Identification of target source tracers. Before the start of garbage classification, 80% of the MSW sent to the LG base was disposed of in landfills. Landfills were undoubtedly one of the main sources of VC emissions; however, garbage classification and the changes it caused also had the greatest impact on landfills. Taking this as an opportunity, we took the landfill source as the target source and identified its tracers and fingerprint spectrum using the K-means clustering algorithm based on massive online data. The k-means clustering algorithm available in the SPSS software was used in this study. The k-means clustering technique divides the multi-dimensional data into a predefined number of subgroups, which are as different as possible from each other, but as coincident as possible within themselves. This was done by iteratively minimizing the sum of the squared Euclidean distances from each member to its cluster centroid.35,36
The clustering results are presented in Table S4. According to the position of the YQ site and landfill A, when the wind direction was from 0° to 90°, VOCs data could best represent the pollution emission from landfill A. Cluster 2 in stage 1, cluster 1 in stage 2 and cluster 2 in stage 3 satisfied the above wind direction conditions, and the final clustering of their wind direction was 32.51°, 32.08°, and 35.43°, respectively. The final data results of the three stages after the filtering process are listed in Table S5. Although the number of cases has been greatly reduced, these filtered data were the most representative of local source emissions from landfill A. The TVOC concentrations in the stages 1–3 were 38.63, 23.74, and 12.87 ppb respectively. The overall downward trend was more pronounced than that before filtering. Moreover, TVOC concentrations in the stages 1 and 2 were higher than those in the unscreened data from the same period, proving that these data were more representative. In general, the k-means clustering method has shown good results when sifting through a large amount of continuous observational data for more representative cases.
As shown in Table S5, the overall concentration of VOCs discharged from landfill A gradually decreased with the implementation of garbage classification and reduction in the scale of landfill waste disposal. It can be inferred that the concentration and proportion of VOC species discharged from single-source landfill A in the data filtered by k-means clustering should also conform to the overall downward trend. From the variation in proportion of the 24 VOC species, only n-butane, i-butane, and i-pentane exhibited gradual downward trend from stage 1 to stage 3 (Fig. S7) and their proportions in stage 1 were 20.1%, 13.5%, and 12.1% respectively. As the contribution of landfill sources decreased, the proportions of n-butane, i-butane and i-pentane decreased to 15.9%, 10.9%, and 7.4%, respectively. Their proportions were similar to those of the synchronous decline during stage 3.
Identification of main sources tracers. In general, there are several emission sources in addition to landfills as a large comprehensive MSW disposal base. To explore the variation in characteristics of different emission sources after the implementation of garbage classification, appropriate characteristic tracers from 28 VOC species were selected based on correlation analysis.
The observational data of 11 VOC species with relatively high concentration levels (ethane, n-butane, i-butane, i-pentane, ethylene, acetylene, benzene, toluene, ethylbenzene, o-xylene, and m,p-xylene) before garbage classification were used for correlation analysis using the SPSS software (IBM). The calculated Pearson correlation coefficients are shown in Fig. S8. This figure shows the following: (1) There was a pole-strong positive correlation (r > 0.9) between n-butane, i-butane and i-pentane. A good mutual correlation among the species indicates that they may primarily originate from the same source.37 Combined with the results of the identification above, these three VOC species can be considered as tracers of landfill emission sources. (2) A strong positive correlation (r > 0.6) was observed between benzene, ethane, ethylene and acetylene. These four VOC species are typical indicators of combustion and transportation sources. 38–40 (3) The strong positive correlation (r > 0.7) between toluene, ethylbenzene, o-xylene and m,p-xylene implied that the four aromatic hydrocarbons possibly shared similar sources. Most aromatics are xenobiotic and originate through direct volatilization from disposed waste. Such compounds can be released from paper waste, plastic packaging, solvent-containing waste, and petroleum products.41 Therefore, toluene, ethylbenzene, o-xylene and m,p-xylene could be considered tracers of chemical volatilization in disposed waste.
Based on the correlation analysis, we determined the tracers and fingerprint spectrum database of three relatively independent potential emission sources (landfill, transportation and combustion, and volatilization from chemicals) in the LG base.
Quantification of emission variation characteristics of major sources. In this study, the relative changes in concentrations among different source tracers were used to analyze the variation characteristics of different emission sources at the base. The contribution of the emission source to the ambient air is represented by the sum of the concentrations of tracers belonging to the same source. In addition, all concentration data were normalized to ensure comparability. Ternary diagrams of the three sources are shown in Fig. 4, where each dot represents a group of 1-hr data.
As shown in Fig. 4(a), most of the data before garbage classification were distributed in the bottom half of the subfigure. Because the data points were scattered, the median ratio of each source was used as the basis for the judgment. In stage 1, the median ratios of transportation and combustion, landfill and volatilization were 0.46, 0.35 and 0.12, respectively. After garbage classification, the contribution of transportation and combustion sources increased to 0.60, whereas the median ratio of landfills decreased to 0.22. The change in the ratios clearly showed that landfill emissions were effectively restrained owing to the decreased moisture content in the disposed MSW and the reduction in landfill scale. After entering stage 3, it was evident that the points were concentrated in the lower right corner of the ternary diagram, and the trend was similar to that in stage 2. The ratio of landfill sources gradually decreased relative to the other two emission sources. Interestingly, the median ratio of volatilization from chemicals did not change significantly after garbage classification, which were 0.12 and 0.14. In summary, landfill pollution was effectively controlled after garbage classification based on the analysis results of tracers and fingerprints, however, the contribution of combustion sources to the base increased substantially.
To further explore the variation tendency of traffic and combustion sources, tracers other than benzene (benzene concentration levels remained low throughout the observation period) were separately drawn using a ternary diagram (Fig. 4b). In stage 1, the points clustered in the lower-right corner of the subfigure (the median ratios of ethane, ethylene and acetylene were 0.53, 0.23, and 0.24, respectively). Previous studies have shown that ethane is one of the most important alkanes in urban ambient air VOCs because of its wide range of sources, including fossil fuel use, biomass combustion and vehicle emissions.42–44 After garbage classification, it was evident that the points were more scattered, with multiple points gathering on the left side of the subfigure. Surprisingly, the median ratios of ethylene and acetylene increased to 0.36 and 0.28. In addition to the fact that landfills were replaced by other disposal facilities, the transportation of waste within the base also changed because of garbage classification. Most importantly, the requirement for dry and wet waste to be transported separately to specific treatment units for disposal has significantly increased the frequency of trucks and garbage vessels in the base. These large trucks and garbage vessels are powered by diesel engines, and ethylene and acetylene are characteristic VOCs in diesel engines.45,46 As landfill A was sealed after reaching the treatment load, the garbage transportation frequency around the YQ site decreased significantly. Simultaneously, the increased disposal scale of incineration plants and the implementation of landfill gas collection power generation projects caused the contribution of ethylene and ethane to be dominant (the points were clustered along the right side of the subfigure) in stage 3, while the median ratio of acetylene as the main tracer for diesel decreased to 0.10.
Figure S9 shows the effects of wind direction on the concentrations of the 11 source tracers. First, when the monitoring site was downwind of landfill A, the average concentrations of the three landfill tracers (n-butane, i-butane and i-pentane) were generally higher; however, there was less obvious directional dependence after garbage classification. The dependence disappeared completely after landfill A was closed. The average concentrations of four volatilization tracers (toluene, ethylbenzene, o-xylene and m,p-xylene) were higher when the monitoring sites were upwind and downwind of landfill A. In contrast, whether before or after garbage classification, pollutant roses of transportation and combustion tracers other than acetylene indicated nearly even contributions from all directions. The concentration level of acetylene was higher in stage 2 than in the other periods because of more vehicle emissions along the way when winds originated from the southwest.
Implications. The main purpose of this study was to explore the impact of the substantial change in the method of urban garbage collection on the air pollution in a super-large MSW disposal base located in the city. Garbage classification greatly reduced the content of kitchen waste in landfill disposal, and also resulted in the appropriate and harmless disposal of dry and wet waste, which provided conditions for the gradual replacement of landfills by diversified end disposal technologies. Eventually these important factors improved the air quality in the LG base. Thus, we concluded that garbage classification may be an effective means of improving air quality. The new quantification method for the impacts of the major source proposed in this study showed good applicability in a complex multi-source environment similar to the LG base. The results show that while landfill emissions are well controlled by garbage classification, the resulting increase in emissions from transport sources cannot be ignored. In addition, emissions from other waste treatment units such as waste incineration and landfill gas power plants, require more attention. The above problems of pollution source transfer must to be comprehensively considered by internal managers from two aspects: local air quality benefits and waste disposal efficiency. Finally, we believe that the combination of high-resolution online monitoring technology and analysis methods based on the fingerprint spectrum of various emission sources will play a more important role in future research in other fields.