A statistical model for predicting carbon monoxide levels
This paper presents a statistical model that is able to predict carbon monoxide (CO) concentrations as a function of meteorological conditions and various air quality parameters. The experimental work was conducted in an urban atmosphere, where the emissions from cars are prevalent. A mobile air pollution monitoring laboratory was used to collect data, which were divided into two groups: a development group and a testing group. Only the development dataset was used for developing the model. The model was determined by using a stepwise multiple regression modelling procedure. Thirteen independent variables were selected as inputs: non-methane hydrocarbon (NMHC), methane (CH4), suspended dust, carbon dioxide (CO2), nitrogen oxide (NO), nitrogen dioxide (NO2), sulfur dioxide (SO2), ozone (O3), wind speed, wind direction, temperature, relative humidity and solar energy. It was found that NO has the most effect on the predicted CO concentration. The contribution of NO to the CO concentration variations was 91.3%. Adding in the terms for NO2), NMHC and CH4 improved the model by only a further 2.3%. The derived model was shown to be statistically significant, and model predictions and experimental observations were shown to be consistent.
Keywords: carbon monoxide, correlations, regression models, statistical analysis, urban atmosphere, vehicular emissions