Abstract
<jats:p>Tornadoes and waterspouts in the western Mediterranean and Iberian region are low-frequency but potentially damaging hazards, whose spatial occurrence remains difficult to characterize using fixed geographical or climatological regions. This limitation is especially relevant for risk-sensitive applications, where local exceedance probabilities and return-period estimates require physically meaningful regional occurrence rates.This work presents an objective framework for the environmental regionalization and probabilistic hazard assessment of tornado and waterspout occurrence over peninsular Spain and the Balearic Islands. A unified and quality-controlled event catalogue was built by merging national (SINOBAS/AEMET) and European (ESWD/ESSL) severe weather records, yielding 1185 tornado and waterspout events for the period 1992–2025 after duplicate removal. For each event, pre-event atmospheric environments were characterized using ERA5-derived thermodynamic, kinematic, and composite instability parameters extracted around the event location and time.To move beyond fixed a priori regions, event environments were spatially aggregated over a hexagonal grid of approximately 50 km and summarized through robust environmental signatures. Dimensionality reduction via principal component analysis, followed by Ward hierarchical clustering with spatial connectivity, identified three objective environmental regions. These clusters provide a physically based spatial partition of tornado-favourable environments and are evaluated against a priori subregional divisions using supervised machine learning classifiers.Building on this regionalization, the study develops an occurrence, point-scale exceedance, and 50-km neighbourhood return-period framework for tornadoes. Regional tornado rates are estimated by intensity class and combined with Monte Carlo path-geometry simulations and internal wind-speed exceedance fractions. These were used to derive exceedance curves EP(v,T) and threshold-based hazard maps consistent with the observed spatial heterogeneity of tornadic environments.Results show that the data-driven regions capture coherent environmental regimes — shear-dominated in the west and southwest, thermodynamically charged in the central-southeast, and weakly forced in the north and east — that are not fully reproduced by fixed geographical subdivisions. A key outcome is a decoupling between occurrence and severity: the regimes that produce most tornadoes are not those that dominate the high-end wind hazard, since the least active, thermodynamically charged regime concentrates the most intense events and controls exceedance at the highest thresholds. The proposed framework provides a reproducible basis for tornado hazard mapping in Spain and the Balearic Islands, with potential applications to civil protection, territorial planning, and the assessment of risk-sensitive infrastructures.</jats:p>