Prediction and spatiotemporal analysis of atmospheric Fine Particles and their effect on temperature and vegetation cover in Iran: Using Exponential Smoothing approach in Python
Document Type : Research Paper
Authors
1 PhD candidate, Department of Civil-Environmental Engineering, School of Environment, University of Tehran, Tehran, Iran
2 Associate professor, Department of Environmental Design Engineering, School of Environment, University of Tehran, Tehran, Iran
3 PhD in Environmental Engineering, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran Head of the General Department of Environment and Sustainable Development, Deputy of Municipal Service
Abstract
A relatively accurate prediction of the concentration of pollutants and environmental variables, short-term and long-term, is an important step in reducing the damage caused by poor air quality. In this article, first by spatiotemporal analysis and then by using prediction techniques, to predict the concentration of fine atmospheric particles (PM2.5), temperature and vegetation cover index (NDVI) on the trend of PM2.5 in a period of 5 years (2017-2022) was discussed at the level of Iran. The data of PM2.5 concentration, temperature and vegetation index were extracted based on MERRA-2, FLDAS and MODIS satellite models. In the five-year study period, a somewhat downward trend was observed for the air concentration of PM2.5. The results showed the lowest annual average concentration of fine atmospheric particles in 2019 and 2020. Also, a strong correlation between PM2.5 concentration and temperature was obtained. The highest average concentration of PM2.5 occurred in the northwest, west, and southwest of Iran. In the next step, in order to predict the future concentration of PM2.5 air particles, temperature and vegetation index, the Exponential Smoothing approach was used in the Python statistical library (Statsmodels) to model monthly time series. Evaluation of the models with two criteria of root mean square error (RMSE) and coefficient of determination (R2) was done in order to minimize the estimation error and find the most suitable model among the eleven predicted models. The results show that double exponential smoothing models are more suitable for predicting PM2.5 concentration and triple exponential smoothing models with Holt-Winter trend are more suitable for predicting temperature and NDVI data. This study can help public and private institutions to better understand economic, health and environmental condition affected by air pollution effects by predicting the period when air pollution levels may be particularly high.
Keywords
)