Effects of soil freezing and thawing on vegetation

carbon density in Siberia: A modeling analysis with

the Lund-Potsdam-Jena Dynamic Global Vegetation

Model (LPJ-DGVM)

C. Beer,1,2 W. Lucht,3 D. Gerten,3 K. Thonicke,4 and C. Schmullius1

Received 18 May 2006; revised 8 November 2006; accepted 16 November 2006; published 20 February 2007.

[1] The current latitudinal gradient in biomass suggests a climate-driven limitation ofbiomass in high latitudes. Understanding of the underlying processes, and quantificationof their relative importance, is required to assess the potential carbon uptake of thebiosphere in response to anticipated warming and related changes in tree growth andforest extent in these regions. We analyze the hydrological effects of thawing and freezingof soil on vegetation carbon density (VCD) in permafrost-dominated regions ofSiberia using a process-based biogeochemistry-biogeography model, the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM). The analysis is based onspatially explicit simulations of coupled daily thaw depth, site hydrology, vegetationdistribution, and carbon fluxes influencing VCD subject to climate, soil texture,and atmospheric CO2 concentration. LPJ represents the observed high spring peak ofrunoff of large Arctic rivers, and simulates a realistic fire return interval of 100 to200 years in Siberia. The simulated VCD changeover from taiga to tundra is comparableto inventory-based information. Without the consideration of freeze-thaw processesVCD would be overestimated by a factor of 2 in southern taiga to a factor of 5 in northernforest tundra, mainly because available soil water would be overestimated with majoreffects on fire occurrence and net primary productivity. This suggests that forest growth inhigh latitudes is not only limited by temperature, radiation, and nutrient availabilitybut also by the availability of liquid soil water.

Citation: Beer, C., W. Lucht, D. Gerten, K. Thonicke, and C. Schmullius (2007), Effects of soil freezing and thawing on vegetation

carbon density in Siberia: A modeling analysis with the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM),

Global Biogeochem. Cycles, 21, GB1012, doi:10.1029/2006GB002760.

1. Introduction

[2] High-latitude ecosystems currently experience variouschanges due to warming, for example, increasing netprimary production (NPP) [Lucht et al., 2002; Nemani etal., 2003] that is attended by northward migration of the treeline [Serreze et al., 2000; Esper and Schweingruber, 2004],but also increasing heterotrophic soil respiration [Oechel etal., 1993; Zimov et al., 1993]. Owing to rising temperature,a large fraction of the current extend of permafrost soils ofabout half of the territories of Canada and the former SovietUnion [Bonan and Shugart, 1989] is assumed to degradeuntil 2100 [Anisimov and Nelson, 1997; Stendel andChristensen, 2002]. This accompanied a risk of an addi-tional radiative forcing of the climate system by CO2

respired from carbon that is currently blocked in perma-nently frozen grounds [Goulden et al., 1998].[3] The ability of increasing carbon uptake by expanding

forests to partly counteract these carbon emissions is un-clear. To assess a potential carbon uptake by vegetation inthe high latitudes, an understanding of the environmentalfactors and processes limiting the current amount of bio-mass as well as their relative importance is required. Theobserved gradient in vegetation carbon density (VCD) from0.6 kg carbon per m2 (kgC/m2) in the Russian tundra to3.6 kgC/m2 in the southern taiga [Shvidenko et al., 2000]suggests a climatic driven limitation of biomass in northerntaiga and tundra. The relationship of biomass to the depth ofthe active layer [Bonan, 1989], which is the uppermost partof the soil in permafrost areas that is thawing each summerand therefore provides plants with liquid water andnutrients, supports this assumption. Temperature, solarradiation and supply of nutrient and water are reported tolimit generally the establishment and survive of seedlings,carbon assimilation and plant growth [e.g., Bonan andShugart, 1989; Woodward, 1995; Korner, 2003]. However,Briffa et al. [1998] found that tree growth was less sensitive

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 21, GB1012, doi:10.1029/2006GB002760, 2007ClickHere

for

FullArticle

1Institute of Geography, Friedrich Schiller University, Jena, Germany.2Now at Max Planck Institute for Biogeochemistry, Jena, Germany.3Potsdam Institute for Climate Impact Research, Potsdam, Germany.4School of Geographical Sciences, University of Bristol, Bristol, UK.

Copyright 2007 by the American Geophysical Union.0886-6236/07/2006GB002760$12.00

GB1012 1 of 14

to increasing temperature during the last decades, and thatthis could not be explained readily from soil moisture data.In addition, disturbances like fire and damage by insects orwind had the capability to counteract an increasing produc-tivity due to warming. Fire is the major disturbance agent inthe boreal zone, where between 3 and 23 million ha areburned annually mainly as crown fires in boreal NorthAmerica, but apart from extreme fire years as surface firesin boreal Eurasia [Kasischke et al., 2005], thereby releasingapproximately 106–209 TgC/a.[4] The processes controlling biomass take place in par-

allel or are even coupled, especially in regions with perma-nently frozen grounds. For instance, low temperaturesduring the growing season limit not only photosynthesisbut also active-layer thickness (ALT), which impacts wateravailability [Benninghoff, 1952]. Both effects of tempera-ture impact carbon assimilation by the vegetation. More-over, low temperatures lead to slow litter decomposition,hence nutrient limitation [Bonan and Shugart, 1989]. Thedeep litter layer also isolates the underlying soil leading to alow thaw depth [Hinzman et al., 1991], thus low wateravailability in summer again. In addition, litter moisturedetermines the probability of fire. Such linkages make itimpossible to separate the even most important effects ofenvironmental influences on the amount of biomass by fieldstudies alone, in particular because biomass is the result ofdifferent processes over decades. Experiments and observa-tions are critical to providing the inputs and functions formodels, but it would probably be impossible to set up anexperiment to answer the questions posed here.[5] In this paper, we investigate the effects of the alter-

ations in the soil water balance due to freeze-thaw processesin the soil on the amount of VCD in permafrost-dominatedregions of Siberia. The impact of soil water content which iscontrolled by phase change on the productivity and firereturn interval is studied. For this purpose, we perform twosimulation experiments with the LPJ-DGVM, in whichfreeze-thaw processes are excluded in the control run. Bothsimulations are driven by the same observations of climaticvariables and atmospheric CO2 content. The comparison ofVCD results by these simulation experiments with forestinventory estimates over a large transect swath in centralSiberia allows for the investigation of the working hypoth-esis that besides direct influences of temperature, solarradiation and nutrient availability on the productivity, alsolimitations in liquid soil water availability during the yearand the disturbance dynamics substantially contribute tolow VCD values in high-latitude permafrost regions, i.e.,that soil thermal dynamics are crucial for the changeover ofVCD from taiga to tundra. In order to isolate better theinfluence of fire disturbance and water stress, the results ofthese experiments are further compared to results of twoLPJ-P model runs excluding the simulation of fire and waterstress.

2. Methods

2.1. LPJ-DGVM Overview

[6] The LPJ-DGVM is a coupled dynamic biogeography-biogeochemistry model which combines process-based rep-

resentations of terrestrial vegetation dynamics and land-atmosphere carbon and water exchanges in a modularframework. For a detailed description and evaluation ofthe model see Sitch et al. [2003]. LPJ explicitly considerskey ecosystem processes such as photosynthesis, carbonallocation, mortality, resource competition, fire disturbanceand soil heterotrophic respiration. To account for the varietyof structure and functioning among plants, four plantfunctional types (PFTs) are distinguished for the borealzone (see auxiliary material1). Gross primary production iscalculated on the basis of a coupled photosynthesis-waterbalance scheme after Farquhar et al. [1980] with leaf-leveloptimized nitrogen allocation [Haxeltine and Prentice,1996]. The underlying model of light use efficiencyassumes no limitation of nitrogen uptake [Haxeltine andPrentice, 1996], thus no nitrogen cycle is represented in themodel. Fire disturbance is mainly driven by litter moistureand dead fuel load as well as PFT specific fire resistancesThonicke et al. [2001].[7] The LPJ-DGVM used in this study has been inten-

sively validated against observations of key ecosystemparameters on different scales including land-atmosphereCO2 exchange [McGuire et al., 2001; Sitch et al., 2003;Peylin et al., 2005], dominant vegetation distribution[Cramer et al., 2001; Sitch et al., 2003; Beer, 2005], fluxesof water [Gerten et al., 2004, 2005], global patterns of soilmoisture [Wagner et al., 2003], NPP enhancement due tocarbon fertilization [Gerten et al., 2005], and satellite-observed northern latitude leaf area index anomalies [Luchtet al., 2002].[8] In this study specific leaf area (SLA, in m2/kg) is

related to leaf longevity (aleaf, in years) after Reich et al.[1997] as follows:

SLA ¼ 0:22 � exp 5:6� 0:46 � ln aleaf � 12� �� �

: ð1Þ

Moreover, parameters of boreal vegetation are slightlyadjusted in this study (see auxiliary material). With theseadjustments, a more valid distribution of dominant woodyvegetation in Siberia is achieved [Beer, 2005].

2.2. Input Data and Modeling Protocol

[9] LPJ was driven by gridded monthly fields of airtemperature, precipitation, number of rainy days, and cloudcover at a 0.5� pixel size. This station-based data set wasprovided by the Climatic Research Unit (CRU) of theUniversity of East Anglia, UK [Mitchell and Jones, 2005]but revised and extended by the Potsdam Institute forClimate Impact Research (PIK), Germany [Oesterle et al.,2003]. As a nongridded global input, annual CO2 concen-trations were used that were derived from ice-core measure-ments and atmospheric observations, provided by theCarbon Cycle Model Linkage Project [McGuire et al.,2001]. In addition, eight soil texture classes (see auxiliarymaterial) were used which are also available globally on a0.5� grid [Food and Agriculture Organization, 1991]. LPJwas run from 1901 to 2003 with a pseudo-daily time step,preceded by a 1000-year spinup period that brought the

1Auxiliary material data sets are available at ftp://ftp.agu.org/apend/gb/2006gb002760. Other auxiliary material files are in the HTML.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

2 of 14

GB1012

carbon pools and vegetation cover in equilibrium withclimate.[10] Two model experiments were performed for the land

area of Northern Eurasia. In both, the same model version isused but (1) with (LPJ-P) and (2) without permafrost (LPJ-noP). In order to isolate the influence of fire disturbance andwater stress, the results of these experiments are furthercompared to results of LPJ-P model runs excluding fire(LPJ-noFire), and excluding water stress (LPJ-noStress).

2.3. Soil Water Balance

[11] Soil hydrology is modeled as described by Sitch et al.[2003] and Gerten et al. [2004] following a semi-empiricalapproach. There are two soil layers of constant depth(0.5 and 1 m) and spatially varying soil texture. Plantavailable soil moisture w is expressed as the fraction ofavailable water holding capacity Wplant, which is defined asthe difference between volumetric soil water content at fieldcapacity and permanent wilting point Wpwp. For calculatingsoil thermal properties, the whole water content is required,hence Wpwp is added to Wplant (see auxiliary material).[12] The water content of each layer is updated daily by

taking into account rainfall, snowmelt, canopy interceptionof precipitation, percolation, soil evaporation, transpirationand runoff. Monthly values of precipitation, P are distrib-uted stochastically to provide pseudo-daily values. Below athreshold value of 0�C precipitation is defined as snow.Snowmelt (M, mm/d) is calculated using an index equation[Choudhury and DiGirolamo, 1998]:

M ¼ 1:5þ 0:007 � Pð Þ � Ta; ð2Þ

where Ta is the air temperature above 0�C. Percolation rate( p, mm/d) through the soil layers depends on soil textureand layer thickness,

p ¼ kperc � w1; ð3Þ

where w1 is the plant available soil moisture of the upperlayer and kperc the texture-dependent conductivity, varyingfrom 5 mm/d for sandy soils to 0.2 mm/d for vertisols.Surface runoff and drainage are calculated as the excesswater above field capacity in both soil layers [Gerten et al.,2004]; that is, all water above field capacity that does notpercolate to the respective underlying layer will run off.[13] Under conditions of permafrost, the water balance is

altered as follows.[14] 1. Thaw depth within the active layer changes

dynamically for each simulation day. Wplant is adjusteddaily in accordance, which directly influences runoff gen-eration and water availability. For instance, if only a verythin part of the soil is already thawed in spring, the storagecapacity of the soil will be small hence most of snow meltor rainfall will run off. Lateral flow of water, which mayplay an important role in permafrost regions, is not consid-ered in the present model. This may be a shortcoming forsmall-scale investigations, but implies that lateral flows arecanceled out at the large scale considered here (0.5�), whichis not an unrealistic assumption. Lateral flow among gridcells, however, will be enabled in a forthcoming version ofthe LPJ model (Stefanie Jachner, unpublished data, 2006).

[15] 2. Water is only allowed to percolate through a soillayer if it is completely thawed. This will increase soilmoisture of the upper layer after precipitation if the transi-tion to the underlying layer remains frozen.[16] 3. In addition to w a new vector stores frozen water

for both layers also as a fraction of plant available soilwater. Thus precipitation in autumn is conserved until thenext spring.[17] Water supply to plants is the product of the plant

root-weighted soil moisture availability and a maximumtranspiration rate [Sitch et al., 2003]. The fractions of fineroots in the upper and lower soil layer, z1 and z2 areoptimized daily for boreal needleleaved summergreen veg-etation (BoNS) as a function of mm thawed water (w1,upper layer; w2, lower layer) to give BoNS the opportunityto access as much water from both layers as possible,

z1 ¼ w1 w1 þ w2ð Þ�1

z2 ¼ 1� z1;

while z1 and z2 are set to 0 if w1 + w2 = 0. This mechanismgives BoNS an advantage over other woody PFTs inpermafrost regions by allowing them to better managedroughts (e.g., in East Siberia). Larix gmelinii uses soilwater of the upper 15 cm under wet conditions, while rootsobserved at 30–70 cm depth are possibly essential for theuptake of thawed water during drought [Sugimoto et al.,2002].[18] Links between site hydrology and the carbon cycle

are simulated by the LPJ-DGVM on several timescales. Thefast processes, photosynthesis and transpiration are coupledby the dynamics of canopy conductance after Farquhar etal. [1980], which strongly depend on soil moisture [Gertenet al., 2005]. On a longer timescale, moisture-related adjust-ments in the allocation of carbon to different plant compo-nents [Bongarten and Teskey, 1987; Mencuccini and Grace,1995] lead to moisture-dependent leaf area [Grier andRunning, 1977; Gholz, 1982]. The ratio between atmo-spheric demand and supply of water (water stress, w) ismodeled to influence the carbon allocation between leavesand fine roots [Sitch et al., 2003],

cleaf ¼ w � croot: ð4Þ

2.4. Fire

[19] A full description of the LPJ fire module Glob-FIRMis given by Thonicke et al. [2001]. The probability p of afire increases with decreasing litter moisture content asdescribed by

p wð Þ ¼ exp �p w � w�1e

� �� �; ð5Þ

w is the PFT-weighted moisture of extinction, above whichlitter cannot ignite, because all the ignition energy isconsumed to vaporize the water it contains [see Sitch et al.,2003, Table 3]. The fire season s is calculated as the annualarithmetic mean of the daily fire probability. A constantnonlinear relationship between s and annual area burned A,assuming that the fractional area burned increases the longerthe burning conditions persist, is applied [Thonicke et al.,2001]. Fire effects are described by fire resistances, which

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

3 of 14

GB1012

combine fire intensity and fire severity in a PFT-specificparameter (see auxiliary material).

2.5. Freeze-Thaw Processes in the Active Layer OverPermanently Frozen Ground

[20] In a first step we determine whether a given geo-graphic location is characterized by permanently frozenground. Permafrost status is applied to the whole of a gridcell, neglecting discontinuous permafrost, since DGVMs donot resolve subgrid localizations of spatial heterogeneity(this is a tradeoff with their continental-scale applicability inthe absence of detailed spatial data). Surface temperaturebelow snow and litter (including near-surface carbon, e.g.,of grass) is calculated by treating snow and vegetationlayers separately. If a particular grid cell is determined tobe located in a permafrost region, thaw depth is computedby using this altered surface temperature. The simulation ofhydrological processes follows as described in section 2.3.2.5.1. Permafrost Distribution[21] Spatial permafrost distribution is modeled at annual

time steps using the normalized frost index F [Nelson andOutcalt, 1987],

F ¼ffiffiffiffiffiffiDf

p ffiffiffiffiffiDt

pþ

ffiffiffiffiffiffiDf

p� ��1

: ð6Þ

Its reliability can be assumed to also hold in a changingclimate [Anisimov and Nelson, 1997; Christensen andKuhry, 2000; Stendel and Christensen, 2002]. Df and Dt

represent the annual freezing and thawing degree-days (sumof pseudo-daily temperatures that are derived from inter-polated monthly data). In case of snow coverage thetemperature below snow is used for the calculation of Dt

and Df (see section 2.5.2). Thus vegetation has an impact onthe calculation of the frost index since canopy interceptionof snow is taken into account in LPJ [Gerten et al., 2004].Values of F that represent the threshold boundaries ofcontinuous, discontinuous and sporadic permafrost are 0.67,0.6 and 0.5, respectively [Nelson and Outcalt, 1987;Christensen and Kuhry, 2000]. In this study F = 0.6 isapplied to distinguish between permafrost and nonperma-frost situations.2.5.2. Daily Thaw Depth[22] In this work ‘thaw depth’ represents the soil depth z

of the 0�C isotherm. We assume only one isotherm insteadof two phase interfaces in autumn since the effects of theprecise discrimination of the location of the interfacesbetween the two soil layers on the carbon fluxes are smallwhile air temperature is below the freezing point. In otherwords, in autumn our variable ‘thaw depth’ corresponds tothe thawed water column length. During the growing seasonwhen air temperature is above 0�C simulated thaw depthmatches the definition used for the temperature-measuredvariable in the field.[23] Soil temperature in �C is assumed to follow an

annual sinusoidal cycle of air temperature with a dampedoscillation [Campbell and Norman, 2000],

T z; tð Þ ¼ Tave þ A � exp �z � d�1� �

sin W �Dt � z � d�1� �

; ð7Þ

where Tave is the mean annual temperature above therespective layer and A is the amplitude of this temperaturefluctuation. d =

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2k � W�1

prepresents the damping depth

with thermal diffusivity k = l � c�1, where c is volumetricheat capacity and l thermal conductivity. W = 2p � t�1

represents the angular frequency of oscillation withoscillation period t. The method of solving equation (7)for each simulation day is described in the appendix of[Sitch et al., 2003]. The concept of a running mean isapplied for calculating Tave; that is, the model is not requiredto have knowledge of future climate. Snow cover and anear-surface carbon layers (aboveground litter and grasscarbon density) are considered to be above the soil in thisstudy. The temperature above snow corresponds to the airtemperature. Equation 7 is applied to calculate the followingtemperatures for each simulation day step by step:temperature below snow, temperature below near-surfacecarbon, and temperature in 0.25 m soil depth. Each of theseresults is taken as input for calculating the subsequent value.Mean depth of snow and near-surface carbon layers of thelast 31 days as well as their mean thermal diffusivity areupdated daily (see auxiliary material).[24] Once temperature below snow or near-surface carbon

layers is estimated, soil temperature in 0.25 m depth andthaw depth can be calculated in parallel. In doing so, soilvolumetric heat capacity c and thermal conductivity l arerequired for each day, which are derived from meanvolumetric water content of the thawed fraction of the soilof the last 31 days, w, after Campbell and Norman [2000].The nonlinear relationship l = l(w) is approximated line-arly for w 2 [0..0.1), w 2 [0.1..0.25) and w 2 [0.25..1),respectively. While a phase change the latent heat Q isconsidered as apparent heat capacity [Jumikis, 1966] in theexponential damping term of equation (7),

Q ¼ k � wthaw � Qw: ð8Þ

wthaw represents the amount of thawed soil water contentaveraged over the last 31 days and Qw = 334.94 kJ/m3 is theheat of fusion of water. We approximate k to 0.5 byassuming a nearly linear change of the 0�C isotherm duringthe short interval of a 1-day time step.[25] Figure 1 summarizes the scheme of the thaw depth

module and visualizes feedbacks of vegetation on thawdepth through soil moisture and insulating snow and near-surface carbon layers. Thaw depth is finally computed byfinding the null of equation (7) numerically via Newton’smethod for each simulation day,

znþ1 ¼ zn þ T zn; tð Þ T 0 zn; tð Þð Þ�1: ð9Þ

If the calculated thaw depth indicates a thawing or freezingevent, the associated soil water will be shifted between theseparate pools for water and ice for each layer, and a newvalue for soil moisture is derived by calculating the sitehydrology. In doing so, soil thermal properties of the nextsimulation day are influenced.

2.6. Freeze-Thaw Processes Outside the PermafrostZone

[26] Outside of the permafrost zone (determined asdescribed in section 2.5.1) the soil freezes during winter

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

4 of 14

GB1012

to a certain depth and thaws in spring at two phase interfacesuntil the whole soil column is thawed. In contrast to the thawdepth over permafrost soils during summer, the winter frostdepth has less of an importance for the vegetation function-ing in the boreal zone. Trees are hardly able to make use ofwater from deeper soils while the xylem is frozen in winter[Woodward, 1995]. In addition, photosynthesis can beneglected while temperature is clearly below the freezingpoint. Thus we do not attempt to estimate frost depth duringwinter but consider the whole of the soil column to be frozenwhen the temperature (as calculated by equation (7)) fallsbelow 0�C in 10 cm depth, following Zhuang et al. [2003].In doing so, we make sure that snow melt water in springdoes not infiltrate immediately into the soil, that is, the waterpool available to vegetation.

2.7. Study Region for Evaluation of VegetationCarbon Density

[27] Simulated VCD values are compared to a recentlyupdated inventory-based VCD data from an ecosysteminformation system on a 1:1 million polygon basis, thatwas made available by the International Institute for AppliedSystems Analysis (IIASA), Laxenburg, Austria. It covers a3 million km2 broad transect swath through central Siberiareaching from the Arctic Ocean to the southern steppe andthe Yenisey river in the West to Lake Baikal in the east(approx. 79�E–119�E, 51�N–78�N). This ecosystem database [Nilsson et al., 2005] was created as part of theEuropean Commission’s SIBERIA-II project [Schmulliusand Hese, 2002].

3. Results

3.1. Permafrost Model Evaluation

[28] For site-specific evaluation, LPJ was run with the soiltextures of the observation points. Simulated daily thawdepth is compared to seasonal thaw depth at four stations inAlaska from 1995 until 1997 [Nelson, 1996] and one site inthe Siberian Far East from 1981 until 1985 (S. Zimov, Thaw

depth measurements near River Kolyma 1981–1985, per-sonal communication, 2005). Thaw depth simulated by LPJfollows the seasonal cycle of measured thaw depth valuesfairly well (Figure 2). Discrepancies between modeled andobserved values are within the range of standard deviationsof observations. In many cases the model slightly under-estimates the maximum thaw depth. However, this isobviously the case only for the selection of locations shownhere for which seasonal thaw depth are available. Otherlocations show as many cases of overestimation as under-estimation of maximal thaw depth, as can be seen fromFigure 3. It compares LPJ-simulated ALT with 178 obser-vations collected by the Circumpolar Active Layer Moni-toring (CALM) project [Brown et al., 2003] from 29 stationsin Alaska, Canada and Russia between 1992 and 2002.Most simulation results range between 50 and 200% dis-crepancy to the observations, the coefficient of determina-tion and root mean square error are 0.3 and 21 cm,respectively. These discrepancies are of the same magnitudeas the natural variability of ALT over only tens of meters[Nelson et al., 1997]. Thus results by LPJ that were drivenby coarse climate and soil data are considered reasonable,particularly as this study is focused not on the details ofpermafrost dynamics but on general effects of permafrost onthe vegetation growing on permafrost soils.[29] In addition to the seasonal cycle of thaw depth and its

maximum, Figure 4 provides an evaluation of the interan-nual variability of ALT at two stations in Alaska and two inRussia. As can be seen, changes in simulated ALT fromyear to year correspond well to observations. This demon-strates the ability of the algorithm to simulate the impacts ofair temperature, soil moisture and insulations on soil tem-perature, which act on a daily basis or even finer temporalresolution but are here recovered from a coarser resolutionmodel.[30] Spatial details of simulated ALT within permafrost

areas are shown in the auxiliary material. Continental-scaleobservations of ALT are not available, hence spatial patternof ALT are compared to results from a state-of-the-artfreeze-thaw model, the Goodrich model [Oelke et al.,2003]. Although, in contrast to LPJ, this model is drivenby NCEP/NCAR reanalysis climate data with a topography-based correction of surface air temperature and a radar-based information on snow water equivalent, both modelresults match in their general pattern (see auxiliary material).They agree in the increase of ALT from north to south aswell as the decrease from west to east in Northern Eurasiathat is a consequence of a more continental climate inEastern Siberia. LPJ, however, shows a stronger spatialgradient in the northern taiga. ALT mostly ranges from25 cm in the north or east up to 100 cm in the south orwest. In a few regions in Western Siberia ALTwas computedto reach 250 cm. A band of 75–250 cm values as computedby LPJ is situated more to the north than is the case for theGoodrich model, especially in Western Eurasia. In addition,ALT as computed by LPJ is lower in the Far East, withvalues between 25 and 50 cm. The strong impact of soilparameters on ALT can be seen on the border betweendifferent soil texture classes, for example, around 75�N,95�E.

Figure 1. Principal steps in calculating daily thaw depth inLPJ.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

5 of 14

GB1012

[31] By comparing the simulated permafrost distributionagainst the permafrost map by Brown et al. [1998] thegeneral applicability of the frost index by Nelson andOutcalt [1987] can be confirmed (see auxiliary material).However, by setting F in equation (6) to 0.6, which wasfound to represent the border between discontinuous andsporadic permafrost [Nelson and Outcalt, 1987; Christensenand Kuhry, 2000], it seems that rather the zonation ofcontinuous permafrost was delineated, as can be seen fromdiscrepancies south of the Ob Bay (see auxiliary material).

Reasons for this could also be the usage of different climateinput data or different formulations for snow properties likedensity and thermal conductivity.[32] Given that the discrepancies noted are discrepancies

between models, that the comparison with observations ison average satisfactory though with large deviations inparticular cases, and that the spatial patterns between themodels are mostly similar, we conclude that the model issufficiently accurate for the present study, which is focusedon the changes caused in vegetation by changes permafrost-

Figure 2. Mean seasonal thaw depth (crosses) and their standard deviation (bars) at five stations inAlaska between 1995 and 1997 [Nelson, 1996] and one station in the Russian Far East between 1981 and1985 (S. Zimov, personal communication, 2005) in comparison to LPJ simulation results of daily thawdepth (line).

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

6 of 14

GB1012

induced changes in site hydrology. In the absence of moredetailed data and considerably more expensive computa-tional effort, both of which are currently unrealistic for thecontinental scale, we argue that the present model issufficient for use in a DGVM.

3.2. Impacts of Freeze-Thaw Processes on the WaterBalance

[33] Effects of permafrost on the water balance arequantified by analyzing runoff simulated by the LPJ-noPand LPJ-P model runs for a number of large river water-sheds (more than 200,000 km2) that are located (partially)within the permafrost zone, and for which discharge meas-urements over a period of more than 10 years were available[Vorosmarty et al., 1996]. The model results shown inFigure 5 demonstrate the importance of freeze-thaw pro-cesses on a continental scale for the water balance in theboreal zone. Without the consideration of these processes(LPJ-noP model) runoff is clearly underestimated in springby 30–60% because the assumed high available waterholding capacity of the soil does not reflect the reality wherepartly frozen grounds additionally constrain the holdingcapacity (see section 2.3). By contrast, the enhanced modelversion (LPJ-P model) leads to runoff results over basins oflarge Arctic rivers that compares well with the observationsin spring. Snow melt and precipitation over a still frozenground during spring cause the observed runoff peak. Thisprocess is represented by the model now.

Figure 3. Comparison of simulated maximum thaw depthwith observations at 29 stations from Alaska, Canada, andRussia from 1992 until 2002 [Brown et al., 2003].Observations represent mean values of measurementswithin 100 m2 to 1 km2 large areas with standard deviationsin the range of 5–20 cm [Brown et al., 2003].

Figure 4. Time series of active-layer thickness anomalies at four CALM [Brown et al., 2003] stations incomparison to LPJ-P model results.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

7 of 14

GB1012

[34] However, simulated runoff remains too low in latesummer and autumn. Additional analysis (see auxiliarymaterial) revealed that this is due to an overestimation ofthe amount of water transpired or intercepted by vegetationin the summer and in autumn, which is a consequence of theDGVM simulating dense forest cover (foliage cover 95%)throughout the region when the actual foliage cover is lessdense (20–75% following Hansen et al. [2003]) andspatially heterogeneous. Model runs with an improvedlower vegetation density showed little change in the springrunoff but correctly increased runoff in autumn, in accor-dance with observations [Beer et al., 2006].[35] Despite the analysis of simulated river runoff, which

allows for conclusions about impacts of freeze-thaw pro-cesses on the water balance at continental scale, an evalu-ation of simulated soil moisture dynamics at stations locatedin various ecosystems within the permafrost zone (Figure 6)is required to assess the reliability of the hydrologicalscheme. During the first days of soil thawing modeledrelative water content oscillate highly in contrast to obser-vations. This is due to a very thin layer of thawed soilsimulated by LPJ during this period while observations areconsidered after some centimeter of soil has already thawed.Except for this fact, modeled soil moisture is within therange of standard deviation of different observations at thesame site (Figure 6b) and its dynamics is represented by

model results (most notably Figure 6a and Figure 6d). Asfor the thaw depth calculation, highest uncertainty inmodeled soil moisture is due to assumed homogenous soilproperties with depth. In comparison to mean observationsin the first meter of soil, LPJ seems to overestimate soilwater content in spring at Bayelva, Svalbard (Figure 6d).The comparison to the observation at 6 cm soil depth,however, clarifies this fact (Figure 6d). During this timeperiod, LPJ results represent only the water content of theupper thin layer that is thawed.

3.3. Consequences for the Fire Return Interval

[36] As can be seen in Figure 7, simulation results of thefire return interval (FRI) notably differ between the bothmodel versions LPJ-P and LPJ-noP. FRI values simulatedby LPJ-noP are considerably too long. With the consider-ation of freeze-thaw processes (LPJ-P) FRI is mostlyreduced to values of 100 to 200 years in Northern Eurasia,in accordance with observations [Bonan and Shugart, 1989;Kharuk et al., 2005].[37] Shorter FRI values are associated with higher carbon

emissions. The LPJ-P model estimates twofold emissions incentral Siberia compared to LPJ-noP results, which are upto fourfold in Western Siberia. Values ranges from 25 to45 gC/m2/a on average between 1993 and 2003.

Figure 5. Comparison of simulated monthly long-term means of fresh water runoff in Siberian Arcticriver basins with observations of river discharge [Vorosmarty et al., 1996] between 1936 and 1984.Simulated runoff values are shifted by 1 month for the two large rivers Yenisey and Lena and by0.5 months for the mesoscale river basins Yana and Indirgirka since no river routing scheme is embeddedin the LPJ-DGVM. LPJ-PIK denotes to results by the original LPJ model [Sitch et al., 2003].

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

8 of 14

GB1012

[38] The reason for this increase in fire activity is asfollows. Increased high runoff of water from snow melt andprecipitation on the surface of predominantly frozengrounds during spring leads to dry soils. Vegetation con-tributes to this drying by rapidly taking up water after springgreenup from that thin fraction of the active layer that isalready thawed. As a consequence, fire probability and thusfire occurrence in Siberia is highest in spring [Korovin,1996]. The global fire module embedded in the LPJ-DGVM[Thonicke et al., 2001] uses a statistical approach that doesnot resolve the interannual variability of fire occurrence, butthe demonstrated combined effect of permafrost on runoff(see section 3.2) and FRI emphasizes the strong linkbetween permafrost, water balance and fire probability, alink represented by the model LPJ-P with consequences forthe carbon density of standing vegetation.

3.4. Permafrost Effects on Growth

[39] In permafrost regions, a large amount of water is lostas runoff during spring (section 3.2) and is hence notavailable for vegetation later in the growing season. Inaddition, soil water content is constrained by thaw depthduring early summer [Benninghoff, 1952]. Besides theeffects of soil moisture on fire, water availability limitsNPP in a twofold manner. As a fast process, GPP isconstrained by stomatal conductance [Farquhar et al.,1980]. On the other hand, water availability influences the

Figure 6. Dynamics of volumetric soil water content for the year (a–c) 2003 or (d) 1999. The LPJ-Pmodel is driven by local soil texture information and precipitation (except for Figure 6c) for thiscomparison. Asterisks denote daily precipitation (mm). In Figure 6b, means and standard deviations offour plots are shown. Data were provided by Overduin [2005] (a), A. Rodionov, personal communication,2006 (Figure 6b), S. Zimov, personal communication, 2006 (Figure 6c), and Roth and Boike [2001](Figure 6d).

Figure 7. Mean fire return interval in Northern Eurasia1993–2003 as simulated by (middle) LPJ-noP and (bottom)LPJ-P. (top) The result by the original LPJ model [Sitch etal., 2003] is shown for comparison.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

9 of 14

GB1012

allocation of carbon to the components of plants [Bongartenand Teskey, 1987; Mencuccini and Grace, 1995] leading tolower leaf area under drier conditions [Grier and Running,1977; Gholz, 1982].[40] To differentiate between these effects, two additional

simulations excluding fire, LPJ-noFire, or excluding water-stress impacts on carbon allocation (equation (4)),LPJ-noStress, are compared to LPJ-noP and LPJ-P for alarch-dominated permafrost area around 68.25�N and120.25�E in the first 300 simulation years (Figure 8).LPJ-P results of leaf area index (LAI), foliage projectedcover (FPC), NPP and finally VCD are lowest and bestcomparable to independent estimates. Without the consid-eration of permafrost, these parameters acquire considerablyhigher (unrealistic) values. LPJ-noFire results in only slightlyhigher values than LPJ-P during the first 100 simulationyears, revealing that the direct impacts of fires are not thedominant effect.[41] By contrast, LPJ-noStress shows higher values than

LPJ-P already after about 50 years since no extra investmentin roots is assumed (equation (4)). This illustrates the impactof higher carbon allocation to roots on LAI and NPP, thusVCD in permafrost regions. However, LPJ-noStress stillproduces lower LAI, NPP and VCD than LPJ-noP (Figure 8).This is due to the fact that lower water availability alsodirectly impacts productivity [e.g., Gerten et al., 2005].

[42] During the simulation period, fire also has a directimpact on VCD, as can be seen by the comparison of LPJ-Pand LPJ-noFire results in Figure 8d. However, the resultsshown in Figure 8 also demonstrate the impact of wateravailability on growth through canopy conductance andcarbon allocation to roots. For the same location, Figure 9clarifies the impact of plant available soil water content(SWC) on NPP that is mediated by thaw depth. It showsLPJ-noP and LPJ-P simulation results of NPP, thaw depthand SWC on a daily time step for the simulation year 50.Without the representation of freeze-thaw processes, a verydeep and saturated soil would be assumed with SWC ofabout 200 mm throughout the whole year. By contrast,SWC is constrained by thaw depth in the LPJ-P experimentleading to reduced NPP. In addition, the timing of the firstday with photosynthesis is altered forward in time.

3.5. Vegetation Carbon Density

[43] Simulated VCD approximates inventory estimatesmuch better in the LPJ-P simulation as compared to LPJ-noP (Figure 10a), where the deviation is considerable. VCDis estimated by LPJ-P to range between 6 and 7 kgC/m2

south of 63�N decreasing to 1–2 kgC/m2 north of 69�N.Freeze-thaw processes account for a 1.5 to 2.5-fold reduc-tion in VCD in the South (Figure 10b) which increased with

Figure 8. Ecosystem variables in the region around 68.25�N, 120.25�E (larch forest, northern taiga) forthe first 300 simulation years. Simulation results are with (LPJ-P) or without (LPJ-noP) the considerationof freeze-thaw processes, or with it but without the consideration of fire (LPJ-P-noFire) or water stress(LPJ-P-noStress). Independent estimates are shown in grey and are from (a) Schulze et al. [1999],(b) Hansen et al. [2003], and (c, d) Stolbovoi and McCallum [2002]. The width of grey bars indicates therange of observations (Figure 8a), the standard deviation (Figure 8b), and the reported uncertainty range[see Shvidenko et al., 2001] of forest inventory estimates (Figures 8c and 8d).

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

10 of 14

GB1012

latitude to a fivefold reduction in high latitudes. Thisdemonstrates the impact of freeze-thaw processes on bio-mass through disturbances and water availability. Soilthermal dynamics are primary processes of boreal ecosys-tems that cannot be neglected, as has commonly been thecase, in DGVM simulations. Besides the fact that VCD islower in magnitude, its changeover between 60 and 70�E issmoother in LPJ-P (Figure 10a) than in LPJ-noP owing tothe effect of a decreasing ALT on biomass [Bonan, 1989].

4. Discussion

[44] Freeze-thaw processes strongly codetermine the sea-sonality of soil moisture in the boreal zone, particularly inthe permafrost zone, where the thermal dynamics in theactive layer are significantly altered. Owing to the intrinsicphysiological coupling of the carbon and water cycles invegetation, with an additional important soil moisturefeedback on vegetation mediated through fire, the most

important disturbance in the boreal forest, freeze-thawprocesses cannot be neglected in process-based analyzesof the continental-scale carbon and water balance at boreallatitudes.[45] The permafrost module we therefore implemented

into the LPJ-DGVM reflects the general seasonal cycle ofthawing depth and the large-scale spatial patterns of itsmaximum. We are aware of the limited significance ofcomparing simulated thaw depth based on 0.5� � 0.5�averages of climate and soil data with observations at muchsmaller scales (1 ha or 1 km2) because of the spatiallyhighly heterogeneous nature of thaw depth processes, whichare influenced by variations in soil properties, slope andslope aspect [Washburn, 1979; Nelson et al., 1997]. None-theless, such a comparison allows to demonstrate thereliability of computed first-order patterns. Our evaluationof simulated thaw depth seasonality and spatial patterns ofALTwith observations and the results of a recent permafrostmodel, the Goodrich model, shows the capability of ourmethod to generically represent freeze-thaw processes.Discrepancies to observations are within the range ofvariability of thaw depth that occurs naturally on a muchsmaller spatial scale than applied in this study. A limitationof the method applied is the assumption of homogenous soilproperties with depth, a limitation that equally applies to allother large-scale permafrost models. The comparison ofresults obtained from different approaches is a problem thatcan currently not be resolved satisfactorily and places thestrongest constraints on our current ability to demonstrateand validate results.[46] It should be noted that the current snow module

embedded into the LPJ model does not take into accountvegetation effects on solar radiation extinction and atmo-spheric turbulence which are a far greater influence on snowcover than interception. This could lead to an underestima-tion of snow depth in forest areas with effects on the soilthermal regime. In addition to this, DGVMs do not resolvesubgrid localizations of spatial heterogeneity hence are notable to represent discontinuous permafrost. We acknowl-edge that the greatest change in permafrost is likely to

Figure 9. Daily estimates of woody NPP simulated by theLPJ-noP and LPJ-P models, thaw depth simulated by LPJ-P,and soil water content (Wplant) simulated by LPJ-noP andLPJ-P around 68.25�N, 120.25�E (larch forest, northerntaiga) in 2003.

Figure 10. (a) Vegetation carbon density as a function of latitude in the SIBERIA-II study transectswath in 2003 as derived by the land-ecosystem inventory approach of the International Institute forApplied Systems Analysis, Austria [Nilsson et al., 2005], and as simulated by LPJ-noP and LPJ-P. LPJ-PIK denotes to the result by the original LPJ model [Sitch et al., 2003]. (b) Ratio between LPJ-noP andLPJ-P results shown in Figure 10a.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

11 of 14

GB1012

happen in this zone. Further, regional-scale studies are thusdesirable to explore these effects in more detail.[47] Given the permafrost module, our evaluation of

runoff, fire return interval and VCD as computed in re-sponse to its dynamics demonstrates the importance offreeze-thaw processes for the functioning of boreal ecosys-tems. High runoff of snow melt and precipitation on a stillfrozen ground in spring subsequently leads to dry soilconditions. In permafrost regions, this effect is amplifiedby the fact that vegetation is able to draw water only fromthe shallow fraction of the active layer which had beenalready thawed in spring and early summer. The ensuinglow soil water content, which rapidly translates into lowlitter moisture, strongly increases fire probability reducingthe fire return interval. Our model results show that freeze-thaw processes and the related site hydrology are essentialfor realistic fire disturbance modeling on a continental scalein the boreal zone. The actual area burned, and its spatialand temporal variability, however, depend on additionalfactors such as ignition, tree allometry and structure, andlitter quality. In addition to disturbances, NPP is the maindeterminant of biomass as it quantifies vegetation growthand background mortality. The relatively low soil wateravailability during the boreal summer causes low values ofNPP, hence plant growth depressed owing to low LAI andcanopy conductance in permafrost regions. This remains thecase despite various adaptations of boreal vegetation topermafrost conditions, represented in our model, such as arapid greenup and shallow rooting systems. Even if theresulting low foliage cover would be partly compensated onthe ecosystem level by more seedlings during the time, stemdensity increased, also contributing to lower VCD values.Our results support the fact that plant growth in the northerntaiga is limited by water availability in addition to otherfactors, such as nitrogen uptake.[48] The dynamic ecosystem model still simulates too

high VCD north of 60�N, where values may be up to twiceas large as independent estimates produced by IIASA fromforest inventory data (Figure 10). A number of limitations totree establishment and growth in high latitudes that are notrepresented in the simulation model are a likely explanationfor this overestimation, for example temperature depen-dence of cell division [Korner, 2003] or of seed qualityand quantity [Black and Bliss, 1980; Sveinbjornsson et al.,2002], limitations by nutrient availability [Bonan andShugart, 1989] or limitation of tree establishment by stand-ing water on the permafrost [Sveinbjornsson et al., 2002]. Inthe boreal zone, a considerable amount of heterotrophicrespiration was detected during winter [Monson et al., 2006]with consequences on nitrogen mineralization, hence nitro-gen availability during the growing season. This positiveeffect of winter soil respiration on productivity is notconsidered in this study but LPJ even assumes no limitationof nitrogen uptake. Future modeling studies which aim atclarifying the nutrient limitation of VCD will have toconsider this positive effect of winter soil respiration onnutrient availability.[49] However, our results clearly point out that freeze-

thaw processes are of first order and should not be neglectedin a process-based modeling framework. The increase of

ALT and the disappearance of permafrost under warmingconditions [Anisimov and Nelson, 1997; Stendel andChristensen, 2002] will lead to a substantial alteration ofthe site hydrology [Kane et al., 1992]. The consideration ofthe most important consequences of freeze-thaw processeson the hydrological regime in high latitudes by the LPJ-Pmodel give us an opportunity to arrive at more reliableestimates of the future dynamics of the global carbon andwater balances, which are both strongly influenced by thehigh latitudes. Uncertainties exist with respect to thedynamics, for example, of peat, cryoturbation, methano-genesis and methanotropy, of which large-scale observationsare not currently available. In addition, the impact of fire onnutrient availability should be considered for projections ofthe future carbon balance in the boreal zone. Furthermore,feedbacks between changing vegetation coverage and cli-mate through albedo and carbon and water fluxes can belarge [e.g., Betts, 2000; Chapin et al., 2005]. Nonetheless,the inclusion of freeze-thaw processes is a significant steptoward a more comprehensive understanding of ecosystemprocesses in the boreal zone and Arctic tundra. Our freeze-thaw algorithm of medium complexity also is suitable forapplication in the land surface scheme of a General Circu-lation Model since additional computation cost is minimal.

5. Conclusion

[50] The combined evaluation of river runoff, fire returninterval and vegetation carbon density, as simulated by thepermafrost-enhanced LPJ-DGVM, demonstrates that bothlimitation in water availability during the growing seasonand the disturbance dynamics of boreal vegetation substan-tially contribute to the low vegetation carbon density observedin permafrost regions of the boreal zone. This supports theconclusion that the availability of liquid water and soilmoisture seasonality are also major limitations of biomasswithin the permafrost zone, in addition to temperature andnitrogen availability. Since moisture dynamics are stronglydetermined by the thermal dynamics in these regions, freeze-thaw processes are found to be ecosystem processes of firstorder. This conclusion demonstrates the potential of vegeta-tion in high latitudes to sequester more carbon than itcurrently does owing to warming that leads to increasingALT or even the complete disappearance of permafrost.

[51] Acknowledgments. The authors thank B. Smith, S. Schaphoff,and S. Sitch for support regarding the LPJ-DGVM code, and M. Fink,D. Knorr, W. Cramer, V. Romanovsky, D. Riseborough, O. Anisimov, andP. Ciais for valuable notes. A. Shvidenko and I. McCallum from theInternational Institute for Applied Systems Analysis, Austria made the forestinventory data available. We are grateful to G. and S. Zimov who providedseasonal thaw depth and soil moisture measurements, to P. Overduin,A. Rodionov, and J. Boike, who provided seasonal soil moisture observa-tions, and to all members of the SIBERIA-II consortium for discussions.Financial support was provided by the European Commission under theSIBERIA-II project, contract EVG2-2001-00008. W. L. and D. G. werefunded by the German Ministry of Education and Research under theDEKLIM project CVECA. We thank Nigel Roulet and an anonymousreviewer for helpful comments on an earlier manuscript of this paper.

ReferencesAnisimov, O. A., and F. E. Nelson (1997), Permafrost zonation and climatechange in the Northern Hemisphere: Results from the transient generalcirculation models, Clim. Change, 35, 241–258.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

12 of 14

GB1012

Beer, C. (2005), Klimatische Ursachen der Entwicklung der Biomasse derWalder Russlands von 1981 bis 1999 - Eine Analyse mittels eines an dieboreale Zone angepassten und durch Satellitenprodukte beschranktendynamischen globalen Vegetationsmodells, 88 pp., Ph.D. thesis,Friedrich-Schiller-Univ., Jena, Germany.

Beer, C., W. Lucht, C. Schmullius, and A. Shvidenko (2006), Small netcarbon dioxide uptake by Russian forests during 1981–1999, Geophys.Res. Lett., 33, L15403, doi:10.1029/2006GL026919.

Benninghoff, W. S. (1952), Interaction of vegetation and soil frost phenom-ena, Arctic, 5, 34–44.

Betts, R. A. (2000), Offset of the potential carbon sink from boreal foresta-tion by decreases in surface albedo, Nature, 408, 187–190.

Black, R., and L. Bliss (1980), Reproductive ecology of Picea mariana(Mill.) at the tree line near Inuvik, Northwest Territories, Canada, Ecol.Monogr., 50(3), 331–354.

Bonan, G. B. (1989), Environmental factors and ecological processes con-trolling vegetation paterns in boreal forests, Landscape Ecol., 3(2), 111–130.

Bonan, G., and H. Shugart (1989), Environmental factors and ecologicalprocesses in boreal forests, Annu. Rev. Ecol. Syst., 20, 1–28.

Bongarten, B. C., and R. O. Teskey (1987), Dry weight partitioning and itsrelationship to productivity in loblolly pine seedlings from seven sources,For. Sci., 33(2), 255–267.

Briffa, K., F. H. Schweingruber, P. D. Jones, T. J. Osborn, S. G. Shiyatov,and E. A. Vaganov (1998), Reduced sensitivity of recent tree-growth totemperature at high northern latitudes, Nature, 391, 678–682.

Brown, J., O. Ferrians Jr., J. Heginbottom, and E. Melnikov (1998), Circum-Arctic map of permafrost and ground-ice conditions, http://nsidc.org/data/docs/fgdc/ggd318_map_circumarctic/index.html, Natl. Snow and IceData Cent., Boulder, Colo.

Brown, J., K. Hinkel, and F. Nelson (2003), Circumpolar Active LayerMonitoring (CALM) program network: Description and data, in Interna-tional Permafrost Association Standing Committee on Data Informationand Communication (comp.) Circumpolar Active-Layer Permafrost Sys-tem, Version 2.0 [CD-ROM], edited by M. Parsons and T. Zhang, Natl.Snow and Ice Data Cent., Boulder, Colo.

Campbell, G. S., and J. M. Norman (2000), An Introduction to Environ-mental Biophysics, 2nd ed., Springer, New York.

Chapin, F. S., III, et. al. (2005), Role of land-surface changes in arcticsummer warming, Science, 310, 657–660.

Choudhury, B., and N. E. DiGirolamo (1998), A biophysical process-basedestimate of global land surface evaporation using satellite and ancillarydata - I. Model description and comparison with observations, J. Hydrol.,205, 186–204.

Christensen, J. H., and P. Kuhry (2000), High-resolution regional climatemodel validation and permafrost simulation for the East European Rus-sian Arctic, J. Geophys. Res., 105(D24), 29,647–29,658.

Cramer, W., et al. (2001), Global response of terrestrial ecosystem structureand function to CO2 and climate change: Results from six dynamic globalvegetation models, Global Change Biol., 7, 357–373.

Esper, J., and F. H. Schweingruber (2004), Large-scale treeline changesrecorded in Siberia, Geophys. Res. Lett., 31, L06202, doi:10.1029/2003GL019178.

Farquhar, G., S. Caemmerer, and J. Berry (1980), A biochemical model ofphotosynthetic CO2 assimilation in leafs of C3 plants, Planta, 149(1),78–90.

Food and Agriculture Organization (1991), The Digitized Soil Map of theWorld (Release 1.0) [Disks], Rome.

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.

Gerten, D., W. Lucht, S. Schaphoff, W. Cramer, T. Hickler, and W. Wagner(2005), Hydrologic resilience of the terrestrial biosphere, Geophys. Res.Lett., 32, L21408, doi:10.1029/2005GL024247.

Gholz, H. (1982), Environmental limits on aboveground net primary pro-duction, leaf area, and biomass in vegetation zones of the Parzific North-west, Ecology, 63, 469–481.

Goulden, M. L., et al. (1998), Sensitivity of boreal forest carbon balance tosoil thaw, Science, 279, 214–217.

Grier, C., and W. Running (1977), Leaf area of mature northwesternconiferous forests: Relation to site water balance, Ecology, 58, 893–899.

Hansen, M., R. DeFries, J. Townshend, M. Carroll, C. Dimiceli, andR. Sohlberg (2003), Global percent tree cover at a spatial resolution of500 meters: First results of the MODIS Vegetation Continuous Fieldsalgorithm, Earth Interact., 7(10), 1–15.

Haxeltine, A., and I. C. Prentice (1996), A general model for the light-useefficiency of primary production, Funct. Ecol., 10(5), 551–561.

Hinzman, L. D., D. L. Kane, R. E. Gieck, and K. R. Everett (1991),Hydrologic and thermal properties of the active layer in the AlaskanArctic, Cold Reg. Sci. Technol., 19, 95–110.

Jumikis, A. (1966), Thermal Soil Mechanics, Rutgers Univ. Press, NewBrunswick, N. J.

Kane, D., L. Hinzman, M. Woo, and K. Everett (1992), Arctic hydrologyand climate change, in Arctic Ecosystems in a Changing Climate: AnEcophysiological Perspective, edited by F. Chapin III et al., pp. 35–57,Elsevier, New York.

Kasischke, E. S., E. J. Hyer, P. C. Novelli, L. P. Bruhwiler, N. H. F. French,A. I. Sukhinin, J. H. Hewson, and B. J. Stocks (2005), Influences ofboreal fire emissions on Northern Hemisphere atmospheric carbon andcarbon monoxide, Global Biogeochem. Cycles, 19 , GB1012,doi:10.1029/2004GB002300.

Kharuk, V. I., M. L. Dvinskaya, and K. J. Ranson (2005), The spatiotem-poral pattern of fires in northern taiga larch forests of central Siberia,Russ. J. Ecol., 36(5), 302–311.

Korner, C. (2003), Carbon limitations in trees, J. Ecol., 91, 4–17.Korovin, G. (1996), Analysis of the distribution of forest fires in Russia, inFire in Ecosystems of Boreal Eurasia, edited by J. Goldammer andV. Furyaev, pp. 112–128, Springer, New York.

Lucht, W., I. C. Prentice, R. B. Myneni, S. Sitch, P. Friedlingstein,W. Cramer, P. Bousquet, W. Buermann, and B. Smith (2002), Climaticcontrol of the high-latitude vegetation greening trend and Pinatubo effect,Science, 296, 1687–1689.

McGuire, A. D., et al. (2001), Carbon balance of the terrestrial biosphere inthe twentieth century: Analyses of CO2, climate and land use effects withfour process-based ecosystem models, Global Biogeochem. Cycles,15(1), 183–206.

Mencuccini, M., and J. Grace (1995), Climate influences the leaf/sapwoodarea ratio in Scots pine, Tree Physiol., 15, 1–10.

Mitchell, T. D., and P. D. Jones (2005), An improved method of construct-ing a database of monthly climate observations and associated high-resolution grids, Int. J. Climatol., 25(6), 693–712.

Monson, R. K., D. L. Lipson, S. P. Burns, A. A. Turnipseed, A. C. Delany,M. W. Williams, and S. K. Schmidt (2006), Winter forest soil respirationcontrolled by climate and microbial community composition, Nature,439, 711–714.

Nelson, F. (1996), Air temperature and thaw depth, north slope, Alaska:Field measurements (1995–97) and computed values, in InternationalPermafrost Association Standing Committee on Data Information andCommunication (comp.) Circumpolar Active-Layer Permafrost System,Version 2.0, [CD-ROM], edited by M. Parsons and T. Zhang, Natl. Snowand Ice Data Cent., Boulder, Colo.

Nelson, F. E., and S. I. Outcalt (1987), A computational method for pre-diction and regionalization of permafrost, Arct. Alp. Res., 19(3), 279–288.

Nelson, F. E., N. I. Shiklomanov, G. Mueller, K. M. Hinkel, D. A. Walker,and J. G. Bockheim (1997), Estimating active-layer thickness over a largeregion: Kuparuk river basin, Alaska, U. S. A., Arct. Alp. Res., 29(4),367–378.

Nemani, R. R., C. D. Keeling, H. Hashimoto, W. M. Jolly, S. C. Piper, C. J.Tucker, R. B. Myneni, and S. W. Running (2003), Climate-drivenincreases in global terrestrial net primary production from 1982 to1999, Science, 300, 1560–1563.

Nilsson, S., et al. (2005), Attempting a verified regional terrestrial biota fullcarbon account: Experiences from central Siberia, paper presented at 7thInternational Conference on CO2, Clim. Monit. and Diag. Lab., NOAA,Denver, Colo.

Oechel, W. C., S. J. Hastings, G. Vourlitis, M. Jenkins, G. Richers, andN. Gruike (1993), Recent change of Arctic tundra ecosystems from a netcarbon dioxide sink to a source, Nature, 361, 520–523.

Oelke, C., T. Zhang, M. C. Serreze, and R. L. Armstrong (2003), Regional-scale modelling of soil freeze/thaw over the Artic drainage basin,J. Geophys. Res., 108(D10), 4314, doi:10.1029/2002JD002722.

Oesterle, H., F.-W. Gerstengarbe, and P. Werner (2003), Homogenisierungund Aktualisierung des Klimadatensatzes der Climate Research Unit derUniversitat of East Anglia, Norwich, Terra Nostra, 6, 326–329.

Overduin, P. P. (2005), The physical dynamics of patterned ground in thenorthern foothills of the Brooks Range, Alaska, Ph.D. dissertation, Univ.of Alaska Fairbanks, Fairbanks.

Peylin, P., P. Bousquet, C. Le Quere, S. Sitch, P. Friedlingstein, G. McKinley,N. Gruber, P. Rayner, and P. Ciais (2005), Multiple constraints on regionalCO2 flux variations over land and oceans,Global Biogeochem. Cycles, 19,GB1011, doi:10.1029/2003GB002214.

Reich, P., M. Walters, and D. Ellsworth (1997), From tropics to tundra:Global convergence in plant functioning, Proc. Natl. Acad. Sci. U. S. A.,94, 13,730–13,734.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

13 of 14

GB1012

Roth, K., and J. Boike (2001), Quantifying the thermal dynamics of apermafrost site near Ny-Alesund, Svalbard, Water Resour. Res., 37(12),2901–2914.

Schmullius, C., and S. Hese (2002), SIBERIA-II: Sensor systems and dataproducts for greenhouse gas accounting, paper presented at DFD Nut-zerseminar, Dtsch. Fernerkundungsdatenzentrum, Oberpfaffenhofen,Germany. (Available at http://www.caf.dlr.de/caf/aktuelles/veranstaltungen/nutzerseminar/dfd_19/publikationen/papers/schmullius.pdf)

Schulze, E. D., et al. (1999), Productivity of forests in the Eurosiberianboreal region and their potential to act as a carbon sink—A synthesis,Global Change Biol., 5, 703–722.

Serreze, M. C., et al. (2000), Observational evidence of recent change in thenorthern high-latitude environment, Clim. Change, 46, 159–207.

Shvidenko, A., S. Nilsson, V. Stolbovoi, M. Gluck, D. Schepaschenko, andV. Rozhkov (2000), Aggregated estimation of the basic parameters ofbiological production and the carbon budget of Russian terrestrial eco-systems: 1. Stocks of plant organic mass, Russ. J. Ecol., 31(6), 371–378.

Shvidenko, A., S. Nilsson, V. Stolbovoi, V. Rozhkov, and M. Gluck (2001),Aggregated estimation of the basic parameters of biological productionand the carbon budget of Russian terrestrial ecosystems: 2. Net primaryproduction, Russ. J. Ecol., 32(2), 71–77.

Sitch, S., et al. (2003), Evaluation of ecosystem dynamics, plant geographyand terrestrial carbon cycling in the LPJ dynamic global vegetation model,Global Change Biol., 9, 161–185.

Stendel, M., and J. Christensen (2002), Impact of global warming on per-mafrost conditions in a coupled GCM, Geophys. Res. Lett., 29(13), 1632,doi:10.1029/2001GL014345.

Stolbovoi, V., and I. McCallum (2002), Land Resources of Russia [CD-ROM], Int. Inst. for Appl. Syst. Anal., Laxenburg, Austria.

Sugimoto, A., N. Yanagisawa, D. Naito, N. Fujita, and T. C. Maximov(2002), Importance of permafrost as a source of water for plants in eastSiberian taiga, Ecol. Res., 17, 493–503.

Sveinbjornsson, B., A. Hofgaard, and A. Lloyd (2002), Natural causes ofthe tundra-taiga boundary, Ambio Spec. Rep., 12, 23–29.

Thonicke, K., S. Venevsky, S. Sitch, and W. Cramer (2001), The role of firedisturbance for global vegetation dynamics: Coupling fire into a dynamicglobal vegetation model, Global Ecol. Biogeogr., 10, 661–677.

Vorosmarty, C., B. Fekete, and B. Tucker (1996), River Discharge Data-base, Version 1.0 (RivDIS v1.0): A Contribution to IHP-V Theme 1,UNESCO Press, Paris.

Wagner, W., K. Scipal, C. Pathe, D. Gerten, W. Lucht, and B. Rudolf(2003), Evaluation of the agreement between the first global remotelysensed soil moisture data with model and precipitation data, J. Geophys.Res., 108(D19), 4611, doi:10.1029/2003JD003663.

Washburn, A. L. (1979), Geocryology: A Survey of Periglacial Processesand Environments, Edward Arnold, London.

Woodward, F. I. (1995), Ecophysiological controls of conifer distributions,in Ecophysiology of Coniferous Forests, edited by W. K. Smith and I. M.Hinckley, pp. 79–94, Elsevier, New York.

Zhuang, Q., et al. (2003), Carbon cycling in extratropical terrestrial eco-systems of the Northern Hemisphere during the 20th century: A modelinganalysis of the influences of soil thermal dynamics, Tellus, Ser. B, 55(3),751–776.

Zimov, S. A., G. M. Zimova, S. P. Daviodov, Y. V. Voropaeva, S. F.Prosiannikov, O. V. Prosiannikova, I. V. Semiletova, and I. P. Semiletov(1993), Winter biotic activity and production of CO2 in Siberian soils: Afactor in the greenhouse effect, J. Geophys. Res., 98(D3), 5017–5023.

�������������������������C. Beer, Max Planck Institute for Biogeochemistry, Hans-Knoll-Str. 10,

D-07745 Jena, Germany. (cbeer@bgc-jena.mpg.de)D. Gerten and W. Lucht, Potsdam Institute for Climate Impact Research,

P.O. Box 601203, D-14412 Potsdam, Germany.C. Schmullius, Friedrich Schiller University, Institute of Geography,

Lobdergraben 32, D-07743 Jena, Germany.K. Thonicke, School of Geographical Sciences, University of Bristol,

Bristol BS8 1KJ, UK.

GB1012 BEER ET AL.: FREEZE-THAW EFFECTS ON BIOMASS

14 of 14

GB1012