AI-Driven Public Health Surveillance: Analyzing Vulnerable Areas in Brazil Using Remote Sensing and Socioeconomic Data

Abstract

Urban vulnerability assessment is essential for identifying deprived areas and associated public health risks, especially under the accelerating effects of climate change and pollution. This paper introduces a novel AI-driven methodology that integrates remote sensing imagery, socioeconomic data, and machine learning models to analyze urban vulnerability in Brazil. Using data sources including Sentinel-2, Sentinel-5P, IBGE indicators, and OpenStreetMap, we develop a vulnerability index that captures environmental and infrastructural conditions. The model is further applied to predict pollution indices such as SO₂, NO₂, O₃, and CO using regression methods including Random Forest, XGBoost, and Linear Regression. Our results reveal that favelas exhibit significantly higher vulnerability scores, though non-slum areas show considerable heterogeneity. The proposed framework demonstrates high performance in pollution prediction and offers a valuable decision-support tool for public health and urban planning. We also release an open-access dataset to support further research in this area.