Application of Variable Selection Method (Stepwise Regression Analysis) to Forecast the PM₁₀ Concentrations in Agadir City (Morocco)

Many countries are preoccupied with the high concentrations of particulate matter PM10, due to the rapid industrialization which can be harmful to people's health and the environment. The capacity to predict these particulate matters in advance allows for better planning to avoid and decrease their impact. This is possible and attainable if we have a good grasp of the contaminants present in the air and the factors that determine their patterns. A great deal of research indicates that pollutant pattern is influenced by meteorological parameters such as wind speed (WS), relative humidity (H), and temperature (T). Statistically, the correlation between these parameters is significant in most cases. The present study aims to use the stepwise regression method for forecasting ambient PM10 in the city of Agadir, Morocco, based on the data collected from a fixed station placed at a school, in Agadir city center during two years (2009-2010). The goodness of fit, which implies a comparison of the observed and the predicted values, is studied through performance indicators using the Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and coefficient of determination (R2). In the face of the difficulty in synthesizing and interpreting certain scores due to the diversity of the parameters used in their computations, and the fact that the scores do not all point in the same direction, we were able to come up with satisfactory results in the evaluation of the used model <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(R2 = 0.64,RMSE = 9.224,MSE = 85.088,MAE = 6.687)</tex> . This shows the ability of the stepwise regression model to predict the next day PM10 concentration.