ORIGINAL ARTICLE

Sinkyu Kang Æ Dowon Lee Æ Jangho Lee

Steven W. Running

Topographic and climatic controls on soil environments and netprimary production in a rugged temperate hardwood forest in Korea

Received: 19 April 2004 / Accepted: 16 June 2005 / Published online: 6 September 2005� The Ecological Society of Japan 2005

Abstract Eight years (1994–2001) of field data and abiogeochemical process model, BIOME-BGC, were usedto examine effects of local topography and inter-annualclimatic variability on soil physical (i.e., soil moistureand temperature) and biogeochemical (i.e., organicmatter content, soil respiration, and leaf litter produc-tion) variables in a temperate hardwood forest in Korea.The field data were collected from adjacent south-facing(S) and north-facing (N) slopes, respectively, to examineeffects of local topography, and were utilized to validatepredictability according to BIOME-BGC which wasapplied to model unmeasured hydro-ecological pro-cesses [i.e., evapotranspiration, net primary production(NPP), and net ecosystem exchange of carbon]. Ourfield-data analyses indicated that soil-related variablesincluding soil temperature, water content, organic mat-ter, soil respiration, and floor leaf litter store signifi-cantly differed between the S and N slopes, while leaflitter production did not differ as significantly as the soil-related variables. The BIOME-BGC predictions showedgood agreement with the mean field data aggregatedacross the slopes. Our simulation results and fieldobservations indicated that the inter-annual variationsof leaf litter production and maximum leaf area indexwere best explained by precipitation, both at a 1-yearlag, while variation in annual NPP was well correlatedwith precipitation without a temporal lag. Our results

imply that: (1) local topography needs to be explicitlyconsidered in ecosystem studies as a forcing functiongenerating spatial heterogeneity in soil physical andbiogeochemical variables within a rugged landscape, and(2) water limits vegetation productivity in our studyforest, in spite of a relatively high annual precipitationrate (1,579 mm year�1).

Keywords Climate variability Æ Topography Æ Primaryproduction Æ Temperate hardwood forest

Introduction

The temperate mixed hardwood forest biome is a majorecosystem component of the mid-latitude northernhemisphere. For the past decade, this biome, togetherwith the boreal forest biome, has been suggested as apotential terrestrial carbon sink (Kauppi et al. 1992;Wofsy et al. 1993; Fang et al. 2001; Pacala et al. 2001;Lee et al. 2002). One of the key concerns about theseforests in global carbon cycling research is an under-standing of seasonal and inter-annual variability offorest production, and particularly those processesassociated with climatic change (Keeling et al. 1995;Battle et al. 2000; Bousquet et al. 2000; Cavender-Baresand Bazzaz 2000).

Recent field evidence from deciduous forests in thecentral region of the Korean Peninsula indicated limita-tions of soil water availability on ecosystem carbon pro-cesses: e.g., for net primary production (NPP) ofvegetation (Eum et al. 2005) and for soil respiration rates(Kang et al. 2003). In contrast, Churkina and Running(1998) showed, in their modeling study on global vegeta-tion production, that East Asia including the KoreanPeninsula is a radiation-limited ecosystem for NPP.Similarly, Nemani et al. (2003) suggested, in their remotesensing and modeling study, that air temperature andsolar radiation increased in the Korean Peninsula andnorthernChina in the period of 1982–1999,which resulted

S. Kang (&)Department of Environmental Science,Kangwon National University,Chunchon, Kangwon-do, 200-701, Republic of KoreaE-mail: kangsk@kangwon.ac.krTel.: +82-33-2508578Fax: +82-33-2513991

D. Lee Æ J. LeeGraduate School of Environmental Studies,Seoul National University, Seoul, 151-742, Republic of Korea

S. W. RunningNumerical Teradynamic Simulation Group, School of Forestry,University of Montana, Missoula, MT 59812, USA

Ecol Res (2006) 21:64–74DOI 10.1007/s11284-005-0095-0

in NPP increases in this area. The inconsistency betweenthe results from field measurements and modeling studiesis partly due to an insufficient field sampling period and/orincorrect model parameterization applied to the studyarea. Hence, an integrative study using both long-termfield measurements and locally parameterized modelsimulation is necessary in order to understand primaryclimatic variables controlling the vegetation productivityof the temperate deciduous forest biome in East Asia.

In a recent study based on long-term forest inventorydata, both Lee et al. (2002) and Choi et al. (2002) re-ported large uptake rates of atmospheric carbon dioxideby forest plantations in Korea—a consequence of stronggovernmental efforts in the national reforestation cam-paign since the 1960s. A significant portion of theKorean landscape consists of temperate hardwood for-ests (65%, together with evergreen needleleaf forests)(Korean Forest Service 1997). Most forests are withinrugged montane areas, because of deforestation, highpopulation density, and intensive land use. The effects oflocal topography on microclimate and soil physical andbiogeochemical processes, therefore, need to be consid-ered explicitly for a better understanding of the rates andspatial distributions of forest production in Koreanforested landscapes.

Compared to flat landscapes, local topography inrugged forested landscapes often complicates soil-vege-tation-atmosphere (SVAT) interactions by generatingslope-driven spatial heterogeneities in diverse physicaland biogeochemical variables of the SVAT system. Forexample, local topography generates considerable spa-tial variability in soil temperature (Kang et al. 2000),incoming solar radiation (Running et al. 1987; Kanget al. 2002), soil water content (SWC) (Band et al. 1993;Kang et al. 2004), and soil respiration (Kang et al. 2003).Even within a small watershed, local topography affectssnow cover and airflow that alters forest floor leaf litterredistribution across watershed boundaries, resulting inleaf litter accumulation and high nutrient availability innorth-facing slopes (N) (Lee et al. 1999; Oh et al. 1999;Yoo et al. 2001). However, these studies have not pro-vided conclusive evidence on the role of topography inforest production, because data are often inconsistentconcerning forest production: e.g., lower solar radiationand higher soil fertility on N slopes are compared withhigher radiation and lower soil fertility on south-facing(S) slopes.

There has been a variety of previous research on thehardwood forest ecosystem. Burrows et al. (2003)

reported significant spatial patterns in wood, foliage,and understory NPP of deciduous forests in northernWisconsin. Bolstad et al. (2001) provided field evidenceindicating that both leaf area index (LAI) and above-ground NPP decreased significantly from low to highelevation in southern Appalachian deciduous forests.Similarly, several studies using the geographic informa-tion system showed that topographic indices derivedfrom local topographic features were useful indicators topredict site productivity (McNab 1993; Iverson et al.1997). Very few studies, however, directly measuredboth soil physical and biogeochemical variables andforest production within a rugged forested watershed atthe same time.

This study was designed for two purposes: first, toexamine the effects of local topography on soil physicaland biogeochemical variables by using 8 years of fieldobservations at a rugged temperate mixed-hardwoodforest; second, to investigate primary climatic variablescontrolling vegetation production by using both fieldmeasurements and model simulation. Stand-level bio-geochemical modeling was introduced for predictingecosystem variables related to forest carbon process notmeasured by our field data. The field data collected fromadjacent S and N slopes were utilized for both dataanalysis and model parameterization and validation,respectively.

Materials and methods

Site description and field measurements

The study site is located in a forested watershed of Mt.Jumbong (38�02¢N, 128�26¢E), Kangwon Province,Korea. Soils within the region are classified as sandyloam, as shown in Table 1. Vegetation is predominantlytemperate mixed-hardwood forest composed of Mon-golian oak (Quercus mongolica), Korean maple (Acerpseudo-sieboldianum), and heartleaf hornbeam (Carpi-nus cordata). Our sampling strategy was to have twosampling plots adjacent to each other on S and N slopes,respectively. Both slopes were similar with respect toslope reach lengths from ridges to valleys (�20 m), ele-vations (�1,100 m), overstory species, and basal area,but differed in understory species (Table 1). A moredetailed description of the site can be found in otherpapers (Lee et al. 1999).

Table 1 Site characteristics of the south-facing (S) and north-facing (N) plots. Value in parenthesis is SD. SOM Soil organic matter, TNtotal soil nitrogen, LAI leaf area index

Site Aspect Texture(%)

SOM(%)

TN(%)

Basalarea(m2 ha�1)

Leaf litterproduction(g m�2 year�1)

UnderstoryLAI(m2 m�2)

Sand Silt Clay

S SW 67 24 9 16 (4) 0.51 (0.02) 33.1 333 (53) 0.8 (0.3)N NE 60 31 9 17 (2) 0.57 (0.03) 31.8 360 (47) 1.2 (0.2)

65

Monthly field sampling was conducted from August1994 to December 2001. We collected data at multiplelocations within the slopes and measured leaf litterproduction, floor leaf litter store (FLS), soil tempera-ture, SWC, soil organic matter (SOM), and soil respi-ration. For observing leaf litter production, five andthree leaf litter traps with circular shape and surface areaof 1 m2 were installed at the S and N slopes, respectively.The traps were uniformly located from ridge to bottomof each slope. The collected leaf litter was dried at 75�Cfor 48 h and twig, fruit, and leaf material were eachweighed separately. At each sampling, five replicatemeasurements of soil temperature (�C), gravimetricwater content (%), organic matter content (%), and soilrespiration (g C m�2 h�1) were obtained along transectswithin each slope.

A fixed transect was established parallel to each studyslope and extended from ridge top to valley bottom.Sampling nodes were then designated at 5-m intervalsalong each transect. Measurement points were selectedaround each node, but at least 1 m away from sur-rounding overstory trunks. Measurements were con-ducted around each node on a monthly basis, but notalways at identical points, since destructive soil samplingmethods were employed following soil respiration mea-surements. By marking previous measurement points,we were able to collocate new measurement points onundisturbed locations adjacent to former samplingpoints.

Bulk mineral soil and root respiration rates weremeasured by removing surface litter, inserting a closedgas chamber (EGM2 soil respiration meter, PPSystem)to 1-cm soil depth, and recording carbon dioxide evo-lution for 2 min. Surface litter was removed to minimizeerrors associated with alteration of chamber air volume.Independently, the daily soil respiration rate wassimultaneously measured by using a soda-lime[NaOH+Ca(OH)2] method with litter layers undis-turbed (Londo et al. 1999). Five static chambers wereplaced along the transect in each slope, and five controlchambers were sealed to estimate the amount of ambientcarbon dioxide inside the chamber. Pre-dried soda lime(30 g) was inserted into each chamber and left for 24 h,and then the soda lime was dried again and reweighed inthe laboratory to measure the amount of absorbed car-bon dioxide (Londo et al. 1999).

Soil respiration measurements were conductedsimultaneously with soil temperature measurements at10-cm depth (�C). After soil respiration measurementshad been completed, the chamber was removed, and thearea of soil covered by the chamber was homogenized to10-cm soil depth. Approximately, 100 g (fresh) of thehomogenized soil was removed and transported to thelaboratory within 24 h for the following analyses.Subsamples of the collected soils were used to determinegravimetric SWC (% of g H2O g soil�1), SOM content(% of g OM g soil�1), total soil nitrogen (TN, %), etc.The SWC and organic matter content were determinedin triplicate for each soil sample. Soil texture was

determined by using the soil hydrometer method andwas classified by USDA criteria. Continuous soil tem-peratures at 10-cm depth were recorded hourly at themid-point of the S and N slopes, respectively, by usingtemperature data loggers (HOBO; Onset Computer,Bourne, Mass.). The LAI was measured by using a LI-COR 2000 plant canopy analyzer (LI-COR, Neb.).

The field variables measured in this study were notcollected in consistent time periods, because we selectedthe field variables according to funding requirementsbased on the 8-year study period (1994–2001). It should,therefore, be noted that this study analyzed a differentobservation period for each field variable: leaf litterproduction for 1994–2001; soil temperature for 1996–2000; SWC for 1996–2000; SOM for 1996–2000; soilrespiration for 1998–2000; floor leaf-litter store for1997–1999.

Meteorological data and site water balance

Meteorological data including air temperature and pre-cipitation (PRCP) were collected hourly from April 1998to December 2001 by using a weather monitoring system(Weatherlink II; Davis Instruments) at the study site.Because the observed meteorological data do not coverthe entire study period from 1994 to 2001, we used theobserved meteorological records to extrapolate dailymeteorological data for the earlier study period fromJanuary 1994 to March 1998, using nearby long-termrecords from National Weather Stations (NWS) assuggested by Daly et al. (2000). After examining thePearson correlation coefficients between our observeddaily records and the daily data from nearby NWS, twoNWS were finally chosen for the extrapolations: Dae-kwanryung NWS (r=0.99, P<0.01) for daily air tem-perature, and Sokcho NWS (r=0.93, P<0.01) fordailyPRCP. By using daily data from both NWS sites,linear regression models between our daily records andNWS records were developed and utilized to extrapolatedaily air temperature and PRCP from January 1994 toMarch 1998. The 8-year daily meteorological data(Table 2) were used as input for model simulation.

Site water balance (WB), that is the difference be-tween PRCP and potential evapotranspiration (PET), isa potential limiting factor in forest productivity, asillustrated by Churkina and Running (1998). We calcu-lated WB by using the Priestley–Taylor PET (Priestleyand Taylor, 1972) given by Federer et al. (1996) whichwas utilized as a site water availability index in this study(Eq. 1):

PET(mm/day) ¼ 1:26� DRn

Lvqw(Dþ c)ð1Þ

where Rn is daily net radiation above the surface(W m�2), Lv the latent heat of vaporization(2,448.0 MJ Mg�1), D the rate of change in vaporpressure with temperature (kPa K�1), q w the density of

66

water (1.0 mg m�3), and c is the psychrometer constant(0.067 kPa K�1).

BIOME-BGC simulation

In the current study, BIOME-BGC is used to predictdaily evapotranspiration (ET) and NPP on the S and Nslopes, to examine the effects of topography and climateon forest productivity. The BIOME-BGC (version 4.11)is a general ecosystem process model designed to simu-late flux and storage of water, carbon, and nitrogen forterrestrial biomes ranging from single plot to globalscales. Details of the model are presented elsewhere andinclude applications for multiple biome types and spatialscales (e.g., Thornton et al. 2002; White et al. 2000). Themodel is designed to realistically simulate soil-plantcarbon and nitrogen cycling, but with simplifyingassumptions to facilitate application at regional scales byusing a limited number (34) of biome-specific physio-logical constants. All plant, litter, and soil carbon,nitrogen, and water pools and fluxes are entirely prog-nostic. The plant/ecosystem surface is represented by asingle, homogeneous canopy, snow (when present) andsoil layers, where understory vegetation is not distin-guished from the aggregate canopy layer. The model alsouses a daily time-step with daily maximum and minimumair temperature and PRCP as the primary inputs fromwhich mean daily short-wave radiation, vapor pressuredeficit (VPD), and day/night average temperatures areestimated. Biophysical processes represented by themodel include: photosynthetic carbon fixation fromatmospheric carbon dioxide; nitrogen uptake from theatmosphere and soil; carbon/nitrogen allocation togrowing plant parts; seasonal phenology, decompositionof fresh plant litter and SOM; plant mortality, growth,litter production, decomposition, and disturbance (i.e.,fire and management); solar radiation interception andpartitioning into sunlit and shaded leaf fractions; rainfallrouting to leaves and soil; snow accumulation andmelting; drainage and runoff of soil water; evaporationof water from soil and wet leaves; and ET partitioninginto transpiration, snow, soil, and canopy evaporation

components. The BIOME-BGC model has been suc-cessfully applied and validated over a range of diversebiomes, spatial scales, and climate regimes (e.g., Whiteet al. 2000), including individual boreal forest stands andsub-regions within the BOREAS study region (Kimballet al. 1997; Amthor et al. 2001.

For BIOME-BGC simulations, daily meteorologicaldata were prepared by running a topoclimatic model,MT-CLIM version 4.3 (Running et al. 1987; http://www.ntsg.umt.edu/) for each S and N slope, respec-tively. The daily air temperature and PRCP prepared inthe earlier section were used as inputs of MT-CLIM tocalculate daily atmospheric VPD and solar radiation, allof which were then utilized as input meteorological datafor BIOME-BGC simulations. Among the meteorolog-ical variables, solar radiation is the only variable to havea slope-specific topographic effect, because slope aspectwas utilized to calculate the radiation, while elevationwas considered only in calculating other meteorologicalvariables (i.e., air temperature, PRCP, and VPD).Consequently, both slopes were differentiated on thebasis of slope-specific soil texture and solar radiation inour simulation.

The BIOME-BGC simulations were conducted byfirst running the model to steady-state ecosystem carbonand nitrogen pool conditions under continuous prepared8-year (1994–2001) daily meteorological data cycles andfixed atmospheric carbon dioxide concentration levels(360 ppm). The results of these model spin-up runs werethen used as the initial conditions for additional simu-lations that were conducted from 1994 to 2001. Thesemodeling procedures were repeated by tuning a param-eter of ‘‘fraction of leaf nitrogen content in Rubisco’’(0.12) within a range suggested by White et al. (2000),which resulted in minimal error between the predictedand measured maximum LAI for both slopes. Then, thecalibrated model simulations were validated by usingindependent measurements of soil physical and biogeo-chemical variables (i.e., soil temperature, SWC, SOM,soil respiration, floor leaf-litter store, and leaf litterproduction) from the S and N slopes, respectively.

We used ecophysiological parameters suggested byWhite et al. (2000) in this study, with the exception of

Table 2 Summary of inter-annual climatic variables at the Mt.Jumbong study site. Tavg Annual mean air temperature, Tmax an-nual mean daily maximum air temperature, Tmin annual mean dailyminimum air temperature, PRCP annual precipitation, RAD mean

daily solar radiation, WB annual water balance, WBs summerwater balance from June to August after considering runoff inBIOME-BGC

Year Tavg (�C) Tmax (�C) Tmin (�C) RAD (MJ m�2 day�1) PRCP (mm year�1) WB (mm year�1) WBs (mm t�1)

1993 5.9 12.7 0.9 14.2 1,630 726 3271994 6.4 11.7 1.2 14.2 1,287 332 �1901995 5.2 10.2 0.2 13.8 1,273 377 341996 5.1 10.2 0.1 14.4 1,450 516 341997 5.8 10.9 0.6 14.1 1,627 697 �2931998 6.6 11.2 1.9 13.3 2,085 1175 2941999 6.1 11.0 1.2 13.7 1,998 1073 292000 6.0 11.2 0.9 13.9 1,560 641 142001 6.6 11.9 1.3 13.6 1,350 436 �166Mean 6.0 11.0 0.9 14.2 1,579 656 �30

67

specific leaf area, which was set to 35 m2 kg C�1, basedon the field-measured LAI and leaf litter production.Soil texture (Table 1) and soil depth (0.7 and 0.8 m forthe S and N slopes, respectively) measured were utilizedfor site soil characteristic parameters.

Model validation and statistical analysis

For validating daily predictions from the model simu-lation, we compared model estimations with fieldobservations, which included soil respiration, SWC,SOM, soil respiration, leaf litter production, and FLSs.Daily mean soil temperature was calculated by usinghourly soil temperature. Because we have no data ondiurnal measurements, it was assumed that diurnalvariation could be ignored for SWC and SOM. Weused the instantaneously measured values as dailymeans. In this study, diurnal variation of soil respira-tion was also ignored and hourly units of the measuredsoil respiration (g C m�2 h�1) were converted to dailysoil respiration (g C m�2 day�1) by simply multiplyingthe initial measurement by 24. As Kang et al. (2003)discussed, this conversion method overestimates dailysoil respiration by ignoring diurnal soil respirationvariation, which reaches a maximum rate during earlyafternoon. In the Results section, we discussed thematter of using daily-converted soil respiration formodel validation.

A two-tailed paired-sample t-test was used to testfor differences between variables from the S and Nslopes. The BIOME-BGC outputs were compared withavailable field observations at daily and annual timescales. The analyses were used to test the ability ofBIOME-BGC to simulate slope-specific differences insoil physical and biogeochemical variables found infield observations. Paired variables included daily soiltemperature, soil respiration, SWC, SOM, FLSs, andannual leaf litter production. Next, a Pearson correla-tion analysis was utilized to examine the relationshipsbetween inter-annual variations in leaf litter produc-tion, maximum LAI and NPP, and meteorologicalvariables. The statistical analysis was done by usingSPSS for Windows, Release 10.0.5 (SPSS., Chicago,Ill.).

Results

Effects of topography

Our field observations showed that differences betweenthe variables measured from both slopes were statisti-cally significant for every variable, except for leaf litterproduction (Fig. 1). Soil temperatures were significantly(P<0.01) higher in the S slope (9.7�C) than the N slope(9.1�C) (Fig. 1). Interestingly, this difference in soiltemperature was most distinct during two periods withinthe year: early spring and late autumn (Fig. 2a). During

early spring, thawing was delayed by �16 days (range:13–19 days), resulting in considerably lower soiltemperatures on the N slope than the S slope. Duringwinter and growing seasons, however, snow cover, FLS,and canopy mitigate the effect of local topography onatmospheric thermal flux to soil (Kang et al. 2000),which resulted in an insignificant difference in soil tem-peratures between the slopes as observed in this study. Incontrast, SWC, SOM, and FLS were significantly(P<0.02) higher at the N slope (SWC=75%;SOM=8.8%; FLS=188 g C m�2) than at the S slope(SWC=66%; SOM=8.0%; FLS=54 g C m�2),respectively (Fig. 1). Leaf litter production was higher,but not statistically significantly, on the N slope (162 gC m�2 year�1) than on the S slope (150 gC m�2 year�1).

Soil respiration was higher (P<0.05) on the N slope(4.4 g C m�2 day�1) than on the S slope (3.7 gC m�2 day�1). Soil temperature explained about 75% ofthe seasonal variations in soil respiration for both slopes(Fig. 3a). This figure also shows that soil respiration ofthe N slope was greater than that of the S slope at asimilar soil temperature, which implies that higher SWCand SOM (Fig. 1) stimulated greater microbial activityon the N slope than on the S slope, as suggested by Kanget al. (2003). These results indicate that slope/aspect is agood indicator with which to identify spatial differencesin soil physical and biogeochemical variables, whichimplies an important role of local topography in con-trolling diverse hydro-ecological processes of a ruggedforested landscape. Although our findings are based onfield observations only from a pair of slopes, we foundconsistent results year-by-year and similar results con-ducted in similar study areas were provided by usingmodel simulation (Kang et al. 2004) and field observa-tion (Kang et al. 2003).

0

50

100

150

200

250

300

ST SWC SOM SR FL LP

SW

C (

%),

FL

(gC

m-2

), L

P (

gC m

-2 y

-1)

0

5

10

15

20

25

30

ST

(o C

), S

OM

(%

), S

R (

gC m

-2 d

-1)N S

Fig. 1 Comparison of means of observed variables on the north-facing (N) and south-facing (S) slopes. Vertical bar is SD. *P<0.05, **P<0.01 (two-tailed). ST Soil temperature (�C), SWC soil watercontent (%), SOM soil organic matter (%), SR soil respiration (gC m�2 day�1), FL floor leaf litter store (g C m�2), LP leaf litterproduction (g C m�2 year�1), d day

68

Validation of model predictions

Predictions of soil temperature generally showed goodagreements with the measured soil temperature butconsiderable errors were found during each winter(Fig. 2b). In spite of soil temperature staying around0�C during winters, the predicted temperature showedconsiderable variations. A soil temperature sub-model ofBIOME-BGC follows an algorithm suggested by Zhenget al. (1993) that uses the 11-day moving average of airtemperature without explicitly considering the effect ofsoil surface cover (e.g., snow, floor leaf litter, moss orlichen, and canopy leaves) (Thunholm 1990; Kang et al.2000). Hence, while deep snowpack at our study sitecaused the relatively constant soil temperature duringwinter seasons, the model (BIOME-BGC version 4.11)predicted soil temperature variable depending on airtemperature. The slope-specific difference shown inFig. 2a was not predicted in this study, because we used

identical air temperature on both slopes and becauseBIOME-BGC does not consider effects of surface coveron soil temperature.

Modeled SWC and SOM were higher on the Nslope than on the S slope but were not significantlydifferent when compared to field measurements(Figs. 1, 2). SWC (mass, %) was underestimated withmean biases of �7% and �20% on the N and S slopes,respectively (Fig. 2c). It was expected that a moresandy-textured soil at the S slope would result in alower SWC than on the N slope. The model, however,was too sensitive to soil texture, which predicted muchlower SWC (mean=43%) than the measured SWC(mean=63%) on the S slope. Similarly, modeled SOM(mean=3.5%) underestimated the measured SOM(8.4%), partly because BIOME-BGC averaged soil-column SOM in the entire rooting depth (1.0 m in thisstudy), while we measured SOM from the top 10 cm ofthe soil.

-2

0

2

4

6

8

Soi

l tem

pera

ture

(o C

)

1996 1997 1998 1999 2000

-10

0

10

20

30

Soi

l tem

pera

ture

(o C

)

Predicted Observed

1996 1997 1998 1999 2000

0

30

60

90

120

150

SW

C (

%)

NS

1996 1997 1998 1999 2000

a

b

c

Fig. 2 Time series of a difference insoil temperature (�C) between the Sand N slopes measured at 10-cm soildepth, b soil temperature (�C)measured and predicted at the N slope,and c gravimetric soil water content(SWC, %) measured and predictedfrom 1996 to 2000 on the S slopes(filled square) and N slopes (opentriangle). Vertical bar is SD. For otherabbreviations, see Fig. 1

69

As explained in the earlier section, infra-red gasanalysis (IRGA) measurements were conducted for2 min to measure daytime soil respiration and then,were converted to daily values by simply ignoringdiurnal variations, which likely resulted in overesti-mation (Kang et al. 2003). On the other hand, it is

generally recognized that the soda lime methodunderestimates soil respiration (Raich and Schlesinger1992; Raich et al. 2002). So, we assumed that theIRGA-based daily soil respiration would provide ahypothetical upper bound of actual daily soil respira-tion (Fig. 3b), while soda-lime measurement wouldsuffice as a lower bound of actual daily soil respiration(Fig. 3c). Modeled soil respiration (mean=2.9 gC m�2 day�1) lay between the hypothetical upperbound by the IRGA (4.0 g C m�2 day�1) (Fig. 3b)and the hypothetical lower bound by the soda-limemethod (1.9 g C m�2 day�1) (Fig. 3c). In addition, thepredicted soil respiration showed significant correla-tions with the upper (r2 =0.82 and 0.81, P<0.05, forN and S slopes, respectively) and lower bounds (r2

=0.64 and 0.32, P<0.05, for N and S slopes, re-spectively). We, therefore, consider the modeled soilrespiration to be reasonable in magnitude and a goodrepresentation of seasonal variation in measured soilrespiration.

Lee et al. (1999) showed that FLSs were very sensitiveto local topography, wind direction, and snowpackdynamics, which resulted in much higher accumulationon the N slope than on the S slope (Fig. 4a). TheBIOME-BGC did not simulate this phenomenon, butonly predicted FLS as a residual of leaf litter productionby the microbial decomposition process. Nevertheless,mean seasonal variation of FLSs across both slopes wasreasonably well described by the model (Fig. 4b).Overestimation of the modeled FLSs takes place becausefloor leaf litter moves out of the slopes to bottom val-leys, which were not within the scope of modeling area inthis study. Further, complicated simulation using two-dimensional wind fields would be necessary to mimic thephenomena.

Although the predicted leaf litter production waslower in magnitude (147 g C m�2 year�1) and less var-iable inter-annually (140–156 g C m�2 year�1) than thatobserved (156 g C m�2 year�1; 140–167 g C m�2

year�1), we found excellent correlation between themodeled and observed leaf litter production (r=0.96,P<0.01). As noted in field observations, the leaf litterproduction levels predicted were not significantly dif-ferent between the slopes (145 and 148 g C m�2 year�1

for the S and N slopes, respectively).

Effects of climatic variability

Table 3 summarizes inter-annual variation of selectedmodel variables, averaged across the S and N slopes.The study site was characterized as a carbon sink (12 gC m�2 year�1) for all years, except 1997 and 1998.Approximately, 60% (43–74%) of annual PRCP wasreturned to the atmosphere by ET. The inter-annualvalues of the modeled variables varied by 10–20%.Variables relevant to productivity were minimized in1997 by the long dry summer (Table 2, Fig. 2c). Incontrast, the highest soil respiration was found in 1998,

R2 (N) = 0.64

R2 (S) = 0.32

0

1

2

3

4

5

6

0 1 2 3 4 5 6

Predicted (gC m-2 d-1)

Obs

erve

d (g

C m

-2 d

-1)

R2 (N) = 0.75Q10 = 4.3

R2 (S) = 0.76Q10 = 4.1

0

2

4

6

8

10

0 5 10 15 20 25

Soil temperature (oC)

Soi

l res

pira

tion

(gC

m-2

d-1

) NS

R2 (N) = 0.82

R2 (S) = 0.81

0

2

4

6

8

10

0 2 4 6 8 10

Predicted (gC m-2 d-1)

Obs

erve

d (g

C m

-2 d

-1)

N S

a

b

c

Fig. 3 Regressions between a observed soil respiration and soiltemperature, and scatter plots between the predicted and themeasured soil respiration b by using an infrared gas analyzer and cby using the soda-lime method for the N and S slopes. Vertical baris SD. Solid and dotted lines are the best regressions of the N and Sslopes, respectively. R2 Coefficient of determination, Q10 temper-ature sensitivity of soil respiration to 10�C increase in soiltemperature; for other abbreviations, see Fig. 1

70

when the highest temperature and PRCP for the studyperiod were recorded (Table 2). We hypothesize that lowproductivity and high respiration forced our study forestto act as a carbon source in 1997 and 1998.

In terms of magnitude and temporal (seasonal orinter-annual) variation, BIOME-BGC predicted lesssignificant slope-specific differences than were observed.For examples, the modeled soil respiration was higher,but not statistically significantly (P=0.68), on the Nslope than on the S slope, which compared with signif-icantly (P=0.03) higher soil respiration on the N slopefound in field measurement (Fig. 1). We therefore ig-nored slope-specific differences and used predictionsaveraged across both slopes for carrying out correlationanalysis with climatic variability.

The correlation analysis showed distinct and inter-esting results (Table 4). Among other meteorologicalvariables, PRCP and annual WB showed significantpositive correlation with leaf litter production, maxi-mum LAI, and NPP. Leaf litter production and maxi-mum LAI were significantly correlated with PRCP andWB with a 1-year lag, while NPP showed significantcorrelation without a year lag. Interestingly, radiationconsistently showed negative correlation with the modelvariables, possibly caused by a strong negative correla-tion (r=�0.88, P<0.01) between summer PRCP andradiation (Fig. 5).

Discussion

Effects of local topography on soil physicaland biogeochemical processes

The status of soil physical and biogeochemical variablesis the result of complex interactions among diverseecosystem components including climate, vegetation andfloor structure, soil characteristics, and local topogra-phy. Our field results indicate the important role of localtopography in determining the status of soil physicaland biogeochemical variables. Local topography con-trols surface energy flux (Running et al. 1987; Kanget al. 2002), soil physical variables (Kang et al. 2004),and consequent microbial activity (Kang et al. 2003).Our results indicate that slope/aspect is a good indicatorwith which to identify spatial heterogeneity in the soilphysical and biogeochemical variables.

Local topography produced significant differences insoil-related variables such as soil temperature, watercontent, organic matter, and soil respiration, as well asFLSbetween adjacentNandS slopes (Fig. 1). In contrast,vegetation production, for which we used leaf litter pro-duction as a surrogate, did not differ considerably be-tween the two slopes, despite the fact that the N slopereceived less solar radiation (13.1 MJ m�2 day�1) than

0

100

200

300

400

97/9/1 98/3/1 98/9/1 99/3/1 99/9/1

leaf

litte

r (g

C m

-2 y

r-1) bgc

obs

0

100

200

300

400

0 100 200 300 400

Predicted litterfall (gC m-2)O

bser

ved

litte

rfal

l (gC

m-2

) S

N

Mean

ab

Fig. 4a, b Comparisonsbetween the predicted andmeasured floor leaf litter stores.a Scatter plot between predicted(bgc) and observed (obs) floorleaf litter stores on the S and Nslopes, and b seasonal variationof bgc and obs floor leaf litterstores, averaged across both Sand N slopes. Vertical bar isSD. yr Year; for otherabbreviations, see Fig. 1

Table 3 Summary of model predictions averaged across S and Nslopes; units are g C m�2 year�1, except for evapotranspiration(ET; mm year�1) and LAImax (m

2 m�2). Leaf_bgc Annual leaf litterproduction, NPP net primary production, SR total soil respirationincluding microbial and root respiration, NEE net ecosystem ex-change; for other abbreviations, see Table 1

Year ET Leaf_bgc LAImax NPP SR NEE

1994 929 142 4.8 690 745 201995 852 143 4.7 722 765 241996 922 146 4.8 710 750 281997 879 139 4.6 663 745 �151998 970 147 4.8 786 897 �351999 864 155 5.1 739 755 602000 972 150 4.9 732 799 92001 996 148 4.9 727 778 29Mean 923 146 4.8 721 745 12

r = -0.88 (p < 0.01)

14

15

16

17

18

19

20

0 300 600 900 1200 1500

Precipitation (mm)

Rad

iatio

n (M

J m

-2 d

-1)

Fig. 5 Correlation between June and August mean daily radiation(MJ m�2 day�1) and precipitation (mm) for 1993–2001

71

the S slope (14.6 MJ m�2 day�1) (Kang et al. 2002).Among many other possible explanations, we hereemphasize plant physiological responses to meteorologi-cal and soil physical conditions for each slope.We suspectthat a cooler and wetter microclimate during growingseasons resulted in longer stomatal opening time andhence compensated for the decreased incoming radiationat the N slope. It is also suspected that dry and hot soilconditions on the S slope resulted in shorter stomatalopening time and greater allocation of photosyntheticproducts to belowground parts than aboveground parts.Further studies on the plant physiological response wouldbe necessary to elucidatemechanisms of controlling slope-specific vegetation productivity.

As Lee et al. (1999) indicated, our field results showedthat FLSs were very sensitive to local topography, winddirection, and snowpack dynamics. During late autumnand early winter, frequent southwest wind resulted inmuch higher accumulation on the N slope than on the Sslope (Fig. 4a). In addition, compared to rarely snow-covered S slopes, thick snowpack is usual on the N slopeduring winter seasons, which retains floor leaf litterduring the winter and consequently provides muchgreater organic inputs to the N soil. The proposedmechanisms are very relevant to local heterogeneity interms of slope aspect and could explain why SOM isgreater on the N slope than the S slope in our study area.On the other hand, it is well known that SWC is closelycorrelated with SOM content in forest soils. We, there-fore, suppose that local heterogeneity in slope aspectexplains well why SWC and SOM are greater at the Nslope in our study area.

Previous studies, conducted in similar study areas, byusing model simulation (Kang et al. 2004) and fieldobservation (Kang et al. 2003) provided similar resultsto those obtained from this study, which generalizes ourfindings from a slope level to a rugged forested land-scape. There is tremendous uncertainty in scalingplot-level ecosystem observations up to a landscape orregional scale. However, with rapid improvement ofsatellite remote sensing, SISs, fluxtower eddy-covariancemeasurements, and spatial ecosystem modeling, itbecomes possible to conduct a further enhanced studyon the effect of local topography on hydro-ecological

processes in a rugged forested landscape. The BigFootproject was a recent effort to scale plot-level observa-tions up to a regional scale (Cohen et al. 2003; Turneret al. 2003) but was still limited by incomplete treatmentin dealing with the effect of local topography, comparedto other modeling studies (e.g., Band et al. 1993; Kanget al. 2004). Our study indicates that local topographyneeds to be considered as an important input to generatespatial heterogeneity in future scaling studies.

Climatic controls on vegetation production

Although BIOME-BGC produced less significant slope-specific differences than observations, the model showedgood agreement with the magnitude and temporal var-iation of the observed carbon-process variables. Thisresult indicates that the model predictions could be usedto examine inter-annual carbon processes at the site.Caution should be exercised in modeling conditionsduring extraordinarily dry years like 1997, because themodel overestimated drought effect by predicting muchlower leaf litter production (139 g C m�2 year�1) thanwas observed in the field (152 g C m�2 year�1). Thiscaveat is especially important when the model is appliedin rural montane areas, where soil texture information israrely available, because modeled SWC is sensitive tosoil texture.

Inter-annual variations in both observed and mod-eled leaf litter production and maximum LAI were bestexplained by PRCP and WB with a 1-year lag. Thetemporal lag of climatic control on leaf-related variablesis well recognized in leaf phenology studies, demon-strating that last year’s summer temperature controlshardness of bud dormancy (Kramer 1994; Chuine 2000).In contrast, variation in NPP was well correlated withPRCP without a temporal lag. Our findings that highproductivity in the current year improves canopy bio-mass in the following year are consistent with recognizedtree physiology principles for fixed growth speciesincluding Q. mongolica, a dominant species in our studysite (Kozlowski et al. 1991).

Churkina and Running (1998) showed that East Asiais a radiation-limited ecosystem for NPP. Alternatively,

Table 4 Summary of correlation analysis; P-values in parentheses(not given when P>0.05). Leaf litter production, LAImax, and NPPwere correlated with meteorological variables with and without 1-year time lag effect of meteorology on the carbon-cycling variables.

Lag Presence of a 1-year lag of the variables listed in the correlationanalysis, Leaf_obs observed leaf litter production, Leaf_bgc simu-lated leaf litter production, LAImax simulated LAImax, NPP simu-lated annual NPP; for other abbreviations, see Tables 1 and 3

Lag Tavg(�C) Tmax(�C) Tmin(�C) RAD (MJ m�2 year�1) PRCP (mm year�1) WB (mm year�1)

Leaf_obs (g C m�2 year�1) Yes 0.29 0.72 0.37 0.38 0.82 (0.04)* 0.82 (0.05)*No 0.75 0.88 (0.02)* 0.60 �0.06 �0.18 �0.21

Leaf_bgc (g C m�2 year�1) Yes 0.40 �0.45 0.59 �0.25 0.77 (0.04)* 0.78 (0.04)*No 0.21 0.11 0.28 �0.59 0.67 0.67

LAImax(m2 m�2) Yes 0.44 �0.17 0.65 �0.89 (0.01)** 0.82 (0.02)* 0.84 (0.02)*

No 0.36 0.33 0.35 �0.21 0.52 0.51NPP (g C m�2 year�1) Yes 0.14 �0.44 0.05 �0.09 0.26 0.24

No 0.37 0.03 0.59 �0.87 (0.01)** 0.82 (0.02)* 0.85 (0.02)*

*P £ 0.05, **P £ 0.01

72

our study indicates that radiation is not a limiting factorin NPP, but that soil water availability in the growingseason controls NPP, even in areas with relatively highannual PRCP. We hypothesize that the well-drainedsoils and rugged topography characteristics of our studyregion may cause temporary drought conditions, par-ticularly during the growing season. In fact, the summerWB shown in Table 2 demonstrates the frequency ofdrought conditions at our study site, even without con-sideration of frequent subsurface flow.

In this simulation study, our study site was estimatedto be a carbon sink (12 g C m�2 year�1) for all yearsexcept 1997 and 1998. These exceptions were probablydue to dry summers and hot growing seasons. We sus-pect that our current net ecosystem exchange estimationprobably underestimates the carbon-sequestration abil-ity of our study site because the carbon cycling of theforest was assumed to be in a steady state. We consid-ered neither age-specific growth rate nor disturbanceregimes, which are likely to have an influence on currentcarbon budgets. This fact is especially important in mostKorean forests, since they are still aggrading due to thenational replanting effort following the Korean War(Lee et al. 2002). More extensive field surveys and moreexplicit simulation efforts with regard to anthropogenicdisturbances are required for a more accurate estimationof current carbon budgets in Korean forests.

Acknowledgements We greatly appreciate the help of our col-leagues who shared field work at Mt. Jumbong forest and MonkSungil for his invaluable assistance during our field work. Thisresearch was supported by a grant (1-8-2) from the SustainableWater Resources Research Center of the Twenty-First CenturyFrontier Research Program and from the Eco-technopia 21 Project(Ministry of Environment, Korea). Sinkyu Kang was supported bya grant from Kangwon National University.

References

Amthor JS, Chen JM, Clein JS, Frolking SE, Goulden ML, GrantRF, Kimball JS, King AW, McGuire AD, Nikolov NT,Potter CS, Wang S, Wofsy SC (2001) Boreal forest CO2

exchange and evapotranspiration predicted by nine ecosystemprocess models: intermodel comparisons and relationshipsto field measurements. J Geophys Res Atmos 106:33623–33648

Band LE, Tague CL, Groffman P, Belt K (1993) Forest eco-system processes at the watershed scale: hydrological andecological controls of nitrogen export. Hydrol Process15:2013–2028

Battle M, Bender ML, Tans PP, White JWC, Conway T, FranceyRJ (2000) Global carbon sinks and their variability inferredfrom atmospheric O2 and d13 C. Science 287:2467–2470

Bolstad PV, Vose JM, McNulty SG (2001) Forest productivity, leafarea, and terrain in southern Appalachian deciduous forests.For Sci 47:419–427

Bousquet P, Peylin P, Ciais P, Quere CL, Friedlingstein P, Tans PP(2000) Regional changes in carbon dioxide fluxes of land andoceans since 1980. Science 290:1342–1346

Burrows SN, Gower ST, Norman JM, Diak G, Mackay DS, AhlDE, Clayton MK (2003) Spatial variability of aboveground netprimary production for a forested landscape in northern Wis-consin. Can J For Res 33:2007–2018

Cavender-Bares J, Bazzaz FA (2000) Changes in drought responsestrategies with ontogeny in Quercus rubra: implications forscaling from seedlings to mature trees. Oecologia 124:8–18

Choi SD, Lee K, Chang YS (2002) Large uptake rate of atmo-spheric carbon dioxide by planted forest biomass in Korea.Glob Biogeochem Cycles 16(4):36/1–36/5

Chuine I (2000) A unified model for budburst of trees. J Theor Biol207:337–347

Churkina G, Running SW (1998) Contrasting climatic controls onthe estimated productivity of global terrestrial biomes. Eco-systems 1:206–215

Cohen WB, Maiersperger TK, Yang Z, Gower ST, Turner DP,Ritts WD, Berterretche M, Running SW (2003) Comparisons ofland cover and LAI estimates derived from ETM+ and MO-DIS for four sites in North America: a quality assessment of2000/2001 provisional MODIS products. Remote Sens Environ88:233–255

Daly C, Bachelet D, Lenihan JM, Neilson RP, Parton W, Ojima D(2000) Dynamic simulation of tree-grass interactions for globalchange studies. Ecol Appl 10:449–469

Eum S, Kang S, Lee D (2005) A simulation study to investigateclimatic controls on net primary production of a rugged for-ested landscape in the mid-western Korean Peninsula. KoreanJ Agric For Meteorol 7:66–77

Fang JY, Chen AP, Peng CH, Zhao SQ, Ci L (2001) Changes inforest biomass carbon storage in China between 1949 and 1998.Science 291:2320–2322

Federer CA, Vorosmarty C, Geketer B (1996) Intercomparison ofmethods for calculating potential evaporation in regional andglobal water balance models. Water Resour Res 32:2315–2321

Iverson LR, Dale ME, Scott CT, Prasad A (1997) A GIS-derivedintegrated moisture index to predict forest composition andproductivity of Ohio forests (USA). Landsc Ecol 12:331–348

Kang S, Kim S, Oh S, Lee D (2000) Predicting spatial and temporalpatterns of soil temperature based on topography, surface coverand air temperature. For Ecol Manage 136:173–184

Kang S, Kim S, Lee D (2002) Spatial and temporal patterns ofsolar radiation based on topography and air temperature. CanJ For Res 32:487–497

Kang S, Doh S, Lee DS, Lee D, Jin V, Kimball JS (2003) Topo-graphic and climatic constraints on soil respiration in temperatemixed-hardwood forests in Korea. Glob Change Biol 9:1427–1437

Kang S, Lee D, Kimball JS (2004) The effects of spatial aggregationof complex topography on hydro-ecological process simulationswithin a rugged forest landscape: development and application ofa satellite-based topoclimatic model. Can J For Res 34:519–530

Kauppi PE,Mielikaninen K, Kuusela K (1992) Biomass and carbonbudget of European forests 1971 to 1990. Science 256:70–74

Keeling CD, Whorf TP, Wahlen M, Plicht JV (1995) Interannualextremes in the rate of atmospheric carbon-dioxide since 1980.Nature 375:666–670

Kimball JS, White MA, Running SW (1997) BIOME-BGC simu-lations of stand hydrologic processes for BOREAS. J GeophysRes 102:29043–29051

Korea Forest Service (1997) Statistical yearbook of forestry. Seoul,Republic of Korea

Kozlowski TT, Kramer PJ, Pallardy SG (1991) Physiologicalecology of woody plants. Academic Press, New York

Kramer K (1994) Selecting a model to predict the onset of growthof Fagus sylvatica. J Appl Ecol 31:172–181

Lee D, Yoo G, Oh S, Shim JH, Kang S (1999) Significance ofaspect and understory to leaf litter redistribution in a temperatehardwood forest. Korean J Biol Sci 3:143–147

Lee D, Yook KH, Lee D, Kang S, Kang H, Lim JH, Lee KH(2002) Changes in annual CO2 fluxes estimated from inventorydata in South Korea. Sci China 45[Suppl]:87–96

Londo AJ, Messina MG, Schoenholtz SH (1999) Forest harvestingeffect on soil. Soil Sci Soc Am J 63:637–644

McNab WH (1993) A topographic index to quantify the effect ofmesoscale landform on site productivity. Can J For Res23:1100–1107

73

Nemani RR, Keeling CD, Hashimoto H, Jolly WM, Piper SC,Tucker CJ, Myneni RB, Running SW (2003) Climate-drivenincreases in global terrestrial net primary production from 1982to 1999. Science 300:1560–1563

Oh S, Yoo G, Shim J, Kang S, Lee D (1999) Spatial heterogeneity ofsoilK, Ca,Mg, Fe, andAl governed by leaf litter redistribution ina temperate hardwood forest. Korean J Ecol 22:205–210

Pacala SW et al (2001) Consistent land- and atmosphere-based UScarbon sink estimates. Science 292:2316–2320

Priestley CH, Taylor RJ (1972) On the assessment of surface heatflux and evaporation using large-scale parameters. MonWeather Rev 100:81–91

Raich JW, Schlesinger WH (1992) The global carbon dioxide fluxin soil respiration and its relationship to vegetation and climate.Tellus 44B:81–99

Raich JW, Potter CS, Bhagawati D (2002) Interannual variabilityin global soil respiration. Global Change Biol 8:800–812

Running S, Nemani R, Hungerford R (1987) Extrapolation ofsynoptic meteorological data in mountainus terrain and its usefor simulating forest evapotranspiration and photosynthesis.Can J For Res 17:472–483

Thornton PE, Law BE, Gholz HL, Clark KL, Falge E, EllsworthDS, Goldstein AH, Monson RK, Hollinger D, Falk M, ChenJ, Sparks JP (2002) Modeling and measuring the effects of

disturbance history and climate on carbon and water budgetsin evergreen needleleaf forests. Agric For Meteorol 113:185–222

Thunholm B (1990) A comparison of measured and simulatedsoil temperature using air temperature and soil surfaceenergy balance as boundary condition. Agric For Meteorol53:59–72

Turner DP, Ritts W, Cohen WB, Gower ST, Zhao M, Running SW(2003) Scaling gross primary production (GPP) over boreal anddeciduous forest landscapes in support of MODIS GPP productvalidation. Remote Sens Environ 88(3):256–270

White MA, Thornton PE, Running SW, Nemani RR (2000)Parameterization and sensitivity analysis of the BIOME-BGCterrestrial ecosystem model: net primary production controls.Earth Interact 4:1–85

Wofsy SC et al (1993) Net exchange of CO2 in a mid latitude forest.Science 260:1314–1317

Yoo G, Park EJ, Kim SH, Lee HJ, Kang S, Lee D (2001) Trans-port and decomposition of leaf litter as affected by aspect andunderstory in a temperate hardwood forest. Korean J Biol Sci5:319–325

Zheng D, Hunt ER Jr, Running SW (1993) A daily soil tempera-ture model based air temperature and precipitation for conti-nental applications. Clim Res 2:183–191

74