Using mPING Observations to Verify Surface Precipitation Type from Operational Numerical Models

The mPING app allows the public citizen to submit reports of the weather occurring at their location from anywhere on the globe. This study uses precipitation type reports made through mPING in the continental United States to verify precipitation type forecasts of operational numerical models. The models evaluated are the North American Mesoscale (NAM) model, the Global Forecast System (GFS), and the Rapid Refresh (RAP) model. Strengths and weaknesses of each model's forecast are investigated for freezing rain, ice pellets, rain, and snow.

4x4 contingency tables are constructed to evaluate the models ability to forecast all four precipitation types. Then 2x2 contingency tables are formed to evaluate model performance for each precipitation type. The Heidke and Peirce skills scores are used predominantly, along with other performance measures. The NAM had the best skill scores overall for both the three and six-hour forecasts. However, when examining model performance for each precipitation type no model was superior to any other. The rare precipitation types showed a decrease in skill for all models. Overall, the models show less skill in the rare events of freezing rain and ice pellets, while overcompensating those precipitation types for rain or snow.