Empirical Correction of a Statistical Model for Predicting Low Wind Speed Percentiles in Urban Environments

Accurate prediction of wind speed percentiles is essential for diverse applications, including wind energy assessment, infrastructure resilience to extreme winds, and the reliable operation of wind turbines. Among these, the 25th percentile—commonly referred to as the first quartile—plays a key role in characterizing low-wind conditions. A previous statistical model, originally developed for wind speed probability distributions, showed poor performance in predicting the 25th percentile when applied to urban environments, with significant underestimation and scatter relative to reference data. In this study, an optimization of that statistical model is undertaken to improve its predictive capability specifically for the 25th percentile. Large Eddy Simulation (LES) data from a wind tunnel experiment representing a typical urban morphology are used as a reference. The original model’s predictions are first evaluated against the LES-derived 25th percentiles, confirming persistent underestimation (fractional bias FB = 0.25). To address this, correlations between the 25th percentiles and key statistical descriptors—namely mean, standard deviation, skewness, kurtosis, and maximum wind speed—are systematically examined. Linear and power-law fits are developed, revealing strong linear relationships with the mean and skewness, and a robust power-law relationship with the maximum wind speed. No meaningful correlations are found with standard deviation or kurtosis. The proposed corrections offer a simple yet effective physical scaling framework to mitigate internal model bias for the 25th wind speed percentile within the investigated urban setup. While further out-of-sample validation across varied topologies is required, the modified approach demonstrates promising potential for future integration into urban wind resource assessment, microclimate planning, and environmental risk tools.