Modeling Gross Primary Production in North America with MODIS Images and Reanalysis Climate Data

Accurate estimation of gross primary production (GPP) of terrestrial ecosystems is vital for a better understanding of the spatial-temporal patterns of the global carbon cycle. In this study we estimated GPP in North America (NA), using the satellite-based Vegetation Photosynthesis Model (VPM), MODIS images at 8-day temporal and 500 m spatial resolution, and NCEP-NARR reanalysis climate data. The simulated GPP (GPPVPM) agreed well with the flux tower derived GPP (GPPEC) at 39 AmeriFlux sites (155 site-years). The GPPVPM in 2010 was spatially aggregated to 0.5 by 0.5 degree grid cell, and then evaluated with solar-induced chlorophyll fluorescence (SIF) data from GOME-2, which is often regarded as the proxy of direct measurement of vegetation photosynthesis. There were good agreements in spatial distribution and seasonal dynamics between GPPVPM and SIF. At biome scale, the relationship between GPPVPM and SIF showed strong linear correlations (R2 > 0.95) and small variations of slopes (4.60 - 5.55 g C m-2 year-1 / (mW m-2 nm-1 sr-1)). The total annual GPPVPM in NA in 2010 was approximately 13.53 Pg C year-1, which accounted for ~11.0% of the global terrestrial GPP and was within the range of annual GPP estimates from several other process-based and data-driven models (11.35 - 22.23 Pg C year-1). Forests contributed most (4.17 Pg C year-1) to the annual GPP in NA. Evergreen broadleaf forests were most productive with an average annual GPP exceeding 2000 g C m-2 year-1. The results from this study has demonstrated the reliable performance of VPM at the continental scale, and the resultant GPP product at 500-m spatial resolution provides more opportunities to improve the studies of carbon cycle, model inter-comparison, and benchmarking.