Abstract
<jats:p>Seasonal forecasts represent a promising tool to support water resources management in the Mediterranean region, where water availability is increasingly affected by climate change and the occurrence of extreme events. Knowing in advance the availability of water resources allows for better management of water’s multiple uses, including agriculture and renewable hydropower production.This study presents an assessment of ECMWF SEAS5 seasonal forecasts (51 members, 0.4° grid spacing, 6h time resolution, seasonally aggregated, with lead times from 1 to 7 months) over Italy, with a focus on their potential applicability for the prediction of water resource availability and, ultimately, hydropower production.The forecasts cover the period 2018-2024, while a hindcast period 1994-2016 is considered for bias correction purposes and as baseline reference. The forecast performance has been evaluated for 2 m air temperature and precipitation using a high-resolution observational reference dataset from Ispra (IspraObs). This dataset, covering the period 1951-2024 with a spatial resolution of 1km, has been upscaled to 0.4° of resolution, using a conservative remapping algorithm. The performance metrics used include the Anomaly Correlation Coefficient (ACC), correlating forecast and observed anomalies with respect to the climatology of the observational dataset, the RMSE Skill Score (RMSESS), and the Continuous Ranked Probability Skill Score (CRPSS), which compares the cumulative distribution function of forecasts with the one of the reference climatology.Simple Bias correction approaches, i.e. additive mean adjustment for temperature and multiplicative mean ratio for precipitation, calibrated using hindcast data, have proven to increase the performance of the forecast, particularly for temperature and, to a lesser extent, for precipitation.Results indicate that temperature forecasts show moderate predictive skill over the country, particularly in summer, with ACC average values above 0.3 at a 4-month lead time, suggesting a reasonable ability to reproduce the observed climatic interannual variability. In contrast, precipitation forecasts show lower and spatially heterogeneous skill, reflecting the intrinsic difficulty of predicting precipitation in the Mediterranean region. In this case, the seasonal forecast skill seems no better than climatology for lead times greater than 1 month, with quite poor performances for lead times greater than 4 months. A more specific verification has been conducted over the river basin of Tevere, Central Italy, which is of particular interest for the energy sector in terms of hydropower production. The results show good performances at 3-month lead time in summer: ACC values are in the range 0.6-1, while RMSESS is around 0.4.Building on these results, future work will focus on integrating the bias-corrected seasonal forecasts into a hydrological modeling chain, to provide estimates of water resource availability for Tevere river basin, at a monthly time scale. This enables the assessment of forecast-driven streamflow predictions and their potential use for optimizing hydropower production in Italy. Such an approach is highly relevant in the context of climate change, as improved seasonal predictability could enhance adaptation strategies for water resource management and energy production in the Mediterranean region.</jats:p>