Minimum Requirements for Predicting the Risks of Extreme Weather Two Weeks to a Season Ahead

Given that predictions beyond two weeks are inherently probabilistic, it is essential for forecasting systems to accurately represent the probability density functions (PDFs) of daily anomalies in order to reliably predict the risks of extreme weather. Given that such PDFs are generally skewed and heavy-tailed, a minimum requirement then is to represent not only the mean and variance (the first and second statistical moments) but also the skewness and kurtosis (third and fourth moments) of the PDFs. But how good are current forecasting systems in this regard? We have addressed this issue by examining the first four statistical moments of daily anomalies in ensemble seasonal forecasts made with the System-3 and System-4 versions of the ECMWF coupled seasonal forecasting system over the 1980 to 2010 period. These forecast moments were compared with corresponding observed moments estimated using the ERA-Interim reanalysis dataset. Maps of the mean, variance, skewness and kurtosis of the upper and lower tropospheric vorticity, and of lower tropospheric winds and temperature, were examined. A remarkable and unexpected conclusion from these intercomparisons was that the “difficult” moments of skewness and kurtosis were actually much better represented in the forecasts than the mean and the variance. In other words, both System 3 and System 4 had less trouble capturing the distinctive non-Gaussian shapes of the observed PDFs than their location (mean) and dispersion (variance). Overall, System 4 had generally smaller mean biases, but larger variance errors, than System 3, making the improvement in representing tail probabilities somewhat ambiguous. A similar tendency for skewness and kurtosis to be accurate but variances to become progressively larger was also noted in other uncoupled AMIP-style integrations of the ECMWF atmospheric model run at progressively higher resolutions from T95 to T2047. These results have important implications for current and future capabilities in predicting extreme weather risks that will be elaborated during the talk.