Abstract
Kelp forests are highly productive and diverse coastal marine ecosystems which have high variability in extent and biomass spatially and through time. Mapping and monitoring their distribution is integral to understanding the ecology of kelp forests and to inform marine protected area planning. Canada's Pacific coast presents specific challenges to mapping canopy forming kelp with thousands of kilometers of coastline, complex topography, and a large tidal range. While in situ methods and manual interpretation of aerial photography are commonly used to map canopy-forming kelp at local or regional scales, these methods are prohibitively expensive for continued large-area application. Historical and current available inventories of kelp extent are therefore incomplete for Canada's Pacific Coast. The advent of Google Earth Engine, a cloud-computing platform with a repository of Landsat imagery, provides an opportunity to apply Landsat image archive analyses and integrate temporal compositing and filtering to obtain unprecedented data on historical and current kelp extents. However, the effects of the spatial resolution and temporal coverage for mapping canopy-forming kelp in regions with complex coastlines and large tidal ranges are unknown. The objectives of this study were twofold: (1) develop a tool within Google Earth Engine to automate the detection of kelp using the Landsat satellite image archive, and (2) quantify the associated limitations for mapping the surface extent of canopy-forming kelp by comparing Landsat derived kelp extent with extents derived from WorldView-2 satellite imagery. The results indicate that the overall accuracy of the developed method is high at 80% and very close in accuracy to WorldView-2 data, but no detection is feasible within approximately one 30 m pixel from the shoreline. The interaction between the spatial resolution, kelp bed size, and bed configuration affected the detected kelp extent, as did the tide level during image acquisition. For both sensors, a 2 m increase in tide resulted in a 40% decrease in kelp extent detected. The Google Earth Engine kelp mapping tool presented provides an efficient and transferable method for mapping canopy kelp extent, and the only currently available method to reconstruct coast wide kelp development for the past three decades. The presented tool has applications in conservation and management planning, for example in creating a baseline of past kelp extent and examining trends in kelp forest variability.
Original language | English |
---|---|
Pages (from-to) | 41-50 |
Number of pages | 10 |
Journal | Remote Sensing of Environment |
Volume | 220 |
DOIs | |
Publication status | Published - 1 Jan 2019 |
Funding
This work was funded supported by the Hakai Institute ( www.hakai.org ), Fisheries and Oceans Canada 's National Marine Conservation Targets Program and the Strategic Program for Ecosystem Research and Analysis, and the University of Victoria , Canada. A portion of the WorldView-2 imagery was provided by the DigitalGlobe Foundation.
Keywords
- Canada
- Google earth engine
- Kelp
- Landsat
- Macrocystis pyrifera
- Nereocystis luetkeana
- Pacific
- Worldview 2