From vulnerability mapping to action: artificial intelligence and geospatial methods for urban heat resilience
More details
Hide details
1
Wits Planetary Health Research, University of the Witwatersrand, Johannesburg, South Africa
2
Department of Informatics, University of Oslo, Oslo, Norway
3
DHIS2 Climate and Health, University of Oslo, Oslo, Norway
4
IBM Research Africa, IBM, Johannesburg, South Africa
5
Swiss Tropical and Public Health Institute, University of Basel, Basel, Switzerland
6
Climate System Analysis Group, University of Cape Town, Cape Town, South Africa
Popul. Med. 2026;8(Supplement Supplement 1):A404
ABSTRACT
INTRODUCTION:
African cities face escalating heat exposure, yet fine-scale heat-associated vulnerability data to guide public health interventions remains scarce. The HE2AT Centre aims to reduce heat-related morbidity and mortality across African cities by generating evidence, building capacity, and translating research into actionable public health tools. This workshop explores how heat vulnerability mapping could translate into operational surveillance systems through the DHIS2 Climate & Health platform, an open-source health management information system currently deployed across seven African countries.
METHODS:
We integrated high-resolution satellite data (land surface temperature, vegetation indices), ERA5 climate reanalysis, and GCRO socio-economic variables to create ward-level heat vulnerability indices for 135 Johannesburg wards. Machine learning models with SHAP interpretability methods identified key predictors and threshold effects. Multi-scale spatiotemporal data harmonisation protocols enabled integration across disparate data sources. We developed methods to design vulnerability data layers for integration into DHIS2 to support real-time surveillance.
RESULTS:
Johannesburg analyses revealed extreme intra-urban heterogeneity, with vulnerability concentrated in informal settlements with limited vegetation and healthcare access. Machine learning identified non-linear exposure-response relationships, with SHAP values highlighting critical temperature thresholds. The heat vulnerability index demonstrates how such analyses can identify priority areas for intervention. DHIS2 Climate & Health, deployed across Niger, Senegal, Nigeria, Guinea Bissau, Mozambique, Malawi, Ghana, and DRC, provides a platform for linking vulnerability hotspots to health facilities.
CONCLUSIONS:
This interactive workshop equips participants with practical geospatial and machine learning methods for heat vulnerability assessment and pathways for integration into operational health surveillance systems. Participants gain hands-on experience with DHIS2 Climate tools and connection to a pan-African network advancing climate health equity. We demonstrate the power of collaboration between researchers, technology developers, and public health practitioners across international scales.