Comparison of LUR-based and ANN-based PM2.5 concentration estimation over Taipei metropolis

Fine particulate matter (PM2.5) is an air pollutant that has been becoming one of the major environmental issues in national governments. Air quality monitoring and prediction are thus necessary for management and control. In previous studies, a land-use regression (LUR) model with several factors such as chemical particles, meteorological information, greenness environments, and landmarks combined with interpolation techniques is used to predict PM2.5 concentrations using data from Taipei metropolis, which exhibits typical Asian city characteristics. Recently, a lot of attention was paid to the improvement of methods which are used to predict air quality especially PM2.5. This study proposes utilizing artificial neural networks to predict PM2.5 concentrations and the built PM2.5 prediction model is compared with that using LUR. To obtain the resulted, cross-validation is adopted in the proposed method. 17 air quality monitoring stations established by environmental protection administration of Taiwan with annual average PM2.5 concentrations from 2006-2012 were used for model development. In experiments, quantitative accuracy assessments were conducted to evaluate the performance of proposed methods, in term of determination coefficients (R²) and root means square error (RMSE), compared with LUR-based method. The result show that LUR-based still perform better than ANN-based. The R² of LUR-based was 0.9 while the R² ANN-based was 0.8.