Deriving Gross Primary Productivity (GPP) across different subtropical grasslands of Florida Using Landsat-8 corroborated by onsite Eddy Covariance GPP estimates

Abstract:

Florida subtropical grasslands, containing a prevalence of C4 grasses and rich cyperoid ground cover, serve local cattle ranches as major grazing lands that spans intensive introduced grasslands to native grasslands. Characterized by management intensities, they range from most intensively managed sown pastures to least managed rangelands, and all types include embedded seasonal wetlands. Gross primary productivity (GPP), defined as the photosynthetic rate at which plants capture and store chemical energy, provides essential information for better understanding the forage production of grazing lands. Previous studies have confirmed that the performances of remote sensed GPP models vary among subtypes of grasslands. In understanding of this variation in modeled GPP estimation, we improved a remote sensed GPP model covering multiple subtypes of grazing lands at Buck Island Ranch (BIR), Archbold Biological Station and the University of Florida, Range Cattle Research and Education Center (RCREC) in south central Florida, which are in the Long-Term Agroecosystem Research (LTAR) Network. Our proposed model, Improved Photosynthetic Capacity Model (IPCM), is an entirely satellite-based GPP model free of the input of spatially insufficient meteorological data. Landsat-8 and Eddy Covariance (EC) data were input for model parameterization and validation. Across four grassland subtypes (improved and semi-native pastures, wetlands, and rangelands), the correlation between the IPCM and the EC-derived GPP was relatively better in improved and semi-native pastures, in terms of the starting and ending phenological dates. The accuracy of IPCM was compared with that of the VPM and Landsat GPP products. With higher root mean squared error (RMSE) and the relative error (RE), the accuracy of GPP predicting for Landsat GPP products was lower than the other two models. RMSE ranged from 0.362 to 1.456 for the IPCM and from 0.517 to 1.586 for the VPM, and the largest difference of RMSE between two models was 0.704 in semi-native pastures. The performance of Landsat GPP products was lower than the other two in any terms of comparison. The IPCM was comparable with the VPM and higher than the Landsat GPP Products.