16
Ecological Modelling 189 (2005) 183–198 Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment T. Palosuo a,, J. Liski b , J.A. Trofymow c , B.D. Titus c a European Forest Institute, Torikatu 34, FIN-80100 Joensuu, Finland b Finnish Environment Institute, P.O. Box 140, FIN-00251 Helsinki, Finland c Natural Resources Canada, Canadian Forest Service, Pacific Forestry Centre, 506 West Burnside Road, Victoria, BC V8Z 1M5, Canada Received 25 August 2004; received in revised form 3 February 2005; accepted 29 March 2005 Available online 31 May 2005 Abstract Litterbag experiments provide valuable data for testing the accuracy of predictions of decomposition from soil carbon models. The soil carbon model Yasso describes litter decomposition based on basic climate and litter quality information, and was calibrated using European litterbag data. In this study, we tested the predictive capabilities of Yasso using independent litterbag data for 10 foliage litter types decomposed for 6 years at 18 upland forest sites across Canada (CIDET). The model underestimated mass of leaf litters remaining on CIDET sites, with only a small systematic error in predicting the effects of climate when effective temperature sum was used as the temperature variable in the model. The overall rate of decomposition was predicted correctly when mean annual temperature was used as the temperature variable, but then the model substantially overestimated climatic effects. The model correctly predicted differences in decomposition rates among litter types in the early years of decomposition, but underestimated them in later years. The decomposition rate of the litter type richest in phenolic compounds (larch needles) was systematically overestimated, and that of the litter type richest in O-alkyl compounds (grass leaves) was systematically underestimated. Accounting for these factors would improve the general applicability of the model. However, accounting for the initial nitrogen concentration of litter did not improve the accuracy of the model unless the initial lignin (i.e., acid unhydrolyzable residue) content was also taken into account. We conclude that the model Yasso accounts for most of the effects of climate and initial litter quality on the decomposition of a range of foliage litter types under varying climate conditions. Recalibration of the reference decomposition rates used in the model may improve the accuracy when applying the model outside of Europe. © 2005 Elsevier B.V. All rights reserved. Keywords: Dynamic model; Yasso; Decomposition; Litterbag; Soil carbon; CIDET Corresponding author at: European Forest Institute, c/o Suitia Research Farm, University of Helsinki, FIN-02570 Siuntio, Finland. Tel.: +358 9 8190 8515; fax: +358 9 8190 8528. E-mail address: [email protected] (T. Palosuo). 0304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2005.03.006

Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

Embed Size (px)

Citation preview

Page 1: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

Ecological Modelling 189 (2005) 183–198

Litter decomposition affected by climate and litterquality—Testing the Yasso model with litterbag data from

the Canadian intersite decomposition experiment

T. Palosuoa,∗, J. Liskib, J.A. Trofymowc, B.D. Titusc

a European Forest Institute, Torikatu 34, FIN-80100 Joensuu, Finlandb Finnish Environment Institute, P.O. Box 140, FIN-00251 Helsinki, Finland

c Natural Resources Canada, Canadian Forest Service, Pacific Forestry Centre, 506 West Burnside Road, Victoria, BC V8Z 1M5, Canada

Received 25 August 2004; received in revised form 3 February 2005; accepted 29 March 2005Available online 31 May 2005

Abstract

Litterbag experiments provide valuable data for testing the accuracy of predictions of decomposition from soil carbon models.The soil carbon model Yasso describes litter decomposition based on basic climate and litter quality information, and wascalibrated using European litterbag data. In this study, we tested the predictive capabilities of Yasso using independent litterbagdata for 10 foliage litter types decomposed for 6 years at 18 upland forest sites across Canada (CIDET).

The model underestimated mass of leaf litters remaining on CIDET sites, with only a small systematic error in predictingall rate of

the model

ition, butneedles)allyg for theyzable

ition ofed in the

.

the effects of climate when effective temperature sum was used as the temperature variable in the model. The overdecomposition was predicted correctly when mean annual temperature was used as the temperature variable, but thensubstantially overestimated climatic effects.

The model correctly predicted differences in decomposition rates among litter types in the early years of decomposunderestimated them in later years. The decomposition rate of the litter type richest in phenolic compounds (larchwas systematically overestimated, and that of the litter type richest inO-alkyl compounds (grass leaves) was systematicunderestimated. Accounting for these factors would improve the general applicability of the model. However, accountininitial nitrogen concentration of litter did not improve the accuracy of the model unless the initial lignin (i.e., acid unhydrolresidue) content was also taken into account.

We conclude that the model Yasso accounts for most of the effects of climate and initial litter quality on the decomposa range of foliage litter types under varying climate conditions. Recalibration of the reference decomposition rates usmodel may improve the accuracy when applying the model outside of Europe.© 2005 Elsevier B.V. All rights reserved.

Keywords: Dynamic model; Yasso; Decomposition; Litterbag; Soil carbon; CIDET

∗ Corresponding author at: European Forest Institute, c/o Suitia Research Farm, University of Helsinki, FIN-02570 Siuntio, FinlandTel.: +358 9 8190 8515; fax: +358 9 8190 8528.

E-mail address: [email protected] (T. Palosuo).

0304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolmodel.2005.03.006

Page 2: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

184 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

1. Introduction

The sizes of soil carbon pools and rates of organicmatter decomposition are crucial components of theglobal carbon budget (Prentice et al., 2001). Measuringdecomposition rate of organic matter and carbon accu-mulation in soils is difficult, and models are widely usedto estimate these parameters (Powlson et al., 1996).Evaluating the accuracy of such models is thereforeessential if they are to be used to reliably describe soilcarbon stocks and their dynamics at present and in thefuture.

The dynamic model Yasso was developed to be asimple but widely applicable soil carbon model requir-ing a limited amount of input data (Liski et al., 2005).The model has already been applied as a soil carbonmodule in the CO2FIX model, which is a general modelfor estimating carbon balance and carbon sequestrationcapacity in forest stands and landscapes (Schelhaas andNabuurs, 2001; Masera et al., 2003), in forest stand sim-ulator MOTTI (Hynynen et al., 2005) and in the forestresource projection model EFISCEN (Karjalainen etal., 2002).

Yasso takes into account differences in initial litterquality and the effect of climate on decomposition pro-cesses. The effect of climate is described using equa-tions that include temperature and summer drought asexplanatory variables (Liski et al., 2003, 2005). Eithermean annual temperature (MAT) or the effective tem-perature sum over 0◦C threshold (DD0) can be useda to beaa dilya merd cip-i toS int dif-f

lessd am-p tai note iledm thans rgerm ege-

tation model LPJ (Sitch et al., 2003), Canadian forestsector carbon budget model (Kurz and Apps, 1999) or astand-level forest and wood products model GORGAM(Schlamadinger and Marland, 1996). As Yasso is an in-dependent model, its applicability is not limited to anylarger modelling framework but it may be used withany other modelling system that calculates estimatesof litter production. In structure, Yasso is partly similarto models DocMod (Currie and Aber, 1997) and GEN-DEC (Moorhead and Reynolds, 1991); in these modelslitter entering soil is divided into explicit groups ofchemical compounds. The CENTURY and the RothCmodels are different in this respect as in them this divi-sion of litter depends on chemical indicators related toits decomposability. In summary, the low requirementsof input data, the general structure and the easiness toapply it for different research purposes make Yasso anattractive alternative among soil carbon models. How-ever, as there are a number of factors that Yasso doesnot take into account that also affect decompositionprocesses, such as initial nitrogen and phosphorus con-tents of litter, microclimatic variation created by standstructure, and site factors such as soil chemistry, soiltexture or organisms (Berg and McClaugherty, 2003),it is important to test the validity of this model and toidentify the most important possibilities to improve itsaccuracy.

A number of long-term litterbag experiments haverecently been established in which different litter typesare incubated in the field under a wide range of siteaW a-t tingm ctso Inp po-sW ofY to ac opew pa-r

ETd rbonm peso terc nceo ture

s the temperature variable. DD0 has been foundn effective predictor of decomposition rate (Liski etl., 2003), whereas MAT data are usually more reavailable for different study sites and areas. Sumrought is taken as the difference between the pre

tation and potential evapotranspiration from Mayeptember. The initial litter quality is considered

erms of physical size of the litter and content oferent carbon compounds in the litter.

Compared to other soil carbon models, Yasso isetailed and requires less input data than, for exle, the two widely used models CENTURY (Parton el., 1987) or RothC (Coleman and Jenkinson, 1996). It

s, therefore, possible to use Yasso when there isnough input information to use these more detaodels. On the other hand, Yasso is more general

oil carbon modules developed specifically for laodelling systems such as the dynamic global v

nd climate conditions (e.g.,Trofymow and the CIDETorking Group, 1998). Data from these field incub

ions are particularly valuable for independently tesodels like Yasso for their ability to predict the effef climate and initial litter quality on decomposition.articular, data from the Canadian Intersite Decomition Experiment (CIDET) (Trofymow and the CIDETorking Group, 1998) allow for independent testing

asso under Canadian climatic conditions which,ertain extent, are similar to those of northern Eurhere the data used to derive the original model

ameter values were collected.The objective of this study was to use the CID

ata set to evaluate how accurately the soil caodel Yasso predicts the mass loss of different tyf forest litter over 6 years, based on initial lithemistry and climatic variables. The performaf the model was tested for two different tempera

Page 3: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 185

variables (mean annual temperature and effectivetemperature sum). The potential of additional initiallitter quality variables to enhance the performance ofYasso was also assessed.

2. Materials and methods

2.1. Model

Yasso (Liski et al., 2005) begins leaf litter decompo-sition by dividing initial leaf litter mass into three sep-arate compartments named extractives, celluloses andlignin-like compounds. These groups of compoundsare then decomposed at their own rates (ki) indepen-dent of their origin, losing the proportionki of theirmass over unit time. The decomposition rate of the ex-tractives is, however, different for coniferous and de-ciduous litter. A proportion of the decomposed mass(1− pi) leaves the soil as heterotrophic respiration,while the remainder (pi) forms increasingly recalcitrantcompounds. Extractives and celluloses are transformedto lignin-like compounds, lignin-like compounds aretransformed to resistant humus compounds, and resis-tant humus compounds are transformed to very resis-tant humus compounds.

The climatic dependencies of the decompositionrates are described with linear regression models us-ing the relative change in decomposition compared tor d us-i ;L atedu -o thet ughtv ira-t

k

w merdD vari-a -cβ ngei merd

The reference decomposition rateski0 were deter-mined with data from a litterbag experiment usingScots pine (Pinus sylvestris L.) needles and white birch(Betula pendula Roth.) leaves incubated at Jadraas,Sweden (Berg et al., 1991; Liski et al., 2005) and soilcarbon measurements at 26 Scots pine (P. sylvestris)sites along a 5300-year soil chronosequence in southernFinland (Liski et al., 1998, 2005) (Table 1 inLiski et al.,2005). Climatic conditions used for the reference sitewere annual mean temperature 3.3◦C, DD0 = 1903◦Cdays, summer drought variable−32 mm.

2.2. Test data

Data were collected using litterbags (20 cm× 20 cmbags of 0.25 mm× 0.5 mm mesh) containing 10 g sam-ples of 10 foliar litter types (for details seeTrofymowand the CIDET Working Group, 1998). Ten sets of lit-terbags were placed in each of four replicate plots on18 upland forest sites representing a range of forestedregions (temperate to subarctic) across Canada. Annualmass remaining values (means of four replicate plotsper site) of 10 foliar litters (Table 1) were measured atthe 18 upland forest sites over 6 years (Trofymow etal., 2002). The data were then compared with predictedvalues generated by Yasso.

The results of conventional elemental and proxi-mate analysis of each CIDET litter type inPrestonet al. (2000)were used to derive initial litter con-centration values for the three carbon compartments( r-s blef y-d

rmm pre-c ETs 8wm atew be-c thant ex-a blev imesb . Ina arn ET

eference values. The models have been determineng litterbag data from across Europe (Berg et al., 1993iski et al., 2003). Separate models have been cresing the effective temperature sum with a 0◦C threshld (DD0) or mean annual temperature (MAT) as

emperature variable, together with a summer droariable (precipitation minus potential evapotranspion between May and September):

i(T, D) = ki0(1 + β(T − T0) + γ(D − D0)) (1)

hereT and D are the temperature and the sumrought variables at study sites, respectively,T0 and0 are the temperature and the summer droughtbles at the reference site, andki0 is the reference deomposition rate for each model compartmenti. Theandγ parameters describe the proportional cha

n decomposition rates when temperature or sumrought variables change (Table 2 inLiski et al., 2005).

Table 1), where “extractives” = nonpolar + wateoluble extractives, “celluloses” = acid hydrolyzaraction, and “lignin-like compounds” = acid unhrolyzable residue.

Thirty-year (1951–1980) climate normal (long-teeans) data (i.e., mean monthly temperatures and

ipitation) for the climate stations nearest the CIDites (Trofymow and the CIDET Working Group, 199)ere used as model climate input (Table 2), whicheans that the effect of yearly variation in climas not taken into account. This was reasonableause the variation among the sites was greaterhe variation among the years within one site, formple the deviation of MAT and the drought variaalues among the sites were about five and two tigger, respectively, than the deviation within sitesddition,Trofymow et al. (2002)noted that the 30-yeormals and 6-year actual climatic data for the CID

Page 4: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

186 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

Table 1Initial chemical concentrations of litters used as model input (mg g−1 of total dry litter mass; afterTrofymow and the CIDET Working Group,1998)

Litter type Initial chemical concentrations (mg g−1)

Extractivesa Cellulosesb Ligninsc Ashd

Coniferous treesBlack spruce (Picea mariana (Mill) B.S.P.) 308 370 283 42Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) 218 416 303 67Jack pine (Pinus banksiana Lamb.) 222 424 328 27Tamarack (Larix laricina (Du Roi) K. Koch) 404 301 240 59Western redcedar (Thuja plicata Donn) 212 365 356 72

Deciduous treesAmerican beech (Fagus grandifolia Ehrh.) 201 453 280 71Trembling aspen (Populus tremuloides Michx.) 441 337 144 84White birch (Betula papyrifera Marsh.) 424 303 240 34

OtherBracken fern (Pteridium aquilinum (L.) Kuhn) 113 491 329 72Plains rough fescue (Festuca hallii (Vasey) Piper) 220 585 112 92

a Sum of non-polar extractable and water-soluble extractable compounds.b Acid hydrolyzable fraction.c Acid unhydrolyzable residues.d The initial amount of ash was added to the sum of mass remaining values of carbon compounds to give measured mass remaining values.

Table 2Climate data for CIDET sites used as model input (afterTrofymow and the CIDET Working Group, 1998)

Name CIDET sites Climate variables

Province Latitude Longitude MAT (◦C)a DD0 (◦C days)b Ppt. (mm)c PET (mm)d D (mm)e

Chapleau Ont. 47◦38′ 83◦14′ 1.1 2133 444 362 0Gander Nfld 48◦55′ 54◦34′ 4.3 2120 398 346 0Gillam Man. 56◦19′ 94◦51′ −5.2 1449 270 313 −43Hidden Lake BC 50◦33′ 118◦50′ 6.3 2676 248 383 −135Inuvik NWT 68◦19′ 133◦32′ −9.8 1104 142 267 −125Kananaskis Alta. 51◦00′ 115◦00′ 2.8 1843 367 332 0Montmorency Que. 47◦19′ 71◦08′ 0.6 1778 726 339 0Morgan Arboretum Que. 45◦25′ 73◦57′ 6.1 3104 384 426 −42Nelson House Man. 55◦55′ 98◦37′ −3.9 1613 330 330 0Petawawa Ont. 45◦55′ 77◦35′ 4.3 2697 395 397 −2Port McNeill BC 50◦36′ 127◦20′ 7.9 2900 397 324 0Prince Albert Sask. 53◦13′ 105◦58′ 0.1 2184 265 369 −104Rocky Harbour Nfld 49◦32′ 57◦50′ 4.2 2112 457 343 0Schefferville Que. 54◦52′ 66◦39′ −4.8 1145 402 289 0Shawnigan Lake BC 48◦38′ 123◦42′ 9.3 3386 170 369 −199Termundee Sask. 51◦50′ 104◦55′ 1.8 2446 223 386 −163Topley BC 54◦36′ 126◦18′ 2.5 1873 223 338 −115Whitehorse Yukon 60◦51′ 135◦12′ −1.2 1631 146 333 −187

a Mean annual temperature.b Effective temperature sum with 0◦C threshold calculated from mean monthly temperatures.c Precipitation from May to September.d Potential evapotranspiration from May to September, calculated from mean monthly temperatures using the Thornthwaite method.e Summer drought variable, Ppt− PET if <0.

Page 5: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 187

study sites during the period used in this study werehighly correlated without any systematic differencesbetween them (annual temperature,r2 = 0.98; annualprecipitationr2 = 0.99).

Effective temperature sums (0◦C threshold) werecalculated from mean monthly temperatures by assum-ing that the mean temperatures occurred in the middleof each month, and mean daily values were linearly in-terpolated from these. This is the same procedure thatwas used to calculate the effective temperature sumsfor the empirical models describing the climatic depen-dencies in Yasso (Liski et al., 2003). Accumulated po-tential evapotranspiration between May and Septemberwas calculated from mean monthly temperatures usingthe approximation developed byPalmer and Havens(1958) of the Thornthwaite method (Thornthwaite,1948). The summer drought variable was delimited aszero for the eight sites with positive summer droughtvariable values because it was assumed that there wasno drought effect at these sites.

Because of the lack of radiation data, we could notuse the Priestley–Taylor equation to calculate potentialevapotranspiration at the sites in Canada although itwas used when Yasso was calibrated in Europe (Liskiet al., 2003). While the Thornthwaite method is lessaccurate, it provides reasonable estimates, which areeasy to calculate. Those who apply Yasso for practicalapplications as a part of CO2FIX (Masera et al., 2003),EFISCEN (Karjalainen et al., 2002) or MOTTI models(Hynynen et al., 2005) may often face the same prob-l ul toe ns.

2

cu-r imei tom eret tterq

2n-

t m-p a( fort ta.

Effective temperature sum (DD0) was always used asthe temperature variable, except when it was comparedwith runs using mean annual temperature (MAT).

Yasso calculated the course of decomposition in an-nual time steps. Values of litter mass remaining at theend of each year over the 6-year study period were cal-culated by summing the carbon in the different modelcompartments, and then adding the initial ash content,assuming that the ash content did not decompose orleach out of the litter.

2.3.2. Applicability of European calibrationvalues for Canada

As the climate dependency of Yasso is modelled aschanges in decomposition rates relative to referenceconditions, we first tested whether Yasso predictedlitter mass remaining over time accurately underconditions similar to the model calibration conditionsin Europe. For this comparison we chose CIDETsites with 30-year normal temperature and summerdrought conditions that were closest to those atJadraas (DD0 = 1903◦C days, D =−32 mm). Thesewere Topley (DD0 = 1873◦C days, D =−115 mm)and Kananaskis (DD0 = 1843◦C, D = 0 mm). Wethen compared model-predicted and measured massremaining values (the annual averages and ranges of10 litter types) at these two sites over a 6-year period.

2.3.3. Effect of climate on decompositionThe accuracy of predicting the effect of climate on

d thelt nualm . Thes ea-s ictedv e tos ea-s dento pre-d atet e theo

t per-a delp nceo he

em. The present results are thus particularly usefvaluate the accuracy of Yasso in these applicatio

.3. Testing the model

Data from CIDET were first used to test how acately Yasso predicted litter mass remaining over tn the two study sites with climatic conditions similar

odel calibration conditions in Europe. The data when used to test the effects of climate and initial liuality on model performance.

.3.1. Model runsFor each model run, initial litter quality (i.e., conce

rations of extractives, celluloses and lignin-like coounds in the 10 litter types;Table 1) and climatic dati.e., temperature and summer drought variableshe 18 study sites;Table 2) were used as input da

ecomposition in Yasso was quantified by fittinginear regressionMRmeasured = a + b × MRpredicted be-ween the model-predicted and the measured anean mass remaining values at CIDET study sites

lopes (b) of these regressions (i.e., change in mured values compared to changes in model-predalues in response to changing climate from sitite) show whether the model-predicted and the mured rates of decomposition were similarly depenn climate. Slope values <1 indicate that modelictions were more sensitive to the effects of clim

han was observed, and slope values >1 indicatpposite.

Climate dependencies (Eq.(1)) with the effec-ive temperature sum (DD0) or mean annual temture (MAT) were tested for their effect on moerformance. When testing the climatic dependef the model, the CIDET Inuvik study site in t

Page 6: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

188 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

Arctic was excluded because the MAT in Inuvikwas so low that the model calculated a negativedecomposition rate. The corresponding regression(MRmeasured = a + b × MRpredicted) with DD0 as a tem-perature variable, and with all 18 study sites included,was determined for each litter type separately and theregression slopes (b) were compared to determine howwell the model predicted the effect of climate for indi-vidual litter types, and to determine if there were inter-actions between climate and litter quality.

It was hypothesized that the difference betweenmacroclimatic variables used by the model and in situclimate caused some of the remaining variation in theregressions of measured and model-predicted annualmass remaining values. This hypothesis was tested byselecting specific stand characteristics (stand density,basal area, or their combination;Trofymow and theCIDET Working Group, 1998) and relating these tothe average residual variation in model predictions forthe study sites.

2.3.4. Effect of litter quality on decompositionThe effect of initial litter quality on the accuracy of

model predictions was tested by examining the model’sability to rank litter types by their mass remaining(Spearman’s rank correlation) and by analysing the de-viation of the litter types from the average mass re-maining at the sites over the 6-year study period. DD0was used as a temperature variable and all 18 sites wereincluded in all litter quality tests.

byu sid-u tert ain-i yearo tryv e thei en,p ium)aa C,m cC n-13n ross-pN nm ex-a

3. Results

3.1. Applicability of European calibration valuesfor Canada

The model-predicted values of annual average massremaining for the 10 CIDET litter types at Topleywere near the measured values (Fig. 1a). The model

Fig. 1. Litter mass remaining over time at (a) Topley, and (b)Kananaskis (mean and range of 10 litter types, grey area = therange of model-predicted values, white area between solid blacklines = the range of measured values, grey dashed line = the average ofmodel-predicted values, black dashed line = the average of measuredvalues).

To test if model accuracy could be increasedsing additional litter chemistry variables, the real variation in the model predictions (i.e., the lit

ype-specific deviations of the predicted mass remng values from the measured ones after the firstf decomposition) was compared with litter chemisariables not used in Yasso. These variables wernitial litter element concentrations (carbon, nitroghosphorus, sulphur, calcium, magnesium, potassnd initial lignin-to-nitrogen ratio (fromTrofymow etl., 1995), and organic C chemistry attributes (alkylethoxyl C,O-alkyl, di-O-alkyl, aromatic C, phenoliand carboxyl/carbonyl C) measured using carbo

uclear magnetic resonance spectroscopy with colarisation and magic-angle spinning (13C CPMASMR) (Preston et al., 2000). Correlations betweeodel residuals and litter characteristics were thenmined.

Page 7: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 189

underestimated mean mass remaining at Kananaskis(Fig. 1b), where underestimations in the first 2 yearsled to an underestimation over the whole 6-year pe-riod. The differences in mass remaining values amonglitter types were underestimated at both sites over allyears, the range predicted being 25–55% of the rangemeasured. Average measured mass remaining at thetwo CIDET sites were within 5% of each other overthe first 3 years of the study, but the later mass re-maining values at Topley were slightly lower thanat Kananaskis, indicating faster decomposition ratesat Topley. The model-predicted mass remaining val-ues were, on average, 30% higher at Topley than atKananaskis.

Among the litter types at the Topley site, the model-predicted mass remaining values of three conifer lit-ters (black spruce (Picea mariana), Douglas-fir (Pseu-dotsuga menziesii) and jack pine (Pinus banksiana))and one other litter (white birch (Betula papyrifera))were very similar to measured values (Fig. 2). Themodel-predicted mass remaining values of the twoother conifer litters (tamarack (Larix laricina) andwestern redcedar (Thuja plicata)) and two other litters(American beech (Fagus grandifolia) and tremblingaspen (Populus tremuloides)) were clearly underesti-mated. Plains rough fescue (a grass) (Festuca hallii)was the only litter type for which model-predicted massremaining was clearly overestimated in the early yearsof the study. However, fescue mass remaining was pre-dicted correctly at years 5 and 6. The differences be-t iningv r tot rent(

3

asu ls cesa didnT dif-f massr ATw sitesw ain-i lues,

and underestimated for the lowest mass values), al-though the average mass remaining was predictedcorrectly.

The results for all the individual litter types, otherthan fescue, followed the same pattern as the average oflitters shown inFig. 3(data not shown). There were nosystematic differences between the litters of conifersand non-conifers. The dependence of the decompo-sition of fescue litter on climate was overestimatedthroughout the study period.

We could not relate the residual variation (differ-ences between predicted and measured average littermass remaining values at each study site) to stand char-acteristics that could affect microclimate at the studysites. Over the 6-year study period,r2-values betweenstand density, basal area, and their combination, andresidual variation were low (−0.17 to 0.01, 0.16 to 0.38and−0.13 to 0.06, respectively).

3.3. Effect of litter quality on decomposition

The model correctly (prs < 0.05) predicted the orderof the different litter types at 17 of the 18 sites whenranked by mass remaining at the end of the first year(Table 3). During subsequent years, however, the rank-ing of litter types was less accurately predicted, untilrankings were correctly predicted at only 3 sites at theend of the fifth year.

The standard deviation of predicted values amonglitter types was two-thirds that of the standard deviationo . Them s re-m tioni dv as nos t thee ctedv tiono

-fir,w fern( ert rcha thatb ratesM twol ack

ween measured and model-predicted mass remaalues of different litters at Kananaskis were similahose at Topley, and only the overall level was diffedata not shown).

.2. Effect of climate on decomposition

When the effective temperature sum (DD0) wsed as a temperature variable (Eq.(1)), the modetatistically significantly overestimated the differenmong the sites in year 1, but after this the slopesot deviate statistically significantly from 1 (Fig. 3).here was, however, a tendency to underestimate

erences among sites and underestimate litteremaining as decomposition proceeded. When Mas used in the model, the differences amongere overestimated in all years (i.e., mass rem

ng was overestimated for the highest mass va

f the measured values at the end of the first yearodel therefore underestimated differences in masaining among litter types, and this underestima

ncreased with time (Table 3). Variation in the predictealues decreased in subsequent years, but there wuch trend in the variation of measured values. And of the sixth year, the standard deviation of predialues was only one-third that of the standard deviaf measured values.

The model correctly predicted that Douglasestern redcedar, American beech and bracken

Pteridium aquilinum) litter would decompose slowhan the average for all litter types, that white bind trembling aspen would decompose faster, andlack spruce and jack pine would decompose at aimilar to that for the average of all litters (Fig. 4).odelled and measured data clearly deviated for

itter types, with the model overestimating the tamar

Page 8: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

190T.Palosuo

etal./EcologicalM

odelling189

(2005)183–198

Fig. 2. Percent of original litter mass remaining of 10 litt types at Topley (open dots = measured values, filled dots = model-predicted values; bars indicate±S.D.;n = 4).

er
Page 9: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 191

Fig. 3. Measured over model-predicted litter mass remaining (mean of 10 litter types) at 17 study sites after 1, 3 and 6 years using (a) effectivetemperature sum with 0◦C threshold (DD0,◦C days) or (b) mean annual temperature (MAT,◦C) as the temperature variable in the model. Pointsabove 1:1 line indicate that model underestimated mass remaining; points below 1:1 line indicate that model overestimated mass remaining(Inuvik site excluded because the low subarctic temperatures resulted in negative decomposition rates when MAT was the temperature variable).

Table 3Statistics related to the effect of litter quality on modelled and measured mass remaining

Year 1 Year 2 Year 3 Year 4 Year 5 Year 6

Spearman’s rank correlation coefficientsAverage of all sites 0.66 0.63 0.58 0.50 0.41 0.45Minimum 0.43 0.32 0.38 0.16 0.13 0.15Maximum 0.76 0.80 0.71 0.70 0.71 0.66Number of sites withprs < 0.05 17 13 12 9 3 7

Average standard deviation (% original mass remaining)Model-predicted 6.4 5.5 4.6 4.0 3.5 3.1Measured 9.5 10.9 11.4 10.7 9.5 9.3

Spearman’s rank correlation coefficients between the model-predicted and the measured mass remaining values of the litter types at 18 CIDETstudy sites, and average standard deviation of model-predicted and measured mass remaining among litter types at 18 CIDET sites.

Page 10: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

192T.Palosuo

etal./EcologicalM

odelling189

(2005)183–198

Fig. 4. Deviations in measured and model-predicted mass remaining of each of the 10 litter types from the average litter mass remaining in years 1, 3, and 6. Bars indicate the rangein deviation in mass remaining of each litter type across the 18 sites. Points above or below the 1:1 line indicate the model, respectively, under or over estimated mass remaining.Points in the upper right or lower left quadrant indicate that both the measured and model-predicted litter type decomposed, respectively, faster orslower than the average litter.

Page 11: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 193

decomposition and underestimating the fescue decom-position.

Differences between the model-predicted and mea-sured mass remaining values at the end of the first year(i.e., model residuals) were not correlated with any ofthe litter element concentrations (carbon, phosphorus,sulphur, calcium, magnesium, potassium) (p > 0.05,data not shown), nor initial nitrogen concentrations(Fig. 5a). However, there was a negative correlation

Frfr

(r =−0.65,p = 0.042) between the residuals and initiallignin-to-nitrogen ratio, with the model overestimatingthe mass remaining of litter types with the highestlignin-to-nitrogen ratio, and underestimating the massremaining of litter types with the lowest lignin-to-nitrogen ratio (Fig. 5b). In addition, model residualswere strongly correlated (r =−0.88, p = 0.001) withinitial concentrations of phenolic C in litter: the modeloverestimated the mass remaining of litter types

ig. 5. Deviation of measured and model-predicted percent of origesiduals = modelled–measured) plotted over initial chemical charactor (a) nitrogen (N), (b) lignin-to-nitrogen ratio (KLIG/N), (c) phenolepresents one of the 10 different litter types, with percent of original

inal litter mass remaining after the first year of decomposition (modeleristics of litter (mg g−1 of total dry litter mass; afterPreston et al., 2000)ic carbon (PHEN), and (d)O-alkyl carbon (O-ALKYL). Each pointmass remaining at 1 year averaged across all 18 study sites.

Page 12: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

194 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

with the highest initial concentration of phenoliccompounds and underestimated the mass remainingof litter types with the lowest initial concentrationof phenolic C (Fig. 5c). Initial litter concentrationsof O-alkyl carbon were also well correlated withmodel residuals (r = 0.72, p = 0.018): the modelunderestimated the mass remaining of litter types withhigh initial O-alkyl concentrations, and overestimatedthe mass remaining of litter types with low initialO-alkyl concentrations (Fig. 5d). There were nosignificant correlations with other initial organicC chemical properties of litters (p > 0.05, data notshown).

4. Discussion

4.1. Short-term decomposition

The soil carbon model Yasso describes the dynamicprocess of organic matter decomposition and the resul-tant accumulation of organic matter and carbon in soilunder different climatic conditions for different littertypes. The data used in the current test enabled us totest the ability of the model to predict short-term litterdecomposition using three fast decomposition modelcompartments (i.e., extractives, cellulose and lignin-like compounds). The very slow decomposition of thehumus compartments in the model could not be testedbecause of the relatively short 6-year test period. Thea thes s thed ttera bons

4f

sim-i atJ parew ropeg suredm pleya edt atedt

We think these results indicate that the modeloverestimated, on the one hand, the overall rate ofdecomposition and, on the other hand, the sensitivityof decomposition to summer drought. At Topley, themodel-calculated rate of decomposition was reducedto the correct level by a more severe summer droughtthan at the calibration site Jadraas. At Kananaskis,there was no summer drought and the conditions werethus closer to the conditions that prevailed at Jadraasbut the model overestimated the rate of mass loss.When testing the method of modelling the effects ofclimate alone,Liski et al. (2003)found also that themodel tends to overestimate the effects of droughtin Canada although the effects were not statisticallysignificantly different between Canada and Europe.

The overestimation of the decomposition ratesin Canada could be the result of the larger litterbagmesh openings at Jadraas (1 mm× 1 mm (Berg etal., 1991, 1993) versus 0.25 mm× 0.5 mm in CIDET(Trofymow and the CIDET Working Group, 1998)).This may have led to greater mass loss because ofenhanced faunal impacts on decomposition, modifiedmicroclimate and increased leaching (cf.Edwards andHeath, 1963; Seastedt et al., 1983; Tian et al., 1992;Bradford et al., 2002). However, studies of mesh sizeat one of the CIDET sites (Shawnigan) demonstratedno effect of mesh size (CIDET mesh size versus 5 mm)on decay rates or on soil organism communities inlitterbags with wood chips (Trofymow, 1998; Addisonet al., 2003a,b) or mixed litter (Setala et al., 1996).

de-c entsd tudys haved hicha can-n orthyo

htlyo t be-t sure-m odelw lityu udeda uldh suit-a them

ccurate description of short-term decomposition inoil carbon models is relevant because it determineecomposition of the majority of the mass of fresh lind will therefore affect annual changes in soil cartocks.

.2. Applicability of European calibration valuesor Canada

The Topley and Kananaskis sites had the mostlar climate to the model calibration conditionsadraas in Sweden, and so they were used to comhether the decomposition rates determined in Euave reasonable estimates in Canada. The meaass remaining values were rather similar at Tond Kananaskis (Fig. 1a and b). The model reproduc

hese values accurately at Topley but underestimhem at Kananaskis (Fig. 1b).

Another reason for the overestimated rates ofomposition may be that the calibration measuremid not represent the conditions at the Canadian sites correctly. Examples of such factors that mayeviated are soil chemistry, texture or organisms, wre not accounted for in the model. These factorsot be tested using the present data sets but are wf further study.

Despite the level of decomposition that was sligverestimated in Canada, we concluded that the fiween the model-calculated estimates and the meaents was adequate for continuing to test the mith respect to the effects of climate and litter quasing the European parameter values. We concllso that recalibrating the model using local data woave improved its accuracy in Canada. However,ble data were not available for us to recalibrateodel in this study.

Page 13: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 195

4.3. Effect of climate on decomposition

The effect of climate on model-predicted decom-position rates depended on the temperature variableused in the model. When the effective temperature sum(DD0) was used, model predictions were less biased,although the differences in mass remaining betweenthe sites were overestimated in early years and un-derestimated in later years, and, in addition, overallmass remaining was underestimated. Using mean an-nual temperature (MAT) overestimated differences be-tween the sites probably because of the greater differ-ences in MAT among sites in Canada than in Europebecause of colder Canadian winters.

If, as appears to be the case, reference decompo-sition rates (ki0) determined on European sites over-estimate decomposition rates on Canadian sites withsimilar climate, then this will affect the climate depen-dence as well (see Eq.(1)). To evaluate this effect, wemade some tentative model simulations and decreasedthe reference decomposition rates by 30%. This de-creased the bias of estimating the effect of climate. Theslopes of regression lines fitted to the model-predictedand the measured mass remaining values (Fig. 3, tem-perature variable DD0) changed in the first year from0.56 to 0.73, and in the sixth year from 1.27 to 1.19.

We also tested the effect of increasing the climate de-pendence of decomposition of lignin-like compounds,which has more effect on mass remaining values duringlater years; it was especially during these years whent de-c y as siblyb rredi iod.

lit-t : dif-f andl po-s ored overt e de-c hanp delw ccu-r them thed ally

well and that there did not seem to be any interactionsbetween the litter quality and model bias related to cli-matic dependency among them.

The regressions describing the climatic dependen-cies of the model were created with macroclimatic vari-ables (Liski et al., 2003) that have been found to bethe dominant rate-regulating factors of decomposition(Meentemeyer, 1978; Johansson et al., 1995). We testedwhether or not there was any consistent difference be-tween the macroclimatic variables used and the actualmicroclimate at sites. However, the stand characteris-tics we chose (i.e., stand density, basal area, or theircombination), assuming they affected microclimate atthe sites, did not explain the differences between themodel-predicted and measured mass remaining values.

Although a wide range of climatic conditions in dataenabled extensive testing of the climatic influence ondecomposition in the model, inter-annual variation inclimatic conditions were not taken into account. Weused 30-year climate normal data for each site forthe whole study period, since the climatic equationsof the model were originally based on similar long-term normal data (Liski et al., 2003). In the decompo-sition data, however, yearly variation in decompositioncould clearly be seen (e.g.,Fig. 2). We therefore testedthe applicability of the current model to describe theclimate-driven inter-annual variation in decompositionusing actual monthly climate data measured during theCIDET study period (1992–1998). The results obtainedwere very close to those calculated with normal data( ea-s t thec atis-f cli-m ated porale l ist

4

assr buttp rentd ac-c latede ility

he model underestimated the effects of climate onomposition. This modification had, however, onlmall effect on the slopes of regression lines, posecause quite a small proportion of mass loss occu

n lignin-like compounds over the 6-year study perThe model bias of the climatic dependency of all

er types except fescue followed the same patternerences among sites were initially overestimatedater differences were underestimated. The decomition of fescue was consistently predicted to be mependent on climate than was actually observed

he 6-year study period. This was because fescuomposed quite rapidly in the field, and faster tredicted at all CIDET sites. It seems that the moas not able to estimate decomposition of fescue a

ately. As there were no major differences betweenodel bias of other litter types, we conclude thatecomposition of the other litters was predicted equ

r2 = 0.96) and results of comparisons with the mured data did not change much. This implies thalimatic dependency used in the model may not sactorily describe the effect of annual variation in

ate on decomposition. The development of climependency covering both geographical and temffects of climate on decomposition for the mode

herefore needed.

.4. Effect of litter quality on decomposition

The model ranked the litters quite well by their memaining in the early stages of decomposition,his ability of the model decreased later (Table 3). Thishenomenon can partly be explained by the diffeevelopment of variability between the litter typesording to the measurements and the model-calcustimates. In the results of the model, this variab

Page 14: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

196 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

decreased over time whereas in the measurements itremained stable (Table 3; see alsoFig. 1 for Topleyand Kananaskis sites). There are reasons of concernwith both of these trends. On the one hand, the modelmay not allow enough variability between litter types,especially in the later stages of decomposition, whencarbon accumulates in the compartments of lignin-likeand humus compounds. On the other hand, the litterbag method may overestimate the variability becausethe data are affected by measurement errors and theseerrors may increase with time.

Comparing model performance for different littertypes (Fig. 4) revealed that there were two litters in par-ticular for which decomposition rates were incorrectlypredicted: tamarack (mass remaining consistently un-derestimated) and fescue (mass remaining consistentlyoverestimated). For the model runs, tamarack was con-sidered to be a deciduous litter type because we as-sumed that its deciduous needles would have more incommon with foliage from deciduous broadleaf speciesthan with evergreen conifer species. This meant thatthe decomposition rate of its extractive fraction wasassumed to be greater than for the other coniferousspecies (Table 1 inLiski et al., 2005). This assumptionwas supported by the similarity in the initial proxi-mate chemical fractions of tamarack and white birch(Table 1). However, these faster decomposition ratesof extractives decreased the mass remaining values oftamarack depending on the study site from 5 to 19% inthe first year, and from 1 to 12% in the sixth year, com-p hise n oft % int archa posi-t

s fors hadt on ofp adtW n3p re-d esel oci-a losef

carbon. In our study, the model overestimated the de-composition rates of litters with high concentrations ofphenolic compounds, but underestimated the decom-position rates of litters with high concentrations ofO-alkyl carbon.

The initial nitrogen concentration of the litter typeswas not taken into account by the model. As withinitial concentrations of other litter components, initialnitrogen was not correlated with model residualsafter the first year of decomposition (Fig. 5a). Initiallignin-to-nitrogen ratio, however, was correlated withthe residuals. The model underestimated the mass re-maining values for litters with high lignin-to-nitrogenratios (Fig. 5b). This indicates that if initial nitrogenwas included in the model, its effect on the early stagesof decomposition should be related to the amountof lignin-like compounds. As excluding the effect ofnitrogen caused the model to overestimate the decom-position of litters with high initial lignin-to-nitrogenratios, our findings support the notion that nitrogenlimits the early stages of decomposition of litters withhigh lignin content (Berg and Ekbohm, 1991). Theeffect of nitrogen on model predictions in later stagesof decomposition could not be studied with the 6 yearsof data that was available through CIDET.

4.5. Comparison of model results with regressionmodels

Ther2-values for the linear regressions between them ETd and0( re-g eachy tterqe s-s f thed beenc .66;n icht is isq dedt ac-t ana-d sionm t

ared to runs with slower decomposition rates. Txplains, on average, 70% of the underestimatioamarack mass remaining in the first year, and 10he sixth year. Therefore, we recommend treating ls a coniferous species when calculating decom

ion of its biomass with Yasso.Both tamarack and fescue had extreme value

ome initial chemical characteristics. Tamarackhe highest and fescue had the lowest concentratihenolic carbon (Fig. 5b), and, in addition, fescue h

he highest concentration ofO-alkyl carbon (Fig. 5c).hen comparing the effect of initial litter quality o

-year mass loss,Preston et al. (2000)found that bothhenolic andO-alkyl carbon were associated withuctions in decomposition rate. They attributed th

ow rates to the lignin and tannin components assted with phenolic carbon, and the resistant cellu

ractions of polysaccharides associated withO-alkyl

ass remaining predictions with Yasso and CIData (n = 180) for years 1, 3 and 6 were 0.32, 0.48.56, respectively. These were lower than ther2-values0.76, 0.74 and 0.71) for empirical three-variableression models estimating ln(% mass remaining)ear, which included variables describing both liuality and study site climate (Table 4 inTrofymowt al., 2002). Our r2-values increased with succeive years, mainly because the overall variance oata also increased over time. Values would haveloser to Trofymow’s and others’ (0.50, 0.58 and 0= 144) without tamarack and fescue litters, for wh

he model predictions were consistently biased. Thuite a satisfactory result, given that Yasso is inten

o be a simple model driven by a limited number of fors, and none of its parameters were fitted to this Cian data set in contrary to the empirical regresodels. Furthermore, comparing ther2-values does no

Page 15: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198 197

describe the real success of this type of model, becausethey do not indicate all important features such as thecorrect level of model predictions, climate dependencyand different litter qualities.

5. Conclusions

The model Yasso, which was developed with Eu-ropean data, predicted the decomposition of many dif-ferent leaf litters fairly accurately under subarctic totemperate climatic conditions in Canada. Using themodel parameters determined in Europe overestimateddecomposition rates in Canada. This may have beenthe result of differences in a number of factors in-cluding climate (especially summer drought), soil tex-ture, chemistry, soil organisms, and litterbag mesh size.Recalibration of the model may improve its accuracywhen applying the model outside of Europe.

Model performance under different climatic condi-tions depended on the variable selected to describe thedependence of decomposition rates on temperature inthe model. The effective temperature sum (DD0) was abetter temperature variable than the mean annual tem-perature (MAT), but there were still some systematicerrors in predictions of the effects of climate on de-composition. Decreasing the reference decompositionrates in the current tests would have improved modelperformance under different climatic conditions.

There are certain species for which the performanceo ther nces lingt on-c to-g alc cen-t keni

A

ent( ars , andp ysesr Tro-

fymow, C. Camire, L. Duchesne, J. Fyles, M. Krana-better, L. Kozak, T. Moore, I. Morrison, C. Prescott,M. Siltanen, S. Smith, B. Titus, S. Visser, R. Wein, D.White, L. Kutny, C. Preston, and A. Harris. CIDET isfunded by the Canadian Forest Service and in part by agrant from Panel on Energy Research and Development(PERD). Further information on CIDET is available athttp://www.pfc.cfs.nrcan.gc.ca/climate/cidet.

This study was partly funded by the European Com-mission through ATEAM (EVK2-2000-00075) andCASFOR-II (ICA4-CT-2001-10100) projects, and bythe Academy of Finland through the projects Cross-Disciplinary Approach to Determination of CarbonBalance of Forests (Project 50708) and IntegratedMethod to Estimate the Carbon Budget of Forests(Project 52767).

We thank Raija Laiho, Marcus Lindner and TimGreen for their comments on this paper.

References

Addison, J., Trofymow, J.A., Marshall, V.G., 2003a. Functional roleof Collembola in decomposition in coastal temperate rainforests.Appl. Soil Ecol. 24, 247–261.

Addison, J., Trofymow, J.A., Marshall, V.G., 2003b. Abundance,species diversity, and community structure of Collembola in suc-cessional coastal temperate forests on Vancouver Island, Canada.Appl. Soil Ecol. 24, 233–246.

Berg, B., Booltink, H., Breymeyer, A., Ewertsson, A., Gallardo, A.,Holm, B., Johansson, M.-B., Koivuoja, S., Meentemeyer, V., Ny-

f, H.,ition

nifer-edishand

B com-ong-. 69,

B nta,te-.V.,

sternual-

B Hu-erlin,

B ton,litter

f the current model was consistently biased, andeasons for these differences in model performahould be studied further. According to the modelests performed, incorporating the initial nitrogen centration of litter would only be effective if usedether with initial lignin content. Other litter chemicharacteristics, such as initial phenolic carbon conrations, may be more important and should be tanto account in future model development.

cknowledgements

The Canadian Intersite Decomposition ExperimCIDET) Working Group is cooperating in a 12-yetudy of litter decomposition rates across Canadarovided the mass remaining and chemical analesults reported in this paper. Members include: J.

man, P., Olofsson, J., Pettersson, A.-S., Reurslag, A., StaaStaaf, I. and Uba, L., 1991. Data on needle litter decomposand soil climate as well as site characteristics for some coous forest sites. Part II. Decomposition data. Report 42, SwUniversity of Agricultural Sciences, Department of EcologyEnvironmental Research, Uppsala.

erg, B., Ekbohm, G., 1991. Litter mass-loss rates and deposition patterns in some needle and leaf litter types. Lterm decomposition in a Scots pine forest, VII. Can. J. Bot1449–1456.

erg, B., Berg, M.P., Bottner, P., Box, E., Breymeyer, A., De AR.C., Couteaux, M., Malkonen, E., McClaugherty, C., Meenmeyer, V., Munoz, F., Piussi, P., Remacle, J., De Santo, A1993. Litter mass loss in pine forests of Europe and EaUnited States: some relationships with climate and litter qity. Biogeochemistry 20, 127–159.

erg, B., McClaugherty, C., 2003. Plant Litter. Decomposition,mus Formation, Carbon Sequestration. Springer-Verlag, B286 pp.

radford, M.A., Tordoff, G.M., Eggers, T., Jones, T.H., NewingJ.E., 2002. Microbiota, fauna, and mesh size interactions indecomposition. Oikos 99, 317–323.

Page 16: Litter decomposition affected by climate and litter quality—Testing the Yasso model with litterbag data from the Canadian intersite decomposition experiment

198 T. Palosuo et al. / Ecological Modelling 189 (2005) 183–198

Coleman, K., Jenkinson, D.S., 1996. RothC-26.3—a model for theturnover of carbon in soil. In: Powlson, D.S., Smith, P., Smith,J.U. (Eds.), Evaluation of Soil Organic Matter Models, UsingExisting Long-Term Datasets. Springer-Verlag, Heidelberg, pp.237–246.

Currie, W.S., Aber, J.D., 1997. Modeling leaching as a decompositionprocess in humid, montane forests. Ecology 78, 1844–1860.

Edwards, C.A., Heath, G.W., 1963. The role of soil animals in break-down of leaf material. In: Doeksen, J., van der Drift, J. (Eds.),Soil Organisms. North-Holland Publishing Co., Amsterdam, pp.76–84.

Hynynen, J., Ahtikoski, A., Siitonen, J., Sievanen, R., Liski, J., 2005.Applying the MOTTI simulator to analyze the effects of alterna-tive management schedules on timber and non-timber produc-tion. For. Ecol. Manage. 207, 5–18.

Johansson, M.-B., Berg, B., Meentemeyer, V., 1995. Litter mass-lossrates in late stages of decomposition in a transect of pine forests.Long-term decomposition in a Scots pine forest. IX. Can. J. Bot.73, 1509–1521.

Karjalainen, T., Pussinen, A., Liski, J., Nabuurs, G.-J., Erhard, M.,Eggers, T., Sonntag, M., Mohren, G.M.J., 2002. An approach to-wards an estimate of the impact of forest management and climatechange on the European forest sector carbon budget: Germanyas a case study. For. Ecol. Manage. 162, 87–103.

Kurz, W.A., Apps, J.M., 1999. A 70-year retrospective analysisof carbon fluxes in the Canadian forest sector. Ecol. Appl. 9,526–547.

LIDET, 1995. Meeting the Challenges of Long-Term, Broad-ScaleEcological Experiments. Report 19, US LTER Network Office,Seattle.

Liski, J., Ilvesniemi, H., Makela, A., Starr, M., 1998. Model analysisof the effects of soil age, fires and harvesting on the carbon storageof boreal forest soils. Eur. J. Soil Sci. 49, 407–416.

Liski, J., Nissinen, A., Erhard, M., Taskinen, O., 2003. Climatic ef-fects on litter decomposition from arctic tundra to tropical rain-

L ndel.

M n, T.,hren,tion,

V.2

M tter

M r de-del.

P de-on.

P lysisins

Powlson, D.S., Smith, P., Smith, J.U., 1996. Evaluation of Soil Or-ganic Matter Models. Springer-Verlag, Berlin, 429 pp.

Prentice, I.C., Farquhar, G.D., Fasham, M.J.R., Goulden, M.L.,Heimann, M., Jaramillo, V.J., Kheshgi, H.S., Le Quere, C., Sc-holes, R.J., Wallace, D.W.R., 2001. The carbon cycle and atmo-spheric carbon dioxide. In: Houghton, J.T., Ding, Y., Griggs, D.J.,Noguer, M., van der Linden, P.J., Dai, X., Maskell, K., Johnson,C.A. (Eds.), Climate Change 2001: The Scientific Basis. Con-tribution of Working Group I to the Third Assessment Reportof the Intergovernmental Panel on Climate Change. CambridgeUniversity Press, Cambridge, pp. 183–238.

Preston, C.M., Trofymow, J.A., and the CIDET Working Group,2000. Variability in litter quality and its relationship tolitter decay in Canadian forests. Can. J. Bot. 78, 1269–1287.

Schelhaas, M.J., Nabuurs G.J., 2001. CO2FIX at the LandscapeLevel—An Application for the Veluwe Area, the Netherlands.Alterra-rapport 301, Alterra, Green World Research, Wagenin-gen.

Schlamadinger, B., Marland, G., 1996. The role of forest and bioen-ergy strategies in the global carbon cycle. Biomass Bioenerg. 10,275–300.

Seastedt, T.R., Crossley, D.A., Meentemeyer, V., Waide, J.B., 1983.A two-year study of leaf litter decomposition in relation to macro-climatic factors and microarthropod abundance. Holarct. Ecol. 6,11–61.

Setala, H., Marshall, V.G., Trofymow, J.A., 1996. Influence ofbody size of soil fauna on litter decomposition and 15N up-take by poplar in a pot trial. Soil Biol. Biochem. 28, 1661–1675.

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, plant ge-ography and terrestrial carbon cycling in the LPJ dynamic globalvegetation model. Global Change Biol. 9, 161–185.

T ifica-

T po-fromistry

T alitydian

T vity53.

T na-and-X-

T , I.,,

itterlitter

forest. Global Change Biol. 9, 1–10.iski, J., Palosuo, T., Peltoniemi, M., Sievanen, R., 2005. Carbo

and decomposition model Yasso for forest soils. Ecol. Mo189, 168–182.

asera, O.R., Garza-Caligaris, J.F., Kanninen, M., KarjalaineLiski, J., Nabuurs, G.J., Pussinen, A., de Jong, B.H.J., MoG.M.J., 2003. Modeling carbon sequestration in afforestaagroforestry and forest management projects: the CO2FIXapproach. Ecol. Model. 164, 177–199.

eentemeyer, V., 1978. Macroclimate and lignin control of lidecomposition rates. Ecology 59, 465–472.

oorhead, D.L., Reynolds, J.F., 1991. A general model of littecomposition in the northern Chihuahuan Desert. Ecol. Mo56, 197–219.

almer, W.C., Havens, A.V., 1958. A graphical technique fortermining evapotranspiration by the Thornthwaite method. MWeather Rev. 86, 123–128.

arton, W.J., Schimel, D.S., Cole, C.V., Ojima, D.S., 1987. Anaof factors controlling soil organic matter levels in great plagrasslands. Soil Sci. Soc. Am. J. 51, 1173–1179.

hornthwaite, C.W., 1948. An approach toward a rational classtion of climate. Geogr. Rev. 38, 55–94.

ian, G., Kang, B.T., Brussard, L., 1992. Effects of chemical comsition on N, Ca and Mg release during incubation of leavesselected agroforestry and fallow plant species. Biogeochem16, 103–119.

rofymow, J.A., Preston, C.M., Prescott, C.E., 1995. Litter quand its potential effect on decay rates of materials from Canaforests. Water Air Soil Pollut. 82, 215–226.

rofymow, J.A., 1998. Detrital carbon fluxes and microbial actiin successional Douglas-fir forests. Northwest Sci. 72, 51–

rofymow, J.A. and the CIDET Working Group, 1998. The Cadian Intersite Decomposition Experiment (CIDET): ProjectSite Establishment Report. Report Information Report BC378, Pacific Forestry Centre, Victoria, Canada.

rofymow, J.A., Moore, T.R., Titus, B., Prescott, C.E., MorrisonSiltanen, M., Smith, S., Fyles, J., Wein, R., Camire, C., DuscheneL., Kozak, L., Kranabetter, M., Visser, S., 2002. Rates of ldecomposition over 6 years in Canadian forests: influence ofquality and climate. Can. J. For. Res. 32, 789–804.