Provide methodology to calculate CIIs and their skill
The aim of work package 22 is to assess the value of seasonal forecasts for a collection of user-targeted climate information indices (CIIs). As such, CIIs represent a relatively easy way to produce alternatives to more sophisticated impact models, but may still provide information of more relevance to the user than forecasts of meteorological variables usually produced by seasonal forecasting centres. Forecasts of CIIs evaluated in this work package are used to complement the prototypes and case studies developed as part of the EUPORIAS project.
Producing and assessing long-range forecasts of CIIs can be tricky: CIIs are often defined with respect to absolute thresholds (e.g. frost days are days with minimum temperatures below 0° C) and therefore systematic model errors have to be adjusted before CIIs can be computed. This calibration of daily time series is resource-intensive as such forecast datasets often amount to several Gb in size. An alternative are climate indices that are defined with respect to the local climatology. Such a definition makes these indices less prone to model errors and calibration of daily time series is generally not necessary. In return, such relative indices are not always linking to direct user relevance. While forecast calibration is important, we find that forecast skill is generally independent of the specific calibration method used.
Unfortunately, forecast skill of seasonal CII forecasts in Europe is fairly limited. CII forecasts tend to be more skilful in southern Europe in summer, whereas forecast skill in winter is generally low. In other areas of the world such as the tropics or the western US, seasonal forecasts are more skilful. Also forecasts of CIIs are found to be at most, as skilful as the forecasts of the meteorological variables these CIIs are derived from. Even if skill in forecasts of CIIs is not enhanced compared with the underlying meteorological variables, CII forecasts are potentially more relevant and useful to the users as these allow framing the seasonal forecast in a more application-specific way. As such, CIIs are of more value to users than forecasts of the underlying quantity.
The limited skill of CII forecasts of in Europe is a general limitation for the utility of long-range forecasts. Forecast skill, however, varies by region, time of year, variable, forecast lead time, spatio-temporal aggregation etc. This strong variability of forecast skill offers the possibility to identify periods and regions where skilful forecasts can be made. Further research will be directed towards identifying and understanding such windows and pockets of opportunity.