i

Implied Urban
Nature Vitality

San Francisco, California

Scroll to reveal ↓

Most maps tell you where the parks are. Few tell you whether a city’s nature is actually alive.

So we read San Francisco through three lenses — who lives here, what the street looks like, and how green space is used.

A measure of urban nature not by acreage, but by life: Implied Urban Nature Vitality.

Click to explore the map

Interactive Project · Urban Vitality & Environmental Justice

Most maps tell you where the parks are. Few tell you whether a city's nature is actually alive — seen, used, and felt by the people around it. This project proposes a measure for exactly that, read across San Francisco through three very different lenses.

808Kresidents
240census tracts
26,810street-view points
4,180Flickr photos
442green spaces
12vitality types

The question

Green isn't the same as alive

Conventional green-space studies map distribution in two dimensions — acres, polygons, buffers. But a vibrant tree-lined street and a fenced, empty lawn can look identical from above. Implied Urban Nature Vitality (IUNV) reframes urban nature as a human–urban–nature interaction: not merely where green exists, but how vital it is to the people who live with it.

Three lenses

One city, read three ways

01

Census

Who lives here

Nine socio-economic variables from the ACS 2016–2020 (CDC Social Vulnerability Index): population, poverty, education, age, and ethnicity — summarized for every census tract.

02

Street view

What the street looks like

26,810 Google Street View images sampled every 10 m along the road network, segmented with the ADE20K model (~150 classes) into the share of trees, sky, buildings, and road.

03

Social media

How it's actually used

Geotagged Flickr photos of the city's green spaces (2013–2022), labelled by the Google Vision API and grouped by hierarchical clustering into recurring scenes.

The method

From raw data to vitality

Step 1 · Acquire

Build the dataset

OSM road network & land use, Google Street View imagery, ACS census data, and Flickr photos.

Step 2 · Process

See & cluster

GIS, computer-vision segmentation, and unsupervised hierarchical clustering of image labels.

Step 3 · Model

Test the drivers

Pearson correlation, multilevel OLS, Moran's I, and a spatial error model to find what moves IUNV.

12 vitality types

The scenes of a living city

Hierarchical clustering of 3,434 cleaned Flickr photos surfaced twelve recurring kinds of urban-nature life — from waterfronts and leisure parks to cultural performances and sacred green spaces. Bars show how often each scene appeared.

Leisure Parks
617
Eco-Friendly Gatherings
511
Architecture & Nature
501
Waterfront Living
447
Natural Green Spaces
368
Art in Urban Design
277
Urban Blossom
225
City on Wheels
175
Cultural Performances
80
Monuments Amid Nature
80
Urban Aviary
77
Sacred Green Spaces
76

Reading the map

Where signals align

Switch between Photos, Street view, and Census in the interactive map above to read the city three ways at once. Comparing the social-media scenes against the census layer reveals where social, environmental, and behavioural signals reinforce one another — and where access to vibrant urban nature falls short. Younger neighbourhoods such as South Beach and Mission Bay register higher IUNV, while lower educational attainment tracks with lower vitality.

What we found

Five readings of urban vitality

Macro context rules

Accessibility, land-use mix, road density, and population density were the dominant drivers of IUNV — the city's structure sets the stage.

The street view matters

Among visual features, trees and sculptures were the two most powerful — greenery and public art consistently lifted vitality.

Crowding cuts both ways

More visible people, walls, and fences correlated negatively; open, sittable space (chairs) correlated positively.

Demography leaves a trace

Younger areas showed higher vitality; education and economic constraints emerged as limiting factors.

Vitality is local

A near-zero Moran's I (−0.0136) means vitality doesn't cluster neatly by geography — it is stitched together block by block.

Why it matters

Designing for vitality, not just acreage

By isolating the levers that actually move vitality — canopy, public art, and accessible, sittable open space — the IUNV framework helps planners allocate green investment where it yields the most liveability and equity, rather than counting hectares alone.

Project

San Francisco Nature Vitality

San Francisco Nature Vitality maps where residents, greenery, and everyday activity converge across the city. By combining census demographics, street-view greenery, and geotagged social-media activity, the project builds an implied measure of urban nature vitality — and surfaces where access to vibrant green space is unevenly distributed across neighbourhoods.

Team

Mingze Chen — Research Lead
Yuxuan Cai — Research
Shuying Guo — Research
Mingze Chen — Visualization & Web

A Nature AI Lab project.

Publication

Chen, M., Cai, Y., Guo, S., Sun, R., Song, Y., Shen, X. (2024). Evaluating implied urban nature vitality in San Francisco: An interdisciplinary approach combining census data, street view images, and social media analysis. Urban Forestry & Urban Greening.

Use & Credit

The visualizations on this page may be reused in any publication provided that:

  1. They are duly credited as a project by the Nature AI Lab;
  2. A copy of the publication is sent to mingze.chen@ubc.ca.

For more information, mingze.chen@ubc.ca

NATURE AI LAB natureailab.com
Lab Publications ↗ Lab News ↗