Mapping seaweed forests with IKONOS image based on bottom surface reflectance

Abstract

Seaweed forests are important habitats for many fishery species. However, decrease in seaweed forests is reported in all over Japan. Mapping and monitoring seaweed forest distribution is necessary for understanding their present status and taking measures for their conservation. Since traditional diving visual observation is not efficient for large scale mapping, alternative method is required. Although satellite remote sensing is one of the noteworthy methods, only a few studies have been conducted probably due to two main problems about mapping seaweed forests by remote sensing. The first one is a difficulty to collect field truth data. The second one is a light attenuation effect in water column which makes analysis more difficult. We applied an efficient method to overcome these two problems. We selected the seaweed beds off Shimoda in Izu Peninsula, Japan, as a study area. An IKONOS satellite image was used for analysis because its high spatial and radiometric resolutions are practical for seaweed mapping. We measured spectral reflectance profiles of seaweed and substrates in the study area. The result revealed effective wavelength bands for distinguishing seaweeds from other substrates. Truth data for satellite image analysis and evaluation were collected in the field using the boat and an aquatic video camera. This method allowed us to collect many truth data in short time. Satellite image analysis was conducted using radiometric correction for water column and maximum likelihood classification. The overall accuracy using error matrix reached 97.9%. The results indicate usefulness of the method for seaweed forest mapping.