Air temperature versus thermal indices (UTCI and PET) across Türkiye: reanalysis evidence that thermal comfort and heat stress cannot be inferred from air temperature alone.

Air temperature (Ta) remains the most widely communicated measure of heat, yet human heat stress also depends on air humidity, wind conditions and radiation fluxes. We quantified the difference between Ta and two human-biometeorological indices, the Universal Thermal Climate Index (UTCI) and the Physiologically Equivalent Temperature (PET), across Türkiye using hourly May-September reanalysis data for 1950-2024. ERA5-Land supplied Ta, humidity, wind and pressure at ~ 0.1° resolution, while mean radiant temperature (MRT) was taken from the Copernicus thermal-comfort dataset derived from ERA5 (ERA5-HEAT) and interpolated to the ERA5-Land grid. PET was computed with a steady-state energy-balance model. We analysed long-term mean ΔT = index - Ta, daytime-only ΔT, representative diurnal cycles, regional distributions, non-parametric trends and meteorological drivers. The national 24-h long-term mean ΔTUTCI was close to zero (+ 0.2 °C), but daytime conditions were much warmer than Ta: over 09:00-15:00 local time, mean ΔTUTCI was + 4.6 °C nationally and reached + 9.5 °C locally. PET had a lower 24-h seasonal mean than Ta (-2.4 °C nationally), yet daytime ΔTPET was also positive (+ 3.1 °C nationally). MRT was the dominant correlate of both ΔT indices in every region (Spearman r ≈ 0.80-0.94). Satellite-era Sen-slope estimates for 1979-2024 indicated weak but spatially coherent ΔT increases, especially for PET (+ 0.07 °C decade⁻¹). FDR-significant fractions were 24.2% for UTCI and 73.7% for PET with the standard Mann-Kendall test and 31.6% and 75.3%, respectively, after the Hamed-Rao autocorrelation correction. Overall, Ta gives an incomplete and time-dependent picture of outdoor thermal stress in Türkiye, supporting heat-warning and climate-services approaches based on physiologically meaningful indices.