Application of the teaching–learning-based optimization algorithm to an analytical model of thunderstorm outflows to analyze the variability of the downburst kinematic and geometric parameters

Abstract

Abstract. Downbursts winds, characterized by strong, localized downdrafts and subsequent horizontal straight-line winds, present a significant risk to civil structures. The transient nature and limited spatial extent present measurement challenges, necessitating analytical models for an accurate understanding and predicting their action on structures. This study analyzes the Sânnicolau Mare downburst event in Romania, on 25 June 2021, using a bi-dimensional analytical model coupled with the teaching–learning-based optimization (TLBO) algorithm. The intent is to understand the distinct solutions generated by the optimization algorithm and assess their physical validity. Supporting this examination are a damage survey and wind speed data recorded during the downburst event. Employed techniques include agglomerative hierarchical K-means clustering (AHK-MC) and principal component analysis (PCA) to categorize and interpret the solutions. Three main clusters emerge, each displaying different storm characteristics. Comparing the simulated maximum velocity with hail damage trajectories indicates that the optimal solution offers the best overlap, affirming its effectiveness in reconstructing downburst wind fields. However, these findings are specific to the Sânnicolau Mare event, underlining the need for a similar examination of multiple downburst events for broader validity.