ASSESSING THE ACCURACY OF SHALLOW WATER DEPTH ESTIMATION BY USING MULTISPECTRAL SATELLITE IMAGES

: Timely and accurate bathymetry information is needed to support an effective policy on utilization and management of coastal natural resources. Satellite derived bathymetry (SDB) has been widely considered as an advanced and low-cost method for shallow water depth estimation. This is due to the availability of multi-temporal and multi-resolution satellite data. This study focuses on evaluating the accuracy of satellite derived bathymetry derived from multispectral images recorded by various sensors with various spatial resolution. The study area is located in a small island nearby Morotai Island, Indonesia. Four SDB models were compared. The implementation of the SDB model was carried out by combining echo-sounding measurements and the reflectance of blue, green, red, and near infrared bands of three satellite images (World View 2, Sentinel 2A and Landsat 8). Our findings reveal that all three satellite images performed well in assessing SDB at various spatial and spectral resolution, however, the use of high-resolution imagery did not always improve accuracy, for example when using SVM (Support Vector Machine). When using RF (Random Forest), Sentinel 2A produced the best accuracy and when using GAM (Generalized Additive Model), the most feasible result was generated only by using WorldView 2 image. In all cases, RF performed well and provided the most accurate SDB prediction.