Satellite bathymetry estimation in the optically complex northern Baltic Sea

Abstract

Satellite bathymetry estimation (SBE) is a cost-effective method for producing depth information for shallow marine and aquatic areas. However, most SBE methods have been developed and are refined in clear water conditions and their feasibility in optically complex waters is poorly understood. This hinders the use of SBE in optically complex waters. To test the feasibility of SBE in optically complex waters, we used multi-temporal Sentinel-2 (S-2) imagery and in-situ measured depth information from dive computers to fit bathymetric models for 14 study areas in the northern Baltic Sea. We compared different methods of using multi-temporal S2-scenes, atmospheric correction, and statistical modeling, as well as different band combinations to define the optimal model for each study area. We tested the dependence of the depth estimation on 13 environmental variables with potential influence on the accuracy of the SBE. The optimal bathymetric models differed among the study sites, with the differences manifested in different requirements for atmospheric correction, band combination, statistical model, and the use of multiple S2-scenes. The coefficients of determination (R2) of the optimal models varied between 0.42 and 0.79, with mean (X) R2 equaling to 0.65. The coefficients of determination were positively related to the proportion of sea within the 5 km search kernel and to the exposure of the study site. At the predicted depth of 2 m, the differences between observed and predicted depths varied between 0.11 and 0.60 m (X = 0.35 m). At the predicted depth of 1 m, the differences varied between 0.04 and 0.34 m (X = 0.21 m). SBE provides accurate depth estimates in optically complex waters, even when in-situ calibration data are limited. However, the maximum detectable depth obtained with SBE in optically complex waters is significantly lower than in clear water conditions. The SBE process can be fine-tuned to the specific characteristics of an area, improving its performance in optically complex conditions. In particular, the identification of the optimal band combination is critical for successful SBE. The depth estimates produced in this study can be used in a variety of applications that require accurate information on the seafloor bathymetry.