Lessons learned from the implementation of the near real-time S2S4E Decision Support Tool

Abstract

<p>Under the context of the H2020 S2S4E project, industrial and research partners co-developed a fully-operational Decision Support Tool (DST) providing during 18 months near real-time subseasonal and seasonal&#160; forecasts tailored to the specific needs of the renewable energy sector. The tool aimed to breach the last mile gap between climate information and the end-user by paying attention to the interaction with agents from the sector, already used to work with weather information, and willing to extend their forecasting horizon by incorporating climate predictions into their daily operations.</p><p>With this purpose, the tool gathered a heterogeneous dataset of seven different essential climate variables and nine energy indicators, providing for each of them bias-adjusted probabilistic information paired with a reference skill metric. To achieve this, data from state-of-the-art prediction systems and reanalysis needed to be downloaded and post-processed, fulfilling a set of quality requirements that ensure the proper functioning of the operational service. During the design, implementation, and testing phases, a wide range of scientific and technical choices had to be made, making clear the difficulties of transferring scientific research to a user-oriented real-time service. A brief showcase will be presented, exemplifying the different tools, methodologies, and best practices applied to the data workflow, together with a case study performed in Oracle&#8217;s cloud infrastructure. We expect that by making a clear description of the process and the problems encountered, we will provide a valuable experience for both, upcoming attempts of similar implementations, and the organizations providing data from climate models and reanalysis.</p>