A long-term simulation of forest carbon fluxes over the Qilian Mountains

- The original MOD_17 model was optimized using refined input data and parameters.
- Sensitivity analysis of the ecophysiological parameters of Biome-BGC model was conducted.
- Biome-BGC was calibrated through the integration strategy.
- Long-time series forest carbon fluxes and climatic factors were analyzed.

In this work, we integrated a remote-sensing-based (the MODIS MOD_17 Gross Primary Productivity (GPP) model (MOD_17)) and a process-based (the Biome-BioGeochemical Cycles (Biome-BGC) model) ecological model in order to estimate long-term (from 2000 to 2012) forest carbon fluxes over the Qilian Mountains in northwest China, a cold and arid forest ecosystem. Our goal was to obtain an accurate and quantitative simulation of spatial GPP patterns using the MOD_17 model and a temporal description of forest processes using the Biome-BGC model. The original MOD_17 model was first optimized using a biome-specific parameter, observed meteorological data, and reproduced fPAR at the eddy covariance site. The optimized MOD_17 model performed much better (R2 = 0.91, RMSE = 5.19 gC/m2/8d) than the original model (R2 = 0.47, RMSE = 20.27 gC/m2/8d). The Biome-BGC model was then calibrated using GPP for 30 representative forest plots selected from the optimized MOD_17 model. The calibrated Biome-BGC model was then driven in order to estimate forest GPP, net primary productivity (NPP), and net ecosystem exchange (NEE). GPP and NEE were validated against two-year (2010 and 2011) EC measurements (R2 = 0.79, RMSE = 1.15 gC/m2/d for GPP; and R2 = 0.69, RMSE = 1.087 gC/m2/d for NEE). NPP estimates from 2000 to 2012 were then compared to dendrochronological measurements (R2 = 0.73, RMSE = 24.46 gC/m2/yr). Our results indicated that integration of the two models can be used for estimating carbon fluxes with good accuracy and a high temporal and spatial resolution. Overall, NPP displayed a downward trend, with an average rate of 0.39 gC/m2/yr, from 2000 and 2012 over the Qilian Mountains. Simulated average annual NPP yielded higher values for the southeast as compared to the northwest. The most positive correlative climatic factor to average annual NPP was downward shortwave radiation. The vapor pressure deficit, and mean temperature and precipitation yielded negative correlations to average annual NPP.