Using a land use regression model with machine learning to estimate ground level PM2.5

Ambient fine particulate matter (PM2.5) has been ranked as the sixth leading risk factor globally for death and disability. Modelling methods based on having access to a limited number of monitor stations are required capturing PM2.5 spatial temporal continuous variations with sufficient resolution. This study utilized land use regression (LUR) model machine learning assess spatial-temporal variability PM2.5. Daily average data was collected from 73 fixed air quality monitoring that belonged Taiwan EPA main island Taiwan. Nearly 280,000 observations 2006 2016 were used analysis. Several datasets determine predictor variables, including environmental resources dataset, meteorological land-use inventory, landmark digital road network map, terrain model, MODIS Normalized Difference Vegetation Index (NDVI) database, power plant distribution dataset. First, conventional LUR Hybrid Kriging-LUR identify important variables. Then, deep neural network, random forest, XGBoost algorithms fit prediction variables selected by models. Data splitting, 10-fold cross validation, external verification, seasonal-based county-based validation verify robustness developed The results demonstrated proposed models captured 58% 89% variations, respectively. When algorithm incorporated, explanatory increased 73% 94%, outperformed other integrated methods. demonstrates value combining an estimating exposures.