Conveying Potential Snow Threat in Days 4-7 Given Predictability Limitations Using Multi-Model Ensembles

Forecasts of snow and its impacts on commerce (particularly travel) have generally improved in the short range forecast period (days 1-3) over the last decade. Improved observational networks (satellite, terrestrial, radar), and continued improvements in numerical models and data assimilation techniques, have played a large role in these improvements. The growing availability of guidance from various ensemble systems of models, has allowed development of probabilistic approaches in forecasting wintertime precipitation. Novak et al. (2014) examined the approach used at the National Weather Service (NWS), Weather Prediction Center (WPC). Their approach is based upon developing percentile snow accumulations. These accumulations are derived from probabilistic cumulative distribution functions (CDFs) applied to model ensemble forecasts to produce forecast snowfall range forecasts that vary dynamically, with the resultant ranges specified by the spread of the ensemble forecasts.

During the 2014-2015 winter season, four Weather Forecast Offices (WFOs) in the NWS Eastern Region, leveraged the aforementioned short range probabilistic snowfall guidance from the WPC, to create and disseminate an experimental suite of probabilistic winter snowfall graphical products to external users (e.g., Jacks and Novak 2015). Anecdotal evidence from this effort showed that key user groups (e.g., government officials and transportation departments) benefitted from availability of probabilistic snowfall information.

For the medium range forecast period (days 4-7), snowfall forecasting is more difficult than in the short range, because of the inherent less predictable nature of wintertime precipitation at these longer time ranges. Despite being more challenging, there is increased demand for identifying potential winter impacts in the medium range forecast period.

During 2015-2016 winter season, WPC will make available to WFOs and external users, probabilistic winter weather outlooks for the medium range (days 4-7) twice daily. These outlooks leverage WPC's expertise in providing quantitative precipitation forecast (QPF) guidance by using the deterministic QPF prepared by WPC medium range forecasters. The Day 4-5 and Day 6-7 deterministic QPF is disaggregated into 24-hr amounts and used as a mean in a cumulative distribution function (CDF) where 90 ensemble members from the Global Ensemble Forecast System (GEFS), European Center Ensemble System (ECENS), and Canadian Meteorological Center Ensemble (CMCE) are used as variance, to estimate a probability of WPC QPF exceeding 0.25 inches for a respective 24-hour period on 1x1 degree grid. An ensemble probability of snow and sleet using precipitation type fields from each of the 90 members of the multi-ensemble systems is combined with the QPF threshold probability to arrive at the probability of snow and sleet exceeding 0.25 inches (melted). The precipitation type is computed for each ensemble member using 2-meter temperatures, and several mandatory isobaric levels (925 hPa, 850 hPa, and 700 hPa).

For the 2015-16 winter season, WFO Sterling, Virginia, plans to leverage the WPC medium range probabilistic winter precipitation outlooks to create experimental guidance intended to convey information to its key users on the impacts of potential snow events in days 4-7. This guidance will also be based upon WFO-Sterling's staff expertise and experience working with key users to assess impacts of wintertime precipitation events, and the staff's expertise in local meteorology in the mid-Atlantic region. This presentation will report on the methodology developed along with preliminary results.

REFERENCES: Eli Jacks and David R. Novak, 2015: NWS Efforts to Improve Forecasting and Messaging of Uncertainty. AMS 27th Weather Analysis and Forecast Conf, 10A.5, July 2015, Chicago, IL.

David R. Novak, Keith F. Brill, and Wallace A. Hogsett, 2014: Using Percentiles to Communicate Snowfall Uncertainty. Wea. Forecasting, 29, 1259–1265.