Comparison of empirical remote-sensing based models for the estimation of gross primary productivity using eddy covariance and satellite data over agroecosystem

Estimation of terrestrial gross primary productivity (GPP) is critical for global climate and ecological studies. However, the lack of multi-model studies for GPP estimation over agroecosystem in India limits the carbon budgeting at the regional scales. Satellite-derived parameters [(e.g., land surface temperature (LST), Enhanced Vegetation Index (EVI), Land Surface Water Index (LSWI)] combined with meteorological variables offer a promising tool for regional estimates of the GPP. In this study, site-specific GPP was evaluated based on the eddy-covariance (EC) tower data and satellite-derived parameters. Four satellite-based GPP models, (a) greenness and radiation (GR) model, (b) VI × VI model, (c) photosynthetic capacity model (PCM), and (d) temperature and greenness (TG) model have been compared for the estimation of GPP in Saharanpur Flux tower site (SFS) from April 2014 to April 2015 using meteorological variables from EC and the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite images. Among the predictive GPP models, TG models performed best with the RMSE of 2.03 g C m−2 day−1. The relationship of MODIS-LST with photosynthetically active radiation (PAR), GPP and air temperature (Ta) indicates that the climate variables are imperative for GPP estimation. In the VI × VI model series, the combination of EVI × EVI × PAR provided the best GPP estimates with an RMSE of 2.99 g C m−2 day−1. The comparative analysis of the GPP models has the potential for GPP estimates over agroecosystems and further carbon flux predictions at the regional scale.