Improving NOAA's Water Tools Through Physically Motivated, Data-driven, and Hybrid Modeling Techniques

NOAA’s hydrologic model forecasts support warnings of multiple hydrologic hazards such as droughts and floods, and also support water and ecosystems managers who need hydrologic information for their management decisions. This presentation will describe a project underway at the NOAA Physical Sciences Laboratory to support and contribute to the improvement of NOAA’s water-prediction tools through the development and evaluation of data-driven (i.e., machine learning) and physically motivated hydrological models, working towards hybrid modeling techniques and tools that leverage the strengths of both methods. The project tackles modeling of several different aspects in the hydrological system. These projects include the estimation of snow and the sensitivity of snow-generated streamflow forecasts to snowpack spatial resolution, novel ways to characterize drought via evaporative demand and using the complementary relationship between evaporative demand and evapotranspiration, and the generation of process-based ensemble streamflow forecasts through data-driven post-processing of meteorological forecasts. This presentation will broadly describe each of these projects, including the partnerships that motivated their forecast target. It will end with a discussion on the strengths of taking a holistic approach toward hydrologic system development.