抄録
Multispectral remote sensing methods of water depth using high-resolution satellite imagery can potentially play an important role in coral reef bathymetry. The popular method by Lyzenga et al. predicts water depth by a linear function of variables derived from observed radiance for each band. Up to now, the predictor has been calibrated image to image independently using corresponding in-situ depth measurements. However, a recent study has shown some inter-image compatibility of the predictor.
Based on the result, in this study we tried to build a predictor which can be applied to a wide range of coral reef images. We used four WorldView-2 images of three coral reef areas of different bottom types and water qualities. As a result, we successfully obtained a predictor which has good accuracy over all the images. The generality of the predictor will reduce the amount of the depth measurements necessary for calibration of each image to be predicted.
