Best practices and challenges for urban tree detection, classification, and geolocation with street-level images across North American cities

Accurate and up-to-date catalogs of urban tree populations are crucial for quantifying ecosystem services and enhancing the quality of life in cities. However, identifying and mapping tree species cost-effectively remains a significant challenge. Remote sensing is an active area of research where scientists are developing general-purpose tools that can capture information about plant populations across broad spatial scales. In this study, we developed computer vision models to detect, classify, and map 100 tree genera across 23 cities in North America using Google Street View (GSV) and iNaturalist images. We validated our predictions in independent portions of each city. We then compared our predictions to existing street tree records to evaluate the spatial context of errors using generalized linear mixed-effects models. Overall, the computer vision models identified the majority of independent, ground-truthed street trees (67.1%), and performance varied across the 23 cities (67.4% +- 9.3%) and 100 genera (50.9% +- 23.0%). Matching success increased with local street-view image density, simpler local stand structure, and greater training representation, particularly from the focal city. We found that genus classification performed better in continental cities with lower relative diversity, and that seasonal changes in the appearance of trees provided visual cues that moderate classification rates. Our study demonstrates that using widely available street-level imagery is a generalizable and promising avenue for mapping tree genus distributions across urban environments.