Ensemble forecasting is an emerging technique for improving short-range (0 to 48 hour) weather forecasts, and understanding the relative importance of initial condition uncertainty versus model physics uncertainty is crucial to determining how best to construct ensembles.

This study illustrates the important and beneficial role that model physics uncertainty can play in short-range ensemble forecasting, particularly with respect to warm season precipitation forecasts.

Ensembles are groups of weather forecasts, valid over the same time period, that differ in either their initial state of the atmosphere or in their model physics components or both. Since the atmosphere is a chaotic system, meaning that forecasts are sensitive to small uncertainties in any part of the forecast system, it is very doubtful that any single model forecast will be correct in every detail. Thus, ensembles are one approach to predicting the probability of future weather events. This study examines the relative contributions of the uncertainties in our initial picture of the atmosphere and in our modeling systems to ensembles, and concludes that both sources of uncertainty can be used beneficially in an ensemble approach. The uncertainties in the initial picture of the atmosphere are most important to include in ensembles when the atmospheric signal is strong, whereas the uncertainties in the modeling system are most important to include in ensembles when the atmospheric signal is weak.

Dr. David J. Stensrud, Leader, Models and Assimilation Team,
Research Meteorologist
Forecast Research and Development Division
NOAA/National Severe Storms Laboratory
1313 Halley Circle
Norman, Oklahoma 73069