Toward Automatic Urban Forest Inventories with Remote Sensing

About the Project

Urban forests provide extensive benefits to residents of cities such as controlling microclimate and sequestering carbon. Public street trees are managed by local governments to maximize their benefits. However, public street trees make up only a portion of a city’s urban forest and inventories are not located in a central location. The extent of the benefits provided to residents by their urban trees depends on the number of trees in that city, both public and private. When making policy decisions about where the next tree planting is needed most, cities need to be able to account for both the publicly and privately managed regions of urban forest.

Part of our research has been aggregating and cleaning public street tree inventories into a statewide inventory. This dataset is called the California Urban Forest Inventory (CUFI), and currently has over 7 million points. Using training data based on the CUFI and manually annotated images, we created a neural network that counts trees using aerial data from the National Agriculture Imagery Program (NAIP). Using this network, we created tree counts for cities within California, and created a spatially explicit point file that has a point for every urban tree in the state. This research allows city managers to create better estimates of the ecosystem services that their urban forests provide and can allow for more robust calculations of the urban forest’s ecosystem services and monetary value in the state of California.