389
Verifying Model Output Statistic Variables and Observer Forecasts for Mount Washington, New Hampshire

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 5 January 2015
Michael Dorfman, Mount Washington Observatory, North Conway, NH; and M. A. Carmon and E. P. Kelsey

The nonprofit Mount Washington Observatory located on the summit of Mount Washington (44.3N 71.3W, 1917 m asl), New Hampshire is an official reporting station for the National Weather Service. Model Output Statistics (MOS) numbers play an important role in helping meteorologists at the Mount Washington Observatory (MWO) reporting station (KMWN) produce forecasts for the summit of Mount Washington. In this study, MOS numbers and higher summits forecasts produced by MWO forecasters are compared to conditions recorded on the summit for one winter season (October, 2012-March 2013) and one summer season (April 2013-September 2013). While comparison studies have previously been performed between forecasts, MOS outputs, and actual conditions for most other stations in the US, the impact of the mountainous terrain on Mount Washington's weather presents a unique challenge to forecasters. This study compares GFS MOS, NAM MOS, a non-weighted and weighted GFS-NAM consensus, and MWO forecasts to actual summit conditions for most MOS variables, including daily highs and lows, temperature, dew point, cloud cover, wind direction, wind speed, 6 and 12 hour probability of precipitation, 6 and 12 hour quantitative precipitation, 6 and 12 hour probability of thunderstorms, and 24-hour snowfall categorical forecasts. To calculate the non-weighted GFS-NAM consensus, the mean value for each weather variable is determined between the two models, and in order to determine the weighted consensus, the minimum variance-estimated weights are determined over a training period. The Brier Score is used to rate the accuracy of percentage variables and the mean error is used to rate the accuracy of numerical variables. These statistics are analyzed to determine if biases exist in the evaluated MOS variables and regression methods are used to attempt to correct for such biases. In addition, MOS skill is analyzed during specific atmospheric events, such as frontal passages and coastal storms, to determine if there are any predictable biases during particular weather patterns. Overall, this analysis demonstrates the accuracy of various MOS parameters for the KMWN reporting station and will help to improve meteorologists' skill in forecasting for the higher summits.