Introducing physics to artificial intelligence methods to improve satellite monitoring of the water cycle

The project aims to develop, train, and apply a hybrid neural network model to optimise EO data for a coherent, balanced water cycle at the global scale resulting in a new pixel-resolution datasets for the four water cycle components: precipitation, evapotranspiration, change in water storage, and runoff (or river discharge). These data will cover the entire globe on quarter-degree grid cells and on a monthly time scale.