Carbon Monoxide forecasting with artificial neural networks for Konya (Case Study of Meram)

Abstract

The first use of the term air quality problem, which emerged with the industrial revolution, date back to the 18^th and 19^th centuries. Natural causes such as forest fires and volcanic eruptions caused by air pollution, as well as the effect of increasing human activities on air quality with the industrial revolution, are more than natural effects. Consequences of air pollution; acid rains, climate change, respiratory diseases, occurrence of extreme weather conditions, decrease/increase in the number of species in the ecosystem. Especially in megacities, human health is closely affected due to wrong construction, heavy traffic and population density. For this reason, the preliminary forecast and model of air quality has an important place for possible health problems and global problems. In this study, Carbon Monoxide (CO, µg/m³) records of Meram district of Konya were modeled with three different Artificial Neural Networks (ANN) methods. These are Multilayer, Radial-Based and Generalized Regression ANN. Input parameters in modeling are air quality parameters such as; PM₁₀, SO₂, NO₂, NO_X and periodicity. CO is the output parameter. CO is quite harmful for human health; It is a colorless, odorless gas and is formed when the carbon in fuels is not fully combusted. When the comparison criteria are examined, it is seen that the best result is the input model of the Multilayer ANN model (RMSE=90.361, MAE=74.206, R²= 0.824).