Multi-timescale hydro-meteorological forecasts for the optimal control of the multipurpose Lake Como

Hydrometeorological forecast products have seen an increase in their availability, accuracy and reliability, and they are expected to be an essential tool to support adaptive and robust control of water systems under changing hydroclimatic conditions. In this study, we investigate how the most valuable information (lead time and forecasted variable) from multi-scale forecasts can be selected and used to inform the optimal control of multipurpose water reservoirs. The framework is composed of Input Variable Selection algorithms, supporting the extraction of the most informative policy inputs, from a set of different hydrological forecasts ranging from short to long term, coupled with the Evolutionary Multi-Objective Direct Policy Search method, for designing Pareto optimal control policies conditioned on forecast information. This approach is tested on the Lake Como system, a regulated lake in Northern Italy which is controlled for preventing floods along the lake shores, as well as for providing irrigation supply to downstream users and avoiding low lake levels. We expect to extract the best subset or combinations of lead times from a suite of forecast products, including short-term local deterministic forecasts as well as sub-seasonal and seasonal large-scale ensemble forecasts provided by the European Flood Awareness System (EFAS), part of the Copernicus Emergency Management Service. The candidate variables proposed as inputs for the IVS include different statistics extracted from these forecasts, over a variety of temporal scales and spatial domains. The performance of the designed forecast-informed control policies is contrasted against various benchmarks, including controllers relying on perfect forecasts as well as baseline control policies not informed by any forecast. Beside improving the controller performance, results are expected to provide insights on how to address the intrinsic bias of forecast products and to highlight the role of forecast uncertainty in optimal control design.