Predicting crop yields using remotely sensed observations

Abstract:

Understanding the response of agriculture to environmental stressors is essential to adapt food systems to climate change. In much of the world, though, the response of crop yield to climate is unclear because of data limitations. In data sparse regions, yields are often only reported at the national-scale and weather stations have insufficient coverage. Here, we explore the utility of using remotely sensed observations to estimate both subnational yield and its response to climate. We downscale nationally reported yields using observations of solar-induced fluorescence, which serve as a proxy of vegetation productivity. We evaluate the viability of this method for a range of nutritionally important crops across Sub-Saharan Africa, highlighting case studies in Madagascar and Kenya. For regions and crops in which solar-induced fluorescence well predicts yield, we resolve the relationship between solar-induced fluorescence and water availability using remotely sensed soil moisture observations. These relations can then be used in future work to predict the effect of climate change on agricultural production.