15th AMS Conference on Weather Analysis and Forecasting
Norfolk, Virginia, 19-23 August 1996, in press



Harold E. Brooks, John V. Cortinas Jr., Paul R. Janish, David J. Stensrud

NOAA/ERL/National Severe Storms Laboratory
Norman, Oklahoma


During the winter of 1995-6, an ongoing pilot project to test short-range ensemble forecasting (SREF) being carried out at the National Center for Environ-mental Prediction captured several significant winter events. The details of what days were modelled and characteristics of the ensemble are discussed elsewhere in this volume (see Brooks et al. 1996 and references therein). We provide an example on the potential use of the ensemble for forecasting winter precipitation type , with a comparison to the 29-km Eta (Black 1994), the "meso-Eta", forecast for one case of hazardous winter weather. We will use only the ten 80-km Eta members of the ensemble in this work. Our focus is on the forecast of precipitation type.


The ensemble forecast initialized at 1200 UTC on 31 January 1996 covered a significant winter weather outbreak with the complete assortment of precipitation types and extremely cold temperatures. Beginning on 1 February, freezing rain and drizzle began to fall in a band from Texas into the western Carolinas. Particularly hard hit were areas in Louisiana, Mississippi and Alabama. We're focusing here on the freezing rain part of the forecast at 0000 UTC on 2 February, when freezing rain extended from Texas into extreme northwestern Alabama, with a second area in the Carolinas.

We compare the two sets of forecasts (the meso-Eta and the SREF) made at 1200 UTC on 31 January 1996, valid at 0000 UTC on 2 February. Both forecasts use the same algorithm to forecast a conditional precipitation type (Baldwin et al. 1994). The meso-Eta produces a binary (yes/no) forecast of precipitation type at a large number of points in the United States. For the SREF product, a set of ten forecasts all run with the 80-km horizontal grid spacing version of the Eta model, exists. With the latter, we've "averaged" the output of the precipitation algorithm to get a probability of precipitation type. For example, if all ten forecasts produce freezing rain, the probability from the ensemble would be 100%; if five produce freezing rain, the probability would be 50%. (We have no guarantees that the ensemble is spanning the true space of probabilities, but this represents a simple starting point of potential interest.)

The forecasts have interesting similarities and differences (Fig. 1). In particular, the ensemble indicates a significant probability of freezing rain through Mississippi up to the Alabama border, and a second area in North Carolina. There are six locations forecast for freezing rain in the single run from the meso-Eta, three in Texas, and one each in Louisiana, eastern Tennessee, North Carolina, and Kansas. The two forecasts agree in the southwestern part of the forecast area, although the SREF highest confidence is a little further south, and in the eastern part. The biggest area of disagreement is in Mississippi, where the meso-Eta has no freezing rain forecast, and the ensemble has greater than 30% probability over a large area with a maxima of 50%. In the 2 hours either side of the forecast valid time, freezing precipitation was observed from Texas into extreme northwestern Alabama, with a second area in the southern Appalachians (Fig. 2).

Both forecasts overforecast the threat in North Carolina, while picking up on the threat in Texas and Louisiana. The biggest difference is in Mississippi and northwestern Alabama, where the meso-Eta indicates no threat, while the ensemble has significant probabilities there. The ensemble does seem to have the area of highest probability a little south of the observed region. However, the threat to Mississippi and Alabama is much more obvious. This illustrates one potential value of the ensemble technique---the identification of areas of threat from weather hazards.


Baldwin, M., R. Treadon, and S. Contorno, 1994: Precipitation type prediction using a decision tree approach with NMC's mesoscale Eta model. Preprints, 10th Conf. on Numerical Weather Prediction, Amer. Meteor. Soc., Portland, Oregon, 30-31.

Black, T., 1994: The new NMC mesoscale Eta model: Description and forecast examples. Wea. Forecasting, 9, 265-278.

Brooks, H. E., D. J. Stensrud, and M. S. Tracton, 1996: Short-range ensemble forecasting pilot project: A status report. Preprints, 11th Conf. on Numerical Weather Prediction, Amer. Meteor. Soc., Norfolk, Virginia, this volume.