Satellite-Based Modeling of the Carbon Fluxes in Mature Black Spruce Forests in Alaska: A Synthesis of the Eddy Covariance Data and Satellite Remote Sensing Data

Masahito Ueyama Graduate School of Life and Environmental Sciences, Osaka Prefecture University, Sakai, Japan

Search for other papers by Masahito Ueyama in
Current site
Google Scholar
PubMed
Close
,
Yoshinobu Harazono International Arctic Research Center, University of Alaska Fairbanks, Fairbanks, Alaska

Search for other papers by Yoshinobu Harazono in
Current site
Google Scholar
PubMed
Close
, and
Kazuhito Ichii Faculty of Symbiotic Systems Science, Fukushima University, Fukushima, Japan

Search for other papers by Kazuhito Ichii in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Scaling up of observed point data to estimate regional carbon fluxes is an important issue in the context of the global terrestrial carbon cycle. In this study, the authors proposed a new model to scale up the eddy covariance data to estimate regional carbon fluxes using satellite-derived data. Gross primary productivity (GPP) and ecosystem respiration (RE) were empirically calculated using the normalized difference vegetation index (NDVI) and land surface temperature (LST) from the Moderate Resolution Imaging Spectroradiometer (MODIS). First, the model input is evaluated by comparing with the field data, then established and tested the model at the point scale, and then extended it into a regional scale. At the point scale, the empirical model could reproduce the seasonal and interannual variations in the carbon budget of the mature black spruce forests in Alaska and Canada sites, suggesting that seasonality of the NDVI and LST could explain the carbon fluxes and that the model is robust within mature black spruce forests in North America. Regional-scale analysis showed that the total GPP and RE between 2003 and 2006 were 1.76 ± 0.28 and 1.86 ± 0.26 kg CO2 m−2 yr−1, respectively, in mature black spruce forests in Alaska, indicating that these forests were almost carbon neutral. The authors’ model analysis shows that the proposed method is effective in scaling up point observations to estimate the regional-scale carbon budget and that the mature black spruce forests increased in sink strength during spring warming and decreased in sink strength during summer and autumn warming.

*Corresponding author address: Masahito Ueyama, 1-1, Gakuen-cho Naka-ku, Sakai, Osaka, Japan. miyabi-flux@muh.biglobe.ne.jp

Abstract

Scaling up of observed point data to estimate regional carbon fluxes is an important issue in the context of the global terrestrial carbon cycle. In this study, the authors proposed a new model to scale up the eddy covariance data to estimate regional carbon fluxes using satellite-derived data. Gross primary productivity (GPP) and ecosystem respiration (RE) were empirically calculated using the normalized difference vegetation index (NDVI) and land surface temperature (LST) from the Moderate Resolution Imaging Spectroradiometer (MODIS). First, the model input is evaluated by comparing with the field data, then established and tested the model at the point scale, and then extended it into a regional scale. At the point scale, the empirical model could reproduce the seasonal and interannual variations in the carbon budget of the mature black spruce forests in Alaska and Canada sites, suggesting that seasonality of the NDVI and LST could explain the carbon fluxes and that the model is robust within mature black spruce forests in North America. Regional-scale analysis showed that the total GPP and RE between 2003 and 2006 were 1.76 ± 0.28 and 1.86 ± 0.26 kg CO2 m−2 yr−1, respectively, in mature black spruce forests in Alaska, indicating that these forests were almost carbon neutral. The authors’ model analysis shows that the proposed method is effective in scaling up point observations to estimate the regional-scale carbon budget and that the mature black spruce forests increased in sink strength during spring warming and decreased in sink strength during summer and autumn warming.

*Corresponding author address: Masahito Ueyama, 1-1, Gakuen-cho Naka-ku, Sakai, Osaka, Japan. miyabi-flux@muh.biglobe.ne.jp

1. Introduction

Warming trends in arctic and boreal regions have been observed in the past few decades (ACIA 2004). This warming has resulted in dramatic changes, such as alterations in the distribution of permafrost, vegetation phenology, land cover, and carbon budgets (e.g., Chapin et al. 2005; Hinzman et al. 2005). Boreal forests—the second largest biome of Earth’s surface, covering 39% of the pan-Arctic basin (Kimball et al. 2006)—greatly affect the current and future carbon cycle via changes in vegetation productivity and decomposition of the soil organic matter (Euskirchen et al. 2006; Kimball et al. 2006). This forest ecosystem stores a large portion of global terrestrial carbon (Dixon et al. 1994) and could become a carbon source because of decomposition stimulated by the warming (Goulden et al. 1998; Piao et al. 2008). On the other hand, the warming could enhance vegetation productivity (Euskirchen et al. 2006; Kimball et al. 2006) and potentially increase the carbon sink. Because the warming trend will likely continue in the future (Chapman and Walsh 2007), it is important to elucidate the potential response of the high-latitude carbon cycle to future climate changes.

Tower-based flux measurements using the eddy covariance method have been conducted to reveal the terrestrial carbon fluxes (Baldocchi et al. 2001). Multiyear measurements by the eddy covariance method in black spruce forests have demonstrated that these carbon fluxes are controlled by environmental conditions such as air and/or soil temperature (Goulden et al. 1998; Ueyama et al. 2006; Bergeron et al. 2007; Welp et al. 2007) and water availability (Dunn et al. 2006). Although field observations have improved our understanding of the terrestrial carbon cycle, a limited number of observations in high-latitude ecosystems make it difficult to estimate their spatial and temporal variability. Thus, scaling up of field observations to estimate regional carbon fluxes is important.

Satellite remote sensing is a useful tool to scale up point observations to estimate spatial patterns. Normalized difference vegetation index (NDVI) is a widely used vegetation index observed by satellite-borne sensors for vegetation monitoring (Huete et al. 2002; Huemmrich et al. 1999; Chen et al. 2007). Analysis of the historical NDVI data has revealed recent ecosystem changes in the boreal forest, a recent decreasing trend in NDVI, probably due to the spring warming and summer drought (Angert et al. 2005; Bunn and Goetz 2006). The Moderate Resolution Imaging Spectroradiometer (MODIS) has provided nearly real-time estimates of key parameters of the terrestrial carbon cycle, such as vegetation indices and land surface temperature (LST), with state-of-the-art algorithms to minimize atmospheric and geometric noise. This technology is particularly useful for the development of the scaling-up technique of ground observation to estimate regional carbon fluxes (Running et al. 2004).

A number of process-based ecosystem models have previously been proposed to estimate the spatial distributions of carbon fluxes (e.g., McGuire et al. 2000; Sasai et al. 2005). Although those models were validated in many forest ecosystems (e.g., Kimball et al. 1997; Clein et al. 2002), the model performance for high-latitude forests remains uncertain. For example, Kimball et al. (Kimball et al. 2006) reported that the magnitudes of the simulated fluxes were markedly dissimilar between different ecosystem models for high-latitude forests. Ueyama et al. (Ueyama et al. 2010) also simulated carbon and water cycles at larch ecosystems in a cool-temperate to boreal region by using a biosphere model and found that the model with default parameters simulated biased results. Those biases and uncertainties were caused by limited validation and calibration because of few observed data at high-latitude ecosystems.

Synthesis of ground observation and satellite data by a simple empirical model is a promising method to estimate the spatial distributions of the carbon fluxes, because most of the model parameters and processes are obtained from the observation (Mahadevan et al. 2008). For this reason, several scaling techniques have been developed by integrating observed and satellite data (e.g., Rahman et al. 2001; Vourlitis et al. 2003; Yang et al. 2007; Kitamoto et al. 2007; Yamaji et al. 2008; Date et al. 2009; Huemmrich et al. 2009). Empirical models are useful to understand the terrestrial carbon cycle; however, a limited number of empirical models remain uncertain in their applicability, because the algorithm, suitable input parameters, and applicability could be intrinsic in each ecosystem. Thus, further development, improvement, and applications of empirical algorithms are needed.

To understand the carbon cycle of subarctic forest in Alaska, where pronounced summer warming has been observed (Chapin et al. 2005), we have conducted year-round flux observations using the eddy covariance method in interior Alaska since fall 2002 (Ueyama et al. 2006) and evaluated the important environmental factors to model the carbon fluxes in Alaska (Kitamoto et al. 2007; Date et al. 2009; Ueyama et al. 2009). In this study, we proposed a simple empirical model driven by satellite data. We had the following goals: 1) to scale up eddy covariance data using remote sensing data to estimate regional carbon fluxes and 2) to evaluate the seasonal and interannual variation of the carbon budget throughout mature black spruce forests in Alaska. First, we proposed a satellite-based empirical model driven by MODIS standard products and evaluated the model using ground observations. Then, we calculated regional carbon fluxes from 2003 to 2006 at 16-day intervals using this model and evaluated the response of the regional carbon fluxes to weather conditions, such as seasonal surface air temperature patterns.

2. Method

2.1. Study area

Our study focused on mature black spruce forests of interior Alaska, covering the region of 72°–52°N, 140°–168°W (Figure 1). To identify pixels of dominant black spruce cover within the region, we used a land-cover map (1.1-km-grid spatial resolution provided by Alaska Geospatial Data Clearinghouse; Fleming 1997) and converted it to a 250-m grid spatial resolution for model input assuming the homogeneous in each 1.1-km grid. We selected the following pixels as mature black spruce forests from the vegetation classification: 1) spruce woodland shrubs, 2) open spruce forest–shrub–bog mosaic, and 3) open and closed spruce forest. The climate is continental with large seasonal temperature variability, low humidity, and relatively light and irregular precipitation (Shulski and Wendler 2007; Hinzman et al. 2006). Although summer is warm and sunny, the annual average air temperature in the region is a few degrees below freezing because of extremely cold winters: the temperature ranges from −50° to 30°C with about 250 mm yr−1 of the total annual precipitation.

2.2. Field observations and data processing

We used data from two field observation sites. One set of field data is from Alaska (inside the study area; Figure 1), and the other one is from Canada (outside the study area; about 2800 km southwest of the Alaska site); we used the one additional site from Canada, because there is only one field site available inside Alaska.

The field observation site in the study area is located in a typical taiga forest in Fairbanks, interior Alaska (64°52′N, 147°51′W; Figure 2). The mean monthly air temperature of Fairbanks between 1971 and 2000 is the lowest in January, at −23.2°C, and highest in July, at 16.9°C, with an annual mean of −2.9°C (Shulski and Wendler 2007). The average precipitation was 263 mm yr−1, with approximately 37% as snow and the rest in the form of rain. Black spruce (Picea mariana), the dominant overstory tree species, grows sparsely on discontinuous permafrost. Average canopy height is about 1.5 m, but tall trees of up to 6 m can be found sporadically. The forest floor is completely covered by understory vegetation, such as low and dwarf shrubs, vascular plants, mosses, and lichens (Figure 2a). The leaf area index (LAI) measured by a plant canopy analyzer (Li-Cor, LAI-2000, United States) was 2.0 m2 m−2 above the ground (Ueyama et al. 2006). The forest around the observation tower is old growth, with a mean age of 120 yr (Vogel et al. 2005). The approximate fetch was 180 m in the west, 250 m in the south, and more than 500 m in the north to east directions, and the dominant wind direction was southwest and east. Considering the measurement height of the eddy covariance sensors at 6 m and the dominant wind direction, the tower footprint area represented the black spruce stand categorized as the open spruce forest–shrub–bog mosaic mentioned in section 2.1 (Figure 2d).

The eddy covariance measurements were conducted over a flat area using a 10-m tower (Ueyama et al. 2006). Fluxes of CO2, water vapor, heat, and momentum were measured at 10 Hz from a sampling height of 6 m above ground level, using the eddy covariance method with a sonic anemometer (CSAT3, Campbell Scientific Inc., United States) and an open-path gas analyzer (LI-7500, Li-Cor; Figure 2c). Other climate variables, such as soil and air temperature, vapor pressure deficit (VPD), wind speed, and precipitation, were also measured. Half-hourly turbulent fluxes were calculated from the recorded data by correcting high-frequency loss associated with the pathlength and sensor separation (Moore 1986). CO2 and water vapor fluxes were corrected for density fluctuations arising from air temperature and humidity [Webb–Pearman–Leuning (WPL) correction; Webb et al. 1980]. Because of the calm wind condition and the severe winter climate at this site, limited data were available after the quality control (QC) with a nighttime friction velocity threshold (Ueyama et al. 2006; Ueyama et al. 2009); the percentages of the remaining flux data in the total time period for 2003, 2004, 2005, and 2006 were 45%, 35%, 46%, and 44%, respectively.

Data gaps were filled by using standardized methods: linear interpolation for small gaps shorter than 2.5 h and mean diurnal variation method for longer gaps (Falge et al. 2001). Negative CO2 fluxes derived from the open-path sensor during snow periods (e.g., Hirata et al. 2005) were eliminated in this analysis. Ecosystem respiration (RE) was derived from temperature response functions between nighttime net ecosystem exchange (NEE) and air temperature, and then the relationships were extrapolated to daytime (Ueyama et al. 2006). Gross primary productivity (GPP) was determined by the difference between NEE and RE. For the periods with long data gaps, such as summer in 2004, we recovered GPP by applying an empirical model (Ueyama et al. 2006). Both GPP and RE were calculated every 30 min, and then aggregated daily mean values were derived. Details of the observations, the analysis, and the model parameters are described in Ueyama et al. (Ueyama et al. 2006).

To test the effect of different spatial scales (ground and satellite observations) on vegetation index, incoming and reflected shortwave radiation (SR) and photosynthetically active radiation (PAR) were measured at 6 m by pyranometers (CMP3, Kipps and Zonen, Holland) and quantum sensors (LI-190, Li-Cor), respectively; the view angle of the pyranometer is 180°, whereas that of the quantum sensor is approximately 160°. The incoming and reflected radiation measurements were used to calculate broadband NDVI (bNDVI) as follows (Huemmrich et al. 1999):
i1087-3562-14-13-1-e1
where ESWin, ESWr, EPARin, and EPARr are incoming and reflected SR and incoming and reflected PAR, respectively. We calculated bNDVI every 30 min for clear-sky conditions by using observed incoming, reflected SR and incoming, reflected PAR at the study site, and we chose daily maximum bNDVI for daily values to minimize the effect of the solar zenith angle (Huemmrich et al. 1999).

We additionally used the flux data observed at a mature black spruce forest in Canada as an independent model validation. The site is located in central Manitoba, Canada (55°53′N, 98°29′W), which was established as the Northern Old Black Spruce (NOBS) site for the National Aeronautics and Space Administration (NASA) Boreal Ecosystem-Atmosphere Study (BOREAS; Goulden et al. 1997; Dunn et al. 2006; this is called the NOBS site here). The dominant tree species is black spruce (Picea mariana). The average stand age is 160 years, with a leaf area index of 4.2 m2 m−2. The canopy height ranged from 10 m in the uplands to 1–6 m at the lower elevations. The eddy covariance measurements were conducted with a sonic anemometer (SATI/3K, Applied Technologies Inc., United States) and a closed-path gas analyzer (LI-6262, Li-Cor) at the height of 29 m. The mean annual air temperature and precipitation are −3.2°C and 514.7 mm yr−1, respectively (Bergeron et al. 2007). More detailed information about the NOBS site and the measurement including the gap-filling and flux-partitioning method are shown in Dunn et al. (Dunn et al. 2006).

2.3. Model

To scale up the eddy covariance data to estimate regional carbon fluxes, we developed a satellite-based empirical model, the Scaling Approach of Ecosystem Productivity (SCAPE; Figure 3). The model empirically calculates GPP, RE, and NEE every 16 days, using stepwise regression algorithms with the inputs of the NDVI and LST from satellite-based observations. GPP and RE were separately parameterized by using observed data, and NEE was calculated by the difference between GPP and RE. In this paper, negative and positive NEE means a net sink and source of atmospheric CO2, respectively.

Field studies at black spruce forests (Ueyama et al. 2006; Bergeron et al. 2007; Welp et al. 2007) show that GPP is strongly related to air and/or soil temperature and vegetation status, and the photosynthetic capacity is strongly regulated by leaf area index (Ueyama et al. 2006; Kitamoto et al. 2007), seasonal temperature dynamics (Bergeron et al. 2007; Date et al. 2009), and nighttime freezing temperature (Goulden et al. 1997). Therefore, GPP is modeled by NDVI and daily and nighttime air temperature as follows:
i1087-3562-14-13-1-e2
where fGPP(NDVI), fGPP(Tdaily), and f(Tmin) are empirical functions for NDVI, daily air temperature Tdaily, and minimum air temperature Tmin, respectively (Table 1); f(Tmin) is used to express the suppression effect due to nighttime freezing. We applied the suppression function by Tmin from the algorithm in the MOD17 GPP (Heinsch et al. 2006), and the other functions [fGPP(NDVI) and fGPP(Tdaily)] were empirically determined by using the observed data (section 3.1.2).
RE is the sum of autotrophic and heterotrophic respiration, and they are controlled by lived and dead biomass as well as metabolic rates (Chapin et al. 2002). In black spruce forests on cold soils, RE in early spring is also restricted by low soil temperature, despite warm air temperatures (e.g., Goulden et al. 1997; Ueyama et al. 2006). To simply account for these effects, we modeled RE as follows:
i1087-3562-14-13-1-e3
where fRE(NDVI), fRE(Tdaily), f(Tmin), and fRE(GDD) are functions of NDVI, Tdaily, Tmin, and growing degree days (GDD), respectively (Table 1), and GDD is cumulative daily air temperature above 5°C. These functions are empirically determined using ground observations (see section 3.1.2).

2.4. Model inputs

We used NDVI (MOD13Q1) and daytime and nighttime LST (MOD11A2) data in the MODIS standard products (collection 4; available online at https://lpdaac.usgs.gov/; Huete et al. 2002; Wan et al. 2002) as the model inputs. Because the products are provided at 250-m spatial resolution for MOD13-NDVI and 1-km spatial resolution for MOD11-LST, MOD11-LST was resampled to 250-m spatial resolution by the nearest-neighbor method. Because MOD13-NDVI and MOD11-LST are provided for 16- and 8-day compositing periods, respectively, MOD11-LST was formulated as a composite for every 16-day interval; we simply used the first-half 8-day LST data within each 16-day period. For the calculations, unreliable input, mainly caused by cloud contamination, was removed by using the MODIS QC flag; we only used data with a “good quality” flag for the analysis. The rejected data were filled by the mean value of each pixel for that period, by creating the mean seasonal cycle of MOD13-NDVI and MOD11-LST between 2001 and 2006 for each pixel.

2.5. Experiments

The analyses consist of point analyses at the flux observation sites and the regional-scale analyses in interior Alaska. First, we conducted point-based analyses to test the satellite input using ground and satellite observations, to parameterize the model by using the eddy covariance data at the Alaska site, and to examine the model performance at the sites in Alaska and Canada. Then we extended the model to estimate regional carbon fluxes in mature black spruce forests in Alaska.

2.5.1. Point-scale analyses

First, we tested the potential capability of the scaling-up technique for model input. We compared the ground observed bNDVI and air temperature with the satellite-based NDVI and LST and established the relationships. For applications of MOD11-LST, we examined empirical relationships of MOD11-LST to the daily and minimum air temperature at our observation site and nine weather stations made available by the National Climate Data Center (NCDC; http://www.ncdc.noaa.gov/oa/ncdc.html), located in interior Alaska (hereafter called observed air temperature). In this examination, 1 km × 1 km MOD11-LST pixels centered on each station with a good-quality flag were used. The local time of data acquisition within the study region ranged from 10.6 to 13.3 for daytime LST and from 19.8 and 22.4 for nighttime LST, whereas those for the study site were approximately 11 for daytime LST and approximately 21 for nighttime LST. For the estimation of daily air temperature, we examined relationships between observed daily air temperature and mean of daytime and nighttime MOD11-LST using a total of 815 data points. For minimum air temperature, we calculated the 8-day average of daily minimum air temperature derived from our observations and the NCDC and then examined relationships between the 8-day average of minimum air temperature and nighttime MOD11-LST using a total of 976 data points.

Second, we established and tested the model by using the observed and satellite-derived data at the Alaska site. Because the satellite data were only available at coarse temporal resolutions, we used the observed daily NDVI and air temperature at the study site for the model parameterizations. Observed daily air temperature was used to determine fGPP(Tdaily), fRE(Tdaily), and fRE(GDD) in Equations (2) and (3). We used the observed data between 2004 and 2006 as calibration data and the data of 2003 as validation data for the Alaska site; the observed half-hourly data were averaged every 16 days corresponding to the model time step.

In addition to the observed data at the Alaska site, the model was also validated by using the flux data derived from the mature black spruce forest at the NOBS site. Using MOD13-NDVI and MOD11-LST at a pixel centered on the tower site at 16-day intervals from 2000 and 2005, we executed the model to calculate seasonal and interannual variations of GPP, RE, and NEE and then compared results with the observed ones to assess the model performance and uncertainties.

2.5.2. Regional-scale analyses

After parameterization and validation of the SCAPE model, we calculated regional-scale carbon fluxes using the calibrated model. We analyzed the seasonal and interannual variations in modeled carbon fluxes and satellite-based LST to reveal how seasonal variations in the temperature can affect the carbon budget for mature black spruce forests. The analysis period was from 2003 to 2006, when there are available flux data for model validation.

3. Results

3.1. Point-scale analyses

3.1.1. Linking ground and satellite observations

The observed bNDVI and MOD13-NDVI showed similar seasonal variations at the forest in Alaska (Figure 4); bNDVI was adjusted using a linear relationship between field-based bNDVI and MODIS NDVI (hereafter, bNDVIadjust is referred to as adjusted bNDVI) and averaged over 16 days, and observed bNDVI was converted to MOD13-NDVI by adjusted offset and slope values, where the slope (1.0) and offset (0.2) were determined by a simple linear regression between the observed bNDVI and MOD13-NDVI (r2 = 0.87). During the snow-melting season around the composite period of day of year (DOY) 113, both NDVIs suddenly increased and then gradually increased from 0.4 to 0.8. Then, the NDVIs gradually declined after the peak from June to late August and suddenly declined at the start of the snow period. Consequently, we confirmed that MOD13-NDVI retained phenological characteristics at this site. We used the bNDVIadjust to parameterize the SCAPE model, because use of the bNDVIadjust allowed for the examination of many data points on daily time scales, whereas MOD13-NDVI was limited to a 16-day composite of data. Although the NDVIs were strongly affected by snow cover in addition to the phenological variations, we used the total range of values for model parameterization, because the spring onset of photosynthesis was tightly linked with snow-cover conditions.

Both daily and minimum air temperature strongly correlated with that from MOD11-LST (r2 > 0.9; Figure 5), and we confirmed that the MOD11-LST products could be used to estimate near-surface temperature. The agreements in the field and satellite-based observations indicated the applicability of the conversion functions from NDVI to MOD13-NDVI and from LST to air temperature, although the temporal and spatial resolutions are different.

3.1.2. Parameters estimation for the satellite-based model

The relationship between the bNDVIadjust and observed GPP (GPPobs) was examined during the period 2004–06 (Figure 6a). We found that GPPobs is linearly correlated with the bNDVIadjust, but the relationship became saturated when the bNDVIadjust was approximately 0.8. Thus, we treated fGPP(NDVI) as a saturation at 0.8 of the bNDVIadjust. Then, we examined the model parameters for the temperature dependency of GPP (Figure 6b). The ratio of GPPobs to GPP calculated by the bNDVIadjust (GPPNDVI; GPP calculated by the regression in Figure 6a) exponentially increased with increasing air temperature, and the relationship became unclear when the temperature rose above 15°C; we treated fGPP(Tdaily) as a saturation when the daily air temperature was above 15°C.

For parameterization of RE, we first examined the relationship between observed RE and the bNDVIadjust (Figure 7a), showing that REobs exponentially increased with increase of the bNDVIadjust (r2 = 0.63). The close relationship between the bNDVIadjust and RE likely reflected the close correspondence between GPP and autotrophic respiration, with a favorable correspondence between GPP, litterfall, and heterotrophic respiration. Similarly to GPP, the relationship became unclear when the bNDVIadjust was above 0.8; fRE(NDVI) was treated as a saturation when the bNDVIadjust was above 0.8. Because RE was also highly sensitive to air temperature (Goulden et al. 1997; Lindroth et al. 1998; Ueyama et al. 2006), we found an exponential increase of RE with an increase in daily air temperature (Figure 7b). The result demonstrated that the residual between REobs and fRE(NDVI) exponentially increased with an increase in daily air temperature. Because the examined dependency separated out the effects of vegetation phenology by dividing REobs by fRE(NDVI), the exponential relationship could reflect the temperature dependence of metabolic activities. To include suppression of RE by frozen soil during the spring (Goulden et al. 1997; Ueyama et al. 2006), we examined the suppression ratio (Figure 7c), where GDD was chosen as an index identified early in the growing season. According to the residual between REobs and RENDVI*Tair (RE calculated by the regression in Figures 7a,b), RE was suppressed approximately 10% only for low GDD values (less than 600°C day). The period when GDD was less than 600°C roughly corresponded with the period when the active layer depth was less than 20 cm. The examined model parameters and the equations are summarized in Table 1.

3.1.3. Model validations

The calculated carbon fluxes agreed reasonably well with the observed ones in terms of seasonality and the amplitude during both validation and calibration periods at the Alaska site (Figure 8). Major phenological events, such as the start, end, and peak of the carbon fluxes, were clearly reproduced. During the validation period in 2003, the calculated fluxes also reproduced the observed ones within the estimated range of flux-partitioning error. The close agreement of the NEE in 2003 indicated that the calibrated model could be used in other years. Model outputs of 16-day GPP and RE were highly correlated with field-based estimates; slopes between observed and calculated GPP and RE were 1.03 (r2 = 0.85) and 1.04 (r2 = 0.81), respectively. On the other hand, the slope and correlation between model output and observed 16-day NEE were lower (0.74 and r2 = 0.51) than those of GPP and RE, although the calculated NEE were within the range of the standard deviations of the observed 16-day NEE. This lower correlation was probably because NEE was a small difference between the large two terms of GPP and RE. The RMSE of calculated GPP and RE were 3.0 g CO2 m−2 day−1 and 2.7 g CO2 m−2 day−1, respectively, whereas that of NEE was 2.0 g CO2 m−2 day−1. The observed CO2 budgets of 2003, 2005, and 2006 were 137, 151, and 68 g CO2 m−2 yr−1, whereas the calculated ones were 133, 22, and 22 g CO2 m−2 yr−1 (Figure 9). The interannual variations of calculated NEE were properly reproduced in the model. RMSE of the calculated annual GPP and RE were 153 and 171 g CO2 m−2 yr−1, respectively, whereas that of NEE was 70 g CO2 m−2 yr−1.

As an independent validation, we also compared the model output to the observed results from the other mature black spruce forest at the NOBS site (Figure 10; Goulden et al. 1998; Dunn et al. 2006). The parameterized model properly reproduced the seasonal variations of the carbon fluxes at the NOBS site, as well as for the Alaska site. Model outputs of 16-day GPP and RE were highly correlated with field-based estimates, where slopes were 1.00 (r2 = 0.87) and 1.04 (r2 = 0.84), respectively. The slope and correlation between model output and observed 16-day NEE were lower (0.54 with r2 = 0.29) than those of GPP and RE. The RMSE of calculated fluxes were almost the same as that for the Alaska site: 2.9 g CO2 m−2 day−1 for GPP, 2.6 g CO2 m−2 day−1 for RE, and 2.2 g CO2 m−2 day−1 for NEE. The interannual variations of GPP and RE were also successfully calculated (Figure 11). Interannual variation of calculated NEE was also similar to the observed NEE at the NOBS site. The lower RMSE compared with that for the Alaska site was estimated in the NOBS site on an annual time scale: 20.9 g CO2 m−2 yr−1 for GPP, 20.8 g CO2 m−2 yr−1 for RE, and 3.6 g CO2 m−2 yr−1 for NEE. This concordance in data suggests that the SCAPE model parameterized at the black spruce forest in Alaska is reasonably applicable for estimations of the regional carbon fluxes within mature black spruce forests.

We assessed the model uncertainties in the air temperature estimation from MOD11-LST. Considering the RMSE for the daily air temperature of 3.1°C (Figure 5), we calculated seasonal and annual carbon fluxes at the black spruce forests in Alaska by applying a ±3.1°C bias and found that the RMSE for NEE was 1.8 g CO2 m−2 day−1 at a seasonal scale and 102.8 g CO2 m−2 yr−1 at an annual scale. We also conducted the same sensitivity analysis for the minimum air temperature, which resulted in an RMSE for minimum air temperature that was lower compared with that for daily air temperature. Finally, we concluded that the uncertainties from the air temperature estimation resulted in errors of 284.4 g CO2 m−2 yr−1 for GPP, 261.3 g CO2 m−2 yr−1 for RE, and 102.8 g CO2 m−2 yr−1 for NEE on an annual scale.

3.2. Regional-scale analyses

The spatial distributions of the carbon budget between 2003 and 2006 showed that GPP, RE, and NEE (carbon sink strength) were greater in the western part than in the eastern part of the region (Figure 12). The spatial variations of GPP were similar to that of RE, where the carbon fluxes were also greater in the western part than in the eastern part. Those spatial variations could be due to the distinct climatic region in Alaska. The western region of the Alaska boreal forest, where mean annual air temperature is above freezing and annual precipitation is between 400 and 560 mm, is influenced by the Bering Sea, whereas the northern interior region to the eastern Yukon and Tanana river valleys, where the annual air temperature is −5° to −7°C and annual precipitation is 215 and 200 mm, are at higher elevations than the western region (Hinzman et al. 2006). Thus, the calculated spatial distributions could reflect that the western region with a warm and wetter environment featured high productivity and a small carbon sink, whereas the eastern region with a characteristically cold and dry environment featured low productivity with a small carbon source.

The annual GPP and RE over the study area during the 4-yr period were 1.76 and 1.86 kg CO2 m−2 yr−1. Considering uncertainties in eddy covariance measurements (e.g., Falge et al. 2001) and our SCAPE model (previous section), the mature black spruce forests in Alaska were almost carbon neutral during the study period. The estimated regional carbon budget might show a source or sink, considering the model uncertainties. In spite of the uncertainties, the estimated carbon budget was consistent with another modeling study (Deng et al. 2007). Deng et al. (Deng et al. 2007) estimated the global carbon flux by using an inverse model and found that the Alaska region (regions 2 and 3 in their study) acted as a small carbon source, despite considerable uncertainties. The interannual variations of GPP, RE, and NEE calculated as coefficients of variances during the four years were 9%, 8%, and 20%, respectively. Therefore, GPP and RE almost equally contributed to interannual variations of the carbon budget.

The calculated seasonal cycle of the regional carbon fluxes between 2003 and 2006 are shown in Figure 13. The calculated regional mean of GPP started to increase during late April (DOY 113) in 2003, 2004, and 2005, whereas the start in 2006 was delayed to early May (DOY 129) because of the cooler spring that year (Figure 13e). The peak of calculated GPP ranged from 18.5 to 15.4 g CO2 m−2 day−1 and occurred from late June to mid-July. Afterward, the GPP gradually declined and became closer to zero in early October. The calculated regional mean of RE started to increase in the same period at the start of GPP, but the peak of RE was delayed 16 days. The start of negative NEE, namely, the net uptake of CO2, appeared in May from DOY 129 to DOY 145. The peak of the uptake occurred with a maximum NEE of about −4.8 g CO2 m−2 day−1 in June, from DOY 161 to DOY 177, which was about one month earlier than the peak of GPP. The peak period of NEE and the magnitude were similar to previous observations at mature black spruce forests in Canada (Dunn et al. 2006; Bergeron et al. 2007) and Alaska (Ueyama et al. 2006; Welp et al. 2007). After the peak, NEE rapidly declined, whereas GPP had an increasing trend as RE increased in response to the rise in air temperature. Because NEE was the balance of GPP and RE, an earlier increase in GPP compared with RE resulted in the earlier peak period of NEE.

4. Discussion

4.1. Controlling factors of GPP and RE at seasonal and interannual time scales

Our SCAPE model simplified calculation of the carbon fluxes over mature black spruce forests in interior Alaska by using the empirical relationships and the satellite data. With only two satellite-based inputs, NDVI and LST, the model captured the observed seasonal and interannual variations of the carbon fluxes well (Figures 8, 9). The model established at the mature black spruce forest in Alaska was also applicable to the other monitored mature black spruce forest (Figures 10, 11). This result suggests that seasonality of NDVI and temperature (air and/or surface temperatures) played an important role in the carbon fluxes, which is consistent with the observation studies. Jarvis and Linder (Jarvis and Linder 2000) reported that water supply associated with increase in spring soil temperature is a key variable to start vegetation growth in boreal forests. Our simple model could capture the spring onset by incorporating the suppression effect of nighttime freezing temperature (input of nighttime LST) and linking seasonal variation between NDVI and GPP at the canopy scale. Bergeron et al. (Bergeron et al. 2007) reported that the relationship between the carbon fluxes and environmental factors were similar within three black spruce forests in Canada on a monthly time scale and that the monthly GPP and RE could be mostly explained by the air and soil temperature variations, respectively. Similar results were also reported from a climatological analysis (Suzuki et al. 2006), in which the temperature dominated the activity of boreal and tundra vegetation.

4.2. Sensitivity of regional carbon budget

The responses of boreal forests to the temperature change have received considerable attention, because a significant warming trend in high latitudes has been observed in the past few decades (Chapin et al. 2005). Several modeling studies reported that future warming would increase the terrestrial carbon sequestration, because the spring warming advances the start of growing season (Euskirchen et al. 2006; Kimball et al. 2006). In the warmer spring of 2004 (DOY 129–161 in Figure 13), the calculated sink strength of the black spruce forests in Alaska increased in comparison to the cool spring in 2006. This result is consistent with the observed trend in the black spruce forest in Alaska (Welp et al. 2007) and a process-based modeling study (Ueyama et al. 2009), indicating that spring warming potentially increases the sink strength. On the other hand, our results show that the warmer midsummer in 2004 (DOY 177–209 in Figure 13) decreased the sink strength, because the warming stimulated RE rather than GPP in the summer. The opposite response to spring and midsummer temperature could be explained by the optimal temperatures of the carbon fluxes. The optimal temperature for GPP in boreal forests ranged from 15° to 20°C (e.g., Chen et al. 1999; Goulden et al. 1997; Lindroth et al. 1998), as well as in the black spruce forest in our site (Ueyama et al. 2006), whereas RE increased with rise in air temperature within the range of the observed temperature. Because the daily air temperature in interior Alaska was nearly optimal for GPP in summer months (Simpson et al. 2002; Shulski and Wendler 2007), sensitivity of GPP to air temperature could be lower than that of RE in summer. In the autumn (DOY 273; cool in 2004 and warm in 2006 in Figure 13), calculated results demonstrated that air temperature increased RE rather than GPP. These results suggested that the future carbon budget of mature black spruce forests in Alaska is dependent on the delicate changes in seasonal temperature variation. According to our remote sensing estimation, the mature black spruce forests will increase the sink strength during spring warming and decrease it in summer or autumn warming. These results were also consistent with simulation results from a calibrated BIOME-BGC model at the black spruce forest in Alaska (Ueyama et al. 2009).

4.3. Limitations and future implications

Although our model successfully calculated the seasonal and interannual variations in terrestrial carbon cycle, we examined several limitations of the applicability. First, the observed GPP and RE were derived from the NEE measurements. The eddy covariance data contained several gap-filled values, which potentially induced uncertainty in the model calibration. Because our black spruce forests are located in high latitudes, where the nighttime length was shorter in the summer, the estimated GPP and RE from the observations remain uncertain. The current model could not reproduce the interannual variations of GPP and RE at the black spruce forest in Alaska (Figure 9), whereas the model captured them well at the Canadian black spruce forest (Figure 11), suggesting that the shorter nighttime period could lead to errors in the GPP and RE estimation at the Alaska site. These errors might be the primary cause for misrepresentations of interannual variations for GPP and RE in our black spruce forests (Figure 9). This kind of error due to the flux partitioning is a primary uncertainty in current empirical model calibrations as well as process-based model validation and should be reduced by a more sophisticated partitioning algorithm in a future study.

Second, the modeling of the response to differing water conditions could improve the understanding of the carbon cycle in boreal forests, although the response of the carbon fluxes to water conditions was not clear in Alaska black spruce forests (Welp et al. 2007). Several investigations reported that changes in the water condition affected carbon fluxes and thus the carbon sequestration in boreal forests (Iwashita et al. 2005; Dunn et al. 2006). Our SCAPE model did not include the sophisticated suppression algorithm by the water limitation, but these effects were possibly incorporated in the model algorithm. Bunn and Goetz (Bunn and Goetz 2006) reported that the continental summer drought was detected in North America by the satellite-derived NDVI. Thus, the possible decline of carbon fluxes by drought effect was probably incorporated in the NDVI inputs. To understand the linkage between the carbon and water fluxes, indices related to water conditions (Xiao et al. 2004; Claudio et al. 2006) and improvement of microwave remote sensing will improve our understanding of the carbon fluxes in northern ecosystems.

This study simply linked the observed carbon fluxes with satellite-derived NDVI and LST. Although our algorithm is suitable for homogeneous area, the applicability to heterogeneous area remains uncertain. Further examination combining with high-resolution satellite sensors, such as Landsat, IKONOS, and Advanced Visible and Near Infrared Radiometer-II (AVNIR2), should be conducted to verify our methodology on heterogeneous landscape.

Although our study focused on mature black spruce forests in Alaska, several studies reported that the carbon flux in the North American boreal biome was significantly dependent on the stand age (e.g., Welp et al. 2007; Goulden et al. 2006); young stands could have higher carbon sequestration capacity compared with the mature forests. To accurately understand the carbon budget in the Alaskan boreal forest, further improvements are required, including the effect of stand age and other biomes.

Our study is an early attempt to scale up the eddy covariance data to regional carbon fluxes by using satellite remote sensing data. Because our satellite-based empirical model was driven by only NDVI and LST, this simplicity allows us to apply our model to both temporal and spatial scale analyses combined with various remote sensed data, such as the Advanced Very High Resolution Radiometer (AVHRR). This approach could also be applicable to various terrestrial ecosystems, if the dominant controlling factors are determined by satellite remote sensing; we need further model validation in other ecosystems. Recently, the continuous eddy covariance data has been made available through a global network of micrometeorological flux measurements, FLUXNET (Baldocchi et al. 2001). Integration of our methodology to FLUXNET could provide a useful dataset to improve our understanding of not only the carbon budget in high-latitude regions but also of the global carbon cycle.

5. Conclusions

Synthesis of satellite remote sensing and empirical modeling, based on field observations, has improved the understanding of regional carbon fluxes over mature black spruce forests in Alaska. The model successfully reproduced the seasonal variations of observed carbon fluxes and the interannual carbon budget in the black spruce forests in Alaska and Canada, using satellite-derived NDVI and LST. These results suggest that seasonality of NDVI and temperature (air and/or surface temperatures) played an important role regarding the carbon fluxes of mature black spruce forests and that the sensitivity to the input parameters might be similar within mature black spruce forests.

The estimated regional averages of GPP and RE between 2003 and 2006 were 1.76 ± 0.28 and 1.86 ± 0.26 kg CO2 m−2 yr−1, respectively, such that the forest was almost carbon neutral during the study period. The spatial distributions of calculated carbon fluxes reflected climate and topography over Alaska. According to remote sensing estimations, the future carbon budget of mature black spruce forests in Alaska is sensitive to the delicate changes in warming periods during spring, summer, and autumn; the black spruce forests will increase the sink strength in spring warming and decrease the sink during summer or autumn warming.

The methodology of this study is applicable for scaling up of the observed fluxes to the regional scale by applying satellite-derived parameters related to observed carbon fluxes. Our satellite-based empirical modeling to estimate regional carbon fluxes will be improved by reducing the uncertainties resulting from accuracy of the field measurements and response to water conditions, as well as the effect of stand age.

Acknowledgments

This study was supported by the IJIS (IARC/JAXA information systems) and Carbon Cycle Study Program of IARC/NSF by the U.S. National Science Foundation. We thank Dr. M. Hori and Dr. G. Kadosaki of JAXA; Dr. H. Iwata of University of Alaska Fairbanks; Dr. S. Yamamoto, Dr. T. Iwata, T. Kitamoto, and T. Date of Okayama University; and two anonymous reviewers for their beneficial comments. We also thank T. Suzuki of Fukushima University for supporting data processing and T. Saito, N. Rozell, and other staff members with IARC for their support. We thank personnel of the Dr. S. C. Wofsy Group of Harvard University Atmospheric Sciences and Dr. B. D. Amiro of the University of Manitoba (Fluxnet-Canada Research Network) for providing the observed carbon fluxes at the northern old black spruce forests in Canada.

REFERENCES

  • ACIA 2004. Impacts of a Warming Arctic: Arctic Climate Impact Assessment. Cambridge University Press, 1020 pp.

  • Angert, A. , S. Biraud , C. Bonfils , C. C. Henning , W. Buermann , J. Pinzon , C. J. Tucker , and I. Fung . 2005. Drier summers cancel out the CO2 uptake enhancement induced by warmer springs. Proc. Natl. Acad. Sci. 102:10 82310 827.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. Coauthors 2001. FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bull. Amer. Meteor. Soc. 82:24152434.

    • Search Google Scholar
    • Export Citation
  • Bergeron, O. , H. A. Margolis , A. Black , C. Coursolle , A. L. Dunn , A. G. Barr , and S. C. Wofsy . 2007. Comparison of carbon dioxide fluxes over three boreal black spruce forests in Canada. Global Change Biol. 13:89107.

    • Search Google Scholar
    • Export Citation
  • Bunn, A. G. and S. J. Goetz . 2006. Trends in satellite-observed circumpolar photosynthetic activity from 1982 to 2003: The influence of seasonality, cover type, and vegetation density. Earth Interactions 12.[Available online at http://EarthInteractions.org].

    • Search Google Scholar
    • Export Citation
  • Chapin III, F. S. , P. A. Matson , and H. A. Mooney . 2002. Principles of Terrestrial Ecosystem Ecology. Springer-Verlag, 436 pp.

  • Chapin III, F. S. Coauthors 2005. Role of land-surface changes in arctic summer warming. Science 310:657660.

  • Chapman, W. L. and J. E. Walsh . 2007. Simulations of arctic temperature and pressure by global coupled models. J. Climate 20:609632.

    • Search Google Scholar
    • Export Citation
  • Chen, B. , J. M. Chen , and W. Ju . 2007. Remote sensing-based ecosystem-atmosphere simulation scheme (EASS): Model formulation and test with multiple-year data. Ecol. Modell. 209:277300.

    • Search Google Scholar
    • Export Citation
  • Chen, W. J. Coauthors 1999. Effects of climatic variability on the annual carbon sequestration by a boreal aspen forest. Global Change Biol. 5:4153.

    • Search Google Scholar
    • Export Citation
  • Claudio, H. C. Coauthors 2006. Monitoring drought effects on vegetation water content and fluxes in chaparral with the 970 nm water band index. Remote Sens. Environ. 103:304311.

    • Search Google Scholar
    • Export Citation
  • Clein, J. S. , A. D. McGuire , X. Zhang , D. W. Kicklighter , J. M. Melillo , S. C. Wofsy , P. G. Jarvis , and J. M. Massheder . 2002. Historical and projected carbon balance of mature black spruce ecosystems across North America: The role of carbon-nitrogen interaction. Plant Soil 242:1532.

    • Search Google Scholar
    • Export Citation
  • Date, T. , M. Ueyama , Y. Harazono , Y. Ota , T. Iwata , and S. Yamamoto . 2009. Satellite observations of decadal scale CO2 fluxes over black spruce forests in Alaska associated with climate variability. J. Agric. Meteor. 65:4760.

    • Search Google Scholar
    • Export Citation
  • Deng, F. , J. M. Chen , M. Ishizawa , C-W. Yuen , G. Mo , K. Higuchi , D. Chan , and S. Maksyutov . 2007. Global monthly CO2 flux inversion with a focus over North America. Tellus 59B:179190.

    • Search Google Scholar
    • Export Citation
  • Dixon, R. K. , S. Brown , R. A. Houghton , A. M. Solomon , M. C. Trexler , and J. Wisniewski . 1994. Carbon pools and flux of global forest ecosystems. Science 263:185190.

    • Search Google Scholar
    • Export Citation
  • Dunn, A. , C. C. Barford , S. C. Wofsy , M. L. Goulden , and B. C. Daube . 2006. A long-term record of carbon exchange in a boreal black spruce forest: Means, responses to interannual variability, and decadal trends. Global Change Biol. 12:114.

    • Search Google Scholar
    • Export Citation
  • Euskirchen, E. S. Coauthors 2006. Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Global Change Biol. 12:731750.

    • Search Google Scholar
    • Export Citation
  • Falge, E. Coauthors 2001. Gap filling strategies for defensible annual sums of net ecosystem exchange. Agric. For. Meteor. 107:4369.

    • Search Google Scholar
    • Export Citation
  • Fleming, M. 1997. A statewide vegetation map of Alaska using phenological classification of AVHRR data. Proc. Second Circumpolar Arctic Vegetation Mapping Workshop and the CAVM-North America Workshop, Anchorage, Alaska, U.S. Geological Survey, 25–26.

    • Search Google Scholar
    • Export Citation
  • Goulden, M. L. , B. C. Daube , S-M. Fan , D. J. Sutton , A. Bazzaz , J. W. Munger , and S. C. Wofsy . 1997. Physiological responses of a black spruce forest to weather. J. Geophys. Res. 102:28 98728 996.

    • Search Google Scholar
    • Export Citation
  • Goulden, M. L. Coauthors 1998. Sensitivity of boreal forest carbon balance to soil thaw. Science 279:214217.

  • Goulden, M. L. , G. C. Winston , A. M. S. McMillan , M. E. Litvak , E. L. Read , A. V. Rocha , and R. Elliot . 2006. An eddy covariance mesonet to measure the effect of forest age on land-atmosphere exchange. Global Change Biol. 12:21462162.

    • Search Google Scholar
    • Export Citation
  • Heinsch, F. A. Coauthors 2006. Evaluation of remote sensing based terrestrial productivity from MODIS using regional tower eddy flux network observations. IEEE Trans. Geosci. Remote Sens. 44:19081925.

    • Search Google Scholar
    • Export Citation
  • Hinzman, L. D. Coauthors 2005. Evidence and implications of recent climate change in northern Alaska and other arctic regions. Climatic Change 72:251298.

    • Search Google Scholar
    • Export Citation
  • Hinzman, L. D. , L. A. Viereck , P. C. Adams , V. E. Romanovsky , and K. Yoshikawa . 2006. Climate and permafrost dynamics of the Alaskan boreal forest. Alaska’s Changing Boreal Forest, F. S. Chapin et al., Eds., Oxford Press, 39–61.

    • Search Google Scholar
    • Export Citation
  • Hirata, R. , T. Hirano , J. Mogami , Y. Fujinuma , K. Inukai , N. Saigusa , and S. Yamamoto . 2005. CO2 flux measured by an open-path system over a larch forest during the snow-covered season. Phyton 45:347351.

    • Search Google Scholar
    • Export Citation
  • Huemmrich, K. F. , T. A. Black , P. G. Jarvis , J. H. McCaughey , and F. G. Hall . 1999. High temporal resolution NDVI phenology from micrometeorological radiation sensors. J. Geophys. Res. 104:27 93527 944.

    • Search Google Scholar
    • Export Citation
  • Huemmrich, K. F. Coauthors 2009. Remote sensing of tundra gross ecosystem productivity and light use efficiency under varying temperature and moisture conditions. Remote Sens. Environ. 114:481489.

    • Search Google Scholar
    • Export Citation
  • Huete, A. , K. Didan , T. Miura , E. P. Rodriguez , X. Gao , and L. G. Ferreira . 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83:195213.

    • Search Google Scholar
    • Export Citation
  • Iwashita, H. , N. Saigusa , S. Murayama , H. McCaughery , A. Black , A. Barr , K. Higuchi , and S. Yamamoto . 2005. Effect of soil water content on carbon dioxide flux at a sparse-canopy forest in the Canadian boreal ecosystem. J. Agric. Meteor. 61:131141.

    • Search Google Scholar
    • Export Citation
  • Jarvis, P. and S. Linder . 2000. Constraints to growth of boreal forests. Nature 405:904905.

  • Kimball, J. S. , P. E. Thornton , M. A. White , and S. W. Running . 1997. Simulating forest productivity and surface-atmosphere carbon exchange in the BOREAS study region. Tree Physiol. 17:589599.

    • Search Google Scholar
    • Export Citation
  • Kimball, J. S. , M. Zhao , K. C. McDonald , and S. W. Running . 2006. Satellite remote sensing of terrestrial net primary production of the pan-Arctic basin and Alaska. Mitigation Adapt. Strategies Global Change 11:783804.

    • Search Google Scholar
    • Export Citation
  • Kitamoto, T. , M. Ueyama , Y. Harazono , T. Iwata , and S. Yamamoto . 2007. Applications of NOAA/AVHRR and observed fluxes to estimate carbon budget at Alaska black spruce forests. J. Agric. Meteor. 63:171183.

    • Search Google Scholar
    • Export Citation
  • Lindroth, A. , A. Grelle , and A-S. Morén . 1998. Long-term measurements of boreal forest carbon balance reveal large temperature sensitivity. Global Change Biol. 4:443450.

    • Search Google Scholar
    • Export Citation
  • Mahadevan, P. Coauthors 2008. A satellite-based biosphere parameterization for net ecosystem CO2 exchange: Vegetation Photosynthesis and Respiration Model (VPRM). Global Biogeochem. Cycles 22:GB2005. doi:10.1029/2006GB002735.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D. , J. S. Clein , J. M. Melillo , D. W. Kicklighter , R. A. Meier , C. J. Vorosmaty , and M. C. Serreze . 2000. Modelling carbon responses of tundra ecosystems to historical and projected climate: Sensitivity of pan-Arctic carbon storage to temporal and spatial variation in climate. Global Change Biol. 6:141159.

    • Search Google Scholar
    • Export Citation
  • Moore, C. J. 1986. Frequency response corrections for eddy correlation systems. Bound.-Layer Meteor. 37:1735.

  • Piao, S. Coauthors 2008. Net carbon dioxide losses of northern ecosystems in response to autumn warming. Nature 451:4952.

  • Rahman, A. F. , J. A. Gamon , D. A. Fuentes , D. A. Roberts , and D. Prentiss . 2001. Modeling spatially distributed ecosystem flux of boreal forest using hyperspectral indices from AVIRIS imagery. J. Geophys. Res. 106:33 57933 591.

    • Search Google Scholar
    • Export Citation
  • Running, S. W. , R. R. Namani , F. A. Heinsch , M. Zhao , M. Reeves , and H. Hashimoto . 2004. A continuous satellite-derived measure of global terrestrial primary production. BioScience 54:547560.

    • Search Google Scholar
    • Export Citation
  • Sasai, T. , K. Ichii , Y. Yamaguchi , and R. Nemani . 2005. Simulating terrestrial carbon fluxes using the new biosphere model “biosphere model integrating eco-physiological and mechanistic approaches using satellite data” (BEAMS). J. Geophys. Res. 110:G02014. doi:10.1029/2005JG000045.

    • Search Google Scholar
    • Export Citation
  • Shulski, M. and G. Wendler . 2007. The Climate of Alaska. University of Alaska Press, 216 pp.

  • Simpson, J. J. , G. L. Hufford , M. D. Fleming , J. S. Berg , and J. B. Ashton . 2002. Long-term climate patterns in Alaskan surface temperature and precipitation and their biological consequences. IEEE Trans. Geosci. Remote Sens. 40:11641184.

    • Search Google Scholar
    • Export Citation
  • Suzuki, R. , J. Xu , and K. Motoya . 2006. Global analysis of satellite-derived vegetation index related to climatological wetness and warmth. Int. J. Climatol. 26:425438.

    • Search Google Scholar
    • Export Citation
  • Ueyama, M. , Y. Harazono , E. Ohtaki , and A. Miyata . 2006. Controlling factors on the inter-annual CO2 budget at a subarctic black spruce forest in interior Alaska. Tellus 58B:491501.

    • Search Google Scholar
    • Export Citation
  • Ueyama, M. , Y. Harazono , Y. Kim , and N. Tanaka . 2009. Response of the carbon cycle in sub-arctic black spruce forests to climate change: Reduction of a carbon sink related to the sensitivity of heterotrophic respiration. Agric. For. Meteor. 149:582602.

    • Search Google Scholar
    • Export Citation
  • Ueyama, M. Coauthors 2010. Simulating carbon and water cycles of larch forests in East Asia by the BIOME-BGC model with AsiaFlux data. Biogeosciences 7:956977.

    • Search Google Scholar
    • Export Citation
  • Vogel, J. G. , D. W. Valentine , and R. W. Ruess . 2005. Soil and root respiration in mature Alaskan black spruce forests that vary in soil organic matter decomposition rates. Can. J. For. Res. 35:161174.

    • Search Google Scholar
    • Export Citation
  • Vourlitis, G. L. , J. Aillie , W. C. Oechel , A. Hope , D. Stow , and R. Engstrom . 2003. Spatial variation in regional CO2 exchange for the Kuparuk river basin, Alaska over the summer growing season. Global Change Biol. 9:930041.

    • Search Google Scholar
    • Export Citation
  • Wan, Z. , Y. Zhang , Q. Zhang , and Z-I. Li . 2002. Validation of the land-surface temperature products retrieved from Terra Moderate Resolution Imaging Spectroradiometer data. Remote Sens. Environ. 83:163180.

    • Search Google Scholar
    • Export Citation
  • Webb, E. K. , G. I. Pearman , and R. Leuning . 1980. Correction of flux measurements for density effects due to heat and water vapour transfer. Quart. J. Roy. Meteor. Soc. 106:85100.

    • Search Google Scholar
    • Export Citation
  • Welp, L. R. , J. T. Randerson , and H. P. Liu . 2007. The sensitivity of carbon fluxes to spring warming and summer drought depends on plant functional type in boreal forest ecosystems. Agric. For. Meteor. 147:172185.

    • Search Google Scholar
    • Export Citation
  • Xiao, X. , D. Hollinger , J. Aber , M. Goltz , E. A. Davidson , Q. Zhang , and B. Moore III . 2004. Satellite-based modeling of gross primary production in an evergreen needleleaf forest. Remote Sens. Environ. 89:519534.

    • Search Google Scholar
    • Export Citation
  • Yamaji, T. Coauthors 2008. Scaling-up technique for net ecosystem productivity of deciduous broadleaved forests in Japan using MODIS data. Ecol. Res. 23:765775.

    • Search Google Scholar
    • Export Citation
  • Yang, F. Coauthors 2007. Developing a continental-scale measure of gross primary production by combining MODIS and AmeriFlux data through support vector machine approach. Remote Sens. Environ. 110:109122.

    • Search Google Scholar
    • Export Citation

Figure 1.
Figure 1.

Study area and spatial distributions of black spruce forests in interior Alaska. The star represents the observation site at Fairbanks.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 2.
Figure 2.

Photograph showing the Alaska site in (a) summer and (b) winter and (c) the eddy covariance instruments. (d) Aerial photography of the site obtained from Google Earth, where the dashed circle and box show the typical footprint area, 10 times the measurement height, of tower flux at 6 m and a 250-m grid centered on the tower site, respectively. The dominant wind direction was east and southwest.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 3.
Figure 3.

Schematic diagram of a scaling approach using flux observations and MODIS products.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 4.
Figure 4.

Time variations of adjusted bNDVI and MOD13-NDVI at the Alaska site. The bNDVI was adjusted by using a linear relationship between field-based bNDVI and MOD13-NDVI and averaged over 16 days.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 5.
Figure 5.

Comparison (a) between mean of daytime and nighttime MOD11-LST and observed daily air temperature and (b) between nighttime MOD11-LST and observed minimum air temperature. The observed data are from nine weather stations of the NCDC and the tower site in interior Alaska.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 6.
Figure 6.

Relationships between (a) adjusted bNDVI and GPPobs and (b) daily air temperature and ratio of observed GPP to estimated GPPNDVI [GPP calculated by the regression in (a)] at the Alaska site. (a) The total data points of 532 were binned by bNDVI into classes of 0.1, whereas (b) the total data points of 460 were binned by air temperature into classes of 5°C. The points and vertical bars represent the average and standard deviation in each class, respectively.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 7.
Figure 7.

Relationships between (a) adjusted bNDVI and observed RE (REobs), (b) daily air temperature and ratio of REobs to estimated RE by adjusted bNDVI [RENDVI; RE calculated by the regression in (a)], and (c) GDD and the ratio of observed RE to estimated RE by NDVI and the temperature [RENDVI*Tair; RE calculated by the stepwise regressions in (a),(b)] at the Alaska site. (a) The total data points of 677 were binned by bNDVI into classes of 0.1, and the total data points of 756 and 280 were binned by each (b) air temperature and (c) GDD class, respectively. The points and vertical bars represent the average and standard deviation in each class, respectively.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 8.
Figure 8.

Time variations of observed (solid lines) and modeled (dashed lines with dots) (a) GPP, (b) RE, and (c) NEE at the black spruce forest in Alaska. Gray areas represent estimation error for observed GPP and RE, and standard deviation within each 16-day for observed NEE. The estimation error for GPP and RE were assumed as standard deviations of RE within the each temperature range of Q10 relationship.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 9.
Figure 9.

Interannual variations of (a) observed and modeled GPP, RE, and NEE and (b) cumulative annual NEE between 2003 and 2006 at the black spruce forest in Alaska.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 10.
Figure 10.

Time variations of observed (solid lines) and modeled (dashed lines with dots) (a) GPP, (b) RE, and (c) NEE at the mature black spruce forest in Canada. Gray areas represent standard deviation within each 16-day interval for observed GPP, RE, and NEE.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 11.
Figure 11.

Interannual variation of (a) observed and modeled GPP, RE, and NEE, and (b) cumulative NEE between 2000 and 2005 at the mature black spruce forest in Canada.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 12.
Figure 12.

The spatial distributions of (a) GPP, (b) RE, and (c) NEE over black spruce forests in Alaska between 2003 and 2006 (kg CO2 m−2 yr−1).

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Figure 13.
Figure 13.

The seasonal variations of (a) GPP, (b) RE, (c) NEE, (d) MOD13-NDVI, and (e) estimated daily air temperature by MOD11-LST over the study area between 2003 and 2006.

Citation: Earth Interactions 14, 13; 10.1175/2010EI319.1

Table 1.

List of equations and variables of the SCAPE model.

Table 1.
Save
  • ACIA 2004. Impacts of a Warming Arctic: Arctic Climate Impact Assessment. Cambridge University Press, 1020 pp.

  • Angert, A. , S. Biraud , C. Bonfils , C. C. Henning , W. Buermann , J. Pinzon , C. J. Tucker , and I. Fung . 2005. Drier summers cancel out the CO2 uptake enhancement induced by warmer springs. Proc. Natl. Acad. Sci. 102:10 82310 827.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. Coauthors 2001. FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bull. Amer. Meteor. Soc. 82:24152434.

    • Search Google Scholar
    • Export Citation
  • Bergeron, O. , H. A. Margolis , A. Black , C. Coursolle , A. L. Dunn , A. G. Barr , and S. C. Wofsy . 2007. Comparison of carbon dioxide fluxes over three boreal black spruce forests in Canada. Global Change Biol. 13:89107.

    • Search Google Scholar
    • Export Citation
  • Bunn, A. G. and S. J. Goetz . 2006. Trends in satellite-observed circumpolar photosynthetic activity from 1982 to 2003: The influence of seasonality, cover type, and vegetation density. Earth Interactions 12.[Available online at http://EarthInteractions.org].

    • Search Google Scholar
    • Export Citation
  • Chapin III, F. S. , P. A. Matson , and H. A. Mooney . 2002. Principles of Terrestrial Ecosystem Ecology. Springer-Verlag, 436 pp.

  • Chapin III, F. S. Coauthors 2005. Role of land-surface changes in arctic summer warming. Science 310:657660.

  • Chapman, W. L. and J. E. Walsh . 2007. Simulations of arctic temperature and pressure by global coupled models. J. Climate 20:609632.

    • Search Google Scholar
    • Export Citation
  • Chen, B. , J. M. Chen , and W. Ju . 2007. Remote sensing-based ecosystem-atmosphere simulation scheme (EASS): Model formulation and test with multiple-year data. Ecol. Modell. 209:277300.

    • Search Google Scholar
    • Export Citation
  • Chen, W. J. Coauthors 1999. Effects of climatic variability on the annual carbon sequestration by a boreal aspen forest. Global Change Biol. 5:4153.

    • Search Google Scholar
    • Export Citation
  • Claudio, H. C. Coauthors 2006. Monitoring drought effects on vegetation water content and fluxes in chaparral with the 970 nm water band index. Remote Sens. Environ. 103:304311.

    • Search Google Scholar
    • Export Citation
  • Clein, J. S. , A. D. McGuire , X. Zhang , D. W. Kicklighter , J. M. Melillo , S. C. Wofsy , P. G. Jarvis , and J. M. Massheder . 2002. Historical and projected carbon balance of mature black spruce ecosystems across North America: The role of carbon-nitrogen interaction. Plant Soil 242:1532.

    • Search Google Scholar
    • Export Citation
  • Date, T. , M. Ueyama , Y. Harazono , Y. Ota , T. Iwata , and S. Yamamoto . 2009. Satellite observations of decadal scale CO2 fluxes over black spruce forests in Alaska associated with climate variability. J. Agric. Meteor. 65:4760.

    • Search Google Scholar
    • Export Citation
  • Deng, F. , J. M. Chen , M. Ishizawa , C-W. Yuen , G. Mo , K. Higuchi , D. Chan , and S. Maksyutov . 2007. Global monthly CO2 flux inversion with a focus over North America. Tellus 59B:179190.

    • Search Google Scholar
    • Export Citation
  • Dixon, R. K. , S. Brown , R. A. Houghton , A. M. Solomon , M. C. Trexler , and J. Wisniewski . 1994. Carbon pools and flux of global forest ecosystems. Science 263:185190.

    • Search Google Scholar
    • Export Citation
  • Dunn, A. , C. C. Barford , S. C. Wofsy , M. L. Goulden , and B. C. Daube . 2006. A long-term record of carbon exchange in a boreal black spruce forest: Means, responses to interannual variability, and decadal trends. Global Change Biol. 12:114.

    • Search Google Scholar
    • Export Citation
  • Euskirchen, E. S. Coauthors 2006. Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Global Change Biol. 12:731750.

    • Search Google Scholar
    • Export Citation
  • Falge, E. Coauthors 2001. Gap filling strategies for defensible annual sums of net ecosystem exchange. Agric. For. Meteor. 107:4369.

    • Search Google Scholar
    • Export Citation
  • Fleming, M. 1997. A statewide vegetation map of Alaska using phenological classification of AVHRR data. Proc. Second Circumpolar Arctic Vegetation Mapping Workshop and the CAVM-North America Workshop, Anchorage, Alaska, U.S. Geological Survey, 25–26.

    • Search Google Scholar
    • Export Citation
  • Goulden, M. L. , B. C. Daube , S-M. Fan , D. J. Sutton , A. Bazzaz , J. W. Munger , and S. C. Wofsy . 1997. Physiological responses of a black spruce forest to weather. J. Geophys. Res. 102:28 98728 996.

    • Search Google Scholar
    • Export Citation
  • Goulden, M. L. Coauthors 1998. Sensitivity of boreal forest carbon balance to soil thaw. Science 279:214217.

  • Goulden, M. L. , G. C. Winston , A. M. S. McMillan , M. E. Litvak , E. L. Read , A. V. Rocha , and R. Elliot . 2006. An eddy covariance mesonet to measure the effect of forest age on land-atmosphere exchange. Global Change Biol. 12:21462162.

    • Search Google Scholar
    • Export Citation
  • Heinsch, F. A. Coauthors 2006. Evaluation of remote sensing based terrestrial productivity from MODIS using regional tower eddy flux network observations. IEEE Trans. Geosci. Remote Sens. 44:19081925.

    • Search Google Scholar
    • Export Citation
  • Hinzman, L. D. Coauthors 2005. Evidence and implications of recent climate change in northern Alaska and other arctic regions. Climatic Change 72:251298.

    • Search Google Scholar
    • Export Citation
  • Hinzman, L. D. , L. A. Viereck , P. C. Adams , V. E. Romanovsky , and K. Yoshikawa . 2006. Climate and permafrost dynamics of the Alaskan boreal forest. Alaska’s Changing Boreal Forest, F. S. Chapin et al., Eds., Oxford Press, 39–61.

    • Search Google Scholar
    • Export Citation
  • Hirata, R. , T. Hirano , J. Mogami , Y. Fujinuma , K. Inukai , N. Saigusa , and S. Yamamoto . 2005. CO2 flux measured by an open-path system over a larch forest during the snow-covered season. Phyton 45:347351.

    • Search Google Scholar
    • Export Citation
  • Huemmrich, K. F. , T. A. Black , P. G. Jarvis , J. H. McCaughey , and F. G. Hall . 1999. High temporal resolution NDVI phenology from micrometeorological radiation sensors. J. Geophys. Res. 104:27 93527 944.

    • Search Google Scholar
    • Export Citation
  • Huemmrich, K. F. Coauthors 2009. Remote sensing of tundra gross ecosystem productivity and light use efficiency under varying temperature and moisture conditions. Remote Sens. Environ. 114:481489.

    • Search Google Scholar
    • Export Citation
  • Huete, A. , K. Didan , T. Miura , E. P. Rodriguez , X. Gao , and L. G. Ferreira . 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83:195213.

    • Search Google Scholar
    • Export Citation
  • Iwashita, H. , N. Saigusa , S. Murayama , H. McCaughery , A. Black , A. Barr , K. Higuchi , and S. Yamamoto . 2005. Effect of soil water content on carbon dioxide flux at a sparse-canopy forest in the Canadian boreal ecosystem. J. Agric. Meteor. 61:131141.

    • Search Google Scholar
    • Export Citation
  • Jarvis, P. and S. Linder . 2000. Constraints to growth of boreal forests. Nature 405:904905.

  • Kimball, J. S. , P. E. Thornton , M. A. White , and S. W. Running . 1997. Simulating forest productivity and surface-atmosphere carbon exchange in the BOREAS study region. Tree Physiol. 17:589599.

    • Search Google Scholar
    • Export Citation
  • Kimball, J. S. , M. Zhao , K. C. McDonald , and S. W. Running . 2006. Satellite remote sensing of terrestrial net primary production of the pan-Arctic basin and Alaska. Mitigation Adapt. Strategies Global Change 11:783804.

    • Search Google Scholar
    • Export Citation
  • Kitamoto, T. , M. Ueyama , Y. Harazono , T. Iwata , and S. Yamamoto . 2007. Applications of NOAA/AVHRR and observed fluxes to estimate carbon budget at Alaska black spruce forests. J. Agric. Meteor. 63:171183.

    • Search Google Scholar
    • Export Citation
  • Lindroth, A. , A. Grelle , and A-S. Morén . 1998. Long-term measurements of boreal forest carbon balance reveal large temperature sensitivity. Global Change Biol. 4:443450.

    • Search Google Scholar
    • Export Citation
  • Mahadevan, P. Coauthors 2008. A satellite-based biosphere parameterization for net ecosystem CO2 exchange: Vegetation Photosynthesis and Respiration Model (VPRM). Global Biogeochem. Cycles 22:GB2005. doi:10.1029/2006GB002735.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D. , J. S. Clein , J. M. Melillo , D. W. Kicklighter , R. A. Meier , C. J. Vorosmaty , and M. C. Serreze . 2000. Modelling carbon responses of tundra ecosystems to historical and projected climate: Sensitivity of pan-Arctic carbon storage to temporal and spatial variation in climate. Global Change Biol. 6:141159.

    • Search Google Scholar
    • Export Citation
  • Moore, C. J. 1986. Frequency response corrections for eddy correlation systems. Bound.-Layer Meteor. 37:1735.

  • Piao, S. Coauthors 2008. Net carbon dioxide losses of northern ecosystems in response to autumn warming. Nature 451:4952.

  • Rahman, A. F. , J. A. Gamon , D. A. Fuentes , D. A. Roberts , and D. Prentiss . 2001. Modeling spatially distributed ecosystem flux of boreal forest using hyperspectral indices from AVIRIS imagery. J. Geophys. Res. 106:33 57933 591.

    • Search Google Scholar
    • Export Citation
  • Running, S. W. , R. R. Namani , F. A. Heinsch , M. Zhao , M. Reeves , and H. Hashimoto . 2004. A continuous satellite-derived measure of global terrestrial primary production. BioScience 54:547560.

    • Search Google Scholar
    • Export Citation
  • Sasai, T. , K. Ichii , Y. Yamaguchi , and R. Nemani . 2005. Simulating terrestrial carbon fluxes using the new biosphere model “biosphere model integrating eco-physiological and mechanistic approaches using satellite data” (BEAMS). J. Geophys. Res. 110:G02014. doi:10.1029/2005JG000045.

    • Search Google Scholar
    • Export Citation
  • Shulski, M. and G. Wendler . 2007. The Climate of Alaska. University of Alaska Press, 216 pp.

  • Simpson, J. J. , G. L. Hufford , M. D. Fleming , J. S. Berg , and J. B. Ashton .