Immissions profiling using AI

Abstract

Industrial activity and not only, generates both emissions and immissions of pollutants into the atmosphere. Thus, their magnitude and dynamics will present a specific footprint for each pollutant. The article aims to identify immission profiles using tools specific to artificial intelligence applied to a wide set of recordings of environmental parameters. The first part briefly presents the issue of environmental protection and specific regulations at national and European level, and the second part showcases the database of environmental parameters and the theoretical model of data processing. The last part of the paper is dedicated to results obtained and their analysis which shows the presence of emissions patterns (profiles) measured in different locations and time periods. Seasonality and its impact on emission profiles were also analysed. On this occasion, the use of the Hurst exponent allowed the segregation of various time series of data based on the resulting memory interpreted as a measure of immissions’ persistence. Jumps in the temporal dimension of values allowed the anonymous association of immissions with different locations. Analysis of the topology of clusters associated with immission profiles highlighted the presence of two types: rare clusters and dense clusters. Rare clusters can be associated with immission having accidental dynamics and dense clusters can be associated with systematic immissions. Use of the framed method allows for a classification of pollutants resulting in increased chances of solving the environmental impact.