Using Jensen-Shannon distance to better understand the role of landscape heterogeneity in the relationship between TROPOMI SIF product and Gross Primary Production
Daniel E. Pabon-Moreno
Sun-Induced Fluorescence (SIF) is considered to be a valuable signal detectable from space that provides direct information about Gross Primary Production (GPP). Previous studies have shown a high correlation between SIF estimated from satellite observations and GPP predicted using satellite images and machine learning techniques. Many times, SIF and GPP products are trained and validated from in-situ measurements, however, often a perfect match is assumed between the area sensed by the satellite and the area sensed in-situ. For this reason, it is important to quantify the representativeness of the in-situ observations when compared with satellite products at coarser resolution. In the present work, we evaluated the representativeness of different eddy covariance towers footprints when compared with TROPOMI SIF ungridded product. To quantify the representativeness, we quantify the amount of information shared by the vegetation around the tower and the vegetation sensed by the Sentinel-5p satellite based on Sentinel-2 data cubes and Jensen-Shannon distance. We expect that characterizing the mismatch with this Jensen-Shannon distance will help improve the correlation between SIF from the satellite and GPP estimations from the tower. Finally, to guarantee that our analysis fulfills the FAIR principles, we will also present a general workflow to run the analysis on-demand using the Copernicus Data Space Ecosystem infrastructure.
