Harmonizing Multi-Source Sonar Backscatter Datasets for Seabed Mapping Using Bulk Shift Approaches

Abstract

The development of multibeam echosounders (MBES) as a seabed mapping tool has resulted in the widespread uptake of backscatter intensity as an indicator of seabed substrate properties. Though increasingly common, the lack of standard calibration and the characteristics of individual sonars generally produce backscatter measurements that are relative to a given survey, presenting major challenges for seabed mapping in areas that comprise multiple MBES surveys. Here, we explore methods for backscatter dataset harmonization that leverage areas of mutual overlap between surveys for relative statistical calibration—referred to as “bulk shift” approaches. We use three multispectral MBES datasets to simulate the harmonization of backscatter collected over multiple years, and using multiple operating frequencies. Results suggest that relatively simple statistical models are adequate for bulk shift harmonization procedures, and that more flexible approaches may produce inconsistent results that risk statistical overfitting. While harmonizing datasets collected using the same operating frequency from separate surveys is generally feasible given reasonable temporal limitations, results suggest that the success at harmonizing datasets of different operating frequencies partly depends on the extent to which the frequencies differ. We recommend approaches and diagnostics for ensuring the quality of harmonized backscatter mosaics, and provide an R function for implementing the methods presented here.

Limitations of Predicting Substrate Classes on a Sedimentary Complex but Morphologically Simple Seabed

“…Promising new possibilities at various stages of the classification process are currently being developed. These include developments of remote sensing technology and improvements of backscatter processing approaches, including the merging of uncalibrated backscatter mosaics [101], calibrated backscatter [102], multispectral backscatter [103], ARA [104] and HAC [21], acquiring more representative samples by applying rigorous sampling designs [105,106] and the derivation of novel features from the primary MBES data with increased discriminatory power for sediment classification [107,108]. We expect progress to mainly occur through refinement at each stage of the mapping process as discussed above.…”

Examining the Links between Multi-Frequency Multibeam Backscatter Data and Sediment Grain Size

“…Using only a selection of incidence angles while excluding the near-nadir and outer beams' regions, and dealing with a single survey line, has been proposed as a solution to these limitations mentioned; especially when discriminating large seafloor features [30,40]. Nevertheless, several other studies that use a mosaic approach instead of portions of incidence angles exist [7,15,55].…”

“…While other studies have proposed acquisition of multifrequency backscatter from a single source [7,9,55], our study uses individual information from each single-frequency system to generate a multifrequency dataset using the PCA approach [39]. Multifrequency acquisition from a single source eliminates the potential problems associated with differences in sonar configuration, survey geometry, and signal penetration variation, which might influence the final output [2,17].…”

Seabed Mapping: A Brief History from Meaningful Words

“…The word "datum" positioning itself as the most important term over the last several decades, consistent with an increase in data availability and quality. This demonstrates how laborious the challenge of building data repositories will be, besides the difficulties associated with compiling and harmonizing datasets across surveys (as approached with backscatter data in Misiuk et al, 2020 [42]).…”