FSC economic section.

• Typically, this data includes clearcuts, rate of harvest, biotopes, and maps of

productive zones. EO products support the transferability of such data. • Using EO data CBs can better identify critical points and improve quality of

audits, while amount field work won’t decrease: o efficiency of HCVF conservation o legality of wood cutting o quality of loggings (retention of biotopes in clear cuts, intensity of se-

lective cuttings)

Overall ranking of EO Products in terms of each of the key product utility criteria

Technical barriers High Medium Low System continuity Poor Fair Good Excellent Data Transferability Poor Fair Good Excellent




Metria – Sweden





Page 44


4.2 Summary of Portfolio Utility

Evaluation Criteria All products Frequency Optical acquisitions restricted to one per year. Flexibility on timing of acquisition using Radar im-

portant. Future EO acquisition programme would reflect cost. e.g.:

• Worldview 2 – every 10 years • CSM – every year for logging areas • Palsar – on demand for selected areas

Spatial Coverage Currently: • MondiSYK has 2,1m ha of concessions and forest activities within 250km radius of Syktyvkar

mill. 100% of the leased area needs regular monitoring. Monitoring using ground data sources only is not enough.

• Airborne mapping is not permitted in Russia • Existing forest inventory and infrastructure maps are out of date Demonstrated products provide 100% spatial coverage of all MondiSYK leased concessions

Spatial and Biophysical Data accuracy

• DSM would be sufficient resolution for preliminary indoor planning. For detailed planning ground data is necessary. Demonstrated product good for forest and non forest monitoring

• CSM and Palsar, supported by SPOT provide a good range of spatial resolution. • CSM higher resolution offers great potential for monitoring clear fell areas, seed trees • and regrowth

Independence • Equivalent to 3rd party independently generated data – this is valuable • There is a lot of additional value in having additional data sources and layers in the ForestIS

system

Accessibility • DSM may require formatting restrictions to meet Russia security regulations

Timeliness • Products could meet temporal specification for all data uses: o Annual planning, o FSC certification o Long term (10 year) strategic plans.

• EO products very useful for monitoring requirements under FSC • Acquisition of SAR in all weather all year round in Komi is big advantage over optical

Overall Technology Integration (select one of following statements):

This technology could replace existing conventional data source without further development or

delay for Products - none

This technology could be used to reinforce existing conventional data source because it adds to the

quality of data for Products 5, 10.

This technology could replace or reinforce existing sources, but will not be for other company spe-

cific reasons (e.g. investments made, data acquired, contractual commitments etc) for Products

0 (DSM).

This technology could not replace existing practice but could subject to improvements (please spec-

ify below) for Products 1, 13, 14.

This technology could not be considered as a replacement or to reinforce existing data sources (ex-

plain below) for Products 12.

This technology could provide a completely new product option for Products - None




Metria – Sweden





Page 45


Any Comments:

For Products 1, 13, 14 the following improvements would have been of value:

• species separation,

• volume calculation (e.g. QGIS Toolkit),

• wet forest detection,

Product 12 - Does not correspond to required detail level for detecting human made objects.

4.3 Potential for cost saving in generating in situ or conventional data

Issues for consideration Response and comments from Trial Partner

What impact could EO have on current conventional data collection and enumeration costs?

Taking in to consideration the fact that EO data cannot substitute conventional data and can only give additional information all costs related to EO application will be also additional.

What impact could it have on operational efficiency at the FMU level?

It can help in updating existing forest and infrastructure maps, which can improve planning accuracy.

Could EO reduce the cost of the certification process if the demonstrated products were used within the audit?

• It can improve quality of field checks, but can not reduce cost, as soon as amount of field work is usually increasing.

• With improvements and regular use (date stamping) it is likely to build confi-dence of auditors and operations and make the audits more efficient.

Do you think it could it reduce audit manpower and site visit costs?

It can decrease amount of indoor work by reducing a number of papers and docu-ments being reviewed for defining critical points.

Could the portfolio of services improve the quality or reduce costs of the measurement and valuation of growing assets as required under accounting stan-dard IAS 41?

It could give information about quantitative changes of resources (deforestation volumes, standing volume increment, regrowth, etc.), but without qualitative estima-tion (changes in species composition, assortments structure, etc.)

If you do not consider EO data cost effective. please indicate what the barrier to adoption might be:

• cost of imagery; • cost of EO processing software; • capacity building; • GIS software cost;

Cost of SAR images will be additional, as soon as optical data will be used anyway. Thus, additional cost can be a barrier.

4.4 Overall Value for money

Based on their analysis, MondiSYK was asked to rate the overall cost effectiveness of the Trial Product Port-folio. Cost effectiveness assessment and grading is based on the indicated service cost (see Section 2.3 above) and estimated operational cost savings and summarised on the following scale:

• Excellent value for money - 4

• Very good value for money - 3

• Good value for money - 2

• Poor value for money - 1

MondiSYK: Good value for money (2) – i.e. not enough approbation and validation of products in forestry,

in order to have independent and sufficient products more application in real operations is needed.

It is expected that the quality and effectiveness of the technology and applications will improve rapidly with




Metria – Sweden





Page 46


more use. The use of open source GIS holds much promise as a “front-end” communication/education tool

for company-community cooperation monitoring of natural resources.

4.5 Technology Risk and Reward

The technology risk/reward of the overall service package has been assessed by the trial partner as High,

Medium or Low where:

Reward Risk

High Certain long term operational benefits Likely operational risks or high uncertainty of benefit

Medium Possible long term operational benefits Possible operational risks or some uncertainty of benefit

Low Few probable long term benefits No discernable operational risks or uncertainty of benefit

Using the following grid, we also asked all stakeholders’ to provide an overall indication of what they consid-

ered the potential risk-reward of technology adoption might be. In the table below, we have provided a

summary of each stakeholder’s selection.

Stakeholder selection Reward

Low Medium High

Low

Medium ▲MondiSYK Risk

High




Metria – Sweden





Page 47


Annex 1 – Product Evaluation Criteria – Issues to be considered

Evaluation Criteria Issues for consideration

Acquisition Frequency

• Is the data sufficiently current?

• Can the date of acquisition be sufficiently accurately specified to a required point in time?

• Is the repeat pass interval acceptable for monitoring changes?

• Are there other factors that limit data availability; are you satisfied that alternative sensors can be used

to increase frequency; is cloud cover an issue in your area?

Spatial Coverage • Does sensor cover geographic areas of interest?

• Are there topographical issues (e.g. slope) that limit performance?

Product accuracy (spa-tial and biophysical)

• How does the accuracy of conventional sources of information compare to the supplied product?

• Does mapping resolution match your spatial and mapping accuracy requirements?

• Is classification sufficiently detailed for task?

• Is classification sufficiently consistent across sensors?

• Is classification consistent across satellite sensor and software?

• Have you identified any unexplained differences between EO derived information and your conventionally

derived maps?

• Is there an acceptable level of correlation between conventional data and EO data? If not, is there an

explanation to delta?

• Is the extractable data of value in reducing forest management costs or operations?

• Is the extractable data of value in forest accounting (IAS-41) (Biophysical product only)

Product Dependency • Does this product rely on another source of EO or ground data for generation which is problematic

• Is the utility of this product linked to another product or service which is problematic

Independence (3rd party data)

• Is data independence important?

• Is data transparency important?

• Is data repeatability important?

• Are conventional sources of this data subject to potential bias which might be overcome with this prod-

uct?

Product accessibility and transferability

• Can you foresee operational barriers that could restrict data access or product utility?

• Is data easily transferrable to other FSC stakeholders – especially CBs?

• Can you foresee restrictions on data or technology transfer which neutralise this value?

Product Timeliness

• Can the product be delivered sufficiently quickly to meet your needs for forest management and certifi-

cation?

• Is the speed of EO data processing and delivery of classified images critical?




Metria – Sweden





Page 48


Annex 2 - ALOS PALSAR-1 failure

An important development since the inception of the project, and completion of the demonstration trials in the three trial areas, has been the announcement that the ALOS sensor was terminated on April 23 due to a malfunction in the solar power generation system. In the event power cannot be restored, Sarmap has pre-pared the following overview to establish the potential for other sources of radar data to be used as alterna-tives for all mapping and monitoring services, except for Timber Volume estimation which requires an L band radar sensor for which there is no current replacement though replacement sensors are scheduled over the next few years3.

ENVISAT ASAR AP data acquisitions have been requested for May 2011 and subsequent months over Komi project area in order to understand the potential contribution of ASAR AP data for the demonstrated prod-ucts. In the meantime, some ENVISAT ASAR AP data is available to Sarmap over an earlier ESA funded pro-ject (see RD-4) the following overview serves as an interim demonstration of data substitution and mapping compatibility in South African plantation conditions.

In addition, Cosmo-SkyMed data was acquired over both the Sappi and MondiSYK project areas, and the data consistency and comparability demonstrated in both these areas, particularly for precise mapping of clearfell areas.

The following compares the above data sources with Also Palsar data, and confirms their functional com-patibility.

3 Three additional L Band sensors will be launched in the next 2 years - ALOS-2 (3m resolution) will be launched in 2012 and SAOCOM-1/2 (10m resolution) will be launched 2013/2014




Metria – Sweden





Page 49


Figure 1 illustrates, for comparison purposes, ALOS PALSAR-1 FBD (L-band, HH/HV polarization), ENVISAT ASAR AP (C-band, HH/HV polarization), and Cosmo-SkyMed (X-band) at HH/HV polarization and acquired in an interferometric way at HH polarization. A detail, including a forest map based on L- and X-band data, is shown in Figure 2. Note that all data, by considering the different sensor's characteristics and acquisition modes, have been processed in the same way. Figure 3, finally, shows the regressions for various biophysi-cal parameters used for the generation of (validated) biophysical maps for Jessivale, South Africa (ESA, DUE II, 2010).

Figure 1 – South Africa, Jessivale – ALOS PALSAR-1 FBD 15m, August 2008 (top left); ENVISAT ASAR AP 15m, February 2010 (top right); Cosmo-SkyMed (HH & HV) 3m, Jan 2010 (bottom left); Cosmo-SkyMed 3m InSAR HH, December 2009 (bottom right).




Metria – Sweden





Page 50


Figure 2 – South Africa, Jessivale – ALOS PALSAR FBD data acquired on August 2008 and correspond-ing forest map (left) and interferometric Cosmo-SkyMed HH data acquired on December 2009 and corresponding forest map (right).




Metria – Sweden





Page 51


Figure 3 – South Africa, Jessivale.

Top - Relationship between selected biophysical parameters and ALOS PALSAR HV data ac-quired on August 2009.

Bottom - Relationship between selected biophysical parameters and ENVISAT ASAR HV data acquired on February 2010.




Metria – Sweden





Page 52


Based on these key examples and on the experiences done within this project and in the previous DUE II one, following conclusions can be made:

– Clear evidence of usefulness of PALSAR FBD thematic products for thematic purposes. In overall, the level of detail (in terms of information) is slightly increased in non-forest areas, if HH coherence is additionally considered. In general, ALOS PALSAR FBD data provide similar information compared to SPOT-4 data.

– Due to the higher spatial resolution of Cosmo-SkyMed (3 meter) and complementary X-band fre-quency, the information content is significantly increased particularly in non forest area, when 1-day interferometric Cosmo-SkyMed HH data are used. Clear-felling, as shown in the Komi trial, is accurately detected.

– The use of ENVISAT ASAR AP (HH/HV polarization) data, in particular based on time-series, is doubtless valuable in particular for forest area changes and for non forest areas.

– In South Africa, a clear relationships were established between L-band HV data and field based, geo-referenced plot level standing volume, average tree height, average DBH and basal area. This is not possible for C- and X-band data.




Metria – Sweden





Page 53


Annex 3 Forest Succession Table – Komi

The following table which provides an overview of the forest dynamics, has been developed from a report by Ministry of Natural Resources of the Russian Federation, Forestry Agency for the Komi Republic, Silver Taiga Foundation, Priluzje State Forestry Unit entitled “Recommendations on logging operations with conservation of ecological values in intact (pristine) forests in the Komi Republic” (July 2005).

Indicator DECIDUOUS-TO-SPRUCE PINE-TO-SPUCE PINE TO PINE SPRUCE NON-FIRE PINE FORESTS ON

BOGGY SOILS

BIRCH FORESTS ON

BOGGY SOILS

Fire frequency On average, 250-500 years. On average 200-350 years. On average 50-200 years. Extremely rare < once

every millennium.

Very rarely, 1-2 times

in a millennium.

Very rarely, 1-2 times

in a millennium.

Tree species com-position:

From pioneer deciduous to coniferous stands of every

tree age, including all

transition succession stages.

From deciduous stands with different part of pine to spruce

stands with little share of de-

ciduous species and different part of pine.

Pine stands with little share of birch, and some-

times non-productive

spruce never in the first layer.

Spruce forest with possible small share of birch.

Boggy pine forests,

different share of

birch trees.

Boggy birch forests,

sometimes with very

depressed spruce undergrowth.

Typical location:

Forest near fluvial terraces

of small rivers and streams, low lands near

fluvial terraces, boggy

watershed.

Drained places with loamy soils.

Fluvial terrace with sandy

soils.

Along rivers and streams,

on boggy watersheds.

Swamped places.

Swamped places with

stagnant, circulating or seasonal moisten-

ing.

Processes and

stand structure:

• Total destruction of a

forest stand after fire;

• Forestation begins from

birch and/or spruce;

• Spruce gradually re-

places deciduous;

• Fire interrupts formation

of uneven-age spruce

stand.

• Total destruction of spruce

and deciduous trees after fire;

• Part of pine trees, some-

times of several generations,

survive a fire;

• Forestation begins from pine

and deciduous species;

• Spruce appears under a for-

est canopy, gradually comes to

the first layer and forms a stand of absolutely different

tree age;

• destruction of deciduous

generations;

• pine is permanently pre-

sented in a stand and trees are

of a different age frequently.

• destruction of some

part of pine trees after

fire;

• forestation begins with

pine;

• uneven-aged pine forest

is formed with clear pine

generations;

• spruce if appears is

never dominating because of low soil productivity

and high frequency of

fires.

• formation and foresta-

tion of gaps – determining

process;

• stand is of different

tree age;

• trees are of various

diameter;

• vertical crown density;

• permanent deadwood;

• depressed spruce un-

dergrowth.

• rare and low pro-

ductive stand;

• absence of fires;

• absence of spruce.

• rare and low pro-

ductive stand;

• very depressed

spruce.




Metria – Sweden





Page 54


Indicator DECIDUOUS-TO-SPRUCE PINE-TO-SPUCE PINE TO PINE SPRUCE NON-FIRE PINE FORESTS ON

BOGGY SOILS

BIRCH FORESTS ON

BOGGY SOILS

Main feature:

REPLACEMENT of post-fire

generation of deciduous trees by spruce and DE-

STRUCTION of deciduous.

REPLACEMENT of deciduous by spruce. DESTRUCTION of post-

fire deciduous species. CON-

TINUOUS PRESENCE OF PINE.

FREQUENT FIRES form and

support DOMINATION OF PINE and constancy of ALL

AFTERFIRE STRUCTIRES.

spatial and temporal con-

stancy of ecological struc-tures and processes.

Extremely LOW PRO-

DUCTIVITY in the PERMANENTLY over

moisturized environ-

ment.

Extremely LOW PRO-

DUCTIVITY in the PERMANENTLY over

moisturized environ-

ment.




Metria – Sweden





Page 55


Annex 4 – Additional validation for Forest Area (Product 1) and Land Cover (Product 13)

following the confusion matrix procedure

This additional validation is to be understood as complementary to the extensive and 'user-driven' validation (and evaluation) work performed by the MondiSYK team. In that validation, the work was carried out, based on the current practice and commercial objectives of local foresters, who know the operational terrain and forest environment.

A complementary way to assess accuracy is given by a general tool for assessing the thematic accuracy, regardless of the definition used for the total accuracy, is the error matrix or the confusion matrix, from which different accuracy measurements can be derived. It is worth mentioning, that of key importance, among others, for the accuracy assessment are:

1. The reference data that is used for comparison with the actual product. 2. The sampling strategy that will be used to get a representative and robust comparison of the product

with the reference data. 3. Reference data must be collected using the same classification scheme that was used for the remotely

sensed data, and should also be applied to the same minimum mapping unit as was applied to the re-motely sensed data.

Since in this specific case, the available information at MondiSYK over the trial area is scattered in terms of time and location, we have opted - for the provision of the confusion matrix - to use two high resolution optical images (hence independent sources at approximately the same resolution as ALOS PALSAR-1) ac-quired as close in acquisition time as possible to the ALOS PALSAR-1 scene used for the generation of the product (Figure 1). In summary, the reference data set consists of:

– SPOT-4 scene (20m) acquired on May 13th 2010;

– Landsat TM-5 (30m) acquired on August 12nd 2009;

while the Product (Figure 1) has been generated using:

– ALOS PALSAR-1 FBD data (15m) acquired on February 2nd 2009.

In order to not select points which are convenient to obtain a priori a good assessment, we opted to overlay on the common area showing the highest spatial heterogeneity (hence not only forest, which would lead to a very good accuracy) a regular grid with a sampling of 2 km. Figure 2 shows the grid overlaid to the three scenes for the overall selected area, while Figure 3 illustrates a detail.




Metria – Sweden





Page 56


Figure 1 - Left: Colour composite in GeoTif format of ALOS PALSAR-1 FBD data

Right: Forest area cover in GeoTif format and/or shape format

Cartographic and geodetic system: UTM-39 North, WGS-84




Metria – Sweden





Page 57


Figure 2 - Top left Landsat TM-5, Top right SPOT-4, Bottom ALOS PALSAR-1. The total number of points, in yellow colour, is 288 (16 times 18).




Metria – Sweden





Page 58


Figure 3 - Top left Landsat TM-5, Top right SPOT-4, Bottom left ALOS PALSAR-1, Bottom right Product. Se-lected points are in yellow.




Metria – Sweden





Page 59


The matrix below illustrates the confusion matrix obtained by comparing a visual interpretation the Product based on ALOS PALSAR-1 and the two reference scenes point by point. It is worth mentioning that:

– 20 points (of 288) have not been considered, because covered by clouds or shadow.

– In general, it was often not easy to interpret the Landsat TM-5 image, because areas are af-fected by haze. This image has been used, were the SPOT-4 were affected by clouds or shadow (around 60 points corresponding to 20% of the total points).

– 5 points have been adjusted, e.g. classified as forest in the Product but as bare soil in the SPOT-4 scene (and hardly identifiable in the Landsat TM-5). Since SPOT-4 has been acquired around 7 months later, and they show evident clear fell, they have been moved from forest to bare soil.

Reference scenes - Landsat TM-and & SPOT-4

Forest Bare Soil Low Vegetation River Wetlands Total

Forest 161 3 1 2 167

Bare Soil 6 59 5 1 71

Low Vegetation 2 24 26

River 0 0

Wetlands 1 3 4

Total 169 63 30 3 3 268

Pro

duct based o

n P

ALSAR-1

Producer's Accuracy 95% 94% 80% 0% 100%

Overall Producer's Accuracy 74%

Overall Producer's Accuracy omitting River 92%

In summary: the confusion matrix confirms the assessment provided by the MondiSYK team. Nevertheless, it has to be pointed out that this type of assessment must be treated with caution: Numbers are only a com-ponent of the real world; human knowledge and feeling is also valuable.




Metria – Sweden





Page 60


Annex 5 - Clarification on the sensitivity of the L Band data to saturation when estimating the

timber volume above certain values.

Figure 1 illustrates the relationship between Timber Volume (m3/ha) vs. L-band HV radar backscatter for Zululand (top) and Komi (bottom), while Figure 2 the results obtained by Peregon in boreal forest.

Based on these three independent trials, the following conclusions can be made:

− In boreal forest conditions such as those in Komi, the L band Palsar saturation level is around 150 (m3/ha). There are three major factors limiting the use of L-band data at these geographic latitudes:

o The complex structure and highly varied nature of the forest (confirmed by the field data);

o The conditions of the understory biomass (younger lower story trees, dense bushes (e.g. blueberries) and fallen branches and dead trees;

o This area has cloud cover for some 60% of the year and the damp soil conditions are a near permanent feature of large areas of these forests. As summarised in Annex 3 of the Mondi-SYK Komi report, the complex biodiversity of the these forests require these wet conditions.

These factors significantly contribute to a limited dynamic range of the radar backscatter.

− In dry conditions and single species and single age class forest compartments as in South Africa (re-fer also to the ESA DUE II project), the saturation level is around 400 (m3/ha). The dry conditions, the forest homogeneity and the hard dry soil significantly contribute to a large dynamic range of the radar backscatter.

Figure 1 - Timber Volume (m3/ha) vs. L-band HV radar backscatter for Zululand (top) and Komi (bottom).




Metria – Sweden





Page 61


Figure 2 - Timber Volume (tons/ha) vs. L-band radar backscatter for boreal forest (Anna Peregon, JAXA K&C meeting 15, Tokyo, 2011).