| Abstract: | Statistical regression models are presented that explain the observed variations, across urban areas, in the concentrations of two major pollutants, ozone and carbon monoxide. Model specification and estimation are based on an explicit and new spatial framework derived from the theoretical concept of well‐mixed cells, whereby the basic Fickian system of diffusion equations is integrated over the regional space partitioned into a grid of large cells. The concentration in each cell results from the balance of pollutant flows into and out of this cell and of pollutant emissions and removal within that cell, and is expressed as the sum of two concentration contributions: (1) the local effect, dependent upon pollution‐related factors around the measuring station, and (2) the regional effect, dependent upon pollutant flows originating outside the local area. A large database is developed, making extensive use of GIS technology, to spatially relate such data as pollution measurements, meteorological factors, land‐use characteristics, census socioeconomic data, and major highway network characteristics. The results confirm the appropriateness of the well‐mixed cell framework, are in line with general knowledge regarding the determinants of ozone and carbon monoxide concentrations, and clarify the role of transportation, residential fuel use, economic activities, natural environments, and meteorological factors such as temperature and solar radiation. About SO percent of the variations in concentrations are explained by these models. Several areas of further research are outlined. |
