Differentiating Eelgrass (Z. marina), Finger Kelp (L. digitata), and Sugar Kelp (S. latissima) at the Eastern Shore Islands, Nova Scotia, Using PRISMA Hyperspectral Satellite Imagery

Abstract

Laminaria digitata (finger kelp), Saccharina latissima (sugar kelp), and Zostera marina (eelgrass) are three marine macrophytes whose habitats are known as blue carbon ecosystems due to their outstanding carbon sequestration capabilities. All three species are found at the Eastern Shore Islands (ESI), Nova Scotia, an Area of Interest (AOI) for ecological and biological importance. It is important to distinguish these species, as they have different suitable water temperature ranges, contribute to the blue carbon ecosystem in different ways, and are at threat due to climate change. Previous regional studies using multispectral satellite sensors to map marine macrophytes were unsuccessful at differentiating eelgrass from brown algae, but a hyperspectral satellite sensor offers the high spectral resolution (many bands <15nm wide) that is needed for detecting subtle differences between species. This thesis assessed the ability of PRISMA 30m spatial resolution hyperspectral satellite imagery to map eelgrass, finger kelp, and sugar kelp using an area near Sheet Harbour, Nova Scotia, within the ESI AOI, as a study site. In July 2025, in-situ data was collected by drop camera survey to confirm absence or presence of target species, then combined with the spectral information from a PRISMA image taken on August 7, 2025. Six different RF models were trained and tested on the PRISMA data. We also trained and tested two models using a Landsat-9 multispectral satellite image from July 27, 2025. A >10% increase in accuracy was achieved by using the PRISMA hyperspectral satellite image over the Landsat-9 image (up to 49.2% balanced accuracy compared to 36.0%). The two highest performing PRISMA models both included depth information, either directly or through a correction to obtain bottom reflectance. This study represents an improvement in differentiating brown algae from eelgrass in the region, as eelgrass habitat was predicted at up to 57% recall, and finger kelp habitat was predicted at up to 71% recall. Further studies should focus on better mapping sugar kelp (recall <5%), as the heterogeneity of species in this region likely affects the models’ accuracy.