Advancing Science to Better Characterize Drought and Groundwater Driven Low-Flow Conditions in NOAA & USGS National-Scale Models

The National Oceanic and Atmospheric Administration’s National Water Center on the campus of the University of Alabama has developed a numerical model called the “National Water Model” (NWM) that generates forecasts of streamflow in all 2.7M reaches in the United States. To date, the primary objective of the NWM has been to predict flooding events to aid federal, state, and local agencies in emergency preparedness and response. Another aspect of the NWM forecasting that has received little attention to date is streamflow under extended periods with little or no precipitation. This portion of streamflow which is not directly related to a specific precipitation event is commonly called base flow and is primarily fed by seepage from near surface soil moisture and groundwater. Baseflow is important for a variety of reasons. In agricultural regions that rely on irrigation, severe reductions in baseflow can be highly problematic. Power generation facilities and various types of industry often rely on baseflow. Furthermore, boats and barges that travel in shallow rivers depend on a certain amount of baseflow for these rivers to be navigable.

Our team of researchers from Brigham Young University (BYU), The University of Alabama (UA), and the University of Vermont (UVM) is developing state-of-the-art methods to improve continental United States (CONUS)-scale operational streamflow forecasts for low-flow conditions. We will do this by training machine learning algorithms using multiple datasets including landcover/land-use, gravity anomalies, rainfall, soil moisture, and groundwater level datasets to identify regions where low flows are influenced by groundwater. We will leverage the on-going USGS efforts to develop machine learning models and enhance performance with feature selection. Once we identify regions where low flows are influenced by groundwater, we will characterize the interactions between groundwater and flow in gauged basins and develop methods to predict baseflows more accurately in regional streams. We will refine and augment groundwater data using Earth observations to compensate for sparse and missing groundwater data. We will develop physical models to calibrate and improve our machine learning algorithms and use the resulting prediction methods to generate input for the new version of the national water model called the NextGen National Water Model which is currently in development.