Evaluation of the Street Canyon Level Air Pollution Distribution Pattern in a Typical City Block in Baoding, China
3.1. Effects of Urban Morphological Factors on Dispersion of Traffic Pollutants
C
=
β
0
+
β
1
∗
E
+
β
2
∗
DR
+
β
3
∗
(
H
/
W
)
+
β
4
∗
(
H
1
/
H
2
)
+
β
5
∗
BCR
+
β
6
∗
Rugosity
+
β
7
∗
Porosity
+
β
8
∗
Occlusivity
+
ϵ
(1)
−2·s−1”, and units for DR and Rugosity are meters (m) Other variables are ratios and do not have units.
The spatial distribution of the air pollutants can be found in Section S4 in the SI. An EOF analysis is also conducted to determine the contribution of various factors on the studied area. From the results, ground-level distribution of air pollution is determined not only by traffic-related emissions, but also by other factors, including background transport and urban morphology (please refer to Section S4 and Figures S10–S18 in the SI for detailed discussion). To quantitatively understand the contribution from individual factors, we constructed a multi-linear regression model and used traffic emissions and the abovementioned seven urban morphological parameters as the independent variables to explore their association to the mean concentration distribution of individual air pollutants. For each grid at the ground level over the study domain, we have a group of data, including mean pollution concentration ( Figure S9 ), emissions and seven building morphological parameters ( Figure S2 ). The grids occupied by buildings are omitted and we finally obtain 23,338 sets of data. The regression results are shown in Table 2 (see Table S5 for the standard error). The algebraic equation of the regression model in Table 2 takes the following form:where C refers to concentration and E refers to emission. Other abbreviations can be found in Table 1 . The unit for emission is “µg·m·s”, and units for DR and Rugosity are meters (m) Other variables are ratios and do not have units.
3 have an R2 over 70%, showing that the selected eight parameters can explain most of the spatial distribution of traffic pollutants. O3 has an R2 of 47%, much lower than the other pollutants, indicating that O3, as a secondary pollutant, is only indirectly related to traffic emissions (through the NOx titration process). These traffic-related parameters, e.g., emissions and DR, are generally having opposite effects on O3 compared to other pollutants.
From Table 2 , all pollutants except Ohave an Rover 70%, showing that the selected eight parameters can explain most of the spatial distribution of traffic pollutants. Ohas an Rof 47%, much lower than the other pollutants, indicating that O, as a secondary pollutant, is only indirectly related to traffic emissions (through the NOtitration process). These traffic-related parameters, e.g., emissions and DR, are generally having opposite effects on Ocompared to other pollutants.
2 and CO, the correlation for all parameters is generally significant, except for the correlation between “Rugosity” and CO. Among the eight parameters, “DR”, “Rugosity” and “Occlusivity” are negatively correlated with the spatial distribution of air pollution. The negative correlation of “DR” shows that the longer the distance from the road, the lower the ground-level pollution concentration, mainly reflecting the dispersion pattern of traffic emissions. The coefficient of NO2 is about twice that of CO, showing that NO2 has a larger change when going away from the main roads. Increased building heights can lead to deeper canyons, where the ground-level wind speed becomes larger and contributes to a quicker dispersion, resulting in lower pollution levels. This is why “Rugosity” has a negative correlation. As previously discussed, “Occlusivity” indicates the openness of the building groups, and the negative correlation with pollution distribution indicates that areas with higher openness have higher pollution levels. However, by common sense, higher “Occlusivity”, i.e., lower openness, should give rise to worse ventilation and, finally, lead to higher pollution. Therefore, to explain this contradiction and reveal the true effect of “Occlusivity”, we also conducted an additional correlation analysis to differentiate the effects between road and non-road areas (
For NOand CO, the correlation for all parameters is generally significant, except for the correlation between “Rugosity” and CO. Among the eight parameters, “DR”, “Rugosity” and “Occlusivity” are negatively correlated with the spatial distribution of air pollution. The negative correlation of “DR” shows that the longer the distance from the road, the lower the ground-level pollution concentration, mainly reflecting the dispersion pattern of traffic emissions. The coefficient of NOis about twice that of CO, showing that NOhas a larger change when going away from the main roads. Increased building heights can lead to deeper canyons, where the ground-level wind speed becomes larger and contributes to a quicker dispersion, resulting in lower pollution levels. This is why “Rugosity” has a negative correlation. As previously discussed, “Occlusivity” indicates the openness of the building groups, and the negative correlation with pollution distribution indicates that areas with higher openness have higher pollution levels. However, by common sense, higher “Occlusivity”, i.e., lower openness, should give rise to worse ventilation and, finally, lead to higher pollution. Therefore, to explain this contradiction and reveal the true effect of “Occlusivity”, we also conducted an additional correlation analysis to differentiate the effects between road and non-road areas ( Table S2 ). The results show that “Occlusivity” is the only parameter with opposite correlation results between road and non-road areas. Over the non-road areas, “Occlusivity” has a positive correlation, just as expected. The negative correlation in road areas, however, is because the lower openness in the roads is related to narrow parts of the streets, where traffic numbers and its related emissions are lower. Since traffic emissions are very important in the studied block, the correlations with road areas are more significant in the correlation, and the combination of the different effects over road and non-road areas finally results in the negative correlation.
The remaining five parameters are all positively correlated with pollution distribution. “Emission” is the source of pollutions, and this result is expected. As for “Aspect Ratio” and “Asymmetry”, which describe the shape of street canyons, higher “Aspect Ratio” (which implies a deeper street canyon) and “Asymmetry” values (in this case closer to 1, which implies a more asymmetric canyon) lead to a higher pollution concentration within the street canyon, which is consistent with previous studies. “BCR” is related to building density, and areas with higher “BCR” values usually have worse ventilation, resulting in ground-level pollution accumulation. “Porosity” reflects the open volume ratios of the building groups. Higher “Porosity” values mean there is less space occupied by buildings, which can enhance the air flow and lead to stronger ventilation. However, another effect, which is that traffic-emitted pollutants are carried by the air flow into the residential areas, also have significant influence and override the effects of dispersion in this case, and finally results in the positive correlation.
3 are generally opposite to the results of NO2 (x consumes O3 and increases NO2 concentrations at the same time. Therefore, in areas where NO2 have high concentration, O3 has relatively low concentration. This also indicates that controlling traffic emissions over a small city block may lower local NOx and CO concentrations, but has little effect on O3. The most effective way to control O3 concentration over the urban area is to reduce the background contribution, which requires sophistically controlling both VOCs and NOx emissions over a much broader area.
The correlation patterns for Oare generally opposite to the results of NO Table S2 ), as the titration by NOconsumes Oand increases NOconcentrations at the same time. Therefore, in areas where NOhave high concentration, Ohas relatively low concentration. This also indicates that controlling traffic emissions over a small city block may lower local NOand CO concentrations, but has little effect on O. The most effective way to control Oconcentration over the urban area is to reduce the background contribution, which requires sophistically controlling both VOCs and NOemissions over a much broader area.
C
=
β
0
+
β
1
∗
DR
+
β
2
∗
(
H
/
W
)
+
β
3
∗
(
H
1
/
H
2
)
+
β
4
∗
BCR
+
β
5
∗
Rugosity
+
β
6
∗
Porosity
+
β
7
∗
Occlusivity
+
ϵ
(2)
As previously discussed, traffic emissions are the most important factor in the area. Therefore, it accounts for a large proportion of the pollution distribution. Nevertheless, the effects of urban form also have important effects. To further explore the effects of city form on pollution distribution, we also conducted a correlation with the emission term excluded. The results are shown in Table 3 (see Table S6 for the standard error). The algebraic equation of the regression model in Table 3 takes the following form:where C refers to concentration. Other abbreviations can be found in Table 1 . The units for DR and Rugosity are meters (m). Other variables are ratios and do not have units.
Without the effects of emission, the regression models for NO2 and CO still have an R2 over 50%, showing that urban form still has reasonably large effects on pollution distribution. Compared to the previous regression, the results for individual morphology parameters are similar, except that “Rugosity” becomes insignificant in this case. These results indicate that a better designed urban form may be helpful for enhancing air ventilation and reducing pollution accumulation, resulting in a better urban air quality and a lower human exposure with the same traffic emissions.
2 concentration and the standardized NO2 emission over the road area is 0.46, which means that every additional percentage of emission leads to a 0.46% increase in NO2 concentration.
We also calculated the association rates between the dependent and each independent variable in both road areas and non-road areas. All the independent and dependent variables are normalized into the range of 0 to 1. The simple correlation results are shown in Table 4 (see Table S7 for the standard error). The regression coefficient of each independent variable here means the rate of the change of the dependent variable is associated to one unit change in the corresponding independent variable. For example, the regression coefficient between the standardized NOconcentration and the standardized NOemission over the road area is 0.46, which means that every additional percentage of emission leads to a 0.46% increase in NOconcentration.
The results show that the correlation coefficients over the road areas are generally larger than those in the non-road areas. This is mainly because the emissions in the road areas can lead to larger rangeability in the concentration than in the non-road areas. From the results, “Emission”, “Aspect Ratio”, “Asymmetry” and “Rugosity” have major impacts, i.e., an additional increase/decrease of over 10% occurs when these factors have a 1% change. Therefore, the above indices should be given more attention in future urban planning. Another interesting finding is that the correlation coefficient between O3 and “Rugosity” (which is 0.73), indicating that high buildings should be avoided alongside the streets, as they cause large O3 accumulation in the street canyons. As for the non-road areas, the “DR” is much larger than other parameters, which indicates that the distance to the pollution source is the dominant factor over the non-road areas. The other parameters, e.g., BCR, Rugosity, Occlusivity and Porosity, though associated with smaller coefficients compared to “DR”, can also significantly affect the traffic pollution dispersion, and thus could serve as the fine-tuning parameters in urban canopy design.
