Identification of seagrass beds using high-resolution satellite imagery to basic inventory data supporting coastal ecosystem management (Case study: Lancang Island)

Abstract

Abstract Seagrass habitat is an important element of the coastal marine ecosystem, as it influences all major fields in the coastal environments. Due to anthropogenic and natural pressure, such identification of habitat distribution becomes crucial. Remote sensing can provide the best solution regarding time and cost to extract seagrass beds’ spatial and temporal information. When the low spatial resolution satellite image provides low accuracy and high uncertainty of seagrass distribution, we assess the high spatial resolution image product, the WorldView-2 image, to identify the seagrass bed distribution in Lancang Island. We confidently found that higher resolution images could leverage the mapping quality, especially on the seagrass beds, which could be highly sensitive to miss-detection. We can produce 83,75% overall accuracy using pixel-based classification; this is better among other low-resolution images for Instants, Landsat, and Sentinel. Our results suggest that the WorldView-2 image could perform better on seagrass beds identification, yet further improvement on more robust classification methods and algorithms could amplify the mapping and monitoring strategy. Based on this, it can be used as basic data to support future coastal ecosystem management on Lancang Island.