View article

There is a critical need to enhance the accuracy and utility of national soil moisture products by integrating new data sources and downscaling them to farm-scale. This research develops and validates national field-scale soil moisture products by using machine-learning approaches to integrate satellite, in situ and model-derived data and downscale them to field scale. Here we focus on comparing algorithms that were developed for agricultural environments (crops and pastures) and forested environments. Specifically, we compare the best machine learning algorithms and identify which features are most important for agriculture versus forests. In addition, we compare the out-of-sample validation of these national products using independent in situ measurements. Our results demonstrate that it is possible to accurately estimate soil moisture in unsampled agricultural and forested environments. These data are …