Estimating global maximum gross primary productivity of vegetation based on the combination of MODIS greenness and temperature data

Abstract

Accurate estimation of the spatial-temporal variation of the maximum gross primary productivity (GPPmax) of vegetation is of great significance for predicting carbon fluxes and vegetation-climate feedback. In this study, the GPPmax estimation models were constructed based on driving variables derived from the MODIS enhanced vegetation index (EVI) and land surface temperature (LST) time series data using a stepwise regression analysis. The models were calibrated and validated with the observed GPPmax from 145 FLUXNET sites with 734 site-year data from 2000 to 2014. The GPPmax estimates and the changes in their trends at the global scale were also compared with other GPPmax products from the vegetation photosynthesis model and the eddy covariance-light use efficiency model. The results showed that the GPPmax for both forest and non-forest vegetation types were estimated well, with R2 of 0.47–0.86 and 0.47–0.95 and relative root mean square error of 10.14–35.14% and 11.25–30.02% for calibration and validation data, respectively. Summer EVI and spring EVI and LST played important roles in explaining the variation in GPPmax. The GPPmax estimates from this study and the changes in their trends were highly correlated with GPPmax estimates from the vegetation photosynthesis model, with R2 > 0.70 for most vegetation types. The GPPmax significantly increased in western North America, northern Europe, and eastern China, but decreased in tropical regions. This study concluded that the variation in GPPmax for various vegetation types on a global scale can be accurately estimated based on MODIS EVI and LST time series data, which provides a simple but effective way for large-scale estimation of GPPmax.