Evaluation of Ensemble‐Based Extreme Forecast Diagnostics for Zonda Wind Prediction

Abstract

Zonda is a downslope windstorm on the eastern slopes of the Andes in western Argentina. This Foehn‐like phenomenon can produce hazardous, strong, and gusty downslope winds, leading to extreme and rapid warming with severe societal impacts. Despite substantial advances in numerical weather prediction, forecasting Zonda winds remains challenging. This study presents a systematic evaluation of ensemble‐based diagnostics for predicting Zonda winds in Mendoza city. The extreme forecast index (EFI), the shift of tails (SOT), and crossing‐point forecast (CPF), are evaluated using the ECMWF Integrated Forecasting System ensemble (ENS IFS). The ability to discriminate between Zonda and non‐Zonda conditions is assessed for four surface variables (2‐m temperature, maximum temperature, 10‐m wind speed, and wind gusts) at 24‐hr lead‐time intervals up to 5 days. Results show that all diagnostics exhibit statistically significant skill well beyond typical short‐range deterministic forecast horizons. SOT provides the best discrimination for wind‐related variables, consistent with its sensitivity to ensemble tail behavior. For temperature‐based predictors, CPF outperforms EFI and SOT by effectively capturing ensemble‐wide and spatially coherent warm anomalies characteristic of Zonda events. EFI shows its strongest performance at mid‐range lead times, outperforming both SOT and CPF mainly in terms of true positive rates. Events are also stratified by severity, revealing that the ensemble diagnostics respond coherently to variations in both intensity and duration, with longer and stronger events exhibiting sharper and more consistent signals than weaker events. These results highlight the potential of ensemble‐based diagnostics (EFI, SOT, and CPF) to support medium‐range probabilistic assessment of Zonda events.