192 New Visualization Techniques, Verification Tools, and Results from the NWS Probabilistic Snowfall Experiment

Monday, 13 January 2020
Hall B (Boston Convention and Exhibition Center)
Jeff S. Waldstreicher, NOAA/NWS, Bohemia, NY; and D. B. Radell

National Weather Service (NWS) forecast offices across the country have been producing and disseminating experimental probabilistic snowfall (ProbSnow) forecasts during the past several years in an attempt to communicate forecast uncertainty to core stakeholders. This effort expanded to 90 NWS offices during the 2018-2019 winter season. The forecasts originate from a 46-member ensemble generated by the Weather Prediction Center (WPC), with a gridded field of probability distribution functions (PDF) provided to NWS weather forecast offices (WFOs). The WFOs use this probabilistic information, along with their official snowfall forecast, to produce “low end”, “expected/most likely” and “high end” snowfall accumulations, along with a suite of exceedance probability information. This information is posted to the web and used by the WFO in support of Decision Support Services.

Effectively visualizing, interpreting, and ultimately communicating probabilistic information (e.g., probability distribution functions) in an operational environment is a challenge. A recent effort utilizing forecast box and whisker plots, in conjunction with ensemble plumes of snow accumulations at various ProbSnow forecast points, has shown promise in providing forecasters insight on the ensemble envelope, including shape and skew of the PDF. Perhaps more importantly, box and whisker plots allow for the quick visualization of “where” the official snowfall forecast is positioned within the ensemble distribution. When viewed over successive forecast updates, this information can provide some context into the changing forecast “uncertainty” relative to the official forecast.

Verification consistently shows that the ProbSnow forecasts are statistically reliable, and user feedback indicates that the information is used by core partners in their decision making in planning for the impacts of snow. New verification visualization tools such as the Gridded Automated Zonal Precipitation and Complete Hi-Res Output (GAZPACHO) tool and ARCGIS have allowed for detailed spatial analysis of ProbSnow output. For example, these tools can easily highlight areas where the observed snowfall amount falls within, or outside, the 10th-90th percentile envelope, as a way to measure spatial reliability. Moreover, the addition of the ProbSnow output to the experimental National Digital Forecast Database in early 2019 will make gridded verification of these probabilistic forecasts much simpler.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner