Cornell Tech Cleanup Crew

Inspiration

The inspiration behind this project comes from the lack of integrated data that connects wildfires with their long-term health impacts, social vulnerability, and geographic factors. Despite existing datasets on wildfire perimeters, air quality, social vulnerability indices, and disease prevalence, these sources remain siloed, preventing a holistic understanding. With wildfire smoke causing approximately 40,000 deaths annually in the US and projections rising to 70,000 deaths by 2050, it was clear that policymakers, researchers, and citizens needed a clearer, interactive way to visualize and comprehend how wildfires impact health across different communities, especially in wildfire-prone California.

What it does

The project offers an interactive map dashboard that aggregates various data sources related to wildfires, health outcomes, social vulnerability, and geography into clear, county-level health risk profiles for California. Users can:

View composite risk scores that combine wildfire exposure, air quality, and social vulnerability data.
Access social vulnerability overlays showing demographics like poverty, age distribution, and insurance coverage gaps.

This helps raise awareness about the long-term health consequences of wildfires, supports research, and informs policymakers and local governments for resource allocation, wildfire response, and targeted healthcare support.

How we built it

The solution integrates multiple data sources:

Historic wildfire data from the National Interagency Fire Center (NIFC)
Social Vulnerability Index data from CDC
Healthcare utilization summaries from Milliman
Research linkages between wildfire exposure and health outcomes from Systems.com

These datasets are cleaned and preprocessed at the county level for California. A composite wildfire health risk score is generated using machine learning models such as XGBoost, which learns optimal weights by balancing wildfire exposure, social vulnerability, and health data. The results are visualized in an interactive Next.js dashboard that allows users to click on counties for detailed health impact insights linked to wildfires.

Challenges we ran into

Data limitations: Available data lacked granularity and completeness, which led to challenges such as model overfitting.
Fragmented data sources: Integrating siloed datasets from different domains required substantial preprocessing and normalization.
Complexity of weighting factors: Determining appropriate weights to represent wildfire risk, social vulnerability, and health outcomes required ML modeling and tuning.
Limited personal-level data: Lack of detailed information such as household vegetation, personal medical conditions, or vehicle ownership constrained prediction accuracy.

Accomplishments that we're proud of

Successfully built an integrated multi-source dataset linking wildfire exposure to health and social vulnerability at the county level.
Developed a composite health risk scoring model driven by machine learning.
Delivered an easy-to-understand, interactive map dashboard that enables diverse users—citizens, researchers, and policymakers—to explore wildfire health risks by county.
Raised awareness about the sustained and unequal impact of wildfires on vulnerable populations, emphasizing public health equity.

What we learned

Interactive visualizations empower different stakeholders with actionable insights to better prepare and respond to wildfire health challenges.

What's next for CTCC

Expanding data coverage by collecting richer, longer-term, and proprietary datasets, including chronic disease prevalence and more granular demographic info.
Scaling the tool beyond California to national coverage.