HYDROLOGICAL ASPECTS OF THE CARBON BALANCE IN A BOREAL CATCHMENT: A MODEL STUDY .

Document nameDOCTORAL DISSERTATION Date of issue 2007-10-01

OrganizationLUND UNIVERSITY Geobiosphere Science CentreDep. of Physical Geography and Ecosystems Analysis,Sölvegatan 12, 223 62 Lund Sweden

Author(s)Yurova Alla Yu.

Sponsoring organizationNECC

Title and subtitle Hydrological aspects of the carbon balance in a boreal catchment: A model study. Meddelanden från Lunds Universitets Geogrfiska Institution. Avhandlingar 173

AbstractThe cycling of carbon in its gaseous, solid and dissolved forms is of central importance to ecosystemscience, particularly, when the natural carbon cycle has been perturbed by human activity. Mathematicalmodels are commonly used to asses and predict global and regional patterns of terrestrial carbon exchange.At the same time, processes operating on smaller (sub-grid) scales are usually outside the scope of the large-scale models exploring climate–biosphere interactions. However, if a sub-grid feature is found repeatedly in a large number of particular realizations, we may expect that its characteristic dynamics are likely to be important even at larger scales. This work examines some of these phenomena. First, a boreal landscape differentiated into forests and peatlands is investigated. Second, the export of dissolved organic carbon(DOC) in peatland streams is assessed. The work described herein mainly addresses methodology, i.e. how approaches to hydrology and ecosystemmodelling can be combined. Specifically, the general ecosystem model GUESS was modified and supplied with new modules to be used for peatland simulations. The model was also designed to reproduce soil moisture and related ecosystem properties along a topographical catena. In addition, a completely new model of DOC production, transport and transformation in peat was developed.The data for model development and validation came from boreal catchments in the Vindelns Forest Research Parks, The Swedish University of Agricultural Sciences, Northern Sweden. A relatively good fitbetween the modelled and measured NEE at the Degerö Stormyr peatland was demonstrated for 2001-2003.Further, the model was used to demonstrate that the NEE was quite insensitive to changes in water table level within the ranges typical for the dominant mire plant community. The newly constructed model was able to reproduce DOC concentrations measured at the outlet of the Kallkälsmyren peatland for 1993-2001.The model suggested that the main drivers for interannual variability in DOC concentrations were the rates of microbially mediated DOC production and mineralization and the flow intensity in the active surface layer of the peatland. The amount of DOC in the peat water at a particular time was mainly determined by(much larger) amounts of soluble carbon sorbed on the peat matrix. A dynamic, rather than static,description of the adsorption and desorption processes was advocated.The work makes some contribution to filling the gap that currently exists between field observations at the plot scale and large-scale model analysis of the climate–biosphere interaction.

Keywords : model, ecohydrology, boreal peatland, dissolved organic carbon

Classification system and/or index terms (if any)Supplementary bibliographical information Language

English

ISSN and key title

0346-6787

ISBN

978-91-85793-03-7

Recipient´s notes Number of pages 50

Price

Security classificationDistribution by (name and address) Alla Yurova, alla.yurova@nateko.lu.se

I, the undersigned, being the copyright owner of the abstract of the above-mentioned dissertation, hereby grantto all reference sources permission to publish and disseminate the abstract of the above-mentioneddissertation.

Signature: Date: 2007-10-01

Hydrological aspects of the carbon balance in a boreal catchment: A model study

Alla Yurova

An academic thesis in fulfillment of the degree of Doctor of Philosophy in Geobiosphere Science, discipline of Physical Geography and Ecosystem Analysis, at the Faculty of Science of Lund University, which will be publically defended onFriday 26th October 2007 at 10:15 in Världen, Geocentrum I, Sölvegatan 12, Lund

Faculty opponent: Dr. Per-Erik Jansson, The Royal Institute of Technology,Stockholm, Sweden

A doctoral thesis at a university in Sweden is produced either as a monograph or as a

collection of papers. In the latter case, the introductory part constitutes the formal thesis,

which summarizes the accompanying papers. These have either already been published

or are manuscripts at various stages (in press, submitted or in preparation).

Yurova, A. (2007). Hydrological aspects of the carbon balance in a boreal catchment: A model study. Scripta Academica Lundensia. Meddelanden från Lunds Universitets Geogrfiska Institution. Avhandlingar 173. 50 pp. Lund.

MEDDELANDEN FRÅN LUNDS UNIVERSITETS

GEOGRAFISKA INSTITUTION. AVHANDLINGAR

Avhandlingsnr. 173

ISBN: 978-91-85793-03-7

ISSN: 0346-6787

2007 Alla Yurova

Print: KFS AB, Lund, 2007

To my teachers

CONTENTS

List of papers

Acknowledgements

INTRODUCTION……………………………………………………………………...9

THE SUBJECT AND AIMS OF THE STUDY………………………...…………...14

METHODS…………………………………………………………………………….15

Model approach………………………………………………………………………15

Field and laboratory materials………………………………………………………..15

Modelling framework………………………………………………………………...19

Model formulation…………………………………………………………………19

Parameterization and validation…………………………………………………..24

Model experiments…………………………………………………………………24

Sensitivity and uncertainty issues………………………………………………….24

RESULTS…………………………………………………………………………..….26

Spatial variation in carbon storage along the catena. Dynamic aspects

of the soil moisture–soil C relationship (Paper I)……………………………………26

Mechanism of NEE regulation by WT level on a mire (Paper II)…………………...29

Hydrological and physico-chemical control of DOC production and

export from the mire (Papers III, IV)………………………………………………...29

CONCLUSIONS……………………………………………………………………....31

POSSIBLE APPLICATIONS AND FURTHER DEVELOPMENT………………32

APPENDICES……………………………………………………………………..…..35

REFERENCES………………………………………………………………………..42

POPULÄRVETENSKAPLIG SAMMANFATTNING……………….…………….48

Papers I-IV

List of doctoral thesis published in the present series

LIST OF PAPERS

This thesis is based on studies reported in the following papers, referred to by their

Roman numerals. Published papers are reprinted by kind permission from the copyright

holder

I. Yurova A., Lankreijer H. 2007. Carbon storage in the organic layers of boreal

forest soils under various moisture conditions: a model study for Northern

Sweden sites. Ecological Modelling, 204, 475-484

II. Yurova A., Wolf A., Sagerfors J., Nilsson M. 2007. Variations in Net

Ecosystem Exchange of Carbon Dioxide in a Boreal Mire: Modelling

Mechanisms Linked to Water Table Position. Journal of Geophysical

Research: Biogeosciences, 112, G02025, doi: 10.1029/2006JG000342

III. Yurova A., Buffam I., Bishop, K. A model of dissolved organic carbon

concentrations in peat and stream waters. I: Development and parameterization

(Submitted)

IV. Yurova A., Sirin A., Bishop K., Laudon, H. A model of dissolved organic

carbon concentrations in peat and stream waters. II: Application and the

sensitivity analysis (Submitted)

ACKNOWLEDGEMENTS

I greatly acknowledge generous financial support that came from NECC project and Lund University. Special thanks to Anders Lindroth and Harry Lankreijer for providing me with an excellent opportunity to be fully devoted to research during the last four years.

Coming to Sweden I was amazed how well organized, esthetic and cozy a scientific institution can be, and how easy it can be to solve all practical problems in a friendly atmosphere. Thanks for everyone who keep it running, not least to administration, librarians and technical staff at Lund University.

During these years I was much enriched by scientific communication with many people; to list few who influenced me most: Annett Wolf, Mats Nilsson, Kevin Bishop, Ben Smith, Oleg Chertov and Andrej Sirin. The great variety of things you told me was overwhelming, but it was always a pleasure to note how understandable and well-structured things get after I have discussed them with you. Thanks to all and each of you!

Special thanks to many people at the Department of Forest Ecology at Umeå University (and others involved into Krycklan activity) who have inspired me to work on various aspects of C modeling on a catchment scale and to look closer at peatland ecosystems. Your data and expertise formed the basis for this PhD thesis. Thanks to Jörgen, Ishi and Hjalmar for careful and almost immediate response to all my data queries and for sharing your knowledge of an area. Ishi is also greatly acknowledged for setting and running long-term laboratory peat incubations that were so needed for parameterization of our DOC model.

I thank Harry and Ben for providing me an opportunity to participate in teaching “Ecosystem modeling” course. I’ve learned a lot from being beside such experienced mentors as you are; and motivated and curious students have forced me to think more carefully about many issues.

Sees-editing Ltd is acknowledged for quick, thoughtful and efficient English correction of the manuscripts.

Thanks to my friends- Annett and Christian, Anna-Maria, Ejiang and Jin Chuan-with you I felt like at home. Thanks also to my friends in Moscow, especially to Natasha U. and Semionovs’ family-great that you know how to deal with me. Distant but so familiar voices, reminding ironically that most of the ecosystems are in fact managed by humans and the book from 1930s, which was finally found after some adventures in Moscow libraries, may still be not exactly the one I need- thank you for all help and support.

This thesis is dedicated to my teachers, whose deep knowledge, patience, enthusiasm, fundamental view on science and abilities to ruin all unjustified “common sense” thinking and yet to demonstrate elegant and integral theories instead have always been an encouraging example for me. It becomes really so much easier to think things through and to face difficult problems when one was trained by you, and finishing a PhD is just another occasion to say thanks.

ABBREVIATIONS

GUESS general ecosystem simulator [Smith et al., 2001]

LPJ Lund-Potsdam-Jena dynamic global vegetation model [Sitch et al., 2003]

ROMUL model of raw humus, moder and mull [Chertov et al., 2001]

MMWH mixed mire water and heat model [Granberg et al., 1999]

GLUE generalized likelihood uncertainty estimation [Beven and Binley, 1992]

REA representative elementary area

DOC dissolved organic carbon

SSOC sorbed soluble organic carbon

O layers organic (soil) layers

NEE net ecosystem exchange

PET potential evapotranspiration

NECC Nordic Center for Studies of Ecosystem Carbon exchange and its

interaction with the Climate system

IPCC Intergovernmental Panel of Climate Change

INTRODUCTION

The rapid increase in atmospheric CO2 concentration resulting from human activities

has raised concerns in the scientific community and society in general about possible

climate changes induced by altering the global carbon cycle. It has been shown that

even though anthropogenic CO2 emission is insignificant on the geological time-scale, it

will affect the Earth’s climate for hundreds of years [e.g., Manabe, 1967; IPCC reports

1996, 2001]. At the same time, our current understanding of the climate system is

limited and it is not possible to predict future climate with any certainty. One remaining

scientific challenge has its origin in the nature of greenhouse gases that affect the

Earth’s energy balance. Water vapour, carbon dioxide and methane molecules not only

obey physical and chemical laws, but also cycle continuously through living organisms

and the products of their activity and decay. The complexity of the “sphere of life” – the

biosphere [Vernadsky, 1926] – provides numerous possibilities for feedbacks between

its components, biogeochemical cycles and climate. An understanding of the numerous

links between different ecosystem components, as well as reliable quantitative estimates

of carbon cycling in the biosphere, are needed to improve prediction of climate system

behaviour with increasing CO2.

Climate can affect an ecosystem by altering (i) radiative balance, (ii) heat balance and

therefore temperature regime and (iii) water balance. The focus of this work is on the

hydrological aspects of ecosystem functioning. Water provides environment for all

organisms that inhabit the oceans, freshwater ecosystems and soil and sediment

solutions. CO2, CH4 and various organic molecules, including humic substances,

dissolve in it. Water also plays a critical role, from the molecular to the biome level, in

terrestrial ecosystems. Soil moisture has a significant effect on vegetation composition,

growth and disturbance regimes. Direct coupling between photosynthesis and

transpiration in plants suggests a close link between the carbon and water balances. The

process of decomposition and further transformation of soil organic matter is largely

dependent on the soil moisture [e.g., Davidson et al., 2006; Carrasco et al., 2006].

Water provides specific conditions for certain groups of microorganisms, both aerobic

and anaerobic. Water content in the soil therefore determines the dominant ecosystem

type, forest or wetland, in an area. Wetlands are of particular concern in the context of

greenhouse gas balance because of specific C exchange. During the Holocene they have

constantly sequestered C in peat, but, at the same time, they emit CH4 and may, under

certain conditions, emit large quantities of CO2 [e.g., Moore and Knowles, 1987;

Gorham, 1991]. Water and carbon cycling in wetlands are closely interconnected [e.g.,

Ivanov, 1981; Hilbert et al., 2000]. While the net peat accumulation is controlled by the

balance between the rates of net C uptake and release, small changes in either of these

processes caused by changes in the water balance may substantially alter the rate of net

peat accumulation.

9

Besides shaping the landscape, water in the soil acts as a conduit between terrestrial and

aquatic ecosystems. Dissolved organic carbon (DOC) is formed as a product of the

decomposition and transformation of organic matter in soils and peat. DOC is

transported in moving water from the land to aquatic ecosystems, where it has many

important functions; including providing an energy source for heterotrophic

microorganisms [e.g., McKnight et al., 1998; Steinberg et al., 2006]. Thus, C originally

taken up by plants may be either deposited in ocean and freshwater sediments [e.g.,

Dean and Gorham, 1998] or mineralised and emitted back to the atmosphere in

significant amounts [e.g., Cole et al., 1994; Cole et al., 2007].

If the climate changes, it will affect water balance, and therefore certain ecosystem

processes. The responses of ecosystems may in turn affect climate, and could involve

both physical and biochemical aspects. Surface reflectance, latent and sensible heat

fluxes, local atmospheric instability, cloudiness, and precipitation can be altered due to

changes in plant physiology, stand level canopy structure, phenology and vegetation

composition [e.g., Charney, 1975; Shukla and Mintz, 1982; Bonan et al., 1992; Bonan

2002]. Biochemical effects may caused by changes in photosynthetic CO2 uptake

affected by soil moisture and changes in the rate of soil respiration. The latter is likely if

relatively liable C, which is currently protected from decomposition by high soil

moisture, is exposed to drier conditions [e.g., Carrasco et al., 2006]; alternatively,

drought can inhibit decomposition in water-limited regions [e.g., Schaphoff et al.,

2005]. Potential peatland feedback mechanisms deserve special consideration, because

of the possibility of autogenic non-reversible lateral peatland expansion [e.g., Ivanov,

1981; Inisheva et al., 2006] and abrupt changes in C balance that these ecosystems may

exhibit if the water balance is altered [e.g., Hilbert et al., 2000]. Changes in the amount

of DOC transported from soils and peat can influence DOC exports to the ocean

[Freeman et al., 2001b] and CO2 emissions from rivers and lakes [e.g., Sobek, 2005].

Our knowledge of the structure and functioning of ecosystems can be formalized in

mathematical equations, thus providing opportunities to study interconnected and non-

linear internal dynamics of the biosphere and its responses to external drivers. Today,

terrestrial vegetation models are widely used in ecosystem studies at various scales from

individual plots to the entire globe. The current generation of dynamic vegetation

models has the ability to capture the effects of environmental drivers on plant

physiology and biochemistry, vegetation composition and soil organic matter dynamics

[e.g., Cramer et al., 2001; Sykes et al., 2001]. Global versions of dynamic vegetation

models are now being coupled to general atmospheric circulation models and used for

improving future climate predictions [e.g., Foley et al., 1998; Cox et al., 2000]. Results

produced using such coupled models have stimulated a growing interest in the impact of

the biosphere on the Earth’s climate. However, those results are not without controversy

[e.g., Moorcroft, 2006]. Future changes predicted by biosphere–atmosphere modelling,

such as the disappearance of the Amazonian forests [Cox et al., 2000], may seem too

10

drastic. The intuitive feeling is that ecosystems are more adaptive than the models

sometimes suggest. Whether the soil C pool will be influenced by future climate change

to such an extent and in such a manner that models predict, is also open to debate [e.g.,

Agren and Hyvonen, 2003; Davidson and Janssens, 2006; Carrasco et al., 2006;

Moorcroft, 2006]. One way to test the validity of any model is to investigate how well it

describes the current situation. However global models are very generalized, while the

landscape we observe in reality is much more diverse. It is not an easy or

straightforward task to choose an “average” or “standard” ecosystem from the wide

variety of ecosystem types. In addition, valuable information about the ecosystem’s

flexibility can be derived by studying the adaptations of vegetation and soils to the local

climatic, hydrological or geochemical factors acting at scales much smaller than the

model resolution. Downscaling of models is, therefore, one way to check a model’s

value and is necessary for better understanding and prediction of the biosphere–

atmosphere interaction.

Models at sub-grid scales are valuable on their own, and can be used as a tool for

studying particular ecosystem types, e.g., forests and wetlands. However, it is also

important to examine how information derived from such models can be assimilated

into those representing a larger-scale. Many variants are possible, and the simplest

approach is to divide the large-scale model cell into portions dominated by a particular

ecosystem type. Another approach is to try to use a representative elementary area

[REA, Wood et al., 1988]. It has been suggested that at the scale of an REA, the

variability between different catchments is minimised and it is statistical distribution,

not actual spatial patterns, of soil characteristics within the REA that determines runoff

and, potentially, other associated parameters (soil C storage, DOC fluxes). At smaller

scales, variability is much greater and model results are strongly dependent on the

choice of scale. At the REA scale, the catchment area is large enough so that all

important characteristics are sufficiently sampled to represent the mixture of landscape

units. At larger scales, variability may become important due to differences in geology,

geomorphology and other large-scale factors.

The focus of this work was on various effects that hydrological processes may have on

C cycling, and landscape differentiation as a result of soil moisture was of interest.

Some approaches to the coupled modelling of C and water balances were considered as

potentially useful for this study (Table 1).

11

Table 1. Models of carbon dynamics at different scales, noting how hydrological

processes are incorporated

Can models simulate…? Model type How hydrology is

modelled Veget.

C

Soil

C

Wetland

C

DOC

No.

of

puba.

Global Phenomenological eq. for infiltration and runoff

Yes Yes Yesb No 740

Regional 1D solution of Richards or another eq. for vertical flow, lateral flow parameterized or not included

Yes Yes Yesc ? 81

Catchment various, from complete 3D solution of flow eq. to lumped phenomenological runoff eq., many combinations of intermediate complexity possible

Yes ? ? Yes 16

Catena Various, mainly analytical or numerical solutions of kinematic wave or other 1D hillslope flow eq.

Yes ? No Yes 3

Site 1D solution of Richards or another eq. for vertical flow, lateral flow parameterized or not included

Yes Yes Yes ? 858

“?” in the table indicates that the author of this thesis is not aware of models that include the specified

component. The possibility that someone has undertaken such a study is, however, not excluded

a: 1969-2007, ISI Web of KnowledgeSM search from 16/08/07

b: on-going work Wania et al. (LPJ model), de Noblet et al. (ORCHIDEE model)

c: on-going work Roulet et al. (Canadian Land Surface Scheme)

Due to their development history and the scale at which they operate, models of

different types found in the literature have different ways of describing hydrological

processes (Table 1). Many models based on 1D differential equations of water flow are

designed to be used at the site scale, and direct measurements of hydrological variables

can be easily obtained for validation. In contrast, many models operating at a large scale

utilize rather generalized phenomenological expressions to simulate an area-average

runoff and soil moisture. However, it can be seen (Table 1) that there is a gap between 12

the field scale (m) and the regional-to-global grid cells (typically more than 50 50 km).

The scale least represented in the modern literature devoted to ecosystem modelling is

small or medium catchments (from 1km2). Yet, this scale is one of the most interesting

from many perspectives. In the boreal region in particular, a number of factors support

this contention:

Much of the spatial variability in soil carbon storage can be found within small

catchments. The gradient from dry sites, usually with poor O layers (typical mor),

to wet sites on fully developed humic podzol (rich O layers, hydromor) or peat

soil, is a characteristic, repeated feature of such catchments

Peatlands often occupy a significant proportion of a catchment, saturated due to

the high water table position. Peatlands contain the greatest amounts of carbon per

unit area of all the terrestrial ecosystems, and as much as 20-30% of the global soil

C is currently stored in the boreal peatlands [Schlesinger, 1997]. Although the

overall likelihood of peatland formation and growth in the region is a factor of

climate, the presence of peatland at a certain location is largely determined by the

configuration of a particular catchment [Ivanov, 1981]

DOC fluxes are strongly controlled by water routing at this scale. Export of DOC

is likely to occur mainly from the areas where flowing water can bypass mineral

soils, i.e. surface and subsurface flow through riparian zones and peatlands [e.g.,

McKnight et al., 1998; Qualls, 2000; Bishop et al., 2004].

The size of the representative elementary area (REA) in till soils in Northern

Europe is 1 to 10 km2. [Beldring et al., 1999; Temnerud et al., 2007].

It is therefore important to (i) look at specific landscape units, including the peatlands,

and (ii) try to find methods for quantifying moisture-related gradients for ecosystem

properties, starting with small experimental catchments (1-2km2). Such studies will

foster understanding and will help with the parameterization of processes that are of

great importance at much larger scales.

13

THE SUBJECT AND AIMS OF THE STUDY

This PhD thesis was supported by NECC (Nordic Center for Studies of Ecosystem

Carbon exchange and its interaction with the Climate system), with the goal of

investigating the features characteristic of Nordic ecosystems and integrating the data

collected into regional C balance assessments. The PhD research involved a numbers of

steps aimed at constructing a process-based model, which would:

describe the C balance of a small boreal catchment

do this explicitly, so that each main terrestrial ecosystem type, including peatland,

is simulated separately

use a dynamic hydrological approach allowing moisture accounting and solute

(DOC) tracing

The following research questions were addressed as model development progressed:

How does soil C storage in organic layers differ with respect to local soil moisture

conditions within a landscape? Are there new explanations for this spatial

variation? (Paper I)

How is net ecosystem CO2 exchange in a Sphagnum-dominated peatland related

to water table position? (Paper II)

How large and how variable is the export of DOC from a peatland? Which factors

determine temporal variability in DOC fluxes at different time scales? (Paper IV)

Three of the four papers included in this thesis (Papers II, III, IV) were devoted to a

particular ecosystem type, namely peatland. In addition, as presented in Paper I and

partly here in the thesis summary, an attempt was made to consider the full range of

ecosystem types along the soil moisture gradient.

14

METHODS

Model approach

Models were the main tools of this study; they were modified and developed, as

follows:

The dynamic vegetation model LPJ-GUESS was coupled to a dynamic soil

organic matter model, ROMUL (Paper I)

The combined GUESS-ROMUL model was linked to a simple hydrological

scheme allowing simulation of vegetation and soil dynamics along a forested

catena (Paper I, thesis)

Modifications were made to the model in order to quantify the CO2 flux in a

Sphagnum-dominated peatland. The MMWH model was then used to simulate

hydrology and the peat thermal regime (Paper II)

A completely new model of DOC concentration and export from a peatland was

developed (Papers III and IV).

Field and laboratory materials

Although there was no geographical reference in the model formulation (the scheme

was kept rather general), this study was closely linked to a few geographical locations

situated in the Vindeln Forest Research Park, Northern Sweden. Field data used in the

thesis for model development and validation (summary in Table 2, p.18) were kindly

provided by the Department of Forest Ecology at SLU (Swedish University of

Agricultural Sciences) in Umeå.

Study area

The Vindelns Forest Research Park is situated 60 km northwest of Umeå, Northern

Sweden (Figure 1). There is a long history of geomorphological, hydrological, botanical

and ecological studies in an area, and mires have always been of a particular interest.

The area is located in the boreal zone with a cold temperate humid climate. The

reference mean annual temperature, measured at the nearby Kulbäcksliden

meteorological station, is +1.2oC [Alexandersson et al., 1991] and the mean annual

precipitation is 523 mm, 34% of which falls as snow [Alexandersson et al., 1991]. The

spring flood is the dominant hydrological event of the year, accounting for up to 60% of

the annual stream discharge [Bishop et al., 1995]. Precipitation and runoff have been

higher than normal in a last decade [Lindström and Alexandersson, 2004]. The

underlying geology is blocky glacial till from the quaternary; the landscape includes

moraine features with hills and former melt water beds. Soil types are iron-podzol in dry

upslope areas, iron humic podzol in wetter locations and humic podzol with a distinct

peat horizon in areas with high ground water levels. The most common tree species are

15

Scots pine (Pinus sylvestris) in dry areas and Norway spruce (Picea abies) in wetter

areas. The most extensive mire type is aapa (a mire complex differentiated into

oligotrophic and minerotrophic sections). Mires are rather common and can be found on

gentle slopes, in drained and undrained topographical depressions and in riparian zones

of streams and rivers. The peat largely consists of Sphagnum remains, but sedge

remains are also commonly found.

Two experimental catchments were studied: the drainage basin of the Degerö Stormyr

peatland (Figure 1b) and the Nyänget catchment that includes the Kallkälsmyren

peatland (Figure 1c). A topographical transect selected to represent soil C storage in

organic layers along the catena is situated within the Nyänget catchment. This was

drawn on a national topographical map (Lantmäteriet, 2002), across the Kallkälsmyren

peatland, perpendicular to the stream (Figure 1c). Transect crosses the main landscape

elements: a dry upland forest, a peatland, mesic-moist and mesic forests on a slope and

a stream riparian zone. Field data collected from two peatlands (Degerö Stormyr and

Kallkälsmyren) have been used for model development and validation.

Degerö Stormyr (64 11 N, 19 33 E) is a large (6.5km2) oligotrophic–minerotrophic

mire complex. Peat depth is mostly between 3 and 4 m, with maximum depths up to 8 m

[Malmström, 1923]. The hydrology of the system is rather complex, with ground water

being routed into pipe flow and springs [Malmström, 1923]. The mire has distinct

microlandscape differentiation and only the lawn plant community was studied here.

The water table position usually varies from 5 to 19 cm below the peat surface (90%

percentile-10% percentile), but flooding occurs at every snowmelt while the peat is still

frozen. Species dominating the lawn plant community are Sphagnum balticum Russ. C.

Jens. with sparse Sphagnum lindbergii. The vascular plant community is dominated by

Eriophorum vaginatum L., Vaccinium oxycoccos L., Andromeda polifolia L., with

Scheuchzeria palustris L occurring more sparsely.

Kallkälsmyren (64 14 N, 19 46 E) is a small (9ha) complex mire occupying a

topographic depression. The mire is 2 to 5m deep. It is drained by a single stream,

which originates on its southern edge. Flow in the surface peat layer is accompanied by

a minor, but not insignificant, ground water component [Sirin et al., 1998]. The main

section of the mire is raised, sloping towards the edges, and has distinct ombrotrophic

features, with vegetation composed of a sparse Scots pine stand with dwarf-shrubs and

Sphagnum mosses, with Carex on the edges. Mineratrophic fen vegetation occupies the

southern area of the mire adjacent to the stream and is found in two belts situated on the

eastern and western sides of the raised section.

16

64oN, 19oE

Figure 1. (a) Map of Sweden showing the location of the study area. (b) Map of the area near Vindeln, Västerbotten in Northern Sweden, showing the locations of the Degerö mire, the eddy tower measurement site, and the area of a plot (c). (c) Map showing the location of the Kallkällsmyren mire, the measurement site at the stream origin, and the location of transect AB used to simulate soil C storage along the catena. Peatland areas are shown in dark gray.* MarkInfo [2006] http://www-markinfo.sluPart of the dataset used in this publication was made available by the Swedish National Survey of Forest Soils and Vegetation performed by the Department of For est Soils, SLU. The author are solely responsible for the interpretation of data.

**Monthly averaged air temperature, precipitation and runoff measured at Svartberget meteo station in 1981-2001 [Ottosson-Löfvenius et al., 2003].

17

Tab

le 2

. Mea

sure

d da

ta

Dat

a ty

pe

Dat

abas

eS

ourc

e an

d co

ntac

ts

Var

iabl

esIn

stru

men

tati

on

Tim

e fr

ame

How

use

d in

the

m

odel

(P

aper

nr.

) M

eteo

rolo

gica

lC

RU

Air

tem

pera

ture

In

terp

olat

ed u

sing

all

met

. st

atio

ns i

n a

grid

0.5

0.5

1901

-199

8S

pinu

p (I

, II)

Pre

cipi

tati

on-

1901

-199

8S

pinu

p (I

, II)

Sch

ool

of

Env

iron

men

tal

Sci

ence

s U

nive

rsit

y of

Eas

t A

ngli

acr

u@ue

a.ac

.uk

Per

cent

age

of

full

sun

shin

e -

1901

-199

8S

pinu

p (I

, II)

Deg

erö

Air

tem

pera

ture

M

P10

0 te

mpe

ratu

re

sens

or

(Rot

roni

c A

G)

2001

-200

3In

put

(II)

Dep

t. o

f F

ores

t E

colo

gy,

SL

U,

Um

eå,

Sw

eden

Jorg

en.S

ager

fors

@se

k.sl

u.se

Mat

s.B

.Nil

sson

@se

k.sl

u.se

Pre

cipi

tati

on

tipp

ing-

buck

et

AR

G

100

(Cam

pbel

l S

cien

tifi

c)20

01-2

003

Inpu

t (I

I)

Glo

bal

radi

atio

nL

i 20

0 sz

( L

I C

OR

) 20

01-2

003

Inpu

t (I

I)

Soi

lte

mpe

ratu

re

term

isto

rs i

n a

seal

ed t

ube

(TO

3R,

TO

JO

skog

stek

nik,

)20

01-2

003

Val

idat

ion

Sva

rtbe

rget

Air

tem

pera

ture

fo

llow

ing

WM

O s

tand

ards

19

86-2

005

Inpu

t (I

V)

Pre

cipi

tati

on-

1986

-200

5In

put

(IV

)P

ET

si

te-a

djus

ted

calc

ulat

ion

usin

g P

enm

an-

met

hod.

19

86-2

005

Inpu

t (I

V)

Sva

rtbe

rget

F

ield

S

tati

on,

Vin

deln

E

xper

imen

tal

For

ests

SL

U, U

meå

m

ikae

ll.o

ttos

son.

lofv

eniu

s@se

k.sl

u.se

Sno

w d

epth

M

ean

valu

e of

5 g

auge

s 20

04, 2

005

Val

idat

ion

Hyd

rolo

gica

lD

eger

öas

abo

veW

ater

ta

ble

dept

hfl

oat

and

coun

terw

eigh

t sy

stem

att

ache

d to

a p

oten

tiom

eter

[R

oule

t, 19

91]

2001

-200

3V

alid

atio

n (I

I)

Kal

lsäl

smyr

enD

ept.

of

Eco

logy

and

Env

iron

men

tal

Sci

ence

s, U

meå

Uni

vers

ity,

Sw

eden

(i

n K

ryck

lan)

hj

alm

ar.l

audo

n@em

g.um

u.se

Str

eam

di

scha

rge

Hei

ght

by p

ress

ure

tran

sduc

er+

rati

ng

curv

es (

buck

et f

low

or

salt

dil

utio

n m

easu

rem

ents

).

1993

-200

1C

alib

rati

on (

IV)

Bio

chem

ical

Deg

erö

as a

bove

NE

Eed

dy c

ovar

ianc

e. 3

D s

onic

ane

mom

eter

10

12R

2 (S

olen

t,

Inst

rum

ents

)+cl

osed

-pat

h in

frar

ed

gas

anal

yzer

IR

GA

626

2 (L

I C

OR

)

Gil

l20

01-2

003

Val

idat

ion

(II)

Kal

lsäl

smyr

enas

abo

ve

(in

Kry

ckla

n)

DO

Cco

ncen

trat

ion

Doh

rman

n (1

993-

1994

), S

him

adzu

TO

C-

5000

(19

95-2

001)

19

93-2

001

Val

idat

ion

(IV

)

18

Measurements (Table 2)

Net ecosystem exchange was measured with an eddy-flux technique in a lawn section of

Degerö Stormyr peatland since 2001 [Sagerfors et al., 2007] (site location on Figure

1b). DOC concentrations and stream discharge were observed at the outlet of the

Kallkälsmyren peatland during the long-term monitoring program started in 1993

[Bishop and Laudon, unpublished data] (site location on Figure 1c). Unfortunately, an

integral assessment of both vertical (CO2) and horizontal (DOC) carbon fluxes for

Degerö Stormyran [Sagerfors et al., 2007] has been published only at a late stage of this

PhD work, when recently established measurements of DOC concentrations at this

peatland became available [Laudon et al., unpublished data]. Those data were not

included in the thesis.

Laboratory materials

To obtain estimates of microbial DOC production and mineralization rates in peat, long-

term laboratory peat incubations were designed (Paper III). Experiments were done in

cooperation with the laboratory of the Department of Forest Ecology at SLU.

Methodology and information related to the laboratory experiments can be found in

Appendix of Paper III.

Modelling framework

Model formulation

The model supporting this work can be presented in two variants: forest and peatland

(Figure 2). Coupled to the peatland version, but not integral to it, is a model of DOC

concentration in peat water; in the course of this study, this was both formulated and the

computer programme developed (Figure 3).

The GUESS model [Smith et al., 2001; Sitch et al., 2003] was used here as a framework

for ecosystem simulations. GUESS (also known in its global version as LPJ) is a model

that describes population dynamics, with the focus on plant functional types competition

and replacement. GUESS is based on gap theory and simulates individual plant growth

in a number of replicate patches, the patch average representing the forest stand. For

each patch, stochastic generic disturbance is applied. The model follows generally

accepted schemes to simulate photosynthesis [Farquhar et al., 1980; Halexeltine and

Prentice, 1996a] and plant respiration [Sprugel et al., 1995; Halexeltine and Prentice,

1996a]. Soil biochemistry in the model is somewhat simplified: three pools of soil

organic matter are specified and the turnover rates are fixed as a simple function of

temperature. Intermediate complexity is assumed for other important ecosystem

processes, such as plant carbon allocation and light interception [Sitch et al., 2003]. The

hydrology subroutine utilises an empirical infiltration model [Haxeltine and Prentice,

1996b] and runoff is defined as any excess over the field capacity of a soil. Coupling

19

between evapotranspiration and photosynthesis is explicitly modelled [e.g., Sitch et al.,

2003]. The LPJ-GUESS model develops dynamically and is valuable in global and

regional C balance assessments [e.g., Morales et al., 2007; Prentice et al., 2007; Zaehle

et al., 2007], on a national level, particularly for Sweden [Koca et al., 2006], in research

projects related to plant population dynamics in changing environmental conditions

[e.g., Lucht et al., 2002], in paleostudies and for investigating site-specific ecosystem

phenomena. Recently, [Smith et al., on-going work] the model has also been coupled to

the regional version of the Swedish climate model RCA [latest version. Kjellström et

al., 2005].

In this study the ROMUL model replaced the standard formulation of soil biochemistry

in GUESS based on a simple compartment model. ROMUL [Chertov et al., 2001] is

also a compartment model, but the compartments are specified using the concept of

“humus type”. The humus of organic and mineral soil horizons is distinguished. The

rate of mineralization and the flow of matter between different compartments are

described by kinetic coefficients. Each coefficient is specifically dependent on both the

organic matter quality (C/N ration) and the environmental conditions (temperature and

moisture). The coefficients used for litter and humified forest floor were derived from

long-term laboratory studies [Chertov, 1985], and the rates of organic matter turnover in

the mineral horizon were deduced from the data on the activity of specific groups of soil

organisms [Chertov and Komarov, 1997]. The model has been evaluated against

numerous long-term experimental observations and provides good results for real forest

data sets [e.g., Chertov et al., 1997]

Forest modification for transect simulations (Figure 2a)

In this set up (Paper I, thesis) GUESS-ROMUL was run for a number of sites along the

topographical transect. A steady-state ground water level along the transect was

estimated and scaling method was used to correct the surface moisture to correspond

with the ground water depth (for a description of the methodology, see below). Each

simulated site, therefore, had a specific moisture regime, ranging from dry to

waterlogged. If the simulated steady-state ground water level was at or above the land

surface, indicating permanently waterlogged conditions, the peatland modification of

the model was used. Other environmental conditions, such as air temperature and

incoming solar radiation were the same for all the simulated sites, which is of course a

simplification.

Data collected from till soils within the Nordic region have demonstrated that it is

appropriate (at least in a statistical sense) to estimate surface soil moisture from the

local depth to the ground water table [Beldring et al., 1999]. Following Beldring et al.

[1999], a modified Brooks and Corey equation [Brooks and Corey, 1966] was used in

this study to calculate the surface soil moisture at a given site, based on (i) ground water

level and (ii) surface soil moisture at a reference ground water level (derivation and

20

Precipitation

TemperatureLight availability

CO2

Geology,topography

Water availability

establ.:tree and grass PFT’s

NPP

Leavesandroots

decomp.litter

mortality

wood

Water table

CO2

soilOM

Soil texture

Precipitation

TemperatureLight availability

CO2

Water content

SphagnumNPP

decomp.litter

Watertable

CO2

peatTopography

DOC

sorption

SSOC

export

a: forest model

b: peatland model

Figure 2. Schematic diagram of the model structure . Model core is an ecosystem simulator GUESS . Shaded are extensions applied in this study: A-scheme for simulation of ground water depth and soil moisture for selected topographical transect; B-ROMUL soil organic matter dynamics model; C-Mixed Mire Water and Heat model, MMWH; D-Model of DOC concentration in peatland. Model abbreviations are explained in the text.

A

B

C

D

21

A gradient of the the DOC flux density

Zero-orderproduction, first-ordermineralization

Linear kinetics

How modelled

D0 ,P, 1

ks

Q10,, kanP , kan

KD , desParameters

HydrologicalMicrobialSorptionProcesses

Any, dependent on flow rate

Long/weeks-years

Short/hoursTime scale

MMWH model

Paper III. Implementation: numerical or analytical solution of CDE

Paper IV. Implementation: numericalsolution of CDE

Wat

er f

lux

T, water content

Figure 3. A schematic representation of formulation and structure of the peatland DOCmodel. Source: Paper III.

Modifier to account for the effect of oxygen supply on 1kan

Dispersivity

Molecular diffusion coefficient D0,,

Modifier to account for the effect of oxygen supply on PkanP

Modifier to account for the effect of temperature on P and 1Q10

The constant reducing the microbial mineralization rate when the soluble OC is sorbedks

The rate of the microbial mineralization of DOC at 200C1basal

The rate of microbial DOC production at 200CPbasal

Partitioning coefficient describing the sorption equilibriumKD

Desorption kinetic constant des

22

discussion are presented in Appendix C). The reference ground water depth

corresponded to the standard GUESS soil moisture simulations. Corrections then

decreased or increased soil moisture at locations where ground water differed from the

reference, but the daily variability in soil moisture remained the same as in the standard

simulation.

Depth to the ground water table was estimated using an equation describing the steady-

state ground water level in a topographical transect between two aquatic bodies. The

method was originally proposed by Ivanov [Ivanov and Shumkova, 1967; Ivanov, 1981]

as a rough tool for peatland delineation and for predicting changes in peatland area in

response to altered water regime. Here a derivation of this equation (eq. 12 in Paper I)

starting from the Boussinesq equation for the subsurface one-dimensional flow

[Boussinesq, 1904] is presented for two cases: horizontal and tilted bedrock (Appendix

B). Surface topography, depth to the bedrock, effective precipitation, hydraulic

conductivity and porosity are required for the calculations. Sample calculations for the

profile crossing the studied peatland, Kallkälsmyren, are included in this thesis. Bedrock

depth was taken from Sirin et al. [1998] and the hydraulic conductivity was chosen

from the normal range observed in Scandinavian tills [Lind and Lundin, 1990].

Peatland modification (Figure 2b)

To make GUESS-ROMUL applicable to peatland simulations, the MMWH model

[Granberg et al., 1999] was used to reconstruct soil thermal and water regimes.

Designed for boreal peatlands, the model provides a process-based description of heat

transfer and snow dynamics. The model also simulates daily fluctuations in water table

depth. An empirical, but justifiable, function [Weiss et al., 1998] was used here to

describe the steady-state moisture distribution in the unsaturated zone above the water

table.

Sphagnum moss was included as a new PFT within the model and equations that allow

moss photosynthesis rates to be controlled by variation in moss water content were

added [Williams and Flanagan, 1998]. Specific coefficients were used to describe

decomposition of the Sphagnum litter and peat and a simple parameterization of

decomposition under anoxic conditions was applied to a soil organic matter dynamics

module.

Model of DOC concentration in peatland water

The model was designed to reconstruct DOC concentration and fluxes in peatland

water. The starting point for model construction was a mass balance for dissolved

organic carbon in the form of convection–dispersion equation. In conjunction, a mass

balance equation was also written for the soluble organic carbon sorbed on a peat matrix

at a given time. Microbial DOC production and mineralization, sorptive exchange

between the solid and dissolved phases and hydrological transport with flowing water

are the main processes included in the model (Figure 3). Unlike previous DOC models

23

[e.g., Grieve, 1991; Boyer et al., 1996; Michalzik et al., 2003] the model proposed here

is based on a dynamic rather than a static description of sorption. The state variables

included in the simulation of DOC concentration, namely water fluxes (vertical and

lateral), water content in the unsaturated zone and peat temperature, were modelled

using the MMWH. The DOC model was applied to two situations: one to describe

stagnant water conditions in a laboratory peat incubation (Paper III), and another for the

field situation with flowing water in a peatland (Paper IV).

The derivations of two analytical solutions for the mass balance equations for dissolved

and sorbed soluble organic carbon in stagnant water conditions (eq. 5 and 6, Paper III)

are presented in Appendix A.

Parameterization and validation

Some of the DOC model parameters were derived from long-term laboratory peat

incubations that were conducted using peat collected from the Kallkälsmyren peatland

(Paper III).

All model variables that were validated (Papers II, IV) are included in Table 2, p.18.

Mean square error, mean bias error and the Willmott index of agreement [Willmott,

1982] were chosen to characterize the model performance.

Model experiments

In this work, model tests were widely used. In such cases, the model was not run to

represent a “real” situation, but in order to study a specified factor. For example, it was

investigated how the simulated response of NEE to water table depth in a mire depends

on the parameterization of (i) photosynthesis and (ii) heterotrophic respiration in

relation to water content (Paper II). The model was also used to partition the integral

fluxes into contributory components (photosynthesis and heterotrophic respiration;

terms contributing to DOC balance).

Sensitivity and uncertainty issues

All the papers in this thesis address the issue of model sensitivity and uncertainty, but

more advanced methodology is used in the later papers. In Paper I it was found that the

disturbance regime regulates long-term litter input into the soil and, therefore, affects

soil C storage. A single sensitivity test was conducted to demonstrate the dependency of

simulated soil C storage in O horizons on a specified fire return interval (Figure 2, Paper

I). Simulating a large number of patches, with disturbance occurring randomly, also

gave some estimate of the uncertainty associated with the mean C storage (Figure 1,

Paper I). In addition, the paper included an illustration (Figure 5, Paper I) of the

24

potential size of the uncertainty associated with the moisture response function for the

soil organic matter transformation rates close to and above water saturation.

In Paper II global sensitivity analysis [e.g., Saltelli, 2000] was examined using Monte-

Carlo analysis for all model parameters introduced or modified for the peatland

simulations (Table 2, Paper II). It was both proposed and proved that the studied

variable, NEE, is characterized by a non-linear response to parameters (Table 5, Paper

II). The variance-based Sobol sensitivity indices were therefore chosen as a measure of

the parameters’ importance [Sobol’, 1990; Saltelli, 2000]. One important result of the

sensitivity analysis was to show, through parameterizations, the key role that

autotrophic respiration plays in determining NEE.

In Paper III the kinetics of DOC sorption were studied by differential sensitivity

analysis [e.g., Saltelli, 2000]. The partial derivatives (local sensitivity coefficients) of

the DOC concentration with respect to the model parameters were expressed from the

original model equations. Normalized sensitivity coefficients were then plotted and

compared on the basis of hourly time intervals (Figure 4, Paper III). It was then possible

to separate two phases in the laboratory peat incubation: one ruled mostly by sorption

kinetics (the kinetic sorption coefficient has the highest effect on model results) and

another when microbial processes played the dominant role, though modulated by the

steady-state sorption partitioning.

In Paper IV the GLUE methodology [Beven and Binley, 1992] was used to estimate

uncertainty in DOC simulations associated with uncertainties in parameter values.

Outcomes of the Monte Carlo simulations were compared to the observed DOC

concentrations and a Willmott index calculated for each run as a measure of model

performance with a particular set of parameters. The prediction bounds (upper and

lower 10% quantiles) were then estimated for a daily time interval from the distribution

of all model outputs weighted by the Willmott index of agreement (Figure 3, Paper IV).

Global sensitivity analysis was also conducted, as described in Paper IV (Table 2, Paper

IV).

25

RESULTS

The main goal of this thesis was to explore different aspects of the interactions between

hydrology and carbon dynamics in boreal catchments. To achieve this goal a general

ecosystem model, GUESS, was linked to a dynamic soil organic matter model,

ROMUL, and modified so that new applications of the model to wet and water-logged

conditions were possible. A novel approach to modelling DOC concentrations was also

introduced and developed into an original model of DOC in peatland water. One

important result of this work was that relatively good model performance was

demonstrated both for NEE simulated at Degerö Stormyr mire for 2001-2003 (Paper II,

Table 3, Figure 2) and for DOC concentration at the outlet of Kallkälsmyren mire for

1993-2001 (Paper IV, Table 3, Figure 3). More specific results are described below.

Spatial variation in carbon storage along the catena. Dynamic aspects of the soil

moisture–soil C relationship (Paper I)

In Paper I the GUESS-ROMUL model was run for two contrasting soil moisture

conditions (mesic and mesic-to-wet). Some unpublished results from the model runs for

a number of sites (from dry to waterlogged) along the chosen topographical transect are

presented here in Figure 4.

1. Soil C storage in organic layers (litter, humified litter and peat) differs notably

depending on local soil moisture conditions. In areas where the water table

comes close to the surface (WT depth<1m), the GUESS-ROMUL model

predicts significantly greater O layer accumulation than in mesic and dry areas

(WT depth>1m) (Figure 4). Considering the high proportion of wet areas in the

boreal zone, simulations for “well drained” or average soil moisture conditions

may not provide the optimal estimate of soil C storage

2. Both organic matter decomposition rate and fire disturbance frequency

contribute to the difference in soil C storage in organic horizons between sites

with different moisture conditions (Figure 4). The model shows that, with a

natural fire regime, as much as 40% of the difference between mesic and

mesic-to-wet sites can be explained by the difference in litter input related to

fire frequency. Drier forests burn more often, and due to frequent disturbance

they are generally less productive because the age distribution shifts to

younger, less productive classes.

3. It is important to study, both under laboratory conditions and in the field, how

water content affects decomposers and their ability to utilize and transform

different fractions of litter and soil organic matter. The subject is

underrepresented in the current literature.

26

300 400 500 600 700 800 900 100010

20

30

40

50

Distance, m

Ele

vati

on, m

2

4

6

8

C i

n O

lay

ers,

kgC

m-2

100 300 5003

4

5

6

7

C i

n O

lay

ers,

kgC

m-2

Fire return interval, years

Dry Peatland Mesic Mesic- Mesic moist

Site class Fire return * interval____________________

Mesic-moist 300 years

Mesic 100 years

Dry 50 years

140 a

c

water table depth,sim.

b

peatland border, map

Figure 4. (a) Simulated soil C storage in organic horizons for sites with various moisture conditions along the catena illustrated in (b); (b) steady-state estimates of average ground water elevations; (c) relationship between fire-return intervals and C storage in O soil layers at the mesic site. The lines of boxes in plots (a) and (c) indicate the lower, median, and upper quartiles, and the whiskers show the lowest and highest values. The location of topographical transect AB is shown in Figure 1c. Mire borders taken from the map are shown in (b), and calculated locations of points where the water table rises above the land surface roughly match them. *Engelmark [1987]. Source: (a), (b)-unpuplished model results, (c)-Paper I.

27

-30 -20 -10 0

-2

-1

0

1

CO

2 fl

ux,

gCm

-2da

y-1

-30 -20 -10 0

-2

-1

0

1

-30 -20 -10 0

-2

-1

0

1

-30 -20 -10 0

-2

-1

0

1

WT depth, cm

stand modeltest1(psn)test2(resp)

obsmodel

a b

dc

Figure 5. (a) Relationship between simulated daily photosynthesis rateand water table depth for the lawn at Degero Stormyr. (b) Relationship between simulated daily heterotrophic respiration and water table depth (c)-(d) Relationship between NEE and water table depth (c) Trend lines (best-fits by second-order polynomial regression) basedon measured data (solid) and standard model simulations (dashed);(d) Trend lines based on the data from the standard model simulation (bold solid), test run 1 (thin solid with crosses), and test run 2 (dash-dotted). Test run 1: no effect of water table depth on photosynthesis. Test run 2: decay constant, corresponding to anoxic conditions, regardless of the water table depth. Source: (a), (b)-unpublished model results, (c),(d)-Paper II.

psn model resp model

28

Mechanism of NEE regulation by WT level on a mire (Paper II)

In Paper II, net ecosystem exchange at Degerö Stormyr was modelled using a modified

version of GUESS-ROMUL. The relationship between NEE and water table depth was

investigated

1. Model simulations showed that any changes in the water balance that do not

cause the water table to shift outside the normal range (10-20 cm) are likely to

have only minor effects on the annual rates of land–atmosphere CO2 exchange

in the studied peatland

2. When the water table level was either higher (> - 10 cm) or lower (< - 20 cm)

than normal the mire net CO2 uptake decreased (Figure 5c)

3. According to model simulations the reductions in net CO2 uptake rates that

occurred when the water table level was higher than -10 cm are attributable to

reductions in photosynthetic rate (Figure 5a, d)

4. The reductions in net CO2 uptake when the lower water table level was lower

than -20 cm resulted from increases in heterotrophic respiration, rather than

reductions in photosynthesis (Figure 5b, d).

Hydrological and physico-chemical control of DOC production and export from

the mire (Papers III, IV)

Paper III introduces the concept of modelling DOC concentrations with a convection–

dispersion equation. Some model parameters were derived on the basis of long-term

laboratory peat incubations and model behaviour for stagnant water conditions was

studied. In Paper IV DOC concentrations and fluxes at the outlet of Kallkälsmyren were

simulated using a newly constructed model.

1. Soluble organic carbon is present in the mire system in two states: dissolved

(DOC) and sorbed, potentially soluble, but solid at a time (SSOC). SSOC

contributes a much higher proportion to the total store than does DOC (the

latter accounts for no more then 20%).

2. Sorption is one of the major processes determining DOC concentration and

therefore fluxes. Externally induced changes in the DOC concentration in the

peat pore water may be counteracted by sorption or desorption. Release of

DOC from the SSOC store occurs if DOC concentration is decreased by

dilution, for example, if precipitation water is added. DOC removal may occur

if the DOC concentration is increased, for example, if DOC-rich water from

the surface peat is advected into the layers below.

3. The balance between the two phases (DOC and SSOC) is time-dependent, and

adsorption and desorption have to be described as dynamic rather than static.

29

1993 1995 1997 1999 200110

20

30

40

50

60

70

80

SS

OC

&D

OC

sto

red,

gm

-2 D

OC

, mgL

-1

1993 1995 1997 1999 20010

10

20

30

40

50

60

year

SS

OC

&D

OC

sto

red,

gm

-2

SSOC&DOC stored previous year, modelvol.weighted stream DOC, modelvol.weighted stream DOC, obs.

SSOC&DOC stored current year, model

net production, model

stream export, model

a

b

Figure 6. (a) Interannual variations in the amounts of SSOC and DOC stored before the start of the hydrological year (spring flood) simulatedby the model. Interannual variation in the volume-weighted annual stream DOC concentration, simulated by the model and calculated fromthe interpolated observations. (b) Simulated components of the annual DOC-SSOC balance: microbially-mediated net production and DOC stream export. Changes from year to year in the SSOC amounts stored in the acrotelm are also shown (estimated for the dates before the first autumn frost).

30

4. Interannual variation in the average volume-weighted DOC concentration

generally resembles that of the SSOC and DOC store within the acrotelm, as

reproduced by the model (Figure 6a).

5. According to the model, some “memory” is characteristic of the system: the

store in the preceding year affects the concentration and the fluxes in the

following year (Figure 6a). The same is true when the seasons are considered

separately. The correlation between the simulated volume-weighted stream

DOC concentration in a particular season and the store of DOC and SSOC in

the preceding season is high for the winter (r=0.75, P=5.8 10-5) and summer

(r=0.87, P=1.9 10-5), but lower for the spring (r=0.40, P=0.023)

6. Interannual variability in the SSOC and DOC stores within the acrotelm, in

turn, depends on the conditions for microbially mediated DOC production and

mineralization during the current year (temperature, aeration) (Figure 6b: net

production) and flow intensity (Figure 6b: stream export).

7. It is important to include vertical peatland inhomogeneity in the model to give

an adequate description of the processes that are either strongly dependent on

depth (production, horizontal transport) or driven by vertical inhomogeneity

(sorption, dispersion)

CONCLUSIONS

The following conclusions can be drawn from this work:

Patterns of soil C storage in a boreal landscape resulting from differences in soil

moisture are advantageous objects to model. Repeated and spatially predictable,

such patterns are helpful in a broader understanding of ecosystem dynamics and

functioning; particularly so, if we expect water balance to change in response to

altered precipitation patterns.

Peatlands that occupy extreme habitats and exhibit specific self-regulating

mechanisms can be described using the same simulation framework as other

ecosystem types. It is important, however, to incorporate an adequate model of

peatland hydrology. With general models such as GUESS, that were not originally

designed for peatland simulations, predictions for one vegetation type are feasible,

but more work is required to parameterize competition between plant types.

Wherever peatlands are present and drained by surface flow, a substantial flux of

DOC always accompanies the water flow, but seasonal and annual variations in

DOC concentration can be large. According to our model, the amount of soluble

organic carbon sorbed on a peat matrix strongly influences the annual volume-

weighted DOC concentration at a peatland outlet. The rate of microbial organic

matter transformation determines net production of soluble C, and sorption

partitions it between the solid and liquid phases. Most of the soluble C is stored in

a sorbed state and a part of it is released into flowing water when the sorption

equilibrium is shifted.

31

POSSIBLE APPLICATIONS AND FURTHER DEVELOPMENT.

The model developed in this study could be applied as a diagnostic and predictive tool.

Potentially, it can also offer an approach to the parameterization of smaller scale

processes in C balance models to be applied at larger scales. A few possible model

applications are presented below.

Diagnostic applications

It has been proposed [H. Grip, personal communication] that the approach used here to

reconstruct the depth to the water table along the selected topographical profile [Ivanov,

1981] could be applied to a study of ecosystem development in systems of newly

emerging islands. Specifically, this approach provides a rough estimate of the steady

state distribution of peatlands, defined as locations where the calculated water table is

situated above ground level. It is, therefore, possible to examine whether newly formed

landscape features gradually develop towards this theoretical steady state and how fast

this happens. Even more interesting could be to run the whole GUESS-ROMUL model

to see if it can reconstruct vegetation and soil dynamics over longer time intervals.

Peatland modification of the model can be used to study the relationship between NEE

and the depth to the water table for other peatland microtopographical features, in

addition to the lawn type studied here, but only under the assumption of unchanging

plant species composition.

New tests could be conducted using the DOC model to see how changes in peat sorption

properties affect DOC concentration [K. Bishop, personal communication]. Additional

development is needed, however, to introduce the dependency of kinetic sorption

coefficients on chemical characteristics of the soil solution, such as ionic strength [e.g.,

Skyllberg and Magnusson, 1995] and SO42- concentration [e.g., Clark et al., 2006].

Predictive applications

Some estimates of the effect of increased precipitation on the soil C storage in the

organic horizons were presented in Paper I. To evaluate the net change in the soil C

store more sites and climate scenarios need to be included in the analysis. The influence

of moisture conditions on ecosystem functions may be as important as that of

temperature. Although there is great uncertainty associated with predicting precipitation

using climate models [e.g., Palmer et al., 2005], it is very important to further evaluate

the role of changed soil moisture conditions on the soil C store. Such a study could also

consider the effect of precipitation change on DOC fluxes.

The peatland modification of the model in its current version does not describe all

aspects of plant competition, and long-term changes in NEE related to shifts in species

32

composition cannot be predicted by the model. It is, however, possible to run the model

for various climate scenarios and see whether the present vegetation types are still stable

under new conditions, i.e. if NEE stays within its normal range in the specific plant

community [M. Nilsson, personal communication].

The DOC model can be used to predict the effects of climate and human activity on

DOC fluxes from peatlands. The model could also be used to test some of the widely

discussed hypotheses such as the “enzymatic latch” mechanism, i.e. the increase in

anaerobic peat decomposition resulting from a temperature increase [Freeman et al.,

2001a, 2001b; questioned by Tranvik and Jansson, 2002], to eliminate the role of

resistance time, intensity and pathways of water flow.

Scaling

The question is sometimes posed: “Are big basins just the sum of small catchments?”

[e.g., Shaman et al., 2004]. The question is not only relevant to runoff modelling, but

also to all aspects of C balance modelling where local water flow and surface moisture

patterns become important (identification of wetlands and wet areas emitting methane,

modelling carbon accumulation in wet soils, studying cycling and loss of DOC etc). It is

important to know whether small basins already contain most of the variability in the

properties studied, and thus whether statistical information about the distribution of

these properties on smaller scales is enough to extrapolate to larger scale assessments.

The existence of a representative elementary area (REA) has yet to be theoretically

proven and may not be possible [e.g., Fan and Bras, 1995]. Intuitively, however, it is

clear that patterns of soil C storage in a boreal catchment, regulated by soil moisture, are

spatially predictable and repetitious. The spatial variability in DOC concentration in

aquatic bodies, which is a reflection of water routing through the patterns of organic

soils [e.g., Bishop et al., 2004], could be expected to have a characteristic pattern too.

Some information about these features at the catchment scale could, therefore, improve

regional and global C predictions. Descriptions of a large-scale picture can be built on

generalizations, rather than neglecting small-scale complexity. Potentially, there are two

alternatives, presented below, in which the model developed in this study could be

placed in the broader context and applied to scaling issues

Alternative I. Explicit catchment hydrology model

In this option the GUESS-ROMUL model could be coupled to the catchment hydrology

model. Important criteria for model choice are: explicit moisture accounting in the

unsaturated zone; water table fluctuations; the interaction between the saturated and

unsaturated zones; and explicit solutions of the flow equation in the modelling domain

(the latter is important for DOC predictions). The TOPOG model [e.g., Vertessy et al.,

1993; Zhang et al., 1999] may be a good option here. It uses an analytical solution of

the Richards equation to describe vertical flow in the unsaturated zone and an explicit

33

kinematic wave solution for the lateral fluxes in the saturated zone. The TOPOG-

Dynamic model [Silberstein et al., 1999] can also simulate plant growth. In addition, a

package that solves the convection–dispersion equation is available, and can be used to

simulate DOC concentrations. One advantage of this approach is that it is possible to

describe explicitly both flow and surface soil moisture within a catchment at a relatively

good spatial resolution. Therefore, straightforward validation of the model is possible

where data is available. One disadvantage, however, may be that a lot of computer time

is needed to run such a model. It is doubtful whether on-line nesting of the catchment-

scale model into the larger-scale C balance models is practically possible.

Alternative II. Statistical generalization of the catena approach

In a statistical sense it is valid to relate the surface soil moisture to the depth to the

ground water [Beldring et al., 1999]. If the area is drained by a series of parallel streams

then the approach used here to estimate, roughly, the steady-state ground water depth is

potentially applicable. It is then possible to collect sufficient information about patches

with various water table depths by sampling a number of transects perpendicular to the

streams in a catchment. Ensemble averaging or other statistical procedures could be

used to generate a catena that is not a real transect, but rather a statistical generalization

for the whole area or a whole cell within a large-scale model. Instead of a number of

random patches currently simulated by GUESS, patches with characteristic soil

moisture and fire disturbance regimes could be modelled. Changes in the distribution of

patches due to changes in the water budget can be predicted using the equation for the

average water table depth. Analytical solutions for the kinematic wave equation [Troch

et al., 2002] can be used to simulate daily water fluxes at specific locations along the

catena if daily accounting of DOC fluxes is required. The accompanying convection–

dispersion equation can also be applied to the DOC, and an analytical solution is

possible under some simplifying assumptions. It has yet to be proven whether the

proposed generalization is mathematically correct. However, such an intuitive approach

is found in classical ecological studies, where patterns and relationships found in sample

catena studies are extrapolated to the whole geographical region of interest.

34

APPENDICES

Appendix A

A1. Derivation of Equation (5) in Paper III

Starting from equations (3)-(4) in Paper III (symbol definitions in the paper)

,aMPaV

MaK

t

ac1sdes

cDdes

c

(1)

,aaV

MaK

t

as2sdes

cDdes

s

(2)

(1)+(2)

,aaMPt

)aa(s2c1

sc

(3) by definition ,k 1s2 (4)

using (4) in (3)

),aka(MPt

)aa(ssc1

sc

(5)

at the sorption steady-state

,aV

MaK sdes

cDdes

,V

MKaa Dcs

(6)using (6) in (5)

),V

MKk1(aMP

t

)V

MK1(a

Dsc1

Dc

introducingV

MK11K D and )

V

MKk1(2K Ds1 , (7)

cc

cc

a1K

2KM

1K

P

t

a

,a2KPMt

a1K

Introducing1K

PM3K and

1K

2K4K , (8)

35

,4K

3K))tt(4Kexp()a4K3K(a

)),tt(4Kexp(a4K3K

a4K3K

,t4Ka4K3K

)a4K3K(

,a4K3Kt4K

)a4K3K(

,a4K3Kt

a

11cc

11c

c

t

1t

t

1t c

c

cc

cc

,4K

3K))1tt(4Kexp()

4K

3Ka(a 1cc (9)

from (8) and (7)

V

MK1

)V

MKk1(

1K

2K4K

D

Ds1

,

)V

MKk1(

PM

2K

PM

4K

3K

Ds1

,

by definition V)t(cac and Vca 11c ,

,

V

MKk1

MP))tt(

V

MK1

V

MKk1

exp()

V

MKk1

MPVc(V)t(c

Ds1

1

D

Ds

1

Ds1

1

A2. Derivation of Equation (6) in Paper III

If volume of water, V, and mass of peat, M, are constant during the experiment, thenfrom the mass conservation:

VcMsVcMs 0011 ,

where c0 and c1 are DOC concentrationand s0 and s1 are SSOC concentration at the start of experiment and at t1, respectively

0011 cV

Msc

V

Ms ,

At steady-state 1D1 cKs ,

where KD is an equilibrium sorption partitioning constant

00D1 cV

Ms)1

V

MK(c ,

36

,1

V

MK

cV

Ms

c

D

00

1

Appendix B. Derivation of the equation for the annual average steady-state depth to the water table on a 1D hillslope

Subsurface flow from a one-dimensional hillslope with a sloping aquifer can be described by the Boussinesq equation [Boussinesq, 1904]

f

P

x

hisin

x

hh

xicos

f

k

t

h, (1)

where h is elevation of the water table measured perpendicular to the underlyingimpermeable bedrock i is a slope angle of the bedrockk is hydraulic conductivityf is drainable porosityP is recharge to the saturated area.When averaged over a year, the changes in unsaturated zone storage can be ignored. In till soils surface discharge by overland flow is seldom observed. P then equals effective precipitation, the difference between precipitation and actual evapotranspiration.Equation (1) is written using Cartesian coordinates, xh (Figure A), so that the x-axis runs parallel to the bedrock For slightly sloping bedrock ,icosisin

f

P

x

hh

xicos

f

k

t

h,

A steady-state water table depth can be calculated from:

icosk

P

x

hh

x,

x

0

x

0

dxicosk

P

x

hh ,

37

,c)xxc(icosk

Ph

,cxcicosk2

Px

2

h

,x)cicosk

P(hh

,cxicosk

P

x

hh

42

32

21

22

x

0

x

0

1

1

,c)xxc(icosk

Ph 4

23 (2)

where c1, c2, c3 and c4 are integration constantsIn the case when elevation of the water level above the bedrock is the same in left and right draining streams (Figure A1), equation (2) is symmetrical. The point with thehighest elevation of the water table along the hillslope is an equal distance, a, from the left and right streams. Constants in (2) can then be determined:

At the left stream location ,hc,ch,0x 2044

Where ho is the average elevation of the stream water level above the bedrock

At the right stream location ,hh,a2x 0

,a2c

0))a2(a2c(icosk

P

3

23

Replacing c3 and c4 in (2) ,h)xax2(icosk

Ph 2

02 (3)

Equation (3) for the case of horizontal bedrock (cos i=1) was used in Paper I with reference to Ivanov [1981], who published it without derivation, but referring toKamensky [1935].

If another coordinate system h~

x~ is chosen so that elevation h~

is calculated from a fixed reference depth, which is absolute rather then related to the bedrock (Figure A1), thenequation (3) can be transformed:

,itan)xa2(icos

hh~

Similar to the symmetrical case is a derivation for the non-symmetrical case (FigureA2), when beds associated with the left and right streams have different elevations fromthe bedrock. Separate equations are then written for the portions of the hillslopedraining into the left and right streams respectively:

38

x

xh

hh0

h0

hmax

a h0

i

h0l

L

h0r

hl

h0l hl

xl

xl

h0r

hr

hr

hmax

al a

rx

r

xr

i

Fig.A. Catchment characteristics used for the water table depth calculations (Appendix B). Symbol definitions are in the text.

A1: symmetric case a=L/2; left and right channels have the same water level relative to the impermeable bed

A2: asymmetric case al+ar=L;left and right channels have different water level relative to the impermeable bed

free ground water surface positions

impermeable horizon

39

Left portion:

,h)xxa2(icosk

Ph 2

l02

llll (4)

Right portion:

,h)xxa2(icosk

Ph 2

r02

rrrr (5)

The divide location rrll ax,ax (Figure A2) can be found from

rl hh , and Laa rl

2

L

icoskPL2

hha

2l0

2r0

l , (6)

If another coordinate system h~

x~ is chosen so that elevation h~

is calculated from thefixed reference depth, which is absolute rather then related to the bedrock (Figure A2),then equations (5)-(6) can be transformed:

,itanxicos

hh~

,itan)xL(icos

hh~

rr

ll

Equations (5)-(6) were used in the thesis to reconstruct the water table position along the sample catena (Figure 4 thesis).

Appendix C. Derivation of the equation for the correction factor for soil moisture related to the local water table depth

The Brooks and Corey equation [1966] gives the following expression to connectcapillary potential ( ) and soil water content ( ):

,n

a

ro

r (1)

where n0 -porosity

a -air entry tension head

- a parameter that depends on the pore size distribution of the soil-residual water content

Application of the Brooks-Corey characteristic function for simulating surface moisture( 0 ) from the depth to the ground water (zWT) assumes the steady-state distribution of

soil moisture in the unsaturated zone above the water table*. In steady-state conditions

40

(no vertical flow) the total potential is zero and the capillary potential equals thegravitational potential. At the soil surface: WTz

Equation (1) can then be modified for the surface water content, assuming zero residual water content:

aWTWT

a00 zif,

zn (2)

Surface water content, 0 , simulated by phenomenological equations in ecosystem

model GUESS can be related to a certain average water table depth. This can be achieved by fitting simulated 0 in (1) and estimating zWT for a longer time interval (a

year), with subsequent averaging. The reference water table depth, zR, obtained this waycan then be used to introduce a correction factor to the surface water content for sites with different local water table depths.

aWTWT

RR

RWT

Ra0 zif,

z

z

zz

zn

where is the water content at a specified location and R is the surface water content

simulated by GUESS, assuming free drainage. This correction does not change the dailyvariability in simulated water content, but rather the absolute value becomes smaller fordry locations with a deep water table and larger for wet and semi-wet areas with impeded drainage and a shallow water table. Estimates are rough, but soil moistureobservations in till soils at sites with various water table depths [Beldring et.al, 1999]show that they are not without value.

For aWTz surface water content is assumed to be equal to porosity

* The following factors have to be considered to justify the validity of the steady-stateapproximation [e.g., Grip and Rodhe, 1994]:

Hydraulic conductivity should be high enough so that any shift from the steady-state moisture distribution caused by precipitation or evapotranspiration wouldlead to a new steady-state being established within a relatively short time periodGroundwater should not be too deep. If the groundwater is deep, a zone with lowwater content and low water conductivity may form between the surface layer andthe capillary fridge above the water table. In this case, no equilibration between the two would occur.

Till soils generally satisfy the above conditions. Because of relatively low effectiveporosity, the water table in till soils is usually shallow [Grip and Rodhe, 1994]. Watertable depth also varies greatly over the seasons [Grip and Rodhe, 1994]. No drasticreduction in hydraulic conductivity occurs as a result of drying, and, in general,equilibration within the unsaturated zone is relatively fast [hours according to numericalexperiments using Richards’ equation]. The capillary fridge usually reaches the soil surface and variations in the surface soil moisture follow variations in water table depth [e.g., Beldring et al., 1999].

41

42

REFERENCES

Ågren, G. I., and R. Hyvonen (2003), Changes in carbon stores in Swedish forest soils due to increased biomass harvest and increased temperatures analysed with a semi-empirical model, Forest Ecology and Management, 174, 25-37.

Alexandersson, H., C. Karlström, and S. Larsson-Mccann (1991), Temperature and precipitation in Sweden 1961-1990. Reference normals., 87 pp., SMHI, Norrköping.

Beldring, S., L. Gottschalk, J. Seibert, and L. M. Tallaksen (1999), Distribution of soil moisture and groundwater levels at patch and catchment scales, Agricultural and Forest Meteorology, 98-99, 305-324.

Beven, K. J., and A. M. Binley (1992), The future of distributed models-model calibration and uncertainty prediction, Hydrological Processes, 6, 279-298.

Bigelow, N. H., L. B. Brubaker, M. E. Edwards, S. P. Harrison, I. C. Prentice, P. M. Anderson, A. A. Andreev, et al. (2003), Climate change and Arctic ecosystems: 1. Vegetation changes north of 55°N between the last glacial maximum, mid-Holocene, and present, Journal of Geophysical Research 108, 8170.

Bishop, K., J. Seibert, S. Kohler, and H. Laudon (2004), Resolving the Double Paradox of rapidly mobilized old water with highly variable responses in runoff chemistry, Hydrological Processes, 18, 185-189.

Bishop, K. H., Y. H. Lee, C. Pettersson, and B. Allard (1995), Methylmercury output from the Svartberget Catchment in Northern Sweden during spring flood, Water,Air, and Soil Pollution, 80, 445-454.

Bonan, G. B. (2002), Ecological Climatology, Cambridge University Press, Cambridge, UK.

Bonan, G. B., D. Pollard, and S. L. Thompson (1992), Effects of boreal forest vegetation on global climate, Nature, 359, 716-718.

Boussinesq, J. (1904), Recherches theoriques sur l'ecoulement des nappes d'eau infiltrees dans le sol., J. Math Pures Appl 5me Ser, 10, 5-78.

Boyer, E. W., G. M. Hornberger, K. E. Bencala, and D. McKnight (1996), Overview of a simple model describing variation of dissolved organic carbon in an upland catchment, Ecological Modelling 86, 183-188

Brooks, R. H., and A. T. Corey (1966), Properties of porous media affecting fluid flow, Proc. Am.Soc.Civ. Eng. J. Irrigation ans Drainage Division, 92, 61-88.

Buffam, I., H. Laudon, J. Temnerud, C.-M. Mörth, and K. Bishop (2007), Landscape-scale variability of acidity and dissolved organic carbon during spring flood in a boreal stream network, Journal of Geophysical Research-Biogeosciences, 112,doi:10.1029/2006JG000218.

Carrasco, J. J., J. C. Neff, and J. W. Harden (2006), Modeling physical and biochemical controls over carbon accumulation in a boreal forest soil. , Journal of Geophysical Research, 111:G02004, doi:02010.01029/02005JG000087.

Charney, J. G. (1975), Dynamics of deserts and droughts in the Sahel, Q. J. R. Meteorol. Soc., 101, 193–202.

Chertov, O. G. (1985), Simulation model of mineralization and humification of the forest litter and forest floor. (Imitatsionnaya model mineralizatsii i gumifikatsii

43

lesnogo opada i podstilki). Zh. Obshch. Biol., 46, 794–804 (in Russian). Chertov, O. G., and A. S. Komarov (1997), SOMM: A model of soil organic matter

dynamics, Ecological Modelling, 94, 177-189. Chertov, O. G., A. S. Komarov, G. Crocker, P. Grace, J. Klir, M. Korschens, P. R.

Poulton, et al. (1997), Simulating trends of soil organic carbon in seven long-term experiments using the SOMM model of the humus types. , Geoderma, 81, 121-135.

Chertov, O. G., A. S. Komarov, M. Nadporozhskaya, S. S. Bykhovets, and S. L. Zudin (2001), ROMUL - a model of forest soil organic matter dynamics as a substantial tool for forest ecosystem modeling, Ecological Modelling, 138, 289-308.

Clark , J., P. J. Chapman , A. L. Heathwaite, and J. Adamson (2006), Suppression of Dissolved Organic Carbon by Sulfate Induced Acidification during Simulated Droughts, Environmental Science and Technology 40, 1776-1783.

Cole, J. J., N. F. Caraco, G. W. Kling, and T. K. Kratz (1994), Carbon dioxide supersaturation in the surface waters of lakes, Science, 265, 1568-1570.

Cole, J. J., Y. T. Prairie, N. F. Caraco, W. H. McDowell, L. J. Tranvik, R. G. Striegl, C. M. Duarte, et al. (2007), Plumbing the Global Carbon Cycle:Integrating Inland Waters into theTerrestrial Carbon Budget, Ecosystems 10, 171–184

Cox, P. M., R. A. Betts, C. D. Jones, S. A. Spall, and I. J. Totterdell (2000), Acceleration of global warming due to carboncycle feedbacks in a coupled climate model, Nature, 408, 184–187.

Cramer, W., F. I. Bondeau, F. I. Woodward, I. C. Prentice, R. A. Betts, V. Brovkin, P. M. Cox, et al. (2001), Global response of terrestrial ecosystem structure and functioning to CO2 and climate change: results from six dynamic global vegetation models, Global Change Biology, 7, 357-374.

Davidson, E. A., I. A. Janssen, and Y. Luo (2006), On the variability of respiration in terrestrial ecosystems: moving beyond Q10, Global Change Biology, 12, 154–164

Dean, W. E., and E. Gorham (1998), Magnitude and significance of carbon burial in lakes, reservoirs, and peatlands, Geology, 26, 535-538.

Engelmark, O. (1987), Fire history correlations to forest type and topography in northern Sweden, Annales Botanici Fennici 24, 317-324.

Fan, Y., and R. L. Bras (1995), On the concept of representative elementary area in catchment runoff., Hydrological Processes, 9, 821-832.

Farquhar, G. D., S. von Caemmerer, and J. A. Berry (1980), A biochemical model of photosynthetic CO2 assimilation in Leaves of C3 species, Planta, 149, 78-90.

Foley, J. A., S. Levis, I. C. Prentice, D. Pollard, and S. L. Thompson (1998), Coupling dynamic models of climate and vegetation, Global Change Biology, 4, 561–579.

Freeman, C., C. D. Evans, and D. T. Monteith (2001b), Export of organic carbon from peat soils, Nature, 412, 785.

Freeman, C., N. Ostle, and H. Kang (2001a), An enzymic 'latch' on a global carbon store, Nature, 409, 149.

Gorham, E. (1991), Northern Peatlands: Role in the Carbon Cycle and Probable Responses to Climatic Warming, Ecological Applications, 1, 182-195.

Granberg, G., H. Grip, M. O. Lofvenius, I. Sundh, B. H. Svensson, and M. Nilsson (1999), A simple model for simulation of water content, soil frost, and soil

44

temperatures in boreal mixed mires, Water resources research, 35, 3771-3782. Grieve, I. C. (1991), A model of dissolved organic carbon concentrations in soil and

stream waters, Hydrological Processes, 5, 301-307. Grip, H., and K. H. Bishop (1990), Chemical dynamics of an acid stream rich in

dissolved organics, in The Surface Water Acidification Programme, edited by B. J. Mason, pp. 75-84, London: Royal Society, London.

Grip, H., and A. Rodhe (1994), Vattens väg från regn till bäck, Hallgren & Fallgren Studieförlag ab, Uppsala.

Halexeltine, A., and I. C. Prentice (1996a), A general model for the light-use efficiency of primary production, Functional Ecology, 10, 551-561.

Haxeltine, A., and I. C. Prentice (1996b), BIOME3: An equilibrium terrestrial biosphere model based on ecophysiological constrains, resource availability, and competition among plant functional types, Global Biogeochemical Cycles, 10, 693-709.

Hilbert, D. W., N. Roulet, and T. Moore (2000), Modelling and analysis of peatlands as dynamical systems, Journal of ecology, 88, 230-242.

Inisheva, L. I. (2003), Spatial and temporal features of peatland development (Prostranstvenno-vremennye zakonomernosti razvitia bolotnykh sistem), in Self-organization and dynamics of geomorphosystems under anthropogenic and climate forcing edited, Tomsk (In Russian).

IPCC (1996), Climate Change 1995—The Science of Climate Change: Contribution of Working Group I to the Second Assessment Report of the Intergovernmental Panel on Climate Change., New York: Cambridge University Press.

IPCC (2001), IPCC Third Assessment Report: Climate Change 2001: Synthesis Report,184 pp., IPCC, Geneva, Switzerland.

Ivanov, K. E. (1981), Water movement in Mirelands., Academic Press, London. Ivanov, K. E., and E. L. Shumkova (1967), Hydrological motivation and simulation of

forest disappearance and natural paludification that accompany river damming (Gidrologicheskoe obosnovanie i raschet vypadeniia lesov i rasshireniia ploshchadey estestvennogo zabolachivaniia pri podtoplenii v rechnykh sistemakh), Proceeding of GGI, 145, 3-26 (In Russian).

Kamensky, G. N. (1935), Foundations of groundwater dynamics. Part.2, 279 pp., ONTI NKTP, Moscow (In Russian).

Kjellström, E., L. Bärring, S. Gollvik, U. Hansson, C. Jones, P. Samuelsson, M. Rummukainen, et al. (2005), A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3), Norrköping.

Koca, D., B. Smith, and M. T. Sykes (2006), Modelling regional climate change effects on Swedish ecosystems, Climatic Change, 78, 381-406.

Lantmäteriet (3-2002), Terrängkartan , 21J NO Vindeln Lind, B. B., and L. Lundin (1990), Saturated Hydraulic Conductivity of Scandinavian

Tills Nordic Hydrology, 21 107-118.Lindström, G., and H. Alexandersson (2004), Recent mild and wet years in relation to

long observation records and future climate change in Sweden, Ambio, 33, 183-186. Lucht, W., I. C. Prentice, R. B. Myneni, S. Sitch, P. Friedlingstein, W. Cramer, P.

45

Bousquet, et al. (2002), Climatic control of the high-latitude vegetation greening trend and Pinatubo effect, Science 296, 1687-1689.

Malmström, C. (1923), Degerö stormyr: en botanisk, hydrologisk och utvecklinghistorisk undersökning över ett nordsvenskt myrkomplex (in Swedish), 176 pp., Stockholm.

Manabe, S., and R. T. Weatherald (1967), Thermal Equilibrium of the Atmosphere with a Given Distribution of Relative Humidity, Journal of the Atmospheric Sciences, 24,241-259.

MarkInfo (2006), Soil characteristics and vegetation of Swedish forest soils. The complete/Part of the dataset used in this publication was made available by the Swedish National Survey of Forest Soils and Vegetation performed by the Department of For est Soils, SLU. The author/authors are solely responsible for the interpretation of data, edited.

McKnight, D., and G. R. Aiken (1998), Sources and Age of Aquatic Humus, in AquaticHumic Substances Ecology and Biogeochemistry, edited by D. O. Hessen and L. J. Tranvik, Springer.

Michalzik, B., E. Tipping, J. Mulder, J. F. G. Lancho, E. Matzner, C. L. Bryant, N. Clarke, et al. (2003), Modelling the production and transport of dissolved organic carbon in forest soils, Biogeochemistry, 66, 241-264.

Moorcroft, P. R. (2006), How close are we to a predictive science of the biosphere?, Trends in Ecology and Evolution 21

Moore, T. R., and R. Knoweles (1987), Methane and carbon dioxide evolution from subarctic fens, Canadian Journal of Soil Science, 67, 77-81.

Morales, P., T. Hickler, D. P. Rowell, B. Smith, and M. T. Sykes (2007), Changes in European ecosystem productivity and carbon balance driven by Regional Climate Model output., Global Change Biology 13, 108-122.

New, M., M. Hulme, and P. D. Jones (1999), Representing twentieth century space-time climate variability. Part 2: development of 1901 1996 monthly grids of terrestrial surface climate. , unpublished.

Ottosson-Löfvenius, M. O., M. Kluge, and T. Lundmark (2003), Snow and soil frost depth in two types of shelterwood and a clear-cut area. , Scand. J. Forest Res., 18,54-63.

Palmer, T. N., G. J. Shutts, R. Hagedorn, F. J. Doblas-Reyes, T. Jung, and M. Leutbecher (2005), Representing model uncertainty in weather and climate prediction 33, 163-193.

Pastor, J. J., and W. M. Post (1988), Response of northern forest to CO2-induced climate change, Nature, 334, 55-58.

Prentice, I. C., A. Bondeau, W. Cramer, S. P. Harrison, T. Hickler, W. Lucht, S. Sitch, et al. (2007), Dynamic global vegetation modelling: quantifying terrestrial ecosystem responses to large-scale environmental change, in Terrestrial Ecosystems in a Changing World, edited by J. G. Canadell, et al., pp. 175-192., Springer Verlag, Springer Verlag.

Qualls, R. G. (2000), Comparison of the behavior of soluble organic and inorganic nutrients in forest soils, Forest Ecology and Management, 138, 29-50.

Roulet, N., S. Hardill, and N. Comer (1991), Continuous Measurement of the Depth of

46

Water-Table (Inundation) in Wetlands with Fluctuating Surfaces, HydrologicalProcesses, 5, 399-403.

Sagerfors, J. (2007), Land-Atmosphere exchange of CO2, water and energy at a boreal minerotrophic mire. Doctoral Dissertation. Acta Universitatis agriculturae Sueciae vol. 2007:4., Umeå.

Sagerfors, J., A. Lindroth, A. Grelle, L. Klemedtsson, P. Weslien, and M. Nilsson (2007), Annual Carbon Exchange Between a Nutrient Poor, Mineritrophic, Boreal Mire and the Atmosphere, Journal of Geophysical Research- Biogeosciences 112,doi:10.1029/2006JG000306.

Saltelli, A. (Ed.) (2000), Sensitivity analysis, 475 pp., Wiley, Chichester. Schaphoff, S., W. Lucht, D. Gerten, S. Sitch, W. Cramer, and I. C. Prentice (2006),

Terrestrial biosphere carbon storage under alternative climate projections., Climaticchange, 74 7-122

Schlesinger, W. H. (1997), Biogeochemistry, An Analysis of Global Change, Academic Press.

Shaman, J., M. Stieglitz, and D. Burns (2004), Are big basins just the sum of small catchments? , Hydrological Processes, 18, 3195–3206.

Shukla, J., and Y. Mintz (1982), Influence of land-surface evapotranspiration on the Earth’s climate, Science 215, 1498–1501.

Silberstein , R. P., R. A. Vertessy , J. Morris, P. M. Feikma, and (1999), Modelling the effects of soil moisture and solute conditions on long term tree growth and water use: a case study from the Shepparton irrigation area, Australia. , Agricultural Water Management, 39, 283-315.

Sirin, A., S. Kohler, and K. Bishop (1998), Resolving flow pathways and geochemistry in a headwater forested wetland with multiple tracers, Hydrology, Water Resources and Ecology in Headwaters.

Sitch, S., B. Smith, I. C. Prentice, A. Arneth, A. Bondeau, W. Cramer, J. O. Kaplan, et al. (2003), Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ Dynamic Global Vegetation Model, Global Change Biology, 9,161-185.

Skyllberg, U., and T. Magnusson (1995), Cation adsorbed to soil organic matter-a regulatory factor for the release of organic carbon and hydrogen ions from soils to water Water, Air and Soil Pollution, 85, 1095-1100.

Smith, B., I. C. Prentice, and M. T. Sykes (2001), Representation of vegetation dynamics in the modelling of terrestrial ecosystems: comparing two contrasting approaches within European climate space, Global Ecology & Biogeography, 10,621-637.

Sobek, S. (2005), Carbon Dioxide Supersaturation in Lakes-Causes, Consequences and Sensitivity to Climate Change, 42 pp, Uppsala University, Uppsala.

Sobol', I. M. (1990), Sensitivity estimates for nonlinear mathematical models (in Russian), Matematicheskoe Modelirovanie, 2, 112-118.

Sprugel, D. G., M. Ryan, J. R. Brooks, K. A. Vogt, and T. A. Martin (1995), Respiration from the Organ Level to the Stand, in Resource physiology of conifers. Acquisition, allocation and utilization, edited by W. K. Smith and T. H. Hinckley,

47

Academic Press, San Diego, California. Steinberg, C. E. W., S. Kamara, V. Y. Prokhotskaya, L. Manusadrianas, T. A.

Karasyova, M. A. Timofeyev, Z. Jie, et al. (2006), Dissolved humic substances-ecological driving forces from the individual to the ecosystem level?, FreshwaterBiology, 51, 1189-1210.

Sykes, M. T., I. C. Prentice, B. Smith, W. Cramer, and S. Venevsky (2001), An introduction to the European Terrestrial Ecosystem Modelling Activity, GlobalEcology and Biogeography, 10, 621-637.

Temnerud, J., J. Seibert, M. Jansson, and K. Bishop (2007), Spatial variation in discharge and concentrations of organic carbon in a

catchment network of boreal streams in northern Sweden Journal of Hydrology,doi:10.1016/j.jhydrol.2007.1005.1015.

Tranvik, L. J., and M. Jansson (2002), Terrestrial export of organic carbon, Nature, 415,861-862.

Troch, P., E. van Loon, and A. Hilberts (2002), Analytical solution to a hillslope-storage kinematic wave equation for subsurface flow, Advances in Water Resources, 25,637-649.

Walker, S., J. B. Wilson, J. B. Steel, G. L. Rapson, B. Smith, W. M. King, and Y. H. Cottam (2003), Properties of ecotones: Evidence from five ecotones objectively determined from a coastal vegetation gradient. , Journal of Vegetation Science, 14,579-590.

Weiss, R., J. Alm, R. Laiho, and J. Laine (1998), Modeling moisture retention in peat soils, Soil science society of America journal, 62, 305-313.

Vernadsky, V. I. (1926, in Russian; 1998, English translation), The biosphere, 192 pp., Copernicus, New York.

Vertessy, R., T. J. Hatton, P. J. O'Shaughnessy, and M. D. A. Jayasuriya (1993), Predicting water yield from a mountain ash forest using terrain analysis based catchment model, Journal of Hydrology, 150, 665-700.

Williams, T. G., and L. B. Flanagan (1998), Measuring and modelling environmental influences on photosynthetic gas exchange in Sphagnum and Pleurozium, PLant,Cell and Environment, 21, 555-564.

Willmott, C. J. (1982), Some comments on the evaluation of model performance, Bulletin of the American Meteorological Society, 63, 1309-1313.

Wood, E. F., M. Sivapalan, K. Beven, and L. Band (1988), Effects of spatial variability and scale with implication to hydrologic modelling, Journal of Hydrology, 102, 29-47.

Zaehle, S., A. Bondeau, T. R. Carter, W. Cramer, M. Erhard, I. C. Prentice, I. Reginster, et al. (in press), Projected changes in terrestrial carbon storage in Europe under climate and land use change, 1990-2100. , Ecosystems.

Zhang, L., W. R. Dawes, T. J. Hatton, P. H. Reece, G. T. H. Beale, and Z. Packer (1999), Estimation of soil moisture and groundwater recharge using the TOPOG-IRM model, Water Resources Research, 35, 149-161.

48

POPULÄRVETENSKAPLIG SAMMANFATTNING

Koncentrationen av växthusgaser i atmosfären har ökat kontinuerligt under de senaste

200 åren, och ökningen tillskrivs huvudsakligen antropogena utsläpp. FN´s senaste

klimatrapport slår vidare fast att den ökade halten av växthusgaser i atmosfären leder till

klimatförändringar. Både förhöjda halter av växthusgaser i atmosfären och ett ändrat

klimat påverkar dock omsättningen av växthusgaser i olika ekosystem, s.k.

återkopplingsmekanismer. Förändringarna i ekosystemens funktion kan leda till

antingen ökad eller minskad produktion eller upptag av växthusgaser. D.v.s att

förändringarna i ekosystemens funktion kan antingen ytterligare accelerera eller bromsa

växthuseffekten. Ett av de viktigaste redskapen för att förstå hur olika ekosystem svarar

på klimatförändringar är modeller.

Syftet med den här avhandlingen har varit att utveckla och att anpassa

ekosystemmodeller till boreala ekosystem. Arbetet har utgått ifrån en generell

ekosystemmodell, GUESS, till vilken en särskild modell för markens biogeokemiska

processer, ROMUL, har kopplats. Arbetet har huvudsakligen inriktats på att utveckla

modeller för att studera interaktioner mellan vattnets omsättning, hydrologin, och

omsättningen av kol. Fältdata för såväl parametrisering som testning av modellerna

kommer från Vindelns Skogliga Försöksparker, Sveriges Lantbruksuniversitet, ca 6 mil

från Umeå, norra Sverige, och har tillhandahållits av olika forskargrupper. Mer precist

har följande frågeställningar behandlats:

Hur varierar mängden organiskt material i marken som en funktion av

varierande markfuktighet i ett skogslandskap?

Hur påverkas utbytet av koldioxid mellan en myr och atmosfären av

vattenytans position, d.v.s. vad händer om vattenbalansen ändras?

Vad kontrollerar mängd och variabilitet av löst organiskt kol i avrinningen från

torvmarker? Även vilka faktorer som styr variationen av löst organiskt material

på olika tidsskalor har studerats.

Simuleringarna av mängden organiskt material som en funktion grundvattenytans läge

utmed olika transekter i ett skogslandskap genererar samma variation i organiskt

material som återfinns i landskapet. Dock visar modellresultaten att olikheter i mängd

organiskt material bara till viss del kan förklaras med enbart variation i markfuktighet.

En väl så viktig faktor är förekomsten av skogsbrand. Om man antar naturlig förekomst

av skogsbrand så härrör ca 40% av variationen i humuslagrets tjocklek från brand. I och

med att människan i dag, i Sverige, mycket framgångsrikt bekämpar skogsbränder tyder

detta på att humustäckets tjocklek i Svenska skogar sannolikt kommer att öka.

Ytterligare ett viktigt bidrag i avhandlingen har varit inkorporeringen av torvbildande

ekosystem, myrar, i den generella ekosystemmodellen GUESS. Den nya versionen av

modellen användes b.la. för att studera hur varierande vattenytenivå samt temperatur

påverkar utbytet av koldioxid mellan myren och atmosfären. Modellen visar tydligt att

så länge vattenytan varierar inom det intervall som är typiskt för det dominerande

49

växtsamhället så är flödet av koldioxid relativt stabilt. Om vattenytan blir betydligt

högre eller lägre än normalt så påverkas utbytet av koldioxid i första hand genom att

fotosyntesen påverkas och i mindre utsträckning genom ändrad nedbrytning.

Ytterligare en utveckling av modellen gjordes för att kunna modellera förekomsten av

löst organiskt material i markvatten i myrar samt i avrinningen från myrar.

Modelleringen av koncentrationen av löst organiskt material baseras på en ekvation för

advektion och dispersion. Modellen inkluderar vidare både produktion, nedbrytning

samt adsorption av det lösta organiska materialet. Eftersom mängden humus i sjöar och

vattendrag har stor betydelse för många av vattnens biogeokemiska funktioner och dess

kvalité är det av stor vikt kunna modellera mängden löst organiskt material och hur

denna påverkas av olika mänskliga aktiviteter såväl som av klimatförändringar.

Modellkörningar visar bl.a. att de huvudsakliga faktorerna som bestämmer nivån på hur

mycket löst organiskt material som kan finnas i vattnet är hastigheten på bildningen

samt nedbrytningen av det lösta organiska materialet. Hur mycket löst organiskt

material som finns vid en given tidpunkt bestäms däremot huvudsakligen av

adsorptionsprocesser till det fasta organiska materialet.

De olika utvecklingarna och anpassningarna av ekosystemmodellen GUESS-ROMUL

är ett viktigt bidrag både för att ytterligare kunna förstå kolomsättningen i olika

ekosystem men också för att göra uppskattningar av storleken på olika processer eller

flöden över större områden.

50

I

I

e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

avai lab le at www.sc iencedi rec t .com

journa l homepage: www.e lsev ier .com/ locate /eco lmodel

Carbon storage in the organic layers of boreal forest soilsunder various moisture conditions: A model study forNorthern Sweden sites

Alla Yu. Yurova ∗, Harry LankreijerDepartment of Physical Geography and Ecosystems Analysis, Lund University, Solvegatan 12, 223 62 Lund, Sweden

a r t i c l e i n f o

Article history:

Received 31 October 2005

Received in revised form

29 January 2007

Accepted 1 February 2007

Published on line 23 March 2007

Keywords:

Soil moisture

Carbon sequestration

Boreal forest

Vegetation dynamics

Model

ROMUL

GUESS

a b s t r a c t

A typical feature of the boreal forest landscape is a gradient from dry to wet sites, with

associated increases in the depth of the soil organic layers. In this study, the coupled

ecosystem–soil biogeochemistry model GUESS–ROMUL is used to explore how the spe-

cific features of soil organic matter decomposition and vegetation dynamics account for an

observed difference between the soils formed under contrasting moisture conditions. Two

sites, one mesic and one mesic-to-wet, representative of the natural forest in Northern Swe-

den, are simulated. In addition to the assumptions underlying the GUESS–ROMUL model,

it is assumed that the fire frequency was higher at the mesic site. The model shows that

with a natural fire regime, the soil organic layers at the mesic-to-wet site store 6.0 kg C m−2

compared to 3.1 kg C m−2 at the mesic site. Forty-seven percent of the difference between

the sites in this respect is explained by suppressed decomposition under higher moisture

conditions, 37% by the decreased litter input into the soil (more frequently disturbed ecosys-

tems have lower productivity) and 16% by direct consumption of the forest floor in fires. It

is predicted that due to anthropogenic fire suppression the organic soil layers of mesic sites

will, in the future, sequester carbon at an average rate of 0.0103 kg C m−2 year−1 and have

an equilibrium storage capacity of 5.4 kg C m−2. For the mesic-to-wet site, the model pre-

dicts an extremely slow sequestration rate of 0.0022 kg C m−2 year−1. The effect of increased

precipitation on the carbon storage at the landscape level is also investigated.

© 2007 Elsevier B.V. All rights reserved.

1. Introduction

The soil, which represents a large and dynamic carbon pool,is a key element that must be considered when discussinghow carbon storage in boreal ecosystems may be influencedby future climate change. Some fractions of soil carbon, suchas humic substances bound to mineral matrices, have a veryslow turnover rate and therefore changes in carbon storagecapacity should be considered over a long time scale (e.g.Trumbore and Harden, 1997). However, the soil organic lay-

∗ Corresponding author. Tel.: +46 46 2223749; fax: +46 46 2220321.E-mail address: alla.yurova@nateko.lu.se (A.Yu. Yurova).

ers (O layers) are much more labile and the carbon content inthem can vary considerably, even within a decade (Bormannet al., 1995; Yu et al., 2002). Among the factors affecting soil Cdynamics, changes in litter production and soil temperatureare most often considered. Following anticipated increasesin atmospheric CO2 levels and temperature, primary produc-tion is expected to increase, leading to increases in organicmaterial inputs to the soil. On the other hand, rises in soiltemperature may compensate for, or even overbalance thiseffect by increasing respiratory losses from the soil. Currently,

0304-3800/$ – see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolmodel.2007.02.003

476 e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

there is no general agreement regarding how these two forcesare balanced in the boreal region, but soil organic matter mod-els are widely used in attempts to explore the relationships,and mechanisms, involved (Liski et al., 1999; Hyvonen et al.,2002; Agren and Hyvonen, 2003; Davidson and Janssens, 2006;Carrasco et al., 2006).

Model simulations typically assume that soil moistureconditions are average, or optimal, despite the fact that obser-vations suggest that there is a strong correlation between theamount of carbon in the O layers and soil moisture content.A typical boreal landscape can be divided into the followingstructural units according to the soil moisture regime: (i) thedriest watershed parts and convex slopes, (ii) mesic sites, sit-uated mainly on the straight slopes, (iii) the hollows, lowerparts of the catchment adjacent to the streams and the con-cave slopes with favourable moisture conditions and (iv) mireswith water-logged conditions. The mires store the greatestamounts of carbon per unit area, but they represent ecosys-tems with distinct vegetation and hydrological regimes thatare beyond the range of the moisture gradient explored in thisstudy, so they are not considered here. Instead, we focus onthe gradient encompassing the other three units – from drysites, with usually poor O layers (typical mor), to wet sites, onfully developed humic podzol (rich O layers, hydromor) or peatsoil – that is a basic feature of boreal landscapes where mireshave not developed.

The well-drained and wet soil types are so quantitativelyand qualitatively different that they should be expected torespond differently to a changing environment. Therefore, inorder to reliably predict the future carbon storage capacity ofthe boreal zone, soil organic matter models should be testedand further developed by incorporating the effects of varia-tions in moisture conditions.

A crucial issue is why the O layers of moist forest soils accu-mulate more C than those of well-drained soils. The acceptedexplanation is that decomposition is suppressed under highmoisture conditions. However, there is an alternative hypoth-esis linking the observed differences to the historical patternsof fire disturbance (M. Nilsson, pers. commun., 2004). Moistsections of the boreal forest used to burn less frequently thandry sections (e.g. Engelmark, 1987). Consequently they tendedto have higher proportions of mature trees (e.g. Wardle et al.,2003) and, hypothetically, higher long-term average rates oflitter production. The fact that the amount of carbon stored inthe O layers can be as high as 27 kg C m−2 (equivalent to a depthof ca. 0.5 m) in boreal ecosystems where the disturbance fre-quency is apparently low provides support for this hypothesis(Hesselman, 1926; Wardle et al., 2003). These unique soils werefound in old nature reserves, or unmanaged and rarely burnedforests on islands, and were not characterised by high soilmoisture content. For comparison, a mean estimate of the car-bon storage capacity for the global boreal forest is 3.3 kg C m−2

(Vogt et al., 1986).It has been reported that as the disturbance frequency

in natural forest is reduced, soil begins to act as a C sink(Kurz and Apps, 1996; Bhatti et al., 2002). Therefore, we couldspeculate that anthropogenic fire suppression could lead tothe increased sequestration of carbon by the O soil layers,thus reducing the differences between moist and well-drainedsoils.

The aim of the study presented here was to use the cou-pled ecosystem–soil organic matter model GUESS–ROMUL toexamine the effects of two moisture conditions, mesic andmesic-to-wet, on carbon storage in organic layers of the natu-ral Northern Sweden boreal forest soils. The model simulatesvegetation dynamics, including the disturbance regime, anddecomposition of various types of organic matter in relationto soil/litter moisture levels. Analysis of the present situa-tion and future predictions are made. Some rough calculationsconcerning the effect of current changes in precipitation oncarbon storage in the soil are also included in the study, bychanging the relative proportions of sites with different mois-ture regimes in the simulated landscape.

2. Description of the modelling system

In the present study the general ecosystem simulator GUESS(Smith et al., 2001) has been fully coupled with the soil organicmatter dynamics model ROMUL (Chertov et al., 2001); thecombined modelling system is called GUESS–ROMUL. It isimportant to note that the GUESS model incorporates a formu-lation of litter and soil biogeochemistry that was not appliedhere.

GUESS is a terrestrial ecosystem model that simulates veg-etation dynamics and carbon and water exchange. The keyprocesses in the model are vegetation establishment, growth,mortality and the competition for light among plant individ-uals, which determines vegetation composition and naturalsuccession. Each simulated woody individual belongs to a cer-tain species or plant functional type (PFT) with a specific setof parameters controlling its potential climatic space, abilityto survive under various stress conditions, scaling factors forphotosynthesis and respiration, allometry and carbon alloca-tion, phenology, establishment and mortality. GUESS can beapplied to any given ecosystem by specifying these generalparameters rather than by using a direct validation proce-dure. The driving force is soil texture and climate, providedas model inputs, therefore there is always a generic connec-tion between the variability of environmental parameters andchanges in ecosystem structure and functioning (concerningthe latter, changes in productivity are the most frequentlydiscussed). The model has been recommended for, and suc-cessfully applied to, landscape-to-regional scale simulationsof various systems, including the Swedish boreal forest (Kocaet al., 2006). For a comprehensive description the reader isreferred to Smith et al. (2001). Here only the features of themodel that are directly linked to the simulation of soil carbondynamics (such as litter production, the disturbance regime,soil temperature and moisture) are outlined.

2.1. Litter production

Net primary production (NPP) is modelled in GUESS by explic-itly representing photosynthesis and maintenance respiration(Haxeltine and Prentice, 1996; Haxeltine et al., 1996; Sitch etal., 2003). At the end of the simulation year, after subtract-ing the reproduction costs from NPP, the remaining fraction ispartitioned amongst leaf mass, fine root mass and sapwoodusing allometric relationships. Leaf and fine root biomass are

e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484 477

transferred into the litter pool with a given turnover rate. Themortality of leaves (needles), root and woody biomass alsoprovides inputs into the litter pool. Mortality in GUESS is astochastic process with a background rate related to plantlongevity and growth efficiency.

2.2. Disturbance

GUESS simulates individual plant growth in a number of repli-cate patches, the patch average representing the forest stand.For each patch, stochastic generic disturbance is applied witha prescribed frequency, with disturbed biomass being trans-ferred to the litter pool. This accounts for disturbances suchas wind or insect damage but does not include fire, which ismodelled explicitly in GUESS as described by Thonicke et al.(2001). Two basic assumptions are made in the model: (i) afire can only occur if the amount of litter stored in a forestfloor exceeds a threshold value, and (ii) soil moisture is a keyenvironmental variable determining the probability of fire. Fireresistance prescribed for the PFT or plant species indicates thepercentage of individuals consumed in a fire. A fraction of thelitter is also burned if the fire occurs, and the fire resistance ofthe litter is assumed to be the same as for living individuals.For comparison with historical data and reconstructions thefire probability estimated in the model can be recalculated asan average fire return interval.

2.3. Soil temperature and water balance

Soil temperature is assumed to follow surface temperatureaccording to a Fourier annual sinusoidal cycle with a dampedoscillation amplitude and a temporal lag. Soil thermal diffu-sivity is dependent on both the soil texture and water content.

Representation of the soil moisture in GUESS is basedon Haxeltine and Prentice (1996). A two-layer, simple waterbalance scheme includes evapotranspiration, snowmelt, per-colation and runoff. An excess of soil moisture over theavailable water holding capacity in the first and secondlayers is attributed to surface runoff and drainage to theground water, respectively. The percolation coefficient andfield capacity are texture-dependent constants.

It has been demonstrated that LPJ (a global version of themodel, sharing all of the components except the vegetationdynamics component with GUESS; Smith et al., 2001) accu-rately reproduces the spatial patterns of soil moisture on acoarse scale (Wagner et al., 2003), as well as the seasonalityof soil moisture for specific sites (Sitch et al., 2003). Furtherdetails about the hydrological evaluation of the model can befound in Gerten et al. (2004). Sensitivity analysis of the LPJmodel (Zaehle et al., 2005) has revealed that, in water-limitedregions, parameters controlling the evapotranspiration rateare of considerable importance for the simulated net pri-mary production. Wolf et al. (in revision) have studied theuncertainty of GUESS model simulations arising from themathematical formulation of the water balance. The simu-lations for the region in Northern Russia produced by (i) thestandard model, and (ii) GUESS, driven by the soil moistureestimates produced by the mechanistic model JULES (Cox etal., 1999), were compared. It was shown that differences insoil moisture had a minor influence on the vegetation carbon

fluxes simulated with GUESS, whereas the soil carbon fluxeswere very sensitive to the soil moisture estimates.

ROMUL (Chertov et al., 2001) is a dynamic soil organicmatter model based on the concept of “humus type”. Themodel was originally called SOMM (Chertov, 1985; Chertov andKomarov, 1997) and was developed with data from long-termlaboratory experiments on the decomposition of plant litter inrelation to the chemical composition of the material (nitrogenand ash content), litter/soil temperature and moisture. Theprocess of decomposition is treated in the model as sequentialstages of organic matter transformation (humification) per-formed by different functional groups of soil organisms. Thestages are: L, fresh litter; F, complex of humus substanceswith underdecomposed plant debris (CHS); HO, organic layerhumus substances (OLH) in the later stages of decomposition(visually manifested as peat); HM, humus of the mineral soillayer or “true humus” closely bound to the mineral matrix. Themass balance for each of these organic matter types (pools) isdescribed by the system of linear differential equations withvarying coefficients presented below

∂L

∂t= LO − (k1 + k3)L, (1)

∂F

∂t= k3L − (k2 + k4O + k4M + k5)F, (2)

∂HO

∂t= k4OF − k7HO, (3)

∂HM

∂t= (k4M + k5)F − k6HM, (4)

where LO is a litter input into the soil per unit time, whileL, F, HO and HM are the current sizes of the respective pools(kg C m−2).

Coefficients ki (k1, . . ., k7) (day−1) account either for com-plete mineralization or transformation of the specific organicmatter type into another type. Table 1 describes the processesincluded in the ki coefficients and gives some information onthe groups of soil organisms performing these processes. Eachvalue of ki is presented as a function of the initial N and ashcontents in plant litter, multiplied by the temperature andmoisture response functions:

ki = ki(N, ash)fi(T)gi(W), (10)

The functions fi(T) and gi(W) are empirical and differ forthe different processes, to account for the fact that responsesto environmental parameters are not universal for all typesof organic matter. Here, the dependence of ki on nitrogen andash contents and soil/litter moisture were taken from Chertovet al. (2001), while temperature response curves were updatedaccording to Chertov et al. (2005). Equations parameterizingthe moisture response curves (5)–(9) are given in Table 1 asthey are used further in the discussion. Different ki coefficientswere used for the above and belowground litter compartments(omitted from the above equation system for simplicity). Fur-thermore, in comparison with a published version of ROMUL,an additional pool OLH (HO) was added to the model (Chertov,pers. commun., 2004). This accounts for the fact that whenlitter quality is poor and/or its acidity is high, the soil Arthro-

478 e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

Table 1 – Kinetic coefficients ki (i = 1, . . ., 7) used in the model

Coefficient Process described Group of the soil organismsperforming it

Dependence on the soil moisture, W(mass%)

k1 Litter mineralization (L↑) Fungi and micro-faunacomplex

(5) g1 =

⎧⎪⎨⎪⎩

0 W ≤ 7, W > 6000.0435W − 0.304 7 < W ≤ 301 30 < W ≤ 300−0.003W + 2 300 < W ≤ 600

k2 CHSa mineralization (F↑) Fungi and micro-faunacomplex

(6)

g2 =

⎧⎪⎨⎪⎩

0 W ≤ 7, W > 12000.0233W − 0.163 7 < W ≤ 50−0.00625W + 1.31 50 < W ≤ 90−0.00068W + 0.811 90 < W ≤ 1200

k3 Litter to CHS transformation (L → F) Fungi and micro-faunacomplex

g3 = g1

k4O CHS to OLHb transformation (F → HO) Bacteria and Arthropoda (7) g4 =

⎧⎪⎨⎪⎩

0.025W W ≤ 401 40 < W ≤ 400−0.0033W + 2.333 400 < W ≤ 7000 W > 700

k4M CHS to “true humus” transformation(F → HM)

k5 CHS to “true humus” transformation(F → HM)

Earthworms (8) g5 =

⎧⎪⎨⎪⎩

0 W ≤ 2, W > 1200.0769W − 0.1538 2 < W ≤ 151 15 < W ≤ 70−0.02W + 2.4 70 < W ≤ 120

k6 “True humus” mineralization (HM↑) Soil bacteria and fungi (9) g6 ={

0.025W W ≤ 401 W > 40

k7 OLH mineralization (HO↑) Soil bacteria and fungi g7 = g2

a Complex of humus substances with underdecomposed plant debris (humified forest floor).b Organic layer humus substances (peat).

poda do not transport organic material down to the mineralpart of the soil profile, but rather leave it in the organic layer.The richer is CHS in nitrogen (lower C/N ratio) the higher pro-portion is going to mineral topsoil (the more active and largerare Artropoda and Insecta larvae). Therefore, coefficient k4

describing the transformation of CHS can be divided into twocomponents: the flow to OLH (kO) and the flow to the mineralsoil layer (kM):

k4M = 0, k4O = k4,CN

≤ 10, k4M = k4

(1.4 − 0.02

CN

),

k4O = k4 − k4M, 10 <CN

≤ 35, k4O = k4, k4M = 0,CN

> 35,

(11)

The mineralization rate for OLH is described by coefficientk7 = 0.00015f2(T)g2(W) (temperature and moisture responsecurves are taken as those for CHS).

The model has been evaluated against long-term exper-imental observations and provides good results for actualforest data sets (Chertov et al., 1997)

3. Model set-up

In the present study simulations were performed for a geo-graphical location representative of the Northern Swedishboreal forest (64.0◦N, 19.5◦W). Soil texture data were takenfrom the FAO data set (Zobler, 1986; FAO, 1991) and average

monthly values for air temperature, precipitation and cloudcover, for the closest 0.5◦ × 0.5◦ grid square derived from ameteorological observation network, kindly provided by theClimate Research Unit (CRU), University of East Anglia (New etal., 1999, unpublished), were used to run the model. Althoughthe simulations were extended into the future, neither climatepredictions nor measured data for the period after 1970 wereused since the potential effects of climate change were notthe focus of this study. The series of inter-annually varyingmeteorological parameters, detrended in the case of temper-ature for the period 1901–1970, were repeated to representfluctuations about constant long-term climatic parameters.Simulations with GUESS–ROMUL started from “bare ground”,but results from the initial period (before the first equilibriumwith respect to vegetation and carbon pool) were not analysed.Plant species were described explicitly and all model param-eters were taken from the validated simulations with GUESSfor the Swedish forest (Koca et al., 2006).

Generic disturbance (other than fire) was applied with aprescribed frequency of 100 years. Plant litter C/N ratios inabove and belowground compartments for the given specieswere taken from Berg and McClaugherty (2003). According toobservations (T. Persson, pers. commun., 2004) no earthwormsare found in Swedish forests at the latitude simulated here,therefore coefficient k5, characterizing the activity of this typeof soil fauna, was set to zero.

The overall dynamics for a stand were derived by simu-lating a number (in this case 100) of 1000 m2 patches andaveraging the results. Between-patch variability resulted from

e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484 479

stochastic variation in establishment, mortality, fire, and othertypes of disturbance. Due to averaging over a large number ofpatches, stand species and age composition were relativelystable and well mixed, although dependent to a large degreeon the assumed disturbance interval.

The sum of carbon stored in all organic soil layers (L + F + HO)was analysed; the humus of the mineral layer (HM) was notincluded in this study.

Soil moisture in mass% was calculated separately for theorganic and mineral soil layers. Bulk density and wilting pointdata for the mineral layer were taken from the DAAC database(Global Soil Data Task Group, 2000). For the organic layer, roughestimates of bulk density of 0.05 g cm−3 were derived from theregression of organic layer depth on relative mass presentedby Chertov (1981).

Two contrasting soil moisture regimes, mesic and mesic-to-wet, were considered to investigate the effect of soil moistureon organic soil layer accumulation. The Swedish NationalForest inventory divides the soils in the country accordingto the water table depth (WT) in the following classes: dry(WT > 2 m), mesic (1 < WT < 2), mesic-to-wet, wet and water-logged (for the last three WT < 1 m). Beldring et al. (1999) reporta regression linking an average water table depth with soilmoisture, which was developed using data for the Swedishboreal forest. According to water moisture values producedby GUESS–ROMUL the simulation site falls into the categorymesic. That is in accordance with the model soil hydrologyformulation, which assumes free drainage (Section 2). To sim-ulate mesic-to-wet sites, soil moisture for the mesic site wassimply multiplied by the scaling factor so that WT = 0.6 m. Thisimplies that annual variability is the same for sites with dif-ferent levels of soil moisture, which is not strictly true, but isrealistic as shown by Beldring et al. (1999). The climatic aver-age of the modelled O soil layer moisture (mass%) is 220% and440% for the mesic and mesic-to-wet sites, respectively.

It is important to point out that soil moisture influences notonly the process of decomposition but also the occurrence offorest fires, which in turn influences vegetation dynamics. Toassess the effect of fire frequency on the size of the O layers,different fire return intervals were applied for the mesic andmesic-to-wet sites. First, the results from a model fire rou-tine were examined but the fire return intervals of 360 yearsfor mesic, and 1000 years for mesic-to-wet sites appeared tobe overestimated. Therefore, estimates found in the literaturefor the pre-industrial frequency of fire in the simulation areawere used instead. Fire frequency was taken as 50 years forthe mesic, and 300 years for the mesic-to-wet sites, based onestimates made by examining records of fire scarring on treesin a relatively dry pine forest area (Niklasson and Granstrom,2000) and a variety of forest types (Engelmark, 1987) in North-ern Sweden. The data from the latter study support the ideathat the fire return interval varies (from 48 to 270 years) andcorrelates with forest type, while on mires traces of fire arevery rare or absent. A study on Russian forests with naturalfire disturbance regimes (Wirth et al., 2002) found that, on aver-age, 50% of the carbon in the O soil layers is consumed by fire,which was integrated into the model. After the year 1870 allfires are assumed to be suppressed (Niklasson and Granstrom,2000). However, the effects of other forest management prac-tices were not included in this study.

The analyses presented below are based on the followingmodel runs:

• Run I. Natural fire frequency. Simulations continue 2000years from the time when the C pool in the O layers reachesa steady state.

• Run II. No fire disturbance. “Other” types of disturbance witha prescribed frequency of 100 years are included. Simula-tions continue from 1870 until the C pool in the O layersreaches a new steady state.

4. Results and discussion

4.1. Comparison of mesic and mesic-to-wet sitesunder natural fire conditions (model run I)

A mixed stand of spruce, pine and birch was predicted as thenatural vegetation for both the mesic, and for the mesic-to-wet sites. However, the average contribution of birch biomassin a stand varied by 7% on the mesic site but only by 3% on themesic-to-wet site. Birch is an early successional species so itscontribution increases with increases in fire frequency. For themesic-to-wet site, the long-term average annual abovegroundlitter production was 0.06 kg C m−2 year−1. This is consistentwith an estimate obtained from empirical regressions forspruce and pine litter fall as a function of actual evapotran-spiration, in mature forest (Akselsson et al., 2005). Soils in themore frequently disturbed forest at the mesic site had a muchlower average litter input of 0.045 kg C m−2 year−1 due to thefact that the productivity of the young forest was lower thanthat of the mature one. The model results agree with empir-ical observations (Malkonen, 1974; Berg and McClaugherty,2003) showing that litter production is a function of tree age,generally increasing and reaching a steady state as the standmatures. Fine root litter in the model was assumed to be equalto the leaf litter.

At the beginning of the simulation, the slowest-developingO soil layers (HO) reached an approximate steady state (withoscillations due to disturbance) after about 200 years in themesic site and 500 years in the mesic-to-wet site. Simulationsfrom the following 2000 years were used to investigate thefluctuations of the pool sizes around the steady state due todisturbance of particular patches. We calculated average car-bon storage at the steady state to be 3.1 and 6.0 kg C m−2 forthe mesic and mesic-to-wet stands, respectively (see Fig. 1a, inwhich the distribution of the pool sizes within the 2000-yeartime series is also shown). This difference in carbon storagecapacity could be explained by:

• Organic matter decomposition being suppressed withincreasing soil moisture.

• The higher fire frequency in the mesic sites causing a sig-nificant proportion of the soil carbon to be burned.

• The overall litter input into the soil being lower at the mesicsite. The higher fire frequency would cause young forest,which is less productive than mature stands, to dominate.

480 e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

Fig. 1 – Modelled carbon storage in forest soil O layersunder mesic (MES) and mesic-to-wet (MTW) moistureconditions and a natural fire regime. (a) Box plotcharacterizing the distribution of the total stand C stored insoil O layers. The boxes have lines at the lower, median(bold), and upper quartiles and the whiskers show thelowest and highest values. Statistics are based on a2000-year time series with climate anddisturbance-induced fluctuations around the steady state.(b) The relative contribution of peat (HO) and the other Olayers (L + F), where L is fresh litter and F is the humifiedforest floor.

Model tests have shown that 47% (95% confidence inter-val, CI, 45–49%), 16% (CI = 11–21%) and 37% (CI = 33–41%) of thedifference can be explained by the first, second and third rea-sons, respectively. Therefore, it is suggested that soil moisturemodulates carbon storage in O layers, not only by its directeffect on organic matter decomposition but also, and equallyimportantly, by determining the frequency of the disturbanceevents (fires).

The relative ratio of the peat pool (HO) and other soil O layerpools (L + F) in the mesic-to-wet site is about twice as high asin the mesic site (Fig. 1b).

4.2. Sensitivity of C storage to the frequency of firedisturbance

An estimate of the fire return interval in the mesic-to-wet for-est contains a large level of uncertainty, therefore the modelsensitivity to this parameter was tested. The results, pre-sented in Fig. 2 show that increasing the fire return intervalfrom 100 to 300 years had a statistically significant effect on Cstorage (the medians of the two groups differ at the 5% signif-icance level if the inner bounds of the boxes do not overlap),while the difference in the values between 300 and 500 yearswas not significant.

Fig. 2 – Relationship between fire return interval and Cstorage in O soil layers in the mesic-to-wet site. The boxeshave lines at the lower, median (bold), and upper quartilesand the whiskers show the lowest and highest values.Statistics are based on a 2000-year time series with climateand disturbance-induced fluctuations around the steadystate.

There was no direct effect of increased water content ongross primary production according to the GUESS–ROMULmodel, presumably because there was no water stress in thearea. Changing the relative contribution of different plantspecies in a stand had a negligible effect on the C dynamics ofthe O layers.

Increases in the amount of carbon (or organic matter) in soilO horizons with increases in stand age have been observed invarious locations within the boreal region (e.g. Bormann et al.,1995; Yu et al., 2002). Berg and McClaugherty (2003) argue thatsuch an accumulation can continue for up to a millenniumin undisturbed ecosystems. Furthermore, Wardle et al. (2003)speculates that forest fires are the main factors responsible forreductions in the depth of the organic soil layer in the borealregion.

The present model study shows that the lower fire fre-quency in temporarily water-logged soils may be one of themain reasons for the larger C accumulation in the past.

4.3. Comparison of mesic and mesic-to-wet sitesunder the conditions of anthropogenic fire suppression(model run II)

The model predicts that O layers have accumulated carbonfrom the time fire began to be suppressed (in 1870), and willcontinue to accumulate it until approximately 2100. However,whilst this increase is expected to be large for the mesic site,it should be almost negligible for the mesic-to-wet site. Oneexplanation for this difference is that the O layers of the soilat the mesic site are further from steady state than thoseat the mesic-to-wet site due to past disturbance. Values forthe prospective equilibrium size of O layers and estimatedsequestration rates are given in Table 2. The sequestrationrate for the mesic site corresponds well with an estimate pro-vided by the limit value concept (Akselsson et al., 2005) for thegiven geographical location. Peltoniemi et al. (2004) developeda modelling method, validated with empirical data, whichallowed changes in soil organic layer C storage following stand

e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484 481

Table 2 – Changes in carbon storage in O soil layers due to fire suppression from 1870 to 2100 per year and as aproportion of the initial storage ((O1870 − O2100)/O1870)

Site Increase in O layers C for the period 1870–2100 Steady-state O layer Cstorage O2100 (kg C m−2)

kg C m−2 year−1 O2100 − O1870/O1870 (%)

Mesic 0.0103 77.8 5.4 (4.7–6.1)Mesic-to-wet 0.0022 8.4 6.5 (6.0–7.2)

The end of the period indicates new steady state in respect to O soil layers. The size of a new equilibrium pool is given in the third column. O1870:stand C storage in O soil layers in 1870 representing the pre-industrial steady state. O2100: stand C storage in O soil layers in 2100 representingthe steady state under the conditions of anthropogenic fire suppression.

disturbance (harvesting) in southern Finland to be predicted.According to their estimate, an average annual increase of5.8 ± 1.0 g m−2 year−1 in the O soil layers should be expectedfor a stand greater than 20 years of age. However, those resultsand the model estimate presented here should not be com-pared directly, not only because initial C storage and actuallitter production differ, but also because GUESS–ROMUL wasused to simulate the average response over a number of forestpatches.

4.4. Changes due to precipitation

According to the Swedish National Forest Inventory about 59%of the forest in the country can be characterised as mesicand about 31% as mesic-to-wet. Therefore, mesic-to-wet soils,which differ from mesic soils, both in the amount of car-bon stored and their ability to sequester carbon in the future,should not be disregarded. An interesting question there-fore, is whether the proportion of wet soil in a landscaperemains constant or whether it may be altered by changingenvironmental parameters. Analysis of data from the mete-orological observation network presented in Lindstrom andAlexandersson (2004) has shown that there was an 8% increasein precipitation in the boreal region of Northern Sweden from1901–1950 to 1950–2000. At the same time no significant trendwas observed in temperature.

To assess roughly how this affected water input into the soilcould change the relative proportion of wet soils in a land-scape, a very simplified approach has been applied (Ivanov,1981). The method was previously used to obtain a dynamicestimate of the wetland area in river catchments and providedgood results. It is assumed that the elevation y of a free groundwater surface above the impermeable bed at a given point xalong the topography transect in a catchment can be describedby the parabolic equation (m):

y =√

p

k(2ax − x2) + h2, (12)

where p = P − E (cm year−1) is the climatic average excess ofprecipitation over evapotranspiration, k (cm year−1) the infil-tration coefficient, which is dependent on the soil physicalproperties, h (m) the depth from the river water surface to theimpermeable bed, and a (m) is the width of the catchmentfrom the border of the river to the watershed divide. Variablesused in (12) are illustrated in Fig. 3.

Water table depth is calculated as the difference betweenthe actual topography elevation and the height of the freeground water surface (y).

The topographical profile used here was derived from anational topographical map (Lantmateriet, 2002) of the samearea as the modelled sites. The depth of the impermeable soilhorizon was fixed at 3 m. Although the filtration coefficientk has a physical meaning it is rather empirical and here thevalue was tuned to produce reasonable values of the watertable depth. The fact that k chosen in that way falls withinthe range of observed values shows that the approach wasvalid.

Two calculations of the water table depth were made:the first with an average precipitation value for the period1901–1950 (P1, Fig. 3) and the second with an 8% increasein precipitation (P2, Fig. 3). In both cases the same value forevapotranspiration produced by the GUESS model was used,assuming no trends in temperature. The areas of the catch-ment where the water table were below and above 1 m werecalculated. With enhanced precipitation the proportion of thecatchment area with a WT < 1 m increased from 18% to 29%. Byassigning sectors with a WT < 1 m as mesic-to-wet and thosewith a WT > 1 m as mesic, the data from Table 2, combinedwith the initial C storage in O layers estimated from run I,can be used to calculate an average C sequestration valuefor the whole catchment. If the proportion of mesic-to-wetareas remains constant at 18%, the C sequestration rate will

Fig. 3 – Catchment characteristics used for the water tabledepth calculations (Eq. (12)). Dashed lines illustrate two freeground water surface positions (x, y) assuming an averageprecipitation for 1901–1950 (P1) and an 8% increase inprecipitation (P2). h is the depth from the river water surfaceto the impermeable bed and a is the width of the catchmentfrom the border of the river to the watershed divide.

482 e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

be 0.0089 kg C m−2 year−1, while an increase up to 29% of thearea will result in a higher rate of 0.0094 kg C m−2 year−1.

5. Suggestions for the extension of themodel to broader applications

A logical continuation of this work would be to model car-bon storage in the O layers of the boreal landscape across thewhole gradient, from the dry upland to wet forest. Soil mois-ture can be estimated by means of hydrological modelling ona catchment scale, or by indirect observations, for example,based on ground vegetation type. However, an attempt to testthe model for the whole range of soil moisture conditions hasrevealed some limitations. Fig. 4 shows the moisture responsefunctions gi(W) (Eqs. (5)–(9) in Table 1) for the kinetic coef-ficients, ki. As mentioned in an original SOMM formulation(Chertov, 1985), no data were found to parameterise g3 (trans-formation of plant litter into CHS, regulated by moisture) andan assumption was made that g3 = g1. Moisture response func-tions for the mineralization of litter (g1) and CHS (g2) are bothbased on data from long-term laboratory experiments and aninteresting observation is that the range of optimum moisturediffers: with a rather broad interval (30–300%) for the litterand a single peak (50%) for the CHS. The same finding, thatlow levels (<100%) of moisture decrease decomposition ratesin the CHS, was also reported by Douglas and Tedrow (1959).Another important feature, seen in Fig. 4, is that g1 decreasesto zero as the litter moisture approaches 600% (around com-plete water saturation). Therefore, for soil that is saturated forlong periods, the GUESS–ROMUL model erroneously predictsan infinite accumulation of organic matter.

The literature was searched for measured rates of lit-ter decomposition in the range of soil moisture that wouldencompass complete saturation, but with little success. In asynthesis paper, Walse et al. (1998), reviewing experimentaldata on organic matter decomposition with respect to mois-ture, included no data for moisture levels above field capacity,mainly due to a lack of such information in the original pub-lications. The curves presented in Fig. 5 were derived fromFlanagan and Veum (1974), summarising weight loss and res-piration experiments performed at tundra field sites. Largedifferences between the curves can be explained by varia-tions in litter quality and temperature. However, in most cases,

Fig. 4 – Moisture response curves for the mineralization ofplant litter (g1), and CHS (g2), transformation of litter to CHS(g3), and CHS to OLH (g4) used in the model. W is litter/soilmoisture expressed in mass%, CHS is a complex of humussubstances with underdecomposed plant debris, and OLHis organic layer humus substances.

Fig. 5 – Observed moisture response curves for plant littermineralization derived from various experimental sites atdifferent temperatures (Flanagan and Veum, 1974). W islitter/soil moisture expressed in mass%. Complete watersaturation for plant litter is around 600%.

no dramatic reductions in litter decomposition near satura-tion and even far above it were observed. In comparison withoptimum conditions a relative mineralization rate at W = 600%varied from 0.1 to 1.

Trettin et al. (2001) evaluated 12 commonly used soilorganic matter models and found that they did not describeanaerobic conditions adequately. Here the question was,how does decomposition change when the soil approacheswater saturation? None of the models, except CENTURY(Parton et al., 1987) and Wetland–DNDC (Zhang et al., 2002),extrapolate their simulations across the full range of soilmoisture conditions. However, this was done in the originalSOMM for the moisture level above field capacity (30 < W < 600,g1 = g3 = 1) and the results it provided are probably valid firstapproximations. Alternatively a linear function similar tothat presented in the Wetland–DNDC model with g1 = 0.75 atsaturation, could be used.

The model test has shown that for two moisture regimes(mesic and mesic-to-wet) compared in this paper, long-termcarbon accumulation in the O layers is not sensitive to g1 or g3

and is determined only by g2. The results presented here willtherefore also be valid with a new g1 parameterisation.

6. Concluding remarks

The importance of soil moisture for the simulation of soilorganic matter dynamics was shown in this study and shouldnot be neglected when climate change issues are considered.The GUESS–ROMUL model is applicable for further investi-gation in this area and coupling with catchment hydrologymodels appears to be promising. However, the model requiressome improvement concerning the dependence of the kineticcoefficients ki on soil moisture levels.

r e f e r e n c e s

Agren, G.I., Hyvonen, R., 2003. Changes in carbon stores inSwedish forest soils due to increased biomass harvest andincreased temperatures analysed with a semi-empiricalmodel. For. Ecol. Manage. 174 (1–3), 25–37.

e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484 483

Akselsson, C., Berg, B., Meentemeyer, V., Westling, O., 2005.Carbon sequestration rates in organic layers of boreal andtemperate forest soils-Sweden as a case study. Global Ecol.Biogeogr. 14 (1), 77–84.

Beldring, S., Gottschalk, L., Seibert, J., Tallaksen, L.M., 1999.Distribution of soil moisture and groundwater levels at patchand catchment scales. Agric. For. Meteorol. 98/99, 305–324.

Berg, B., McClaugherty, C., 2003. Plant litter. In: Decomposition,Humus Formation, Carbon Sequestration. Springer Verlag,Heidelberg, Berlin, p. 286.

Bhatti, J.S., Apps, M.J., Jiang, H., 2002. Influence of nutrients,disturbances and site conditions on carbon stocks along aboreal forest transect in Central Canada. Plant Soil 242 (1),1–14.

Bormann, B.T., Spaltenstein, H., McClellan, M.H., Ugolini, F.C.,Cromackjr, K., Nay, S.M., 1995. Rapid soil development afterwindthrow disturbance in pristine forests. J. Ecol. 83, 747–757.

Carrasco, J.J., Neff, J.C., Harden, J.W., 2006. Modeling physical andbiochemical controls over carbon accumulation in a borealforest soil. J. Geophys. Res. 111, G02004,doi:10.1029/02005JG000087.

Chertov, O.G., 1981. Ecology of Forest Lands. Nauka, Leningrad, p.192 (in Russian).

Chertov, O.G., 1985. Imitatsionnaya model mineralizatsii igumifikatsii lesnogo opada i podstilki. Zh. Obshch. Biol. 46 (6),794–804 (in Russian).

Chertov, O.G., Komarov, A.S., 1997. SOMM: A model of soil organicmatter dynamics. Ecol. Modell. 94 (2/3), 177–189.

Chertov, O.G., Komarov, A.S., Crocker, G., Grace, P., Klir, J.,Korschens, M., Poulton, P.R., Richter, D., 1997. Simulatingtrends of soil organic carbon in seven long-term experimentsusing the SOMM model of the humus types. Geoderma 81(1/2), 121–135.

Chertov, O.G., Komarov, A.S., Nadporozhskaya, M., Bykhovets,S.S., Zudin, S.L., 2001. ROMUL—a model of forest soil organicmatter dynamics as a substantial tool for forest ecosystemmodeling. Ecol. Modell. 138 (1–3), 289–308.

Chertov, O., Lauren, A., Nadporozhskaya, M., Palviainen, M.,Tanskanen, H., 2005. Modelling soil organic matter dynamicsin organic horizons in natural ecosystems: some missedfunctions. In: Proceedings of the Fifth European Conferenceon Ecological Modelling, Pushchino, Russia, September 19–23,pp. 37–38.

Cox, P.M., Betts, R.A., Bunton, C.B., Essery, R.L.H., Rowntree, P.R.,Smith, J., 1999. The impact of new land surface physics on theGCM simulation of climate and climate sensitivity. ClimateDynamics 15, 183–203.

Davidson, E.A., Janssens, I.A., 2006. Temperature sensitivity ofsoil carbon decomposition and feedbacks to climate change.Nature 440, 165–173, doi:10.1038/nature04514.

Douglas, L.A., Tedrow, J.C., 1959. Organic matter decompositionrates in arctic soils. Soil Sci. 88, 305–312.

Engelmark, O., 1987. Fire history correlations to forest type andtopography in northern Sweden. Ann. Bot. Fenn. 24, 317–324.

FAO, 1991. The Digitized Soil Map of the World (Release 1.0), vol.67/1. Food and Agriculture Organization of the United Nations.

Flanagan, P.W., Veum, A.K., 1974. Relationships betweenrespiration, weight loss, temperature and moisture in organicresidues on tundra. In: Holdring, A.J. (Ed.), Soil Organisms andDecomposition in Tundra: Proceedings of the Microbiology,Decomposition and Invertebrate Working Groups Meeting.University of Alaska, Fairbanks, AK, USA, August 1973.LiberTryck, Stockholm, pp. 249–277.

Gerten, D., Schaphoff, S., Haberlandt, U., Lucht, W., Sitch, S., 2004.Terrestrial vegetation and water balance—hydrologicalevaluation of a dynamic global vegetation model. J. Hydrol.286, 249–270.

Global Soil Data Task Group, 2000. Global girded surfaces ofselected soil characteristics (IGBP-DIS). [Global girded surfacesof selected soil characteristics (InternationalGeosphere-Biosphere Programme—Data and InformationSystem)]. Data set. Available on-line[http://www.daac.ornl.gov/] from Oak Ridge NationalLaboratory Distributed Active Archive Center, Oak Ridge, TN,USA.

Haxeltine, A., Prentice, I.C., 1996. BIOME3: an equilibriumterrestrial biosphere model based on ecophysiologicalconstraints, resource availability, and competition amongplant functional types. Global Biogeochem. Cycles 10, 693–709.

Haxeltine, A., Prentice, I.C., Cresswell, I.D., 1996. A coupledcarbon and water flux model to predict vegetation structure. J.Vegetat. Sci. 7 (5), 651–666.

Hesselman, H., 1926. Studier over barrskogens humustacke, dessegenskaper och beroende av skogsvarden. Stockholm, 552 pp.(in Swedish).

Hyvonen, R., Berg, M.P., Agren, G.I., 2002. Modelling carbondynamics in coniferous forest soils in a temperature gradient.Plant Soil 242, 33–39.

Ivanov, K.E., 1981. Water Movement in Mirelands. AcademicPress, London, p. 276.

Koca, D., Sykes, M., Smith, B., 2006. Modelling regional climatechange effects on Swedish ecosystems. Clim. Change 78 (2),381–406.

Kurz, W.A., Apps, M.J., 1996. Retrospective assessment of carbonflows in Canadian boreal forest. In: Apps, M.J., Price, D.T.(Eds.), Forest ecosystems, Forest Management and the GlobalCarbon Cycle: Proceedings of the NATO Advanced ResearchWorkshop “The Role of Global Forest Ecosystems and ForestResource Management in the Global Cycle”. Banff, Canada,September 12–16, 1994. Springer, Berlin, pp. 173–182.

Lantmateriet, 2002. Terrangkartan, 21J NO Vindeln, 1:50000.Lindstrom, G., Alexandersson, H., 2004. Recent mild and wet

years in relation to long observation records and futureclimate change in Sweden. AMBIO 33 (4/5), 183–186.

Liski, J., Ilvesniemi, H., Makela, A., Westman, C.-J., 1999. CO2

emissions from soil in response to climatic warming areoverestimated—the decomposition of old soil organic matteris tolerant of temperature. AMBIO 28 (2), 171–174.

Malkonen, E., 1974. Annual primary production and nutrientcycle in some Scots pine stands. Commun. Instit. For. Fenn. 84(5), 1–87.

New, M., Hulme, M., Jones, P.D., 1999. Representing twentiethcentury space-time climate variability. Part 2. Development of1901–1996 monthly grids of terrestrial surface climate,unpublished data.

Niklasson, M., Granstrom, A., 2000. Number and sizes of fires:long-term spatially explicit fire history in a Swedish boreallandscape. Ecology 81 (6), 1484–1499.

Parton, W.J., Stewart, J.W.B., Cole, C.V., 1987. Dynamics of C, N, S,and P in grassland soils: a model. Biogeochemistry 5, 109–131.

Peltoniemi, M., Makipaa, R., Liski, J., Tamminen, P., 2004. Changesin soil carbon with stand age—an evaluation of a modellingmethod with empirical data. Global Change Biol. 10,2078–2091.

Sitch, S., Smith, B., Prentice, I.C., Arneth, A., Bondeau, A., Cramer,W., Kaplan, J.O., Levis, S., Lucht, W., Sykes, M.T., Thonicke, K.,Venevsky, S., 2003. Evaluation of ecosystem dynamics, plantgeography and terrestrial carbon cycling in the LPJ DynamicGlobal Vegetation Model. Global Change Biol. 9, 161–185.

Smith, B., Prentice, I.C., Sykes, M.T., 2001. Representation ofvegetation dynamics in the modelling of terrestrialecosystems: comparing two contrasting approaches withinEuropean climate space. Global Ecol. Biogeogr. 10 (6), 621–637.

Thonicke, K., Venevsky, S., Sitch, S., Cramer, W., 2001. The role offire disturbance for global vegetation dynamics: coupling fire

484 e c o l o g i c a l m o d e l l i n g 2 0 4 ( 2 0 0 7 ) 475–484

into a Dynamic Global Vegetation Model. Global Ecol.Biogeogr. 10 (6), 661–677.

Trettin, C.C., Song, B., Jurgenses, M.F., Li, C., 2001. Existing soilcarbon models do not apply to forested wetlands. GeneralTechnical Report SRS-46, USDA Forest Service, SouthernResearch Station.

Trumbore, S.E., Harden, J.W., 1997. Accumulation and turnover ofcarbon in organic and mineral soils of the BOREAS northernstudy area. J. Geophys. Res. 102, 817–830.

Wagner, W., Scipal, K., Pathe, C., Gerten, D., Lucht, W., Rudolf, B.,2003. Evaluation of the agreement between the first globalremotely sensed soil moisture data with model andprecipitation data. J. Geophys. Res. 108 (D19), 4611,doi:10.1029/2003JD003663.

Walse, C., Berg, B., Sverdrup, H., 1998. Review and synthesis ofexperimental data on organic matter decomposition withrespect to the effect of temperature, moisture, and acidity.Environ. Rev. 6, 25–40.

Wardle, D.A., Hornberg, G., Zackrisson, O., Kalela-Brundin, M.,Coomes, D.A., 2003. Long-term effect of wildfire on ecosystemproperties across an island area gradient. Science 300 (5621),972–975.

Wirth, C., Czimczik, C.I., Schulze, E.-D., 2002. Beyond annualbudgets: carbon flux at different temporal scales in fire-proneSiberian Scots pine forests. Tellus B 54 (5), 611–630.

Wolf, A., Blyth, E., Harding, R., Jacob, D., Keup, E., Goettel, H.Callaghan, T., in revision. Sensitivity of an ecosystem modelto hydrology and temperature. Submitted to ClimaticChange—Special Issue on climate change in the BarentsRegion.

Yu, Z., Apps, M.J., Bhatti, J.S., 2002. Implication of floristic andenvironmental variation for carbon cycle dynamics in borealforest ecosystems of Central Canada. J. Vegetat. Sci. 13,327–340.

Zaehle, S., Sitch, S., Smith, B., Hatterman, F., 2005. Effects ofparameter uncertainties on the modeling of terrestrialbiosphere dynamics. Global Biogeochem. Cycles 19, GB3020,doi:10.1029/2004GB002395.

Zhang, Y., Li, C., Trettin, C.C., Li, H., Sun, G., 2002. An integratedmodel of soil, hydrology, and vegetation for carbon dynamicsin wetland ecosystems. Global Biogeochem. Cycles 16, 1061,doi:10.1029/2001GB001838.

Zobler, L., 1986. A world soil file for global climate modelling.NASA Tech. Memorandum 87802, 32.

I

II

I

Variations in net ecosystem exchange of carbon dioxide in a boreal

mire: Modeling mechanisms linked to water table position

Alla Yurova,1 Annett Wolf,1,2 Jorgen Sagerfors,3 and Mats Nilsson3

Received 5 October 2006; revised 13 February 2007; accepted 7 March 2007; published 19 May 2007.

[1] In mires, which occupy large areas of the boreal region, net ecosystem CO2 exchange(NEE) rates vary significantly over various timescales. In order to examine the effect ofone of the most influencing variables, the water table depth, on NEE the generalecosystem model GUESS-ROMUL was modified to predict mire daily CO2 exchangerates. A simulation was conducted for a lawn, the most common microtopographicalfeature of boreal oligotrophic minerotrophic mires. The results were validated againsteddy covariance CO2 flux measurements from Degero Stormyr, northern Sweden,obtained during the period 2001–2003. Both measurements and model simulationsrevealed that CO2 uptake was clearly controlled by interactions between water table depthand temperature. Maximum uptake occurred when the water table level was between10 and 20 cm and the air temperature was above 15�C. When the water table was higher,the CO2 uptake rate was lower, owing to reduced rates of photosynthetic carbon fixation.When the water table was lower, NEE decreased owing to the increased rate ofdecomposition of organic matter. When the water table level was between 10 and 20 cm,the NEE was quite stable and relatively insensitive to both changes within this rangeand any air temperature changes above +15�C. The optimal water table level range forNEE corresponds to that characteristic of mire lawn plant communities, indicatingthat the annual NEE will not change dramatically if climatic conditions remain within theoptimal range for the current plant community.

Citation: Yurova, A., A. Wolf, J. Sagerfors, and M. Nilsson (2007), Variations in net ecosystem exchange of carbon dioxide in a

boreal mire: Modeling mechanisms linked to water table position, J. Geophys. Res., 112, G02025, doi:10.1029/2006JG000342.

1. Introduction

[2] Boreal mires contain a large carbon pool, accumulatedover millennia, which is significant at both regional andglobal scales [e.g., Gorham, 1991; Turunen et al., 2002].Water balance plays a crucial role among the factorsdetermining mire growth and development [e.g., Ivanov,1981; Hilbert et al., 2000]. Therefore climatic variations atdifferent timescales may affect the rate of CO2 uptake andthe overall C dynamics of mires [Moore et al., 1998]. In thiscontext, it is very important to incorporate mire ecosystemsinto the framework of dynamic vegetation modeling, a state-of-the-art approach to carbon balance assessments andpredictions.[3] The fundamental ecosystem processes (including pho-

tosynthesis, respiration, and soil organic matter dynamics)are common to mires and other terrestrial ecosystems, so itshould be possible to adapt a general ecosystem model to

mires. However, mires have highly specific self-regulatoryproperties that arise from the interactions amongst theirhydrological and biological processes on all scales from themicrotopographical unit to the whole ecosystem [e.g.,Belyea and Baird, 2006]. Direct measurements of changesin NEE of individual microforms associated with changes inthe water table level are used to investigate the hydrologicalcontrol of mire CO2 exchange rates. A number of studies[e.g., Alm et al., 1999; Waddington and Roulet, 2000;Bubier et al., 2003; Lafleur et al., 2003; Nykanen et al.,2003; Syed et al., 2006] have demonstrated, using chamberor eddy covariance measurements over different microtopo-graphical features, that responses of NEE to changes in thewater table depth can vary widely. Analysis of data using anecophysiological model might help to explain the variabil-ity. However, to date, relatively few process-based generalecosystem models, particularly ones specific to mires andrun using daily time steps, have been applied to predict NEEand validated by field measurements from actual mire sites[Potter et al., 2001; Frolking et al., 2002; Zhang et al.,2002].[4] Here we present a modified general ecosystem simu-

lator, GUESS [Smith et al., 2001], which has been previ-ously used to simulate forest and grassland ecosystemdynamics, coupled to the soil organic matter dynamicsmodel ROMUL [Chertov et al., 2001]. The GUESS-ROMUL model [Yurova and Lankreijer, 2007] was restruc-

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, G02025, doi:10.1029/2006JG000342, 2007ClickHere

for

FullArticle

1Department of Physical Geography and Ecosystems Analysis, LundUniversity, Lund, Sweden.

2Now at the Department of Forest Ecology, Institute of TerrestrialEcosystems, Zurich, Switzerland.

3Department of Forest Ecology, Swedish University of AgriculturalSciences, Umea Sweden.

Copyright 2007 by the American Geophysical Union.0148-0227/07/2006JG000342$09.00

G02025 1 of 13

tured so that it could be applied to Sphagnum-dominatedmicrotopographical features of mires. We focused on mod-eling the CO2 exchange of an area dominated by lawn plantcommunities at Degero Stormyr, a minerotrophic mire innorthern Sweden. The results were compared with dailyCO2 fluxes for 2001–2003 obtained from eddy-covariancetower measurements. Furthermore, we derived and analyzedNEE response functions, from both the model and empiricaldata, in relation to the water table depth.

2. Model

[5] The modeling framework used here was the coupleddynamic vegetation-soil organic matter model GUESS-ROMUL. A description of the GUESS is given by Smithet al. [2001], the ROMUL model is summarized by Chertovet al. [2001], and the combined GUESS-ROMUL system isdescribed by Yurova and Lankreijer [2007]. Here we onlydescribe those processes within the model that we revised ormodified in order to simulate mire ecosystems.[6] The basic unit in the GUESS model is the plant

functional type (PFT), characterized by its physiological,morphological, and phenological attributes. To simulateboreal mires, we introduced Sphagnum mosses as a newPFT within the model. Major changes to the model were asfollows:[7] 1. A new water balance scheme was introduced.[8] 2. The photosynthesis model was modified to account

for (1) regulation of Rubisco activity by the tissue N contentand (2) the dependency of moss photosynthetic activity onwater content. Parameters in the equations describing auto-trophic respiration were also revised.[9] 3. Specific coefficients were used to describe decom-

position of the Sphagnum litter and peat. Parameterizationof decomposition under anoxic conditions was included in asoil organic matter dynamics model.[10] The changes are described in more detail below.

Model parameters and constants are summarized in Table 1.

2.1. Hydrology

[11] An important feature of the mire water regime is thatthe water table is so close to the surface that there is aconstant supply (via capillary action) of water to theunsaturated zone; that is, water can move from the watertable up to the peat surface.[12] Two layers can be distinguished within the peat

profile: the upper, temporarily aerated layer consistinglargely of living and lightly decomposed plant material(acrotelm), and the permanently saturated, lower zone,consisting largely of compact, relatively highly decomposedplant material (catotelm). Owing to the very strong reduc-tion in hydraulic conductivity from the acrotelm to thecatotelm [Ivanov, 1981] it can be assumed that the move-ment of water in the catotelm is negligible compared to itsmovement in the acrotelm.[13] Here we used the set of equations from the Mixed

Water and Heat model (MMWH) [Granberg et al., 1999],with minor modifications, to describe the water balance inthe acrotelm. The MMWH model was specifically devel-oped to reconstruct the water table position in a borealmixed mire and was validated for the lawn plant communityat Degero Stormyr (the mire used in the present study).

[14] The total volume of water in the profile, from thepeat surface to the depth of the permanently saturated zone(catotelm, zcat,, cm) at the current water table depth (zwt,,cm), Vtot (cm), was defined as

Vtot ¼ �f zcat � zwtð Þ þZ0zwt

qus zð Þdz; ð1Þ

where the first term describes water volume in the saturatedzone (cm) in relation to the thickness of the saturated layer -(zcat � zwt) and porosity (f). The second term is the watercontent in the unsaturated zone (qus(z)) integrated from thecurrent water table depth to the surface. (The signconvention in this paper is that water table depth is notedwith a negative value when water table is situated below thepeat surface). The function qus(z) describes the gradualchange in the capillary water content from its maximumnear the water table (supply) to its lowest value at theevaporating and transpiring vegetated surface. Here we usedan equation developed by van Genuchten [1980], simplifiedby setting the residual water content to zero [Weiss et al.,1998],

qus zð Þ ¼ f 1þ a � zwt � zð Þð Þð Þn½ ��1þ1=n: ð2Þ

[15] The pressure head (in cm H2O) in the van Genuchtenequation was replaced by the distance to the water table�(zwt � z), assuming that the water profile in the unsatu-rated zone is in equilibrium. The empirical parameters n anda were derived from a study of a large number of peat corescollected in Finland [Weiss et al., 1998],

n ¼ 1:49� 3:56 � rþ 13:2 � r2 þ 0:0027 � ca� 0:037 � l1 ð3Þ

log10 a ¼ �51:8 � r2 � 0:0057 � ca� 0:01 � sphþ 0:53 � l1; ð4Þ

where r is peat bulk density (g cm�3), ca and sph are Carexand Sphagnum residue contents in the peat (%), respec-tively, and l1 has a value of 1 for the peat at a depth of�10–0 cm and 0 otherwise.[16] The water balance equation was calculated at daily

intervals, to identify changes in water storage in the profile(DS, cm d�1),

DS ¼ P � E � q; ð5Þ

where P is rain or snowmelt, E is evapotranspiration and qis discharge.[17] The description of saturated flow in the acrotelm is

based on work by Ivanov [1981],

q ¼ lq � i � Tr � zcat � zwtð Þ; ð6Þ

where q is specific discharge (cm d�1), i is a local slope ofthe water table (assumed to be equal to the topographicalslope, dh

dl), Tr (cm d�1) is transmissivity coefficient, and

lq (cm�1) is a lumped parameter. Zero recharge in the areaand no horizontal flow in the unsaturated zone are assumed.The peat transmissivity is approximated as an exponential

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

2 of 13

G02025

function of the water table depth, with coefficients based ondata for the lawn [Ivanov, 1981]: This takes into account thepossibility of a runoff feedback mechanism, for example,the ability of the mire ecosystem to regulate its waterbalance by runoff increasing when the water table is high.[18] Evapotranspiration is simulated differently here

from the original formulation in the MMWH model.First, potential evapotranspiration (Eq cm d�1) is estimatedusing the Priestley-Taylor equation as in the standardGUESS model [e.g., Gerten et al., 2004]. Then an expo-nential function is used to adjust actual evapotranspiration(E, cm d�1), to correspond to the current water table depthin the mire.

E ¼ Eq � exp lE � zwtð Þ: ð7Þ

[19] The empirical parameter lE (cm�1) was derived fromevapotranspiration measurements obtained from three Fin-ish mires, using sample containers with different waterlevels [Virta, 1966].[20] Knowing the change in water storage from equation (5),

we obtain the new value for the total water volume in theprofile (Vtot) and solve equation (1) for zwt to determine thenew water table depth for the current day.[21] Acrotelm depth was assumed to be constant (�30 cm)

over the simulation period (2001–2003).

2.2. Moss Production

2.2.1. Photosynthesis[22] The C3 photosynthesis in GUESS is formulated

following a simplified version of the Farquhar model

[Farquhar et al., 1980; Collatz et al., 1991; Haxeltine andPrentice, 1996]. The rate of photosynthesis can be limitedby either the electron transport rate or Rubisco activity. Inthe standard version of GUESS, nitrogen supply is assumedto be always sufficient to provide optimum Rubisco activityfor the given PAR level [Haxeltine and Prentice, 1996]. Inboreal mires, however, nitrogen is often limiting [e.g.,Malmer and Nihlgard, 1980]. We therefore used an equa-tion presented by Friend [1995] to calculate maximumRubisco activity, Vmax (kgC m�2 d�1) as a function of nitro-gen concentration in moss tissues N.

Vmax ¼ 1036:8 � kc 6:25 � 8=550ð Þ � fN ;Rub � N ; ð8Þ

where kc is the temperature-dependent Rubisco carboxylationturnover rate (molCO2 mol�1sites s�1) and fN,Rub is thefraction of total nitrogen content bound in Rubisco (dimen-sionless). The nitrogen concentration, in (N, kg m�2)recalculated from C:N ratios, and fN,Rub, were specifiedas model input parameters. The constant 550 has units(kg mol�1), and 1036.8 (12�10�3kg mol�186400 s d�1) is aconversion factor from the original units (mol CO2 m

�2 s�1)used by Friend [1995] to kgC m�2 d�1.[23] The photosynthetic activity of mosses also depends

on their water content [Titus and Wagner, 1984; Williamsand Flanagan, 1996; Schipperges and Rydin, 1998]. Whenthe water content of the mosses is low, desiccation adverselyaffects their photosynthetic capacity, and when the watercontent is too high their photosynthetic rates decreasebecause the rate of transport of CO2 is lower in water thanin air [Silvola, 1990]. The threshold values are species-

Table 1. Conventional GUESS-ROMUL Parameters and Constants Used for the Mire Simulation and New Parameters Included in

Additional Model Equations

Function Symbol Value Units Description Source

Photosynthesis wmin 0 gH20/gC determine the photosynthesisresponse curve depending onthe Sphagnum water content

Schipperges andRydin [1998]

wlow 12 gH20/gCwhigh 25 gH20/gCwmax 70 gH20/gCNRub 0.18 proportion of N in Rubisco Friend [1991]cn 50 C:N ratio in Sphagnum tissues Malmer and Wallen [2005]

Moss respiration b 0.03 shaping parameter in equation (11) simultaneous calibration of b, rmand rg based on data fromWilliams and Flanagan [1998]

rm 2.0 shaping parameter in equation (12)rg 0.5 shaping parameter in equation (10)

Turnover tr 1.0 proportion of the biomasstransferred annually to the litter

Nungesser [2003]

Litter and SOM dynamicsa k1 1.6�10�3 day�1 litter(L) mineralization coefficient Verhoeven and Toth [1995]k2 4.0�10�4 day�1 SOM (F) mineralization coefficienta O. G. Chertov

(personal communication, 2004)kan 0.50 ratio of aerobic to anaerobic

decomposition rate in the acrotelmMoore and Dalva [1997]

kcat 1.0�10�4 year�1 decay rate in the catotelm Clymo [1984]rL 0.02 g cm�3 moss litter density Nungesser [2003,

and references therein]rF; rH 0.02; 0.05 g cm�3 peat density at various depths Granberg et al. [1999]

Hydrology f 0.92; 0.98 porosity: acrotelm; catotelm Granberg et al. [1999]

zcat �30 cm acrotelm depth Granberg et al. [1999]sph 90 % proportion of Sphagnum and

Carex remains in the peatM. Nilsson(personal communication, 2006)

ca 3 %lE 0.02 cm�1 shaping parameter in equation (7) based on work by Virta [1966]

aSOM is soil organic matter.

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

3 of 13

G02025

dependent and not all mosses show strong reductions inphotosynthesis when water contents increase to high levels.Here we simulated the potential photosynthetic assimilationrate further multiplying it by the empirical linear functionfwc (dimensionless),

fwc ¼

w� wminð Þ= wlow � wminð Þ; wmin < w � wlow

1; wlow < w � whigh

w� wmaxð Þ= whigh � wmax

� �; whigh < w < wmax

0;w � wmax;w � wmin

8>>>>>>>>>><>>>>>>>>>>:

; ð9Þ

where w is water content (gH2O/gC).[24] Parameters wmin, wlow, whigh and wmax (gH2O/gC)

were derived from the empirically derived curves publishedby Schipperges and Rydin [1998] for the typical lawnspecies Sphagnum balticum, which dominates the lawns atDegero Stormyr. The water content of the apical part of theSphagnum shoots was calculated using equation (2), inte-grated from the peat surface down to a fixed depth (�2 cm).Photosynthesis was assumed to be totally suppressed undersnow cover and when the surface temperature falls belowzero.2.2.2. Respiration[25] For Sphagnum, the differentiation between

photosynthetic and nonphotosynthetic tissues is somewhatobscure. However, Williams and Flanagan [1998] havepostulated that respiration from both compartmentsshould be specified. We therefore calculate total respirationR (kg C m�2 d�1) as follows.

R ¼ rg A� Rnon�phot � Rphot

� �þ Rnon�phot þ Rphot ; ð10Þ

where the first term describes growth respiration and thelast two maintenance respiration from nonphotosyntheticRnon�phot: (kgC m�2 d�1) and photosynthetic Rphot:(kgC m�2 d�1) tissues, respectively. A is the photosyntheticCO2 assimilation rate (kgC m�2 d�1) and rg is a growthrespiration coefficient (1-growth yield).[26] Maintenance respiration is calculated following

Haxeltine and Prentice [1996] and Sitch et al. [2003].

Rphot ¼ b � Vmax ð11Þ

Rnon�phot ¼ rm � kr � g Tairð Þ �M=cn; ð12Þ

where b and rm are scaling coefficient, M is total biomass(kg C m�2), cn is C:N ratio, g(Tair) is a temperatureresponse function described by the modified Arrheniusequation [Lloyd and Taylor, 1994], and kr (day�1) is therespiration rate for a 10�C baseline.[27] We were unable to find any studies in which the

respiration rate of Sphagnum species was treated in theframework of the growth-maintenance paradigm. However,dark respiration rates of Sphagnum samples have beenmeasured simultaneously with net assimilation rates byWilliams and Flanagan [1998]. Using the values of envi-

ronmental parameters presented by the cited authors, weconducted an optimization of three model parameters(b, rg, rm) to achieve the same ratio of respiration to netassimilation as measured by Williams and Flanagan [1998].Different combinations of b, rg and rm were potentiallypossible, so we selected the one that conformed to thestandard parameter ratio used by GUESS (Table 1).2.2.3. Growth[28] At the end of each year, NPP (net primary produc-

tion) was converted to biomass increments after deducting10% to account for reproduction (converted into the litterpool). The rest was allocated to new moss biomass. DailyNPP can be negative for mosses, if respiration exceedsphotosynthesis, for example, on days with low light andsufficiently high temperatures. Annual litter production wasassumed to equal NPP [Nungesser, 2003].

2.3. Soil Organic Matter Dynamics

[29] It is widely recognized that the decomposition ofSphagnum moss litter is slow in comparison to that of otherplant genera [Clymo, 1965]. In addition, decomposition alsoslows under anoxic conditions.[30] The ROMUL model used in this study specifies the

following three pools of soil organic matter: fresh litter (L),complexes of humic substances with recalcitrant remains ofplant debris (F); and humic substances that are in late stagesof decomposition and/or bound to the mineral soil (H). Therate of mineralization and the flow of matter betweendifferent pools is described by the kinetic coefficients ki(i = 1. . .6), which are dependent both on the litter qualityand environmental conditions (temperature and moisture).To adapt the ROMUL model to the mire sites, we consid-ered the F pool to be represented by the acrotelm peat. Notransport of organic material to the mineral horizon wasconsidered. The H pool was interpreted as being thepermanently saturated catotelm peat.[31] In their comprehensive survey of the decomposabil-

ity of the litter of various plant growth forms (Sphagnum,evergreen and deciduous shrubs, graminoids and forbs)from mires in different geographical regions, Dorrepaal etal. [2005] concluded that the standard chemical variables(the C:N ratio as used in ROMUL, or the lignin:N ratio)explain a very small proportion of the variation in decom-position rates, and that life form per se may be a betterpredictor. On the basis of this conclusion, we chose thespecific mineralization rate k1 = 0.0016 day�1 for theSphagnum litter (L pool), derived from data presented byVerhoeven and Toth [1995] from 73-day-long laboratorydecomposition experiments. The surface peat mineralizationrate k2 = 0.0004 d�1 for the F pool was obtained from recentlaboratory experiments with a poor fen peat under aerobicconditions (O. G. Chertov, personal communication, 2004).The model coefficient k3 remained as standard for above-ground tissues [Chertov et al., 2001]. The fixed acrotelmdepth (zcat = �0.3 m) determines the annual amount of thematerial that submerges into the permanently saturated zoneK (kgC m�2),

K ¼ rF �zcat � L=rL þ F=rFð Þð Þ; ð13Þwhere L and F (kgC m�2) are pool sizes, rL, rF (kgC m�3)are the densities of the corresponding pools and zcat isexpressed in meters (m).

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

4 of 13

G02025

[32] The rate of catotelm decay, kcat, was taken from theliterature [e.g., Clymo, 1984; Tolonen et al., 1992; Korholaet al., 1996; Kuhry and Vitt, 1996]. We also introduced anew coefficient, kan, defined as the ratio of anaerobic toaerobic decomposition in the acrotelm. To account for theproportion of the acrotelm peat under the water table at anygiven time, the mineralization rate was reduced by multi-plying it by kan. The value of kan was taken from Moore andDalva [1997].[33] It should be noted that the suppressed decay rates

found under periodically anoxic conditions in the surfacepeat (as described among other things by Moore and Dalva[1997] and Bergman [1998]) are much higher than theextremely low decay rates found in deep, permanentlywaterlogged peat, which, according to peat core analysis,could be up to a thousandfold lower [e.g., Clymo, 1984;Korhola et al., 1996], as hypothesized by Freeman et al.[2001] owing to the inhibition of phenol oxidases by bothlow partial oxygen pressures and high concentrations ofphenolic compounds.[34] The time steps for the soil organic matter simulations

and the transport of material to the catotelm were 1 day and1 year, respectively.

2.4. Modeling Protocol

[35] The model was run for the period 2001–2003,following a 7000-year spinup: roughly the time sinceDegero Stormyr was established [Malmstrom, 1923]. Theclimatic data input for the spinup period (monthly airtemperature, precipitation and cloud cover) were repeated60-year series, using real figures for 1901–1961 derivedfrom the CRU05 database (provided by the Climate ResearchUnit, University of East Anglia; M. New et al., Representingtwentieth century space-time climate variability: Part 2.Development of 19011996 monthly grids of terrestrial sur-face climate, 1999, unpublished manuscript) for the coordi-nates of Degero Stormyr. The daily meteorological data (airtemperature and precipitation) from the station at DegeroStormyr mire were used for the simulation period 2001–2003. Global solar radiation data were used and PAR wascalculated as 50 percent of the global radiation. For the

current study, we excluded all plant functional types exceptmosses from the simulation. The model was run over dailytime intervals.[36] The following model simulations were conducted:[37] 1. A run with a simulated water table to assess

overall model performance.[38] 2. A run with the actual measured water table depths

instead of the modeled values, to estimate the level ofuncertainty added to the model as a result of unknownsassociated with its hydrology section. Data from this runwere also used to analyze the dependency of NEE on thewater table depth.[39] 3. Two test runs excluding the effect of water depth

on (1) photosynthesis and, subsequently, (2) peat decom-position rate.[40] 4. A series of Monte Carlo runs to determine the

model’s sensitivity to specific parameters and constants(Table 2).

3. Site Description

[41] The EC measurement tower for monitoring CO2

exchange at the oligotrophic minerotrophic mire DegeroStormyr (64�110N, 19�330E) is part of the Nordic eddy-fluxmeasurement network, NECC. The mire is located in aboreal forested zone in a cold temperate humid climate. Thelength of the period with air temperatures above 0�C isabout 200 days per year. The mean annual temperature atKulbacksliden, a meteorological station near the mire, is+1.2�C, the mean July temperature is +14.7�C and themean January temperature is �12.4�C (1961–1990)[Alexandersson et al., 1991]. The mean annual precipita-tion is 523 mm (1961–1990) [Alexandersson et al., 1991],34% of which falls as snow.[42] Degero was initiated approximately 7–8 kyrs BP and

the current peat depth is mostly between 3 and 4 m, withmaximum depths up to 8 m [Malmstrom, 1923]. Near thetower the peat depth is 5 m [Granberg et. al., 1999]. Thewater table position usually varies from 5 to 19 cm belowthe peat surface level (90th percentile to 10th percentile),but the mire is flooded every spring during snowmelt while

Table 2. Range of Parameters and Constants Used in the Sensitivity Analysis

Parameter UnitsValue in

Standard Model Lower Range Higher Range Source and Comments

Productionwmin gH20/gC 0 0 4 Nungesser [2003, and references therein],

Schipperges and Rydin [1998]wlow gH20/gC 12 10 16whigh gH20/gC 25 16 30wmax gH20/gC 70 28 70NRub 0.18 0.18 0.22 Friend [1991], not specifically for mossb 0.03 0.01 0.045 assumedrm 2.0 1.0 5.0 assumedrg 0.5 0.15 0.50 assumed

Decompositionk1 day�1 1.6�10�3 4.0�10�4 1.0�10�2 calibration based on work by

Nilsson and Bohlin [2001]k2 day�1 4.0�10�4 9.6�10�6 6.0�10�4 Waddington et al. [2001], McKenzie et al. [1998]k3 day�1 1.1�10�2 6.1�10�3 1.5�10�2 Chertov [1985], not specifically for mosskcat year�1 1.0�10�4 8�10�5 5�10�4 Clymo [1984], Korhola et al. [1996]kan 0.50 0.07 1.0 Bergman [1998], Moore and Dalva [1997]

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

5 of 13

G02025

the peat is still frozen. Species dominating the lawn plantcommunity are Sphagnum balticum Russ. C. Jens. withsparse Sphagnum lindbergii. The vascular plant communityis dominated by Eriophorum vaginatum L., Vacciniumoxycoccos L., Andromeda polifolia L., with Scheuchzeriapalustris L occurring more sparsely.

3.1. Measurements

[43] The eddy covariance system for measuring CO2,water, and energy fluxes is installed on a tower, and has afetch area that is dominated by the lawn microtopographicalfeature. The energy balance closure of the system was 96%over the 3 years, indicating that the data are reliable[Sagerfors et al., 2007]. The water table relative to the peatlevel is monitored automatically using a float and counter-weight system [Roulet et al., 1991]. In addition, sets ofstandard meteorological observations (global and net solarradiation, air and peat temperature, air humidity and snowdepth) are recorded at the site [Sagerfors et al., 2007]. Dataseries for three full years, 2001–2003, were used for modelvalidation.

3.2. Evaluation of the Model Performance, StatisticalData Processing, and Sensitivity Analysis

[44] Simulated water table depth (WT) and net ecosystemexchange (NEE) data were compared with observed valueson a daily basis over the period 2001–2003. In accordancewith the recommendations of Willmott [1982] we calculatedthe following quantitative measures to evaluate the modelperformance: the mean bias error (MBE), the mean squareerror (MSE), which was also partitioned into ‘‘systematic’’(MSEs) and ‘‘unsystematic’’ (MSEu) components, and thesummary measure the Willmott index of agreement(d, defined in Appendix A).[45] To analyze the relationship between the response

variable NEE (y) and the explanatory variable, depth tothe water table in the mire (x), a least squares second-orderregression (y = ax2 + bx + c +e) was applied as the bestpolynomial model. The t-statistic and corresponding p-valuewere used to estimate the significance of the second-orderregression coefficient a (Ho: a = 0).[46] Regression lines based on the observed data and

outcomes of the test runs were compared with the standardmodel simulation to test the hypothesis of coincidence. Amodel, y = a1x

2 + b1x + c1 + a2x2 � z + b2x � z + c2 � z + e,

was fitted to the data from both of the data sets to becompared, where the dummy variable z took the value 1for the first subset (standard simulation) and 0 for the other.The F-statistic from the ANOVA tables was then used to testthe significance of z, x � z, and x2 � z simultaneously.[47] Quantile regression [e.g., Koenker, 2005] as a rec-

ommended method for estimating effects of limiting factorsin ecology [e.g., Cade et al., 1999], was applied to definethe upper and lower bounds to the limitation posed on NEEby the water table position. In opposite to classical linearregression methods, which base their statistics on thesquared deviation from a mean function, quantile regressionmethods offer estimating models for median and quantilefunctions (Appendix A). The second-order polynomialmodel used for the least squares regression above waslinearized by introducing x2 = x2 and the regression quan-tiles {0.05, 0.20, 0.35, 0.5, 0.65, 0.80, 0.95} were calculated

for the resulting model y = ax2 + bx + c + e. If wehypothesize that if other factors confound the influence ofthe water table depth on the net CO2 uptake, then the lowestquantiles must exhibit the highest rate of change in NEEwith changes in water table depth. (Conventionally, upperquantiles would be used, but since we are operating withnegative values of NEE, values below the lower regressionquantiles correspond to the maximum net CO2 uptake rates,i.e., situations where the effects of other constraints arelowest). Ninety percent confidence intervals for the quantileregression coefficients a and b were computed by the rankinversion method [Koenker, 2005] to estimate if there weresignificant differences in the rates of change in NEE towater table depth calculated for the different quantileregression lines. All the statistical calculations were imple-mented in the R programming environment [R DevelopmentCore Team, 2003].[48] In order to evaluate the model parameter-based

uncertainty concerning the NEE predictions, we performeda multiparameter sensitivity analysis. Uncertainties in thekey parameters of the LPJ model, which is identical toGUESS in its representation of physiological and biogeo-chemical processes, have been determined and investigatedpreviously using Monte Carlo regression-based analysis[Zaehle et al., 2005]. In this study we focused only on theparameters added or modified in GUESS-ROMUL formodeling the mire ecosystems. The output variable usedin the analysis is NEE. Since NEE is not a monotonousfunction of the input parameters, we were not able to use aregression-based method; hence the variance-based Sobol’sensitivity indices were chosen instead (Appendix A). Thefirst-order sensitivity coefficient Sj and the total effect termTSj for the parameter of interest were compared usingStudent’s t-test. Significant differences between the twowere treated as indications of nonlinear interaction betweenthe model parameters. We used Latin hypercube sampling(LHS) [McKay et al., 1979] to generate sets of parametersfor the Monte Carlo runs, and where possible the parameterranges were derived from the literature (Table 2).

4. Results

4.1. Hydrology

[49] Good agreement between the modeled and measuredwater table depths with a mean deviation (MBE) of onlyaround 1 cm was found for 2001–2003 (Figure 1), compa-rable to previously published validation data for 1998[Granberg et al., 1999]. The Willmott index of agreementwas high (d = 0.92). The root mean square error (RMSE)was 3.5 cm and the systematic portion accounted for 8% ofthis. The difference between the measured and modeledwater table depths was greatest during the spring flood,reaching up to 10 cm on some days.

4.2. Net Ecosystem Exchange of CO2

[50] About 4.5 m of the catotelm peat was accumulatedduring the 7000-year model initialization period, whichcorresponds well with the measured peat depth, 5 m, atthe study site [Granberg et al., 1999].[51] The model was able to capture the general patterns of

CO2 exchange observed by the eddy covariance tower atDegero Stormyr mire (Figure 2). The predicted NEE

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

6 of 13

G02025

(gCm�2 yr�1) values for 2001, 2002 and 2003 were �46,�81 and �57, while the measured values were �48, �61,and �59, respectively. (Net carbon uptake by the mire isrepresented by negative NEE values in this paper). Themodeled NEE values were the differences between thesimulated net photosynthesis rates of �162, �244, and�197 gCm�2 yr�1 and simulated heterotrophic respirationrates of 116, 163, and 140 gCm�2 yr�1 for the three re-spective years. The deep (catotelm) peat accounted for 10–20% of the total modeled decomposition. The seasonalswitch of the mire between a sink and a source of CO2

was clearly reproduced by the model (Figure 2a).

[52] Some quantitative measures of the model’s perfor-mance at daily time intervals are presented in Table 3. Usingthe simulated water table depth instead of the measureddepth did not influence the quality of NEE predictionssignificantly (Table 3). In both cases the Willmott indexof agreement (d) was high and the model bias small. Thesystematic portion of the mean square error (MSE) wasrather low, but the MSE itself was relatively large. Duringthe first part of the vegetation period the model simulationalways resulted in a larger net uptake (lower NEE values)compared to the measured values. During the second part ofthe vegetation period there were both positive and negativedeviations (Figure 2b). In general, the simulations for the

Figure 2. (a) Simulated (black line) and measured (crosses) NEE (gC m �2 d�1) for the lawn at DegeroStormyr for 2001–2003. Negative NEE, carbon uptake; positive NEE, carbon release. (b) Deviationsbetween simulated and measured NEE based on 5-day averages.

Figure 1. Simulated (dash-dotted line) and measured (solid line) water table depth (WT, cm) from thelawn at Degero Stormyr for 2001–2003.

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

7 of 13

G02025

days with negative modeled NEE were poorer than thosefor days when no CO2 fixation took place, but there wasno clear relationship between productivity and modelperformance.

4.3. Relationship Between Water Table and NEEfor Modeled and Measured Data

[53] The two least squares regression lines, based onmodel simulations and direct measurements for the periodwith air temperatures above 5�C, are very close to eachother (Figure 3a and Table 4). Both show that the maximumnet CO2 uptake occurs when the water table is in the range10 to 20 cm. This range coincides with that for the averagewater table depth of this microtopographical unit [e.g.,Rydin, 1986]. The response function is also relativelysmooth, indicating that within this range the NEE is ratherinsensitive to any change in water table depth.[54] Quantile regression was used to investigate if the

water table depth is an active constraint on NEE across thewhole range of data. The quantile regression lines(Figures 3b and 3c) run parallel in the statistical sense asindicated by the 90% confidence intervals (not shown) forthe regression coefficients a and b. Confidence intervalsoverlap for all the quantile regression lines based on theobserved values and all but the 95th based on the modelsimulations. This means that the effects of the water tablelevel on NEE were similar across the range of uptake ratesfrom strongly positive to strongly negative. Only simulatedvalues comprising the 5% most positive (net release) NEEvalues related differently to the water table level; that is,the water table did not influence NEE within this range(Figure 3b).[55] The effects of the interactions between water table

depth and (1) global radiation and (2) air temperature onNEE were also studied. For this, 2D diagrams of isolinesof NEE versus water table depth and global radiation(Figures 4a and 4b), and water table depth versus temper-ature (Figures 4c and 4d) were used. If the water table depthhad no effect on the NEE within a certain temperature orsolar radiation range, the NEE isolines should run parallel tothe x axis (water table depth) within that range. However, atany given temperature or solar radiation level, the maximumCO2 uptake was found to be in the water table depth range10 to 20 cm; outside this range the CO2 uptake was reduced(Figure 4). The same pattern can be seen for both themeasured (Figures 4a and 4c) and modeled (Figures 4band 4d) values.[56] In order to separate the effects of the water table

depth on photosynthesis and peat decomposition, twoadditional simulations were performed.[57] 1. In the first run (Test run 1, psn) it was assumed

that photosynthesis was independent of water content; that

is, the photosynthetic rate was maximal with any water tabledepth. This resulted in strong increases in net CO2 uptakewhen the water table was high, but no change in NEE whenthe water table was low (Figures 5a and 5c and Table 4).[58] 2. In the second run (Test run 2, resp) it was assumed

that all the acrotelm litter and peat decayed at a ratecorresponding to the rate of anaerobic decay of Sphagnumplant material of the same degree of degradation, indepen-dently of variations in water table depth. This resulted in

Table 3. Quantitative Measures of Model Performance With Respect to Daily NEE Valuesa

MBE, gCm�2 d�1 MSE, gCm�2 d�1 MSEs, % MSEu, % d

Using measured water table �0.057 0.17 9 91 0.80Using simulated water table �0.075 0.20 15 85 0.79

aDaily NEE varies between –2.2 gCm�2 d�1 (uptake) and +1.7 gCm�2 d�1 (release). MBE, mean bias error; MSE, mean square error (apportioned intoMSEs, systematic mean square error, and MSEu, unsystematic mean square error); d, Willmott index of agreement (d = 0, no agreement; d = 1, perfectmatch).

Figure 3. Relationship between the NEE (gC m�2 d�1)and water table depth (WT, cm) for the lawn at DegeroStormyr. (a) Measured data (grey diamonds and solid trendline) and model simulations (crosses and dashed trend line)analyzed by the least squares second-order polynomialmodel. The 95th, 80th, 65th, 50th, 35th, 20th, and 5thquantile regression lines (b) for measured data and (c) formodel simulations.

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

8 of 13

G02025

stronger net CO2 uptake by the mire, due to the decrease indecomposition (Figures 5b and 5c and Table 4). The mostpronounced differences from those obtained with the orig-inal parameter settings were observed when the water tablewas at a low level, and conditions are normally oxic. Leastsquares regression lines based on the results of test runs 1and 2 are significantly different from the regression linebased on the results from the standard model simulation(Table 4).

4.4. Sensitivity Analysis

[59] The sensitivity analysis revealed the input parametersthat had strong effects on simulated NEE. The five most

important parameters were: whigh; wmax, describing theinhibition in photosynthesis when the water content of themoss is too high; rm, the maintenance respiration coeffi-cient; kan, describing the degree of reduction in heterotro-phic respiration under anoxic conditions; and k2, the rate ofpeat decomposition (Table 5). This is in good agreementwith our analysis of the key factors governing NEE atchanging water table depths (section 4.3). The sensitivityanalysis also showed that the model is nonlinear withrespect to the state variable NEE, since the total effect termTSj and the first-order sensitivity coefficient Sj for the modelinput parameters differ because of the significant contribu-tion of the interaction terms to the total variance (Table 5).

5. Discussion

[60] We found good agreements between the measuredvalues and values simulated by GUESS-ROMUL for bothwater table depth and NEE (Figures 1 and 2). However, theMBE of the NEE estimates showed seasonal variability,which, we believe, reflects model uncertainty in biomassestimates. Vascular plants (which have strong, seasonalphenological patterns) were excluded from the simulation,under the assumption (grounded in the observation that thenet primary production in the tower footprint is almostcompletely dominated by the Sphagnum species) that theircontribution to NEE is negligible. The gross primaryproductivity (GPP) of vascular plants peaks in July Augustwhen it appears to contribute up to 20% of total GPP (work

Figure 4. Isoline plots for the NEE (gC m�2 d�1) as functions of water table depth and either airtemperature or global radiation. Each contour connects the points with equal NEE corresponding to aparticular combination of water table depth and (a, b) global radiation or (c, d) air temperature. Figures 4aand 4c are based on measurements and Figures 4b and 4d are based on model outputs.

Table 4. Estimates of the Regression Coefficients for the Model

y = ax2 + bx + c + ea

Coefficients

pa b c

Measurements 0.00370b 0.134 0.401 0.12Simulation, standard 0.00424b 0.145 0.337 . . .Test run 1 (psn) 0.00315b 0.090 �0.289 <0.001Test run 2 (resp) 0.00375b 0.147 0.165 <0.001

aHere y is NEE (gCm�2 d�1) and x is the depth to the water table(centimeters), and significance P for Ho: The regression line coincides withthe regression line based on the results from the standard model simulation;psn denotes photosynthesis and resp denotes respiration.

bSignificance is <0.001.

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

9 of 13

G02025

in progress), but much lower proportions of GPP areattributable to the vascular plants during the rest of thegrowing season. Even during the driest periods in 2002 and2003 no sign of Sphagnum desiccation was noted on themire.[61] The importance of autotrophic respiration as the

second most important factor in determining the sensitivityof NEE provides further evidence of the need to increaseour understanding of plant respiration in ecosystem CO2

exchange models [e.g., Cannell and Thornley, 2000; Dewar,2000].[62] Both the modeled and observed data showed that

NEE varied with changes in the position of the water table.NEE was minimal (i.e., net CO2 uptake was highest) whenthe water table was 10–20 cm below the surface, and itincreased (i.e., net uptake of CO2 declined) when the water

table either rose or fell beyond this range (Figure 3),regardless of other constraints (temperature or solar radia-tion), as shown in the contour plots (Figure 4). Theparallelism of the quantile regression lines in Figures 3band 3c also indicates that the role of the water table on NEEwas very similar for all ranges of NEE, and thus that wereno interactions (multiplicative effects) between water tabledepth and other variables. However, it does not exclude thepossibility that such variables may have additive effects, asshown in Figure 4.[63] Reductions in the mire’s net uptake of CO2 when the

water table falls below this range may be due to increases intemperature stimulating heterotrophic respiration and/orincreases in the proportions of the peat that are decompos-ing under aerobic conditions [e.g., Silvola et al., 1996].Reductions in CO2 uptake when the water table is close toor above the surface are probably due to the inhibition ofphotosynthesis [e.g., Titus et al., 1983; Tuittila et al., 2004],caused by reductions in rates of CO2 supply due to itsrestricted transport in water and/or reductions in levels ofincoming photosynthetically active radiation.[64] Using the presented model we attempted to separate

the effects of changes in water table depth on (1) mossphotosynthesis (test run 1, psn), and (2) heterotrophicrespiration (test run 2, resp). When the water table levelwas high, less than 10 cm from the peat surface, themodeled NEE of the lawn was influenced most stronglyby changes in photosynthesis related to the water content ofthe Sphagnum shoots (Figures 5a and 5c). When the watertable was low, the increased heterotrophic respiration due tothe proportionally larger aerobic decomposition was respon-sible for the decrease in NEE (Figures 5b and 5c), whilephotosynthesis continued at close to optimal conditions; thatis, despite low water table levels, the capillary action in thesurface peat was sufficient to supply the photosyntheticallyactive layer with water (Figures 5a and 5c).[65] The results of this study indicate that there are

unlikely to be any great changes in NEE so long as theaverage water table level remains in the wide range from 10to 20 cm deep. However, if the climate changes there couldbe significant long-term changes in the water table position.If so, the model predicts that the NEE of the lawn commu-nity will decrease; that is, less carbon will be taken up bythe mire ecosystem. The amplitude of the changes in NEE

Table 5. Sensitivity Indexes for the Five Most Important

Parameters, With Respect to Daily NEEa

Parameter

WholeSimulation Period

Days With Air TemperaturesGreater Than 5�C

Sj TSj Sj TSj

whigh 0.150 0.264 0.166 0.280rm 0.134 0.216 0.055 0.183kan 0.089 0.197 0.084 0.184wmax 0.074 0.160 0.091 0.187k2 0.043 0.070 0.041 0.074

aSj, first-order sensitivity coefficient of the parameter j; TSj, total effectterm of the parameter j (sum of the first-order sensitivity Sj and all higher-order coefficients). Differences between Sj and TSj indicate nonlinearinteractions between the parameters. All the differences presented in thetable are statistically significant (Student’s t-test) at the P < 0.001 level.

Figure 5. Relationship between the NEE (gC m�2 d�1)and water table depth (WT, cm) from the simulation withthe standard parameter settings and two different test runs.(a) Test run 1: no effect of water table depth onphotosynthesis. (b) Test run 2: decay constant, correspond-ing to anoxic conditions independent of water table depth.(c) Trend lines (best-fits by second-order polynomialregression) based on the data from the standard simulation(bold solid lines), test run 1 (thin solid line with crosses),and test run 2 (dash-dotted line).

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

10 of 13

G02025

would depend on the combination of climatic factors, butunder certain circumstances carbon may even be releasedfrom the lawn. The hydrological feedback mechanismwould mitigate these changes by evaporation decreasingwhen the water table is low and runoff increasing (due tohigher peat transmissivity) when it is high. Since long-termchanges in species composition cannot be predicted with thecurrent version of the GUESS-ROMUL model, long-termchanges in NEE are likely to differ from the projectionsbased on the 3-year period studied here.[66] Our future goal is to extend the current homogeneous

model by including the possibility of modeling shiftsbetween different microcommunities. Other vegetation clas-ses that may be present in mires have already been param-eterized for the GUESS model, for example, shrubs (A. Wolfet al., Future changes in vegetation and ecosystem functionof the Barents Region, submitted to Climate Change, 2007)and graminoids (R. Wania et al., Integrating peatlands andpermafrost in a dynamic global vegetation model: Evalua-tion, sensitivity and application, manuscript in preparation,2007). Following requirements for modeling mires listed byBelyea and Baird [2006] we can summarize the advantagesand disadvantages of our modeling approach, as follows.[67] Advantages of the GUESS-ROMUL model are:

(1) process-based representation of the photosynthesis andtranspiration allowing explicit response to climatic forcing;(2) direct link between the fast ecosystem processes andother processes as competition for light and water, tissueturnover and population dynamics; (3) explicit representa-tion of spatial heterogeneity in vegetation realized bysimulating a number of repetitive patches; and (4) nonlinearinteraction between the net primary production and soilorganic matter dynamics demonstrated in this study bynonlinear interaction of the key parameters.[68] Some disadvantages of the model, in comparison

with more advanced features of certain other mire models(cited in parentheses), are that it does not account for:(1) localized flows of water and nutrients [e.g., Nungesser,2003; Swanson and Grigal, 1988; Rietkerk et al., 2004];(2) plant competition for nutrients [Pastor et al., 2002];(3) disturbance of the strict division between the acrotelmand catotelm mediated by oxygen supplied via roots ofsedges and dwarf shrubs [e.g., Frolking et al., 2002; Zhanget al., 2002]; (4) division of the vertical structure of the peatinto cohorts [e.g., Frolking et al., 2002]; and (5) feedbacksbetween the peat accumulation and the hydrological pro-cesses [Hilbert et al., 2000].

6. Conclusions

[69] In this study the modified GUESS-ROMUL modelof vegetation and soil CO2 exchange was used to investigatethe effect of the water table depth on net CO2 uptake rates ofa Sphagnum-dominated mire. NEE data obtained fromeddy-tower measurements over the lawn plant communitiesat the oligotrophic minerotrophic mire Degero Stormyr wereused for the validation, and the model was shown toperform well with daily time steps. The results of the studyindicate the following.[70] 1. Rates of net CO2 uptake at Degero Stormyr mire

were relatively constant during the 2001–2003 growing

seasons when the depth of the water table was between 10and 20 cm. Model simulations show that any changes in thewater balance that do not cause the water table to shift fromwithin this normal range are likely to have minor effects onthe annual rates of land-atmosphere CO2 exchange.[71] 2. When the water table level was either higher

(>�10 cm) or lower (<�20 cm) than normal the mire netCO2 uptake decreased.[72] 3. According to model simulations the reductions in

net CO2 uptake rates that occurred when the water tablelevel was higher than �10 cm are attributable to reductionsin photosynthetic rates that are probably due to reductions inthe rate of CO2 assimilation associated with higher thanoptimal Sphagnum water contents and increased CO2 lim-itations due to the lower transport rate of CO2 in water.[73] 4. The reductions in net CO2 uptake when the lower

water table level was lower than �20 cm resulted fromincreases in heterotrophic respiration, rather than reductionsin photosynthesis since no Sphagnum desiccation was notedon the mire or predicted by the model during 2001–2003.[74] These results may be also applied to other mire sites

with similar vegetation, but only under the assumption ofconstant plant species composition.

Appendix A

A1. Willmott Index of Agreement d

[75] This dimensionless measure is related to the absolutesize of the difference between the modeled and observedvalues.

d ¼ 1�XNi¼1

Mi � Oið Þ2=XNi¼1

jM 0i j þ jO0

ij� �2" #

; ðA1Þ

where Mi0 = Mi � O, Oi

0 = Oi � O, M-modeled and O-observed values.[76] The Willmott index of agreement varies between 0

and 1, with d = 1 indicating a perfect match.

A2. Quantile Regression

[77] Quantile regression [Koenker and Bassett, 1978] is astatistical technique that estimates the full range of condi-tional quantile functions y � Xb(t). Estimation of thecoefficients of b(t) is based on minimization of a weightedsum (with weights depending on the order of the quantiles)of absolute values of residuals,

minXni¼1

rt yi �Xpj¼0

bjxij

!" #; ðA2Þ

where n is sample size, p is the number of parameters,rt (e) = e(t � I(e < 0)), and I(�) is the indicator function.

A3. Variance-Based Sobol’ Sensitivity Indices

[78] The method proposed by Sobol’ [1990] is based ondecomposition of the function and construction of a modeloutput into sums of increasing dimensions. The variance D

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

11 of 13

G02025

of the function can be broken down in a similar manner tothe ANOVA scheme,

D ¼Xkj¼1

Dj þX

1�i<j�k

Dij þ . . .þ D12:::k ; ðA3Þ

where j indicates one of the k model input parameters ( j =1,. . ., k). Both the total variance D and each term of thisdecomposition can be computed using Monte Carlointegrals [see Saltelli et al., 2000]. A first-order sensitivitycoefficient, showing the individual effect of the particularparameter j on the model output, is defined as

Sj ¼ Dj

D: ðA4Þ

[79] The total effect term TSj [Homma and Saltelli, 1996]is defined as the sum of the first-order sensitivity coefficient(Sj) and all higher-order coefficients containing the param-eter j, to account for the interaction of j with otherparameters. If the model is nonlinear, then Sj and TSj differbecause of the significant contribution of the interactionterms to the total variance.

[80] Acknowledgments. We are grateful to three anonymous refereesfor their comments on an earlier version of the manuscript. This work hasbeen supported by the NECC project funded by the Nordic Centre ofExcellence (NCoE) Pilot Programme, the EU project BALANCE, EU grantEVK2-2002-00169, NCCR-Climate, a research initiative of the SwissNational Science Foundation and the Swedish Research Council forEnvironment, Agricultural Sciences and Spatial Planning (grant 21.4/2003-0876 to M. N.). The Kempe Foundation is also gratefully acknowl-edged for grants for the micrometeorological instruments.

ReferencesAlexandersson, H., C. Karlstrom, and S. Larsson-Mccann (Eds.) (1991),Temperature and Precipitation in Sweden 1961–1990: ReferenceNormals, 87 pp., Swedish Meteorol. and Hydrol. Inst., Norrkoping, Sweden.

Alm, J., L. Schulman, J. Walden, H. Nykanen, P. J. Martikainen, andJ. Silvola (1999), Carbon balance of a boreal bog during a year with anexceptionally dry summer, Ecology, 80, 161–174.

Belyea, L. R., and A. J. Baird (2006), Beyond ‘‘The limits to peat boggrowth’’: Cross-scale feedback in peatland development, Ecol. Monogr.,76, 299–322.

Bergman, I. (1998), Microbial carbon mineralisation in peat from a borealmixed mire: Regulation by environmental factors, Ph.D. thesis, 33 pp.,Swedish Univ. of Agric. Sci., Uppsala, Sweden.

Bubier, J., P. Crill, A. Mosedale, S. Frolking, and E. Linder (2003), Peat-land responses to varying interannual moisture conditions as measured byautomatic CO2 chambers, Global Biogeochem. Cycles, 17(2), 1066,doi:10.1029/2002GB001946.

Cade, B. S., et al. (1999), Estimating effects of limiting factors with regres-sion quantiles, Ecology, 80, 311–323.

Cannell, M. G. R., and J. H. M. Thornley (2000), Modelling the compo-nents of plant respiration: Some guiding principles, Ann. Bot., 85, 45–54.

Chertov, O. G. (1985), Imitatsionnaya model mineralizatsii i gumifikatsiilesnogo opada i podstilki, (in Russian), Zh. Obshch. Biol., 46(6), 794–804.

Chertov, O. G., A. S. Komarov, M. Nadporozhskaya, S. S. Bykhovets, andS. L. Zudin (2001), ROMUL—A model of forest soil organic matterdynamics as a substantial tool for forest ecosystem modeling, Ecol.Modell., 138, 289–308.

Clymo, R. S. (1965), Experiments on breakdown of sphagnum in two bogs,J. Ecol., 53, 747–758.

Clymo, R. S. (1984), The limits to peat bog growth, Philos. Trans. R. Soc.,303, 605–654.

Collatz, G.-J., J. T. Ball, C. Grivet, and J. A. Berry (1991), Physiologicaland environmental regulation of stomatal conductance, photosynthesisand transpiration: A model that includes a laminar boundary layer, Agric.For. Meteorol., 54, 107–136.

Dewar, R. C. (2000), A model of the coupling between respiration, activeprocesses and passive transport, Ann. Bot., 86, 279–286.

Dorrepaal, H., J. H. C. Cornelissen, R. Aerts, B. Wallen, and R. S. P. vanLogtestijn (2005), Are growth forms consistent predictors of leaf litterquality and decomposibility across peatlands along a latitudinal gradient?,J. Ecol., 93, 817–828.

Farquhar, G. D., S. von Caemmerer, and J. A. Berry (1980), A biochemicalmodel of photosynthetic CO2 assimilation in leaves of C3 species, Planta,149, 78–90.

Freeman, C., C. D. Evans, D. T. Monteith, B. Reynolds, and N. Fenner(2001), Export of organic carbon from peat soils, Nature, 412, 785.

Friend, A. D. (1991), Use of a model of photosynthesis and leaf micro-environment to predict optimal stomatal conductance and leaf nitrogenpartitioning, Plant Cell Environ., 14, 895–905.

Friend, A. D. (1995), An integrated model of leaf photosynthesis, transpira-tion, and conductance, Ecol. Modell., 77, 233–255.

Frolking, S., N. T. Roulet, T. R. Moore, P. M. Lafleur, J. L. Bubier, andP. M. Crill (2002), Modeling seasonal to annual carbon balance of MerBleue Bog, Ontario, Canada, Global Biogeochem. Cycles, 16(3), 1030,doi:10.1029/2001GB001457.

Gerten, D., S. Schaphoff, U. Haberlandt, W. Lucht, and S. Sitch (2004),Terrestrial vegetation and water balance-hydrological evaluation of adynamic global vegetation model, J. Hydrol., 286, 249–270.

Gorham, E. (1991), Northern peatlands: Role in the carbon cycle andprobable responses to climatic warming, Ecol. Appl., 1, 182–195.

Granberg, G., H. Grip, M. O. Lofvenius, I. Sundh, B. H. Svensson, andM. Nilsson (1999), A simple model for simulation of water content, soilfrost, and soil temperatures in boreal mixed mires,Water Resour. Res., 35,3771–3782.

Haxeltine, A., and I. C. Prentice (1996), A general model for the light-useefficiency of primary production, Funct. Ecol., 10, 551–561.

Hilbert, D. W., N. Roulet, and T. Moore (2000), Modelling and analysis ofpeatlands as dynamical systems, J. Ecol., 88, 230–242.

Homma, T., and A. Saltelli (1996), Importance measures in global sensi-tivity analysis of nonlinear models, Reliab. Eng. Syst. Safety, 52, 1–17.

Ivanov, K. E. (1981),Water Movement in Mirelands, 276 pp., Elsevier, NewYork.

Koenker, R. (2005), Quantile Regression, Cambridge Univ. Press, NewYork.

Koenker, R., and G. Bassett (1978), Regression quantiles, Econometrica,46, 33–50.

Korhola, A., J. Alm, K. Tolonen, J. Turunen, and H. Jungner (1996), Three-dimensional reconstruction of carbon accumulation and CH4 emissionduring nine millennia in a raised mire, J. Quat. Sci., 11, 161–165.

Kuhry, P., and D. H. Vitt (1996), Fossil carbon/nitrogen as a measure ofpeat decomposition, Ecology, 77, 271–275.

Lafleur, P. M., N. Roulet, J. Bubier, S. Frolking, and T. R. Moore (2003),Interannual variability in the peatland-atmosphere carbon dioxide ex-change at an ombrotrophic bog, Global Biogeochem. Cycles, 17(2),1036, doi:10.1029/2002GB001983.

Lloyd, J., and J. A. Taylor (1994), On the temperature dependence of soilrespiration, Funct. Ecol., 8, 315–323.

Malmer, N., and B. Nihlgard (1980), Supply and transport of mineralnutrients in a sub-arctic mire, in Ecology of a Subarctic Mire, editedby M. Sonesson, pp. 63 – 95, Naturvetenskapliga forskningsradet(NFR), Stockholm.

Malmer, N., and B. Wallen (2005), Nitrogen and phosphorous in mireplants: Variation during 50 years in relation to supply rate and vegetationtype, Oikos, 109, 539–554.

Malmstrom, C. (1923), Degero stormyr: En botanisk, hydrologisk och ut-vecklinghistorisk undersokning over ett nordsvenskt myrkomplex (inSwedish), 176 pp., Statens Skogsforsoksanstalt, Stockholm.

McKay, M. D., W. J. Conover, and R. J. Beckman (1979), A comparison ofthree methods for selecting values of input variables in the analysis ofoutput from a computer code, Technometrics, 21, 239–245.

McKenzie, C., S. Schiff, R. Aravena, C. Kelly, and V. St. Louis (1998),Effect of temperature on production of CH4 and CO2 production frompeat natural and flooded boreal wetland, Clim. Change, 40, 247–266.

Moore, T. R., and M. Dalva (1997), Methane and carbon dioxide exchangepotentials of peat soils in aerobic and anaerobic laboratory incubations,Soil Biol. Biochem., 29, 1157–1164.

Moore, T. R., N. T. Roulet, and J. M. Waddington (1998), Uncertainty inpredicting the effect of climatic change on the carbon cycling of Canadianpeatlands, Clim. Change, 40, 229–245.

Nilsson, M., and E. Bohlin (2001), Betydelsen av vitmossa (Sphagnum) forboreala myrars kolbalans: Slutrapport for projekt 216 100-2 ‘‘Fixeringoch ackumulering av kol i torvbildande vaxtsamhallen’’, 50 pp., SverigesLantbrukuniversitet, Umea, Sweden.

Nungesser, M. K. (2003), Modelling microtopography in boreal peatlands:Hummocks and hollows, Ecol. Modell., 165, 175–207.

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

12 of 13

G02025

Nykanen, H., J. E. P. Heikkinen, L. Pirinen, K. Tiilikainen, and P. J.Martikainen (2003), Annual CO2 exchange and CH4 fluxes on a subarcticpalsa mire during climatically different years,Global Biogeochem. Cycles,17(1), 1018, doi:10.1029/2002GB001861.

Pastor, J., B. Peckham, S. Bridgham, J. Weltzin, and J. Chen (2002), Plantcommunity dynamics, nutrient cycling, and alternative stable equilibria inpeatland, Am. Nat., 160, 553–568.

Potter, C., J. Bubier, P. M. Crill, and P. M. Lafleur (2001), Ecosystemmodeling of methane and carbon dioxide fluxes for boreal forest sites,Can. J. For. Res., 31, 208–223.

R Development Core Team (2003), R: A Language and Environment forStatistical Computing, R. Foundation, Vienna. (Available at http://www.R-project.org)

Rietkerk, M., S. C. Dekker, M. J. Wassen, A. W. M. Verkroost, and M. F. P.Bierkens (2004), A putative mechanism for bog patterning, Am. Nat.,163, 699–708.

Roulet, N., S. Hardill, and N. Comer (1991), Continuous measurement ofthe depth of water-table (inundation) in wetlands with fluctuating sur-faces, Hydrol. Processes, 5, 399–403.

Rydin, H. (1986), Competition and niche separation in Sphagnum, Can. J.Bot., 64, 1817–1824.

Sagerfors, J. K., A. Lindroth, A. Grelle, L. Klemedtsson, P. Weslien, andM. Nilsson (2007), Annual carbon exchange between a nutrient poor,minerotrophic, boreal mire and the atmosphere, J. Geophys. Res,doi:10.1029/2006JG000306, in press.

Saltelli, A., K. Chan, and E. M. Scott (Eds.) (2000), Sensitivity Analysis,504 pp., John Wiley, Hoboken, N. J.

Schipperges, B., and H. Rydin (1998), Response of photosynthesis ofSphagnum species from contrasting microhabitats to tissue water contentand repeated desiccation, New Phytol., 140, 677–684.

Silvola, J. (1990), Combined effects of varying water content and CO2

concentration on photosynthesis in Sphagnum fuscum, Holarctic Ecol.,13, 224–228.

Silvola, J., J. Alm, U. Ahlholm, H. Nykanen, and P. J. Martikainen (1996),Fluxes from peat in boreal mires under varying temperature and moistureconditions, J. Ecol., 84, 219–228.

Sitch, S., et al. (2003), Evaluation of ecosystem dynamics, plant geographyand terrestrial carbon cycling in the LPJ Dynamic Global VegetationModel, Global Change Biol., 9, 161–185.

Smith, B., I. C. Prentice, and M. T. Sykes (2001), Representation of vege-tation dynamics in the modelling of terrestrial ecosystems: Comparingtwo contrasting approaches within European climate space, Global Ecol.Biogeogr., 10, 621–637.

Sobol’, I. M. (1990), Sensitivity estimates for nonlinear mathematical mod-els (in Russian), Matematicheskoe Modelirovanie, 2, 112–118.

Swanson, D. K., and D. F. Grigal (1988), A simulation model of mirepatterning, Oikos, 53, 309–314.

Syed, K. H., L. B. Flanagan, P. J. Carlson, A. J. Glenn, and K. E. vanGaalen (2006), Environmental control of net ecosystem CO2 exchangeand carbon balance in a treed, moderately rich fen in northern Alberta,Agric. For. Meteorol., 140, 97–114.

Titus, J. E., and D. J. Wagner (1984), Carbon balance for two Sphagnummosses: Water balance resolves a physiological paradox, Ecology, 65,1765–1774.

Titus, J. E., D. J. Wagner, and M. D. Stephens (1983), Contrasting waterrelations of photosynthesis for two Sphagnum mosses, Ecology, 64,1109–1115.

Tolonen, K., H. Vasander, A. W. H. Damman, and R. S. Clymo (1992), Fateof apparent and true carbon accumulation in boreal peatlands, paperpresented at the 9th International Peat Congress, Int. Peat Soc., Uppsala,Sweden.

Tuittila, E.-S., H. Vasander, and J. Laine (2004), Sensitivity of C sequestra-tion in reintroduced Sphagnum to water-level variation in a cutawaypeatland, Restoration Ecol., 12, 483–493.

Turunen, J., E. Tomppo, K. Tolonen, and A. Reinikainen (2002), Estimatingcarbon accumulation rates of undrained mires in Finland—Application toboreal and subarctic regions, Holocene, 12, 69–80.

van Genuchten, M. T. (1980), A closed form equation for predicting thehydraulic conductivity of unsaturated soils, Soil Sci. Soc. Am. J., 44,892–898.

Verhoeven, J. T. A., and E. Toth (1995), Decomposition of Carex andSphagnum litter in fens: Effect of litter quality and inhibition by livingtissue homogenates, Soil Biol. Biochem., 27, 271–275.

Virta, J. (1966), Measurement of evapotranspiration and computation ofwater budget in treeless peatlands in the natural state, Commentat. Phys.Math., 32(11), 1–70.

Waddington, J. M., and N. T. Roulet (2000), Carbon balance of a borealpatterned peatland, Global Change Biol., 6, 87–97.

Waddington, J. M., P. A. Rotenberg, and F. J. Warren (2001), Peat CO2

production in a natural and cutover peatland: Implications for restoration,Biogeochemistry, 54, 115–130.

Weiss, R., J. Alm, R. Laiho, and J. Laine (1998), Modeling moistureretention in peat soils, Soil Sci. Soc. Am. J., 62, 305–313.

Williams, T. G., and L. B. Flanagan (1996), Effects of changes in watercontent on photosynthesis, transpiration and discrimination against13CO2 and C18O16O in Pleurozium and Sphagnum, Oecologia, 108,38–46.

Williams, T. G., and L. B. Flanagan (1998), Measuring and modellingenvironmental influences on photosynthetic gas exchange in Sphagnumand Pleurozium, Plant Cell Environ., 21, 555–564.

Willmott, C. J. (1982), Some comments on the evaluation of model perfor-mance, Bull. Am. Meteorol. Soc., 63, 1309–1313.

Yurova, A. Y., and H. Lankreijer (2007), Carbon storage in organic layersof boreal forest soils under various moisture conditions. Model study forthe northern Sweden sites, Ecol. Modell, in press.

Zaehle, S., S. Sitch, B. Smith, and F. Hatterman (2005), Effects ofparameter uncertainties on the modeling of terrestrial biospheredynamics, Global Biogeochem. Cycles, 19, GB3020, doi:10.1029/2004GB002395.

Zhang, Y., C. Li, C. C. Trettin, H. Li, and G. Sun (2002), An integratedmodel of soil, hydrology, and vegetation for carbon dynamics in wetlandecosystems, Global Biogeochem. Cycles, 16(4), 1061, doi:10.1029/2001GB001838.

�����������������������M. Nilsson and J. Sagerfors, Department of Forest Ecology, Swedish

University of Agricultural Sciences, Umea, Sweden.A. Wolf, Department of Forest Ecology, Institute of Terrestrial

Ecosystems, Zurich, Switzerland.A. Yurova, Department of Physical Geography and Ecosystems Analysis,

Lund University, Solvegatan 12, Lund SE-22362, Sweden. (alla.yurova@nateko.lu.se)

G02025 YUROVA ET AL.: MODELING CO2 EXCHANGE IN A BOREAL MIRE

13 of 13

G02025

I

I

III

I

Submitted to Water Resources Research

_____________________________________________________________________

A model of dissolved organic carbon concentrations in peat and stream

waters. I: Development and parameterization

Alla Yurova1, Ishi Buffam2 and Kevin Bishop3

1Department of Physical Geography and Ecosystems Analysis, Lund University, Lund, Sweden

2Department of Forest Ecology, Swedish University of Agricultural Sciences, Umeå, Sweden

3Department of Environmental Assessment, Swedish University of Agricultural Sciences, Uppsala,

Sweden

_____________________________________________________________________

Abstract

In this paper we present a model of the DOC concentrations and fluxes in peatland water based on the convection-dispersion equation. The dynamics of sorbed soluble organic carbon (SSOC) in the peat matrix are simulated in parallel with DOC. In a companion paper (Part II), the model is fully applied to simulate the DOC concentrations in the outlet of the steam draining a small headwater mire in Northern Sweden during the period 1993-2001. Here, in Part I, the model is applied solely to stagnant water conditions, in order to interpret the results of laboratory peat incubations, with the focus on sorption processes. Some important model parameters are derived using literature data complemented by information from new incubation experiments. A dynamic approach is chosen for the representation of sorption kinetics in the model. Differential sensitivity analysis accompanied by term-by-term comparison of the processes contributing to changes in DOC are used to demonstrate the leading role of sorption kinetics in determining DOC concentration during the first days after the system is “reset” with a new concentration. Previous models of DOC have usually been based on static equilibrium representations of sorption, and thus may only be fully applicable for a rather coarse time resolution. Analysis of the model highlights the importance of experimental design: when using laboratory peat incubations to determine DOC production rate, the peat mass to water ratio and initial SSOC concentration must be standardized or taking into account. _____________________________________________________________________Keywords: dissolved organic carbon, peatland, model

* Author for correspondence (alla.yurova@nateko.lu.se)

Introduction

If considered separately, the carbon cycle is not closed by exchange with the

atmosphere in either terrestrial or aquatic ecosystems, since the lateral flux of carbon

from the land to aquatic bodies links the two. In the boreal zone, the flux of dissolved

organic carbon (DOC) is much greater then the particulate organic carbon (POC) and

dissolved inorganic carbon (DIC) fluxes [Cole et al., 2007; Jonsson et al., 2007].

Export of DOC is likely to occur mainly from areas where flowing water bypasses the

adsorbing mineral soils, i.e. in surface and subsurface flows from littoral zones and

wetlands [e.g. McKnight and Aiken, 1998; Qualls, 2000]. Field data collected from

various locations within the boreal region have highlighted the importance of

peatland, a boreal type of wetland, as one of the major sources of terrestrially-derived

dissolved organic carbon (DOC) reaching aquatic ecosystems [e.g. Freeman et al.,

2001a; Tranvik and Jansson, 2002]. The amounts of DOC exported from peatlands

are sufficiently large to be included in regional C balances [e.g. Jonsson et al., 2007]

and to contribute to the peatland C balance [Moore et al., 1998; Fraser et al., 2001;

Billett et al., 2004, Roulet et al., 2007]. Aquatic scientists widely recognize the role of

DOC in providing energy for microbial growth [e.g. Jansson et al., 2000], regulating

UV light attenuation [e.g. Schindler et al., 1996], pH buffering [e.g. Kortelainen et

al., 1989] and metal complexation [e.g. McKnight and Aiken, 1998]. Other effects of

DOC, such as interference with photosynthesis, mild stress on aquatic organisms,

promotion and suppression of microbial growth etc. are less well known but equally

important [Steinberg et al., 2006].

2

The concentration of DOC in peat pore water is a product of inputs (production of

soluble organic matter from living plants and decomposing organic remains, imports

via flowing water, dispersion), outputs (mineralization, hydrological export,

dispersion) and sorptive exchanges with solid organic matter [see, for instance Qualls

and Richardson, 2003]. The complexity of the problem has encouraged the

development of several process-based models of DOC in soils and stream water

[Grieve, 1991; Neff and Asner, 2001; Michalzik et al., 2003; Futter et al., 2007].

Despite the inevitable simplifications when describing chemical and biochemical

DOC transformations, such models have been relatively successful for predicting

DOC concentration and fluxes over daily, weekly or monthly time intervals. Within

the existing models, there is general consensus about which processes to include,

namely microbially-mediated transformations, sorptive exchange and hydrological

transport of the DOC. However, the equations used to describe each process differ

from model to model.

For our study (Parts I and II) we selected a mathematical framework that, to our

knowledge, has not previously been applied to the DOC problem. Solving the

classical convection–dispersion equation [e.g., van Genuchten and Waganet, 1989],

we included two important capabilities that are new to DOC models: (i) accounting

for the non-steady state behavior of organic carbon with respect to sorption and (ii)

dynamically and simultaneously tracing transport and biochemistry in the

heterogeneous peat profile.

The model is built up in two separate stages, the ultimate goal being to simulate DOC

export from boreal mires. In Part I (this paper), the model is fully developed. In

3

addition, the version applicable to a situation with stagnant-water and constant

temperature is used to explore the implications of the model for laboratory incubation

studies. We also performed long-term laboratory incubations with peat samples

collected from the mire studied in Part II to help derive the model parameters. In Part

II (the companion paper), the predictive form of the model is used to simulate DOC

build-up and transport from a small boreal peatland in Northern Sweden during the

period 1993–2001.

In addition to model parameterization, our main purpose in Part I was to explore the

role of sorption as a major mechanism regulating DOC concentration and release

rates. This was achieved by analyzing the proposed equations and by applying

differential sensitivity analysis to the key model parameters. While laboratory studies

have shown that DOC sorption is time-dependent [e.g. Qualls, 2000; Rasse et al., in

revision], previous DOC models have treated it as an instantaneous process. In the

model developed by Grieve [Grieve, 1991; also applied in Boyer et al., 1996], the

ratio between the store of soluble C and DOC concentration in water is an empirically

fitted constant, while the models of Neff and Asner [2001] and Michalzik et al. [2003]

make use of an equilibrium partition constant, based on laboratory soil incubations, to

relate the amounts of dissolved and sorbed OC. Here we aim to demonstrate the

advantage of the dynamic approach over a static one when modeling DOC sorption.

We use the data from the long-term peat incubations and perform calculations

designed to interpret the data, focusing on sorption action over different time intervals

and under a range of experimental conditions. This approach is particularly important

when trying to use laboratory studies to parameterize models.

4

Model

The model presented here is based on two main premises:

Soluble organic carbon is present in the mire system in two states: dissolved

(DOC) and sorbed, potentially soluble, but currently solid (SSOC). The balance

between the two phases varies over time, and adsorption and desorption can be

described using first-order kinetics.

A convection–dispersion equation can provide a suitable model of DOC

transport. Including the terms accounting for the adsorption–desorption as well

as microbial production and mineralization in this equation, accounts for the full

mass balance for DOC.

The underlying concept of the model is the schematic representation of processes that

generate and consume DOC in a mire ecosystem, as presented in Qualls and

Richardson [2003]. The processes that contribute to in situ changes in pore water

DOC concentration (sorption, microbial transformation) are distinguished from the

various transport processes, including dispersion and advection by both vertical and

horizontal flow (Figure 1). We model sorption and microbially-mediated

transformations of the DOC according to the scheme presented in Figure 2. This part

of the model (shaded in Figure 1) can be used alone to study situations with no water

flow (e.g. systems with stagnant water in laboratory conditions, Part I – this paper).

Transport processes are modeled as presented in Figure 3, which is relevant to

application of the model under field conditions (Part II). The hydrological model used

to simulate the flow rate and water content is the Mixed Mire Water and Heat Model

[MMWH, Granberg et al., 1999].

5

Sorption

In our model we use a linear kinetic equation to simulate the adsorption of soluble

organic matter to, and its desorption from, solid organic matter. Sorption plays a key

role in controlling the DOC concentration in pore water. As shown by Qualls [2000],

peat always contains substantial quantities of potentially soluble OC sorbed to solid

organic matter (SSOC); there is always more of this than the amount dissolved during

any single leaching event. In laboratory peat incubations, it is possible to observe the

rapid release of DOC into newly added pore water. This occurs as previously sorbed

DOC is desorbed, and during the period before DOC concentrations in the sorbed and

dissolved phases reach equilibrium [Qualls, 2000]. Equilibration is not achieved

immediately, and the rate of change in the DOC concentration due to sorption is

highest at the beginning of experiment, decreases during incubation, and is negligible

only after 2-4 days [e.g. Rasse et al., in revision]. Equilibrium sorption models do not

describe the dynamics of the process adequately, since they are based on the

assumption that exchange between the sorbed and dissolved phases is so rapid that

there is always a fixed ratio between the two. Hence, a linear kinetic model is a better

when modeling DOC with a relatively high time resolution. Sorption occurs not only

in cases in which the pore water is completely replaced (such as laboratory

experiments), but also in any situation where dilution or advection alter DOC

concentrations in pore water.

Microbially-mediated transformations

Metabolic transformations mediated by microorganisms also strongly affect the DOC

balance. Acting at slower rates than adsorption/desorption processes, such

transformations determine the total amount of soluble OC produced by exoenzymatic

6

dissolution of insoluble organic matter. In addition, they control the rate of

mineralization of DOC and SSOC. In our model we only consider the recalcitrant

fraction of DOC. The DOC present in the pore water of a boreal peatland is almost

exclusively in the form of refractory biomolecules, such as fulvic acids and their

precursors [e.g., McKnight et al., 1985] formed mainly from the decomposition of

Sphagnum. Easily decomposable substances, such as root exudates and simple

products of decomposition, are readily consumed by microorganisms, and have high

turnover rates and very short lifetimes in solution [e.g., Qualls, 2000; Crow and

Wieder, 2005]. They therefore comprise only a small proportion of the DOC in

solution [e.g., McKnight et al., 1985]. Out model does not aim to describe the

dynamics of such short-lived substances, and the parameterization presented below is

based solely on data from the laboratory peat incubations in which the bulk DOC of

the peat solution was measured.

In our simplified model, like many models of soil organic matter dynamics [e.g.,

reviewed by Ågren et al., 1991], microbial biomass is not explicitly included. Instead,

we simulate microbially-mediated transformations of the DOC (production and

mineralization) that are affected by temperature and the presence of O2. We assume

first order kinetics for the mineralization of DOC and SSOC. A zero order production

constant is hypothesized, based on the assumption that substrates for the formation of

DOC are not limited in the peat organic matter.

Bulk DOC concentration is simulated without specifying chemical fractions; we

deliberately avoided separating organic matter into different fractions. Instead, we

believe that a new modeling approach to soil organic matter turnover, based on

7

tracing changes in the continuous distribution of quality of the organic molecules,

known as Q theory [Ågren and Bosatta, 1998], provides a more promising approach.

However, the application of Q theory to soil solutions has many associated

challenges, such as defining the relationships between quality and solubility, or

between quality and sorption properties, so this work has yet to be undertaken.

Mathematical formulation

The mass balance equation for the concentration of DOC in pore water, c, can be

written in the form of convection-dispersion equation (symbol definition and units in

Table 1):

,cP)cKs()c()c(t

)c(1Ddes [1]

where is volumetric water content, is peat bulk density, vector q qx, qy, qz is

composed of water flux in horizontal and vertical directions, D Dx, Dy, Dz is a vector

of dispersion coefficients, des is the kinetic rate of desorption (or adsorption if the

third term in equation [1] becomes negative), KD is the partitioning constant,

characterizing an equilibrium between the sorbed and dissolved phases, P is the

microbial production rate of DOC, and 1 is the first-order microbial DOC

mineralization constant. Mathematical operators are:

),z

cD(

z)

y

cD(

y)

x

cD(

x)c( zyx

z

)cq(

y

)cq(

x

)cq()c( zyx

8

The dispersion coefficient is defined as

,q

DD 0

where D0 is the molecular diffusion coefficient, and is dispersivity.

A mass balance for the concentration of the sorbed soluble organic carbon (SSOC), s,

leads to:

,s)cKs(t

s2Ddes [2]

where 2 is the first-order microbial mineralization constant for the SSOC. We use the

term “concentration” rather than “pool size” to describe amounts of SSOC present in

the systems considered in this paper, even though sorbed soluble OC is in a solid state

and thus it is not a wholly appropriate term. The intention is to emphasize that the

term refers to the amounts sorbed per unit mass of peat.

The system (1)-(2) describes DOC concentration at a given time, assuming that the

initial DOC and SSOC concentrations (c0 and s0) are known.

A modified van’t Hoff equation is used to account for the temperature dependency of

DOC production and mineralization:

9

)10/)((

1011

)10/)((

10 ,TbasalT

basal

TTbasal

Q

QPP basal

where Q10 is the relative rate of increase in metabolic rates per 10°C increase in

temperature, and Tbasal is a reference temperature (20°C was used here) at which basal

rates of microbial DOC production (Pbasal) and mineralization ( 1basal) are estimated.

The constant fractions kanP and kan are used to relate the anoxic rates of DOC

production and mineralization to the corresponding DOC production and

mineralization rates under aerobic conditions:

an11

anP

k

,kPPconditionsanaerobicfor

In our model we made use of separate mineralization rate coefficients for DOC and

sorbed soluble organic carbon (SSOC), to account for the stabilizing effect of sorption

on microbial decomposition [e.g., Kalbitz et al., 2005]. The parameter ks is defined as

1

2sk ,

i.e. the mineralization rate of SSOC ( 2) is ks times less then the mineralization rate of

DOC ( 1).

10

Hydrology

Two layers can be distinguished within the peat profile: the upper, periodically

aerated layer consisting largely of living and lightly decomposed plant material

(acrotelm), which is not usually more than 60 cm deep, and the permanently saturated,

lower zone, consisting largely of compacted, relatively highly decomposed plant

material (catotelm). The hydraulic conductivity is high in the upper acrotelm and

declines strongly with depth [Ivanov, 1981]. Consequently, water in the acrotelm

moves rapidly and usually accounts for most of the discharge from the peatland. If the

peatland is formed over minerals soils with low permeability, the water in the

catotelm is mostly stagnant; if, however, the underlying mineral soil is highly

permeable, significant vertical movements can occur [e.g., Reeve et al., 2000]. It has

also been suggested that the acrotelm is the major source of DOC within peatlands

[e.g., McKnight et al., 1985], since decomposition is much slower in the permanently

anaerobic layers, where organic matter is aged and the enzymes regulating organic

matter transformation are suppressed [Freeman et al., 2001a, b], in addition

accumulations of CH4 and CO2 further inhibit decomposition [Blodau et al., 2004].

For the field application of our model (Part II) we selected the Mixed Mire Water and

Heat (MMWH) hydrological model [Granberg et al., 1999] to simulate water flux,

water content, water table depth and temperature in the acrotelm. The MMWH model

is designed specifically for peatlands and is based on the method of steady-state

moisture distribution curves [Romanov, 1961]. Utilizing continuous functions for the

moisture profile and hydraulic conductivity, the model gives accurate accounting of

temperature, moisture, and the water table depth in the acrotelm with high vertical

11

resolution. Simplification of the system of equations (1)-(2) for the 1D flow system

described by the MMVH is given below.

Addition: Convection–dispersion equation for the simplified hydrological schemes.

If a 1D hydrological scheme (vertical discreetization) is used and there is no lateral

import of DOC into the peatland from surrounding areas, the finite-difference form of

system (1)-(2) can be written as.

,cP)cKs(z

)cq(

z

cq

z

)z

c(

Dt

)c(1Ddes

zsz [1a]

,s)cKs(t

s2Ddes [2a]

where z is the depth of a specific layer, qz is the vertical water flux and qs is the

specific (area) lateral discharge from the layer z.

In this simple model, the DOC export flux from the peatland is the sum of the

products of the concentration and specific discharge from all of the vertical peat

layers.

Model parameterization

The model parameters and initial conditions needed to solve equations (1)-(2) in a

predictive way are summarized in Table 2. Only the parameters relevant when

applying the model to laboratory conditions are specified here in Part I (the other

12

parameters are listed, but their values and ranges are presented in Part II, where the

model is applied to field conditions). Four parameters – the sorption equilibrium

partitioning coefficient (KD), the initial SSOC concentration (s0), and the rates of

microbial DOC production (P), and DOC mineralization ( 1) – were estimated using

data obtained from our long-term peat and water incubations (experimental design in

Appendix). In addition, we reviewed the available literature relating to the rates of

DOC release into water during long-term experiments to obtain ranges of estimates

for the two parameters: the DOC production rate (P) and the DOC mineralization rate

( 1).

Method for obtaining values for model parameters

The required model parameters were determined by a two-step process. First, we had

to accept statements (i) and (ii) presented below. Second, when the numerical values

of the model coefficients had been determined, these assumptions were tested by

means of the analysis of the equations and sensitivity study described under “Model

analysis”.

We hypothesized that the process of DOC release into the solution could be divided

into two phases, this is conceptually similar to the “tea bag” approach presented in

Aguilar and Thibodeaux [2005].

(i) During the initial phase, the process is dominated by rapid sorption–

desorption kinetics and not detectably influenced by microbially-mediated

rates of production and mineralization [Qualls, 2000]. This phase can take

up to 4 days [e.g. Rasse et al., in revision] until an equilibrium is reached

13

between the DOC in the pore water and the amount of potentially soluble

DOC sorbed on the peat matrix.

(ii) The second phase involves much slower DOC release; this is governed by

microbial DOC production and mineralization, although also mediated by

sorption.

What we refer to here as the “first steady state” or “sorption equilibrium” is in fact a

local equilibrium between the DOC solution and the initial SSOC concentration in the

peat matrix; this is reached after the rapid release of DOC in the initial phase of the

experiment. Subsequently, the DOC concentration changes more gradually, as a result

of microbial production and mineralization. It can be shown that a global steady state

is also eventually reached, i.e. the system eventually evolves to a state in which the

production of DOC is balanced by the sum of the mineralization of both DOC and

SSOC. However, this global steady state is mostly of theoretical interest, because it is

only reached after a very long time.

Under laboratory conditions there is no advective water flow and diffusive mixing

within the peat sample can be considered instantaneous. Therefore, equations 1-2 can

be simplified as:

,aMPaV

MaK

t

ac1sdes

cDdes

c [3]

,aaV

MaK

t

as2sdes

cDdes

s [4]

14

where M is the mass of the peat sample, V the volume of water added, while ac and as

are the amounts of DOC and SSOC, respectively, defined as ac=cV and as=sM.

We assumed that, in a continuous laboratory incubation of peat with stagnant water,

after the first steady state between DOC and SSOC is reached the sorption reaction

system is in quasi-equilibrium if examined at a daily time step. That is because the

change in total soluble OC is so small that the system continuously adjusts to the state

when s(t)=KDc(t). The analytical solution of equations (3)-(4) can then be obtained:

,

V

MKk1

MP))tt(

V

MK1

V

MKk1

exp()

V

MKk1

MPVc(V)t(c

Ds1

1

D

Ds

1

Ds1

1

[5]

where c1 is the DOC concentration at time t1, the end of the initial phase of incubation

when the first steady-state is established between the sorbed and dissolved phases.

Assuming that the steady-state has been reached, c1 can be calculated from the mass

conservation as follows:

,1

V

MK

cV

Ms

c

D

00

1 [6]

where c0 and s0 are the initial DOC and SSOC concentrations, respectively.

The initial DOC concentration, c0, and the pore water DOC concentration in the first

sorption steady state, c1, can be measured in a laboratory peat incubation. Our peat

15

incubation experiments were performed with two types of water: water from the

stream at the outlet of the mire studied in Part II (c0 =44 mg L-1 DOC), and tap water

(deep groundwater) from a research station near the site (c0 =0.3 mg L-1 DOC) (details

in Appendix). The stream water (44 mg L-1 DOC) was also incubated on its own to

measure DOC mineralization in the water.

KD, s0 : data from laboratory incubations

The time t1 was determined from the DOC vs. time curve as the critical point at which

the first local steady state is reached. Using two measurements of the DOC

concentration (c1) at time t1, one for each of the two different initial DOC

concentrations (c0), we used equation (6) to estimate, simultaneously, the partitioning

coefficient KD and the initial SSOC concentration, s0.

P. 1: data from laboratory peat and water incubations

DOC concentrations from the readings starting from t1 were inserted into equation (5)

and the values of P and 1, optimized using the least squares method. All experiments

were performed at 20°C, so that we could estimate the basal rates Pbasal and 1basal.

The values of parameters des and ks were as presented in Table 2. The data from the

incubations of water without peat were used to calculate the DOC mineralization rate

(first order decay coefficient) in water alone (Table 2). Our original intention was to

compare the resulting value with the corresponding DOC mineralization coefficient,

1basal, obtained by optimization, using the data from peat incubations to determine

whether the presence of a peat matrix changes the DOC solution mineralization rate.

16

However we discovered that the optimization procedure was insufficiently sensitive to

measure 1basal, and the lower range of this parameter was not detectable, presumably

because the mineralization rate is an order of magnitude lower than the rate of

production determined at the same time during the optimization. Therefore, only the

upper range of 1bsal was determined by optimization. The range of the DOC

production rate, presented in Table 2, reflects experimental uncertainty.

P, 1: literature data

The least square optimization for equation (5) was also utilized to process the data

relating to DOC release rates from the long-term laboratory peat incubations found in

the literature [Moore and Dalva, 2001; Aguilar and Thibodeaux, 2005; Chow et al.,

2006]. The listed studies did not contain information about the values of model

coefficients KD (sorption partitioning coefficient) or ks (degree of reduction in the

microbial mineralization when the soluble OC is sorbed) required to calculate the

DOC production and mineralization rates in accordance with equation (5). The

original concentration of SSOC (amount present sorbed on the peat matrix, s0) was

not reported either. To overcome this problem we used a range of parameter values:

KD=0.019-0.091 Lg-1 (minimum taken from Qualls [2000], maximum from Rasse et

al. [in revision]); ks=1/6-1/3 [Kalbitz et al., 2005]; s0=1-10 mgg-1(assumed)). The

results of the calculations therefore present possible ranges for the parameters of

interest (Table 3). When appropriate, the mineralization rate was estimated from the

optimization, otherwise the literature range of the decay constant of the slowly

decaying DOC fraction in throughfall, soil and stream water ( 1=1 10-5 -4 10-5 h-1,

[Qualls 2000]) was used. The range of the calculated production rate here reflects

17

both experimental uncertainty and uncertainty included in the estimates of the other

parameters that were used.

Model analysis

To test the validity of our hypothesis that long-term laboratory incubations could be

considered as two discrete phases, we examined the partitioning between net sorption

(adsorption minus desorption) and net microbial production (production minus

mineralization) for the day-to-day DOC concentrations during the incubations .This

was done by numerically estimating each term in equation (3). Hourly simulated

outputs were accumulated within daily time intervals and two starting DOC

concentrations (0.3 and 44 mgL-1) were considered, in accordance with our

experimental design.

The local sensitivity coefficients for the model parameters KD, des, P and 1 were

calculated by differential analysis [e.g., Saltelli, 2000]. Differential analysis is often

used in studies of chemical kinetics [e.g. Saltelli, 2000]. Equations are derived from

the original equation system to simulate the partial derivatives (local sensitivity

coefficients) of the variable of interest (e.g. DOC concentration) with respect to the

model parameters The direct method, utilizing the simultaneous numerical integration

of a combined system of equations (3)-(4) and forward sensitivity equations, was then

applied here using the CVODES programming package [Cohen and Hindmarsh,

1996; program codes and documentation available from

http://www.llnl.gov/casc/sundials/].

18

Two important characteristics: (i) the DOC during the first steady-state (c1), and (ii)

the rate of DOC release into the water during the second stage of the incubation, were

further evaluated. First, we examined the dependency of c1 on the initial SSOC

concentration (s0) and the peat-to-water ratio (M/V) given by equation (6). In addition,

the net DOC release per unit peat mass per unit time,Mt

Vc, was derived by

applying equation (5) to the time interval t. By changing the input parameters that

are used in the resulting equation, we analysed the influence of such factors as the

initial SSOC concentration (s0) and peat-to-water ratio (M/V) on the rate of DOC

release observed during the laboratory incubations.

Results

Parameterization results

Our estimate of the sorption partitioning coefficient KD was 0.033 Lg-1; lower than

that estimated by Rasse et al. [in revision] based on a surface peat sample from a mire

in Norway (KD=0.091 Lg-1) and slightly higher than 0.019 Lg-1, the value we

calculated using the data by Qualls [2000], who analysed a surface peat sample from a

marsh in the Everglades (Florida, USA). Our estimate of the initial SSOC

concentration s0=2.0 mgg-1 is close to that in a Lamnela peat sample from a drained

black spruce forest soils in Alaska (s0=2.7 mgg-1 [Qualls and Richardson, 2003]).

As summarized in Table 3, the rate of microbial production can be substantially

higher than the rate of release of DOC from peat into water. As interpreted by our

model, the rate of release is less than the rate of production because soluble OC

produced by microbial transformations of the peat is distributed by sorption between

19

the solid and dissolved phases. In experiments, measurements of DOC in solution will

only record the increases in the dissolved phase.

The estimated rate of microbial production we found in our incubation of surface

Sphagnum peat is of the same order as that from the sapric Sphagnum peat sample

incubations of Moore and Dalva [2001] (Table 3). A fibric peat sample [Moore and

Dalva, 2001] and a peat sample from a river delta [Aguilar and Thibodeaux, 2005]

had higher rates of microbial production, while a different river delta sample [Chow et

al., 2006] showed much lower production, resulting in net mineralization. The upper

limit for the mineralization rate in our sample, defined by the optimization, was 1.0

10-4 h-1. This is of the same order as the upper range of the decay constant of the

slowly decaying DOC fraction in throughfall, soil and stream water [0.4 10-4 h-1,

Qualls, 2000].

The role of sorption as a regulator of DOC concentrations at different time

intervals during peat incubations

The role of sorption can be illustrated by comparing modeled net sorption and net

microbial production (Figure 4 a). For the chosen parameter range, sorption strongly

dominated over production during the first 39 h of incubation. It was difficult to

estimate the partitioning later in the experiment as here we compared very small

numbers so that numerical errors became an important issue. To overcome this

problem we compared the weighted local sensitivity coefficients (i

ip

cp ) calculated

for the main model parameters (pi) with hourly time resolution and high precision

(Figure 4b). The kinetic constant of sorption, des, was the most important parameter

for the first 14 hours of incubation, after that the sorption partitioning coefficient KD

20

was most important. After 39h the sensitivity coefficient for the production rate P

became higher than that for the kinetic desorption des coefficient.

The two coefficients related to sorption have distinctly different meanings. While des

is indicative of the time-dependent kinetics of the process, KD characterizes the

partitioning between the solid and dissolved phase when the system is in a steady state

in relation to sorption. Both methods that we used to distinguish sorption from

microbial processes (Figure 4a,b) indicated that partitioning the DOC release into two

phases was valid. There was a clear division between the first phase when sorption

kinetics dominated and the second phase when the observed DOC concentration was

in equilibrium, with adsorptive partitioning of microbially-produced DOC.

Abiotic factors affecting the first (local) steady-state DOC concentration and the

rate of DOC release during the second phase of the long-term peat incubation

As expected according to the model, the amount of DOC that was rapidly released

during the first phase of incubation (c1) exhibited a linear relationship with the

amounts of soluble OC originally sorbed on the peat matrix (Figure 5a). Increases in

the ratio of peat mass to water volume resulted in non-linear increases in the amount

of DOC released (Figure 5b). There was also a notable difference in the DOC release

rate during the second phase of incubation depending on the experimental set-up:

The DOC release rate has a negative linear relationship to the initial SSOC

concentration, s0 (Figure 5c). The initial SSOC concentration presumably varies

naturally among peat samples of different origins and is also affected by the

rinsing or washing of the sample before the start of the experiment. Within the

21

expected natural range of SSOC concentrations the effect on DOC release rate is

presumably not high (up to 10%), but it is significant.

The DOC release rate is non-linearly negatively related to the peat mass-to-water

ratio in the experiment, M/V (Figure 5d). For example, DOC release per mass

unit is much smaller for an unsaturated sample than for a saturated or flooded

sample. This is noteworthy, since here we only deal with the

adsorption/desorption processes affected by the relative proportion of the solid

phase and the effect of moisture on microbial activities is not considered.

Summary and discussion

The convection–dispersion equation used in this paper was developed to describe the

transport and transformations of DOC in a peatland system. To our knowledge this

mathematical approach to the problem of DOC dynamics is new, although in

analogous fields of hydrology, such as pesticide transport in soil, solving the

convection-dispersion equation has become almost standard practice [Lapidus and

Amundson, 1952; van Genuchten and Wagenet, 1989; Baskaran et al., 1996].

The dynamic, rather than static, representation of sorption kinetics is one of the

strengths of the proposed model. In this paper we evaluated mathematically the

mechanism of DOC regulation by sorption kinetics studied in the laboratory

experiments [e.g., Qualls 2000, Rasse et al., in revision]. It is known and often cited

in relation to laboratory DOC studies that (i) the initially rapid release of DOC from

the peat matrix into newly added water is mainly due to abiotic processes [e.g.,

Qualls, 2000, Moore and Dalva, 2001; Aguilar and Thibodeaux, 2005; Chow et al.,

22

2006] and (ii) depending on the initial concentration, DOC can be either added to or

removed from the solution as a result of sorption [Qualls, 2000]. In the current study

we provide equations to explicitly describe these mechanisms.

By comparing the terms in the differential equation for DOC concentration and

performing differential sensitivity analysis we have been able to establish that there is

a clear division between the two phases of DOC release into water during long-term

peat incubations. The first phase was short (39 h for the model fitted to our laboratory

study) and governed solely by sorption, while further, much slower, increases in DOC

concentration during the second phase were due to microbial activity. In addition, the

derived equation (6) illustrated that, depending on the initial DOC and SSOC

concentrations, DOC concentration can either increase or decrease during the first

phase of incubation.

The model also suggests that, during the second phase, the additional soluble OC

produced slowly due to microbial transformations of the peat is distributed by

sorption between the solid and dissolved phases. Only the DOC contribution is

recorded when the change in solution concentration is measured, which means that the

“true” production rate is likely to be higher, and includes the fraction that becomes

sorbed (illustrated by comparing columns three and four in Table 3).

Using the model equations we were able to show how the release of DOC into water

during the long-term incubations can vary substantially according to the experimental

set-up. The DOC release rate into solution was affected by the peat-to-water ratio

(Figure 5d) and the initial amounts of soluble DOC sorbed on the peat matrix (Figure

23

5c). Both effects persisted for the whole time period we considered in our simulations

(five years). It was not only the amounts of DOC released initially, but also the

subsequent release rate that was affected by the initial SSOC concentrations and the

peat mass to water volume ratio. Different peat-to-water ratios resulted in different

rates of DOC release because the same net production was partitioned differently

between the sorbed and solid phases. This consideration is of great importance when

conclusions about microbial processes are drawn by comparing the results of different

laboratory incubations. Usually only changes in DOC concentration are measured and

directly converted to production rates, without taking into account the peat-to-water

ratio. The effect of sorption, rather than other factors such as, for example, the effect

of moisture on microbial activity, could account for the different DOC release rates if

data from two experiments with mixtures of peat and water in different ratios are

compared. The effect of sorption would be even stronger if water was flowing through

the peat, significantly changing the total amount of SSOC; in this paper, however, we

deal only with stationary water.

In Part I we only apply the model to laboratory conditions with stagnant water.

However, the general model formulation we presented here is sufficiently broad to use

in field applications. In the accompanying paper the model is applied to simulate the

variations in DOC concentrations observed during 1993-2001 in a stream draining a

small headwater mire in Northern Sweden. The sensitivity of the results to the

parameter choice is also discussed.

24

Appendix

Method for the long-term (165 day) incubation of a peat-water slurry with two

different initial DOC concentrations.

Two peat samples were collected, using a 5 × 5 × 70 cm peat corer, from

Kallkällsmyren, an 8 ha peat mire in northern Sweden (see Sirin et al. 1998 for a

detailed description) on 21 September 2005. Core 1 was taken from an area

dominated by cottongrass and sedge, while core 2 was taken from an area with

cottongrass, sedge and Sphagnum. A 10 cm3 plug of peat was taken from depths of 40

cm and 60 cm and combined to give a 20 cm3 plug (approximately 35 g wet wt, 3 g

dry wt). Peat plugs were then incubated in 300 mL pre-acid washed high-density

polyethelene bottles filled with water. Two types of water, with different DOC

concentrations, were used: tap water (GW = deep well groundwater) from a research

station near the site (0.3 mg L-1 DOC) and stream water (S) from the mire outlet (44

mg L-1 DOC). Each combination of water type (S and GW) and core (1 and 2) were

incubated in triplicate in the dark at 20ºC.

Periodic measurements (after 1h, 2d, 43d, and 165d) of absorbance at 254 nm were

made in order to calculate DOC concentration in the aqueous phase of the water–peat

slurry. At the time of each measurement, sample bottles were shaken vigorously for 1

minute, then allowed to stand for 30 minutes, after which 5 mL of the sample was

filtered (disposable Millipore 0.45 m MCE membrane filters) and analyzed for

absorbance using an HP 8452-UV Spectrophotometer. Incubation bottles were topped

up with sample water, leaving no headspace during the subsequent incubation. In

addition, bottles of water without peat were incubated as control treatments under the

25

same conditions. The stream water incubation (44 mg L-1 DOC) was used to measure

the rate of DOC mineralization in water alone. Absorbance at 254 nm was used to

estimate DOC concentration using a well-established rating curve for samples from

the stream at the Kallkällsmyren mire outlet:

25423 Ac N = 83; R2 = 0.85; RMSE = 4 mg L-1 (13%) (1.1A)

Where c is DOC concentration (mg L-1) and A254 = absorbance at 254 nm (AU)

measured in a 1 cm quartz cuvette. Measurements of DOC (Combustion and

measurement as CO2 using a Shimadzu TOC-VCPH analyzer) for selected samples

from the incubation experiment gave values always within 20% of the value estimated

from eq. 1.1A.

26

Acknowledgements

We thank Daniel Rasse for insightful comments and suggestions, as well as for

sharing the data from his on-going research. This work has been supported by the

NECC project funded by the Nordic Centre of Excellence (NCoE) Pilot Programme

and by the Swedish Research Council for Environment, Agricultural Sciences and

Spatial Planning.

27

References

Ågren, G. I., and E. Bosatta (1998), Theoretical Ecosystem Ecology -Understanding Element Cycles, Cambridge University Press.

Ågren, G. I., R. McMurtie, W. J. Pastor, J. Pastor, and H. H. Shugart (1991), State-of-the-art of the models of production-decomposition linkages in conifer and grassland ecosystems, Ecological Applications, 1, 118-138.

Aguilar, L., and L. J. Thibodeaux (2005), Kinetics of the peat soil dissolved organic carbon release from bed sediment to water. Part 1. Laboratory simulation,Chemosphere, 58, 1309-1318.

Baskaran, S., N. S. Bolan, A. Rahman, and R. W. Tillman (1996), Non-equilibriumsorption during the movement of pesticides in soils, Pesticide Science, 46, 333-343.

Billett, M. F., S. M. Palmer, D. Hope, C. Deacon, R. Storeton-West, K. J. Hargreaves, C. Flechard, et al. (2004), Linking land-atmosphere-stream carbon fluxes in a lowland peatland system, Global Biogeochemical Cycles 18 GB1024,doi:1010.1029/2003GB002058.

Blodau, C., N. Basiliko, and T. R. Moore (2004), Carbon turnover in peatland mesocosms exposed to different water table levels, Biogeochemistry 67, 331-351.

Boyer, E. W., G. M. Hornberger, K. E. Bencala, and D. McKnight (1996), Overview of a simple model describing variation of dissolved organic carbon in an upland catchment, Ecological Modelling 86, 183-188

Chow, A. T., K. T. Kenneth, S. Gao, and R. A. Dahlgren (2006), Temperature, water content and wet-dry cycle effect on DOC production and carbon mineralization in agricultural peat soils, Soil Biology and Biochemistry, 38, 477-488.

Cohen, S. D., and A. S. Hindmarsh (1996), CVODE, a Stiff/nonstiff ODE Solver in C, Computers in Physics, 10, 138-143.

Cole, J. J., Y. T. Prairie, N. F. Caraco, W. H. McDowell, L. J. Tranvik, R. G. Striegl,C. M. Duarte, et al. (2007), Plumbing the global carbon cycle: Integrating inland waters into the terrestrial carbon budget, Ecosystems 10 171-184

Crow, S. E., and R. K. Wieder (2005), Sources of Co-2 emission from a northern peatland: Root respiration, exudation, and decomposition, Ecology, 86 1825-1834.

Fraser, C. J. D., N. T. Roulet, and T. R. Moore (2001), Hydrology and dissolved organic carbon biogeochemistry in an ombrotrophic bog, Hydrological Processes,15, 3151-3166.

Freeman, C., C. D. Evans, and D. T. Monteith (2001a), Export of organic carbon frompeat soils, Nature, 412, 785.

Freeman, C., N. Ostle, and H. Kang (2001b), An enzymic 'latch' on a global carbon store, Nature, 409, 149.

Futter, M. N., D. Butterfield, B. J. Cosby, P. J. Dillon, A. J. Wade, and P. G. Whitehead (2007), Modeling the mechanisms that control in-stream dissolved organic carbon dynamics in upland and forested catchments

Water Resources Research, 43, W02424.Granberg , G., H. Grip, M. Ottosson Lofvenius, I. Sundh, B. H. Svensson, and M.

Nillsson (1999), A simple model for simulation of water content, soil frost, and soil temperatures in boreal mixed mires, Water Resources Research, 35, 3771-3782.

Grieve, I. C. (1991), A model of dissolved organic carbon concentrations in soil and stream waters, Hydrological Processes, 5, 301-307.

Ivanov, K. E. (1981), Water movement in Mirelands., Academic Press, London. Jansson, M., A. K. Bergstrom, P. Blomqvist, and S. Drakare (2000), Allochthonous

28

organic carbon and phytoplankton/bacterioplankton production relationships in lakes., Ecology 81, 3250-3255.

Jonsson, A., G. Algesten, A.-K. Bergström, K. Bishop, S. Sobek, L. J. Tranvik, and M. Jansson (2007), Integrating aquatic carbon fluxes in a boreal catchment carbon budget, Journal of hydrology, in press.

Kalbitz, K., D. Schwesig, J. Retnemeyer, and E. Matzer (2005), Stabilization of the dissolved organic matter by sorption to the mineral soil, Soil Biology and Biochemistry, 37, 1319-1331.

Kortelainen, P., J. Mannio, M. Forsius, J. Kämäri, and M. Verta (1989), Finnish Lake Survey: The Role of Organic and Antropogenic Acidity, Water, Air, and Soil Pollution, 46, 235-249.

Lapidus, L., and N. R. Amundson (1952), Mathematics of Adsorption in Beds. VI. The Effect of Longitudinal Diffusion in Ion Exchange and ChromatographicColumns, J. Phys. Chem., 56, 984-988.

McKnight, D., and G. R. Aiken (1998), Sources and Age of Aquatic Humus, in Aquatic Humic Substances Ecology and Biogeochemistry, edited by D. O. Hessen and L. J. Tranvik, Springer.

McKnight, D., E. M. Thurman, R. L. Wershaw, and H. Hemond (1985), Biogeochemistry of Aquatic Humic Substances in Thoreau's Bog, Concord, Massachusetts, Ecology, 66, 1339-1352.

Michalzik, B., E. Tipping, J. Mulder, J. F. G. Lancho, E. Matzner, C. L. Bryant, N. Clarke, et al. (2003), Modelling the production and transport of dissolved organic carbon in forest soils, Biogeochemistry, 66, 241-264.

Moore, T. R., and M. Dalva (2001), Some controls on the release of dissolved organic carbon by plant tissues and soils Soil Science, 166, 38-47.

Moore, T. R., N. T. Roulet, and J. M. Waddington (1998), Uncertainty in Predicting the Effect of Climatic Change on the Carbon Cycling of Canadian Peatlands, Climatic change, 40, 229-245.

Neff, J. C., and G. P. Asner (2001), Dissolved organic carbon in terrestrial ecosystems: Synthesis and a model, Ecosystems 4, 29-48.

Qualls, R. G. (2000), Comparison of the behavior of soluble organic and inorganic nutrients in forest soils, Forest Ecology and Management, 138, 29-50.

Qualls, R. G., and C. J. Richardson (2003), Factors controlling concentration, export, and decomposition of dissolved organic nutrients in the Everglades of Florida, Biogeochemistry, 62, 197-229.

Rasse, D. P., H. L. Stap, and J. Mulder (in revision), Sorption and diffusion kinetics of DOC in acid forest soils.

Reeve, A. S., D. I. Siegel, and P. H. Glaser (2000), Simulating vertical flow in large peatlands, Journal of Hydrology, 227, 207-217.

Romanov, V. V. (1961), Mire hydrophysics (in Russian), Leningrad. Roulet, N. T., P. M. Lafleur, P. J. H. Richard, T. R. Moore, E. R. Humphreys, and J.

Bubier (2007), Contemporary carbon balance and late Holocene carbon accumulation in a northern peatland, Global Change Biology, 13, 397-411.

Saltelli, A. (Ed.) (2000), Sensitivity analysis, 475 pp., Wiley, Chichester. Schindler, D. W., P. J. Curtis, B. R. Parker, and M. P. Stainton (1996), Consequences

of climate warming and lake acidification for UV-B penetration in North Americanboreal lakes. , Nature 379, 705-708.

Steinberg, C. E. W., S. Kamara, V. Y. Prokhotskaya, L. Manusadrianas, T. A. Karasyova, M. A. Timofeyev, Z. Jie, et al. (2006), Dissolved humic substances-ecological driving forces from the individual to the ecosystem level?, Freshwater

29

Biology, 51, 1189-1210. Tranvik, L. J., and M. Jansson (2002), Terrestrial export of organic carbon, Nature,

415, 861-862. van Genuchten, M. T., and R. J. Wagenet (1989), Two-Site/Two-Region Models for

Pesticide Transport and Degradation: Theoretical Development and Analytical Solutions, Soil Science Society of America Journal, 53, 1303-1310.

30

Figure captions:

Figure 1. Schematic diagram of the model formulation and structure.

Figure 2. Schematic diagram of the modeled pools of soluble organic matter and

processes that control microbially-mediated transformation (production and

mineralization) and abiotic exchanges (adsorption and desorption) between the solid

and dissolved phases. Parameters are described in detail in Table 1.

Figure 3. Schematic diagram of DOC transport within the peat profile, based on the

convection-dispersion framework.

Figure 4. (a) Modeled estimates of partitioning of the DOC release rate observed in

laboratory peat incubations into net sorption (desorption minus adsorption) and net

production (microbial production minus mineralization) components in incubations

with two starting DOC concentrations (c0) in water. (b) Normalized first-order local

sensitivity coefficients computed by the forward direct method for the main

parameters in the model applied to laboratory incubations (calculated for an

experiment with c0=0.0 mgL-1).

Figure 5. Simulated effect of the initial soluble solid organic carbon (SSOC)

concentration, and peat-to-water ratio (M/V) on the first steady state DOC

concentration (a, b); and on the DOC release rate during the second phase of a long-

term peat incubation (c, d) (KD=0.033 Lg-1, c0=0.0 mg L-1, P=1.0 10-3 mgg-1h-1, 1=4

10-5h-1, ks=0.3 ; (a, c) M/V=10.0 gL-1; (b, d) s0=2.0 mg g-1). A Sphagnum peat sample

with density of 0.02 gcm-3 and porosity 0.98 is saturated at M/V=20gL-1 and

31

oversaturated when M/V is below 20gL-1. Lines “20 days” and “60 days” in subplot

(d) coincide.

32

Table 1. Model variables: symbol definitions and units

Symbol Definition Dimension Units

c DOC solution concentration M L-3 g cm-3 (eq.1,1a);mg L-1(eq. 3-6)

s Sorbed soluble organic carbon (SSOC) concentration

M0 g g-1 (eq2, 2a);mg g-1 (eq.3-6)

ac Amount of DOC in the solution M mg

as Amount of soluble organic carbon sorbed

M mg

Volumetric water content L0 -

Peat bulk density M L-3 g cm-3

q qx, qy,qz Water flux (Darcian velocity) L T-1 cm h-1

D Dx, Dy,Dz Dispersion coefficient L2 T-1 cm2 h-1

M Mass of peat M g

V Volume of water L3 L

qs Specific discharge LT-1 cm h-1

33

Table 2. Model parameters. The list includes all model parameters, but only the parameters relevant to the model application for laboratory conditions are specified. The other parameters are specified in Part II (Table 1) where the model is applied to field conditions.

Parameter Symbol Estimated/ taken from Units Value/ rangeSorptionDesorption kinetic constant des Rasse et al., in revision

Surface peat from the mire in Norway

h-1 0.078

Partitioning coefficientdescribing the sorptionequilibrium

KD Estimated from eq. 6 L g-1 0.033

Initial DOC concentration inwater

c0 Measured mg L-1 Tap water: 0.3 Stream water: 44

Initial SSOC concentration s0 Estimated from eq. 6 mg g-1 2.0MicrobialThe rate of microbial DOC production at 20°C

Pbasal Estimated from least square optimizationof eq. 5

mggh-10.4-2.4 10-3

The rate of the microbialmineralization of DOCat 20°C

1basal Estimated fromwater incubationleast square

optimization of eq. 5

h-1

0.4 10-4

…-1.0 10-4

The constant reducing themicrobial mineralizationrate when the soluble OC is sorbed

ks Kalbitz et al. [2005]organic horizon of a podzol soil

- 1/6-1/3

Modifier to account for theeffect of temperatureon P and 1

Q10 Not used here: constanttemperature 200C

- -

Modifier to account for theeffect of oxygen supply on P

kanP Not used here: aerobicconditions only

- -

Modifier to account for theeffect of oxygen supplyon 1

kan Not used here: aerobicconditions only

- -

HydrologicalMolecular diffusion coefficient

D0 Not used here: assumedthat the sample is homogeneous

cm2 h-1 -

Dispersivity Not used here: assumedthat the sample is homogeneous

cm -

34

Table 3. The rates of DOC production, Pbasal, and mineralization, 1basal, derived fromthe long-term laboratory peat incubations at approximately 20°C.

Pbasal,Sample/location Peat-to-waterratio,gL-1

Net DOC releasemeasured,10-2mgg-1d-1

10-2mgg-1d-1 mgg-1h-1

1basal, h-1

Source and comments

SurfaceSphagnum peat / small headwatermire

10-14 1.2-2.3 1.8-5.8 7.5 10-4-2.4 10-3

a…-1.0 10-4

This study(c0=0.3 mg L-1)

SurfaceSphagnum peat / small headwatermire

14 0.7 1.0-1.7 4.2 10-4-7.1 10-4

a - This study(c0=44 mg L-1)

FibricSphagnum peat / bog

10-100b 4.8 7.4-43.3 3.1 10-4-1.8 10-2

a - Moore and Dalva,[2001]

SapricSphagnum peat / bog

10-100b 3.0 5.0-26.4 2.1 10-3-1.1 10-2

a - Moore and Dalva[2001]

Surface peaty soil with 19%OC/ river delta

140 5.3 18.2-56.2 7.6 10-3-2.3 10-2

a - Aguilar and Thibodeaux [2005]

Subsurfacefibrous peat /river delta

3.3 103 -6.6 10-4 c 1.4 10-3-8.6 10-3

3.3 10-6-1.9 10-5

1.9 10-4-6.7 10-4

Chow et al. [2006] d

Subsurfacefibrous peat /river delta

1.4 103 -7.0 10-3 c 7.9 10-3-4.6 10-2

1.4 10-6-6.7 10-6

1.1 10-3-3.7 10-3

Chow et al. [2006] d

Subsurfacefibrous peat /river delta

500 -3.1 10-2 c 3.4 10-3-1.6 10-2

4.6 10-4-1.4 10-3

Chow et al. [2006] d

a Not detectable by the parameter optimizationb Presented as a range in the source publication [Moore and Dalva, 2001]. c From day 7 to day 60 d this study uses different techniques to determine the DOC release rate, so the data on total soluble OC (SSOC+DOC) were used for the calculations here

35

Processes Sorption Microbial Hydrological

Time scale Short/hours Long/weeks-years Any, dependent on flow rate

How modelled

Linear kinetics Zero-orderproduction, first- ordermineralization

A gradient of the the DOC flux density

Parameters KD,, τdes P, µ1ks

Q10, kanP , kanµ

D0 , λ

MMWH modelPaper I. Implementation: numerical

or analytical solution of CDE

Paper II. Implementation: numericalsolution of CDE

Wat

er flu

x

T, water content

Figure 1. Schematic diagram of the model formulation and structure.

36

Solid OCMs≈M>>ρs+θc

Soluble OC sorbedon solid OCSOC(ρs)

Dissolved OCDOC(θc)

Sorption-τdesρKDc

Desorptionτdesρs

ProductionPρ

Mineralizationµ1ksρs

Mineralizationµ1θc

Figure 2. Schematic diagram of the modeled pools of soluble organic matterand processes that control microbially-mediated transformation (production andmineralization) and abiotic exchanges (adsorption and desorption) between the solid and dissolved phases. Parameters are described in detail in Table 1.

37

0.4 0.6 0.8 1

Water content, θ

z

qzc

qxc qxc

-θDdc/dz

dilution

advectivefluxdispersiveflux

Water table

Depth to thecatotelm

Figure 3. Schematic diagram of DOC transport within the peat profile, based on the convection-dispersion framework.

38

0

0.1

0.2

0

0.1

0.2

0 24 48 72 96

0

2

4

6

8

Incubation time, h

KD

des

P

net productionnet sorption

a

b

c0=0 mgL-1

Figure 4. (a) Modeled estimates of partitioning of the DOC release rate observed in laboratory peat incubations into net sorption (desorption minus adsorption) and net production (microbial production minus mineralization) components in incubations with two starting DOC concentrations (c0) in water. (b) Normalizedfirst-order local sensitivity coefficients computed by the forward direct methodfor the main parameters in the model applied to laboratory incubations (calculatedfor an experiment with c0=0.0 mgL-1).

c0=44 mgL-1p id

c/dp

i, m

gL-1

h-1D

OC

rel

ease

, mgg

-1h-1

39

0

20

40

60

80

0 5 100.0

0.5

1.0

1.5

2.0

2.5

0 50 100 150 200

after 20 daysafter 60 daysafter 5 years

after 20 daysafter 60 daysafter 5 years

a b

c d

Figure 5. Simulated effect of the initial soluble solid organic carbon (SSOC)concentration, and peat-to-water ratio (M/V) on the first steady state DOC concentration (a, b); and on the DOC release rate during the second phase of a long-term peat incubation (c, d) (KD=0.033 Lg-1, c0=0.0 mgL-1,Pbasal=1.0.10-3 mgg-1h-1, 1basal=4.10-5h-1, ks=0.3 ; (a, c) M/V=10.0 gL-1;(b, d) s0=2.0 mgg-1). A Sphagnum peat sample with density of 0.02 gcm-3

and porosity 0.98 is saturated at M/V=20gL-1 and oversaturated when M/V is below. Lines “after 20 days” and “after 60 days” in subplot (d) coincide.

Initial SSOC concentration, s0, mgg-1 M/V, gL-1

DO

C r

elea

se, 1

0-2m

gg-1

day-1

Fir

st s

tead

y st

ate

[DO

C],

c1,

mgL

-1

40

I

IV

I

Submitted to Water Resources Research

_____________________________________________________________________

A model of dissolved organic carbon concentrations in peat and stream waters.

II: Application and sensitivity analysis

Alla Yurova1, Andrey Sirin2, Kevin Bishop3 and Hjalmar Laudon4

1Department of Physical Geography and Ecosystems Analysis, Lund University, Lund, Sweden

2Laboratory of Peatland Forestry and Hydrology, Institute of Forest Science RAS, Uspennskoye,

Russia

3Department of Environmental Assessment, Swedish University of Agricultural Sciences, Uppsala,

Sweden

4 Department of Ecology and Environmental Sciences, Umeå University, Umeå, Sweden

_____________________________________________________________________

Abstract:

A new model of DOC concentrations and fluxes in peatland water was introduced in Part I of this study. Here, in Part II, the model is used to simulate DOC concentrations in the outlet of the steam draining a small headwater mire in Northern Sweden during the period 1993–2001. A relatively good model fit (MBE=-0.5-2.5 mgL-1, Willmottindex of agreement d>0.7) was found for all the categories of stream discharge, except periods with very low flow (q<0.3 mm day-1). Heterogeneity within the vertical peat profile simulated by the model was sufficiently strong to introduce significant fluxes of DOC due to sorption and dispersion. Previous models of DOC rely on bucket representations of the water movement with only a few vertical layers, so it is worth examining whether such simplifications could result in the loss of someepisodic variations in DOC concentrations. When seeking explanations for the interannual variability in DOC concentrations we, like previous authors, could find the influence of temperature, flow path and intensity. However, the model has helped to demonstrate that the system may have some “memory”: the store of sorbed soluble organic carbon (SSOC) in one year affects the DOC concentrations and fluxes in the following year._____________________________________________________________________

Keywords: dissolved organic carbon, peatland, model, uncertainty

* Author for correspondence (alla.yurova@nateko.lu.se)

Introduction

Peatlands are, collectively, one of the major sources of dissolved organic carbon

(DOC) in boreal watercourses [Urban, 1988; Dillon and Molot, 1997; Hope et al.,

1997; Gergel et al., 1999; Laudon, 2004; Ågren et al., 2007]. Changes in DOC

outputs from peatlands are of interest because (inter alia) they affect both regional

carbon balances [e.g., Jonsson et al., 2007] and aquatic biota [e.g., Laudon et al.,

2005; Steinberg et al., 2006]. Recent analyses have also emphasized the importance of

including DOC fluxes when resolving the total C balance of a peatland [Roulet et al.,

2007; Sagerfors et al., 2007]. Not only average DOC fluxes are significant, but also

interannual variation is comparable to that of vertical net carbon exchange [Roulet et

al., 2007]. Models are needed to test hypotheses regarding the causes of the seasonal

and interannual variability that has been recorded, and to predict DOC fluxes from

peatland catchments under different climate change scenarios and other human

impacts [e.g. Futter et. al, 2007].

Given the strong variations of DOC concentrations from peatlands, modeling them is

challenging. It was found that DOC is negatively correlated with water flow through a

system for a number of sites within the boreal region [Laudon et al., 2004; Fraser et

al., 2001]. The relationship is, however, demonstrable only if data from the major

hydrological event of the year, the spring flood, are included in the calculations

[Fraser et al., 2001]. Across the range of flow intensities the DOC concentration and

its relationship to flow can vary both seasonally and annually. The role of particular

environmental factors, including temperature and precipitation, in determining the

variability of the DOC concentration and fluxes is still open to debate [e.g., Freeman

et al., 2001a; Tranvik and Jansson, 2002; Worrall et al., 2004; Erlandsson et al., in

2

review]. It is widely appreciated that both in situ microbial DOC transformation

(production and mineralization) and hydrological transport are important influences

on DOC concentrations [e.g., Tranvik and Jansson, 2002], but the relative

contribution of each factor remains largely unknown. Experimental evidence supports

the view that the sorption equilibrium, based only on physico-chemical properties, is

likely to be as important for the export of the DOC from the wetland as biotic and

hydrological processes [Qualls and Richardson, 2003].

In Part I of this study we introduced a novel (to our knowledge) application of the

convection–dispersion equation to model DOC concentration and transport in a

peatland. The main processes included in the proposed model were: microbially

mediated DOC production and mineralization, hydrological transport and adsorptive

exchanges between two components of the system, namely the solution DOC and

sorbed soluble organic carbon (SSOC) of the peat matrix. In Part I we also discussed

the importance of sorption as a major process regulating the DOC concentration in

laboratory incubations. It was shown that kinetics equations describe the behavior of

such systems during the first 1-2 days of incubation better than steady state

approximations. An examination of the sorption kinetics associated with the water

flow in a vertically heterogeneous peat profile was left for Part II, this paper. This is a

key study to determine whether the model is valid for field situations. Here we apply

the convection–dispersion equation, at a relatively fine vertical resolution, to the

processes within the acrotelm. The model output should describe a degree of

heterogeneity that current DOC models do not consider, since they rely on bucket

[Neff and Asner, 2001; Michalzik et al., 2003] or semi-distributed [Boyer et al., 1996;

Futter et. al., 2007] hydrological approaches, incorporating very few vertical layers.

3

We tested the proposed model using long-term discharge and DOC concentration data

from the stream draining a small headwater mire, Kallkällsmyren, in Northern

Sweden [Sirin et al., 1998] collected during the period 1993–2001. The main

objective of Part II was to assess the applicability of the proposed model and

determine the main driving parameters by means of sensitivity and uncertainty

analyses. We addressed three specific questions when studying the mire:

(i) What are the components (production, storage and output) in the annual DOC

balance of the mire for the period 1993–2001?

(ii) How important is sorbed soluble organic carbon as a regulator of DOC

concentration over different time scales?

(iii) How significant is the vertical heterogeneity and exchange between the different

peat layers?

Methods

Study site

The Kallkällsmyren mire is situated in the Vindelns Forest Research Park, 60 km

northwest of Umeå, Northern Sweden (64 14 N, 19 46 E). The mire catchment is one

of the headwaters in the 68 km Krycklan catchment [e.g., Buffam et al., 2007]. It is

situated upstream of the Nyänget research catchment, which has been monitored since

1980 [Grip et al., 1990]. The site is located in the boreal zone with a cold temperate

humid climate. The mean annual temperature for the period 1981–2001, measured at

the nearby Svartberget Forest Research Station, was 1.5oC [Ottosson-Löfvenius et al.,

2003]. The average annual precipitation for the same period was 660 mm, with a

4

runoff of around 330 mm. About 45% of the precipitation falls as snow, and up to

60% of the annual runoff from the studied catchment can be snowmelt [Bishop et al.,

1995].

Kallkällsmyren is a complex mire occupying a topographic depression. The mire is 2

to 5 m deep and the underlying sediments are blocky glacial tills. On the eastern and

western sides the mire is surrounded by rather steep slopes (Figure 1). The slope to

the north is very gentle, and the southern part of the depression opens into the valley.

The slopes are afforested with spruce and pine and the soils are podzols. The mire is

drained by a single stream, which originates from the mire southern edge. A ditch was

dug for the stream in the 1920s, but the mire edge was left untouched.

The main section of the mire is raised, sloping down towards the edges; it has distinct

ombrotrophic features with vegetation composed of a sparse Scots pine stand with

dwarf-shrubs and Sphagnum mosses, together with Carex at the edges (Figure 1).

Mineratrophic fen vegetation occupies the southern area of the mire adjacent to the

stream, and there are also two belts of this vegetation situated on the eastern and

western sides of the raised section (Figure 1). Surface peat in both sections of the mire

is formed from Sphagnum remains with some Carex in the mineratrophic parts. The

degree of decomposition of the surface peat is less than 27% in the main section and

about 5% at the downslope edge.

The hydrological response at the outlet of the mire is very rapid, with peaks delayed

by no more than a few hours from the maximum precipitation or snowmelt inputs.

Isotope hydrograph separation has shown that 70-90% of the runoff during rain-driven

5

events is pre-event water [Bishop, 1991], while during spring floods, the pre-event

water component is 50–70%, including 5–15% derived from the deep ground water

[Laudon et al., 2007]. Sirin et al. [1998] and Petrone et al. [2007] provide more

detailed descriptions of the Kallkällsmyren mire.

The measurements

The stream water samples and discharge measurements were obtained from the stream

draining the mire about 50 m below its origin. The catchment area draining into the

sampling point is 18 ha, of which 9ha is occupied by the mire. During snow-free

periods of the year, discharge was computed hourly, based on water level

measurements made using a pressure transducer. Rating curves were derived using

timed measurements of discharge volume over the weir. Water sampling was

conducted weekly to biweekly during periods with low flow and up to daily during

snowmelt and a number of rainfall episodes since 1993. Samples were collected

monthly during 1991-1993 in acid-washed, 250 ml HDPE bottles and frozen until

analyzed. TOC was measured with a Shimadzu TOC-5000 total organic compounds

analyzer using catalytic combustion. As shown by Ivarsson and Jansson [1994] and

Gadmar et al. [2002] the particulate fraction of TOC is negligible in the type of water

studied here. Thus, while TOC was measured and is reported here, the results actually

reflect the DOC. The site has been the subject of intensive fieldwork [Bishop, et al.,

1990; Bishop and Pettersson, 1996; Laudon et al., 1999; Hruska et al., 2001] and

additional parameters, such as O18 isotopes and major ion concentrations have been

periodically measured in water samples collected at the stream origin as well as from

piezometers installed at various depths in the mire [Sirin et al., 1998].

6

To calculate annual DOC fluxes from measured DOC concentrations we used daily

average values of stream discharge and linearly interpolated the DOC concentrations.

The model

The main model has three key components: Heat flux, Hydrology and DOC Mass

Balance (Figure 2). We use the Mixed Mire Water and Heat model [MMWH;

Granberg et al., 1999] as modified by Yurova et al. [2007] to describe water fluxes,

vertical distribution of water content in the acrotelm, snow dynamics and heat transfer

in the peat profile. The system of two mass balance equations, one for the sorbed and

one for the dissolved OC(the latter in the form of a convection–dispersion equation),

is solved to estimate DOC concentration in the pore water and the flux entering the

stream, as described in Part I of this paper. Numerical values of the model parameters

were measured or calibrated (Table 1).

The model was developed in C++ code (Microsoft Visual C++) and the numerical

solution of the system of DOC and SSOC mass balance equations was calculated

using the CVODE program package [Cohen and Hindmarsh, 1996; program codes

and documentation available at http://www.llnl.gov/casc/sundials/]. The vertical

resolution of the model is 5 cm in the acrotelm and 10 cm in the catotelm. The time

steps within the model are hourly, and the results were analyzed using daily

averaging. Below are important details for the particular application of the model.

Hydrology and heat transfer

The MMWH model was designed to simulate the flow and moisture dynamics in the

upper, active, intermittently aerated zone of a peatland, the acrotelm. A zero flux

7

(impermeable) lower boundary is placed at the level of constant saturation, and it is

assumed that no flow occurs below this, i.e. in the catotelm [see Reeve et al., 2000 for

an examination of the validity of this assumption]. The model is based on steady-state

vertical moisture distribution curves [Romanov, 1961], which, as we discuss below,

was found to be applicable to the parameter values and time step used in this study.

The model considers variable lateral flow at different levels, to account for reduction

in hydraulic conductivity with depth [e.g., Ivanov, 1981]. The MMWH model was

first developed for the hydrology of an area with prescribed geometry and a single

vegetation type. Here it was applied to the whole mire and model parameters

characterizing the geometry, hydraulic conductivity and surface roughness of the mire

were combined into the calibrated parameters. The vertical peat profile was

discreetized into 5 cm deep layers, to partition the lateral flow. Water content was

integrated within each layer. It was assumed that vertical drainage is not restricted by

ice, but when the ice content of a layer is exceeds 0.7 lateral flow ceases.

Model equations and parameters are given in Granberg et al. [1999] and recent

changes are documented in Yurova et al. [2007]. Here, the model was applied to a

number of layers, rather than the whole active layer. Hence, layer-specific, rather than

integral hydraulic conductivity is used. The general form of the equation used to

calculate the specific discharge then becomes:

z

zcat zdzkq , [1]

where z (cm) is depth to the water table (negative when below the peat surface), zcat

(cm) is the acrotelm depth, (cm-1) is a lumped parameter and the hydraulic

8

conductivity, kz (cmh-1), is approximated as an exponential function of the depth

(kz=exp( z), where is a lumped parameter based on data presented in Ivanov [1981],

and the linear scaling coefficient before the exponent is included in ).

Evapotranspiration is assumed to decrease exponentially with depth (E=EPTexp(lEz),

where EPT (cmday-1) is potential evapotranspiration and lE is a lumped parameter.

During rain or snowmelt events it was assumed that drainage is vertical and that the

vertical water flux, qv, equals precipitation intensity, corrected for the change in water

content due to a rise in the water table. In addition, a generalized form of the Manning

equation [e.g., Beven, 2001] was applied here to describe the overland flow:

67.1sSO hq , [2]

where qS0 (cm h-1) is specific (area) discharge via surface overland flow, h (cm) is

water table depth above the peat surface and s is a lumped parameter.

The Heat balance equation was solved in a slightly different way than in the original

MMWH, mainly for practical reasons. We used a solution presented by Wania et al.

[in preparation] and implemented it using a C++ program. One important difference

from the formulation of Granberg et al. [1999] is that the snow layer and meltwater

pooled on the mire surface are included as additional layers in the vertical profile

when simulating the heat balance.

9

DOC concentration

To simulate DOC and sorbed soluble organic carbon (SSOC) concentrations in the

acrotelm, we applied equations (1a)–(2a), presented in Part I. The initial SSOC

concentration in the acrotelm peat was determined utilizing data from laboratory

incubations of surface peat from the studied mire, as described in Part I. For the

catotelm a more simple approach was taken. We assumed that a very slow production

rate [e.g., Freeman, 2001b] does not change the total DOC amount significantly on

the decadal time scale that we are studying here. We, therefore, assumed a constant

SSOC concentration in equilibrium with the dissolved phase, affected only by

diffusive flux to and from the acrotelm. Data from the sampling wells in the mire

were used to estimate SSOC concentration in the catotelm. A few model parameters

relevant to the field application are discussed in some detail below, the rest of the

parameters remained unchanged from those derived from the laboratory experiments

or the literature, as described in Part I of this study.

In our model we use a single value of Q10 [Moore and Dalva, 2001] as a modifier of

the model parameters Pbasal, 1basal and 2basal ( 2 is included in the model as

2= 1 ks). This approach should only be regarded as a first approximation. Chow et al.

[2006] demonstrated that Q10 is different for the mineralization of soil organic matter

(thought to be one step in DOC production) and mineralization of DOC. Similarly, it

is likely that Q10 also differs for the mineralization of DOC and SSOC. Unfortunately

no published data could be found that specifies Q10 for DOC production, DOC

mineralization or SSOC mineralization.

10

Two important model parameters are the ratios of anaerobic to aerobic DOC

production rate (kanP) and of anaerobic to aerobic mineralization rate (kan ). Anaerobic

to aerobic mineralization rate ratio may also differ between the organic matter, which

is a source of the DOC, and the DOC itself. Some studies have indicated that the net

DOC production rate is the same in oxic and anoxic conditions [Moore and Dalva,

2001], while others have demonstrated the effect of anoxia on DOC production and

mineralization [e.g., Freeman et al., 2001b]. In the absence of validated data from the

site, with which to quantify the parameters, we defined them based on the range of

values available within the literature. The upper limit was set at 1.0 (no effect of

anoxia) and the lower limit at 0.07 [Bergman, 1998; Moore and Dalva, 1997:

decomposition of surface peat in anoxic vs. oxic conditions in laboratory experiments]

for both coefficients (kanP and kan ).

In a current study, living vegetation is not modeled explicitly as a separate layer.

Instead DOC microbial production in the uppermost peat layer is increased by

multiplying the basal production rate, P, by fPs. fPs is a constant ratio between DOC

production from living tissues to that from peat. Is does not depend on the current

NPP, but is calculated assuming an average NPP (taken from eddy-tower estimates at

the near-by Degerö mire [Yurova et al., 2007]) and the ratio of DOC release from

Sphagnum to that from peat presented in Moore and Dalva, [2001]. Here, we ignore

the release of easily decomposable substances, such as root exudates, that have a short

lifetime in solution [e. g., Crow and Wieder, 2005]. The model was formulated for the

bulk of relatively refractory DOC reaching the stream, and not for the labile fractions

with short turnover times. Fractionation of DOC in water samples from the mire

stream and peat extractions (laboratory study in Part I) using the UV spectral

11

deconvolution method [Hassouna et al., 2007] demonstrated that 50% of the DOC is

hydrophobic, 18% transic and 22% hydrophilic. The proportion of these fractions is

the same for the stream and the peat sample extracts [M. Hassouna, unpublished

data]. Therefore, approximately 10% of the DOC is non-humic; this corresponds well

with earlier estimates from the study site [Bishop and Pettersson, 1996]. As shown by

McKnight et al. [1985] the hydrophobic acids in Sphagnum bog waters are mainly

fulvic; the hydrophilic acids present are intermediate compounds in the formation of

fulvic acids. Both are known to be relatively recalcitrant.

When water floods over the saturated peat profile, there is DOC exchange between

the surface water and the pore water in the uppermost peat layer. This is described in

the model using a formulation similar to the dispersion term in eq. (1), Part I, but

with a specific dispersion coefficient, Ds.

Modeling protocol

The model was run for the Kallkällsmyren mire site using a series of climate data for

temperature, precipitation and potential evapotranspiration recorded at the nearby

Svartberget research station [Ottosson-Löfvenius et al., 2003]. The parameters and

constants used are defined and listed in Table 1. The parameters , , s and lE, used

in the hydrology submodel, were optimized by minimizing the mean squared

difference between the modeled and observed specific discharge. The only three

parameters that were adjusted in the DOC concentration submodel were: the

dispersion coefficient for the overland flow, Ds, the ratio of anoxic to oxic DOC

production rate, kan ., and the ratio of anoxic to oxic DOC mineralization rate, kan ..

12

The following model simulations were conducted:

Model initialization was made using climate data from 1986 to 1993, the start of

the actual simulation period. Initial SSOC concentrations were varied from 1 to 5

mg g-1 in order to study the sensitivity of the system to initial conditions

A model run for the whole simulation period (1993–2001)

A series of Monte Carlo runs (1000) to determine the model sensitivity to

specific parameters and constants (Table 2).

The model results were analyzed on the basis of daily averages. Observed annual

DOC fluxes were estimated on the basis of interpolated DOC concentrations

measured at the stream outlet. To evaluate the annual DOC balance predicted by the

model, the total production and mineralization in the acrotelm were simulated

separately and the net production was estimated as the difference between the two

summed over a year. Annual net production in the acrotelm was then compared with

the stream export (flux) of DOC and the total amounts of SSOC and DOC stored in

the acrotelm as predicted by the model. In addition, the annual volume-weighted DOC

concentrations were calculated, based on the observed data and model simulations, to

eliminate the dominance of flow in the DOC fluxes. This reveals how the

concentrations vary between years. Volume-weighted concentration (both observed

and modeled) was then correlated to the modeled amounts of SSOC and DOC stored

in the acrotelm before the spring flood event, to determine whether the latter is related

to the former. Simulated concentrations weighted for different seasons were correlated

to the simulated stored amounts of SSOC and DOC in the preceding seasons. The

seasons were defined arbitrarily as: “winter”, the period when the surface temperature

13

of the peat was at or below 0°C; “spring”, the period from the start of snow melt till

the leveling off of the spring flood flow rate; and “summer” as the rest of the year.

To test the applicability of the steady-state vertical moisture distribution curves, the

1D Richards equations were solved for the observed range of precipitation amounts

and water table depths. The code of Bolster and Raffensperger [1996] based on

Hornberger et al. [1998] was used. van Genuchten soil moisture functions [van

Genuchten, 1980] were simplified, as they are in the current version of the MMWH

that is based on parameterization [Weiss et al., 1998]. The range for the saturated

hydraulic conductivity of the surface peat was derived from the literature [Romanov,

1961; Letts et al., 2000; Reeve et al., 2000].

To investigate the dynamics of the stream DOC concentration during hydrological

episodes when DOC concentration was high, two “rain-driven” flow events were

chosen to characterize the system behavior. The episodes differed in the intensity of

the rain event. In both cases the stream DOC concentration increased with the first

peak in the discharge. After the second (larger) runoff peak, the concentration

stabilized at approximately the same level as prior to the first event (12 July–30

August 1993, Figure 4a) and dropped about 10 mgL-1 during the second event (27

August–28 September 2001, Figure 4b). There was more rainfall during the second

event than the first. To illustrate the effect of vertical heterogeneity within the peat

profile during the second rainfall event, we partitioned the six layers within the

acrotelm with respect to: (i) net microbial production (mineralization) of DOC; (ii) net

release (removal) of DOC due to sorption; (iii) stream export; (iv) vertical advective

fluxes from (and to) adjacent layers; and (v) vertical diffusive fluxes. The sum of the

14

listed components for the whole of the 27 August–28 September 2001 event was

compared within each layer and between the layers.

Evaluation of the model performance, uncertainty, and sensitivity

Following the recommendations of Willmott [1982], we calculated the following

quantitative measures to evaluate the model performance: the mean bias error (MBE),

the root mean square error (RMSE) and the Willmott index of agreement, d. The

Willmott index of agreement is a dimensionless measure that varies between 0 and 1,

with d =1 indicating a perfect match. It is calculated as follows:

,|)'||'(|/)(11 1

22N

i

N

iiiii OMOMd [3]

where OMM ii ' , OOO ii ' , M-modeled and O-observed values.

In order to evaluate the model parameter-based uncertainty, we performed a global

multi-parameter sensitivity analysis [e.g., Saltelli, 2000]. Latin hypercube sampling

[LHS, McKay et al., 1979] was used to generate sets of parameters for use in the

Monte Carlo runs. Where possible, the parameter range was derived from the

literature (Table 2). Uncertainty and sensitivity analysis was performed on the

simulated DOC concentrations. Uncertainty bounds (lower and upper 10% quantiles)

were calculated based on the results of the Monte Carlo runs for each day during the

simulation period, following the GLUE methodology [Beven, 2001]. The Willmott

index of agreement was chosen as a likelihood measure. In the sensitivity analysis, the

15

rank partial correlation coefficient (RPCC) was chosen as a measure of the parameter

importance, indicating the extent to which a particular model parameter determines

the model output (range for the parameter vs. range in the output).

Results

General agreement between simulated and measured discharge, stream DOC

concentrations and fluxes: Error assessment

The simulated discharge corresponded well with the measurements at the mire outlet

(MBE=-0.11 mm, RMSE=1.5 mm d=0.85 mean=1.1 mm) and no systematic bias was

found for particular periods or discharge categories. The largest discrepancies (both

negative and positive) between the modeled and measured discharge occurred during

the spring flood.

Seven years after the start of the simulation (1986) the model performance was no

longer influenced by the initial conditions. The simulated values for DOC

concentration from the runs with different initial SSOC concentrations converged

after about five years.

We found relatively good agreement between the simulated and measured DOC

concentrations both overall (Figure 3) and for particular events (Figure 4). However,

the model performance varied greatly between the categories for total discharge

(Table 3). The model performed very poorly when the discharge was below 0.3 mm

day-1 (Table 3 and “cross” symbols on Figure 3). Above this threshold, the model

performance increased sharply and the best fit between the modeled and measured

16

DOC concentrations was found for the highest discharge category (Table 3). As a

result of the high correlation between discharge and DOC fluxes, when averaged over

a year, the agreement between the measured (interpolated) and simulated annual DOC

fluxes was very high (d=0.95; MBE=1.1 gm-2 year-1 RMSE=2.0 gm-2year-1).

The parameter associated with DOC microbial production, kanP, achieved optimal

model performance when set to a value of 0.07. This is the lower limit of the

anaerobic-to-aerobic decomposition rates for organic matter in the surface peat [e.g.,

Bergman, 1998; Moore and Dalva, 1997]. In contrast, the model performed best when

kan , the parameter associated with DOC mineralization, was allocated a value of one,

i.e. no reduction in DOC mineralization rate due to anoxia.

Uncertainty and sensitivity analysis

Prediction bounds determined by the uncertainly analysis covered most of the

observed DOC concentrations. The most common exceptions were the very low

concentrations recorded during periods of very low flow, below 0.3 mmday-1 (“cross”

symbols on Figure 3), which the model overestimated.

Sensitivity analysis (Table 2) revealed that the most influential model parameters are

the DOC production rate (Pbasal), Q10 , and the parameter accounting for the decease in

the DOC production rate under anaerobic conditions (kanP.). Also important, but less

so, are the DOC mineralization rate ( 1basal) and the sorption constants ( des and KD).

The other parameters exerted less influence over the model performance.

17

Model study of the vertical peat profile of the acrotelm

Estimates produced using Richards equations (not shown) indicated that the steady-

state moisture profile within the unsaturated zone is always reached within an hour of

the end of a rainfall event, even during high-intensity events. This is because of the

relatively shallow unsaturated layer, but, more importantly, the high hydraulic

conductivity of the surface peat. The result is promising as it indicates that the

MMWH model based on steady-state moisture distribution curves is likely to be a

valid approximation for the time scale studied here.

Figure 5 illustrates the simulated fluxes that contribute to the change in DOC

concentration in pore water; the values for the individual layers are summed over the

duration of the event being investigated, i.e. 27 August–28 September 2001 (Figure

4b). The figure does not show the entire summed DOC since it does not include

SSOC mineralization; it represents the total solute DOC concentration only. The first

interesting feature to note is a strong vertical heterogeneity between the 5cm layers.

Microbial DOC production decreases with depth under anoxic conditions and as the

temperature decreases. The stream export term also decreases, due to reduction in

hydraulic conductivity and therefore flow intensity with depth. The DOC is released

into the solution as a result of desorption in the three uppermost layers, but it is

removed from the solution by sorption in the layers below. Therefore, by sorption, the

layers more than 15 cm below the surface accumulate soluble OC carbon produced in

the layers above. Due to vertical advective transport, DOC is lost from the uppermost

layer and transferred to the layers below that were permanently saturated during the

events studied. Dispersion operates throughout the whole acrotelm and its input to the

18

total DOC balance is not insignificant. Another important feature to note from Figure

5 is that, for all the layers except the uppermost, the contribution of sorptive release or

removal of DOC is similar to or even exceeds the net microbial production.

Components of the annual DOC-SSOC balance

Two main simulated components of the annual DOC–SSOC balance, net microbial

production within the acrotelm and stream export, are of a similar order of magnitude

but are not equal for the year under consideration; in fact, the difference between the

two can be large (Figure 6). Interannual variation in the volume-weighted DOC

concentration reflects the variation in the amounts of DOC and SSOC stored within

the acrotelm in the previous year (Figure 7a, the DOC and SSOC are the model

estimates at the start of the current “hydrological” year, and are also equal to those at

the end of the preceding year). The correlation between the volume-weighted stream

DOC concentration in any given season and the store of DOC and SSOC in the

preceding season is high for the winter (r=0.75, P=5.8 10-5) and summer (r=0.87,

P=1.910-5), but lower for the spring (r=0.40, P=0.023) (Figure 7b).

Discussion

The relatively good overall results of the simulations indicate that the model is able to

describe the main mechanisms responsible for the variability in DOC concentration.

Very generally, concentrations rise to their highest levels (e.g. the episode in

September 1996 on Figure 3) when conditions favorable for DOC production (good

aeration and suitable temperatures) are combined with and/or shift the flow depth to

the layers that were not previously exposed to water flow. Some precipitation is,

therefore, always needed to mobilize the DOC present in the unsaturated layer and to

19

shift the sorption equilibrium by adding low DOC water so that additional amounts of

SSOC are released into the water being discharged. According to the model, this

explains the correspondence in timing of the initial increase in stream DOC

concentration and the increase in stream discharge following precipitation events

(e.g., Figure 4 a, b). Strong and prolonged events may, however, reduce the DOC

concentrations (e.g., the 27 August–28 September event shown in Figure 4b) through

the removal of significant amounts of SSOC from a particular layer. After the flow

ceases, there is a build up of new DOC in the pore water; simultaneously, SSOC may

accumulate on the peat matrix as a result of both in situ production and dispersive and

advective exchange with the adjacent layers. The model shows that the vertical peat

profile is strongly heterogeneous (Figure 5) and the role of dispersion is important

down to the bottom of the acrotelm. Sorption may release additional DOC in some

layers and remove it in others (Figure 5), depending on the sorption equilibrium.

Sorption makes a large contribution to DOC – in a particular layer similar to that of

microbial production (Figure 5). The lowest DOC concentrations were observed in the

stream during the spring flood (Figure 3). The flow is so intensive during this period

that significant amounts of SSOC are removed from the acrotelm while the

concentrations remain low. This could also explain the relatively slow recovery of

concentrations after the spring flood (up to one month).

No deep flow mechanism is considered in the current model, which describes only the

water balance in the acrotelm. We hypothesize that this is the main reason for the poor

model performance during periods of very low flow (Table 3 and “cross” symbols in

Figure 3). The water discharged during such periods appears to be groundwater from

the podzols surrounding the mire, which has bypassed the peat soils of the mire. In

20

terms of the total DOC budget, the ground water contribution is rather low, since the

flow itself is low; this input is, however, persistent. During periods of low flow, when

flow out of the mire virtually ceases, the ground water bypass can strongly influence

the DOC concentrations at the outlet. This bypass may be related to the

hydrochemical anomaly observed by Sirin et al. [1998] at a depth of 2 m in the

catotelm.

Results of the sensitivity study show that only a few parameters strongly influence the

model output. They are the DOC production rate and its modifiers that account for the

effects of temperature and anoxic conditions (Table 2). In addition, the model

predictions are also largely dependent on the chosen values of DOC mineralization

rate and the two sorption constants. These parameters require further investigation to

improve model generalization.

The balance between the net production of DOC and SSOC and the DOC

hydrological export, as predicted by the model (Figure 6), may go some way to

explaining the interannual variability in the total amounts of SSOC and DOC stored in

the acrotelm. Although the amount of stored SSOC and DOC in the acrotelm is

several times larger than the annual export, the variation in the amount stored as a

result of the dominance of either production or export in a given year is significant

and may persist in subsequent years. The system therefore has a “memory”, as

demonstrated by the good correspondence between the simulated storage and the

observed annual volume-weighted stream DOC concentration (Figure 7a). An

example of system “memory” has been described by Worrall et al. [2006] discussing

the role of drought in a mire in the UK. Unlike the example in Worrall et al. [2006],

21

the mire we studied exhibits a link between the summer concentrations and the

intensity and duration of the spring flood event, as shown by the very high correlation

between the storage of DOC and SSOC after the spring flood and the volume-

weighted summer stream DOC concentration [Figure 7b: summer]. The opposite is

also true: winter concentrations may be determined by the changes in SSOC amounts

during the previous summer. The situation during the spring flood is more

complicated and the amount of DOC and SSOC stored in the acrotelm before the

event is not such a large determinant of the stream DOC concentration (Figure 7b:

spring).

Net microbially mediated DOC production is influenced first by temperature, but also

by the proportion of the acrotelm in which aerobic conditions favor the exoenzymatic

hydrolysis of organic matter, a critical step in DOC formation. The model is

parameterized so that the DOC production rate is higher in aerobic than in anaerobic

conditions, but DOC mineralization is not. That resulted purely from the model

calibration. We can, however, speculate that the increased DOC production rate in

aerobic (vs. anaerobic) conditions is associated with enrichment of the solution with

phenolic compounds that may impair DOC mineralization [Freeman et al., 2001b].

The suppressed DOC production leads to the low net production simulated by the

model in the wet years (e.g., 1993, Figure 6). At this stage we cannot justify how valid

the assumption is, but it certainly deserves further attention.

Conclusions

In Part I of this study we described the model of DOC concentration and export from

a peatland. We highlighted the importance of modeling sorptive exchange as a

22

dynamic process, rather than a simple steady state, when considering data from

laboratory peat incubations. We also demonstrated how some abiotic factors may,

through the action of sorption, strongly affect the rate of DOC release into water, as

observed during long-term laboratory incubations.

In this paper the DOC concentrations recorded during 1993–2001 in an outlet of the

steam draining a small headwater mire in Northern Sweden were used to validate the

model. Relatively good agreement between the measured and simulated DOC

concentrations was found both for the whole simulation period and for particular

events (Figures 3 and 4). The results of this study indicate the following:

Observed concentrations at the outlet of the stream draining the studied mire

when the flow of intensity exceeds 0.3 mm may, largely, be modeled by the

combination of production, sorption and transport processes within the layered

acrotelm, i.e. by the convection–dispersion equation.

It is important to include vertical heterogeneity in the model in order to

adequately describe the processes that are either strongly influenced by depth

(production, horizontal transport) or driven by the vertical heterogeneity

(sorption, dispersion).

Sorption is an important process in determining DOC concentration and fluxes. It

may cause a release of DOC into the pore water in some layers and removal in

others, that is similar to the amounts of microbially produced DOC. During each

event, part of the DOC in the stream export originates from that which was stored

in the layer prior to the event, part is released into the solution from the peat

matrix due to desorption and the remainder is newly produced.

23

Interannual variability in the SSOC and DOC stores within the acrotelm depends

on the conditions for microbially mediated DOC production and mineralization

during the current year (temperature, aeration) and flow intensity. Some

“memory” is, however, characteristic of the system, meaning that the store in one

year affects the concentrations of fluxes in the following year. DOC export in the

studied mire is very closely linked to the total annual discharge

Though more detailed formulation of the subsurface flow and the ground water flow

may be needed to account fully for DOC transport within the mire, we contend that

even the relatively simple MMWH model can serve well, when the water movement

in the acrotelm dominates the total discharge. We also hope that the current model

formulation will be suitable for ecosystem models designed to examine the C balance

of peatlands.

Acknowledgements

This work has been supported by the NECC project funded by the Nordic Center of

Excellence (NCoE) Pilot Programme and by the Swedish Research Council for

Environment, Agricultural Sciences and Spatial Planning. Ishi Buffam is

acknowledged for preparing the map used in Figure 1.

24

References

Ågren, A., I. Buffam, M. Jansson, and H. Laudon (2007), Importance of seasonality and small streams for the landscape regulation of DOC export, Journal of Geophisical Research-Biogeosciences, accepted.

Bergman, I. (1998), Microbial carbon mineralisation in peat from a boreal mixed mire: regulation by environmental factors. PhD thesis, 33 pp, Swedish Univ. of Agricultural Sciences, Umeå.

Beven, K. J. (2001), Rainfall-runoff Modelling. The Primer, 360 pp., John Wiley & Sons, LTD, Chichester.

Bishop, K. H. (1991), Episodic increases in stream acidity, catchment flow pathways and hydrograph separation. Doctorial Dissertation, University of Cambridge,Cambridge, UK.

Bishop, K. H., H. Grip, and A. O'Neill (1990), The origins of acid runoff in a hillslope during storm events, Journal of Hydrology, 116, 35-61.

Bishop, K. H., Y. H. Lee, C. Pettersson, and B. Allard (1995), Methylmercury output from the Svartberget Catchment in Northern Sweden during spring flood, Water,Air, and Soil Pollution, 80, 445-454.

Bishop, K. H., and C. Pettersson (1996), Organic carbon in the boreal spring fllod from adjacent subcatchments, Environment International, 22, 553-540.

Bolster, C. H., and J. P. Raffensperger (1996), A finite difference code for solving Richards' equation, describing vertical one-dimensional flow in a partially saturated porous medium., edited.

Boyer, E. W., G. M. Hornberger, K. E. Bencala, and D. McKnight (1996), Overview of a simple model describing variation of dissolved organic carbon in an upland catchment, Ecological Modelling 86, 183-188

Buffam, I., H. Laudon, J. Temnerud, C.-M. Mörth, and K. Bishop (2007), Landscape-scale variability of acidity and dissolved organic carbon during spring flood in a boreal stream network, Journal of Geophysical Research-Biogeosciences, 112,doi:10.1029/2006JG000218.

Chow, A. T., K. T. Kenneth, S. Gao, and R. A. Dahlgren (2006), Temperature, water content and wet-dry cycle effect on DOC production and carbon mineralization in agricultural peat soils, Soil Biology and Biochemistry, 38, Soil Biology and Biochemistry.

Cohen, S. D., and A. S. Hindmarsh (1996), CVODE, a Stiff/nonstiff ODE Solver in C, Computers in Physics, 10, 138-143.

Crow, S. E., and R. K. Wieder (2005), Sources of CO2 emission from a northern peatland: Root respiration, exudation, and decomposition, Ecology, 86, 1825-1834.

Dillon, P. J., and L. A. Molot (1997), Effect of landscape form on export of dissolved organic carbon, iron, and phosphorus from forested stream catchments, WaterResources Research, 33, 2591-2600.

Erlandsson, M., I. Buffam, J. Fölster, H. Laudon, J. Temnerud, G. A. Weyhenmeyer,and K. Bishop (in review), 35 years of synchrony in riverine organic matterconcentrations explained by variation in flow and sulfate, Global Change Biology.

Fraser, C. J. D., N. T. Roulet, and T. R. Moore (2001), Hydrology and dissolved organic carbon biogeochemistry in an ombrotrophic bog, Hydrological Processes,15, 3151-3166.

Freeman, C., C. D. Evans, and D. T. Monteith (2001a), Export of organic carbon frompeat soils, Nature, 412, 785.

Freeman, C., N. Ostle, and H. Kang (2001b), An enzymic 'latch' on a global carbon

25

store, Nature, 409, 149. Futter, M. N., D. Butterfield, B. J. Cosby, P. J. Dillon, A. J. Wade, and P. G.

Whitehead (2007), Modeling the mechanisms that control in-stream dissolved organic carbon dynamics in upland and forested catchments, Water Resources Research, 43, W02424.

Gadmar, T. C., R. D. Vogt, and B. Østerhus (2002), The merits of the high temperature combustion method for determining the amount of natural organic carbon in surface freshwater samples International Journal of Environmental Analytical Chemistry, 82 451-461.

Gergel, S. E., M. G. Turner, and T. K. Kratz (1999), Dissolved organic carbon as an indicator of the scale of watershed influence on lakes and rivers, EcologicalApplications, 9, 1377-1390.

Granberg , G., H. Grip, M. Ottosson-Lofvenius, I. Sundh, B. H. Svensson, and M. Nillsson (1999), A simple model for simulation of water content, soil frost, and soil temperatures in boreal mixed mires, Water resources research, 35, 3771-3782.

Grip, H., and K. H. Bishop (1990), Chemical dynamics of an acid stream rich in dissolved organics, in The Surface Water Acidification Programme, edited by B. J. Mason, pp. 75-84, London: Royal Society, London.

Hassouna, M., F. Theraular, and C. Massiani (2007), Direct estimation of nitrate, total and fractionated water extractable organic carbon (WEOC) in an agricultural soil using direct UV absorbance deconvolution, Talanta, 71, 861-867.

Hope, D., M. F. Billett, R. Milne, and T. A. W. Brown (1997), Export of organic carbon in British rivers, Hydrological processes, 11, 325-344.

Hornberger, G. M., J. P. Raffensperger, P. L. Wiberg, and K. N. Eshleman (1998), Elements of Physical Hydrology, The Johns Hopkins University Press.

Hruška, J., H. Laudon, S. Kohler, C. E. Jonhson, and K. Bishop (2001), Acid/base character of organic acids in a boreal stream during snowmelt., Water Resources Research 37, 1043-1056.

Ivanov, K. E. (1981), Water movement in Mirelands., Academic Press, London. Ivarsson, H., and M. Jansson (1994), Regional Variation Of Dissolved Organic-Matter

In Running Waters In Central Northern Sweden, Hydrobiologia 286, 37-51. Jonsson, A., G. Algesten, A.-K. Bergström, K. Bishop, S. Sobek, L. J. Tranvik, and

M. Jansson (in press), Integrating aquatic carbon fluxes in a boreal catchmentcarbon budget, Journal of Hydrology.

Karlstr m, H. (1995), Peat characteristics : based on multivariate data analysis of magnetic resonance spectroscopy data 56 pp, Umeå University, Umeå.

Laudon, H., S. Kohler, and I. Buffam (2004), Seasonal TOC export from seven boreal catchments in northern Sweden, Aquatic Sciences, 66, 223-230.

Laudon, H., S. Köhler, and K. Bishop (1999), Natural Acidity or Anthropogenic Acidification in Spring Flood of Northern Sweden, Science of the Total Environment 234, 63-73.

Laudon, H., V. Sjöblom, I. Buffam, J. Seibert, and M. Mörth (2007), The role of catchment scale and landscape characteristics for runoff generation of boreal streams, Journal of Hydrology, in press.

Laudon, H., L. A. Vøllestad, A. B. S. Poléo, and K. Bishop (2005), Aluminumtoxicity to brown trout during spring flood in DOC-rich streams of northern Sweden: The effect of present and modeled pre-industrial water quality, Environmental Pollution, 135, 121-130.

Letts, M. G., N. T. Roulet , and N. T. Comer (2000), Parameterization of peatland hydraulic properties for the Canadian Land Surface Scheme, Atmosphere-Ocean,

26

38 141-160.McKay, M. D., W. J. Conover, and R. J. Beckman (1979), A comparison of three

methods for selecting values of input variables in the analysis of output from a computer code

Technometrics, 21, 239-245. McKnight, D., E. M. Thurman, R. L. Wershaw, and H. Hemond (1985),

Biogeochemistry of Aquatic Humic Substances in Thoreau's Bog, Concord, Massachusetts, Ecology, 66, 1339-1352.

Michalzik, B., E. Tipping, J. Mulder, J. F. G. Lancho, E. Matzner, C. L. Bryant, N. Clarke, et al. (2003), Modelling the production and transport of dissolved organic carbon in forest soils, Biogeochemistry, 66, 241-264.

Moore, T. R., and M. Dalva (1997), Methane and carbon dioxide exchange potentials of peat soils in aerobic and anaerobic laboratory incubations, Soil Biology and Biochemistry, 29, 1157-1164.

Moore, T. R., and M. Dalva (2001), Some controls on the release of dissolved organic carbon by plant tissues and soils Soil Science, 166, 38-47.

Neff, J. C., and G. P. Asner (2001), Dissolved organic carbon in terrestrial ecosystems: Synthesis and a model, Ecosystems, 4, 29-48.

Nungesser, M. K. (2003), Modelling microtopography in boreal peatlands: hummocksand hollows, 165, 175-207.

Ottosson-Löfvenius, M. O., M. Kluge, and T. Lundmark (2003), Snow and soil frost depth in two types of shelterwood and a clear-cut area. , Scand. J. Forest Res., 18,54-63.

Ours, D. P., D. I. Siegel, and P. H. Glaser (1997), Chemical dilarion and the dual porosity of humified bog peat, Journal of Hydrology, 196, 348-360.

Petrone, K., I. Buffam, and H. Laudon (2007), Hydrologic and biotic control of nitrogen export during snowmelt: a combined conservative and reactive tracer approach. , Water Resour. Res., in press.

Qualls, R. G., and C. J. Richardson (2003), Factors controlling concentration, export, and decomposition of dissolved organic nutrients in the Everglades of Florida, Biogeochemistry, 62, 197-229.

Rasse, D. P., H. L. Stap, and J. Mulder (in revision), Sorption and diffusion kinetics of DOC in acid forest soils., submitted.

Reeve, A. S., D. I. Siegel, and P. H. Glaser (2000), Simulating vertical flow in large peatlands, Journal of Hydrology, 227, 207-217.

Reeve, A. S., D. I. Siegel, and P. H. Glaser (2001), Simulating dispersive mixing in large peatlands, Journal of Hydrology, 242, 103-114.

Romanov, V. V. (1961), Mire hydrophysics (in Russian), Leningrad. Roulet, N. T., P. M. Lafleur, P. J. H. Richard, T. R. Moore, E. R. Humphreys, and J.

Bubier (2007), Contemporary carbon balance and late Holocene carbon accumulation in a northern peatland, Global Change Biology, 13, 397-411.

Sagerfors, J. (2007), Land-Atmosphere exchange of CO2, water and energy at a boreal minerotrophic mire. Doctoral Dissertation. Acta Universitatis agriculturae Sueciae vol. 2007:4., Umeå.

Saltelli, A. (Ed.) (2000), Sensitivity analysis, 475 pp., Wiley, Chichester. Sirin, A., S. Kohler, and K. Bishop (1998), Resolving flow pathways and

geochemistry in a headwater forested wetland with multiple tracers, Hydrology,Water Resources and Ecology in Headwaters.

Steinberg, C. E. W., S. Kamara, V. Y. Prokhotskaya, L. Manusadrianas, T. A. Karasyova, M. A. Timofeyev, Z. Jie, et al. (2006), Dissolved humic substances-

27

ecological driving forces from the individual to the ecosystem level?, FreshwaterBiology, 51, 1189-1210.

Tranvik, L. J., and M. Jansson (2002), Terrestrial export of organic carbon, Nature,415, 861-862.

Urban, N. R., S. E. Bayley, and S. J. Eisenreich (1988), Export of dissolved organic carbon and acidity from peatlands, Water Resource Research, 25, 1619-1628.

van Genuchten, M. T. (1980), Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils. , Soil Science Society of America Journal, 44892-898.

Wania, R., I. Ross, and I. C. Prentice (in preparation), Integrating peatlands and permafrost in a dynamic global vegetation model: Evaluation, sensitivity and application

Weiss, R., J. Alm, R. Laiho, and J. Laine (1998), Modeling Moisture Retention in Peat Soils, Soil Science Society of America Journal, 62, 305-313.

Willmott, C. J. (1982), Some Comments on the Evaluation of Model Performance,Bulletin American Meteorological Society, 63, 1309-1313.

Virta, J. (1966), Measurement of evapotranspiration and computation of water budget in treeless peatlands in the natural state, Commentationes physico-mathematicae,32.

Worral, F., T. Burt, and J. Adamson (2004), Can climate change explain increase in DOC flux from upland peat catchments?, Science of the Total Environment, 326,95-112.

Worrall, F., T. Burt, and J. Adamson (2006), Long-term changes in hydrological pathways in an upland peat catchment-recovery from severe drought?, Journal of Hydrology, 321, 5-20.

Yurova, A., A. Wolf, J. Sagerfors, and M. Nilsson (2007), Variations in net ecosystemexchange of the carbon dioxide in a boreal mire: Modeling mechanisms linked to water table position, Journal of Geophysical Research-Biogeosciences, 112, doi: 10.1029/2006JG000342.

28

Figure captions:

Figure 1. (a) Map of Sweden with study site location. (b) Upper portion of the

Nyänget drainage basin at Svartberget near Vindeln, Västerbotten in Northern

Sweden, showing the location of the Kallkällsmyren mire and the measurement site at

the stream origin. (c) Vegetation map of the Kallkällsmyren mire.

Figure 2. Schematic diagram of the main model components

Figure 3. Simulated (line) vs. measured (symbols) DOC concentrations at the outlet of

Kallkällsmyren mire. ( ) DOC measurements corresponding to stream discharge

values less than 0.3 mm, ( ) DOC measurements corresponding to stream discharge

value greater than 0.3 mm. The gray area indicates the upper and lower 10%

prediction bounds obtained from 1000 Monte-Carlo runs processed using the GLUE

methodology. Panel above: measured stream discharge plotted with y axis inverted

(i.e. with peak flows pointing downwards).

Figure 4. Simulated and measured DOC concentrations for two episodes: (a) 12 July-

30 August, 1993, and (b) 27 August-28 September, 2001. Upper panels illustrate the

measured total stream discharges plotted with y axis inverted (i.e. with peak flows

pointing downwards).

Figure 5. Modeled estimates of partitioning between the processes contributing to

changes in pore water DOC concentrations during the event from 27 August to 28

29

September 2001 in each of the six vertical peat layers of the acrotelm. Layer 1 is the

uppermost layer.

Figure 6. Simulated components of the annual DOC-SSOC balance: microbially

mediated net production and DOC stream export. Changes from year to year in the

amounts of SSOC stored in the acrotelm are also shown (estimated for the dates

before the first autumn frost).

Figure 7. (a) Interannual variations in the amounts of SSOC and DOC stored before

the start of the hydrological year (spring flood) simulated by the model. Interannual

variation in the volume-weighted annual stream DOC concentration, simulated by the

model and calculated from the interpolated observations. (b) Correlation between the

amounts of SSOC and DOC stored in the preceding season (winter and spring,

respectively), and the volume-weighted seasonal stream DOC concentrations in spring

and summer simulated by the model.

30

Table 1. Model parameters and constants Function Symbol Value Units Description Source

Hydrology 0.045 cm-1 lumped discharge parameter,eq. 1

Calibrated

0.20 cm-1 lumped transmitivity parameter Calibrated from the datarange in Ivanov [1981]

s 0.11 cm-0.67h-1 lumped surface discharge parameter, eq.2

Calibrated

lE 0.03 - lumped evapotranspirationparameter

Calibrated from the datarange in Virta [1966],Romanov [1961]

0.980.92;

- porosity: acrotelm; catotelm By analogy, e.g.Granberg et al [1999]

zcat -30 cm acrotelm depth By analogy, as average

sph 90 % proportion of Sphagnum remainsin the peat

Measured

ca 5 % proportion of Carex remains inthe peat

Measured

0.02;0.06

g cm-3 peat density, acrotelm; catotelm By analogy, e.g. Nungesser [2003]

DOCdynamics

KD 0.033 Lg-1 sorption partitioning constant Part I, Table 2

des 0.078 h-1 desorption kinetic constant Rasse et al., in revision:surface peat from the mirein Norway

Pbasal 1.4 10-3 mgg-1h-1 microbial DOC production ratein the acrotelm at 20oC

Part I, average Table 2

fPs 1.8 - correction for the DOC production by vegetation

Moore and Dalva [2001],comp. Sphagnum andfibric peat

1basal 0.4 10-4 h-1 DOC and SSOC mineralizationrate in the acrotelm at 20oC

Part I, water incubation Table 2

ks 1/6 - The constant reducing themicrobial mineralization ratewhen the soluble OC is sorbed

Kalbitz et al. [2005]:podzol organic horizon

Q10 1.7 - Modifier to account for theeffect of temperature onP and 1

Moore and Dalva [2001]:fibric and sapric peat

kanP 1 - ratio of anaerobic to aerobicDOC production

Calibrated from the ragein Moore and Dalva[1997]

kan 0.07 - ratio of anaerobic to aerobicDOC mineralization

Calibrated from the ragein Moore and Dalva[1997]

D0 4.3 10-2 cm2h-1 molecular diffusion coefficient Karlstr m [1995]

10 cm dispersivity Reeve et al. [2001]

Ds 0.19 cm2h-1 surface dispersion coefficient Calibrated

31

Table 2. Range of parameters and constants used in the sensitivity analysis.

Parameter range* Parameter RPCC

min max

Source and comments

Pbasal 0.84 0.4 10-3 2.4 10-3 Table 2 Part I

Q10 -0.78 1.6 4 Chow et al. [2006], Moore and Dalva[1997]

kanP 0.64 0.07 1 Bergman [1998};Moore and Dalva[1997]

basal -0.54 0.4 10-4 1.0 10-4 Table 2 Part I

des 0.53 -50% 50% -

KD -0.51 0.019 0.091 Qualls [2000], Rasse et al., in revision

kan -0.33 0.07 1 Moore and Dalva[1997]

0.21 12 34 From optimization

ks -0.19 1/6 1/3 Kalbitz et al. [2005]:

lE 0.13 0.01 0.03 From optimization

0.09 0.09 0.6 From optimization

-0.07 0.0113 0.0693 From optimization

D0 -0.06 1.2 10-2 7.2 10-2 Karlstr m [1995]

Ds 0.05 0.01 1 -

-0.04 0.2 100 Reeve et al. [2001],Ours et al. [1997]

fPs 0.01 1 3 -

* units as in the Table 1

32

Table 3. Quantitative measures of model performance with respect to DOC

concentrations. MBE, mean bias error; RMSE, root mean square error; d, Willmott

index of agreement (d=0, no agreement; d=1, perfect match) calculated for the whole

simulation period and for particular discharge categories.

Daily discharge interval (N=657), mm

MBE RMSE d

0-0.15 12.8 16.3 -0.65

0.16-0.3 3.8 11.9 0.32

0.3-0.49 2.5 7.4 0.70

0.5-1.4 1.1 8.9 0.74

1.5-23 -0.5 7.2 0.89

All 3.4 10.6 0.68

33

Figure 1. (a) Map of Sweden with study site location. (b) Upper portion of the Nyänget drainage basin at Svartberget near Vindeln, Västerbotten in NorthernSweden, showing the location of the Kallkällsmyren mire and the measurementsite at the stream origin. (c) Vegetation map of the Kallkällsmyren mire.

21 ºE69 ºN

13 ºE55 ºN

a

1 2 3 4

5 6 7 8

100m

Vegetation types:1. Pine (H 7-8m), shrub, sphagnum2. Pine (H 3-4m), shrub, cotton grass, sphagnum3. Cotton grass, sphagnum, some pines4. Pine and spruce (H 4-5m), sedge in the hollows, shrub and sphagnum in the hummocks5. Cotton grass and sedge6. Sedge, swampy vegetation7. Hollows with open peat and water8. Cane (reed)-sedge, swampy vegetation

c

Catchment boundary

StreamStream sampling site

Topo lines

Peat mire

Lake

100m

b

300 290

280

34

...........................

Snow dynamics

Soil temperature

Freezing andthawing

.............................

Evapotranspiration

Runoff in saturatedzoneand overland flow

Water content inunsaturated zone

................................

Convection anddispersion

Sorption/desorption

Metabolic productionand degradation

Mixing in streamwater

Heat transfer Hydrology DOC mass balance

air temperature, precipitation, potential evapotranspiration,peat type, composition and physical properties, acrotelm depth

runoff,DOC concentration in peat pore and stream water,

DOC flux out of the mire

Inputs:

Outputs:

Figure 2. Schematic diagram of the main model components

35

20

10

0 di

scha

rge,

mm

Apr93 Apr94 Apr95 Apr96 Apr97 Apr98 Apr99 Apr00 Apr010

10

20

30

40

DO

C c

once

ntra

tion,

mgL

-1

lower 10% upper 10% obs(q>0.3mm) obs(q<0.3mm) stand sim

Epi

sode

Epi

sode

Figure 3. Simulated (line) vs. measured (symbols) DOC concentrations at the outlet of Kallkällsmyren mire. (x) DOC measurements corresponding to stream discharge values lessthan 0.3 mm, ( ) DOC measurements corresponding to stream discharge value greater than 0.3 mm. The gray area indicates the upper and lower 10% prediction bounds obtained from 1000 Monte-Carlo runs processed using the GLUE methodology. Panel above: measured stream discharge plotted with y axis inverted (i.e. with peak flows pointing downwards).

36

0102030

q, m

m

07/12 07/26 08/09 08/23

20

40

60[D

OC

], m

gL−1

0102030

q, m

m

08/28 09/04 09/11 09/18 09/25

20

40

60

[DO

C],

mgL

−1

modelled DOC obs. DOC

a: 12July−30August 1993

b: 27August−28September 2001

Figure 4. Simulated and measured DOC concentrations for two episodes: (a) 12 July-30 August, 1993, and (b) 27 August-28 September, 2001. Upperpanels illustrate the measured total stream discharges plotted with y axis inverted (i.e. with peak flows pointing downwards).

37

layer1 layer2 layer3 layer4 layer5 layer6

com

pone

nts

of th

e D

OC

bal

ance

, gm

−2

−10

12

3

net productionsorptive releasestream exportadvective fluxdispersive flux

Figure 5. Modeled estimates of partitioning between the processes contributing to changes in pore water DOC concentrations during the event from 27 August to 28 September 2001in each of the six vertical peat layers of the acrotelm. Layer 1 is the uppermost layer.

38

1993 1994 1995 1996 1997 1998 1999 2000 20010

10

20

30

40

50

60

net production stream export SSOC&DOC stored

com

pone

nts

of t

he b

alan

ce, g

m-2ye

ar-1

;

stor

e gm

-2

Figure 6. Simulated components of the annual DOC-SSOC balance: microbiallymediated net production and DOC stream export. Changes from year to year in the amounts of SSOC stored in the acrotelm are also shown (estimated for thedates before the first autumn frost).

39

1993 1995 1997 1999 2001

20

40

60

80

SSO

C&

DO

C s

tore

d, g

m-2

D

OC

, mgL

-1

mod.SSOC&DOCstored before spr.floodmod. vol.weighted annual stream DOCobs. vol.weighted annual stream DOC

10

15

20

25

30

30 40 50 6030

40

50

60

mod. SSOC&DOC stored season before, gm-2

mod

. vol

. wei

ghte

d se

ason

al s

trea

m D

OC

, mgL

-1

a

b: spring

b:summer

r=0.40

r=0.87

Figure 7. (a) Interannual variations in the amounts of SSOC and DOC stored before the start of the hydrological year (spring flood) simulated by the model.Interannual variation in the volume-weighted annual stream DOC concentration, simulated by the model and calculated from the interpolated observations. (b) Correlation between the amounts of SSOC and DOC stored in the preceding season (winter and spring, respectively), and the volume-weighted seasonal stream DOC concentrations in spring and summer simulated by the model.

40

MEDDELANDEN FRÅN LUNDS UNIVERSITETS GEOGRAFISKA INSTITUTION

Serie Avnandlingar

I. Herman Richter: Skånes karta från mitten av 1500-talet till omkring 1700: bidrag till en historisk-kartografisk undersökning. (1929)

II. Josef Westin: Kulturgeografiska studier inom Nätra-, Näske- och Utbyåarnas flodområden samt angränsande kusttrakter. (1930)

III. Herman Richter och Wilhelm Norlind: Orbis Arctoi Nova et Accurata Delineatio Auctore Andrea Bureo Sueco 1626. (1936)

IV. Sven Björnsson: Sommen-Åsundenområdet : en geomorfologisk studie. (1937)

V. Arne Sandell: Tektonik och morfologi inom dalformationen med omgivande urbergsterräng. (1941)

VI. Sven Dahl: Torna och Bara : studier i Skånes bebyggelse- och näringsgeografi före 1860. (1942)

VII. Karl Erik Bergsten: Isälvsfält kring norra Vättern : fysisk-geografiska studier. (1943)

VIII. Carl Erik Nordenskjöld: Morfologiska studier inom övergångsområdet mellan Kalmarslätten och Tjust. (1944)

IX. Sven Björnsson: Blekinge : en studie av det blekingska kulturlandskapet. (1946)

X. Karl Erik Bergsten: Östergötlands bergslag : en geografisk studie. (1946) XI. Tor Holmquist: Den halländska vinterfiskehamnsfrågan. (1947) XII. Olof Ängeby: Landformerna i nordvästra Jämtland och angränsande delar

av Nord-Tröndelag. (1947) XIII. Axel Wennberg: Lantbebyggelsen i nordöstra Östergötland 1600-1875.

(1947)XIV. Lars Bjerning: Skånes jord- och stenindustri : dess utveckling,

lokalisering och betydelse ur näringsgeografisk synvinkel. (1947) XV. Allan Weinhagen: Norbergs bergslag samt Gunnilbo och Ramnäs till

omkring 1820 : studier i områdets närings- och bebyggelsegeografi. (1947)

XVI. Helge Stålberg: Smålands skogs- och träförädlingsindustrier : en närings-geografisk studie. (1947)

XVII. Folke Lägnert: Veteodlingen i södra och mellersta Sverige. (1949) XVIII. Yngve Nilsson: Bygd och näringsliv i norra Värmland : en

kulturgeografisk studie. (1950) XIX. Olof Ängeby: Evorsionen i recenta vattenfall. (1951)XX. Karl Erik Bergsten: Sydsvenska födelseortsfält. (1951) XXI. Folke Lägnert: Valmanskåren på skånes landsbygd 1911-1948. (1952) XXII. Olof Nordström: Relationer mellan bruk och omland i östra Småland

1750-1900. (1952)XXIII. Arvid Bergdahl: Israndsbildningar i östra Syd- och Mellansverige med

särskild hänsyn till åsarna. (1953)XXIV. Sven E Behrens: Morfometriska, morfogenetiska och tektoniska studier

av de nordvästskånska urbergsåsarna, särskilt Kullaberg. (1953)XXV. Torsten Hägerstrand: Innovationsförloppet ur korologisk synpunkt.

(1953)

XXVI. Gunhild Weimarck: Studier över landskapets förändring inom Lönsboda, Örkeneds socken, nordöstra Skåne. (1953)

XXVII. Ingemar Larsson: Structure and landscape in western Blekinge, southeast Sweden. (1954)

XXVIII. Sven Godlund: Busstrafikens framväxt och funktion i de urbana influensfälten. (1954)

XXIX. Folke Lägnert: Syd- och mellansvenska växtföljder. Del I : de äldre bruknings-systemens upplösning under 1800-talet. (1955)

XXX. Olof Ängeby: Toppkonstans, erosionsytor och passdalar i Jämtland och Tröndelag. (1955)

XXXI. Gunnar Johnsson: Glacialmorfologiska studier i södra Sverige. (1956) XXXII. Folke Lägnert: Syd- och mellansvenska växtföljder. Del II : 1900-talet.

(1956)XXXIII. Olof Nordström: Befolkningsutveckling och arbetskraftsproblem i östra

Småland 1800-1955. (1957) XXXIV. Sven Godlund: Befolkning - regionsjukhus - resmöjligheter - regioner.

(1958)XXXV. Elis Pålsson: Gymnasiers rekrytering och lokalisering. (1958) XXXVI. Harald Svensson: Glaciation och morfologi : en glacialgeografisk studie i

ett tvärsnitt genom Skanderna mellan södra Helgelandskusten och Kultsjödalen. (1959)

XXXVII. Erik Ljungner: Nahuel Huapi : ein geographischer Querschnitt durch die Anden in Patagonien. (1959)

XXXVIII. Nils Lewan: Om pendling mellan bostad och arbetsplats : en undersökning med material från sydvästra Skåne. (1960)

XXXIX. Åke Mattsson: Morphologische Studien in Südschweden und auf Bornholm über die nichtglaziale Formenwelt der Felsenskulptur. (1962)

XL. Stig Nordbeck: Framställning av kartor med hjälp av siffermaskiner. (1964)

XLI. Olof Nordström: Svensk glasindustri 1550-1960. (1962) XLII. Jan Davidsson: Littoral processes and morphology on scandian flatcoasts.

(1963)XLIII. Martin Markgren: Detaljmorfologiska studier i fast berg och

blockmaterial : geomorfologisk studie inom Fennoskandia med Skåne. (1962-1963)

XLIV. Martin Markgren: Geomorphological studies in Fennoscandia. II : chute slopes in northern Fennoscandia. A : regional studies. (1964)

XLV. Martin Markgren: Geomorphological studies in Fennoscandia. II : chute slopes in northern Fennoscandia. B : systematic studies. (1964)

XLVI. Lennart Améen: Stadsbebyggelse och domänstruktur. (1964) XLVII. Arvid Bergdahl: Det glaciala landskapet. (1961) XLVIII. Olof Nordström - Solveig Mårtensson: Turism på Öland (1966) XLIX. Jan O Mattsson: The temperature climate of potato crops. (1966) L. Nils Lewan: Landsbebyggelse i förvandling. (1967) LI. Gösta Nordholm: Skånes äldre ekonomiska geografi. (1967) LII. Sven Godlund - Torsten Hägerstrand - Bengt Svanström: Samhälls-

utvecklingen och samhällsplaneringen. (1967) LIII. Tor Fr Rasmussen: Storbyutvikling og arbeidsreiser. (1966) LIV. Erik Fagerlund - Harald Svensson - Sven Lindqvist m fl:

Infrarödtermografi : principer och naturgeografiska tillämpningar. (1967) LV. Lars Eldblom: Structure foncière, organisation et structure sociale. (1968)

LVI. Knut Norborg: Jordbruksbefolkningen i Sverige. (1968) LVII. Gunhild Weimarck: Ulfshult (1968) LVIII. Rune Järvenstedt - Sven Lindqvist - Jan O Mattsson m fl:

Televisionssystem i naturgeografisk forskning. (1968) LIX. Arne Jakobsson: Omflyttningen i Sverige 1950-1960. (1969) LX. Åke Hillefors: Västsveriges glaciala historia och morfologi. (1969) LXI. Sven Lindqvist: Bebyggelseklimatiska studier. (1970) LXII. Torsten Hägerstrand, Gunnar Törnqvist m fl: Urbaniseringen i Sverige.

(SOU 1970:14) LXIII. Bengt Sahlberg: Interregionala kontaktmönster : personkontakter inom

svenskt näringsliv – en flygpassagerarstudie. (1970) LXIV. Björn Hedberg: Kontaktsystem inom svenskt näringsliv : en studie av

organisationers externa personkontakter. (1970) LXV. Mats G Engström: Regional arbetsfördelning : nya drag i förvärvsarbetets

geografiska organisation i Sverige. (1970) LXVI. Torsten Persson: Geomorphological studies in the south-Swedish

highlands. (1972) LXVII. Dewitt Davis Jr: A factoral ecology of Malmö 1965 : the social

geography of a city. (1972) LXVIII. Zoltan Petery: Studier i svensk regionplanering : regionplanering enligt

byggnadslagen i mindre regioner. (1972) LXIX. Tommy Book: Stadsplan och järnväg i Norden. (1974) LXX. Hans Abrahamson: Studier i jordbrukets omstrukturering. (1974) LXXI. Christer Persson: Kontaktarbete och framtida lokaliseringsförändringar.

(1974)LXXII. Ulf Helldén: Karst : en studie av Artfjällets karstområde samt jämförande

korrosionsanalyser från Västspetsbergen och Tjeckoslovakien. (1974) LXXIII. Jànos Szegö: Befolkningstäthet, markanvändning, planering (vol 1 o 2).

(1974)LXXIV. Raul Nino Guerrero: Rural to urban drift of the unemployed in Colombia.

(1975)LXXV. Ulf Erlandsson: Företagsutveckling och utrymmesbehov. (1975) LXXVI. Sture Öberg: Methods of describing physical access to supply points.

(1976)LXXVII. Bo Lenntorp: Paths in space-time environments : a time-geographic study

of movement possibilities of individuals. (1976)LXXVIII. Richard Åhman: Palsar i Nordnorge : en studie av palsars morfologi,

utbredning och klimatiska förutsättningar i Finnmarks och Troms fylke. (1977)

LXXIX. Björn Gyllström: The organization of production as a space-modelling mechanism in underdeveloped countries. (1977)

LXXX. Anders Järnegren - Fosco Ventura: Tre samhällens förändringshistoria : exploateringen av den fysiska miljön i historisk belysning. (1977)

LXXXI. Tommy Book: Stadsplan och järnväg i Storbritannien och Tyskland. (1978)

LXXXII. Jan O Mattsson - Leif Börjesson: Lokalklimatiska temperaturstudier inom ett skånskt fruktodlingsdistrikt med särskilt beaktande av frostläntheten. (1978)

LXXXIII. Bjørn Terje Asheim: Regionale ulikheter i levekår. (1979) LXXXIV. Solveig Mårtensson: On the formation of biographies in space-time

environments. (1979)

LXXXV. Erik Wallin: Vardagslivets generativa grammatik - vid gränsen mellan natur och kultur. (1980)

LXXXVI. Reinhold Castensson: Välja för framtid - om markanvändningsval och förtroendemannainflytande i kommunal planering. (1980)

LXXXVII. Kerstin Montal: Industri och vatten : den vattenförorenande industrins lokaliseringsproblem i Malmöhus län. (1980)

LXXXVIII. Tommy Carlstein: Time resources, society and eecology : on the capacity for human interaction in space and time in preindustrial societies. (1980)

LXXXIX. Jonas Åkerman: Studies on periglacial geomorphology in west Spitsbergen. (1980)

XC. Leif Engh: Karstområdet vid Lummelunds bruk, Gotland, med speciell hänsyn till Lummelundagrottan. (1980)

XCI. Karna Lidmar-Bergström: Pre-quaternary geomorphological evolution in southern Fennoscandia. (1982)

XCII. Lars-Olof Olander: Staten, kommunerna och servicen : tiden kring kommun-reformen i ett ekonomiskt – geografiskt perspektiv. (1984)

XCIII. Bo Malmström och Owe Palmér: Glacial och periglacial geomorfologi på Varangerhalvön, Nordnorge : geomorfologisk kartering med analys av glaciala former och blockhav. (1984)

XCIV. Franz-Michael Rundquist: Hybrid maize diffusion in Kenya : policies, diffusion patterns and consequences. (1984)

XCV. Girma Yadeta: Dynamic processes of development in marginal areas : a case study from the pokot of north west Kenya. (1985)

XCVI. Anders Sporrek: Food marketing and urban growth in Dar Es Salaam. (1985)

XCVII. Rolf Nyberg: Debris flows and slush avalanches in northern Swedish Lappland distribution and geomorphological significance. (1985)

XCVIII. Lennart Olsson: An integrated study of desertification - applications of remote sensing, GIS and spatial models in semi-arid Sudan. (1985)

XCIX. Mikael Stern: Census from heaven? : population estimates with remote sensing techniques. (1985)

C. Katarina Olsson: Remote sensing for fuelwood resources and land degradation studies in Kordofan, the Sudan. (1985)

CI. Göran Loman: The climate of a sugar beet stand - dynamics, impact on the crop and possibilities of improvement. (1986)

CII. Eric Clark: The rent gap and urban change : case studies in Malmö 1860-1985. (1987)

CIII. Karin Hall-Könyves: Remote sensing of cultivated lands in the south of Sweden. (1988)

CIV. Eva Ahlcrona: The impact of climate and man on land transformation in central Sudan : applications of remote sensing. (1988)

CV. Kerstin Cederlund: Offentlig verksamhet : sysselsättning territoriellt och funktionellt. (1988)

CVI. Per Olof Hallin: Tid för omställning : om hushålls anpassningsstrategier vid en förändrad energisituation. (1989)

CVII. Jens Möller: Godsen och den agrara revolutionen : arbetsorganisation, domän-struktur och kulturlandskap på skånska gods under 1800-talet. (1989)

CVIII. Juha Uitto: The Kenyan conundrum : a regional analysis of population growth and primary education in Kenya. (1989)

CIX. Ola Jonsson: Informationsteknologi och arbete : fallstudier inom svensk sjukvård. (1989)

CX. Tora Friberg: Kvinnors vardag. Kvinnors arbete och liv : anpassnings-strategier i tid och rum. (1990)

CXI. Tomas Nihlén: Eolian processes in southern Scandinavia and the Mediterra-nean area. (1990)

CXII. Anders Löfgren: Att flytta hemifrån : boendets roll i ungdomars vuxenblivande ur ett situationsanalytiskt perspektiv. (1990)

CXIII. Irma Guillén: Cuidad Guayana – en stad, två världar : en studie av ett regionalt utvecklingsprojekts lokala effekter. (1991)

CXIV. Hans Holmén: Building organizations for rural development : state and cooperatives in Egypt. (1991)

CXV. Petter Pilesjö: GIS and remote sensing for soil erosion studies in semi-arid environments : estimation of soil erosion parameters at different scales. (1992)

CXVI. Ann-Cathrine Åquist: Tidsgeografi i samspel med samhällsteori. (1992) CXVII. José da Cruz: Disaster and society : the 1985 Mexican earthquakes.

(1993)CXVIII. Tomas Germundsson: Landsbygdens egnahem : egnahemsrörelsen,

småbruket och landskapet i sydsvenskt perspektiv. (1993) CXIX. Ann-Katrin Bäcklund: JUST-IN-TIME : hur industriella

rationaliseringsstrate-gier formar arbetsdelning och kompetens. (1994) CXX. Jon Knudsen: Kulturspredning i et strukturelt perspektiv : eksemplifisert

ved politisk og religiøs endring under moderniseringen av det norske samfunn. (1994)

CXXI. Tone Haraldsen: Teknologi, økonomi og rom : en teoretisk analyse av relasjoner mellom industrielle og territorielle endringsprosesser. (1994)

CXXII. Peter Jönsson: Wind climate during the instumental period and recent wind erosion in southern Scandinavia. (1994)

CXXIII. Peter Schlyter. Palaeo-wind abrasion in southern Scandinavia : field and laboratory studies. (1995)

CXXIV. Jun Yamashita: Spatial interaction and spatial structure : a study of public facility location. (1995)

CXXV. Mats Riddersporre: Bymarker i backspegel : odlingslandskapet före kartornas tid. (1995)

CXXVI. Anders Schærström: Pathogenic paths? : a time geographical approach in medical geography. (1996)

CXXVII. Lars Eklundh: AVHRR NDVI for monitoring and mapping of vegetation and drought in east African environments. (1996)

CXXVIII. Magnus Jirström: In the wake of the green revolution : environmental and socio-economic consequences of intensive rice agriculture – the problem of weeds in Muda, Malaysia. (1996)

CXXIX. Stefan Anderberg: Flödesanalys i den hållbara utvecklingens tjänst : reflektioner kring en ”metabolism” – studie av Rhenområdets utveckling. (1996)

CXXX. Karl-Johan Lundquist: Företag, regioner och internationell konkurrens : om regionala resursers betydelse. (1996)

CXXXI. Badr-Eldin Taha Osman: GIS-hydrological modelling in aridlands : a geographical synthesis of surface waters for the African Red Sea region in the Sudan. (1996)

CXXXII. Marie Stenseke: Bonden och landskapet : ägares och brukares relationer till markerna och förutsättningarna för en uthållig markanvänding. (1997)

CXXXIII. Kristina Blennow: Spatial variation in near-ground radiation and low temperature – interactions with forest vegetation. (1997)

CXXXIV. Lennart Runesson: Tomträtt : ett markpolitiskt instrument i upplösning. (1997)

CXXXV. Johan Hultman: The eco-ghost in the machine : reflexions on space, place and time in environmental geography. (1998)

CXXXVI. Jonas Ardö: Remote sensing of forest decline in the Czech Republic. (1998)

CXXXVII. Per Hillbur: The knowledge arena : approaching agroforestry and competing knowledge systems – a challenge for agricultural extensionl (1998)

CXXXVIII. Tom Mels: Wild landscapes : the cultural nature of Swedish national parks. (1999)

CXXXIX. Carolyn Hannan-Andersson: Promoting equality between women and men in bilateral development cooperation : concepts, goals, rationales and institutional arrangements. (2000)

CXL. Nikolaus Solakius: The parnassus zone, central Greece. (2000)CXLI. Jonathan Seaquist: Mapping primary production for the west African

Sahel using satellite data. (2001)CXLII. Karin Book och Lena Eskilsson: Stadens struktur : varför och hur? (2001) CXLIII. Maria Wikhall: Universiteten och kompetenslandskapet : effekter av den

högre utbildningens tillväxt och regionala spridning i Sverige. (2001)CXLIV. Rannveig Olafsdottir: Land degradation and climate in Iceland : a spatial

and temporal assessment. (2002) CXLV. Marie Ekström: Relationships between atmospheric circulation and wind

erosion in southern Sweden and Australia. (2002) CXLVI. Maj-Lena Finnander Linderson: The spatial distribution of precipitation

in Scania, southern Sweden : observations, model simulations and statistical downscaling. (2002)

CXLVII. Richard Ek: Öresundsregion - bli till! : de geografiska visionernas diskursiva rytm. (2003)

CXLVIII. Olivia Louw: Exploring the culture of non-payment in the post-apartheid South Africa. (2003)

CXLIX. Cecilia Kjellman: Ta plats eller få plats? : studier av marginaliserade människors förändrade vardagsliv (2003)

CL. Christina Scholten: Kvinnors försörjningsrum : hegemonins förvaltare och murbräckor (2003)

CLI. Michael Runnström: Land degradation and mitigation in northern China : evaluated from the biological production (2003)

CLII. Sara Brogaard: Recent changes in land use and productivity in agro-pastoral Inner Mongolia, China (2003)

CLIII. Jan_Henrik Nilsson: Östersjöområdet : studier av interaktion och barriärer (2003)

CLIV. Thomas Hickler: Towards an integrated ecology through mechanistic modelling of ecosystem structure and functioning (2004)

CLV. Andreas Persson: Hydrological modelling, topographical influence and yield mapping in precision agriculture (2004)

CLVI. Maria Olsrud: Mechanisms of below-ground carbon cycling in subarctic ecosystems (2004)

CLVII. Sandra C. Fernández: Farewell to the peasantry? : (Post)modernising rural Mexico : the case of the ejido peasants in the Isthmus of Tehuantepec (2004)

CLVIII. Andrés P. Leskó: To own the phone : spatial diffusion, ownership and regulation of telephone service in Argentina, 1878-1990 (2004)

CLIX. Henrik Svensson: Öppna och slutna rum : enskiftet och de utsattas geografi : husmän, bönder och gods på den skånska landsbygden under 1800-talets första hälft (2005)

CLX. Pablo Morales: Modeling carbon and water fluxes in European terrestrial ecosystems (2006)

CLXI. Emmanuel S. Gritti: Global changes and European terrestrial ecosystems (2006)

CLXII. Ola Thufvesson: Kreativitetens yttre villkor : miljöer, rörlighet och nobelpristagare (2006)

CLXIII. Deniz Koca: Impacts of regional climate change on Swedish forests : an evaluation using process-based regional ecosystem modelling approach (2006)

CLXIV. Bodil Elmqvist: Livelihood diversification and land use change in the Sahel : an interdisciplinary analysis of gum arabic in Sudan (2006)

CLXV. Jan Vang-Lauridsen: Industrial dynamics and the spatial organization of industries (2006)

CLXVI. Heidi Wiig Aslesen: Innovation in an urban context (2006) CLXVII. Torbjörn Johansson: Temporal and spatial variability of carbon cycling in

a subarctic landscape (2006) CLXVIII. Anders Lund Hansen: Space wars and the new urban imperialism (2006) CLXIX. Lars Coenen: Faraway, so close! : the changing geographies of regional

innovation (2006) CLXX. Pontus Olofsson: Remote sensing of carbon balance across Scandinavian

forests (2007) CLXXI. Margareta Rämgård: The power of place : existential crises and place

security in the context of pregnancy (2006) CLXXII. Helena Eriksson: Leaf area index of Scandinavian forests : methods using

in situ and remotely sensed data (2007) CLXXIII. Ransom Lekunze: Corporate social responsibility and development : the

case of the Chad Cameroon oil pipeline project (2007) CLXXIV. Alla Yurova: Hydrological aspects of the carbon balance in a boreal

catchment : a model study (2007)

I