Risk assessment of atmospheric emissions using machine learning

Abstract

Abstract. Supervised and unsupervised machine learning algorithms are used to perform statistical and logical analysis of several transport and dispersion model runs which simulate emissions from a fixed source under different atmospheric conditions.First, a clustering algorithm is used to automatically group the results of different transport and dispersion simulations according to specific cloud characteristics. Then, a symbolic classification algorithm is employed to find complex nonlinear relationships between the meteorological input conditions and each cluster of clouds. The patterns discovered are provided in the form of probabilistic measures of contamination, thus suitable for result interpretation and dissemination.The learned patterns can be used for quick assessment of the areas at risk and of the fate of potentially hazardous contaminants released in the atmosphere.

Algorithm quasi‐optimal (AQ) learning

“…A program called concept association grapth (CAG) was developed by the first author to automatically display such graphs. Figure 1 is a graphical illustration of the rules discovered from an atmospheric release problem 17. Representing relationships with nodes and links is not new nor unique to AQ and has been used in many applications in statistics and mathematics.…”

“…Potential atmospheric hazards include toxic industrial chemical spills, forest fires, intentional or accidental releases of chemical and biological agents, nuclear power plants accidents, and release of radiological material. Risk assessment of contamination from a known source can be computed by performing multiple forward numerical simulations for different meteorological conditions and by analyzing the simulated contaminant clouds with clustering and classification algorithms to identify the areas with highest risk 17…”

Toxicity prediction based on artificial intelligence: A multidisciplinary overview

“…It is worth noting that, in terms of air quality and response to hazards, areas at risk from potentially harmful air pollutants need to be identified. ML has shown to be useful to assess atmospheric emissions risks which precisely evaluate the transport and dispersion of polluting particles according to the complex nonlinear relationships between the specific characteristics of the clouds and the meteorological input conditions 191 . Methods for data collection are evolving to new, more efficient, techniques based on analysis from social media with environmental data, providing accurate emissions inventories 192 .…”

Combined approach of a couple fire model with atmospheric releases: the case of the 2003 Glacier wildfires

“…As a result, aerosols play a role in both the cooling and warming of the Earth. The study of atmospheric emissions is also crucial to identify high risk areas of contamination harmful to human health [Cervone et al, 2008]. Once injected in the atmosphere, the aerosol particles have varying resident times, which are affected by the size of the particles, and weather condition such as precipitation and wind patterns [Albriet et al, 2010].…”