C:N driven modeling process, two major factors may have contributed to the
uncertainty of the NPP prediction: the constant C that is used for C:N ratio
prediction and the N prediction model. As a result of unavailable field C and N
data, this study only contributes on narrowing the uncertainty in NPP simulation
that is introduced by the constant C:N ratio for Canadian mixed grasslands.
8.5 Conclusions
This study describes a method for integrating a critical remote sensing derived
model parameter with the spatial BIOME-BGC model to estimate NPP in Canadian
mixed grasslands. Consistent with previous findings in the literature (Psomas et al.
2008), the spatial BIOME-BGC model was also found to be highly sensitive to the
critical model parameter foliar C:N ratio. Given the importance of the foliar C:N
ratio in terrestrial biochemical cycles and the ability of remote sensing in providing
spatially distributed foliar C:N, we coupled remote sensing derived C:N maps and
ecosystem modeling in order to increase model accuracy for Canadian mixed
grasslands.
MODIS data were used to derive spatial foliar C:N values for the study area. The
foliar C:N ratio map indicated that MODIS derived C:N ratio has a much lower
mean than the model default C:N ratio. Grassland NPP was simulated using a foliar
C:N ratio map to drive the BIOME-BGC model, and field NPP data collected in
2005 were used to evaluate the model results. We found that simulated NPP based
on spatially-derived foliar C:N parameter is better than that based on the fixed C:N
parameter to reflect actual ground conditions. Further analysis indicated that simulated and observed NPP displayed acceptable correlations and RMSE. In comparison with fixed C:N values, spatial C:N greatly increased the accuracy of modeling
results, although both simulated NPP outputs overestimated observed NPP. These
results demonstrate the importance of using spatially explicated foliar biochemical
parameters as an input to ecosystem process models. The use of spatial foliar C:N
ratio could also lead to a better understanding of local interactions on biogeochemical cycles thus improving model accuracy.
Further work will focus on developing an experimental-based carbon and nitrogen dataset for different vegetation communities in the study area, and establishing
remote sensing based C:N ratio models for these vegetation communities. In the
longer term, we will also investigate other important model parameters and develop
methodologies to provide spatially explicit parameters to further improve model
accuracy.
Acknowledgements This research was supported by ISTPCanada C4507 to Dr. Xulin Guo and
NSERC to Dr. Yuhong He. Grateful thanks to Kristina Trusilova for Max-Planck-Institut fu ¨r
Biogeochemie, Germany for generously sharing the codes for the terrestrial ecosystem model
GBIME-BGCv1.0.
8 Grassland Productivity Simulation: Integrating Remote Sensing. . .
167
uncertainty of the NPP prediction: the constant C that is used for C:N ratio
prediction and the N prediction model. As a result of unavailable field C and N
data, this study only contributes on narrowing the uncertainty in NPP simulation
that is introduced by the constant C:N ratio for Canadian mixed grasslands.
8.5 Conclusions
This study describes a method for integrating a critical remote sensing derived
model parameter with the spatial BIOME-BGC model to estimate NPP in Canadian
mixed grasslands. Consistent with previous findings in the literature (Psomas et al.
2008), the spatial BIOME-BGC model was also found to be highly sensitive to the
critical model parameter foliar C:N ratio. Given the importance of the foliar C:N
ratio in terrestrial biochemical cycles and the ability of remote sensing in providing
spatially distributed foliar C:N, we coupled remote sensing derived C:N maps and
ecosystem modeling in order to increase model accuracy for Canadian mixed
grasslands.
MODIS data were used to derive spatial foliar C:N values for the study area. The
foliar C:N ratio map indicated that MODIS derived C:N ratio has a much lower
mean than the model default C:N ratio. Grassland NPP was simulated using a foliar
C:N ratio map to drive the BIOME-BGC model, and field NPP data collected in
2005 were used to evaluate the model results. We found that simulated NPP based
on spatially-derived foliar C:N parameter is better than that based on the fixed C:N
parameter to reflect actual ground conditions. Further analysis indicated that simulated and observed NPP displayed acceptable correlations and RMSE. In comparison with fixed C:N values, spatial C:N greatly increased the accuracy of modeling
results, although both simulated NPP outputs overestimated observed NPP. These
results demonstrate the importance of using spatially explicated foliar biochemical
parameters as an input to ecosystem process models. The use of spatial foliar C:N
ratio could also lead to a better understanding of local interactions on biogeochemical cycles thus improving model accuracy.
Further work will focus on developing an experimental-based carbon and nitrogen dataset for different vegetation communities in the study area, and establishing
remote sensing based C:N ratio models for these vegetation communities. In the
longer term, we will also investigate other important model parameters and develop
methodologies to provide spatially explicit parameters to further improve model
accuracy.
Acknowledgements This research was supported by ISTPCanada C4507 to Dr. Xulin Guo and
NSERC to Dr. Yuhong He. Grateful thanks to Kristina Trusilova for Max-Planck-Institut fu ¨r
Biogeochemie, Germany for generously sharing the codes for the terrestrial ecosystem model
GBIME-BGCv1.0.
8 Grassland Productivity Simulation: Integrating Remote Sensing. . .
167