Retrievals of Chlorophyll Fapar for Improved Crop Gpp Modeling

Abstract:

Accurate estimation of crop gross primary productivity (GPP) is important. We have recently developed an algorithm to derive fAPAR of chlorophyll (fAPAR_chl), fAPAR of foliage (fAPAR_foliage) and chlorophyll LAI (LAI_chl) with PROSAIL2. The MODIS surface reflectance produced with MAIAC were utilized to retrieve fAPAR_chl, fAPAR_foliage and LAI_chl for three AmeriFlux sites of maize and soybean. MOD15A2 FPAR and the retrieved fAPAR_chl were compared with field fAPAR_canopy and the fraction of PAR absorbed by green leaves of the vegetation (fAPAR_green). MOD15A2 FPAR overestimated field fAPAR_canopy in spring and in fall, and underestimated field fAPAR_canopy in midsummer whereas fAPAR_chl correctly captured the seasonal phenology. The retrieved fAPAR_chl agreed well with field fAPAR_green at early crop growth stage in June, and was less than field fAPAR_green in late July, August and September, which is consistent with crop physiology theory. GPP estimates with fAPAR_chl and with MOD15A2 FPAR were compared to tower flux GPP. GPP simulated with fAPAR_chl was corroborated with tower flux GPP. Improvements in crop GPP estimation were achieved by replacing MOD15A2 FPAR with fAPAR_chl which also reduced uncertainties of crop GPP estimates by 1.12 – 2.37 g C m^-2 d^-1.

NDVI, EVI, WDRVI_green, and CI_green have also been employed to estimate GPP by other scientific teams. We investigated the scaling factors and offsets (i.e., regression slopes and offsets) between fAPAR_chl and the VIs. The results showed that the scaled EVI obtained the best performance. The performance of the scaled NDVI, EVI and WDRVI_green was improved across sites, crop types and soil/background wetness conditions, compared to the original un-scaled VIs. The scaled green band indices (WDRVI_green, CI_green) did not exhibit superior performance to either the scaled EVI or NDVI in estimating crop daily GPP at these agricultural fields.

MOD15A2 LAI and the retrieved LAI_chl were implemented into CLM to simulate GPP, respectively. Simulated GPP series with both MOD15A2 LAI and LAI_chl were compared to tower flux GPP. GPP simulation with LAIchl was significantly improved, compared to GPP simulation with MOD15A2 LAI. Our results show that the fAPAR_chl/LAI_chl products significantly improve the accuracy and reduce the uncertainty of GPP estimates, compared to MOD15A2 FPAR/LAI.