Google’s AI Agent Identifies Disaster and Environmental Risks—Analyzing Everything from Disease Outbreaks to Floods

 Google’s AI Agent Identifies Disaster and Environmental Risks—Analyzing Everything from Disease Outbreaks to Floods

Google is expanding the use of a "geospatial AI agent" capable of analyzing natural disasters, disease outbreaks, and environmental risks. The company has developed a system that integrates vast amounts of Earth-related data, geospatial AI models, and various digital tools to analyze complex geographical challenges.

According to information released by Google Research, this system—dubbed a "geospatial reasoning agent"—can simultaneously analyze data from satellite imagery, weather patterns, flood forecasts, population demographics, and maps. It selects the necessary data and tools based on a user's query and conducts a step-by-step analysis. As noted on the Google Research blog, the goal is to simplify complex geographical analyses that would be difficult for standard AI models to perform on their own.

Identifying Risks for 48 Settlements and Over 45,000 People in Minutes

Google reports that the World Health Organization’s (WHO) Regional Office for Africa tested this system during an Ebola outbreak in the Democratic Republic of the Congo.

The team integrated data on deforestation (derived from satellite imagery), the locations of mining sites, and population demographics to analyze high-risk areas. Google stated that the system identified 48 settlements and more than 45,500 people at risk—all within a matter of minutes. The company noted that, under normal circumstances, analyzing such a large volume of data could take weeks. Google has noted that this has helped in deploying mobile laboratories in affected areas and coordinating surveillance in border regions.

**Use in Natural Disaster Preparedness**

This AI agent has also been utilized to identify the potential impact of natural disasters. According to Google, an organization called ‘People in Need’ used this system to analyze the paths of 14 past cyclones, thereby identifying villages at higher risk.

To achieve this, the system integrated cyclone forecasts, population data, and satellite imagery. The organization reported that analysis which previously took days could now be completed in just a few minutes.

Similarly, ‘GiveDirectly’ used it to estimate the number of people affected by cyclones and tornadoes. It is expected to help in predicting how many individuals might be impacted in specific areas.

**From Crop Yields to Power Outage Forecasts**

Google’s ‘Planetary Prediction Engine’ aims to forecast various events by utilizing localized data. This includes data related to disease prevalence, crop yields, and disruptions to power systems.

The system creates machine learning models by combining available data with population and environmental indicators. It can then make predictions regarding risks, potential changes, and future conditions in a given area.

In health-related research released by Google, it was stated that combining population data with satellite indicators improved the prediction of health outcomes by 12 percent, compared to basic models based solely on demographic data. The company claims that this also helps increase the spatial resolution of the details shown on the map by up to eight times.

New feature in the experimental phase

Google has announced that it is expanding access to this technology by introducing an experimental feature called "Ask Google Earth," which allows for the analysis of complex geographical questions. Users can also incorporate their own data into this process.

Additionally, for developers, a mechanism is being rolled out—via an "agent-to-agent" API—that enables the integration of this system with their own software, data systems, and AI agents.

However, Google does not present this system as a fully autonomous technology capable of making final decisions. According to the company, its analytical outputs should be treated as preliminary results that require review by experts.