A Simple Cloud-native Spectral Transformation Method to Disentangle Optically Shallow and Deep Waters

Kurzfassung

Seagrasses—one of the world’s most productive ecosystems—provide many valuable ecosystem services such as habitat provisioning, biodiversity maintenance, food security, coastal protection, and carbon sequestration. With the projected temperature extremes and sea level rise due to climate change, these important ecosystems are highly threatened. Conserving these important ecosystems requires accurate and efficient mapping of its distribution and trajectories of change, which naturally includes seagrass meadows across their full depth range of growth. Unfortunately, the spectral similarities between the detectable deep seagrass and optically deep water pixels in the satellite images, the so-called dark pixel confusion, causes potential classification errors. Within the context of the Global Seagrass Watch project, funded by DLR and supported by the Group on Earth Observations-Google Earth Engine program, we have developed a novel coastal aerosol-blue-green false colour HSV approach within the Google Earth Engine platform to identify and mask out these optically deep water pixels on open Sentinel-2 satellite imagery, at both Top-of-Atmosphere (L1C) and Bottom-of-Atmosphere level (L2A). We have also explored comparisons with the commonly used band ratio approach. Both approaches are able to separate optically deep and shallow waters using supervised thresholding. While the band ratios as a group feature a slightly better quantitative performance, the performance of the individual band ratios are highly reliant on water quality and thus not always better. Comparatively, the false colour hue and saturation bands are more consistent in performance for the L1C and L2A images, respectively, and therefore easily transferable to another region. By removing the optically deep water pixels, this method helps reducing potential dark pixel confusion to improve seagrass detection pushing the limits of seagrass remote sensing. This could benefit scientists focused on seagrass-related mapping, protection, conservation, and conservation.