Assessing the ability of MODIS EVI to estimate terrestrial ecosystem gross primary production of multiple land cover types

Terrestrial gross primary production (GPP) is the largest component flux in the global carbon cycle. Though various light-use-efficiency (LUE) models based on satellite vegetation greenness indices have been developed, the use of LUE models is usually hindered by difficulties in the parameterization of LUE and reliance on coarse meteorological inputs that have large uncertainties. The enhanced vegetation index (EVI) has been shown to have the potential to directly estimate GPP. However, the GPP-EVI relationship within different biomes has rarely been discussed. Here we explored relationships between annual integrated MODIS EVI (iEVI) and GPP from global eddy covariance measurements from 155 sites. Through categorizing these sites into 12 diverse biomes, the iEVI ability in estimating GPP was considerably improved (R2 from 0.56 to 0.70) in comparison to that without categorization. The biome-specific GPP-iEVI formulas were then performed at the global scale and the GPP based on iEVI agreed with a widely used benchmark dataset in either spatial or biome comparison. The predictive ability of iEVI was better in deciduous biomes than evergreen biomes. A significant negative correlation (R2 = 0.49, p < 0.02) between the strength of GPP-iEVI relationships and mean annual maximum leaf area index (LAImax) was consistently observed, showing the constraint of environmental conditions on the capability of EVI to estimate GPP. LAImax also revealed a scaling effect on the GPP-iEVI relationships, suggesting GPP models entirely based on MODIS EVI and LAI are feasible, particularly at an annual time-scale. This analysis highlights the direct link between EVI and GPP; inclusion of additional vegetation information will aid in improving the EVI performance.