Modelling Wind Speed Extremes Using Extreme Value Theory: A Case Study for Namibia

Extreme wind speed events pose a significant threat to structural safety and are an important consideration in the design of the growing renewable energy sectors. This study models extreme wind speeds in Namibia by applying Extreme Value Theory (EVT) to quality-controlled daily maximum wind speed records from six meteorological stations, spanning 13–21 years per station and covering the 2003–2024 period. A two-rule quality-control procedure, combining a regional plausibility limit with an isolated-spike test supported by cross-station coherence checks, identified and removed 70 spurious automatic weather station records (0.24% of observations) prior to analysis. The Generalized Pareto Distribution (GDP) was fitted to declustered threshold exceedances using the Peaks-Over-Threshold method, with 95% confidence intervals for return levels obtained by profile likelihood. Model comparison based on AIC, BIC, and the negative log-likelihood indicated that the Generalized Pareto Distribution (GPD) provided a better description of the extreme wind speed tails. The results reveal a clear coastal-inland contrast; the coastal station Lüderitz experiences the strongest and most frequent extreme wind events, with a 100-year return level of 49.4 m/s (GPD; 95% profile-likelihood interval 44.9–70.3 m/s) and evidence of a bounded upper tail, while inland stations exhibit more moderate extremes with 100-year return levels 36–45 m/s. A seasonal analysis reveals stronger winter extremes at coastal Walvis Bay, while inland stations experience summer convective peaks. The study provides the first systematic station-level EVT analysis of observed wind extremes for Namibia, offering essential quantitative input for wind-sensitive infrastructure design, renewable energy project siting, and national climate adaptation planning.