Empirical and model-based estimates of spatial and temporal variations in net primary productivity in semi-arid grasslands of Northern China

Spatiotemporal variations in net primary productivity (NPP) reflect the dynamics of water and carbon in the biosphere, and are often closely related to temperature and precipitation. We used the ecosystem model known as the Carnegie-Ames-Stanford Approach (CASA) to estimate NPP of semiarid grassland in northern China counties between 2001 and 2013. Model estimates were strongly linearly correlated with observed values from different counties (slope = 0.76 (p < 0.001), intercept = 34.7 (p < 0.01), R(2) = 0.67, RMSE = 35 g C·m(-2)·year(-1), bias = -0.11 g C·m(-2)·year(-1)). We also quantified inter-annual changes in NPP over the 13-year study period. NPP varied between 141 and 313 g C·m(-2)·year(-1), with a mean of 240 g C·m(-2)·year(-1). NPP increased from west to east each year, and mean precipitation in each county was significantly positively correlated with NPP—annually, and in summer and autumn. Mean precipitation was positively related to NPP in spring, but not significantly so. Annual and summer temperatures were mostly negatively correlated with NPP, but temperature was positively correlated with spring and autumn NPP. Spatial correlation and partial correlation analyses at the pixel scale confirmed precipitation is a major driver of NPP. Temperature was negatively correlated with NPP in 99% of the regions at the annual scale, but after removing the effect of precipitation, temperature was positively correlated with the NPP in 77% of the regions. Our data show that temperature effects on production depend heavily on recent precipitation. Results reported here have significant and far-reaching implications for natural resource management, given the enormous size of these grasslands and the numbers of people dependent on them.