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

In the rapidly changing dynamics of urbanization, urban forests offer numerous benefits to city dwellers. However, the information available on these resources is often outdated or non-existent, leading in part to inequitable access to these benefits for the population. Access to equitable and just green spaces is a challenge for local governments, allowing the city's inhabitants to enjoy a healthy environment. In this context, remote sensing serves as a powerful data source that enables the monitoring of the evolution and dynamics of cities over time, as well as changes in urban forests. For this study, our focus is on large trees, defined as those with a canopy diameter exceeding seven meters. These trees play a vital role in offering ecosystem services that improve the environment, biodiversity, and mental health of the inhabitants of cities. Using deep learning algorithms, we identify the large urban street trees in National Agriculture Imagery Program (NAIP) images and analyze the changes in large street trees over an 18-year period (2005-2022) in six counties of the San Francisco Bay Area. We successfully tracked changes in the presence of large trees in the public right-of-way at the census tract, city, and county levels. We tracked changes in large tree availability at the neighborhood, city, and county levels, revealing socio-demographic disparities. Our analysis found that census tracts with higher household incomes, a greater proportion of individuals who self-identify as white, and more families were positively associated with an increase in large tree canopy. This assessment provides insight into the varying levels of access to ecosystem services offered by large trees across urban environments.