How accurately did two models simulate radiant temperature in extreme heat?

Details

Researchers validated seasonal mean radiant temperature (TMRT) simulations from RayMan and ENVI-met to assess model performance in a sensitivity analysis spanning cold to extremely hot conditions. Human-biometeorological validation data were collected along transects in Tempe, Arizona, during five field campaigns between 2014 and 2017. Transects crossed seven locations at two- to three-hour intervals from 6:00 to 23:00 local standard time, using a Kestrel meter and thermal camera in 2014–2015 and the mobile instrument platform MaRTy in 2017. Across diverse urban forms, sky view factors and seasons, observations covered a wide range of solar radiation regimes, with TMRT ranging from 8.7 °C to 84.9 °C.

Both models produced large errors across the regimes: root mean square error (RMSE) ranged from 8 °C to 12 °C for RayMan and from 11.2 °C to 16.1 °C for ENVI-met, exceeding the suggested TMRT accuracy of ±5 °C for heat stress studies. RayMan errors were largest in engineered enclosed spaces, complex urban forms and extreme heat conditions. ENVI-met could not resolve spatial variability of TMRT within the modeled domain and had an RMSE as high as 25.5 °C for engineered shade. Both models failed to simulate TMRT accurately in hot conditions. Errors varied by season: TMRT was overestimated in summer and underestimated in winter and the shoulder seasons.

The researchers concluded that the models should not be used under micrometeorological or morphological extremes without on-site validation to quantify errors and assess the direction of bias arising from model limitations.