Enhancing precision in evapotranspiration estimation: AI-powered downscaling of VIIRS LST

Abstract

Land Surface Temperature (LST) serves as a keystone in environmental research, offering invaluable insights into the Earth's surface energy balance, climate monitoring, and ecosystem health. The significance of LST is further underscored by its pivotal role in estimating EvapoTranspiration (ET), a fundamental component of the Earth's hydrological cycle and agricultural systems. Accurate ET estimates are indispensable for effective water resource management, optimizing agricultural productivity, and maintaining ecosystem health. Recent leaps in remote sensing technology, coupled with the development of cutting-edge machine learning models, have paved new avenues for downscaling LST data to finer resolutions. These advancements empower researchers with access to LST data at unprecedented granularity, ultimately illuminating the intricate dynamics of Earth's surface temperature. In this context, our primary research objective is to procure high-resolution LST data to refine the precision of evapotranspiration estimation, particularly within the Al Gharb region of Morocco. Our approach involves downscaling Visible Infrared Imaging Radiometer Suite (VIIRS) LST data using predictors derived from Landsat-8, facilitating a comparative analysis and detailed examination. This comparison serves as a stepping-stone, guiding our transition to Sentinel-2 data for further refinement. By harnessing the distinctive capabilities of Sentinel-2 satellite imagery and machine learning algorithms. The fine-scale LST data acquired at a remarkable 10-meter resolution unlocks new possibilities for monitoring and managing evapotranspiration with unprecedented accuracy. Our research contributes significantly to the realms of sustainable agriculture, water resource management, and climate change adaptation, all tailored to the unique environmental conditions of the Al Gharb region.