•  
  •  
 

Abstract

Seaweed and seagrass beds are an important ecosystem in coastal zones. However, they are degrading because of various causes, such as the anthropogenic impacts of coastal development, aquaculture, overharvesting, and climate change. To contribute to the research related to coastal blue carbon and marine biodiversity as well as conservation and sustainable management of natural resources in coastal regions, the spatial distribution of benthic cover derived from satellite images can be the most practical tool for monitoring seaweed and seagrass beds. This study aimed at mapping the latest distribution of seaweed and seagrass in Thailand using Landsat 8 images. Thus, we developed a classification method that includes regional segmentation by ISODATA clustering, analysis of optical and textural properties, and classification using a decision tree. First, a subset of images, including those of the Sirinat National Park in Phuket, Southern Thailand, was extracted from the Landsat 8 full-scene images as a training site for the development of a classification method. Then, the developed method was evaluated by comparing the classification result to a visual interpretation result. The classification and visual interpretation results were found to be consistent to each other with a 98% total accuracy. Next, the method was applied to the Landsat 8 full-scene image, and quality assessment was conducted at two different water-type areas: Patong Beach and Tang Ken Bay. At Patong Beach, which has clear seawater, the classification result was consistent with the result of the training site. However, in the Tang Ken Bay, where the seawater is turbid, misclassification of the result evidently occurred. It is believed that the segmentation sizes were not appropriate for benthic cover distributed over small areas, and that the thresholds utilized in the decision tree were not suitable for turbid water.

Share

COinS