Article Review: Spatial Correlation Between Traffic and Air Pollution in Beijing

Introduction

            Traffic related air pollution (TRAP) is thought to be an increasingly important problem in major cities like Beijing, China.  The researchers in this article wanted to find out if a new method would enhance and simplify the spatial relationship between traffic and air pollution.  By analyzing particulate matter (PM) 2.5 concentrations at various sites in the city, they were able to spatially plot the relationship by involving traffic monitoring information (Jiang et al 2019, 655).  Particulate matter, such as soot, ash, and sulfur oxide (ibid. 655) with a diameter greater than 2.5 micrometers has been shown to be toxic to humans, creating a higher risk for lung diseases when inhaled.  The researchers’ simplification of the methods involved in determining this relationship was not as successful as they concluded.

 

Literature Review

            Studies on the relationship between traffic and air pollution are prolific and consistently affirmative (ibid. 655).  Mathematical modeling has been the primary method of researching it, but the drawback is that it can only apply to observation sites (ibid. 655).  The researchers answered this problem by applying geostatistics and a grid-orientation method instead (ibid. 655).  This would ideally be a cheaper way of doing the research, producing better results.

 

Background and Geographical Context

The research was conducted in central Beijing, where air pollution is notoriously bad.  It has a population of over 20 million people and is surrounded by mountains (ibid. 659) that amplify stagnant air.  Time is another dimension of TRAP research.  The optimal time for TRAP research is during rush hour, which is from 5am to 9am and 5pm to 9pm in Beijing (ibid. 656).  This is when peak emissions from vehicles are most likely to occur, providing for research that is better supported by data.  The traffic data itself was retrieved from “car rental companies who installed location devices on all their vehicles” (ibid. 659).  A gridding system was used across central Beijing to record the results (ibid. 657).

 

Methods Used

            A statistical package involving spatial autocorrelation was implemented to explore the relationship between traffic and PM 2.5 levels.  Spatial autocorrelation is a measure of the dependency of data that is near an observation (Clifford et al 2016, 541).  Moran’s I is the mathematical tool used in spatial autocorrelation; it estimates the level of clustering in a set of data (ibid. 542) by measuring how similar the values are to those nearby.  It is a statistic commonly used to evaluate the significance of a relationship between variables.

            GIS is an optimal program for calculating spatial autocorrelation.  In addition to Moran’s I, it can generate various analyses with powerful precision.  One of these is called a hot spot analysis (ibid. 675-677), which plots on a map the areas of high correlation vs. the areas of low correlation.  Other GIS tools can then be used to enhance the analysis through gridding, which creates a raster of the data.

In the Beijing study, GIS was the primary vehicle of data processing.  First, a hot spot analysis was used to find the concentration of levels of traffic and PM 2.5 concentration (Jiang et al 2019, 657).  Gridding was then used to create a raster of the data based on values of the hot spot analysis (ibid. 657).  To create the grid, the Fishnet tool was used to construct polygons over the point features; Spatial Join counted the number of vehicles in each grid; and the Zonal Statistics tool calculated the grid’s average pixel values (ibid. 657).  Spatial autocorrelation was then needed to measure the significance of each grid cell’s value compared to its neighbors.  Moran’s I was then processed in GIS to extrapolate how clustered the cells were (ibid. 658).  Finally, a correlation between the two variables was calculated using the LISA method (ibid. 658).  This resulted in correlation maps showing each grid’s correlation value at certain times and days of the month.

 

Analysis and Discussion

            It was found that days of rain and strong wind significantly impacted PM 2.5 levels, making it less severe than on sunny days (ibid. 662).  The northwest area of Beijing had the worst TRAP on average, associating the highest levels of PM 2.5 with the highest traffic.  The southeast area of Beijing also showed a strong correlation, with the lowest levels of PM 2.5 being associated with the lightest traffic (ibid. 663).  The pattern would change on weekends, when there were fewer routine commutes to work and more random scattering.

            The researchers state that the most likely reason for there being a lower correlation in southwest Beijing is that the wind generally blows from north to south, making the PM 2.5 concentration higher while the traffic is relatively lower (ibid. 664).  Other areas had a lower correlation, like the south-central.  Several reasons are posited for why, but none of them are convincing.  The researchers favor there being other sources of PM 2.5 in the area than traffic congestion (ibid. 665), which is certainly plausible.  Possible sources of PM 2.5 range from vehicle exhaust to smoke from wildfires, factory emissions, dust and salt.  Based on findings from Karagulian et al (2015), only 25% of urban PM 2.5 comes from traffic, with another 15% coming from industrial activities, 20% by domestic fuel burning, 18% from dust and salt, and 22% from unspecified human activity.  Any of these factors could be skewing data from the Beijing study.

 

Conclusion

            When there isn’t a significant finding from a study, it’s important to state it in the abstract.  This saves the reader’s time and energy.  While it was stated that the methods could be a useful guide for future research, it doesn’t hold sway if the research doesn’t provide an important finding.  The researchers would have made a more powerful statement of the method if it had been successful.  You say more by setting the example, not the possibilities.

Bibliography

 

Clifford, Nicholas, Meghan Cope, Thomas Gillespie and Shaun French. Key Methods in Geography. Third edition. London: SAGE Publications Ltd, 2016.

Jiang, Lili, Ziheng Sun, Qingwen Qi, and An Zhang. “Spatial Correlation Between Traffic and Air Pollution in Beijing”. The Professional Geographer 71, no. 4 (2019): 654-657.

Karagulian, Federico, Claudio A. Belis, Carlos Francisco C. Dora, Annette M. Prüss-Ustün, Sophie Bonjour, Heather Adair-Rohani, and Markus Amann. “Contributions to cities' ambient particulate matter (PM): A systematic review of local source contributions at global level”. Atmospheric Environment 120 (2015): 475-483.  doi:https://doi.org/10.1016/j.atmosenv.2015.08.087.