Improving snowfall accumulation predictions by post-processing ensemble forecasts with an Artificial Neural Network

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 19 January 2010: 2:30 PM
B204 (GWCC)
Tyler C. McCandless, Penn State University, University Park, PA; and S. E. Haupt and G. Young

Presentation PDF (267.8 kB)

Correct forecasting snowfall amount has widespread economic and safety consequences for the Mid-Atlantic and Northeast United States. Due to the complex characteristics and dynamics inherent in winter weather systems, snowfall accumulation forecasts tend to have a large amount of uncertainty associated with them. Numerical Weather Prediction (NWP) ensemble models were developed to address the uncertainty in weather forecasts. However, a deterministic forecast is of utmost importance to the public; therefore, several post-processing methods of combining ensemble members have been developed. This study examines the use of an Artificial Neural Network (ANN) as a post-processing method for improving 24 hour snowfall accumulation predictions from the Global Ensemble Forecast System (GEFS). The ANN relates the observed predictand, 24 hour snowfall, to forecasts from the GEFS, observations, and location information such as elevation. A network of 2060 NCDC Cooperative Summary of the Day (co-op) snowfall accumulation reports for the Mid-Atlantic and Northeast are used in this study.