Mutual Information in the Air Quality Monitoring Network of Bogota - Colombia

Large urban areas in the developing world are characterized by high population density and a great variety of activities responsible for emission of trace gases and particulate matter to the atmosphere. In general, these pollutants are unevenly distributed over cities according to the location of sources, meteorological variability and geographical features. Urban air quality monitoring networks are primarily designed to protect public health. The meteorological and air quality information gathered by monitoring networks can also be used to understand pollutant sources, sinks, and dispersion processes and to assess the spatial coverage of the network itself. Several statistical and numerical simulation methods allow for the identification of the domain that influences observations at each of the stations, i.e, the zone and respective population truly covered by the measurements. We focused on Bogota, Colombia, a dense city of approximately 9.6 million inhabitants in its metropolitan area. We analyzed the measurements obtained by the Bogotá Air Quality Monitoring Network (RMCAB) between the years 1997 and 2010 for TSP, PM₁₀, CO, NO_x and O₃. RMCAB is composed of 16 stations, 13 of which are fixed and measure both atmospheric pollutants and meteorological variables. The method applied consisted of a statistical approach based on the mutual information that each station shares with its complement, i.e. the set formed by the other stations of the network. In order to improve our understanding and interpretation of the results, virtual data created for selected receptors along a simple modeled Gaussian plume spreading throughout Bogotá was analyzed. In this Gaussian model, we accounted for the prevailing weather conditions of this city and for different emission features under which the pollutants are emitted. The spatial location of the monitoring stations and emission sources, and the quality of the measurements are relevant factors when assessing the mutual information of RMCAB. As expected, it was found that the stations with average concentrations close to the network mean tend to have larger mutual information, whereas stations with atypical values share less information. The degree of dispersion around the mean of the RMCAB measurements does not exhibit a strong correlation with the tendencies observed for the mutual information. In general, the stations around the geographical center of Bogota or close to areas of large emissions, i.e. industrial areas, share the most information, while the stations located on the city outskirts are particularly singular. This degree of correlation as well as its underlying variables provides an approach to identifying the distribution of the pollutants over the city, which in turn gives insight into the spatial influence on monitoring networks. Moreover, it has the potential to contribute to the reconfiguration of existing networks in order to both improve their influence and optimize operational costs. Finally, the results of this method shall be compared with those obtained by diagnostic atmospheric dispersion models in order to improve our understanding of the pollution phenomena and to reduce uncertainties. This is an ongoing research topic.