Estimation of Long-Term Changes of Seagrass Habitat in the Korean Coastal Waters Using Multi-Remotely Sensed Data

Abstract

Seagrass beds plays important roles, not only providing food and habitat for various animals but also leading nutrients and carbon cycle in coastal ecosystem. Despite the importance of seagrass beds, studies of seagrass distribution in Korea, have been conducted by direct sampling method in limited depth and area. Recently, remote sensing is considered to be very effective tool to detecting wide-scales habitat. In this study, multi-satellite data were used for detecting seagrass beds in turbid waters and describing temporal changes over the past two decades. All seagrass spectra from different sensors (Landsat TM/ETM+, Aster, Spot-4, and Kompsat-2) have low reflectance in green and high in NIR band, whereas bare seawater showed opposite spectrum features. A training area in the green band and Mahalanobis Distance Classification (MDC) algorithm was adopted for classification and mapping of seagrass beds. Based on the underwater acoustic referencedata, overall accuracy of error matrix for Kompsat-2 image classification was 73% accurate. Despite the high turbidity seawater, this accuracy is considered relatively high precision. The average area of seagrass beds was 4.6 km2 from 1990to 2017, and it was consistently >4.0 km2. Seagrass beds are often inaccurately detected when using different spatial resolutions and turbid water, but the MDC method will be appropriate for turbid water of Korean coast.