Work for a Member organization and need a Member Portal account? Register here with your official email address.
Urban network analysis commonly represents streets as planar graphs, although elevation affects pedestrian route choice and the interpretation of accessibility. Gridnberg incorporates local elevation into New York City’s pedestrian network and provides a reproducible framework for comparing terrain-sensitive routing scenarios.
The project comprises a research dataset, an elevation-assignment workflow, and an interactive web application. Developed by Ariel Noyman at the MIT Media Lab’s City Science group, the release retains 313,184 pedestrian segments and 187,657 endpoint nodes, representing 99.24% of the source network’s segments.
Paper · Open application · Code and documentation · Download data
Move the origin (O) and destination (D) markers to compare routes under three cost profiles. Network controls display reported local slope and vertically exaggerated elevation. The optional reachable boundaries approximate the extent of complete segments reached within each scenario’s own route-cost score; they do not represent equal travel-time budgets.
Gridnberg combines the NYCWalks pedestrian network developed by Andres Sevtsuk and collaborators with selected roadbed and bridge elevation observations from the City of New York’s 2022 Planimetric Database. The source network represents sidewalks, crosswalks, and footpaths. The project uses its geometry, rather than its pedestrian-demand estimates.
For each network-geometry vertex, the workflow assigns the mean of selected elevation observations within a 50-meter radius. Segments are retained only when all their vertices have local elevation support. Direction-specific cumulative ascent and descent are calculated across the full segment geometry and used to derive routing costs.
The analytical release contains 313,184 physical segments, 187,657 endpoint nodes, and 626,368 directed edges. Approximately 12.8% of retained segments contain a reported local grade of at least 5%.
Uphill and downhill movement receive different weights. The costs are illustrative scenario scores, not travel-time estimates. The accessibility-sensitive profile does not certify barrier-free access or exclude every short, steep segment.
In the paper’s St. George–Tompkinsville case study on Staten Island, the comfort-oriented route is approximately 5.5% longer than the distance-minimizing route. Modeled ascent across the full geometry decreases from 61.48 to 43.30 meters, a reduction of approximately 30%. The accessibility-sensitive route follows nearly the same corridor.
The Morningside Heights–Harlem case also demonstrates the dataset’s diagnostic limitations: an implausible reported grade of 209.2% affects the distance-minimizing route, while both slope-sensitive profiles avoid that edge. Such values require inspection of the source geometry and elevation support before practical use.
Noyman, A. (2026). Gridnberg: A Topography-Aware Pedestrian Routing Dataset for New York City. arXiv preprint, submitted July 24, 2026. doi:10.48550/arXiv.2607.22523.
Paper and citation · Full-text PDF
The public CSV tables constitute the analytical release. The application uses derived data optimized for browser visualization.
Gridnberg builds on the pedestrian-network research and datasets of Andres Sevtsuk and collaborators and on elevation data maintained by the City of New York. Please cite both Gridnberg and the source pedestrian network when using the release.