Dynamic simulation of airborne pollutant concentrations associated with the effect of climate change in Batu Muda region, Malaysia

• Published in: Modeling earth systems and environment 2022-03, Vol.8 (1), p.323-338
• Main Authors: Lee, Cedric Lemuel, Jie, Wong Yong, Arumugasamy, Senthil Kumar
• Format: Article
• Language: English
• Subjects: Air monitoring; Air pollution; Air quality; Air temperature; Airborne equipment; Airborne sensing; Algorithms; Ambient temperature; Artificial neural networks; Bayesian analysis; Carbon monoxide; Chemistry and Earth Sciences; Climate change; Climate effects; Computer Science; Computer simulation; Dormancy; Earth and Environmental Science; Earth Sciences; Earth System Sciences; Ecosystems; Environment; Humidity; Indoor air pollution; Math. Appl. in Environmental Science; Mathematical Applications in the Physical Sciences; Mathematical models; Meteorological parameters; Monitoring systems; Neural networks; Nitrogen oxides; Original Article; Outdoor air quality; Ozone; Parameters; Particulate emissions; Particulate matter; Photochemical reactions; Photochemicals; Photochemistry; Physics; Pollutants; Pollution monitoring; Probability theory; Public health; Regularization; Relative humidity; Statistics for Engineering; Sulfur; Sulfur dioxide; Sulphur; Sulphur dioxide; Suspended particulate matter; Training; Wind; Wind direction; Wind effects; Wind speed
• ISSN: 2363-6203; 2363-6211
• DOI: 10.1007/s40808-021-01107-6

Air pollution has been a rising concern of the 21st due to its effects to public health. Air Monitoring Stations are state-of-the-art equipment used to measure airborne pollutants concentration i.e. carbon monoxide, nitrogen oxide, sulphur dioxide, particulate matter (PM 10 ) and ozone (O 3 ), as well as the meteorological parameters (i.e. ambient air temperature, relative humidity, wind speed and wind direction). Effects of climate change will affect the ambient temperature and humidity, which may induce a direct effect on air quality. In light of this, feed forward artificial neural network was employed to simulate the dynamic variations of PM 10 and O 3 with relative humidity, temperature, and windspeed data being the inputs under 12 different training algorithms. Based on the results obtained, Bayesian regularization with 12 hidden neurons is the optimized network structure, with mean absolute percentage error in testing dataset of O 3 and PM 10 at 51.31% and 36.49%, respectively. The models performed better in O 3 prediction as it is a photochemical reaction where ozone concentration varies according to temperature, the effect of meteorological parameters is significant. On the other hand, PM 10 is not heavily dependent on meteorological parameters as the diversity of particulate matter components where most of its sources are dormant to changes in climate.