About — Sentinel Chicago

About Sentinel Chicago

Sentinel Chicago is an open-source environmental health data platform that correlates urban wildlife data with environmental justice indicators across Chicago's 77 community areas.

The platform reveals neighborhood-level patterns in anticoagulant rodenticide exposure risk, wildlife monitoring gaps, and access to wildlife rehabilitation services — using birds of prey as sentinel species for broader environmental health.

This is the v0.3 prototype: a static interactive map built from publicly available data. No backend, no database — all data is baked into GeoJSON files served as static assets.

Methodology

AR Exposure Risk Score

The Anticoagulant Rodenticide (AR) Exposure Risk Score is a composite index predicting where Chicago raptors face the highest risk of secondary rodenticide poisoning. It combines four data layers, each ranked as a percentile across all 77 community areas:

ar_risk_score = mean(rodent_baiting_pctile, raptor_habitat_pctile, ej_burden_pctile, lead_poisoning_pctile)

Component	Source	Rationale
Rodent baiting density	Chicago 311 (2019–2026)	Direct proxy for rodenticide deployment
Raptor habitat (observations)	eBird	Exposure pathway — raptors hunt rodents here
EJ demographic burden	EPA EJScreen 2024	Cumulative environmental vulnerability
Lead poisoning rate	Chicago Health Atlas	Environmental health burden proxy

Tiers: High Risk (>75), Elevated (50–75), Moderate (25–50), Lower (<25).

Published studies from other cities consistently show 60–90%+ AR detection rates in urban raptors (Murray 2011, 2017; Elliott et al. 2022, 2024). Nobody has ever tested raptor rodenticide loads in Chicago. This model generates testable hypotheses.

Rodenticide Risk Score

A weighted prediction model for rodenticide exposure based on:

rodenticide_risk_score = 0.4 × rodent_baiting_density + 0.3 × raptor_observations + 0.15 × pre1960_housing_pct + 0.15 × lead_paint_ej_index

All components normalized to 0–100 before weighting. Older housing correlates with pest burden; lead paint EJ index captures environmental vulnerability.

Monitoring Gap Score

Identifies community areas that are undermonitored relative to their vulnerability:

monitoring_gap_score = hardship_pctile − checklist_density_pctile

Positive scores = undermonitored (high hardship, low birding coverage). Negative = well-monitored. Ranges from approximately −80 to +90 across Chicago.

eBird checklist density is significantly correlated with demographics: positively with income (r = +0.356, p = 0.0015), negatively with % people of color (r = −0.544, p < 0.001), and negatively with hardship index (r = −0.486, p < 0.001).

Choropleth Classification

All sequential choropleth layers use quantile breaks with 5 classes, meaning each color class contains approximately the same number of community areas (~15 each). This ensures good visual distribution regardless of data skew. The monitoring gap layer uses symmetric breaks centered on zero.

Data Sources

Source	Date	License	Notes
eBird Basic Dataset Cornell Lab of Ornithology	Dec 2025 release	Non-commercial research	All Illinois observations, filtered to Chicago community areas via spatial join. 455K+ checklists.
Chicago 311 Service Requests City of Chicago Data Portal	2019–2026	CC BY 4.0	Filtered to "Rodent Baiting/Rat Complaint" request type. 351K+ requests.
EPA EJScreen 2024 Archived at Zenodo	2024	Public domain	Block group level, aggregated to community areas using population-weighted means. EPA tool taken offline Feb 2025.
Chicago Health Atlas chicagohealthatlas.org	Various (2019–2023)	Open access	Lead poisoning, asthma, life expectancy, hardship index, poverty, income.
IDNR Licensed Wildlife Rehabilitator List Illinois Dept. of Natural Resources	Current as of 2025	Public document	PDF parsed and geocoded. 184 permit holders statewide, 10 in Cook County.
Chicago Community Area Boundaries City of Chicago Data Portal	Current	CC BY 4.0	77 community area polygons used as the base geography.

Limitations

eBird observation bias: Bird observations are almost entirely a function of birder presence, not bird presence. Raptor observations correlate with checklist density at r = +0.888. We are blind to raptor presence in neighborhoods where few people bird.
AR exposure model not validated: The risk scores are predictions based on publicly available proxies. No raptor toxicology data from Chicago exists to validate the model. Published studies from other cities support the approach but local validation is needed.
311 data as rodenticide proxy: Rodent baiting 311 requests indicate where the city deploys rodenticide, but not all deployment is reported through 311, and private pest control use is not captured.
EJScreen aggregation: EPA EJScreen data is at the census block group level. Population-weighted aggregation to community areas smooths local variation.
Temporal mismatch: Data sources span different time periods (eBird: all-time through Dec 2025, 311: 2019–2026, Health Atlas: various years). Trends over time are not captured in this static analysis.
Species-level analysis deferred: This prototype does not include species-specific vulnerability profiles. A second pass through the eBird data would enable migration window analysis.
Rehabber geocoding: Rehabbers are geocoded to city centroids, not exact addresses. Distances are approximate.

Open Source

Sentinel Chicago is open source under the MIT License. The codebase includes all data processing scripts, analysis code, and the web application.

Repository: github.com/sentinelchicago

Contact

Sentinel Chicago is maintained by the Sentinel Chicago project. For questions, collaboration inquiries, or to report issues, please use the GitHub repository.