As air pollution becomes highly focused in China, the accurate identification of its influencing factors is critical for achieving effective control and targeted environmental governance. Land-use distribution is one of the key factors affecting air quality, and research on the impact of land-use distribution on air pollution has drawn wide attention. However, considerable studies have mostly used linear regression models, which fail to capture the nonlinear effects of land-use distribution on PM2.5 (fine particulate matter with a diameter less than or equal to 2.5 microns) and to show how impacts on PM2.5 vary with land-use magnitudes. In addition, related studies have generally focused on annual analyses, ignoring the seasonal variability of the impact of land-use distribution on PM2.5, thus leading to possible estimation biases for PM2.5. This study was designed to address these issues and assess the impacts of land-use distribution on PM2.5 in Weifang, China. A machine learning statistical model, the boosted regression tree (BRT), was applied to measure nonlinear effects of land-use distribution on PM2.5, capture how land-use magnitude impacts PM2.5 across different seasons, and explore the policy implications for urban planning. The main conclusions are that the air quality will significantly improve with an increase in grassland and forest area, especially below 8% and 20%, respectively. When the distribution of construction land is greater than around 10%, the PM2.5 pollution can be seriously substantially increased with the increment of their areas. The impact of gardens and farmland presents seasonal characteristics. It is noted that as the weather becomes colder, the inhibitory effect of vegetation distribution on the PM2.5 concentration gradually decreases, while the positive impacts of artificial surface distributions, such as construction land and roads, are aggravated because leaves drop off in autumn (September-November) and winter (December-February). According to the findings of this study, it is recommended that Weifang should strengthen pollution control in winter, for instance, expand the coverage areas of evergreen vegetation like Pinus bungeana Zucc. and Euonymus japonicus Thunb, and increase the width and numbers of branches connecting different main roads. The findings also provide quantitative and optimal land-use planning and strategies to minimize PM2.5 pollution, referring to the status of regional urbanization and greening construction.
Keywords: PM2.5; boosted regression tree model; land-use distribution; seasonal variations.