Unraveling the Prediction of Fine Particulate Matter over Jaipur, India using Long Short-Term Memory Neural Network
Uday Pratap Singh, Vivek Saxena, Anil Kumar, Purushottam Lal Bhari, Deepika Saxena
Abstract
Fine particulate matter (PM2.5) is a perilous air pollutant for human health, especially when present at high airborne concentrations. The national clean air program (NCAP) aims at a 40 % reduction in particulate matter by 2026. Therefore, the prediction of PM2.5 has great significance in order to cope with challenges and taking counter majors to cope with air quality issues. The annual change in PM2.5 in Jaipur from 2018 to 2019 shows a sharp decreasing trend but from 2019 to 2020 gradual increasing trend is evident. On the other hand, a sharp increasing trend from 2020 to 2021 is evident while it again starts to fall between 2021 to 2022. The air quality of Jaipur reached a good level during the lockdown period as in all other parts of the country. But now it is reaching toward Hazardous level. As most natural systems exhibit nonlinearity, the prediction of PM2.5 in this study is accomplished by utilizing the Long Short-Term Memory (LSTM) model. The model employed in this study uses several LSTM layers along with dropout and dense layers. The result of the study shows that the model used in the study captures the internal dynamics of the underlying dynamical system, which governs PM2.5.