Estimation of gross primary production (GPP) from remote sensing data is an important approach to study regional or global carbon cycle. However, for a given algorithm, it usually has its limitation on applications to a wide range of vegetation types and/or under diverse environmental conditions. This study was conducted to compare the performance of two remote sensing GPP algorithms, the MODIS GPP and the vegetation photosynthesis model (VPM), in a semiarid temperate grassland ecosystem.Methods
The study was conducted at a typical grassland site in Ujimuqin of Inner Mongolia, North China, over 2 years in 2006 and 2007. Environmental controls on GPP measured by the eddy covariance (EC) technique at the study site were first investigated with path analysis of meteorological and soil moisture data at a daily and 8-day time steps. The estimates of GPP derived from the MODIS GPP and the VPM with site-specific inputs were then compared with the values of EC measurements as ground truthing at the site. Site-specific εmax (α) was estimated by using rectangular hyperbola function based on the 7-day flux data at 30-min intervals over the peak period of the growing season (May to September).Important Findings
Between the two remote sensing GPP algorithms and various estimates of the fraction of absorbed photosynthetic active radiation (FPAR), the VPM based on FPAR derived from the enhanced vegetation index (EVI) works the best in predicting GPP against the ground truthing of EC GPP. A path analysis indicates that the EC GPP in this semiarid temperate grassland ecosystem is controlled predominantly by both soil water and temperature. The site water condition is slightly better simulated by the moisture multiplier in the VPM than in the MODIS GPP algorithm, which is a most probable explanation for a better performance of the VPM than MODIS GPP algorithm in this semiarid grassland ecosystem.