Comparative validation of ENVI-met versions 5.6 and 5.9 against high-resolution in-situ air temperature measurements

Abstract

Microclimate simulations are critical for assessing urban heat mitigation strategies, yet they face ongoing challenges regarding computational cost and physical parameterization. This study evaluates the performance of the three-dimensional microclimate model ENVI-met, comparing air temperature prediction of the legacy Version 5.6 against the updated Version 5.9. The updated kernel introduces a revised formulation for surface-to-air sensible heat exchange alongside code optimizations. The validation utilizes a high-resolution air temperature dataset collected around a standalone office building in Aalborg, Denmark, covering a 7-day period with varying meteorological conditions. Detailed statistical analysis, including daytime and nighttime stratification, reveals that both model versions reproduce diurnal temperature cycles with a coefficient of determination (R2) exceeding 0.94. The updated Version 5.9 demonstrated a reduction in the aggregated Root Mean Square Error (RMSE) from 0.89 K to 0.78 K and successfully reduced systemic Mean Bias Error (MBE) across the domain, with the most distinct improvements observed in the immediate boundary layer of sun-exposed façades. Additionally, the simulation runtime for the observed period decreased by a factor of approximately three. These findings indicate that the algorithmic updates improve physical consistency in resolving near-surface temperature gradients while significantly lowering the computational resources required for microclimate modeling.