Sunday, January 23, 2005
Winter storm forecast evaluation
Jason Samenow @ 11:30 AM
How consistent and accurate was CapitalWeather.com's forecast?
CapitalWeather.com's stuck to its forecast of 4-8", with a most likely snowfall of 6" originally made 60 hours before the storm. On the morning of the storm, CapitalWeather.com commented that the lower end of the range (4-6") was most likely. The actual snowfall was 3-5" across the area. CapitalWeather correctly identified the time of onset of the precipitation, the period of heaviest precipitation (and worst travel conditions), and end of storm.
How did CapitalWeather.com's forecast compare with other forecasters?
CapitalWeather.com's forecast was the most accurate and consistent among public and private forecasters that I am aware of. The National Weather Service originally called for 4-8" (like CapitalWeather) before strangely increased its forecast to 6-10" at 11:36am during the storm even after models had significantly reduced precipitation totals. Joe Bastardi of AccuWeather called for 10"+. Dave Tolleris of WxRisk.com called for 8-14". Bob Ryan forecasted 5-9" and Topper Shutt forecasted 6-10." Not sure about Doug Hill's or Sue Palka's forecast, but I would be very surprised if their respective forecasts (made Friday night) were for any less than 4-8".
What mistakes did CapitalWeather.com and other forecasters make?
Simply put, we relied on computer model guidance too much. Twelve hours prior to the storm, all of the computer models were forecasting 0.75-1.0" of melted precipitation, which would equate to 8-14" of snow, or perhaps 6-10" of a snow/sleet mix. The actual amount of melted precipitation that materialized was 0.3-0.4" (less than half that amount). The models were too strong and too far south with the clipper. Because the clipper was weaker and tracked further north than depicted in the models, we got a much reduced initial surge from the clipper (Saturday morning) and when it transferred its energy to the coast, we got caught in the dry slot (between it and the coastal low forming off the Virginia Capes).
Did CapitalWeather.com account for the appropriate uncertainties in its forecast? In a word, "yes"! We provided probabilistic forecasts and emphasized that there was a possibility that snowfall amounts could be outside of the "most likely" range. Furthermore, in our post Friday night/Saturday morning, we accurately described the uncertainties that could and did reduce snowfall in the area (namely, a more northerly track of the low and the dry slot).
What to take home from this experience? Not to over rely on computer models. Climatologically, we were aware that the setup was not ideal for a huge snow event in the DC area and should have put more weight on the possibility that amounts would be modest.
Comments are closed for this archived entry | Link | 
Email this post
Email this post
