Redlining and Urban Greenness: Linking HOLC Grades to Sentinel‑2 NDVI in San Francisco

Authors

• Evan Passalacqua
• Andie McClaine
• Sydney Griscavage

Project Overview

This project investigates how historical redlining relates to contemporary urban vegetation in San Francisco by combining HOLC “Mapping Inequality” polygons with satellite‑derived greenness indices. Using Sentinel‑2 Level‑2A imagery, the analysis computes the Normalized Difference Vegetation Index (NDVI) across the city and summarizes greenness within each historically graded neighborhood.

The workflow recreates the kind of patterns highlighted by Schell et al. (2020), who argue that systemic racism and classism shape urban ecological conditions, including access to vegetation and exposure to environmental stressors. By visualizing NDVI alongside HOLC grades, the project illustrates how structurally advantaged neighborhoods tend to retain more greenness than historically redlined areas, reflecting persistent environmental injustice.

Data Sources

Primary Vector Dataset
Mapping Inequality: HOLC redlining polygons for U.S. cities, accessed via the University of Richmond’s “Mapping Inequality: Redlining in New Deal America” project.

Primary Raster Dataset
Sentinel‑2 Level‑2A surface reflectance imagery, accessed via STAC from the earth-search (AWS) catalog and used to compute NDVI for the San Francisco study area.

Scientific Context
Schell, C. J. et al. (2020). “The ecological and evolutionary consequences of systemic racism in urban environments.” Science.

Computational Environment

• Python
• Ibis for spatial vector access via DuckDB
• DuckDB as the analytics backend for HOLC vector data
• GeoPandas for local vector manipulation and export
• pystac-client and odc.stac for STAC search and Sentinel‑2 raster loading
• rioxarray for raster reprojection and COG export
• exactextract for zonal statistics of NDVI within HOLC polygons
• leafmap for interactive web mapping
• Matplotlib (seaborn-style boxplot) for statistical visualization
• Jupyter notebook workflow

Repository Structure

• notebook.ipynb contains the full analysis: vector loading and filtering, Sentinel‑2 NDVI computation, zonal statistics, and plots.
• docs/index.html hosts the interactive map overlaying NDVI with HOLC polygons and is served via GitHub Pages.
• docs/sf_ndvi.tif is a cloud‑optimized GeoTIFF of the NDVI mosaic for the San Francisco study area.
• docs/sf_redlining.geojson stores the HOLC polygons used for spatial joins and mapping.
• docs/sf_redlining_ndvi_greenness_stats.geojson augments each HOLC polygon with summary NDVI statistics (mean, median greenness).

Visualizations

Interactive Map:
https://espm-157.github.io/spatial-evan-andie-sydney/#

• Figure 1: Interactive map with Sentinel‑2 NDVI as the background and HOLC polygons overlaid; clicking a polygon reveals NDVI summary statistics such as mean greenness.
• Figure 2: Boxplot summarizing the distribution of mean NDVI by HOLC grade, allowing direct comparison of greenness across historically graded neighborhoods.

Summary of Findings

The interactive NDVI map and zonal statistics reveal systematic differences in vegetation cover across HOLC grades in San Francisco. Neighborhoods that were historically graded as more desirable tend to show higher mean NDVI than those that were redlined, reflecting greater access to trees and green space in structurally advantaged areas. These patterns are consistent with broader literature documenting that racist housing and planning policies have left lasting ecological legacies, including unequal exposure to heat, pollution, and environmental amenities.