Verification of thunderstorm occurrence using the National Lightning Detection Network

Since the summer of 1983, the University at Albany, State University of New York, has organized a daily thunderstorm probability forecast contest. The contest involves predicting the probability that thunder will be heard during a 24-h period at each of ten locations across the continental United States and the forecasts are verified by standard METAR reports. In recent years however, there have been several instances, particularly during overnight hours, in which a thunderstorm failed to be reported in the METAR observations despite its occurrence. During such instances, the forecast contest was verified by (i) contacting the attendant National Weather Service office directly, and/or (ii) examining Weather Surveillance Radar-1988 Doppler and National Lightning Detection Network (NLDN) data. Given that the NLDN has continuous space and time coverage, an average detection efficiency in excess of 90% and mean location errors of ~500 m for cloud-to-ground (CG) lightning flashes, this presentation will examine the utility of using the NLDN to verify thunderstorm occurrence. In particular, a 15-year thunderstorm climatology (1995–2008) from Global Summary of the Day station observations will be compared to an NLDN lightning climatology at the ten stations currently used in the Albany Thunderstorm Contest and ten other first-order reporting stations distributed across a wide variety of climatological synoptic environments. Questions such as (i) how close to a station does CG lightning need to strike for a thunderstorm to be reported and does this distance vary by station, (ii) are there detectable trends and interannual variability in thunderstorm occurrence, and (iii) how do mean thunderstorm day soundings and standard thunderstorm forecasting variables vary from station to station will be addressed in this presentation.