Multiple linear regression was used to fit two models to the daily average concentration of particulate matter with diameter lower than 10 µm (PM10). The explanatory variables in the first model (M1) were the weather variables (air temperature, relative humidity, rainfall, wind speed and atmospheric pressure) and wind direction index (WDI). The second model (M2) used the same variables as the M1 model plus the concentration of PM10 in the previous day (PM10,i-1). The stepwise technique was used for the selection of the explanatory variable. Measurements of PM10 concentration were made between 05/01/2002 and 08/31/2003 in the city of Rio de Janeiro, RJ, Brazil. The regression coefficient (r²) for the models fitting was satisfactory, with better results for model M2 (r² = 0.557) compared to model M1 (r² = 0.334). The weather variables presented negative correlation with PM10, with the exception of the wind direction index, which, similarly to PM10, i-1, had positive correlation. The air relative humidity and the rainfall were the most significant weather variables in the models. However, PM10, i-1 was the most significant variable, when included in the model. The air temperature was statistically not significant (p > 0.05) for both models. M2 model showed an agreement between the estimated and observed values and a better precision than M1 model. In terms of air quality forecast, both models presented satisfactory results, but model M2 was superior.
Particulate matter; linear regression model; urban pollution