An improved downscaled sun-induced chlorophyll fluorescence (DSIF) product of GOME-2 dataset

Abstract

ABSTRACT Sun-induced chlorophyll fluorescence (SIF) is a proxy for plant photosynthesis. However, available SIF products either have a low spatial resolution or are spatially discontinuous. This paper proposes an improved downscaling method to generate a continuous 0.05-degree SIF dataset from GOME-2 retrievals, covering the period from February 2007 to March 2019. First, a random forest model was developed to predict SIF using training samples from GOME-2 SIF and explanatory variables with a resolution of 0.5 degrees (including reflectance at visible and near-infrared bands, vegetation index, temperature, and cosine values of the sun-zenith angle). Then, a 0.05-degree SIF dataset was predicted using the trained model and the corresponding explanatory variables at a resolution of 0.05 degrees. Subsequently, the predicted 0.05-degree SIF dataset was used as the weighting coefficient to redistribute the original 0.5-degree GOME-2 SIF to a downscaled 0.05-degree SIF dataset (DSIF) based on the energy conservation principle. The results showed that DSIF was more consistent with the original 0.5-degree retrievals than the similar GOME-2 derived spatially extended SIF dataset, which produced spatial details with a resolution of 0.05 degrees. The structure and physiology of SIF information were well represented in the DSIF, which is imperative for assessing global photosynthetic activity.