A self-organizing maps methodology for developing a composite air quality and climatic parameters classification

Urban climate change affects important aspects of urban life (health, urban environment and infrastructure) through considerable fluctuations in the values of both climatic and air quality parameters. At the same time, in recent years, the networks of atmospheric pollution and climatic parameters monitoring stations have become denser, leading to more information which, if presented correctly, can guide policy makers to achieve sustainable solutions. Compοsite environmental classifications are a credible tool to describe in an easily comprehensible manner the complex interactions of gaseous and particulate pollutants with climatic parameters in different land use types and urban topography. The aim of this study is the development and implementation of a composite climate - air quality classification in order to describe and study their combined effects on living conditions and quality of life in urban environments. By employing pollutant observations from surface stations and climatic gridded data from reanalysis databases, the available data will be converted into groups of cases through a process which is based on a non-linear method of clustering and categorization. An artificial neural network methodology and in particular, self-organizing maps will be used to convert non-linear statistical associations of input data into simple geometric relationships of points in a low dimensional map. This method can create classifications of air pollutants and climatic parameters that group days which follow specific patterns, hidden due to non-linear interactions. The results can contribute to finding a relationship between ambient air quality and climatic variables and subsequently gaining important knowledge in this field.