Transforming the Temperature Forecasts into Mortality Predictions
We proved that the predictability of the temperature forecasts can be transformed into skilful Temperature-Attributable Mortality (TAM) predictions. Figure 1 shows how the predictability skill decays as a function of the lead time of the temperature forecasts (left) and TAM predictions (right). We see that the average skill for the ensemble of regions is similar for temperature and TAM, but the spread between regions is higher in TAM. This is associated with the epidemiological models that we used to transform the temperatures into TAMs, which preserve the average skill but widen the range of region-to-region differences.
How Will the Operational System Look Like?
Once operational, the TAM system will generate daily predictions of temperature-attributable deaths at the same lead times than the driving weather or subseasonal-to-seasonal climate forecasting scheme. Figure 2 is a good example of how it will look like, it shows the temperature forecast (left) and the Temperature-Attributable Mortality (TAM) prediction (right) for several lead times (1, 7, 15 days) before the 2012 European cold wave in January 27 - February 17, 2012. The observed cold temperature anomaly could be predicted one week in advance, but not beyond. Similarly, the same pattern is found for TAM, with also skilful predictions one week ahead of the event, but not at longer lead times.