Photovoltaic solar power generation has grown exponentially in recent years and will further accelerate. Solar development induces a land use land cover change which may alter water and energy cycles and increase water stress, but these impacts have been overlooked. On a global scale, two road blocks for studying such impacts on water are: (i) no detailed data exist on where, how much, what kind of solar power exist; (ii) models have not incorporated physical processes related to solar panels. We initiate our effort in addressing these two questions on two ends: (1) from basic, we will create a process-based model to describe the physical energy and water cycles of solar power farms; (2) develop a two-level deep learning method to extract type, density, and configuration information from remote sensing images.
