Incorporation of neighborhood information improves performance of SDB models

Mapping of water depth in nearshore waters is important for safe navigation and other human uses of coastal areas. Optical satellite imagery offers a cost-efficient source of data for this pur-pose. Several methods have been developed for satellite-derived bathymetry (SDB), but the de-tails of their implementation, as well as their relative performance and its dependence on the data and environmental context, is unclear. We compared five SDB models, calibrated and tested against airborne lidar data from southern Corsica. We used a blocked strategy to split data into calibration, validation, and test data, because randomly splitting the data resulted in spatial auto -correlation between these data sets and underestimation of model prediction errors. In general, the non-parametric random forest and neural network models outperformed the parametric multi-band and band-ratio models. The convolutional neural network (CNN) model performed the best, especially in relatively deep (10-20 m) areas, and it could be near-optimally tuned with only & SIM;1000 data points. However, developing this CNN model was an iterative trial-and-error process that took several months. Using the random forest model, we demonstrated that using in-formation from more than two bands in the visible and near-infrared spectrum contributed to im-proving model performance, as did incorporation of information from the local neighborhood. We suggest that the use of more than two bands, and the inclusion of spatial contextual informa-tion in addition to the values of the individual pixel, should become standard practice in SDB.

Automated summary

Mapping water depth in nearshore waters is crucial for navigation safety and human activities in coastal areas. Optical satellite imagery provides a cost-effective data source for this purpose. Five satellite-derived bathymetry models were compared and the convolutional neural network (CNN) model performed the best, especially in relatively deep areas. The use of more than two bands and spatial contextual information should become standard practice in satellite-derived bathymetry.