The Relationship between Spread and Forecast Error in Extended-range Forecasts

Timothy W. Barker Department of Meteorology, University of Utah, Salt Lake City, Utah

Search for other papers by Timothy W. Barker in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The relationship between ensemble forecast spread and ensemble forecast error is examined for a large number of extended range forecasts with a fairly simple, yet realistic model. A “perfect model” approach is used so that systematic modeling errors do not overwhelm errors that grow from initial analysis error. Mean square errors and spreads of the forecasts are computed and discussed for this model. The correlation between spread and forecast error tends toward zero at long forecast times, and an explanation of this tendency is presented. Time averaging has little impact on improving the correlation between spread and forecast error. The correlation between spread and forecast error is sensitive to the region being considered, but it is not significantly different for regional domains versus hemispheric domains.

Abstract

The relationship between ensemble forecast spread and ensemble forecast error is examined for a large number of extended range forecasts with a fairly simple, yet realistic model. A “perfect model” approach is used so that systematic modeling errors do not overwhelm errors that grow from initial analysis error. Mean square errors and spreads of the forecasts are computed and discussed for this model. The correlation between spread and forecast error tends toward zero at long forecast times, and an explanation of this tendency is presented. Time averaging has little impact on improving the correlation between spread and forecast error. The correlation between spread and forecast error is sensitive to the region being considered, but it is not significantly different for regional domains versus hemispheric domains.

Save