CO Emissions Modeling and Prediction using ANN and GIS

Air pollution is considered a global concern due to its impacts on human life and the urban environment. Therefore, precise modeling techniques are necessary to predict air quality in congested areas such as megacities. Recently, machine learning algorithms such as Neural Networks show significant possibilities in air quality studies. This paper proposes a model to estimate air quality in a congested urban area in Baghdad city using Artificial Neural Network (ANN) algorithm and Geospatial Information System (GIS) techniques. Carbon Monoxide (CO) gas is selected as the main air pollutant. The model parameters involve; CO samples, traffic flow, weather data, and land use information collected in the field. The proposed model is implemented in Matlab environment and the results are processed after entering ArcGIS software. Using its spatial analysis tools, the outputs are presented as a map. The final findings indicate the highest value of CO emissions that reached 34 ppm during the daytime. The most polluted areas are located near congested roads and industrial locations in comparison with residential areas. The proposed model is validated by using actual values that are collected from the field, where the model's accuracy is 79%. The proposed model showed feasibility and applicability in a congested urban area due to the integration between the machine learning algorithm and GIS modeling. Therefore, the proposed model in this research can be used as a supportive model for decision making of city managers.