National Weather Service California Nevada River Forecast Center Department of Commerce National Oceanic and Atmospheric Administration National Weather Service

Water Resources Forecast Verification Map Help

These map-based verification displays show the mean absolute percent error for long-range volumetric forecasts.  Either the water year or seasonal (April-July) runoff volume forecast errors can be displayed by checking the appropriate box in the menu.  These displays are available for the first day of any month in a water year.  Select the desired first-of-month date using the dropdown menu.  Low percent error values are ideal and indicate low model error while high percent error values are not ideal and indicate higher model error. Cool colors (blues) designate locations with low percent error values (low error) while warm colors (reds) designate locations with high percent error values (high error)

These verification statistics are not based on an archive of operational forecasts because that data set is pretty limited.  They are based on a hindcast (reforecast) effort where water year and seasonal runoff volume forecasts were generated every day from 1990 through 2023 using the Hydrologic Ensemble Forecast Service (HEFS).  More information on HEFS theory can be found here.  

These hindcasts were run in an automated fashion using current hydrology models, and do not include any user-specified model changes from hydrologic forecasters.  This is different from operational water supply forecasting procedures where forecasts incorporate changes to model inputs such as snowpack, soil moisture, and precipitation based on CNRFC forecaster expertise.  This is important to keep in mind when interpreting these results.

The legend at the bottom of the map labeled ‘Mean Absolute Percent Error’ represents the mean absolute percent error normalized by the 34-year average. These statistics were calculated for each point on the map as follows:

1. For each year of the HEFS hindcast period (1990-2023), the error is calculated:

2. The mean absolute error (MAE) over the 34 years is calculated:

3. The MAE is normalized by the 34-year average and converted to a percentage:

Points can be selected by directly clicking on an MAE icon on the map.  A pop-up box is available by clicking on an individual point. The pop-up box includes the id, location name, percent error and a link to additional reforecast verification information for that specific location.

Example of pop-up box: