From What and When Happen, to Why Happen in Air Pollution Using Open Big Data

Abstract

The air pollution phenomenon has been often studied from an environmental dimension but not from a spatial big data approach and considering social perception analysis. In order to understand such complex phenomenon a multidimensional analysis of heterogeneous environmental data might provide new insights. Notably, the Mexico government has released open data on air quality that contains the historical behavior of air pollution in Mexico City, while social networks data provides rich descriptions regarding regional social problems. In order to take into account the respective contributions of these two data sources from a spatial-temporal perspective, we introduce a multidimensional approach whose objective will be to integrate these heterogeneous data sources in an unified framework. While human perception often embedded in social media is naturally subjective, public data is rather objective and reliable, while they are described at different levels of temporal granularity and scale. Therefore, the search for a sound integration of these data sources is surely a non-straightforward issue. The research presented in this paper introduces a modelling and data mining approach to search for spatial-temporal patterns that can describe not only what happens, but also why such phenomenon happens. The whole framework is applied to the study of air pollution in Mexico City. The idea is to connect unstructured data (social data) and structured spatial data (open data) through the reconciliation of spatial-temporal correspondences between them to discover new geographic knowledge on Air Pollution phenomenon in Mexico City.