Prediction of PM2.5 concentrations several hours in advance using neural networks in Santiago, Chile

Hourly average concentrations of PM2.5 have been measured at a fixed point in the downtown area of Santiago, Chile. We have focused our attention on data for the months that register higher values, from May to September, on years 1994 and 1995. We show that it is possible to predict concentrations at any hour of the day, by fitting a function of the 24 hourly average concentrations measured on the previous day. We have compared the predictions produced by three different methods: multilayer neural networks, linear regression and persistence. Overall, the neural network gives the best results. Prediction errors go from 30% for early hours to 60% for late hours. In order to improve predictions, the effect of noise reduction, rearrangement of the data and explicit consideration of meteorological variables are discussed.

[1]  F. Valero,et al.  STATISTICAL FORECAST MODELS FOR DAILY AIR PARTICULATE IRON AND LEAD CONCENTRATIONS FOR MADRID, SPAIN , 1992 .

[2]  Aldo Cipriano,et al.  Forecasting ozone daily maximum levels at santiago, chile , 1998 .

[3]  Marija Zlata Boznar,et al.  A neural network-based method for short-term predictions of ambient SO2 concentrations in highly polluted industrial areas of complex terrain , 1993 .

[4]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[5]  R. Burnett,et al.  Effects of particulate and gaseous air pollution on cardiorespiratory hospitalizations. , 1999, Archives of environmental health.

[6]  L. Tsimring,et al.  The analysis of observed chaotic data in physical systems , 1993 .

[7]  R. Garreaud,et al.  Meteorological air pollution potential for Santiago, Chile: Towards an objective episode forecasting , 1995, Environmental monitoring and assessment.

[8]  Roy M. Harrison,et al.  Sources and processes affecting concentrations of PM10 and PM2.5 particulate matter in Birmingham (U.K.) , 1997 .

[9]  Margarita Préndez,et al.  Correlation between solar radiation and total syspended particulate matter in Santiago, Chile—Preliminary results , 1995 .

[10]  M. W Gardner,et al.  Artificial neural networks (the multilayer perceptron)—a review of applications in the atmospheric sciences , 1998 .

[11]  D. Loomis,et al.  Mortality and ambient fine particles in southwest Mexico City, 1993-1995. , 1998, Environmental health perspectives.

[12]  Jose Torres-Jimenez,et al.  Short-term ozone forecasting by artificial neural networks , 1995 .

[13]  K. Spurny,et al.  On the physics, chemistry and toxicology of ultrafine anthropogenic, atmospheric aerosols (UAAA): new advances. , 1998, Toxicology Letters.

[14]  D. R. Kulkarni,et al.  SIMULATION OF CHARACTERISTICS AND ARTIFICIAL NEURAL NETWORK MODELING OF ELECTROENCEPHALOGRAPH TIME SERIES , 1997 .

[15]  Andreas S. Weigend,et al.  Time Series Prediction: Forecasting the Future and Understanding the Past , 1994 .