Monitoring temporal changes in large urban street trees using remote sensing and deep learning

Area of Science:

• Urban ecology
• Remote sensing
• Geospatial analysis

Background:

• Urban forests offer critical ecosystem services, yet data on their distribution and accessibility is often lacking.
• Inequitable access to urban green spaces poses challenges for public health and environmental justice.
• Remote sensing provides a scalable solution for monitoring urban forest dynamics.

Purpose of the Study:

• To assess changes in large urban street trees (canopy diameter > 7 meters) over an 18-year period (2005-2022) in the San Francisco Bay Area.
• To identify socio-demographic disparities in access to large tree canopy.
• To leverage deep learning and remote sensing for urban forest monitoring.

Main Methods:

• Utilized deep learning algorithms to identify large urban street trees from National Agriculture Imagery Program (NAIP) satellite images.
• Analyzed changes in tree presence at census tract, city, and county levels over 18 years.
• Correlated tree canopy changes with socio-demographic data.

Main Results:

• Successfully tracked changes in large street tree availability across different geographic scales.
• Revealed significant socio-demographic disparities in access to large tree canopy.
• Found positive associations between increased large tree canopy and higher household income, White population proportion, and family presence.

Conclusions:

• Deep learning and remote sensing are effective tools for monitoring urban forest changes and identifying access inequities.
• Socio-economic factors significantly influence the distribution and increase of large urban trees.
• Findings highlight the need for targeted urban planning to ensure equitable access to the benefits of large trees.