POTENTIAL ECONOMIC VALUE OF ENSEMBLE-BASED SURFACE WEATHER FORECASTS

The possible economic value of the quantification of uncertainty in fu ture ensemble-based surface weather forecasts is investigated using a formal, idealized decision model. Current, or baseline, weather foreca sts are represented by probabilistic forecasts of moderate accuracy, a s measured by the ranked probability score. Hypothetical ensemble-base d forecasts are constructed by supplementing the baseline set of proba bilistic forecasts with lower- and higher-skill forecasts. These are c hosen in such a way that mixtures of the forecasts including the lower - and higher-skill subsets with equal frequency exhibit the same accur acy overall as the moderately accurate (conventional, baseline) foreca sts. For both simple one-time decisions (static situation) and related sequences of decisions (dynamic situation), these hypothetical ensemb le-based forecasts are found to lead to greater economic value in the idealized decision problem when protective actions are relatively inex pensive, corresponding to real-world problems. However, for some decis ion problems considered, the ensemble-based forecasts are slightly les s valuable than the baseline forecasts. This result derives at least i n part from the (probably unrealistic) assumption that the ensemble-ba sed forecasts are no more skillful in aggregate than their conventiona l counterparts, but implies that positive economic value for ensemble forecasts with respect to this baseline will not be automatic. Rather, for ensemble-based forecasts to be at least as valuable for all decis ion problems, they will need to exhibit sufficiently higher skill in a ggregate than the conventional forecasts that could have been produced in their place.