Forecasting Particulate Matter Concentration Using Linear and Non-Linear Approaches for Air Quality Decision Support

Abstract

Air quality status on the East Coast of Peninsular Malaysia is dominated by Particulate Matter (PM10) throughout the years. Studies have affirmed that PM10 influence human health and the environment. Therefore, precise forecasting algorithms are urgently needed to determine the PM10 status for mitigation plan and early warning purposes. This study investigates the forecasting performance of a linear (Multiple Linear Regression) and two non-linear models (Multi-Layer Perceptron and Radial Basis Function) utilizing meteorological and gaseous pollutants variables as input parameters from the year 2000–2014 at four sites with different surrounding activities of urban, sub-urban and rural areas. Non-linear model (Radial Basis Function) outperforms the linear model with the error reduced by 78.9% (urban), 32.1% (sub-urban) and 39.8% (rural). Association between PM10 and its contributing factors are complex and non-linear in nature, best captured by an Artificial Neural Network, which generates more accurate PM10 compared to the linear model. The results are robust enough for precise next day forecasting of PM10 concentration on the East Coast of Peninsular Malaysia.

Development of Multiple Linear Regression for Particulate Matter (PM10) Forecasting during Episodic Transboundary Haze Event in Malaysia

“…The RMSE, NAE, and PA components were chosen in this study to measure the performance of the models. RMSE and NAE are considered as suitable indexes for determining model accuracy in which the model is noted as having a high accuracy when their values are close to zero, while the PA value is nearest to 1 [32,50,51]. According to a previous study [32], a comparison of the best statistical PM 10 forecasting methods with the lowest values of RMSE and NAE and the highest PA value has been conducted to select the best-fit prediction model.…”

“…The models were evaluated based on the model's error and accuracy by using several performance indicators, namely Root Mean Square Error (RMSE), Normalized Absolute Error (NAE), and Prediction of Accuracy (PA). The best-fitted model is chosen when it has high accuracy in which the PA is closer to 1 while the minimal error (i.e., RMSE and NAE) is close to 0 [32,35,40]. Equations (6)- (8) show the performance indicators' formula used in this study.…”

“…The fundamental basis of this model represents the relationship between the dependent variable and several independent variables, such as meteorological factors and gaseous pollutants, by uncomplicated computation and easy implementation [30,31]. Previous studies in Malaysia have developed the MLR model to predict PM 10 concentration, specifically in the East Coast Peninsular Malaysia, based on different site classifications of rural, suburban, urban, and industrial areas and during different types of monsoon to determine its variation during non-haze periods [27,32,33]. Moreover, [34] used the MLR model for PM 10 forecasting during haze events that hit Malaysia based on the data from the affected Air Quality Monitoring Station (AQMS).…”

Application and Comparison of MLR, ANN and CART Models for Predicting PM10 Concentration Level of Guwahati City (India)

“…In MLR, the coefficient of determination (R 2 ) indicates the overall capability of the model to handle variance in data. The regression model was composed following equation 2 [42,43].…”

“…MLR assumes that the residuals have a normal distribution with a zero mean, uncorrelated and constant variance. The stepwise multiple linear regression procedure was used here to derive the mathematical equation [43]. Variance inflation (VIF) was used in this study to evaluate the multicollinearity effect on the variance of the estimated regression coefficient.…”