Foresight

Foresight uses machine learning and environmental data to simulate data center siting impacts, compare alternative locations, and support more transparent approval decisions.

Comment

Inspiration

Foresight came from the uneven distribution of environmental harm. Pollution, extreme heat, diesel emissions, industrial noise, and land-use disruption are not experienced equally. They concentrate in specific neighborhoods, often in communities already shaped by past infrastructure decisions.

As data centers expand, we saw the same pattern repeating. Sites are often chosen because land is cheap, power is nearby, and zoning is easier, but those factors do not always account for the environmental stress a community already carries.

What it does

Foresight is an AI-powered siting simulator that helps local governments find lower-impact locations for data centers.

Most siting decisions ask:

  • Is land available?
  • Is power nearby?
  • Is zoning easy?

Foresight adds the missing environmental question:

Can this project be built somewhere with less harm to the surrounding community?

Users can place a proposed data center on the map, and Foresight generates a neighborhood risk profile using public data from the EPA, CDC, U.S. Census, Georgia EPD permits, and data center location sources.

For each site, Foresight shows:

  • Existing health and pollution burdens
  • Community vulnerability
  • Nearby homes, schools, hospitals, and neighborhoods
  • Lower-impact alternative sites nearby

Using a predictive regression model, Foresight forecasts future neighborhood health burden under three scenarios:

  1. No new facility
  2. The proposed data center site
  3. A lower-impact alternative site

Foresight then generates a plain-language public briefing with projected impacts, site rankings, recommended alternatives, and planning guidance. Users can also submit the briefing directly as a Georgia Public Service Commission public comment, turning analysis into civic action.

How we built it

We built Foresight as a full-stack geospatial simulator for data center siting.

Data pipeline

We combined public datasets from:

  • CDC
  • EPA
  • U.S. Census
  • Census tract boundaries
  • Georgia EPD permits
  • DataCenterMap
  • Science for Georgia

We merge the data by census tract and score each area based on:

  • Environmental burden
  • Community vulnerability
  • Permit density
  • Infrastructure context
  • Projected facility impact
  • Site feasibility

Frontend

Built with:

  • React
  • Vite
  • Mapbox
  • Recharts

Users can drop a proposed site on the map, view a risk heatmap, inspect the neighborhood profile, and compare nearby alternatives.

Backend and simulation

Built with Flask, powering:

  • Tract scoring
  • Site ranking
  • Alternative search
  • Health impact simulation
  • Report generation

For forecasting, Foresight uses a machine learning time-series model built with Prophet to project neighborhood health burden through 2030 across three scenarios:

  1. No facility
  2. Proposed site
  3. Lower-impact alternative

Reports and civic action

Foresight generates a plain-English DOCX briefing and can auto-submit it as a Georgia Public Service Commission public comment using Playwright.

Challenges we ran into

One major challenge was the data pipeline. An EPA data source we initially planned to use had been discontinued, so we had to rebuild our approach using similar variables from other public sources.

We cleaned, merged, standardized, and extrapolated data across different formats and geographic boundaries. That made the project harder, but it also forced us to think carefully about how to compare communities fairly.

Accomplishments that we're proud of

We are proud that Foresight is more than a static map. In a limited amount of time, we built a working machine learning simulation layer that projects neighborhood health burden through 2030 and compares three siting scenarios:

  1. No facility
  2. Proposed site
  3. Lower-impact alternative

We are also proud of the public data pipeline. Foresight brings together environmental, health, demographic, permit, census tract, and data center location data into one usable geospatial system.

Most importantly, we built Foresight around action. It turns analysis into a public briefing and can submit that briefing as a Georgia PSC public comment with one click.

What we learned

We learned that public data is powerful, but messy. The hard part was not just finding data. It was making environmental, health, demographic, permit, census, and infrastructure datasets work together.

Each source had different formats, boundaries, and missing pieces, so we had to clean and standardize everything before it could support real planning decisions.

What's next for Foresight

  1. Add more data. We want to include real-time air quality, water usage, grid capacity, energy demand, zoning constraints, community health indicators, and local infrastructure stress.

  2. Expand beyond Atlanta. The current version focuses on Atlanta and Georgia data, but the same pipeline can scale to other cities and states by connecting local environmental, zoning, permit, infrastructure, and demographic datasets.

  3. Scale to national siting analysis. Our goal is to move from Atlanta coverage to statewide, regional, and eventually national analysis, giving more governments and communities a way to compare data center locations before approval.

Built With

  • cdc-environmental-justice-index
  • cdc-places
  • census-tiger
  • datacentermap
  • epa-ejscreen
  • epa-toxic-release-inventory
  • flask
  • georgia-epd-permit-data
  • mapbox-gl
  • node.js
  • playwright
  • positive-regression-machine-learning
  • prophet
  • python
  • react
  • recharts
  • science-for-georgia-arcgis-layers
  • u.s.-census-acs
  • vite
Share this project:

Updates