• Ahl, D., , S. Gower, , D. Mackay, , S. Burrows, , J. Norman, , and G. Diak, 2004: Heterogeneity of light use efficiency in a northern Wisconsin forest: Implications for modeling net primary production with remote sensing. Remote Sens. Environ., 93, 168178.

    • Search Google Scholar
    • Export Citation
  • Birdsey, R. A., , and L. S. Heath, 1995: Carbon changes in U.S. forests. Productivity of America’s forests and climate change, U.S. Department of Agriculture Forest Service Rocky Mountain Forest Experiment Station General Tech. Rep. RM-GTR-271, 56–70.

  • Birdsey, R. A., , K. Pregitzer, , and A. Lucier, 2006: Forest carbon management in the United States: 1600-2100. J. Environ. Qual., 35, 14611469.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1995: Land-atmosphere CO2 exchange simulated by a land surface process model coupled to an atmospheric general circulation model. J. Geophys. Res., 100, 28172831.

    • Search Google Scholar
    • Export Citation
  • Bondeau, A., , D. W. Kicklighter, , J. Kaduk, , and T. P. O. T. P. N. M. Intercomparison, 1999: Comparing global models of terrestrial net primary productivity (NPP): Importance of vegetation structure on seasonal NPP estimates. Global Change Biol., 5 (S1), 3545.

    • Search Google Scholar
    • Export Citation
  • Bresson, C. C., , A. S. Kowalski, , A. Kremer, , and S. Delzon, 2009: Evidence of altitudinal increase in photosynthetic capacity: Gas exchange measurements at ambient and constant CO2 partial pressures. Ann. For. Sci., 66, 505, doi:10.1051/forest/2009027.

    • Search Google Scholar
    • Export Citation
  • Corder, G. W., , and D. I. Foreman, 2009: Nonparametric Statistics for Non-Statisticians: A Step-by-Step Approach. Wiley, 264 pp.

  • Cramer, W., , D. Kicklighter, , A. Bondeau, , B. Moore III, , G. Churkina, , B. Nemry, , A. Ruimy, , and A. Schloss, 1999: Comparing global models of terrestrial net primary productivity (NPP): Overview and key results. Global Change Biol., 5, 115.

    • Search Google Scholar
    • Export Citation
  • Davi, H., and Coauthors, 2006: Effect of aggregating spatial parameters on modelling forest carbon and water fluxes. Agric. For. Meteor., 139 (3–4), 269287.

    • Search Google Scholar
    • Export Citation
  • De Kauwe, M., , T. Quaife, , P. Lewis, , M. Disney, , and M. Williams, 2008: Estimating the spatial exchange of carbon through the assimilation of Earth observation derived products using an ensemble Kalman filter. Proc. Geoscience and Remote Sensing Symp., Boston, MA, IEEE, Vol. 3, 1044–1047.

  • Dubayah, R. O., , and J. B. Drake, 2000: Lidar remote sensing for forestry. J. For., 98, 4446.

  • FAO, cited 2011: FAO/Unesco Soil map of the world (1:5 000 000), 1971-1981. FAO.

  • Fischer, G., , F. Nachtergaele, , S. Prieler, , H. T. van Velthuizen, , L. Verelst, , and D. Wiberg, 2008: Global Agro-ecological Zones Assessment for Agriculture (GAEZ 2008). IIASA and FAO Rep.

  • Gower, S. T., , R. E. McMurtrie, , and D. Murty, 1996: Aboveground net primary production decline with stand age: Potential causes. Trends Ecol. Evol., 11, 378382.

    • Search Google Scholar
    • Export Citation
  • Heinsch, F., and Coauthors, 2006: Evaluation of remote sensing based terrestrial productivity from MODIS using regional tower eddy flux network observations. IEEE Trans. Geosci. Remote Sens., 44, 19081923.

    • Search Google Scholar
    • Export Citation
  • Houghton, R. A., , J. L. Hackler, , and K. T. Lawrence, 1999: The U.S. carbon budget: Contributions from land-use change. Science, 285, 574578.

    • Search Google Scholar
    • Export Citation
  • Kicklighter, D. W., and Coauthors, 1999: Comparing global models of terrestrial net primary productivity (NPP): Global pattern and differentiation by major biomes. Global Change Biol., 5 (S1), 1624.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267.

    • Search Google Scholar
    • Export Citation
  • Kitajima, K., , S. S. Mulkey, , and S. Wright, 1997: Decline of photosynthetic capacity with leaf age in relation to leaf longevities for five tropical canopy tree species. Amer. J. Bot., 84, 702708.

    • Search Google Scholar
    • Export Citation
  • Knuth, D., 1997: Sorting and Searching. Vol. 3, The Art of Computer Programming, 3rd ed. Addison-Wesley, 723 pp.

  • Kobayashi, Y., , K. Sarabandi, , L. Pierce, , and M. C. Dobson, 2000: An evaluation of the JPL TOPSAR for extracting tree heights. IEEE Trans. Geosci. Remote Sens., 38, 24462454.

    • Search Google Scholar
    • Export Citation
  • LeBauer, D. S., , and K. K. Treseder, 2008: Nitrogen limitation of net primary productivity in terrestrial ecosystems is globally distributed. Ecology, 89, 371379.

    • Search Google Scholar
    • Export Citation
  • Lefsky, M. A., , W. B. Cohen, , G. G. Parker, , and D. J. Harding, 2009: Lidar remote sensing for ecosystem studies. Bioscience, 52, 1930.

  • McCallum, I., , W. Wagner, , C. Schmullius, , A. Shvidenko, , M. Obersteiner, , S. Fritz, , and S. Nilsson, 2009: Satellite-based terrestrial production efficiency modeling. Carbon Balance Manage., 4, doi:10.1186/1750-0680-4-8.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D., , J. M. Melillo, , L. A. Joyce, , D. W. Kicklighter, , A. L. Grace, , B. Moore III, , and C. J. Vorosmarty, 1992: Interactions between carbon and nitrogen dynamics in estimating net primary productivity for potential vegetation in North America. Global Biogeochem. Cycles, 6, 101124.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D., and Coauthors, 2001: Carbon balance of the terrestrial biosphere in the twentieth century: Analyses of CO2, climate and land use effects with four process-based ecosystem models. Global Biogeochem. Cycles, 15, 183206.

    • Search Google Scholar
    • Export Citation
  • Melillo, J. M., , A. D. McGuire, , D. W. Kicklighter, , B. Moore, , C. J. Vorosmarty, , and A. L. Schloss, 1993: Global climate change and terrestrial net primary production. Nature, 363, 234240.

    • Search Google Scholar
    • Export Citation
  • Monsi, M., , and T. Saeki, 2005: On the factor light in plant communities and its importance for matter production. Ann. Bot., 95, 549567.

    • Search Google Scholar
    • Export Citation
  • Pacala, S. W., and Coauthors, 2001: Consistent land- and atmosphere-based U.S. carbon sink estimates. Science, 292, 23162320.

  • Pietsch, S. A., , and H. Hasenauer, 2006: Evaluating the self-initialization procedure for large-scale ecosystem models. Global Change Biol., 12, 16581669, doi:10.1111/j.1365-2486.2006.01211.x.

    • Search Google Scholar
    • Export Citation
  • Plummer, S., 2006: On validation of the MODIS gross primary production product. IEEE Trans. Geosci. Remote Sens., 44, 19361938.

  • Potter, C. S., , J. T. Randerson, , C. B. Field, , P. A. Matson, , P. M. Vitousek, , H. A. Mooney, , and S. A. Klooster, 1993: Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Biogeochem. Cycles, 7, 811841.

    • Search Google Scholar
    • Export Citation
  • Pregitzer, K. S., , and E. S. Euskirchen, 2004: Carbon cycling and storage in world forests: Biome patterns related to forest age. Global Change Biol., 10, 20522077.

    • Search Google Scholar
    • Export Citation
  • Raich, J. W., and Coauthors, 1991: Potential net primary productivity in South America: Application of a global model. Ecol. Appl., 1, 399429.

    • Search Google Scholar
    • Export Citation
  • Ruimy, A., , L. Kergoat, , and A. Bondeau, 1999: Comparing global models of terrestrial net primary productivity (NPP): Analysis of differences in light absorption and light-use efficiency. Global Change Biol., 5, 5664.

    • Search Google Scholar
    • Export Citation
  • Running, S. W., , and J. C. Coughlan, 1988: A general model of forest ecosystem processes for regional applications I. Hydrologic balance, canopy gas exchange and primary production processes. Ecol. Modell., 42, 125154.

    • Search Google Scholar
    • Export Citation
  • Running, S. W., , R. R. Nemani, , 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
  • Schloss, A., , D. Kicklighter, , J. Kaduk, , and U. Wittenberg, 1999: Comparing global models of terrestrial net primary productivity (NPP): Comparison of NPP to climate and the normalized difference vegetation index (NDVI). Global Change Biol., 5, 2534.

    • Search Google Scholar
    • Export Citation
  • Still, C., , J. Randerson, , and I. Fung, 2004: Large-scale plant light-use efficiency inferred from the seasonal cycle of atmospheric CO2. Global Change Biol., 10, 12401252.

    • Search Google Scholar
    • Export Citation
  • Sundquist, E. T., , K. V. Ackerman, , N. B. Bliss, , J. M. Kellndorfer, , M. C. Reeves, , and M. G. Rollins, 2009: Rapid assessment of U.S. forest and soil organic carbon storage and forest biomass carbon sequestration capacity. U.S. Geological Survey Open-File Rep. 2009–1283, 23 pp.

  • Tang, J., , and Q. Zhuang, 2008: Equifinality in parameterization of process-based biogeochemistry models: A significant uncertainty source to the estimation of regional carbon dynamics. J. Geophys. Res., 113, G04010, doi:10.1029/2008JG000757.

    • Search Google Scholar
    • Export Citation
  • Tang, J., , and Q. Zhuang, 2009: A global sensitivity analysis and Bayesian inference framework for improving the parameter estimation and prediction of a process-based Terrestrial Ecosystem Model. J. Geophys. Res., 114, D15303, doi:10.1029/2009JD011724.

    • Search Google Scholar
    • Export Citation
  • Turner, D. P., , G. J. Koerper, , M. E. Harmon, , and J. J. Lee, 1995: A carbon budget for forests of the conterminous United States. Ecol. Appl., 5, 421436.

    • Search Google Scholar
    • Export Citation
  • Turner, D. P., and Coauthors, 2006: Evaluation of MODIS NPP and GPP products across multiple biomes. Remote Sens. Environ., 102 (3–4), 282292.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Agriculture, cited 2011: U.S. General Soil Map (STATSGO2). Natural Resources Conservation Service. [Available online at http://soildatamart.nrcs.usda.gov.]

  • Van Deusen, P., , and L. S. Heath, cited 2010a: COLE web applications suite. NCASI and USDA Forest Service Northern Research Station. [Available online at http://www.ncasi2.org/COLE/.]

  • Van Deusen, P., , and L. S. Heath, 2010b: Weighted analysis methods for mapped plot forest inventory data: Tables, regressions, maps and graphs. For. Ecol. Manage., 260, 16071612.

    • Search Google Scholar
    • Export Citation
  • White, M. A., , P. E. Thornton, , S. W. Running, , and R. R. Nemani, 2000: Parameterization and sensitivity analysis of the BIOME–BGC Terrestrial Ecosystem Model: Net primary production controls. Earth Interact., 4. [Available online at http://EarthInteractions.org.]

    • Search Google Scholar
    • Export Citation
  • Xiao, J., and Coauthors, 2008: Estimation of net ecosystem carbon exchange for the conterminous United States by combining MODIS and AmeriFlux data. Agric. For. Meteor., 148, 18271847.

    • Search Google Scholar
    • Export Citation
  • Xiao, J., and Coauthors, 2010: A continuous measure of gross primary production for the conterminous United States derived from MODIS and AmeriFlux data. Remote Sens. Environ., 114, 576591.

    • Search Google Scholar
    • Export Citation
  • Yang, F., and 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
  • Yu, Y., , S. Saatchi, , L. S. Heath, , E. LaPoint, , R. Myneni, , and Y. Knyazikhin, 2010: Regional distribution of forest height and biomass from multisensor data fusion. J. Geophys. Res., 115, G00E12, doi:10.1029/2009JG000995.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , J. Chen, , W. Ju, , S. Shen, , Y. Pan, , R. Birdsey, , and L. He, 2010: Carbon balance in conterminous U.S. forests based on historic changes in climate, atmospheric composition, and disturbances. Proc. Fall Meeting, San Francisco, CA, Amer. Geophys. Union, B41K-07.

  • Zhao, M., , F. A. Heinsch, , R. R. Nemani, , and S. W. Running, 2005: Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sens. Environ., 95, 164176.

    • Search Google Scholar
    • Export Citation
  • Zhao, M., , S. W. Running, , and R. R. Nemani, 2006: Sensitivity of Moderate Resolution Imaging Spectroradiometer (MODIS) terrestrial primary production to the accuracy of meteorological reanalyses. J. Geophys. Res., 111, G01002, doi:10.1029/2004JG000004.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , V. E. Romanovsky, , and A. D. McGuire, 2001: Incorporation of a permafrost model into a large-scale ecosystem model: Evaluation of temporal and spatial scaling issues in simulating soil thermal dynamics. J. Geophys. Res., 106, 33 64933 670.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , A. D. McGuire, , K. P. O’Neill, , J. W. Harden, , V. E. Romanovsky, , and J. Yarie, 2002: Modeling soil thermal and carbon dynamics of a fire chronosequence in interior Alaska. J. Geophys. Res., 107, 8147, doi:10.1029/2001JD001244.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., and Coauthors, 2003: Carbon cycling in extratropical terrestrial ecosystems of the Northern Hemisphere during the 20th century: A modeling analysis of the influences of soil thermal dynamics. Tellus, 55B, 751776.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , J. M. Melillo, , D. W. Kicklighter, , R. G. Prinn, , A. D. McGuire, , P. A. Steudler, , B. S. Felzer, , and S. Hu, 2004: Methane fluxes between terrestrial ecosystems and the atmosphere at northern high latitudes during the past century: A retrospective analysis with a process-based biogeochemistry model. Global Biogeochem. Cycles, 18, GB3010, doi:10.1029/2004GB002239.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., and Coauthors, 2006: CO2 and CH4 exchanges between land ecosystems and the atmosphere in northern high latitudes over the 21st century. Geophys. Res. Lett., 33, L17403, doi:10.1029/2006GL026972.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Spatially explicit mean carbon pool sizes and their standard errors used for calibration. Units are kg C m−2 and g C m−2 for the mean pool sizes and standard errors, respectively. (a) VEGC and VEGC std and (b) SOLC and SOLC std.

  • View in gallery

    Annual-mean carbon fluxes for (a) GPP and (b) NPP used for calibration. Units are g C m−2 yr−1.

  • View in gallery

    Potential vegetation coverage of the forest regions in the conterminous United States at a resolution of 0.5° × 0.5° (longitude × latitude).

  • View in gallery

    Spatial patterns of the calibrated key parameters: (a) Cmax; (b) Kr; (c) Kd; (d) CFALL; (e) Nmax; and (f) Nup.

  • View in gallery

    Annual carbon fluxes and carbon pool sizes from 1948 to 2000 of the forest ecosystems of the conterminous United States. Variations of (a) GPP, (b) NPP, (c) NEP, (d) RA, (e) RH, (f) VEGC, and (g) SOLC. The variable C in the legend stands for each carbon flux and pool size. The subscript p stands for using the spatial calibration method, whereas t stands for the traditional method. Here, std indicates the standard deviation of the ensemble simulation results. Note that the std is too small to be seen in (a)–(e).

  • View in gallery

    Spatial patterns of NEP, VEGC, and SOLC estimated using TEM with the traditional and spatially explicit parameterization methods during the period of 1948–2000. (a) NEP estimated with traditional parameterization, (b) NEP estimated with spatially explicit parameterization, (c) VEGC estimated with traditional parameterization, (d) VEGC estimated with spatially explicit parameterization, (e) SOLC estimated with traditional parameterization, and (f) SOLC estimated with spatially explicit parameterization.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 25 25 5
PDF Downloads 8 8 0

Spatially Explicit Parameterization of a Terrestrial Ecosystem Model and Its Application to the Quantification of Carbon Dynamics of Forest Ecosystems in the Conterminous United States

View More View Less
  • 1 Department of Earth & Atmospheric Sciences, Purdue University, West Lafayette, Indiana
  • | 2 Department of Earth & Atmospheric Sciences, and Department of Agronomy, Purdue University, West Lafayette, Indiana
© Get Permissions
Full access

Abstract

The authors use a spatially explicit parameterization method and the Terrestrial Ecosystem Model (TEM) to quantify the carbon dynamics of forest ecosystems in the conterminous United States. Six key parameters that govern the rates of carbon and nitrogen dynamics in TEM are selected for calibration. Spatially explicit data for carbon and nitrogen pools and fluxes are used to calibrate the six key parameters to more adequately account for the spatial heterogeneity of ecosystems in estimating regional carbon dynamics. The authors find that a spatially explicit parameterization results in vastly different carbon exchange rates relative to a parameterization conducted for representative ecosystem sites. The new parameterization method estimates that the net ecosystem production (NEP), the annual gross primary production (GPP), and the net primary production (NPP) of the regional forest ecosystems are 61% (0.02 Pg C; 1 Pg = 1015 g) higher and 2% (0.11 Pg C) and 19% (0.45 Pg C) lower, respectively, than the values obtained using the traditional parameterization method for the period 1948–2000. The estimated vegetation carbon and soil organic carbon pool sizes are 51% (18.73 Pg C) lower and 29% (7.40 Pg C) higher. This study suggests that, to more adequately quantify regional carbon dynamics, spatial data for carbon and nitrogen pools and fluxes should be developed and used with the spatially explicit parameterization method.

Corresponding author address: Min Chen, Civil Engineering Building, 550 Stadium Mall Drive, West Lafayette, IN 47907-2051. E-mail address: chenm@purdue.edu

Abstract

The authors use a spatially explicit parameterization method and the Terrestrial Ecosystem Model (TEM) to quantify the carbon dynamics of forest ecosystems in the conterminous United States. Six key parameters that govern the rates of carbon and nitrogen dynamics in TEM are selected for calibration. Spatially explicit data for carbon and nitrogen pools and fluxes are used to calibrate the six key parameters to more adequately account for the spatial heterogeneity of ecosystems in estimating regional carbon dynamics. The authors find that a spatially explicit parameterization results in vastly different carbon exchange rates relative to a parameterization conducted for representative ecosystem sites. The new parameterization method estimates that the net ecosystem production (NEP), the annual gross primary production (GPP), and the net primary production (NPP) of the regional forest ecosystems are 61% (0.02 Pg C; 1 Pg = 1015 g) higher and 2% (0.11 Pg C) and 19% (0.45 Pg C) lower, respectively, than the values obtained using the traditional parameterization method for the period 1948–2000. The estimated vegetation carbon and soil organic carbon pool sizes are 51% (18.73 Pg C) lower and 29% (7.40 Pg C) higher. This study suggests that, to more adequately quantify regional carbon dynamics, spatial data for carbon and nitrogen pools and fluxes should be developed and used with the spatially explicit parameterization method.

Corresponding author address: Min Chen, Civil Engineering Building, 550 Stadium Mall Drive, West Lafayette, IN 47907-2051. E-mail address: chenm@purdue.edu

1. Introduction

The global carbon cycle plays an important role in affecting the climate system (Cramer et al. 1999). Quantifying the dynamics of carbon exchange between the biosphere and the atmosphere is important in the understanding of global climate change. To date, many process-based biogeochemical models have been used to quantify carbon dynamics (Bonan 1995; McGuire et al. 1992; Potter et al. 1993; Running and Coughlan 1988; Zhuang et al. 2003). These models incorporate the biological, physical, and chemical processes of ecosystems and use mathematical equations to represent these processes. These mathematical equations are usually parameterized for representative vegetation types and then extrapolated to regional scales. For example, the Terrestrial Ecosystem Model (TEM) has been widely used to study ecosystem carbon and nitrogen dynamics at different scales since the early 1990s (Kicklighter et al. 1999; McGuire et al. 1992; McGuire et al. 2001; Melillo et al. 1993; Raich et al. 1991; Zhuang et al. 2002; Zhuang et al. 2003; Zhuang et al. 2006; Tang and Zhuang 2008; Tang and Zhuang 2009). In TEM, a number of parameters are used to describe and govern the physical processes of carbon, nitrogen, water, and thermal dynamics in the represented ecosystems. The parameters related to hydrologic and thermal processes are mostly determined by literature review or independent estimation using published data; however, some TEM-specific internal parameters that control the rates of carbon and nitrogen processes cannot be determined directly from the experimental measurement data but rather have to be determined through model parameterization. Traditionally, the parameterization of TEM is conducted with field data for typical ecosystem types. However, especially for the forest ecosystem, this treatment does not address the uncertainty due to the spatial heterogeneity of a region, which arises as a result of variation in stand age, species, and geographic location (Bondeau et al. 1999; Ruimy et al. 1999; Schloss et al. 1999). For instance, the aboveground productivity of forests has been reported to decrease with forest age, which may be a result of the altered carbon balance between photosynthesis and ecosystem respiration, as well as the decreased soil nutrient availability (Gower et al. 1996; Pregitzer and Euskirchen 2004; Pietsch and Hasenauer 2006). Leaf aging can also affect plant photosynthetic capacity (Kitajima et al. 1997). Bresson et al. (Bresson et al. 2009) presented evidence for altitudinal increases in photosynthetic capacity using gas exchange measurements. Turner et al. (Turner et al. 1995) reported that forests with different age classes, geographic locations, and species types had different rates of carbon fluxes and pools in the conterminous United States. Because the rate-controlling parameters in TEM are strongly related to the rates of processes, the above evidence suggests that using a single set of parameters calibrated at representative sites for a specific broad vegetation type, but neglecting the spatial heterogeneity of the ecosystems, can result in significant uncertainty for a regional quantification.

Ideally, spatially explicit parameters for TEM are needed for regional simulations. With more available data reporting vegetation and soil carbon and nitrogen pools (Van Deusen and Heath 2010a; Van Deusen and Heath 2010b) and flux (e.g. Zhao et al. 2005), parameterization of TEM for all grid cells at a regional scale is possible. Here, we conduct a study to obtain such data and to evaluate the carbon dynamics of forest ecosystems in the conterminous United States for the period 1948–2000. We first develop spatial distributions for the key parameters of the model, TEM. Second, we evaluate the differences in carbon fluxes and pool sizes between those determined using a traditional parameterization method and those determined using the spatially explicit parameterization method. TEM is calibrated using the required spatial datasets for the forests of the conterminous United States in order to obtain the spatially explicit parameters. TEM simulations are conducted for the period 1948–2000 with both the spatially explicit parameters and the parameters obtained using the traditional method.

2. Methods

2.1. Terrestrial Ecosystem Model and its calibration

The TEM is a well-documented, process-based ecosystem model that describes the carbon and nitrogen dynamics of plants and soils in terrestrial ecosystems (McGuire et al. 1992; McGuire et al. 2001; Melillo et al. 1993; Raich et al. 1991; Zhuang et al. 2001; Zhuang et al. 2002; Zhuang et al. 2004; Zhuang et al. 2003). TEM uses spatially referenced information on climate, elevation, soils, vegetation, and water availability as well as soil- and vegetation-specific parameters to make monthly estimates of important carbon and nitrogen fluxes and pool sizes for terrestrial ecosystems.

In TEM, specific parameters control the magnitude of carbon and nitrogen fluxes (Table 1). The TEM traditional parameterization method uses carbon and nitrogen pools and annual fluxes from intensively studied sites (McGuire et al. 1992) to estimate the values for each of the rate-controlling parameters. The ecosystem data needed for a site-level calibration of TEM include 1) vegetation and soil organic carbon pool sizes (VEGC and SOLC); 2) vegetation and soil nitrogen carbon pool sizes (VEGN and SOLN); 3) gross primary production (GPP) and net primary production (NPP); 4) NPP without N limitation (NPPsat); 5) inorganic N in soil (Nav); and 6) N uptake of vegetation (NUPTAKE). Parameters associated with carbon and nitrogen fluxes in TEM are sequentially adjusted until all carbon and nitrogen pools, as well as annual GPP and NPP, match the observations. Here, we focus on six key parameters that are identified from our previous sensitivity study (Tang and Zhuang 2009). These parameters are Cmax, representing the maximum photosynthesis rate; Kr and Kd, describing the rate of autotrophic and heterotrophic respiration, respectively; CFALL, indicating the carbon litterfall rate; and Nmax and Nup, which are related to the N feedback for C and N uptake in vegetation. The definition of these parameters and their associated processes are documented in Tables 1 and 2.

Table 1.

Vegetation-specific parameters used in TEM. Here, Cmax, Kr, Kd, CFALL, Nmax, and Nup are selected as the key parameters in the spatially explicit parameterization.

Table 1.
Table 2.

Key processes related to carbon dynamics in TEM.

Table 2.

During model calibrations, the ecosystem data from each site are used to initialize the model, and the model is driven with climate data from the same location. Parameters are determined during the TEM calibration by repetitively adjusting parameters, running the model, and comparing the model estimates and observations. This process is completed according to the following strategy: first, we run the C cycle uncoupled from the N cycle in TEM in order to calculate productivities as if N were not limiting. Therefore, the parameters directly associated with calculating carbon fluxes and pool sizes can be determined in this step. GPP, NPP, and the maximum response of NPP to N fertilization (NPPsat) are used to constrain the maximum rate of C assimilation (Cmax). The parameter Kr is determined by the rate of autotrophic respiration (RA), which is the difference of GPP and NPP. Here, Kd and CFALL are determined by SOLC and by the balance between VEGC and SOLC. Second, we run TEM, this time coupled with the N cycle to activate the C–N interactions and determine N-related parameters, which control N cycling rates and feedbacks on C cycle. The available inorganic N in soil and detritus (Nav) constrains the value of Nup; the annual NUPTAKE and NPP determine Nmax. The more specific steps of the calibration procedures are 1) set up the initial values of parameters and pool sizes (use the values from previous studies); 2) turn off nitrogen feedback effects; 3) manually adjust Cmax and run TEM until annual GPP matches the observed value, then manually adjust Kr and run TEM until annual NPP matches the observed value; 4) manually adjust Kd and run TEM until SOLC matches the observed value, then adjust CFALL and run TEM until VEGC matches the observed value; 5) because adjusting CFALL can change both VEGC and SOLC, do iterations of step 4 until SOLC and VEGC both match their observed values; 6) manually adjust Cmax and run the model until annual NPP matches NPPsat; 7) turn on nitrogen feedback effects; 8) manually adjust Nup until Nav is equal to its observed value, and adjust Nmax until NPP and NUPTAKE are close to their observations; and 9) put the values of carbon and nitrogen pool sizes, as well as the parameter values, into the parameter table for model extrapolation. In each step, we control the differences between model estimates and observations within 1% error tolerance. These calibrated parameters are then used for extrapolation simulations. More details of the calibration and extrapolation methods can be found in Raich et al. (Raich et al. 1991) and McGuire et al. (McGuire et al. 1992).

2.2. Spatially explicit calibration for TEM

To date, the parameterization of TEM has been conducted at the site level of major representative ecosystems in order to conduct regional simulations. The parameterization and model were then extrapolated from these site-level observations onto a regional scale. To have a spatially explicit calibration for a region with TEM, here we develop an automatic calibration program for TEM. This program follows the traditional procedures of calibrating TEM but automatically adjusts the calibrating parameters to fit the model estimates with the observations. The program is designed based on the binary-search algorithm (Knuth 1997, 409–426) to accelerate the efficiency of finding the appropriate parameters. The program is used to parameterize TEM for each forest ecosystems grid cell in the conterminous United States at a 0.5° × 0.5° resolution with available satellite products and forest and soil inventory data. The model and parameters are then extrapolated to the regional scale in order to examine how this forest carbon quantification differs from that determined using the traditional parameterization method. To obtain the spatially explicit parameters, we develop the spatial data of vegetation carbon pool sizes using the Carbon On Line Estimator (COLE) developed by the U.S. Department of Agriculture (USDA) Forest Service (Figure 1). COLE data are based on USDA Forest Service Forest Inventory and Analysis and Resource Planning Assessment data and enhanced by other ecological data. COLE provides county-level carbon storage for the forested regions of the United States using weighted analysis methods (Van Deusen and Heath 2010a; Deusen and Heath 2010b). We first obtain the county-level mean live tree carbon storage (VEGC) and soil organic carbon storage (SOLC) data for the conterminous United States. We then resample them into 0.5° × 0.5° grid cells using the nearest-neighborhood method. Finally, we extract forest grid cells from a global vegetation map (Melillo et al. 1993). The nitrogen pool sizes (VEGN and SOLN) are then estimated by the carbon pool sizes using the C:N ratios (VEGcn and SOLcn) used in previous versions of TEM,
e1
e2
Figure 1.
Figure 1.

Spatially explicit mean carbon pool sizes and their standard errors used for calibration. Units are kg C m−2 and g C m−2 for the mean pool sizes and standard errors, respectively. (a) VEGC and VEGC std and (b) SOLC and SOLC std.

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

Here, the VEGcn and SOLcn are long-term average C:N ratios. They provide an estimation of the nitrogen pool sizes for the spatially explicit parameterization and additionally provide the initial values for the C:N ratios during the parameterization procedure; these ratios will be updated for each time step for each grid cell (McGuire et al. 1992). Values of VEGcn and SOLcn for each ecosystem type are adapted from Zhuang et al. (Zhuang et al. 2003). The available nitrogen (Nav) for each ecosystem type is set as a fixed value as follows (McGuire et al. 1992): 0.5, 1.0, 2.0, and 1.5 g N m−2 for boreal forest, temperate coniferous forest, temperate deciduous forest, and temperate mixed forest, respectively. Additionally, in order to test the uncertainty of the model as a result of uncertain carbon pool sizes in calibration, we use the county-level standard error for carbon pool sizes from COLE to calibrate TEM. We resample and extract the data for forest ecosystems in the conterminous United States.

The GPP and NPP data for calibration are obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) products (Figure 2). We obtain MOD17A3_C5.1 (yearly GPP and NPP) from the Numerical Terradynamic Simulation Group (NTSG) of the University of Montana and then calculate the annual mean within the available years from 2000 to 2007. The 1 km × 1 km pixels are then resampled into 0.5° × 0.5° grid cells to match the spatial resolution of TEM.

Figure 2.
Figure 2.

Annual-mean carbon fluxes for (a) GPP and (b) NPP used for calibration. Units are g C m−2 yr−1.

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

The spatial values of NPPsat are, however, unavailable, and we therefore estimate them using the following empirical relationship, which is suggested by (McGuire et al. 1992):
e3
NUPTAKE is calculated as
e4

The meteorology data used for calibration include the monthly precipitation, air temperature, and cloudiness fraction for the period 1948–2000 (Kistler et al. 2001). The 53-yr average of the meteorology data for each grid cell is used for the spatially explicit parameterization.

The Spearman’s rank correlation coefficient ρ is employed to test the spatial correlation between the calibrated parameter values and the ecosystem carbon and nitrogen pool and flux data used for calibration. The ρ indicates the direction of association between the compared variables. A positive ρ indicates the same direction as the compared variables, whereas a negative ρ indicates the opposite direction. The higher absolute value of ρ suggests a stronger correlation of the monotonic relation.

2.3. Regional simulation

To quantify carbon fluxes over the forested area of the conterminous United States and compare the differences between two simulations, one using the spatially explicit parameterization method and the other the traditional method, we apply TEM to the region at a 0.5° × 0.5° spatial resolution for the period 1948–2000, for a total of 1370 grid cells (Figure 3). The spatially explicit soil texture data describing the percentage of sand, silt, and clay in the soil are originally from the published soil map of Food and Agriculture Organization of the United Nations (FAO 2011); Other auxiliary data, such as elevation, are from our previous studies (Zhuang et al. 2003).The annual atmospheric CO2 concentration data from 1948 to 2000 are also based on data from our previous studies (Zhuang et al. 2003). We first run TEM to equilibrium and then spin up the model for 1000 years to account for the influence of climate interannual variability on the initial conditions of the ecosystems. After that, we run the model with transient climate and annual atmospheric CO2 concentrations from 1948 to 2000.

Figure 3.
Figure 3.

Potential vegetation coverage of the forest regions in the conterminous United States at a resolution of 0.5° × 0.5° (longitude × latitude).

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

The COLE data are supposed to be unbiased, but statistical standard errors exist in association with the mean values (Van Deusen and Heath 2010a; Van Deusen and Heath 2010b). To quantify the possible uncertainty induced by the statistical errors, we conduct ensemble calibrations using the mean carbon pool sizes and their corresponding standard errors. Because we use two carbon pools (the vegetation carbon and soil organic carbon), besides the calibration using mean values of VEGC and SOLC from COLE, four additional sets of calibrations are conducted based on the combinations of mean values (e.g., VEGCmean and SOLCmean) and standard errors (e.g., VEGCstd and SOLCstd) of VEGC and SOLC, which are VEGCmean + VEGCstd and SOLCmean + SOLCstd, VEGCmean + VEGCstd and SOLCmean − SOLCstd, VEGCmean − VEGCstd and SOLCmean + SOLCstd, and VEGCmean − VEGCstd and SOLCmean − SOLCstd. We then extrapolate the five sets of spatially explicit parameters to the region. Standard errors for these five sets of calibrated parameters and simulation results are used for uncertainty analysis.

3. Results

3.1. Spatial patterns of the calibrated key parameters

Parameter values vary significantly in the region (Figure 4). The highest Cmax values are mainly observed along the Appalachian Mountains, the coastal plain, the Ozark plateaus, and the Ouachita Mountains, as well as the Pacific coastal regions. The spatial pattern of Kr is similar to that of Cmax, except for the low values in the Pacific coastal region and the high values in northern Michigan. High Kd, however, is mostly distributed in the Appalachian Mountains area, the Rocky Mountains area, and the California Coast Range area. High CFALL is mostly located in the eastern United States and especially along the Gulf Coast and the northern central plain areas. The N-related parameters have similar spatial patterns: both of which are high in the coastal plains, the northern central lowlands, the Rocky Mountains, and the Pacific coastal regions and low in the east-central United States.

Figure 4.
Figure 4.

Spatial patterns of the calibrated key parameters: (a) Cmax; (b) Kr; (c) Kd; (d) CFALL; (e) Nmax; and (f) Nup.

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

The calibrated parameters are spatially associated with the input variables (Table 3). The Cmax is strongly and positively correlated with the GPP and NPP because it directly affects the carbon assimilation rate; the Kr is positively related to the rate of plant maintenance respiration, a portion of GPP in TEM; thus, it is significantly and positively correlated to the GPP. The Kr also has notably negative correlations with the VEGC in our simulation because respiration directly reduces the vegetation carbon pool size. For a similar reason, CFALL is positively correlated to SOLC but varies inversely with VEGC, as we would expect. The Kd indicates the rate of heterotrophic respiration, which can reduce the SOLC size. Our results show Kd varies inversely compared to the SOLC, as expected. The Kd is also found to be negatively related to air temperature and precipitation, but most likely Kd has the same direction of spatial variation as VEGC. The spatial variations of the N-related parameters (Nmax and Nup) are positively correlated with the variations in NPP.

Table 3.

Spatial correlations between the spatially explicitly calibrated parameters and the variables used for calibration. Here, T stands for air temperature and P stands for precipitation.

Table 3.

3.2. Simulated carbon dynamics with the traditional and spatially explicit parameterization methods

The traditional parameterization estimates that the region was a C sink of 0.03 ± 0.14 Pg C yr−1 with an annual GPP of 4.55 ± 0.24 Pg C yr−1 and NPP of 2.37 ± 0.19 Pg C yr−1 for the period from 1948 to 2000 over the total vegetated area of 3.26 × 106 km2 (Table 4). The RA and RH for this period are 2.17 ± 0.06 and 2.34 ± 0.09 Pg C yr−1, respectively. During this period, climate factors fluctuated frequently and resulted in a significant interannual variability in these carbon fluxes (Figure 5).

Table 4.

TEM-estimated average carbon fluxes and pool sizes over the period of 1948–2000 with the traditional and spatially explicit parameterization methods. Here, t stands for estimations by TEM with the traditional parameterization method; p stands for estimations by TEM with the spatially explicit parameterization method; and %diff = (pt)/t × 100% indicates the difference between the two estimations. Units are Pg C yr−1.

Table 4.
Figure 5.
Figure 5.

Annual carbon fluxes and carbon pool sizes from 1948 to 2000 of the forest ecosystems of the conterminous United States. Variations of (a) GPP, (b) NPP, (c) NEP, (d) RA, (e) RH, (f) VEGC, and (g) SOLC. The variable C in the legend stands for each carbon flux and pool size. The subscript p stands for using the spatial calibration method, whereas t stands for the traditional method. Here, std indicates the standard deviation of the ensemble simulation results. Note that the std is too small to be seen in (a)–(e).

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

Overall, TEM with the spatial parameterization provides similar temporal trends for C dynamics, but with different magnitudes in comparison with the estimates from the traditional parameterization method (Figure 5). With the spatially explicit parameterization, TEM estimates that the regional GPP was 4.46 ± 0.28 Pg C yr−1 for the period 1948–2000, which is slightly lower than the estimates from the traditional parameterization method. The NPP and net ecosystem production (NEP) are, however, estimated to be 1.92 ± 0.19 and 0.05 ± 0.16 Pg C yr−1 for the period from 1948 to 2000, which is 19.1% lower and 61% higher than the estimates from the traditional parameterization (Table 4).

The average spatial patterns of the simulated NEP by the spatially explicitly and traditionally parameterized TEM are different over the period 1948–2000 (Figures 6a,b). With the traditional parameterization, TEM-estimated carbon sinks are generally larger in the Southeast United States but smaller in the Pacific Northwest area and the central lowland area in northern Minnesota and Michigan, compared to the estimates by the spatially explicitly parameterized TEM. The carbon-source areas are also estimated to be larger in Minnesota by the traditionally parameterized TEM.

Figure 6.
Figure 6.

Spatial patterns of NEP, VEGC, and SOLC estimated using TEM with the traditional and spatially explicit parameterization methods during the period of 1948–2000. (a) NEP estimated with traditional parameterization, (b) NEP estimated with spatially explicit parameterization, (c) VEGC estimated with traditional parameterization, (d) VEGC estimated with spatially explicit parameterization, (e) SOLC estimated with traditional parameterization, and (f) SOLC estimated with spatially explicit parameterization.

Citation: Earth Interactions 16, 5; 10.1175/2012EI400.1

Spatial patterns and magnitudes for vegetation and soil organic carbon storage are very different with the two methods of parameterization. Kolmogorov–Smirnov tests (Corder and Foreman 2009) on the estimated carbon pool datasets from the two methods confirm that they have significant differences, rejecting the null hypothesis at the 5% significance level, with p < 10−10 for soil organic carbon pools and p < 10−100 for vegetation carbon pools during the study period.

A single set of parameters for each ecosystem type (the traditional parameterization method) produces more continuously homogeneous spatial patterns for the carbon pools, whereas the spatially explicit method generates more discrete spatial distributions. Both the vegetation carbon and soil organic carbon pools (Figures 6d,f) have spatial patterns similar to the initial carbon pools used for the spatially explicit parameterization. For example, the traditional method estimates the deciduous forests in the middle United States stored more vegetation carbon than the other areas, whereas the results from the spatial parameterization indicate the highest vegetation carbon storage was located in the Pacific Northwest and along the Appalachian Mountains (Figure 6c). The spatial features of the soil organic carbon estimated by the two parameterization methods are more similar to each other when comparing the differences between the vegetation carbon pools. Both predict a large amount of soil organic carbon stored in the northeastern, southeastern, and midwestern forests, although the spatial parameterization method provides higher magnitudes (Figure 6e).

4. Discussion

With the developed spatially explicit parameterization method and simulations, we must further consider four issues. One is the verification of our results for carbon dynamics with the study region. A second focuses on investigating the role of carbon pools and fluxes in quantifying carbon dynamics. A third issue is the role of spatially explicit parameterization in regional carbon quantification. Finally, we consider the possible uncertainties associated with the use of the spatially explicit parameterization method and also future work.

4.1. Verification of the estimated carbon dynamics

There are various independently estimated carbon dynamics for the forests of the conterminous United States. Houghton et al. (Houghton et al. 1999) estimated the carbon sink in the forest of the conterminous United States to be 0.06 Pg C yr−1 in the 1980s (Houghton et al. 1999; Pacala et al. 2001); Birdsey and Heath (Birdsey and Heath 1995) reported the carbon sink of the forested regions to be about 0.10 Pg C yr−1 for the same period (Birdsey and Heath 1995) without considering land-use changes. However, a later study demonstrated the land-use changes have significant effects on the carbon budgets (Birdsey et al. 2006). Pacala et al. (Pacala et al. 2001) estimated the sink to be 0.11–0.15 Pg C yr−1 for the 1980s (Pacala et al. 2001). The historical simulated results of our study suggested that the carbon sink is 0.07 Pg C yr−1 for the 1980s with the spatially explicit parameterization method, which is closer to these independent estimations, compared with the 0.04 Pg C yr−1 estimated by the traditional parameterization method. Zhang et al. (Zhang et al. 2010) showed that the annual NPP of the forests in the conterminous United States increased from 1.5 Pg C yr−1 in the early twentieth century to 1.9 Pg C yr−1 in the early twenty-first century (Zhang et al. 2010). NPP estimated by the spatially explicitly parameterized TEM is about 1.92 Pg C yr−1 in the period of 1948–2000, which is lower than the traditional-method-estimated value of 2.37 Pg C yr−1 but closer to Zhang et al.’s (Zhang et al. 2010) result. The MODIS products start in the year 2000 and, for the period of 2000–07, the MODIS products estimate the average annual GPP and NPP of the study region to be 3.98 and 1.94 Pg C yr−1, respectively, which is closer to our estimates of the twentieth century with the spatially explicit parameterization method. Existing soil organic carbon is estimated to be about 25 Pg C (U.S. Department of Agriculture 2011), which is in between the estimations from the two methods used in our study (Table 4); however, our estimations of vegetation carbon (Table 4) with the spatially explicit parameterization is much closer to 24 Pg C (Sundquist et al. 2009). In summary, the spatially explicit parameterized TEM results presented here are broadly consistent with a wide range of previous studies on carbon dynamics in the same region.

4.2. The role of carbon pool sizes and carbon fluxes in parameterization

Estimates of the amount of carbon storage are important because they are a baseline for assessing potential future carbon storage gains or losses (Sundquist et al. 2009) and affect the net exchange of CO2 between forests and the atmosphere (Pregitzer and Euskirchen 2004). Conceptually, the vegetation carbon pool size determines the vegetation biomass and the leaf biomass and therefore influences the photosynthesis rate as modeled in the GPP formulation in TEM (Table 2; Zhuang et al. 2002) and the autotrophic respiration rate; the soil organic carbon pool size is also significant in the determination of heterotrophic respiration (Table 2). The nitrogen pool sizes (derived from carbon pool sizes in our study) concern the carbon–nitrogen interaction processes and therefore also play an important role in carbon dynamics. The carbon fluxes (annual GPP and NPP) used for parameterization are also of significant importance because they directly reflect the features of the ecosystem carbon assimilation rate and the autotrophic respiration rate. Technically, the related parameters are sequentially adjusted during the parameterization process in order to reach the observed carbon pool sizes and fluxes. The spatially explicit carbon pools and fluxes affect the parameter values and therefore affect spatial patterns and the magnitudes of the simulated carbon dynamics at the regional scale. As shown in our results, TEM, using the traditional parameterization, will predict much higher vegetation carbon storage in comparison with other estimates.

4.3. Importance of using spatially explicit parameters

Quantification of ecosystem carbon dynamics with TEM is influenced by parameters. Traditionally, these parameters are determined by calibrating the model at a number of representative sites. When these parameters are applied to the region, the regional grid cells are therefore assumed to have the same characteristics as the calibration sites. However, ecosystem processes are related not only to ecosystem type but also to various environmental and ecological factors (e.g., stand age, species, and geographic location) (Ahl et al. 2004; Monsi and Saeki 2005; Still et al. 2004; Turner et al. 1995). Therefore, because the model with the traditional parameterization is not able to account for the spatially heterogeneous features of the ecosystems, its estimation may be biased. The parameters for each grid cell of a region are therefore needed to better quantify the regional carbon dynamics.

To date, several studies revealed the importance of spatial parameters in model simulations. For example, De Kauwe et al. (De Kauwe et al. 2008) assimilated the spatial leaf area index (LAI) from MODIS over a coniferous forest site in Oregon into an ecosystem model with an ensemble Kalman filter, showing that assimilating the LAI data improved the NEP estimates. Studies on satellite-based terrestrial production models also suggest that the key parameter, light-use efficiency at the canopy level, varies spatially with different vegetation species, stand age, soil fertility, and climate (McCallum et al. 2009). In contrast, Davi et al. (Davi et al. 2006) tested the sensitivity of a combination of six key parameters: the aboveground wood biomass (B), the soil water reserve (SWR), the canopy clumping factor (CF), the LAI (L), the leaf mass per area of sunlit leaves (Msun), and the leaf nitrogen content (N) with the process-based model CASTANEA. The study suggested a slight difference in the estimation of carbon fluxes and almost no difference in the estimation of water fluxes between using spatially explicit parameters and aggregated parameters from a small study region. Our results, however, suggest there are significantly different magnitudes and spatial patterns (e.g., Figure 6 and Table 4) for carbon dynamics in the forest ecosystems of the United States, for both the past and future, between the estimations determined using the spatially explicit and the traditional parameterization methods. Because we use the same climate forcing data and spatial reference data for the model simulations for the two parameterization methods, the reason for the differences is most likely due to the use of the spatially explicit parameters versus the traditionally calibrated parameters. This suggests that the spatially explicit parameters are important to better quantifying regional ecosystem carbon dynamics.

4.4. Possible uncertainties and the future work

There are various possible uncertainty sources in our estimations of carbon dynamics with the spatially explicitly parameterized TEM. First, the uncertainty of the carbon pool size data used for the spatially explicit parameterization may contribute to the total uncertainty. The COLE data are based on forest inventory data and mapped with weighted analysis and have been suggested as an unbiased estimator of carbon mapping. The COLE-estimated carbon data are validated, but with statistical errors. The ensemble parameterizations are conducted to test how the errors of the carbon pool sizes affect the results. As shown in Figure 5, the standard deviation of the results of the ensemble simulations are within very small ranges, indicating that the carbon pool size error slightly alters the model estimates of carbon fluxes but notably influence the estimates of vegetation and soil organic carbon pool sizes (Figure 5). The uncertainty of the vegetation and soil organic carbon pool size estimation induced by the error in the carbon pool size in calibration is as low as about 3%–5% of the mean estimation. We also use two other types of resampling methods (bilinear and cubic) to assess the uncertainties introduced by the methods used to resample the spatially explicit carbon pools. We find the resampled results by the two methods to be similar to the results produced by the nearest-neighborhood method we used in this study. The discrepancies in both methods are in the ±5% tolerance range of our parameterization procedure, and therefore the uncertainties may not be obvious. The method of estimating NPPsat [Equation (3)] may also contribute to the uncertainty of the model calibration and therefore estimations of carbon dynamics. During the processes of calibration, over- or underestimation of NPPsat will lead to higher or lower values of Cmax (see the description of calibration procedures); the N limitation effect is therefore over- or underestimated in association with higher or lower Nmax values. LeBauer and Treseder (LeBauer and Treseder 2008) reported the ratio of estimated aboveground net primary productivity in the fertilized plots to the control plots for all biome types to be 1.22–1.35, and for temperature forests the ratio was 1.11–1.28, with a meta analysis of 126 nitrogen addition experiments evaluating the N limitation of NPP in terrestrial ecosystems. The ratio was, however, suggested to vary with geographic location and environmental conditions. The number we used in this study is within a reasonable range, but the number does not account for the spatial variability of the ratio. Further study in the future will be needed on this point. Another uncertainty may come from the nitrogen pool size. Because of data limitation, we estimate nitrogen pools using the C:N ratios of soils and vegetation in this study. Previous studies suggest that the C:N ratios may vary spatially (White et al. 2000), but recent sensitivity analysis on TEM parameters suggests TEM is not sensitive to the values of the initial C:N ratios as parameters (Tang and Zhuang 2009). Future effort should be made to provide more accurate spatially explicit pool sizes for nitrogen in both vegetation and soils to better quantify regional carbon dynamics.

In addition, GPP and NPP are directly correlated to the photosynthesis rate; thus, the errors in the GPP and NPP spatial data will result in a bias in the estimation of Cmax (i.e., a higher GPP and NPP will have a higher Cmax). Moreover, the gap between GPP and NPP is controlled by the parameter Kr, and thus an overestimation of the gap will lead to an overestimation of Kr and vice versa. Here, we use the spatial data from the MODIS GPP and NPP product, which were generated with a combination of satellite observations and process-based models (Running et al. 2004; Zhao et al. 2005). Although the satellite products were well validated with eddy flux tower data (Heinsch et al. 2006; Plummer 2006; Turner et al. 2006; Zhao et al. 2006), it might be more proper to use independent satellite-based estimations in the future (e.g. Xiao et al. 2008; Xiao et al. 2010; Yang et al. 2007). For future studies at a larger regional scale, several soil carbon inventory databases are available (Fischer et al. 2008; U.S. Department of Agriculture 2011). However, the spatial distribution of vegetation carbon storage for the conterminous United States or the globe is not available. With the development of lidar and radar remote sensing (Dubayah and Drake 2000; Kobayashi et al. 2000; Lefsky et al. 2009; Yu et al. 2010), it is possible to gradually get a global estimation of vegetation carbon pool sizes, which could be directly used as an input for the spatially explicit parameterization of TEM for a better quantification of ecosystem dynamics at regional and global scales.

Finally, the vegetation map we used in the spatially explicit calibration and the base land-cover map from the MODIS carbon flux products are not exactly matched with one another. The COLE data do not have a base land-cover map but are based on the county level. The potential vegetation map used in this study from Mellilo et al. (Mellilo et al. 1993) is at a 0.5° × 0.5° resolution. At the 1 km × 1 km level, both of these maps are coarser than the MODIS products. Errors therefore occur as a result of rescaling and matching these data onto the same spatial scale and range. In addition, these maps are produced in different periods, during which land-use and land-cover change may have happened. Therefore, mismatch of the vegetation maps of these data could happen, and it may affect the parameterization results and the model extrapolation results, mainly due to the following two reasons: 1) mistake in a grid’s vegetation type will result in the usage of improper parameters (e.g., VEGcn, Nav) during the spatially explicit calibration and thus produce wrong calibrations and 2) because the MODIS GPP and NPP products are based on MODIS land-cover products, the MODIS product may provide improper GPP and NPP data for a grid cell because of the mismatch of the MODIS land cover and the vegetation-cover type in this study. As discussed above, the errors in GPP and NPP may bring uncertainty to the results. In the current study, we rescaled all of the data into 0.5° × 0.5° grid cells because of the limit of computing resources and the consideration of the coarse COLE data, but a future study may conduct the spatially explicit parameterization at a higher spatial resolution and based on a unifying land-cover map to reduce the uncertainty.

5. Summary

Our study uses a spatially explicit parameterization method in our process-based ecosystem model. The new parameterization method is able to more adequately deal with the spatial heterogeneity of the ecosystems for the conterminous United States in estimating the carbon dynamics of forest ecosystems. The model parameters have high spatial variation in concert with highly heterogeneous soil and vegetation carbon distributions. The spatially explicit parameters, therefore, lead to distinct estimates of carbon dynamics of the conterminous United States forest ecosystems compared with the results from a traditional parameterization method. With the new parameterization method, the model estimates the net carbon exchanges between the conterminous United States forest ecosystem and the atmosphere to be 61% higher than the value estimated by the traditional method during the period of 1948–2000. The large difference between the two regional estimates indicates the importance of the spatially explicit parameters. This study suggests that more spatially explicit vegetation and soil carbon, nitrogen, and flux data are needed and should be used to improve future quantification of carbon dynamics at regional and global scales.

Acknowledgments

This research is supported by the NASA Land Use and Land Cover Change program (NASA- NNX09AI26G), Department of Energy (DE-FG02-08ER64599), National Science Foundation (NSF-1028291 and NSF- 0919331), and the NSF Carbon and Water in the Earth Program (NSF-0630319). The computing is supported by Rosen Center of high performance computing at Purdue. We also thank Ms. Jayne Piepenburg for English proofreading and editing.

References

  • Ahl, D., , S. Gower, , D. Mackay, , S. Burrows, , J. Norman, , and G. Diak, 2004: Heterogeneity of light use efficiency in a northern Wisconsin forest: Implications for modeling net primary production with remote sensing. Remote Sens. Environ., 93, 168178.

    • Search Google Scholar
    • Export Citation
  • Birdsey, R. A., , and L. S. Heath, 1995: Carbon changes in U.S. forests. Productivity of America’s forests and climate change, U.S. Department of Agriculture Forest Service Rocky Mountain Forest Experiment Station General Tech. Rep. RM-GTR-271, 56–70.

  • Birdsey, R. A., , K. Pregitzer, , and A. Lucier, 2006: Forest carbon management in the United States: 1600-2100. J. Environ. Qual., 35, 14611469.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1995: Land-atmosphere CO2 exchange simulated by a land surface process model coupled to an atmospheric general circulation model. J. Geophys. Res., 100, 28172831.

    • Search Google Scholar
    • Export Citation
  • Bondeau, A., , D. W. Kicklighter, , J. Kaduk, , and T. P. O. T. P. N. M. Intercomparison, 1999: Comparing global models of terrestrial net primary productivity (NPP): Importance of vegetation structure on seasonal NPP estimates. Global Change Biol., 5 (S1), 3545.

    • Search Google Scholar
    • Export Citation
  • Bresson, C. C., , A. S. Kowalski, , A. Kremer, , and S. Delzon, 2009: Evidence of altitudinal increase in photosynthetic capacity: Gas exchange measurements at ambient and constant CO2 partial pressures. Ann. For. Sci., 66, 505, doi:10.1051/forest/2009027.

    • Search Google Scholar
    • Export Citation
  • Corder, G. W., , and D. I. Foreman, 2009: Nonparametric Statistics for Non-Statisticians: A Step-by-Step Approach. Wiley, 264 pp.

  • Cramer, W., , D. Kicklighter, , A. Bondeau, , B. Moore III, , G. Churkina, , B. Nemry, , A. Ruimy, , and A. Schloss, 1999: Comparing global models of terrestrial net primary productivity (NPP): Overview and key results. Global Change Biol., 5, 115.

    • Search Google Scholar
    • Export Citation
  • Davi, H., and Coauthors, 2006: Effect of aggregating spatial parameters on modelling forest carbon and water fluxes. Agric. For. Meteor., 139 (3–4), 269287.

    • Search Google Scholar
    • Export Citation
  • De Kauwe, M., , T. Quaife, , P. Lewis, , M. Disney, , and M. Williams, 2008: Estimating the spatial exchange of carbon through the assimilation of Earth observation derived products using an ensemble Kalman filter. Proc. Geoscience and Remote Sensing Symp., Boston, MA, IEEE, Vol. 3, 1044–1047.

  • Dubayah, R. O., , and J. B. Drake, 2000: Lidar remote sensing for forestry. J. For., 98, 4446.

  • FAO, cited 2011: FAO/Unesco Soil map of the world (1:5 000 000), 1971-1981. FAO.

  • Fischer, G., , F. Nachtergaele, , S. Prieler, , H. T. van Velthuizen, , L. Verelst, , and D. Wiberg, 2008: Global Agro-ecological Zones Assessment for Agriculture (GAEZ 2008). IIASA and FAO Rep.

  • Gower, S. T., , R. E. McMurtrie, , and D. Murty, 1996: Aboveground net primary production decline with stand age: Potential causes. Trends Ecol. Evol., 11, 378382.

    • Search Google Scholar
    • Export Citation
  • Heinsch, F., and Coauthors, 2006: Evaluation of remote sensing based terrestrial productivity from MODIS using regional tower eddy flux network observations. IEEE Trans. Geosci. Remote Sens., 44, 19081923.

    • Search Google Scholar
    • Export Citation
  • Houghton, R. A., , J. L. Hackler, , and K. T. Lawrence, 1999: The U.S. carbon budget: Contributions from land-use change. Science, 285, 574578.

    • Search Google Scholar
    • Export Citation
  • Kicklighter, D. W., and Coauthors, 1999: Comparing global models of terrestrial net primary productivity (NPP): Global pattern and differentiation by major biomes. Global Change Biol., 5 (S1), 1624.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267.

    • Search Google Scholar
    • Export Citation
  • Kitajima, K., , S. S. Mulkey, , and S. Wright, 1997: Decline of photosynthetic capacity with leaf age in relation to leaf longevities for five tropical canopy tree species. Amer. J. Bot., 84, 702708.

    • Search Google Scholar
    • Export Citation
  • Knuth, D., 1997: Sorting and Searching. Vol. 3, The Art of Computer Programming, 3rd ed. Addison-Wesley, 723 pp.

  • Kobayashi, Y., , K. Sarabandi, , L. Pierce, , and M. C. Dobson, 2000: An evaluation of the JPL TOPSAR for extracting tree heights. IEEE Trans. Geosci. Remote Sens., 38, 24462454.

    • Search Google Scholar
    • Export Citation
  • LeBauer, D. S., , and K. K. Treseder, 2008: Nitrogen limitation of net primary productivity in terrestrial ecosystems is globally distributed. Ecology, 89, 371379.

    • Search Google Scholar
    • Export Citation
  • Lefsky, M. A., , W. B. Cohen, , G. G. Parker, , and D. J. Harding, 2009: Lidar remote sensing for ecosystem studies. Bioscience, 52, 1930.

  • McCallum, I., , W. Wagner, , C. Schmullius, , A. Shvidenko, , M. Obersteiner, , S. Fritz, , and S. Nilsson, 2009: Satellite-based terrestrial production efficiency modeling. Carbon Balance Manage., 4, doi:10.1186/1750-0680-4-8.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D., , J. M. Melillo, , L. A. Joyce, , D. W. Kicklighter, , A. L. Grace, , B. Moore III, , and C. J. Vorosmarty, 1992: Interactions between carbon and nitrogen dynamics in estimating net primary productivity for potential vegetation in North America. Global Biogeochem. Cycles, 6, 101124.

    • Search Google Scholar
    • Export Citation
  • McGuire, A. D., and Coauthors, 2001: Carbon balance of the terrestrial biosphere in the twentieth century: Analyses of CO2, climate and land use effects with four process-based ecosystem models. Global Biogeochem. Cycles, 15, 183206.

    • Search Google Scholar
    • Export Citation
  • Melillo, J. M., , A. D. McGuire, , D. W. Kicklighter, , B. Moore, , C. J. Vorosmarty, , and A. L. Schloss, 1993: Global climate change and terrestrial net primary production. Nature, 363, 234240.

    • Search Google Scholar
    • Export Citation
  • Monsi, M., , and T. Saeki, 2005: On the factor light in plant communities and its importance for matter production. Ann. Bot., 95, 549567.

    • Search Google Scholar
    • Export Citation
  • Pacala, S. W., and Coauthors, 2001: Consistent land- and atmosphere-based U.S. carbon sink estimates. Science, 292, 23162320.

  • Pietsch, S. A., , and H. Hasenauer, 2006: Evaluating the self-initialization procedure for large-scale ecosystem models. Global Change Biol., 12, 16581669, doi:10.1111/j.1365-2486.2006.01211.x.

    • Search Google Scholar
    • Export Citation
  • Plummer, S., 2006: On validation of the MODIS gross primary production product. IEEE Trans. Geosci. Remote Sens., 44, 19361938.

  • Potter, C. S., , J. T. Randerson, , C. B. Field, , P. A. Matson, , P. M. Vitousek, , H. A. Mooney, , and S. A. Klooster, 1993: Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Biogeochem. Cycles, 7, 811841.

    • Search Google Scholar
    • Export Citation
  • Pregitzer, K. S., , and E. S. Euskirchen, 2004: Carbon cycling and storage in world forests: Biome patterns related to forest age. Global Change Biol., 10, 20522077.

    • Search Google Scholar
    • Export Citation
  • Raich, J. W., and Coauthors, 1991: Potential net primary productivity in South America: Application of a global model. Ecol. Appl., 1, 399429.

    • Search Google Scholar
    • Export Citation
  • Ruimy, A., , L. Kergoat, , and A. Bondeau, 1999: Comparing global models of terrestrial net primary productivity (NPP): Analysis of differences in light absorption and light-use efficiency. Global Change Biol., 5, 5664.

    • Search Google Scholar
    • Export Citation
  • Running, S. W., , and J. C. Coughlan, 1988: A general model of forest ecosystem processes for regional applications I. Hydrologic balance, canopy gas exchange and primary production processes. Ecol. Modell., 42, 125154.

    • Search Google Scholar
    • Export Citation
  • Running, S. W., , R. R. Nemani, , 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
  • Schloss, A., , D. Kicklighter, , J. Kaduk, , and U. Wittenberg, 1999: Comparing global models of terrestrial net primary productivity (NPP): Comparison of NPP to climate and the normalized difference vegetation index (NDVI). Global Change Biol., 5, 2534.

    • Search Google Scholar
    • Export Citation
  • Still, C., , J. Randerson, , and I. Fung, 2004: Large-scale plant light-use efficiency inferred from the seasonal cycle of atmospheric CO2. Global Change Biol., 10, 12401252.

    • Search Google Scholar
    • Export Citation
  • Sundquist, E. T., , K. V. Ackerman, , N. B. Bliss, , J. M. Kellndorfer, , M. C. Reeves, , and M. G. Rollins, 2009: Rapid assessment of U.S. forest and soil organic carbon storage and forest biomass carbon sequestration capacity. U.S. Geological Survey Open-File Rep. 2009–1283, 23 pp.

  • Tang, J., , and Q. Zhuang, 2008: Equifinality in parameterization of process-based biogeochemistry models: A significant uncertainty source to the estimation of regional carbon dynamics. J. Geophys. Res., 113, G04010, doi:10.1029/2008JG000757.

    • Search Google Scholar
    • Export Citation
  • Tang, J., , and Q. Zhuang, 2009: A global sensitivity analysis and Bayesian inference framework for improving the parameter estimation and prediction of a process-based Terrestrial Ecosystem Model. J. Geophys. Res., 114, D15303, doi:10.1029/2009JD011724.

    • Search Google Scholar
    • Export Citation
  • Turner, D. P., , G. J. Koerper, , M. E. Harmon, , and J. J. Lee, 1995: A carbon budget for forests of the conterminous United States. Ecol. Appl., 5, 421436.

    • Search Google Scholar
    • Export Citation
  • Turner, D. P., and Coauthors, 2006: Evaluation of MODIS NPP and GPP products across multiple biomes. Remote Sens. Environ., 102 (3–4), 282292.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Agriculture, cited 2011: U.S. General Soil Map (STATSGO2). Natural Resources Conservation Service. [Available online at http://soildatamart.nrcs.usda.gov.]

  • Van Deusen, P., , and L. S. Heath, cited 2010a: COLE web applications suite. NCASI and USDA Forest Service Northern Research Station. [Available online at http://www.ncasi2.org/COLE/.]

  • Van Deusen, P., , and L. S. Heath, 2010b: Weighted analysis methods for mapped plot forest inventory data: Tables, regressions, maps and graphs. For. Ecol. Manage., 260, 16071612.

    • Search Google Scholar
    • Export Citation
  • White, M. A., , P. E. Thornton, , S. W. Running, , and R. R. Nemani, 2000: Parameterization and sensitivity analysis of the BIOME–BGC Terrestrial Ecosystem Model: Net primary production controls. Earth Interact., 4. [Available online at http://EarthInteractions.org.]

    • Search Google Scholar
    • Export Citation
  • Xiao, J., and Coauthors, 2008: Estimation of net ecosystem carbon exchange for the conterminous United States by combining MODIS and AmeriFlux data. Agric. For. Meteor., 148, 18271847.

    • Search Google Scholar
    • Export Citation
  • Xiao, J., and Coauthors, 2010: A continuous measure of gross primary production for the conterminous United States derived from MODIS and AmeriFlux data. Remote Sens. Environ., 114, 576591.

    • Search Google Scholar
    • Export Citation
  • Yang, F., and 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
  • Yu, Y., , S. Saatchi, , L. S. Heath, , E. LaPoint, , R. Myneni, , and Y. Knyazikhin, 2010: Regional distribution of forest height and biomass from multisensor data fusion. J. Geophys. Res., 115, G00E12, doi:10.1029/2009JG000995.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , J. Chen, , W. Ju, , S. Shen, , Y. Pan, , R. Birdsey, , and L. He, 2010: Carbon balance in conterminous U.S. forests based on historic changes in climate, atmospheric composition, and disturbances. Proc. Fall Meeting, San Francisco, CA, Amer. Geophys. Union, B41K-07.

  • Zhao, M., , F. A. Heinsch, , R. R. Nemani, , and S. W. Running, 2005: Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sens. Environ., 95, 164176.

    • Search Google Scholar
    • Export Citation
  • Zhao, M., , S. W. Running, , and R. R. Nemani, 2006: Sensitivity of Moderate Resolution Imaging Spectroradiometer (MODIS) terrestrial primary production to the accuracy of meteorological reanalyses. J. Geophys. Res., 111, G01002, doi:10.1029/2004JG000004.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , V. E. Romanovsky, , and A. D. McGuire, 2001: Incorporation of a permafrost model into a large-scale ecosystem model: Evaluation of temporal and spatial scaling issues in simulating soil thermal dynamics. J. Geophys. Res., 106, 33 64933 670.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , A. D. McGuire, , K. P. O’Neill, , J. W. Harden, , V. E. Romanovsky, , and J. Yarie, 2002: Modeling soil thermal and carbon dynamics of a fire chronosequence in interior Alaska. J. Geophys. Res., 107, 8147, doi:10.1029/2001JD001244.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., and Coauthors, 2003: Carbon cycling in extratropical terrestrial ecosystems of the Northern Hemisphere during the 20th century: A modeling analysis of the influences of soil thermal dynamics. Tellus, 55B, 751776.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., , J. M. Melillo, , D. W. Kicklighter, , R. G. Prinn, , A. D. McGuire, , P. A. Steudler, , B. S. Felzer, , and S. Hu, 2004: Methane fluxes between terrestrial ecosystems and the atmosphere at northern high latitudes during the past century: A retrospective analysis with a process-based biogeochemistry model. Global Biogeochem. Cycles, 18, GB3010, doi:10.1029/2004GB002239.

    • Search Google Scholar
    • Export Citation
  • Zhuang, Q., and Coauthors, 2006: CO2 and CH4 exchanges between land ecosystems and the atmosphere in northern high latitudes over the 21st century. Geophys. Res. Lett., 33, L17403, doi:10.1029/2006GL026972.

    • Search Google Scholar
    • Export Citation
Save