Estimation of PM_2.5 Concentration in DKI Jakarta from Sentinel-5P Imagery by Considering Meteorological Factors Using Random Forest Approach

Poor air quality, caused by high pollutant levels, harms the environment and public health. Fine particulate matter (PM 2.5 ), less than 2.5 μm in diameter, is a major concern in air quality observations and is a major concern due to its ability to penetrate the respiratory system, increasing risks of lung cancer, premature death, and unnatural births. Jakarta faces severe air pollution, yet its air quality monitoring network remains limited. To address this, this study employs machine learning, specifically random forest algorithms, using spatial regression to model PM 2.5 levels. The variables used are meteorological elements and particulates and gasses obtained by utilizing remote sensing. It was found that the R 2 value of 0.793 implies that the accuracy of the variables used reaches 79.3 percent and the RMSE value of 8.28 μg/m3. The spatial pattern formed in this spatial modelling follows the pattern of the rainy season and dry season, where the highest value of the spatial pattern of the PM 2.5 parameter is in the JJA month (June, July and August), and finally at the lowest value in the DJF month (December, January and February).