Davis Urban Canopy GeoAI: NAIP mosaic and derived geospatial products for Davis, California

This dataset provides the geospatial inputs and derived products used for optical GeoAI urban-canopy mapping and neighborhood screening in Davis, California. It accompanies the research article: From crown candidates to neighborhood screening: integrating optical GeoAI and spatial modeling for urban-canopy assessment in Davis, California Study extent. Analyses use the Census TIGER municipal boundary for Davis (~25.92 km2). Optical canopy mapping is based on USDA NAIP 2022 imagery (0.6 m RGB+NIR). Crown candidates were generated with DeepForest, refined with an NDVI gate and non-maximum suppression, and segmented with a box-prompted Segment Anything Model (ViT-B) to produce a crown-anchored canopy surface. Files in this deposit davis_naip_2022.tif — USDA NAIP 2022 four-band (R, G, B, NIR) mosaic clipped to the TIGER city boundary (~581 MB; EPSG:4326). Davis_Urban_Canopy_GeoAI_derived_data.zip — replication archive containing: boundary/ — Census TIGER Davis city boundary derived/ — canopy mask, tree detections (points and boxes), Landsat land-surface temperature, street centerlines, 100 m enriched analysis table, fair USDA comparison metrics, and statistical result tables reference/ — USDA/CAL FIRE 2022 canopy raster (Davis_canopy2022.tif) for same-extent comparison Key locked comparison values (TIGER extent). Optical canopy 2.43 km2 (9.37%); USDA/CAL FIRE canopy 7.11 km2 (27.4%); recovery 34.2%; IoU 0.288; Dice 0.448; detection-center precision 0.974. Intended use. Reproduce the fair reference comparison, 100 m spatial association analyses, and planning-screen surfaces reported in the paper. Products are association and screening layers; they do not replace structural LiDAR canopy inventories, municipal assessments, or an official Tree Equity Score. Companion code. https://github.com/MohammadrezaNarimaniUCDavis/Davis_Urban_Canopy_GeoAI Provenance note. NAIP and USDA/CAL FIRE canopy products retain their original source terms of use and are redistributed here solely to support scientific replication of the reported workflow.