19 Predicting daily PM2.5 in Mexico City: A hybrid spatiotemporal modeling approach

• DOI: 10.1017/cts.2024.710
• Publication Date: 2025-03-26T03:58:20Z
• Authors: Mike He; Ellen Ren; Iván Gutiérrez-Avila; Itai Kloog
• Keywords: R; Medicine

Objectives/Goals: In recent years, there has been growing interest in the development of air pollution prediction models, particularly low- and middle-income countries that are disproportionately impacted by effects pollution. Recent methodological advancements, machine learning, provide novel opportunities for modeling efforts. Methods/Study Population: We estimate daily ground-level fine particulate matter (PM2.5) concentrations Mexico City Metropolitan Area at 1-km2 grids from 2005 to 2023 using a multistage approach. Spatial temporal predictor variables include data moderate resolution imaging spectroradiometer (MODIS), Copernicus Atmosphere Monitoring Service (CAMS), additional meteorological land use variables. employed machine-learning-based approaches (random forest gradient boosting algorithms) downscale satellite measurements incorporate local sources, then utilized generalized additive model (GAM) geographically weight predictions initial models. Model performance was evaluated 10-fold cross-validation. Results/Anticipated Results: On average, random forest, boosting, GAM models explained 75, 82, 83% variations measured PM2.5 concentrations. levels were generally higher densely populated urban centers lower suburban rural areas. Important predictors included wind (both u v components), 2-meter mean temperature, elevation, normalized difference vegetation index (NDVI). Discussion/Significance Impact: Using learning-based approaches, we developed robust with fine-scale spatial (1-km2) (daily) 2023. The predicted can further advance public health research on beyond.