Global Airborne Observatory: Locations of Trees Exhibiting Rapid Ohia Death Symptoms 2016 through 2019 on Hawaiʻi Island

Data Background This dataset contains annual geospatial detections of individual brown tree crowns associated with Rapid ʻŌhiʻa Death (ROD) on Hawaiʻi Island from 2016–2019. The detections were produced from airborne laser-guided imaging spectroscopy collected by the Global Airborne Observatory (GAO; formerly Carnegie Airborne Observatory) and processed using machine-learning methods described in Vaughn et al. (2018) and subsequently analyzed in Vaughn et al. (2023). The primary data products are point detections of individual tree crowns exhibiting canopy browning symptoms consistent with Rapid ʻŌhiʻa Death infection. Brown crown likelihood (range 0 to 1) was derived for each 2m x 2m visible-to-shortwave infrared (VSWIR) imaging spectroscopy pixel using a combined Support Vector Machine (SVM) and Gradient Boosting Machine (GBM) classification framework. Pixels with modelled brown crown likelihood that exceeded the threshold (0.65) were considered affected (i.e. browning from disease onset) canopy. Clusters of touching affected canopy pixels were considered single tree crowns, and cluster centroid location and cluster size were retained for each independent cluster with 2 or more affected pixels. The classification approach achieved high reported precision for browning crown detections (~97%) while maintaining high sensitivity. The dataset includes four annual point layers corresponding to airborne surveys conducted in 2016, 2017, 2018, and 2019. Additional polygon layers provide (1) the spatial extent of GAO airborne coverage for each year and (2) clipped or excluded regions removed from analysis because of elevated uncertainty or potential false detections. These data were used in the landscape epidemiology analysis presented in: Vaughn, N.R., Hughes, R.F., and Asner, G.P. (2023). Multi-scale remote sensing-based landscape epidemiology of the spread of rapid ʻŌhiʻa Death in Hawaiʻi. Forest Ecology and Management 538:120983. The detection methodology was originally developed and validated in: Vaughn, N.R., Asner, G.P., Brodrick, P.G., Martin, R.E., Heckler, J.W., Knapp, D.E., and Hughes, R.F. (2018). An Approach for High-Resolution Mapping of Hawaiian Metrosideros Forest Mortality Using Laser-Guided Imaging Spectroscopy. Remote Sensing 10(4):502. Data Description All files are GeoJSON formatted and provided in local UTM coordinate system: Zone 5N, WGS-84 datum (EPSG:32605). The dataset contains the following three types of GeoJSON files, with each type containing separate fiule entries for each of the 4 years: Dataset Type Description File Name Brown tree crown detections Point locations of individual tree crowns identified as exhibiting browning symptoms associated with Rapid ʻŌhiʻa Death for each survey year. Each point has the following properties: ID - point record number X - Easting Coordinate cluster centroid (m) Y - Northing Coordinate cluster centroid (m) Pixels - Number of pixels in cluster Area - Area of the cluster (Pixels x 4.0 sq. m) GAO_brown_trees_FEMpaper_xxxx.geojson GAO coverage polygons Multipolygon boundaries representing airborne imaging spectroscopy coverage for each survey year GAO_xxxx_Coverage.geojson Clipped/excluded regions Polygon regions removed from analysis because of elevated uncertainty, atmospheric effects, or other mapping artifacts GAO_xxxx_Clipped_Regions.geojson Methods Summary Airborne imaging spectroscopy and LiDAR data were collected by the Global Airborne Observatory across Hawaiʻi Island during annual campaigns between 2016 and 2019. Spectral signatures of brown and leafless ʻŌhiʻa crowns were identified from orthorectified imagery and used to train machine-learning classification models. Detected brown crown pixels were grouped into contiguous clusters representing individual tree crowns. Centroid locations of these clusters were exported as point detections and used for subsequent epidemiological analyses.