Verification of the University of Washington's 12 km Probabilistic Ensemble Forecast for Small Craft Advisories in the Strait of Juan de Fuca
BRYAN M. BURLINGAME and BRADLEY R. COLMAN
National Weather Service, Weather Service Forecast Office, Seattle, WA
The overall idea behind this verification is to determine the capabilities and limitations of the University Washington's (UW) 12 km WRF-Ensemble model in predicting the probability of Small Craft Advisories (SCA, winds greater than 18 knots) and of Gale Force winds (GF, winds greater than 34 knots) through the Strait of Juan de Fuca in Western Washington State. *While the accepted wind speed for a SCA for the NWS in Seattle is > than 21knots, for this exercise that limit was set at > 18 knots to coincide with the UW's model parameters for a SCA.* This verification has been conducted to hopefully enhance the forecasting capabilities of the National Weather Service's (NWS) north-west regional office located in Seattle, WA, by determining the accuracy of an insightful tool developed to focus on the marine push through the Strait of Juan de Fuca.
The verification examined just over one month of data, beginning June 12th, 2013, and continuing through July 22nd, 2013 (Excludes data from June 25-28 due to the absence of this information in the UW archive). The 00z and 12z runs of the model were examined for each day for both the SCA and GF wind probability for the 06, 12, 18, and 24 hour forecasts. This was compared with the observed wind data from the Race Rock, Victoria weather station located at the eastern end of the Strait of Juan de Fuca. The probability was related to the magnitude of observed wind, and was analyzed to determine the number of times each percent category (0-10%, 10-30%, 30-50%, 50-70%, 70-90%, and 90% +) verified with winds of SCA or GF magnitude. This data was also broken into the different forecast hours that were looked at to determine how the ensemble performed at different stages in its forecast.
After completion of this preliminary verification, results have revealed that the ensemble forecast for the probabilities of SCA and GFW lack slightly in calibration. The SCA probability has resulted with each percent category, when forecasted, resulting in more events of winds greater than 18 knots then what is to be expected. For the GF probability, no significant conclusions can be made yet as the analyzed data yielded no forecasts greater than 10% probability.