Investigating Cyclicality and Seasonality in Black Carbon Pollution Sources via Linear Regression

Sammanfattning

The focus of this project was to investigate possible daily and yearly patterns in black carbon emissions throughout Europe based on both actual data obtained from 2018 to 2019 and a transport model created to identify the sources of said black carbon. Previous work with the data had shown large gaps between the observed and predicted values, and our hope was to identify cyclical or seasonal patterns that would explain these gaps. By stratifying the data separately according to time of day and month of the year, wecreated individual linear regression models for each subset of data. From here, we ran two parameter reduction methods to help identify possible significant covariates, checking for both common and unique covariates between each model. Finally, we used the reduced parameters and the original transport model predicted values to create one more set of models to investigate. Analysis by time of day showed little in the way of clear patterns, but did help to confirm key covariates suggested by previous work. The eastern European regions which had been identified as areas with underreported black carbon emissions appeared consistently across models. Monthly analysis provided somewhat more interesting results. Regions suspected of contributing more black carbon than predicted were most often identified by our models during the summer months. This was in contrast to our assumption that their overcontributions would be higher in winter, where the gaps between observed and predicted values were greater.