SECLI-FIRM has developed and refined the approach taken by the nine case study, by identifying and addressing users’ requirements. With the ultimate aim to assess the value of including seasonal forecasts in the decision making of the nine case studies, to augment or replace climatological information, appropriate evaluation methods were designed for each case study.
A key component of the refinement of the case studies was the definition and use of decision trees as a visualisation tool and for engagement with the case-study teams. These have been used at different levels:
• To provide context for the systematic and consistent visualisation of key decisions, especially climate-driven ones, by case-study leads.
• To identify the points/nodes in the decision trees where SECLI-FIRM partners can provide and use improved climate information and where the value of adopting this information can be assessed.
• To assess how best to embed the probabilistic format of seasonal forecasts in current decision-making processes.
SECLI-FIRM developed a purpose-built comprehensive dataset with more than ten independent seasonal forecasts systems on a platform which allowed partners to have a common and unique workplace where they could closely interact. This dataset greatly assisted with the scientific developments in SECLI-FIRM.
Scientific research in SECLI-FIRM explored a variety of approaches for extracting the most signal from the forecast for each case study. Approaches range from tailored downscaling, to the use of weather regimes, to the exploitation of large-scale climatic drivers (such as the North Atlantic Oscillation) to the tuning of multi-model combinations, including by means of machine learning models. Aside from having extensively documented these results in public project reports, several of these have been, or are being, published in journals such as the MDPI Climate Special Issue "Seasonal Forecasting Climate Services for the Energy Industry".
Some of the conclusions from this work are:
• The optimal selection of models is different depending on the region/phenomenon. However, by using independence information it is possible to reduce the number of models and data to produce the optimized forecasts.
• Downscaling is one of the most requested features of seasonal forecasts. To get local forecasts is best to downscale the large scale of the dynamical model rather the local features. Clearly, availability of observations is key to optimising local forecasts
• More advanced post-processing methods (e.g. using non-linear statistical models) generally require more data / statistics than the historical records typically available
• "Classic" daily weather types aren't very useful for seasonal forecasts (due to a lack of skill - they often work in "perfect forecast" conditions). However, a more tailored approach was shown to be useful
• The average skill of weather regimes forecast is not good enough to improve the downscaling in general. But in case of good predictability (window of opportunity), it improves its quality for wind and precipitation in mountainous areas
The co-development of the individual Case Studies into trial climate services is another key learning of SECLI-FIRM towards the development of the services, including early conversations through to the presentation of the forecasts in forms useful to the users, to the write up of deliverables. The use of diverse approaches to communicating the seasonal forecasts, from the sharing of plots with a commentary to online visualisation tools (e.g. Teal,
https://tealtool.earth(s’ouvre dans une nouvelle fenêtre)) is a strength of the SECLI-FIRM trial climate services.
SECLI-FIRM has shown how tailoring seasonal forecasts to specific case studies can enhance its predictive power. Crucially, the interaction, or co-production, at the core of SECLI-FIRM activities, have given users sufficient confidence in the benefit of using seasonal forecasts, particularly as a long-term strategy, as in the case of ENEL.
