Detecting regional GPP variations with statistically downscaled solar-induced chlorophyll fluorescence (SIF) based on GOME-2 and MODIS data.

Solar-Induced chlorophyll Fluorescence (SIF) is associated with vegetation canopy photosynthesis and is potentially used to retrieve Gross Primary Productivity (GPP). However, the coarse resolutions of the currently available SIF satellite data limit their applications. To expand the applicability of the SIF dataset, a framework was developed to disaggregate the Global Ozone Monitoring Experiment-2 (GOME-2) SIF dataset, which was based on statistical relationships between SIF and remotely sensed measurements of the Normalized Difference Vegetation Index (NDVI), the fraction of absorbed photosynthetically active radiation (fPAR), the soil moisture index and Land Surface Temperature (LST). The statistical relationships were established within a zone of n × n pixels (n∈[1, 25]) with a moving window technique. The regression function established within n × n pixels with the smallest Root Mean Square Error (RMSE) and highest coefficient of determination (R2) was selected for downscaling regression. Compared with the fixed window technique (n = 5) and theglobal regression model, the moving window technique presented low residuals and high R2 values. Validated with flux-tower eddy covariance measurements, the GPP retrieved within the downscaled SIF data shows the potential to improve vegetation GPP prediction, and the downscaled SIF could trace the seasonal phenology of evergreen forests. [ABSTRACT FROM AUTHOR]