Satellite-derived bathymetry for selected shallow Maltese coastal zones

Abstract

Bathymetric data acquisition has become paramount to manage and operate coastal regions effectively. The conventional method of conducting in-situ bathymetric surveys through the use of echo sounders is inadequate in shallow water environments and entails considerable logistical expenses. Conversely, lidar mapping offers a proficient approach to surveying coastal regions. Nonetheless, this entails substantial financial expenses for data acquisition. Contrastingly, satellite-derived bathymetry (SDB) represents a more costeffective approach to surveying coastal areas, although exhibiting a lower resolution level. This study combines all three of these techniques to achieve accurate bathymetric depth data of three pocket beaches, Golden Bay, G魔ajn Tuffie魔a Bay and G魔adira Bay, located in the Maltese archipelago. An empirical pre-processing workflow for estimating SDB was developed through the use of collected in-situ depth measurements and satellite data acquired from Google Earth Engine. Four separate machine learning algorithms namely, random forest, k-star, multilayer perceptron, and linear regression, were tested on Golden Bay and G魔ajn Tuffie魔a Bay, each of which produced varied depth accuracies through the calibration of SDBs with depth values derived from alternative techniques. K-star provided the most accurate predicted depth values, followed by random forest, multilayer perceptron and linear regression which produced the least accurate results. The predicted depth outcomes were also spatially visualised using maps, providing a better understanding of the variation in depth both temporally and between the different algorithms. Additionally, G魔adira Bay was used to validate the previous results. Thus, this study offers an understanding of the level of accuracy attainable in determining depths of shallow coastal areas using SDB methods.