UCLA

This dissertation develops remote sensing methods to monitor urban forest canopy cover changes, phenology, removal, and post-fire recovery across California. I combine high-resolution National Agriculture Imagery Program aerial imagery, repeated PlanetScope satellite observations, tree inventory data, parcel information, and field and image-based validation to examine urban forest change across multiple spatial and temporal scales.First, I developed a four-year canopy cover time series for all California census-designated places from 2016 to 2022 using 60-cm aerial imagery and a deep learning classification model. I estimated canopy cover and change using an error-adjusted area estimation, which showed apparent statewide canopy decline was not statistically distinguishable from no change once mapping uncertainty was considered. Residential parcels consistently contained more than half of the canopy within incorporated urban areas, showing that statewide canopy goals will depend on engaging with private landowners. Second, I combined biannual canopy maps with monthly PlanetScope time series to identify likely individual tree removals and estimate when they occurred. Seasonal PlanetScope data classified broad foliage types with 84% accuracy across two California cities. NAIP-derived canopy change identified removals over a two-year interval with 91.7% accuracy, while a PlanetScope-based classifier identified removals from time-series data with 81.0% accuracy. For 36 of 49 manually validated removals, PlanetScope estimated a removal date within a validated removal window. These results show that aerial imagery can identify canopy loss between image dates, while repeated satellite observations can provide more information about when that loss occurred. Finally, I applied these methods to measure individual-tree canopy condition after the 2025 Eaton and Palisades fires. I compared post-fire canopy condition with each tree’s own pre-fire seasonal baseline to distinguish trees that were returning toward typical condition from those that remained below baseline. Approximately 26.7% of large fire-affected trees were within their pre-fire range by April 2025, increasing to 38.5% by April 2026. Burn severity was the strongest and most consistent predictor of post-fire canopy condition, while foliage type, species, property type, and city were also associated with recovery patterns. The methods developed in this dissertation make it possible to monitor urban forest change at scales that are difficult to capture through field assessment alone, both through broad spatial scales and frequent temporal updates. Although developed and applied across California, these methods provide an approach that can be extended to other cities with comparable high-resolution imagery, repeated satellite observations, and tree-location data. Cities can use this information to plan urban forest management better and to track whether canopy goals are being met by showing where canopy is being retained, lost, or recovering after disturbance.