3.2. Spatiotemporal Heterogeneity of Dominant Air Pollutants Distribution
First, the temporal heterogeneity of the dominant air pollutants, in terms of the number of days on which each pollutant was dominant, was examined. The number of days on which PM
10 was the dominant air pollutant was larger than 27, which was the largest among all the pollutants in spring, at all three sites, other than the Sifang site (
Figure 5). In summer, the number of days on which O
3 was the dominant air pollutant was larger than 25, which was the largest among all the pollutants at all three sites, other than the Licang site. The number of days on which PM
2.5 was the dominant air pollutant was larger than 61, which was the largest among all the pollutants in winter at each site. The sum of days on which O
3 and PM
2.5 as the dominant air pollutant was larger than 51, which was the largest among all the pollutants at each site in autumn (
Figure 5). In general, the dominant air pollutants were PM
10 in spring, O
3 in summer, O
3 and PM
2.5 in autumn, and PM
2.5 in winter.
Next, temporal heterogeneity of the diel variation in dominant air pollutant concentration was examined. In
Figure 6, the curves show a single peak in the chart of diel variation of the concentration of PM
10 at each of the four sites, so PM
10 (the dominant air pollutant in spring) appeared to show significant temporal heterogeneity at the hourly scale for each site. At the hourly scale, it reached its minimums between 20:00 and 3:00 (on the next day) and its maximums between 9:00 and 12:00, and the difference was about 25 μg/m
3. It can be inferred that human activities, such as construction and vehicle driving in the day-time, impact PM
10 diel variation to a certain degree [
19,
27].
O
3 (the dominant air pollutant in summer) showed significant temporal heterogeneity at the hourly scale across all four sites (
Figure 8). The concentration of O
3 at the four sites varied consistently during the course of a day, and the curves obviously showed trough-and-peak variation patterns; the concentration of O
3 was significantly higher in the daytime than at night, and the difference was about 50 μg/m
3. It ascended rapidly from 7:00 and reached its peak at 14:00, then gradually descended and finally reached the trough at 7:00 the next day. The rule of diel variation of concentration of O
3 reveals that photochemical reactions are an indispensable condition for O
3 formation.
The dominant air pollutant was PM
2.5 at each of the four sites during the winter, and the concentration of PM
2.5 reached peaks twice a day. The curves were bimodal, and there was significant temporal heterogeneity in terms of the concentration of PM
2.5 across all four sites (
Figure 7). The difference in concentration of PM
2.5 between the maximum and minimum was about 20 μg/m
3, and it ascended rapidly from 5:00 to 7:00 in the morning, reaching its peaks between 11:00 and 12:00 before decreasing quickly and reaching its troughs between 16:00 and 17:00. Then, it increased and reached small peaks between 21:00 to 23:00 before gradually decreasing and reaching small troughs between 5:00 and 7:00 the next day. The time of peaks and troughs is greatly related with urban rush-hours, which indicates that vehicle emission makes some contribution to the formation of PM
2.5.
Next, the spatial heterogeneity of dominant air pollutants was analyzed. To determine the dominant air pollutants at each site in each season, the number of days during which each air pollutant was dominant was calculated (
Table 4 and
Figure 5). The number of days on whichPM
10 was the dominant air pollutant at the Licang site was 148, about two times higher than at the other three sites. The number of days on which PM
2.5 was the dominant air pollutant was very high at all of the sites, but especially at the Sifang site. The number of days on which O
3 was the dominant pollutant was significantly larger at the Laoshan and Sifang sites than at the Shibei and Licang sites (
Table 4). In summary, PM
10 and PM
2.5 were dominant at the Licang site, while PM
2.5 and O
3 were dominant at the other three sites. Therefore, it can be inferred that industrial dust and construction dust were the main factors causing air pollution in the Licang zone, and it was significantly different in the other three zones.
In order to further identify the distribution characteristics of the three dominant air pollutants, the spatial heterogeneity was tested with hourly concentrations of the three dominant air pollutants among the four sites. The Wilcoxon signed-rank test method is suitable for testing these differences, because the concentrations of the three dominant air pollutants have a one-to-one correspondence by time among the sites. Therefore, the Wilcoxon signed-rank test method was applied to test the difference at the significant level of 5%.
For the Wilcoxon signed-rank test of the concentration of PM
10 between the pairs of sites in spring, all of the
p-values were smaller than 0.05 except in the case of the Shifang-Shibei pair (
Table 5). Thus, it can be deduced that there was significant spatial heterogeneity in the concentrations of PM
10 between all of the remaining pairs of sites. For the Wilcoxon signed-rank test of the concentration of O
3 between the pairs of sites in summer, all of the
p-values were less than 0.05 (
Table 5). This indicates that there was significant spatial heterogeneity in the concentrations of O
3 in summer across all four sites. For the Wilcoxon signed-rank test of the concentration of PM
2.5 between the pairs of sites in winter, all of the
p-values were smaller than 0.05, except in the case of the Laoshan-Shibei pair (
Table 5). Thus, it can be inferred that there was significant spatial heterogeneity in PM
2.5 concentration between the remaining pairs of sites.
Finally, the spatial heterogeneity of the diel variation in concentrations of the pollutants was analyzed. As
Figure 6 shows, the difference in the concentration of PM
10 between daytime and night-time was lower at the Shibei site than it was at the other sites, and the concentration of PM
10 remained near its midday high for a longer time at the Shibei site than it did at the other sites. At the Laoshan site, the concentration of PM
10 stayed low at night for a longer time, the initial ascension time was later, and the peak time was later than at the other sites. The difference in the concentration of PM
10 between daytime and night was higher at the Sifang site than at other sites, and the lengths of ascension time and descent time were nearly equal at the Sifang site. At the Licang site, the concentration of PM
10 was significantly higher, the initial rise was earlier, and the peak time was earlier than at the other sites.
The concentration of O
3 was lowest at the Licang site, and highest at the Sifang site (
Figure 8). There was roughly a 10 μg/m
3 difference in the concentration of O
3 among the four sites. The curves in
Figure 8 show that the variation trend in the concentration of O
3 in summer was nearly the same at all four sites, so there was no spatial heterogeneity in terms of the fluctuation of the concentration of O
3 in summer between the sites. Combining
Table 5 and
Figure 8, it can be deduced that there was a strong correlation between the concentration of PM
10 and the intensity of sunlight and temperature. Sunlight intensity and temperature were both very low at night, and the concentration of O
3 was correspondingly very low. The atmospheric photochemical reactions gradually increased with the increase in the intensity of solar radiation. As the secondary pollution product of photochemical reactions, the concentration of O
3 increased rapidly. After the concentration of O
3 reached its peak, it decreased with the decrease of sunshine intensity and temperature.
The curves in
Figure 7 show that the variation trend in the concentration of PM
2.5 in winter was nearly the same at all four sites, but the fluctuation of it was not obvious in the morning at the Laoshan site. The concentration of PM
2.5 was slightly higher at the Licang and Sifang sites, and slightly lower at the Shibei and Laoshan sites.
3.3. The Effect of the Relevant Factors on the Dominant Air Pollutants
The non-parametric method was applied to fit the marginal distribution, because the air pollution data and meteorological data did not obey the normal distribution or the
distribution. The appropriate copula function models were selected based on the characteristics of the binary frequency histograms and binary probability density functions, while the binary cumulative distribution functions of the copula model were fitted respectively. For example, the binary frequency histogram and the graph of the binary probability density function between O
3 and temperature at the Licang site is shown as
Figure 9a,b. According to the histogram, the Clayton Copula model is suitable for the correlation analysis between them.
The Spearman rank correlation coefficient calculated with the copula models are shown in
Table 6. The main driving factors of diel variations of the dominant air pollutants at the four sites were generally other pollutants such as SO
2, NO
2, and CO, while the meteorological factors play diverse roles (
Table 6). However, the effects also showed significant diversity between factors. In detail, the diel variation of wind speed had little effect on the diel variation of PM
10 at each site, and other factors had certain effects on the diel variation of PM
10 at the Shibei and Sifang sites. The diel variation of NO
2 had little effect on the diel variation of PM
10, and other factors had certain effects on the diel variation of PM
10 at the Laoshan site and the Licang site. Humidity had a great effect on the diel variation of O
3 at each site. The diel variation of wind speed had a great effect on the diel variation of O
3 at Licang site, but little effect at the other three sites. The diel variation of NO
2 had little effect on the diel variation of O
3 at the Laoshan site, but a great effect at the other three sites; the diel variation of CO had a greater effect on the diel variation of O
3 at the Sifang site than at the other three sites. The diel variations of wind speed and humidity had little effect on PM
2.5 at all sites, but other factors had greater effects.