Diagnosis of systematic forecast errors dependent on flow pattern

Authors


Abstract

Singular-value decomposition (SVD) analysis is employed to study flow-dependent forecast errors. The results presented are based on ECMWF operational forecasts and verifying analysis of 500 hPa heights for the most recent seven winter periods. Beyond forecast day 3 the flow-dependent errors are mainly localized over the Atlantic sector and are associated with the North Atlantic Oscillation- (NAO)-like circulation anomalies. The forecasts systematically underestimate the intensity of pressure anomalies centred over Iceland with the effect of reducing the anomalous westerly/easterly flow over the eastern north Atlantic/western Europe. The flowdependent component of the errors explains about 10% of the total forecast-error variance. However, since it is associated with the NAO mode that dominates the variability of the European weather on longer time-scales, it is essential to identify the model limitations in predicting this flow anomaly in order to guide future work on improving forecasts over Europe.

The relationship between forecast-error variability and NAO fluctuations is effectively described by an SVD analysis performed on fields of forecast-error magnitude and analyses. Regions with large (small) forecast-error anomalies are located over the maximum of the anomalous westerly (easterly) jet, while on the flanks of the jet the error anomaly is expected to be minimum/maximum. Copyright © 2002 Royal Meteorological Society.

Ancillary