This dataset derived from mosaicked RGB imagery, targeting only exposed kelp in priority sites along the California coastline in Sonoma and Mendocino Counties. The imagery was acquired using DJI Phantom, Matrice, and Mavic series UAVs with the standard RGB sensor; sensor resolution ranged from 12.4 - 20.8 megapixels. The ground sampling distance ranged from 2.5-3.8 cm. Pilots aimed for 75-80% front/side overlap at 400ft (120m) altitude. Flights were coordinated to coincide with annual, peak biomass of Nereocystis luetkeana and the lowest tide of the month as tidal height and surface currents have been shown to impact the amount of kelp canopy exposed on the water surface (2), and these impacts can vary regionally (3). Imagery was mosaicked using a standard mapping software (i.e. DroneDeploy, Pix4D, and PhotoScan)

Kelp pixels were identified in each UAV image using a band combination between the red and blue bands (Red - Blue), as Red - Blue has been shown to best distinguish kelp from water in RGB-UAV imagery relative to other RGB vegetation indices (3). Before applying a threshold to an image, all terrestrial objects (land, intertidal rocks) were manually masked. Due to radiometric and spectral variability present in the imagery, thresholds to distinguish kelp from water were selected manually by expert classifiers. For images with high levels of spectral variability due to turbidity, sun glint, or other artifacts, a single threshold could not be used for kelp identification, as the threshold varied throughout the image. These images were gridded into subsets (ranging from 1000 x 1000 m areas to 5000 x 5000 m areas, depending on the level of variability), and each grid was assigned a unique threshold. The classified grids were mosaicked back to their original extent. All final classification mosaics were manually edited to increase accuracy. Classifications values were binary (“Kelp” or “No Kelp”) with the exception of mixed-species marine algal beds and the occasional blurred image, which were assigned “No Data” values.

The classification process yields three levels: kelp, no kelp, and no data (when applicable), where “no data” denotes areas where it was not possible to confirm the presence/absence of kelp.

The imagery for this site was collected on September 17, 2020 between 5:06- 5:34 pm PT.

DISCLAIMER: (1)Timing of the survey is important, particularly with respect to growing season, conditions in the ocean (e.g. tides and currents), storms, and harvest levels preceding the dates of imagery collection. Seasonal variability may account for differences in surveys which may not reflect a change in the bed's extent, productivity, or harvest level. (2) Statistical significance in change of area should be evaluated. (3) Survey sensors across time and space may not be consistent. (4) Because wind and weather conditions varied significantly throughout the data collection process, surveys were not restricted to a specific tidal height or current speed – and data were collected when field conditions allowed for stable UAV launch and landing. While this limitation may introduce bias into area estimates, these data still inform the location of kelp refugia on fine spatial scales.

(2)Britton-Simmons, K., Eckman, J. E. & Duggins, D. O. Effect of tidal currents and tidal stage on estimates of bed size in the kelp Nereocystis luetkeana. Marine Ecology Progress Series vol. 355 95–105 (2008)

(3)Cavanaugh, K. C., Cavanaugh, K. C., Bell, T. W., & Hockridge, E. G. (2021). An Automated Method for Mapping Giant Kelp Canopy Dynamics from UAV. Front. Environ. Sci, 8, 587354.