Closing prediction gaps in urban canopy models through handling parameter and model-form uncertainties

초록

Urban climate studies often rely on physics-based modeling tools, such as urban canopy models (UCMs), to investigate meteorological conditions across different urban areas. Despite continued advancements, application of UCM at city scale remains challenging due to two major limitations: (1) parameter uncertainties to define model inputs, and (2) model-form uncertainties introduced by empirical parameterizations used to represent aerodynamic processes. Existing model calibration studies for UCM remain limited and often consider few observation sites and parameters, particularly in the context of wind speed prediction. This study addresses both sources of uncertainties by calibrating the vertical city weather generator (VCWG) against wind speed measurements from 72 monitoring stations distributed across Seoul, Republic of Korea. Furthermore, unlike previous efforts that rely on CFD-based parameterizations, this work directly re-parameterize key aerodynamic processes—such as turbulence length scales (lk/Ck), building drag coefficients (CDB), and pressure-gradient (Cdpdx) terms—using extensive set of actual observations. Results show that explicitly addressing model-form uncertainties improve prediction accuracy beyond calibrating parameter uncertainty alone, with error reductions of 10.68% below building height (corresponding to 0.081 m/s) and 18.70% at denser urban areas (λp > 0.3, corresponding to 0.208 m/s) relative to baseline VCWG. These improvements are associated with higher lk/Ck, lower CDB, and weaker Cdpdx near the ground relative to the default VCWG parameterizations. Robustness of the calibrated model is demonstrated through validation on independent time periods. These findings from the Seoul case study highlight the important role of model-form uncertainties in VCWG to improve city-scale urban wind predictions. © 2026 Elsevier Ltd

키워드