Google Earth AI Is Predicting Disease Outbreaks Before They Happen

Key Takeaways

• Google Earth AI improved cholera forecasting accuracy by over 35% compared to standard models using WHO Integrated Disease Surveillance Data
• The Population Dynamics Foundation Model (PDFM) maps vaccination coverage down to ZIP-code level using privacy-preserving, aggregated data
• University of Oxford researchers used Earth AI to significantly improve six-month dengue fever forecasts in Brazil
• Google has active deployments in Malawi, Brazil, Australia, and through the WHO Regional Office for Africa

A mapping platform built on decades of geospatial research is now forecasting where disease outbreaks will strike weeks before they peak. Google Earth AI, published March 13, 2026, combines population dynamics, weather modeling, and satellite intelligence to help public health officials move from reacting to crises to anticipating them. This breakdown covers each active deployment, the models powering them, and the verified outcomes recorded so far.

How Google Earth AI Builds Its Health Intelligence

Google Earth AI applies two core models to public health forecasting. The Population Dynamics Foundation Model (PDFM) captures how populations interact with their environments, including factors like weather, air quality, and flooding. Mobility AI provides additional context on how communities move through and engage with their surroundings.

These models combine environmental data with region-specific health information to produce forecasts that can run weeks or months ahead of a potential crisis. The system is authored by Yossi Matias, VP and GM of Google Research, and Michael Howell, Chief Health Officer at Google, signaling this is a senior-level, research-backed initiative rather than an experimental side project.

Also Read Google’s AI Finds Breast Cancers Doctors Missed. Here’s What the Data Reveals.

Cholera and Dengue Forecasting: Verified Outcomes

Weather drives the pace of many infectious diseases. Summer rains cause dengue fever spikes, while flooding significantly increases cholera outbreaks. Google Earth AI addresses both by pairing population dynamics with predictive weather models to generate health emergency forecasts.

In collaboration with the WHO Regional Office for Africa, Google evaluated a sub-national forecasting model using the WHO’s centralized Integrated Disease Surveillance Data. Combining TimesFM (Google’s time-series forecasting model) with PDFM and weather data improved cholera forecasting accuracy by over 35% compared to standard models. Better accuracy means public health officials can pre-position life-saving rehydration supplies before a crisis erupts rather than after.

At the University of Oxford, researchers applied Earth AI models and datasets to dengue forecasting in Brazil. Including PDFM embeddings significantly raised the predictive accuracy of six-month dengue forecasts, giving local authorities more lead time to deploy preventative measures.

Predicting Clinic Demand in Malawi

In Malawi, Google.org grantee Cooper/Smith combined Earth AI’s PDFM and AlphaEarth satellite embeddings with local health data to predict health service utilization at individual clinics. The system helps decision-makers spot early warning signs of disease outbreaks and allocate limited resources more efficiently before demand surges.

This deployment addresses a fundamental operational gap in low-resource health systems: administrators typically discover shortages only after patients arrive. Accurate forward-looking demand forecasts allow clinic-level planning adjustments while there is still time to act.

Mapping Vaccination Gaps at ZIP-Code Level

To address the rise of measles, researchers at Mount Sinai and Boston Children’s Hospital/Harvard used Earth AI’s PDFM to produce what they call “superresolution” estimates of vaccination coverage. The methodology relies entirely on privacy-preserving, aggregated data, meaning individual health records are never exposed.

The output maps vaccination rates down to the ZIP-code level and identifies localized clusters of undervaccination that align with recent outbreaks. National vaccination averages routinely obscure dangerous pockets of low coverage at the neighborhood scale. This granular mapping gives health agencies a precise targeting tool for outreach campaigns.

Chronic Disease Needs in Rural Australia

Google partnered with the Victor Chang Cardiac Research Institute, Wesfarmers Health, and Latrobe Health Services in Australia to deploy Population Health AI (PHAI). PHAI uses PDFM embeddings alongside air quality data, pollen data, and places insights to uncover the health needs of communities in rural Australia, with a focus on chronic disease needs and prevention.

PHAI is currently available as a proof-of-concept to select partners. The initiative reflects a deliberate expansion of Earth AI beyond infectious disease forecasting into non-communicable conditions, where environmental exposure over long periods shapes health outcomes.

Limitations Worth Noting

Earth AI’s effectiveness depends on the quality of local health data it is combined with. All current deployments are either pilot-stage or proof-of-concept, meaning broad rollout timelines have not been confirmed. The platform’s published impact figures reflect evaluated models in specific regional contexts, not globally generalized performance guarantees.

Also Read Perplexity Search API: Real-Time Web Retrieval That Outperforms Closed Search Systems

Frequently Asked Questions (FAQs)