Predicting climate indicators at the decadal scale using a hybrid prediction system: application to SUEZ water management plans over France

As climate change reshapes hydrological cycles, workers in water management face unprecedented challenges in ensuring resource availability, mitigating flood risks, and maintaining resilient infrastructure. Nowadays, water utilities and authorities rely on long-term climate projections to plan for challenges extending through the end of the century. However, critical gaps persist in actionable information for shorter timescales, such as the decadal scale, which better aligns with political and operational decision-making. In this context, decadal climate predictions can be pivotal to address the needs of the water management sector and develop efficient climate services. However, their added values as compared to projections remained limited up to now.
To better understand user requirements, we collaborate with various teams from SUEZ, a company specializing in water management. Through interviews, we have identified the demand for specific indicators based on climate variables (e.g., precipitation, temperature) and corresponding spatio-temporal scales. Building on this understanding, we also develop in IPSL-EPOC decadal prediction team a new hybrid approach to improve our forecasts. This approach includes identifying a climate index (e.g., NAO, WEPA) derived from Sea Level Pressure (SLP) that correlates with the climate variable of interest. Using all the available decadal climate predictions from the DCPP project, we evaluate the predictability of this index, which is usually high for NAO and WEPA. This index is then employed to subsample a few of member CMIP6 climate projections that are in phase with the prediction of the DCPP ensemble. This latter step allows to inflate the amplitude of the predictable signal, resolving the limitation coming from the signal-to-noise paradox. It is also allowing to perform a proper statistical downscaling, used to refine these forecasts, ensuring their usability for identified needs. The resulting forecasts are designed to integrate seamlessly into SUEZ’s water sector models.
Preliminary work has identified diverse parameters of interest for water management, such as daily precipitation (resource availability forecasting), extreme precipitation events at fine temporal resolution (Combined Sewer Overflows modeling), and the number of very cold or very hot days (linked to risks of water mains and service lines failures, respectively). Early findings also suggest that, for the average precipitation over France, the WEPA index exhibits the largest correlations, unlike the NAO, which has greater influence for other European regions. The production of forecasts is currently underway, and their performance regarding the initially identified parameters will be presented.