Heavy air pollution, especially fine particulate matter (PM2.5), poses serious challenges to environmental sustainability in Beijing. Epidemiological studies and the identification of measures for preventing serious air pollution both require accurate PM2.5 spatial distribution data. Land use regression (LUR) models
[...] Read more.
Heavy air pollution, especially fine particulate matter (PM2.5
), poses serious challenges to environmental sustainability in Beijing. Epidemiological studies and the identification of measures for preventing serious air pollution both require accurate PM2.5
spatial distribution data. Land use regression (LUR) models are promising for estimating the spatial distribution of PM2.5
at a high spatial resolution. However, typical LUR models have a limited sampling point explanation rate (SPER, i.e., the rate of the sampling points with reasonable predicted concentrations to the total number of sampling points) and accuracy. Hence, self-adaptive revised LUR models are proposed in this paper for improving the SPER and accuracy of typical LUR models. The self-adaptive revised LUR model combines a typical LUR model with self-adaptive LUR model groups. The typical LUR model was used to estimate the PM2.5
concentrations, and the self-adaptive LUR model groups were constructed for all of the sampling points removed from the typical LUR model because they were beyond the prediction data range, which was from 60% of the minimum observation to 120% of the maximum observation. The final results were analyzed using three methods, including an accuracy analysis, and were compared with typical LUR model results and the spatial variations in Beijing. The accuracy satisfied the demands of the analysis, and the accuracies at the different monitoring sites indicated spatial variations in the accuracy of the self-adaptive revised LUR model. The accuracy was high in the central area and low in suburban areas. The comparison analysis showed that the self-adaptive LUR model increased the SPER from 75% to 90% and increased the accuracy (based on the root-mean-square error) from 20.643 μg/m3
to 17.443 μg/m3
for the PM2.5
concentrations during the winter of 2014 in Beijing. The spatial variation analysis for Beijing showed that the PM2.5
concentrations were low in the north, especially in the northwest region, and high in the southern and central portions of Beijing. This spatial variation was consistent with the fact that the northern region is mountainous and has fewer people and less traffic, which results in lower air pollution, than in the central region, which has a high population density and heavy traffic. Moreover, the southern region is adjacent to Hebei province, which contains many polluting enterprises; thus, this area exhibits higher air pollution levels than Beijing. Therefore, the self-adaptive revised LUR model is effective and reliable.