Improving seasonal water supply predictions through improved observations and high resolution, physics-based modeling

HydroInspector web mapping service display of NCAR observation stations (white circles), State of Colorado observation stations (green triangles), NRCS SNOTEL sites (blue squares) and observation and model time series (right hand side time series plots) from the Upper Rio Grande observation and modeling project.
Development of accurate seasonal water supply forecasts presents one of the most significant challenges in managing tightly administered waters of the Upper Rio Grande River basin. An additional challenge is presented by the lack of high spatial and temporal resolution of precipitation and snowpack in this relatively remote, high elevation watershed. To address this issue RAL has partnered with the Conejos Water Conservancy District, the State of Colorado, the NOAA Severe Storms Laboratory and NASA’s Jet Propulsion Laboratory to coordinate a comprehensive observational and modeling study, called RIO-SNO-FLO, to explore opportunities for improving seasonal water supply forecasts. A research radar, an airborne lidar and multispectral imager and multiple in situ surface measurement stations were deployed to provide new observations of precipitation, snowpack, soil moisture, streamflow and other meteorological conditions (See Figure 1). A more detailed description of these observations is provided in the Hydrometeorological Observations section of this report. Research conducted during 2016 documented the performance of radar estimated vs. precipitation gauge measured snowfall, and observed vs. modeled snowpack depth and near surface temperature, humidity and incoming solar radiation. These results are summarized in a report to the State of Colorado (Gochis et al., 2016) and a manuscript now in preparation (Karsten et al. 2016). The principal outcomes of this work are that research radars possess significant skill in estimating mountain snowfall as validated by surface precipitation gauges in the southern Colorado region and that when used to drive a physics-based hydrologic model, resulting snowpack and streamflow simulations were significantly improved over simulations using background national analyses of precipitation. As a result, the State of Colorado is currently considering the purchase and deployment of a gap-filling radar in the Upper Rio Grande basin.

Experimental real-time WRF-Hydro (blue diamonds) and operational (NOAA/WGRFC-black circles, NRCS-green triangles) seasonal water supply forecasts. Figure provided by James Heath of the Col. Div. of Water Resources.
During FY2016, improvements in the implementation of WRF-Hydro and in the specification of meteorological forcings gained during the first year of the RIO-SNO-FLO research program were incorporated into a new experimental seasonal water supply prediction system. This capability used an ensemble research version of the WRF-Hydro/National Water Model described above. During the 2016 Water Year, NCAR/RAL produced monthly, seasonal water supply forecasts and compared those real-time forecasts with other operational water supply forecasts from the Natural Resources Conservation Service (NRCS) and the NOAA/West Gulf River Forecast Center. The performance of those forecasts shown in the figure below, highlight that the real-time experimental WRF-Hydro ensemble total seasonal streamflow volume forecasts (blue triangles) compared favorably against the other two operational forecasts as compared with the observe runoff volume (blue line). This system is now being upgraded and expanded for Water Year 2017.
RAL is expanding the ensemble seasonal water supply forecasting domain to encompass all of the central Rocky Mountain headwater basins emanating from Colorado. (See Figure)

Map of expanded central Rocky Mountain seasonal water supply forecasting domain to be used beginning in Water Year 2017. Inset watersheds indicate headwater catchments used in model calibration.
The streamflow prediction development activities associated with this effort include:
- Expansion of the modeling domain to provide a more robust assessment of the ensemble water supply prediction system in regions other than the Rio Grande headwaters.
- Improved WRF-Hydro calibration methods, similar to those being developed for the NWM effort described above.
- Development of new, enhanced forcing and streamflow forecast bias-correction techniques.
- Enhanced evaluation of model forcings and model snowpack with additional observational instrumentation being deployed by project collaborators.
- Deployment of four additional in situ surface observations in the Upper Taylor River basin near Crested Butte, Colorado for enhanced snowpack monitoring and model evaluation. (See the Hydrometeorological Observations section of this report for more details on this activity.)
- Display of model analyses and forecasts on the HydroInspector web mapping service