// METHODOLOGY + STUDY ANALYSIS

This project explored particulate matter levels in New York City by season during the year of 2012. In order to produce interpolation maps of the particulate matter levels , the Geo-Analyst Kriging analysis tool was used in ArcGIS. As to validate this process, the Kriging method was first applied to the winter pollution levels, using the default settings. In exploring this data, the trend analysis suggested an inverted parabola shape, which displayed second order polynomial characteristics. This indicated that the Kriging model was a good choice to use for this study.

The Kriging model was then applied once again to the pollution data for winter, however with some alterations to the original setting. First, the second-order-polynomial trend was used. This brought the model’s standard prediction error closer to one and decreased the root-mean-squared. The data for this analysis indicated that most of the lines from the joined pairs were in a Northeast to Southwest part of the map. A third, and final, Kriging analysis was then done, again with second-order polynomial, but this time turning on the “anisotropy” setting, with a lag of 500. This allowed for a more localized look at the data points. An angle of 29.4 degrees was also chosen to denote the Northeast/Southwest pattern observed in the data. This brought the model’s standard predicted error closer to one and decreased the root-mean-square to close to zero. The root-mean-square indicates how accurate the model is, with higher accuracy levels when it is closer to zero. This same process was applied for the surface modeling procedure of the other three season. The data was symbolized using equivalent classifications for all maps in order to allow for comparison between pollution levels in each season.

After determining the average surface characteristics for each census tract (based on the centroid point), the “Validation/Prediction” tool was used to assign a predicted value of pollution for each centroid. The top five census tracts for each season were then compiled, as shown in the charts below. The pollution levels in each census tract varied based on season; however, summer and autumn shared similar results. In fact, four of the five highest polluting census tracts were the same, exhibiting the highest levels of pollution in summer and in autumn. These tracts were: 244.02, 244.01, 248, and 198. From this analysis, we can infer that pollution levels are similarly concentrated across particular seasons; particularly  in summer and autumn. Additionally, similar particulate matter level patterns can be found in summer and autumn, as well as in spring and winter.

 

 

This was a group project for an advanced GIS course at the Graduate School of Architecture, Planning, and Preservation at Columbia University. Other group members include Olivia Jovine and Kellie Radnis. All graphics were completed by Sharon Moskovits. 

Sources: New York State Department of Environmental Conservation, NYC Department of City Planning

 

download the complete project here.