CW3E National FIRO Expansion Pathfinder | Watershed QPF Skill
Primary support by the U.S. Army Corps of Engineers FIRO Program | Return to CW3E Homepage.
Select a watershed to view charts.
Data and Support:

The data presented on this webpage are part of a study by Cordeira et al. (in prep.) titled "A Summary of U.S. Watershed Precipitation Forecast Skill and the National Forecast Informed Reservoir Operations Expansion Pathfinder Program."

Funding for this project was primarily provided by awards from the U.S. Army Corps of Engineers (W912HZ-19-2-0023 and W912HZ-24-2-0001) and by the State of California Department of Water Resources (#4600014942) at the Center for Western Weather and Water Extremes (CW3E) at the University of California, San Diego Scripps Institution of Oceanography.

Additional funding for this project was provided by the National Oceanic & Atmospheric Administration (NOAA) awarded to the Cooperative Institute for Research to Operations in Hydrology (CIROH) through the NOAA Cooperative Agreement with The University of Alabama (NA22NWS4320003) and the Miramar Charitable Foundation on behalf of Eaton and Margaret Scripps.

Methodology:

The QPF data are derived from the NOAA Global Ensemble Forecast System v12 Reforecast (GEFS-R) dataset (Guan et al. 2022) for the period 2000–2019 over the contiguous U.S. (CONUS). Forecast data from the control member of the ensemble are initialized at 0000 UTC for 1 January 2000 through 31 December 2019 and obtained with a horizontal grid spacing of 0.25-degree latitude × 0.25-degree longitude. The NOAA GEFS-R was accessed on 2 September 2023 from https://registry.opendata.aws/noaa-gefs-reforecast. Six-hour forecasts of precipitation were subsequently aggregated into nine lead times corresponding to 24-hour periods ending at 1200 UTC for forecast hours 12–36 (day 1) through 204–228 (day 9).

The quantitative precipitation estimates (QPE) used for evaluation are derived from the Parameter-elevation Regressions on Independent Slopes Model (PRISM) developed by Oregon State University (Daly et al. 1994, 2002, 2008). These data are a combination of point observations, a digital elevation model, and other geographical datasets that are modeled to generate a native 4-km × 4-km gridded daily QPE ending at 1200 UTC each day for a period of record that both predates and extends beyond the GEFS-R period of record. The previously described choice to evaluate QPF in 24-hour periods ending at forecast hours 36 (day-1) through hour 228 (day-9) is motivated by the 24-h aggregation period of the PRISM QPE analysis ending at 1200 UTC. Paired QPF and QPE data are verified as a function of nine lead times during a period spanning 1200 UTC 10 January 2000 through 0000 UTC 31 December 2019.

Both the QPF and QPE data are converted into mean areal precipitation (MAP) values for 2125 USGS Hydrologic Unit Code (HUC)-8-sized watersheds within the CONUS boundary. The MAP is constructed by downscaling the QPF and slightly upscaling the QPE using bilinear interpolation to an array of points within each watershed with a horizontal spacing of 0.05 degrees latitude and longitude. The geometries of the HUC-8 watershed boundaries were obtained from the USGS Watershed Boundary Dataset (https://www.usgs.gov/national-hydrography/watershed-boundary-dataset).

The QPF skill was assessed using the critical success index derived from a 2 × 2 contingency matrix based on threshold values of MAP. The critical success index measures what fraction of observed and/or forecasted events were correctly predicted and can be thought of as the accuracy when correct negatives have been removed from consideration. The chosen thresholds are similar to those used by Sukovich et al. (2014) and Novak et al. (2014) to assess CONUS QPF skill. These thresholds include a precipitation above the watershed-relative top-10% and top-5% values in 24 hours. The top% thresholds are motivated by previous QPF skill analyses by Ralph et al. (2010).

The top% are chosen based on the upper % of wet days for the 2000–2019 period in each watershed. The critical success index will not be calculated where the number of events in a given watershed meeting the threshold requirements is fewer than 10 in order to control for variable results as a function of limited sample size. Note that the methodology also measures the effectiveness of the relatively coarse resolution GEFS-R in resolving potentially localized extreme precipitation within the PRISM dataset. This resolution discrepancy is reduced by computing the MAP and it should not impact relative comparisons of the values from one region to the next or from one lead time to another.

The locations of 749 sites with USACE regulatory involvement (e.g., dams, levees, etc.) are plotted and analyzed along with the spatial analysis in this study to aid in the discussion of the results. The locations of these sites were obtained from the National Inventory of Dams (https://nid.sec.usace.army.mil). All 749 sites are retained even though all infrastructure is not necessarily capable of supporting FIRO (e.g., a levee does not contain a managed outlet).