Earth and Planetary Science

Extreme heat is one of the most consequential manifestations of anthropogenic climate change, yet quantitative assessments combining trend detection, frequency analysis, and probabilistic risk modelling remain rare for tropical West African cities. This study analyses annual maximum temperature (Tmax) variability, extreme heat frequency, and probabilistic return-level estimation for South-Western Nigeria over 1983–2023, using TerraClimate monthly data validated against ERA5-Land reanalysis. Regional annual Tmax ranged from 33.20 ℃ (1991) to 35.82 ℃ (1990), with a modest but statistically significant upward trend confirmed by the Mann–Kendall test (Tau τ = 0.329, p = 0.0025) and quantified at 0.0256 ℃ per year by Sen's Slope Estimator. An annual-scale exceedance index shows extreme years recurring intermittently, with the clearest sustained run only from 2020–2023; a standard daily ETCCDI TX90p index computed at the state level shows a fairly uniform 18–20% exceedance rate across all six states relative to the 1983–2012 baseline. Extreme Value Theory models fitted to the annual Tmax series show the Generalized Extreme Value (GEV) distribution outperforming Gumbel on goodness-of-fit (Kolmogorov–Smirnov (K–S) statistic: 0.0997 vs. 0.1486), with 100-year return levels of 35.98 ℃ and 36.65 ℃, respectively, a difference of 0.67 ℃. Spatial analysis reveals a more nuanced risk profile than a simple coastal-inland split: the interior state of Oyo records the lowest background warming trend, while Ekiti and Lagos record the highest daily extreme-heat frequency. The study advances a replicable, open-data framework for thermal hazard quantification in data-scarce tropical environments, relevant to Nigeria's National Adaptation Plan, infrastructure standards, and the Sendai Framework for Disaster Risk Reduction.