You are here:
Mapping seagrass across the United States using high-resolution, commercial satellite imagery
Citation:
Coffer, M., D. Graybill, AND B. Schaeffer. Mapping seagrass across the United States using high-resolution, commercial satellite imagery. World Seagrass Conference & International Seagrass Biology Workshop, Annapolis, MD, August 07 - 12, 2022.
Impact/Purpose:
Seagrasses offer a variety of ecosystem services, including reducing wave action and shoreline erosion, promoting sedimentation, providing habitat and food resources for fish, marine megafauna, and invertebrates, mitigating ocean acidification, and both storing and filtering carbon, nitrogen, and phosphorus. This study expands methods for leveraging high spatial resolution, commercial satellites for classifying seagrass presence and absence which was first presented for St. Joseph Bay, FL, to now include eleven study sites across the United States. At each study area, reference data and a single satellite image were acquired and agreement was assessed using a variety of statistical tests depending on the spatial data type and seagrass classification type of the reference data. Agreement was high across all sites, although sparse seagrass beds were difficult to classify with satellite imagery despite the high spatial resolution. Results presented here support the use of previously published methods for large-scale seagrass mapping in coastal waters.
Description:
A recent review recommended the need for a consistent monitoring approach for seagrass habitats in order to adequately protect seagrass meadows. Satellite imagery may offer more consistent monitoring over time compared to traditional photointerpretation methods. This study leverages high spatial resolution, commercial satellite data from DigitalGlobe’s (now Maxar) WorldView-2 and WorldView-3 platforms to classify seagrass presence and absence at eleven coastal sites across the United States, representing three of the six global seagrass bioregions and each of the coastal climate regions defined by the National Centers for Environmental Information. Reference data and a single satellite image were acquired at each study area. Classification agreement was assessed depending on the reference data type; statistical tests included balanced agreement and the nonparametric Mann-Whitney U and Kruskal-Wallis tests. Despite temporal offsets of up to 16 years between satellite imagery acquisition and reference data collection, balanced agreement ranged from 58 to 86% and both the Mann-Whitney U test and the Kruskal-Wallis suggested strong agreement between satellite-indicated seagrass percent cover within reference-delineated seagrass density classes. Additionally, at a single study area, reference data consisting of point observations of seagrass percent cover spanning 0 to 100% was used to estimate WorldView-2's seagrass minimum detection level. Results indicated that a satellite pixel containing a minimum of approximately 43% seagrass will be classified as seagrass. This study offers a large-scale assessment of a standardized seagrass classification approach, demonstrating its performance across various seagrass bioregions, optical water types, and seagrass ecosystems.
