Deriving Vegetation Variables from Satellite Observations using a Data-driven Approach

Abstract

This presentation was shared at the virtual conference of the 2025 CISESS Summer Internship Program held on 20 August 2025. The slides were presented by Alan Wang, an undergraduate student at the University of Maryland, under the mentorship of Heshun Wang (CISESS/UMD). Building on previous research in processing Earth Observation data, the presentation detailed the performance of the XGBoost, Cubist, and random forest regression models in deriving in-situ measurements of vegetation cover fraction (fCover) from satellite observations. Ground measurements of fCover from 43 National Ecological Observatory Network (NEON) sites provided labels for training, validation, and testing, which were then upscaled to align with the high spatial resolution land surface reflectance data provided by the Visible Infrared Imaging Radiometer Suite (VIIRS) daily surface reflectance (VNP09GA) product. When evaluated against unseen data, the random forest regression model demonstrated the best agreement (R-squared = 0.912, MAE = 0.043), followed by the XGBoost regressor (R-squared = 0.910, MAE = 0.043) and lastly the Cubist model (R-squared = 0.904, MAE = 0.047). Applying the random forest model on the 2023 VIIRS data for the East Coast produced estimates consistent with the expected annual phenological cycle.