Forecasting PM10 in Algiers: efficacy of multilayer perceptron networks

Abstract

Air quality forecasting system has acquired high importance in atmospheric pollution due to its negative impacts on the environment and human health. The artificial neural network is one of the most common soft computing methods that can be pragmatic for carving such complex problem. In this paper, we used a multilayer perceptron neural network to forecast the daily averaged concentration of the respirable suspended particulates with aerodynamic diameter of not more than 10 μm (PM10) in Algiers, Algeria. The data for training and testing the network are based on the data sampled from 2002 to 2006 collected by SAMASAFIA network center at El Hamma station. The meteorological data, air temperature, relative humidity, and wind speed, are used as inputs network parameters in the formation of model. The training patterns used correspond to 41 days data. The performance of the developed models was evaluated on the basis index of agreement and other statistical parameters. It was seen that the overall performance of model with 15 neurons is better than the ones with 5 and 10 neurons. The results of multilayer network with as few as one hidden layer and 15 neurons were quite reasonable than the ones with 5 and 10 neurons. Finally, an error around 9% has been reached.

Numerical Model-Based Artificial Neural Network Model and Its Application for Quantifying Impact Factors of Urban Air Quality

“…ANN model slightly underestimates NO 2 and PM 10 under high concentrations. In general, the performance of ANN model developed in our study is statistically comparable with previous studies (Abderrahim et al, 2016;Fernando et al, 2012;Grivas and Chaloulakou, 2006;He et al, 2013;Hooyberghs et al, 2005;Liu et al, 2015;Yu et al, 2008). The good performance of ANN model is the base for quantifying the contribution of different factors to day-to-day variations of pollutant concentrations.…”

“…Previous studies revealed that the statistical performance of ANN air quality model is 0.6-0.9 for R (Abderrahim et al, 2016;Fernando et al, 2012;Grivas and Chaloulakou, 2006;He et al, 2013;Hooyberghs et al, 2005;Liu et al, 2015;Yu et al, 2008). The statistical performance of ANN model depends on selected predictors, pollutant categories, the backward concentration level, pollution characteristics (regional or local), time scale (hourly or daily), and ANN structures.…”

Evaluation of Key Parameters Using Deep Convolutional Neural Networks for Airborne Pollution (PM10) Prediction

“…e multilayer perceptron (MLP) architecture is one of the cases of study of this work because it is one of the most popular architectures today for PM10 forecast [13][14][15][16][17]. MLP usually consists of an input layer, one or two hidden layers, and an output layer.…”

Application of neural networks to the prediction of gas pollution of air

“…A number of works have been dedicated to air quality modelling. Some articles refer to prediction of particulate matter and gas pollution (Abderrahim et al, 2016;Kukkonen et al, 2003;Pawul and Śliwka, 2016). A review of the research in that field was published by Rybarczyk and Zalakeviciute (Rybarczyk and Zalakeviciute, 2018).…”