ForecastWatch May 2010 Customer Newsletter

The May 2010 statistics have been loaded into ForecastWatch Online. The precipitation statistics have changed for May, and will be re-aggregated for previous months.

Previously, the precipitation tab on ForecastWatch Online showed the combination of rain forecasts versus rain events and snow forecasts versus snow events. However, not all ASOS stations report snowfall. They are all, for the most part, good at reporting QPE, or "liquid-equivalent" (though there is some debate over heater performance, etc. when its not liquid). So now, for May 2010 and forward (and April 2010 backward, over time) will show precipitation (rain or snow) forecasts versus precipitation observations.

Additionally, previously a forecast that mentioned a thunderstorm was not considered a precipitation forecast. For example, "Chance of thunderstorm" was not, but "Afternoon shower or thunderstorm" was considered a precipitation forecast. Now, thunderstorm forecasts are considered precipitation events.

There are a few main things you will notice with this new methodology. First, the forecast counts will be half what they used to be. This is because it used to be each forecast was counted twice, first as a rain forecast, then second as a snow forecast. Now a forecast is only counted once, as a "precipitation" forecast. Second, you'll notice the percentage of precipitation forecasts has increased, due to the addition of thunderstorm forecasts as precipitation forecasts. While hit rate, ORSS, and Brier seem to have improved, obviously false alarm rate and bias have noticeably changed.

This change has been made after consulting with a number of you over the past two years, along with others. I believe it is a positive change and an improvement to the way precipitation accuracy is calculated. However, because separate rain, snow, and thunderstorm forecasts are also important, I'm working on separate tabs for those to be added in the future. Please let me know if you are interested in these additional statistics.