---

🚇 Transit Accessibility Analysis in New York City

PostGIS · pgRouting · OSMnx · GTFS · Kepler.gl Live Map:

---

Map Preview

---

Overview

New York City operates one of the world’s largest rapid transit systems, yet significant accessibility gaps remain, particularly in outer boroughs. This project analyzes walking-based access to subway entrances, generates 10- and 15-minute walking isochrones, and identifies transit deserts, census tracts where subway access exceeds a realistic walking threshold.

The analysis integrates:

• OpenStreetMap walking networks (via OSMnx)
• PostGIS + pgRouting for spatial analysis and routing
• U.S. Census TIGER/Line & ACS demographics
• Kepler.gl for interactive visualization

📍 Key outcome: 674 NYC census tracts (≈29%) qualify as transit deserts under a 2,000-ft walking threshold.

---

Research Questions

• How far is each NYC census tract from the nearest subway entrance?
• Which tracts qualify as transit deserts under realistic walking assumptions?
• Do transit deserts disproportionately affect high-population or low-income communities?

---

Methodological Note: Walking Threshold

While 800 m (½ mile) is often used as a standard walkable distance, NYC’s dense station spacing and observed rider behavior suggest a shorter effective catchment.

This study adopts a 2,000 ft (~610 m, ~7–8 minute walk) threshold, reflecting:

• High subway entrance density
• Shorter tolerated walking distances among NYC riders
• More conservative, behavior-aware accessibility modeling

Data Sources

Dataset	Source	CRS	Description
Walking network	OpenStreetMap (OSMnx)	EPSG:2263	~269k nodes, ~865k edges
Subway entrances	OpenStreetMap	EPSG:2263	2,024 entrances
Census tracts (2020)	TIGER/Line + NYC Open Data	EPSG:4326	2,325 tracts
ACS 2021 (5-yr)	U.S. Census (DP03)	—	Population, median income

Walking Network of New York City

Methodology

1. Walking Network Construction

• Downloaded NYC walkable street network using OSMnx.
• Reprojected to EPSG:2263 (feet-based CRS).
• Loaded into PostGIS and built a routable topology.
• Snapped subway entrances to the nearest network vertices.

Scripts: download_osm_data.py, load_osm_to_postgis.py, prepare_network.sql

---

2. Isochrone Generation

• Used pgr_drivingDistance (undirected graph).

• Walking speed: ~3 mph.

• Thresholds:

  • 10 minutes: 2,640 ft
  • 15 minutes: 3,960 ft

• Converted reachable nodes into polygons using ST_ConcaveHull.

• Reprojected to EPSG:4326 for web mapping.

---

3. Transit Desert Identification

• Reprojected census tracts to EPSG:2263.
• Calculated minimum distance from each tract to any subway entrance.
• Flagged tracts as transit deserts if distance > 2,000 ft.

---

4. Demographic Integration

• Joined ACS 2021 DP03 data to tracts via GEO_ID.
• Cleaned and converted numeric fields (population, income).
• Handled suppressed ACS values ("-" → NULL).

---

5. Visualization

• Exported final GeoJSON with:

  • Geometry
  • Distance to subway
  • Desert flag
  • Population
  • Median household income

• Visualized in Kepler.gl:

  • Tracts colored by desert status
  • Heights scaled by population
  • 10- and 15-minute isochrones overlaid

---

Results

Distance Summary

• Total tracts: 2,325
• Average distance to subway: ~2,622 ft (~800 m)
• Minimum: 0 ft
• Maximum: ~58,786 ft (~11 miles)

---

Transit Deserts (2,000 ft Threshold)

• Desert tracts: 674 (≈29%)
• Average population per desert tract: 2,538
• Average median income (deserts): $84,289 (Higher than NYC average of ~$75k–$80k)

---

Distance Buckets

Distance	Tracts	Avg Population	Avg Income
≤1,000 ft	1,315	3,408	$76,777
1,001–2,000 ft	336	2,715	$71,217
2,001–2,625 ft	107	2,567	$75,812
2,626–5,000 ft	243	2,321	$77,805
>5,000 ft	324	2,692	$91,977

Key insight: NYC transit deserts are not concentrated in low-income areas; many occur in wealthier, auto-dependent neighborhoods.

Transit Accessibility map

---

3D Illustration

---

Discussion

• ~29% of NYC tracts qualify as transit deserts at a 2,000-ft threshold.
• Higher-income outer-borough areas dominate extreme distances.
• Unlike many cities, NYC transit deserts reflect urban form and density, not poverty concentration.

Limitations

• Straight-line distance (not full pedestrian routing).
• Station entrances may involve long internal walks.
• Results sensitive to threshold selection.

Future Work

• Integrate buses and GTFS schedules.
• Add race/ethnicity and poverty indicators (ACS DP05).
• Population-weighted accessibility metrics.

---

Conclusion

This project builds a city-scale walking network, generates realistic subway isochrones, and identifies 674 transit desert census tracts in NYC using a behavior-aware 2,000-ft threshold. Results show that NYC’s transit deserts are largely low-density and auto-oriented, rather than income-driven, highlighting structural gaps in subway coverage.

---

Tools & Technologies

PostgreSQL · PostGIS · pgRouting · OSMnx · GeoPandas · QGIS · Kepler.gl

Data Sources

OpenStreetMap · U.S. Census TIGER/Line (2020) · ACS 2021 · NYC Open Data

📌 Live demo: Kepler.gl map (https://bit.ly/3Oh2605)

---