CONNECTIONIST MODELLING FOR ANTHROPOGENIC GREENHOUSE GASES (GHG) EMISSIONS IN URBAN ENVIRONMENTS

Global warming induced by greenhouse gases (GHGs) is already a reality and will continue to increase resulting in a severe climate change. The aim of the paper is twofold. First, to investigate the GHGs emissions between the year of 2004 and 2016 in four major urban cities, representing the residential band of Kuwait. Results showed a clear steady yearly increase in GHGs emissions, with more emissions in summer compared to winter, possibly due to the high consumption rate of fossil fuel for cooling purposes and traffic activities. Results also revealed a diurnal variation in GHGs emissions, plausibly attributed to the combined effects of busy traffic hours as well as respiration by the living organisms and/or from soils. A second objective in this paper is, to develop a reliable connectionist models such as neural networks for predicting GHGs emissions. Radial basis function (RBF) network due to its known approximation capabilities, localization of its transfer functions and its efficient training algorithms, showed a superior performance in predicting GHGs emissions. Parity and time series plots of the predicted concentrations against the observed concentrations demonstrated the appropriateness of connectionist modelling as a fast and precise tool for monitoring and forecasting the GHGs emissions.