The establishment of a 3D geometric model is based on the actual terrain and building height data for the simulated area obtained through drone aerial photography technology. Basic data such as the geometric shape of urban buildings, roof height, and vector data of the top of buildings with high resolution, high timeliness, and accuracy are used to construct a realistic 3D geometric model for driving the STAR-CCM+ simulation. In the process of model construction, the same shape as the actual building was maintained to reduce the influence of model errors on the calculation results (Fig. 3a). The length, width, and height of the STAR-CCM+ calculation domain are 13, 11, and 2 km, respectively, among which nearly two-thirds of the buildings are distributed in the range of 10–40 m, with the average height of the buildings being 32 m. The highest building in the area is 390 m, and the lowest building is 6 m.

The geometric model domain is divided by polyhedral meshes (Fig. 3c). The polyhedral mesh has much fewer cells than the traditional tetrahedral mesh, but with a similar accuracy of calculation. Under the same number of grid cells, the numerical simulation results of polyhedral grid cells are more consistent with experimental data than tetrahedral grid cells (Zhang et al., 2020). The grid cells on the ground and near the buildings are much denser (Fig. 3b) (the minimum resolution is about 1 m), so that the influence of the building on the flow patterns can be described more accurately. In the end, the number of unit grid cells generated is 382 181, and the number of nodes is 1 990 224.

STAR-CCM+ solves the RANS equations with the realizable k-ε turbulence closure scheme in this study (Lei et al., 2004; L. Li et al., 2006; S. Li et al., 2019). The ground and building surfaces are set to be no-slip, and the distribution of fluid velocity and pressure near the ground and the building surface is described by the blended wall function. For the coupling of the WRF model to STAR-CCM+, the values of temperature and wind from the WRF simulation are extracted to establish the initial and boundary conditions for STAR-CCM+. Since the variables obtained by the WRF simulation have a relatively coarse resolution of 1 km, the velocity components (U, V, and W) and the temperature are interpolated to the boundary of the STAR-CCM+ domain using the spline interpolation method and the linear interpolation method, respectively. For the turbulence intensity and turbulence viscosity ratio, the lateral and upper boundaries are set as constants with values of 0.1 and 10, respectively.

Figure 5a and b illustrate the distribution of the average wind streamlines in January at the height of 5 and 40 m, respectively. At the height of 5 m, the wind field structure is more complicated (Fig. 5a) than that at 40 m (Fig. 5b). The wind speed is relatively more intense in the areas where the buildings are sparse and smaller. In addition, the flow fields diverge or converge due to the layout of buildings and streets, causing the wind direction inside blocks to differ from the background wind direction greatly. As the density of buildings gradually decreases with the increases in height, this phenomenon diminishes, reflected by the relatively more consistent wind fields at 40 m (Fig. 5b). The phenomenon was also found in a previous study (Sui et al., 2016).

To clearly show the details of the wind field, sub-domain A (Fig. 3a) with complex building structures is selected from the entire computational domain. The near-surface winds disperse or converge horizontally and rise or subside vertically with the building (Fig. 5c). During the climb or fall with the building, downwash winds with high wind speeds occur (as shown in the red dashed circles). Due to the complexity of the building layout, local circulation is formed on the west side of the BSC monitoring site, making the airflow around the building on the south side of the station accumulate and form an obvious convergence area (Fig. 5c), which is not conducive to the air circulation and pollution transport (as shown in the red box).

Figure 6 shows the distributions of near-surface wind and its divergence under four different background wind directions. In general, the overall wind direction in the area is consistent with the background wind direction, but the airflow near the surface is significantly affected by the building layout, thus forming local circulations with divergence or convergence zones. The wind speeds in the areas with dense buildings are significantly smaller than those in open areas (Fig. 6a-1, b-1, c-1, and d-1), which is attributed to the obvious frictional dragging effect of the dense buildings. The overall wind direction in the area is generally the same as the background wind direction, but the airflow is diverged or converged by the influence of the building layout, resulting in a great difference in wind direction inside the block from the background. When the background wind direction is north or west (Fig. 6b and c), the overall wind speed in the area is relatively large. This is mainly due to the temperate monsoon climate in Zhengzhou, where northwest and west winds prevail in winter and wind speeds are relatively high.

It is found that the windward side of the building is mostly a divergence zone, and the leeward side is mainly a convergence zone, which is more obvious for higher buildings. When the airflow meets the building, the airflow on the windward side of the building is blocked and thus spreads outward, forming a divergence zone. The airflow on the leeward side of the building converges and generates a vortex with lower wind speed, forming a convergence zone. For example, at the BSC monitoring site, when the background wind direction is west, the wind speed on the windward side of the building is higher and diffused outward by the building blockage (Fig. 6c-2), resulting in a significant divergence zone (Fig. 6c-3). High-rise buildings have a greater impact on the wind field and cause a strong degree of convergence and divergence. It can be seen that the degree of divergence or convergence around the high-rise building is more significant than those around low buildings in the area (Fig. 6b-3, c-3, and d-3). In addition, the ventilation is better when the dominant airflow is in the same direction as building layout (Fig. 6c). In the process of urban construction, the influence of prevailing wind direction on the layout of buildings should be considered, which could effectively improve the efficiency of urban ventilation.

In order to analyse the potential contribution of the traffic source to different locations, the receptor sites were selected at different locations and heights, and the overall PCR of all wind directions for January 2019 was calculated by RTA (Table 4). Receptor sites S2–S8, with an identical horizontal location but different heights, are selected to investigate contributions of traffic sources to receptor sites at different heights. The PCR of all four roads are 4.05 %, 4.25 %, 4.33 %, and 4.67 % for receptor sites S2 to S5, with heights of 2, 5, 10, and 15 m, respectively. However, as the receptor height continues to rise, namely from S5 to S8, the PCR of the roads gradually decreases from 4.67 % to 3.55 % (Table 4). It is noteworthy that the contributions from R1 and R3 are primary, especially R1, which may be due to the closer distance to the site and the generally northeast wind field. The potential contribution of the traffic source is the greatest when the receptor site is located at a height of 15 m, suggesting the air quality at that height is most susceptible to traffic emissions under the northeast wind field. In addition, according to density distribution (refers to the number of particles that have stayed in the space of 10m×10 m in the horizontal direction and 5 m from the surface to above in the vertical direction) of all trajectory points that have passed through the traffic roads (Fig. 7), it can be seen that the road section with a large potential contribution to the receptor sites is generally located to their northeast, which might be a result of the combination effect of the background wind field and the building layout. For more details, the vertical structure of winds along the direction of the wind field at the receptor site S2 (Fig. 8b) is also presented. It can be seen that there is a general upward motion in the airflow, showing the position with the greatest PCR from traffic sources is located at a certain height, which is about 15 m over the receptor site S2 in this case.

It can be seen from Table 4 that R1 is the road with the greatest potential contribution to the receptor sites. The horizontal distance between road R1 and the receptor sites is about 300 m, and the peak of the PCR occurs at a height of about 15 m (corresponding to the site S5). However, for the road R3, which is closest to the receptor sites in the horizontal (about 200 m), the contribution ratios are lower than those of the road R1. Figure 8a shows that the near-ground winds are generally northeast, resulting in the probability of traffic contributions from R1 and R3 road sections upwind of the site S2 being roughly the same. Nonetheless, as mentioned in Sect. 3.3, the width for the road R1 is about twice that of the road R3. Therefore, even though R1 was a little farther from the receptor sites than R3, the contribution ratios of R1 to the sites were calculated to be larger than those of R3. For R2 and R4, the distance from the receptor sites is about 1200 and 1500 m, respectively, farther away than those of R1 and R3. In addition, under northeast winds, the traffic source was hardly transported to the receptor sites, rendering the contribution ratios quite small below 50 m (Table 4). It can also be seen, from Table 4, that the corresponding PCR of R2 and R4 may peak at a height over 50 m.

Since the road R1 had the largest potential contribution to the receptor sites, the contribution of R1 to different positions is focused in the subsequent discussion. For the receptor site S1, which is about 400 m from R1, located in a dense building area with the building height at 30 to 40 m, the PCR of the traffic source to the receptor site is calculated to be 1.81 %. For the receptor site S2, which is about 300 m from the traffic road, located in an open area and surrounded by low buildings, the PCR of the traffic source is determined to be 2.38 %. It might be inferred that the wind field difference partially resulted from the influence of building layout and led to the higher contribution ratio to S2. From the average wind field in January 2019 (Fig. 8a), it can be seen that the winds were influenced by high-rise buildings around S1, resulting in a change in transport path of pollutants and thus making it difficult for pollutants to reach the S1 site. However, for the S2 site, the winds were less influenced by the buildings, and pollutants were more easily transported there.

In order to investigate the potential contribution of traffic sources under different background wind directions, the receptor site S2 influenced by R1 under the east, the south, the west, and the north wind directions was classified from the simulation period. The PCRs of the traffic source were estimated to be 2.45 %, 0.07 %, 1.98 %, and 2.97 % for the east, the south, the west, and the north wind directions, respectively, revealing that the difference in potential contribution was largest between the south and north wind directions. When the background wind direction was south, the receptor site was located upwind of the road, and the road traffic source contributed very little to the receptor site. On the contrary, when the receptor site was downwind of the road with northern winds, the contribution ratio of road traffic source to the receptor site was the greatest. When the background wind direction was east and west, the contribution ratio to the receptor point was similar, ranging between the ratios under the south and north wind directions. The lower contribution ratio during westerly winds relative to that under easterly winds might partially be due to the denser distribution of buildings upwind of the receptor site. Complex building layouts changed the structure of the wind field and thus had an impact on the transport of pollutants. The slow air circulation in dense building areas made it unfavourable for pollutants to be transported. In the windward side of the dense building area, the wind was blocked and diverted to both sides of the building. Pollutants were difficult to transport to the leeward side of the building, where the receptor site was located. The results of the potential contribution of traffic sources under different background wind conditions are helpful to understand the street-level pollution transport characteristics and provide effective suggestions for the traffic pollution control strategies.