Assessing atmospheric influences for improving time-varying data-driven decadal predictions of Mediterranean spring discharge

Decadal climate forecasting for spring discharge is challenging due to its complex nature. Data-driven models (DDMs) are a promising approach. One such model, the time-varying Periodic AutoRegressive Moving Average with a Conditional Standard Deviation (PARMAX(TVAR)-CSD), applied to the Caposele spring (southern Italy) as a Mediterranean representative, exemplifies this approach by incorporating exogenous variables. Leveraging past data spanning 123 years (1899–2021) and encompassing both large-scale (Arctic Oscillation) and small-scale (precipitation) climate forcings, this model offers annual forecasts extending to 2060. Notably, a discernible interdecadal pattern emerged, projecting an increase in discharge during the decade 2031–2040, followed by a temporary drought phase from 2041 to 2050, consistent with global model projections. These findings highlight the influence of climate change and variability on precipitation patterns in southern Europe, with implications for global water resource management strategies, and emphasize the need for novel approaches to decadal forecasting of spring discharge in the Mediterranean.