Abstract

This study investigates high-frequency mapping of downward shortwave radiation (DSR) at Earth's surface using the Advanced Baseline Imager (ABI) instrument mounted on Geostationary Operational Environmental Satellite-R Series (GOES-R). The existing GOES-R DSR product (DSR ABI) offers hourly temporal resolution and spatial resolution of 0.25 o. To enhance these resolutions, we explore machine learning (ML) for DSR estimation at the native temporal resolution of GOES-R Level-2 Cloud and Moisture Imagery (CMI) product (five minutes) and its native spatial resolution of two-kilometer at nadir. We compared four common ML regression models through the Leave-One-Out Cross-Validation (LOOCV) algorithm for robust model assessment against ground measurements from AmeriFlux and SURFRAD networks. Results show that Gradient Boosting Regression (GBR) achieves the best performance (R² = 0.916, RMSE = 88.05 W m-2) with efficient computation compared to Long Short-Term Memory (LSTM), which exhibited similar performance. DSR estimates from the GBR model (DSR ALIVE) outperform DSR ABI across various temporal resolutions and sky conditions. DSR ALIVE agreement with ground measurements at SURFRAD networks exhibits high accuracy at high temporal resolutions (five-minute intervals) with R² exceeding 0.85 and RMSE=122 W m-2. We conclude that GBR offers a promising approach for high-frequency DSR mapping from GOES-R, enabling improved applications for near-real-time monitoring of terrestrial carbon and water fluxes.