Abstract
<jats:p>Heatwaves (HWs) are among the most damaging climate extremes affecting the Mediterranean basin, where they could drive excess mortality, agricultural losses, water stress and wildfire risk. The Mediterranean region is warming fast, and the frequency, duration and intensity of HWs are projected to keep rising. Anticipating these events requires understanding their drivers. The complex interaction between large-scale atmospheric circulation, remote teleconnections and local land-surface conditions is difficult to capture with conventional statistical or dynamical approaches, and these drivers may vary markedly from one Mediterranean sub-region to another.This work proposes the application of a general driver identification framework, Spatio-Temporal Cluster-Optimized Feature Selection (STCO-FS), to identify the key short-term and seasonal drivers of HWs across the Mediterranean basin. The method combines clustering algorithms for reducing the spatial dimensionality with an ensemble evolutionary optimization algorithm to perform driver selection jointly in the spatial and temporal domains. In a first phase, gridded predictor fields from the ERA5 reanalysis, such as mean sea level pressure, geopotential height at 500 hPa, sea surface temperature, soil moisture, total precipitation and 2 m temperature, are reduced in dimensionality by grouping grid points with similar temporal behaviour into clusters. Climate variability indices (e.g. NAO, ENSO, IOD) and local variables are added directly. In a second phase, a wrapper feature selection approach based on a multi-method evolutionary algorithm (PCRO-SL) selects the most skilful drivers and identifies, for each one, the optimal time lag and time window, distinguishing short-term precursors (days) from sub-seasonal and seasonal influences (up to several months) of HW occurrence.The framework will be evaluated on representative areas of the Mediterranean. We expect that this approach will allow us to unravel the relative contribution of the different variables, and to characterise how these contributions differ across sub-regions of the basin. By revealing the spatio-temporal structure of HW drivers, this framework aims to improve the physical understanding and sub-seasonal predictability of Mediterranean HWs, supporting more effective early warning and climate adaptation strategies. </jats:p>