Method development for mapping kelp using drones and satellite images: Results from the KELPMAP-Vega project
Gundersen, Hege; Hancke, Kasper; Salberg, Arnt Børre; Poulsen, Robert Nøddebo; Buls, Toms; Liu, Izzie Yi; Ghareeb, Medyan; Christie, Hartvig C; Kile, Maia Røst; Bekkby, Trine; Arvidsson, Karoline Slettebø; Kvile, Kristina Øie
Research report
Published version
Permanent lenke
https://hdl.handle.net/11250/3146564Utgivelsesdato
2024Metadata
Vis full innførselSamlinger
- NIVA-rapporter [7017]
- Publikasjoner fra Cristin - NIVA [2235]
Sammendrag
The KELPMAP study demonstrated that high-resolution multispectral data from drones and satellites, combined with AI-based image analysis, can efficiently map kelp forests and other coastal habitats. The field campaign, conducted in 2022 within Vega and Herøy municipalities, produced orthomosaics with 9 cm GSD for multispectral and 5 cm for RGB images. Based on drone data and AI, between 11% and 65% of the study area was identified as brown algae. Satellites overpredicted kelp forests but aligned with drone data after removing uncertain predictions. In Helgeland’s clear waters, benthic species and habitats were identified down to 10 meters. Using NIVA's statistical model, drones were estimated to map almost 60% of Norwegian kelp forests and 80-90% of total kelp biomass, despite only reaching 10 meters. Upscaling habitat maps using satellite images is possible but limited by satellite resolution. Drone-based training data enhances satellite-derived map accuracy. High-resolution drone maps are ideal for local marine spatial planning, while satellite maps are suitable for national level applications like carbon accounting. More ground truth data are needed for improved species-level mapping and validation of upscaled products. The study also assessed mapping benthic habitats according to NiN 3.0, identifying kelp forests, seaweed beds, eelgrass, and various seabed types.
Beskrivelse
Prosjektleder Hege Gundersen