Localization of Urban Trees across the USA with Generative AI

<p>Understanding the quantity and distribution of urban trees is essential to urban planning and management, climate change mitigation, and environmental justice. In this project, we use deep learning to provide a novel solution to map all trees on both public and private lands across 330 United States (U.S.) cities. We localized and counted over 278 million trees, achieving an average tree count accuracy above 91% and spatial accuracy of 1.5m. Of these trees, approximately 117 million are located on private lands (42%) and 161 million on public lands (58%). Urban tree distribution exhibits strong spatial disparity, with disadvantaged communities having substantially lower tree numbers and canopy cover than other communities. Our approach is accurate, automated, and repeatable, providing rapid urban tree localization at a national scale. The project will transform urban ecology and forestry by allowing city planners and land managers to optimize ecological services and redress longstanding environmental inequity, along with potential applications for weather and climate forecasting. This repository serves as the persistent repository for the dataset of urban tree locations of 330 U.S. cities in <strong>ESRI shapefile format</strong> and the complete code base (based on <strong>Tensorflow framework</strong>). All necessary instructions are provided in every directory of the repository.</p>