Multiscale modelling approaches for estimating wind distributions at pedestrian level in real urban environments

Abstract

Accurately estimating wind distributions at pedestrian level in real urban environments is crucial for predicting pollutant dispersion within street canyons, assessing public exposure levels, and evaluating pedestrian comfort. Field measurements can be conducted to capture pedestrian-level wind conditions. However, experimental methods provide data only at limited sampling points, restricting the ability to obtain comprehensive spatial information of urban wind distributions. Therefore, this study aimed to apply multiscale modelling by combining the Weather Research and Forecasting (WRF) model with computational fluid dynamics (CFD) methods to accurately calculate detailed wind distributions at the pedestrian level in real urban environments. To calculate pedestrian-level wind distributions in real urban environments using the multiscale modelling approaches, it is crucial to investigate the applicability of various turbulence models used in CFD for outdoor wind simulations. In this investigation, the performance of steady-state and unsteady-state Reynolds-averaged Navier-Stokes simulation (SRANS/URANS) and large-eddy simulation (LES) were evaluated in calculating airflow and pollutant dispersion in street canyons with generic and real urban layouts. For each layout, wind tunnel experiments with measured wind speed and pollutant concentration were available as benchmarks. In addition, instantaneous concentration fields were analyzed to assess the transient models. The results showed that in the generic urban layout, URANS with the SST k – ω model captured the large-scale fluctuations, while instantaneous results from URANS with the SST model did not change over time in the real urban layout. In both generic and real urban layouts, the RNG k - ε model and SST k - ω model provided similar results for time-averaged wind speed and concentration distributions in SRANS and URANS simulations. Among all the selected RANS models, SRANS/URANS with the SST k - ω model showed best agreement with measured data in calculating wind speed. LES performed best in calculating wind speed and pollutant dispersion, but it was the most time-consuming model. Apart from the turbulence modelling approaches, setting appropriate inflow wind profiles for CFD simulations is also important for predicting the wind distributions in real urban environments. Wind profiles within the atmospheric boundary layer are significantly affected by local atmosphere circulation and diurnal variation. The WRF model is a powerful mesoscale weather prediction model that can be used to provide realistic inflow boundary conditions for CFD simulations. To investigate the accuracy and applicability of a combined WRF and CityFFD method (WRF-CityFFD) for calculating urban wind distributions, this study first validated the WRF and CFD models and then used the validated models in WRF-CityFFD to calculate the wind distributions in the Kowloon district of Hong Kong within an area of 3.5 km × 2.4 km. The wind speed data at two weather stations were used as a benchmark. To evaluate the performance WRF-CityFFD, a comparison with CityFFD using inflow boundary conditions derived from a commonly used semi-empirical method (semi-empirical-CityFFD) was conducted. In this method, power-law wind profiles were used as the inflow wind profiles for CityFFD simulations. The results showed that, at KP and HKO stations, WRF-CityFFD achieved lower RMSEs (1.31 and 1.26 m/s) compared to 2.24 and 1.50 m/s for the semi-empirical approach. Thus, WRF-CityFFD performed better than semi-empirical-CityFFD in in calculating wind speed in urban microclimates. Moreover, WRF provided more accurate wind profiles in coastal areas with onshore winds, indicating such locations are more suitable for defining inflow boundaries in the combined WRF-CityFFD method. This combined method can help improve the model's ability to reproduce urban wind patterns, which is essential for applications such as urban ventilation assessment, pollutant dispersion modelling, and the evaluation of outdoor comfort. Although the potential of the combined WRF and CFD method for urban wind simulations has been preliminarily demonstrated, previous investigations indicate that mesoscale models such as WRF simulations are not sufficiently precise to predict wind profiles in built-up areas, and existing improvement methods remain computationally expensive and time-consuming. To improve the accuracy and efficiency of estimating wind profiles in built-up areas using WRF simulations, this study proposed a method that combines WRF with a porosity model. WRF provides the wind profile at the urban edge, and the porosity model calculates the airflow pressure drop across the selected urban area using a parametrized urban layout. The urban wind profile is then analytically determined with the momentum integral method. The performance of the proposed method was first evaluated in three generic urban layouts, with validated CFD simulations used as benchmarks. The proposed method was then applied in a real urban layout to demonstrate its performance, and the Kowloon district of Hong Kong, with an area of 2,350 m × 643 m, was selected as the target area. The wind profile measured with a radiosonde in the same region was used as a benchmark, and the WRF-calculated wind profile in the built-up area was also evaluated for comparison. The results showed that the method accurately estimated wind profiles in the generic urban layouts. In the real urban layout, the proposed method estimated the urban wind profile reasonably well in the densely built-up area with complex building configurations and performed better than WRF. Based on the above investigations, a multiscale modelling approach was proposed by this study. To further assess the performance of the proposed approach in predicting pedestrian-level wind conditions, wind profiles estimated by the combined WRF and porosity model were used as inflow boundary conditions for CFD simulations. This implementation is hereafter referred to as the analytical-inlet-CFD method. To assess the accuracy of this multiscale approach, a public housing estate covering an area of 578 m × 560 m was selected as the target area, and field measurements were conducted to collect pedestrian-level wind data in this real urban environment. The measured data then served as the benchmark for evaluating the accuracy of the analytical-inlet-CFD for outdoor wind simulations. Additionally, results from CFD with inflow boundary conditions directly extracted from WRF outputs (WRF-inlet-CFD) were also analyzed for comparison. The results showed that analytical-inlet-CFD performed better than WRF-inlet-CFD in calculating pedestrian-level wind distributions in real urban environments. Overall, this thesis systematically evaluated the combined WRF–CFD method for calculating wind distributions in real urban environments and identified its limitations in accurately predicting wind profiles within densely built-up areas. To address these issues, an analytical method combined WRF and a porosity model was proposed to improve the estimation of inflow wind profiles in urban areas. The resulting multiscale modelling approach incorporates mesoscale WRF simulations, analytical wind profile estimation, and CFD modelling. To assess the performance of the proposed approach, field measurements were conducted in a public housing estate to collect pedestrian-level wind data, which served as benchmark for evaluation of simulations. The results showed that the proposed approach improved both accuracy and computational efficiency compared to conventional methods, providing a reliable tool for calculating urban wind distributions in real urban environment, with further implications for pollutant dispersion modelling and pedestrian thermal comfort assessment.