Artificial Neural Network Modelling to Predict PM2.5 and PM10 Exhaust Emissions from On-Road Vehicles in Addis Ababa, Ethiopia


Transport vehicles are the major sources of air pollution in the urban area. This study aims to investigate the level of roadside vehicularPM2.5 and PM10concentrations and their impact on urban air quality. In addition, artificial neural network model is used to predict the average 24 hours concentrations ofPM2.5 and PM10in the capital city of Ethiopia. For the prediction, the model uses relative humidity, temperature, wind speed, wind direction, traffic volume and data of concentrations ofPM2.5 and PM10collected from 15 different sites in city. This model trained, using Levenberg Marquardt and Scaled Conjugate Gradient Algorithm training functions, to define the finest fractional error between the measured and the predicted value. The performance of the model is determined using coefficient of correlation. It is found that the proposed model could predict exhaust emissions with an average coefficient correlation of 0.948 forPM2.5 and 0.959 for PM10. The results show that Levenberg Marquardt algorithm functions have a better coefficient of correlation and this could be considered as an alternative option to evaluate the exhaust emission concentration. The acquired results indicate that the above input data can be used to accurately predict the particulate matter concentrations in the city.