A New Operational System for Forecasting Precipitation Type

Joseph R. Bocchieri Techniques Development Laboratory, National Weather Service, NOAA, Silver Spring, Md 20910

Search for other papers by Joseph R. Bocchieri in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

A new system is developed which gives conditional probability forecasts for three precipitation type categories: snow or sleet, freezing rain and rain. Also, the probability forecasts are transformed into categorical forecasts so that a “best category”is provided.

The Model Output Statistics (MOS) technique is used with output from the Limited-area Fine Mesh (LFM) model to develop statistical forecast equations for each of several regions in the conterminous United States. To help account for the evaporational cooling effect, predictors such as LFM forecasts of boundary layer and 850 mb wet-bulb temperatures and observed surface and dew-point temperatures are included. Also, joint predictors are designed to help account for predictor interactions. The values of the joint predictors are relative frequencies of the freezing rain or snow categories taken from graphs that show these frequencies as joint functions of various pairs of LFM predictors.

Results from a statistical screening procedure vary by region but generally indicate that the 850 mb temperature and boundary-layer wet-bulb temperature joint predictor accounts for most of the reduction of variance of the snow category. For the freezing rain category, the 850–500 mb thickness and 1000–850 mb thickness, 1000–500 mb thickness and boundary-layer potential temperature, and the 850 mb temperature and boundary-layer potential temperature joint predictors are found to be important along with the observed surface temperature and dew point.

Verification of the new system on developmental and independent data samples indicates that the scores for the snow category are generally very good and stable; the results for the freezing rain forecasts are not newly as good.

Abstract

A new system is developed which gives conditional probability forecasts for three precipitation type categories: snow or sleet, freezing rain and rain. Also, the probability forecasts are transformed into categorical forecasts so that a “best category”is provided.

The Model Output Statistics (MOS) technique is used with output from the Limited-area Fine Mesh (LFM) model to develop statistical forecast equations for each of several regions in the conterminous United States. To help account for the evaporational cooling effect, predictors such as LFM forecasts of boundary layer and 850 mb wet-bulb temperatures and observed surface and dew-point temperatures are included. Also, joint predictors are designed to help account for predictor interactions. The values of the joint predictors are relative frequencies of the freezing rain or snow categories taken from graphs that show these frequencies as joint functions of various pairs of LFM predictors.

Results from a statistical screening procedure vary by region but generally indicate that the 850 mb temperature and boundary-layer wet-bulb temperature joint predictor accounts for most of the reduction of variance of the snow category. For the freezing rain category, the 850–500 mb thickness and 1000–850 mb thickness, 1000–500 mb thickness and boundary-layer potential temperature, and the 850 mb temperature and boundary-layer potential temperature joint predictors are found to be important along with the observed surface temperature and dew point.

Verification of the new system on developmental and independent data samples indicates that the scores for the snow category are generally very good and stable; the results for the freezing rain forecasts are not newly as good.

Save