Air Pollution Forecasting for Ankara with Machine Learning Method


Creative Commons License

Ünal Z. F. , Dinç U., Özen C., Toros H.

Journal of Research in Atmospheric Science, vol.1, no.1, pp.42-48, 2019 (Refereed Journals of Other Institutions)

  • Publication Type: Article / Article
  • Volume: 1 Issue: 1
  • Publication Date: 2019
  • Title of Journal : Journal of Research in Atmospheric Science
  • Page Numbers: pp.42-48

Abstract

Environmental pollution problems have become more serious due to perceivable global climate change effects in the 21st century. Air pollution is one of the major pollution problems like water pollution, soil pollution. To protect people from fatal health diseases and extreme weather events as effects of air pollution, many countries have been trying to reduce their own emissions in keeping with agreements such as Paris Agreement, Kyoto Protocols. It is thought that remarking air pollution problems for Ankara in this study would be so essential for Turkey due to its critical daily air quality index values and also creating enlarging public awareness about air pollution. In this study, parameters like wind, pressure, temperature, humidity and precipitation are used for changeable air pollution parameter values like PM10, SO2 and NO2. WRF model was operated for 2017 and 2018 years by using global NOAA’s GDAS data with 9 km resolution. Air pollution measurement data from the Ministry of Environment and Urbanization for six points in Ankara and meteorological variables obtained from the WRF model were used to predict air quality for Ankara. To train about the relationship between air pollution and meteorological variables, the machine learning package H2O was used in the R software. The educated model was tested with last one month data of 2018 and for the final step, the model performance rates and errors were obtained by RMSE (Root Mean Square Error), MAE (Mean Absolute Error) and correlation values..