Dust Pollution

This project models air pollution dynamics across space and time using machine learning to extend insights beyond sparse monitoring stations. We combine pollutant observations with meteorology, land-use/source proxies, terrain context, and (when available) remote sensing features. Time-aware features capture daily to seasonal variability, while evaluation tests generalization across time windows and regions. Outputs include spatiotemporal exposure maps and hotspot detection to support environmental management and public health.

Main focus

• Predict pollutant concentrations across space and time using ML.
• Improve exposure estimation beyond station networks.
• Identify hotspots and conditions linked to pollution peaks.

Objectives

1. Integrate monitoring data with meteorology, land-use, and remote sensing covariates.
2. Train spatiotemporal ML models and evaluate across time/space generalization.
3. Produce exposure maps and hotspot summaries to guide mitigation and monitoring.

Graphical abstract

:::: {.row .justify-content-sm-center} ::: {.col-sm-10 .mt-3 .mt-md-0}

::: ::::

::: caption Graphical abstract summarizing the workflow and key results. :::

Papers (selected)

Paper 1 — Seasonal Drivers of Dust Pollution Using Game Theory

DOI

Paper 2 — Dust Predictions: Accuracy, Uncertainty, and Interpretability

DOI

Paper 3 — Long-Term Air Pollution Effects on Mortality and Morbidity

DOI