Modelling ozone levels in an arid region - a dynamically evolving soft computing approach

Abstract

The primary pollutants may contribute to the increase of ozone levels in the arid regions. Complex interactions between the pollutants and the meteorological variables make the study of this phenomenon more exigent. The dynamically evolving neural fuzzy inference system (DENFIS), as an example of soft computing models, allows the online evolution of both the knowledge and the inference mechanism. It is suitable for real-time applications in producing fairly reliable forecasts. The proposed DENFIS model for two sites in the Empty Quarter (Rub Al-Khali Desert) of Saudi Arabia was developed using the meteorological data collected during the winter and the summer seasons, and the transformed meteorological data. The concentrations of nitrogen oxide (NOx) and their transformations were incorporated as additional inputs for model performance analyses. The mean absolute percentage errors of the model vary from 9.52% to 11.84% with discretion and appreciation of the limitations of the overall model predictions and its performance analyses indicate the viability of application of the adopted online DENFIS modelling approach in short-term modelling of zone levels in arid regions.