iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Google Earth AI is a collection of geospatial models and datasets that public-health teams can use to estimate where health needs or disease risk may emerge. Its Population Dynamics Foundation Model (PDFM) creates location-level embeddings from aggregated signals such as search-topic frequencies, place activity, weather and air quality. Those embeddings can enrich a separate epidemiological or statistical model; they are not diagnoses, individual health records or an autonomous outbreak-prevention system.
Google Research reported partner evaluations spanning vaccination coverage, mortality, dengue, postpartum-depression risk and cholera in an overview dated October 6, 2026. The results vary by task and place, and do not establish a universal performance advantage. Google also describes separate research prototypes used in an Ebola response. Access to the PDFM embeddings is a distinct offering, available in Preview as Population Dynamics Insights.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Schneider's Introduction to Public Health | $61.28 | Buy on Amazon |
| 2 |
|
Public Health 101: Improving Community Health | $50.50 | Buy on Amazon |
| 3 |
|
Introduction to Public Health: . | $61.99 | Buy on Amazon |
| 4 |
|
Public Health: An Introduction to the Science and Practice of Population Health | $63.20 | Buy on Amazon |
| 5 |
|
Essentials of Epidemiology in Public Health | $109.95 | Buy on Amazon |
How Google Earth AI supports public-health analysis
Google describes Earth AI as a suite connecting geospatial information—including satellite imagery, weather, aggregated search trends, mobility and other population-dynamics signals—to models that can help characterize places. One component, PDFM, uses self-supervised learning to combine privacy-preserving signals into compact location embeddings that Google calls “fingerprints.” Google says those embeddings are refreshed monthly.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Signals described for PDFM include the frequency of aggregated searches about topics, local density and busyness around places such as pharmacies, clinics and parks, and high-resolution weather and air-quality measures. The resulting representation is a feature about a location, not a record about a named person. Researchers or public-health teams can add it to their own machine-learning workflow alongside local health records and conventional inputs.
#1 Best Overall
This distinction matters when interpreting results: PDFM is an input to a model, not a standalone clinical or epidemiological decision-maker. Google says it may complement health data that are sparse, delayed or constrained by administrative boundaries. A richer representation can help a model describe context, but it does not guarantee that the underlying data are complete, locally representative or suitable for a particular decision.
What partner evaluations reported
The figures below are findings Google Research reported on October 6, 2026. They concern different outcomes, geographies, units of analysis and baselines; they should not be combined into one overall accuracy claim.
Vaccination coverage across the U.S.–Canada border
For modeling measles, mumps and rubella (MMR) vaccination across 146 U.S. counties near Canada, the partner evaluation reported that adding Canadian context increased explained variance from 16% to 22%, a 36% relative gain. Google says the refined coverage estimates changed estimates by at least three percentage points for 4.7 million residents in the border region. This is a finding about cross-border coverage modeling, not evidence that the model increased vaccination or measured individual immunity.
Rank #2
- New case studies on folic acid and spina bifida, the opioid epidemic, athletic health injuries, mPox, and the patient safety movement.
- Interactive COVID-19 case studies revised and consolidated to take an historical perspective on the COVID-19 pandemic.
- Updated materials for each chapter including Healthy People 2030, pandemic prevention, new vaccines, environmental justice and climate change, and academic health departments.
- New Boxes in each chapter including misinformation/disinformation, suicide, health advocacy, national dietary guidelines, and open notes.
- Improving Population Health interactive exercises retained from 3rd edition
Cardiovascular mortality nowcasting in the contiguous United States
Across 3,091 contiguous-U.S. counties, PDFM-based estimates had a mean absolute error of 18.7 deaths per county, compared with 19.1 for census-based models. The reported root mean square error was 46.00 versus 57.69, respectively. Google says the differences were not statistically significant. The result suggests a possible timely alternative when census data are outdated or unavailable; it does not establish a statistically significant accuracy advantage.
Dengue forecasting in Mexico
For forecasting dengue across roughly 2,450 Mexican municipalities, integrating TimesFM with PDFM produced a statistically significant one-month improvement in weighted interval score: a difference of −0.0051, with lower scores indicating better forecasts. Accuracy improved in up to 72% of municipalities with active transmission. These are short-horizon forecast results for that evaluation, not a guarantee that a municipality will experience an outbreak or that the same gain will hold elsewhere.
Postpartum-depression risk prediction in the United States
In a model using responses from 332,970 U.S. CDC PRAMS respondents, the reported area-under-the-curve (AUC) gain was 0.0020 over a 0.62 baseline in states represented in training, and 0.0038 in states not represented in training. The overview also says area poverty was encoded with an R² of 0.45 and that the embedding recovered about 15% of the predictive signal from income and insurance records. These results concern model discrimination and area-level context; they do not replace individual socioeconomic information, clinical screening or follow-up, and they do not show that screening gaps were resolved.
Rank #3
- Coverage of pressing issues in public health, including: the opioid crisis; e-cigarettes and vaping; West Nile, Zika and other emerging viruses; impact of recent marijuana laws and drugged driving; the resurgence of measles and anti-vaccination sentiment; the Flint water crisis; and much more.
- Updated statistics and information in every chapter to illustrate the current state of public health.
- The latest on the Affordable Care Act (ACA) and the current state of health insurance in the United States.
- Coverage throughout illustrating the influence of politics on public health policy today.
- Navigate Advantage Access (included with each print text) includes a comprehensive and interactive eBook, student practice activities and assessments, learning analytics, reporting tools and more.
Cholera emergence in the Democratic Republic of the Congo
For 403 health zones in the Democratic Republic of the Congo (DRC) over 89 weeks, Google Research reported statistically significant relative gains of 9.7% in area under the precision-recall curve at a four-week horizon and 18.1% in Precision@5 at an eight-week horizon. In endemic zones, the Precision@5 gain was 19.3%. Precision@5 measures how often the five zones a model ranks as highest risk in a given week later have an outbreak. It is a ranking measure, not a statement that all flagged zones will have cholera.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteHow Google describes the Ebola-response prototypes
Google’s October 6, 2026 account describes work with WHO AFRO and DRC INRB during an Ebola response. A prototype called the Geospatial Reasoning agent mapped 48 exposed settlements and more than 45,500 people considered at risk along remote mining corridors in minutes. Google says a conventional process would normally take weeks. The account also describes weekly spread-risk forecasts for uninfected zones using mobility flows, historical case trends and Earth AI models.
These are operational claims made by Google, not independently verified performance findings in the information available here. The agent and a planetary prediction engine are described as research prototypes, not generally available products. A mapped exposed settlement or a high-risk zone is not the same thing as a confirmed infection; such outputs can inform investigation and planning, but require field information and public-health judgment.
Earlier examples use different studies and measures
Google’s March 13, 2026 overview described additional projects: clinic-utilization prediction in Malawi combining PDFM embeddings with satellite-derived location features and local data; measles vaccination-coverage estimates at ZIP-code level; dengue forecasting in Brazil; and an Australia proof of concept for chronic-disease needs. It also described cholera forecasting that combined TimesFM, PDFM and weather data, reporting an accuracy improvement of more than 35% against standard models.
Those examples come from separate projects and do not share one baseline or evaluation design with the October 2026 findings above. In particular, the reported “over 35%” cholera result should not be substituted for the later DRC cholera metrics or treated as a universal improvement across diseases.
What the findings do—and do not—establish
The evaluations illustrate several distinct public-health tasks. Nowcasting estimates a current condition when conventional data may lag; short-horizon forecasting estimates future risk; risk stratification identifies places or groups for closer attention; and mapping helps teams locate potentially exposed settlements or service needs. Each task has a different target and decision context, so a strong result on one metric cannot be read as evidence of performance on another.
Best Value
- Check the decision and unit. Determine whether a result concerns a county, municipality, health zone, respondent group or settlement, and whether it supports planning, follow-up or another specific action.
- Read the baseline and metric together. Explained variance, MAE, RMSE, AUC, weighted interval score and Precision@5 measure different properties. A lower error, higher ranking precision or small AUC gain does not by itself quantify health impact.
- Validate locally. Patterns in search activity, mobility, facility use, weather and satellite-derived information may not represent every population equally. Teams need to test whether inputs and predictions remain useful in the geography and time period where they intend to act.
- Keep human accountability. Local health authorities must decide how to interpret a prediction, verify it against available evidence and allocate scarce staff or services. The model output should supplement existing public-health information rather than replace it.
- Assess governance and equity. WHO’s guidance on AI for health emphasizes evidence-based adoption, ethics, trust, safety, equity and sustainable implementation. Relevant questions include who is represented in the data, who can challenge a decision, how errors are handled, and whether local institutions have the capacity to use the tool.
The available Google accounts do not independently establish performance across all deployment conditions, fairness across populations, privacy architecture beyond Google’s characterization of the signals as privacy-preserving, or whether earlier predictions produced measured health outcomes. Those are separate questions from whether a model performed well on a specified evaluation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Who can access Population Dynamics Insights?
As of Google’s October 6, 2026 announcement, PDFM embeddings are commercially available in Preview as Population Dynamics Insights, a geospatial embeddings dataset from Google Maps Platform. Google says researchers can request no-cost access for select use cases, and eligible organizations can apply for Google Earth credits through Google Maps Platform Public Programs.
Preview access is not the same as general availability of the Geospatial Reasoning agent or planetary prediction engine; Google describes those as research prototypes. The announcement does not specify universal eligibility, prices, service commitments or processing timelines. Organizations considering access should confirm current terms and eligibility directly with Google.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchA practical way to evaluate a geospatial AI project
- Define the public-health decision. State the target outcome, geography, time horizon and action a team might take if risk is elevated. A prediction with no plausible response pathway may have little operational value.
- Set an appropriate baseline. Compare the model with the data and method actually available to the local team, such as existing surveillance, census-based estimates or established forecasting practice. Document what data are current and what data arrive late.
- Test the specific population and period. Evaluate across places and time periods beyond the training data where possible. Examine missed events as well as false alarms, and check whether performance differs among communities.
- Plan interpretation and response before deployment. Decide who reviews alerts, how they are checked against local knowledge, how uncertainty is communicated and what resources can be mobilized.
- Monitor and govern use. Track data changes, model drift, operational effects and unintended distributional consequences. Establish accountability for decisions and a process to revise or stop use if the system no longer serves the intended purpose.
WHO Director-General Tedros Adhanom Ghebreyesus has said: “AI is already playing a role in diagnosis and clinical care, drug development, disease surveillance, outbreak response, and health systems management … The future of healthcare is digital, and we must do what we can to promote universal access to these innovations and prevent them from becoming another driver for inequity.” That principle applies to geospatial prediction as much as to other health technologies: technical performance matters, but so do evidence, access, oversight and the capacity to act.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
