38HYDRO Estimating 3-m Evapotranspiration Using Planet, OpenET, and Machine Learning Techniques

Given the growing challenges facing agriculture due to population growth, climate changes, and declining water supply, monitoring evapotranspiration (ET) at high spatiotemporal resolutions is critical to maximizing production while using limited water resources effectively. Efforts have been made to obtain a suitable and accurate ET product. For example, an open and easily accessible satellite-based ET dataset (OpenET) is available online to provide ET estimations at the field scale at daily and monthly time steps. The OpenET data have been widely used in irrigation scheduling and water resource management. However, the 30-m spatial resolution limits its utility for precision agriculture management for heterogeneous and complex croplands such as vineyards and orchards. Unmanned aerial vehicles (UAVs) can support high resolution mapping (sub-meter scale), but image acquisitions will be limited in time and space and can be labor- and cost-intensive. In contrast, the high-resolution (~4 m) satellite data provided daily by Planet’s multispectral satellite constellation may provide the necessary information to generate high-resolution ET products. This research explores obtaining ET at the Planet image spatial scale based on using OpenET data via machine learning approaches. The proposed Planet-scale ET is compared with UAV-based ET generated at 3.6 m resolution and eddy-covariance (EC) flux tower observations collected as a part of GRAPEX at several California vineyards representing significant range in climatological conditions where winegrapes are grown. The proposed method to obtain high spatial resolution ET takes advantage of the well-established OpenET program and provides critical crop water use information for water resource management and precision agricultural applications. The developed method may be further implemented and complement the existing OpenET operational system to provide high-resolution ET data in more heterogeneous areas.