Recently, the rapid development of the seabed mapping industry has allowed researchers to collect hydroacoustic data in shallow, nearshore environments. Progress in marine habitat mapping has also helped to distinguish the seafloor areas of varied acoustic properties. As a result of these new developments, we have collected a multi-frequency, multibeam echosounder dataset from the valuable nearshore environment of the southern Baltic Sea using two frequencies: 150 kHz and 400 kHz. Despite its small size, the Rowy area is characterized by diverse habitat conditions and the presence of red algae, unique on the Polish coast of the Baltic Sea. This study focused on the utilization of multibeam bathymetry and multi-frequency backscatter data to create reliable maps of the seafloor. Our approach consisted of the extraction of 70 secondary features of bathymetric and backscatter data, including statistic and textural attributes of different scales. Based on ground-truth samples, we have identified six habitat classes and selected the most relevant features of the bathymetric and backscatter data. Additionally, five types of image processing pixel-based and object-based classifiers were tested. We also evaluated the performance of algorithms using an accuracy assessment based on the validation subset of the ground-truth samples. Our best results reached 93% overall accuracy and a kappa coefficient of 0.90, confirming that nearshore seabed habitats can be accurately distinguished based on multi-frequency, multibeam echosounder measurements. Our predictive habitat mapping of shallow euphotic zones creates a new scientific perspective and provides relevant data for the management of natural resources. Object-based approaches previously used in various environments and areas suggest that methodology presented in this study may be scalable.

Glacial landforms are a significant element of landscape in many regions of Earth. The increasing availability of high-resolution digital elevation models (DEMs) provides an opportunity to develop automated methods of glacial landscape exploration and classification. In this study, we aimed to: 1) identify glacial landforms based on high-resolution DEM datasets; 2) determine relevant geomorphometric and spectral parameters and object-based features for the mapping of glacial landforms; and 3) develop an accurate workflow for glacial landform classification based on DEM. The developed methodology included the extraction of secondary features from DEM, feature selection with the Boruta algorithm, object-based image analysis, and random forest supervised classification. We applied the workflow for three study sites: one in Svalbard and two in Poland. It allowed the identification of six categories of glacial landforms: till plains, end moraines, hummocky moraines, outwash/glaciolacustrine plains, valleys, and kettle holes. The majority of relevant secondary features represented DEM spectral parameters calculated from 2-D Fourier analysis. The supervised classification models with the highest performance exhibited up to 96% overall accuracy with regard to a groundtruth dataset. This study showed that glacial landforms can be identified using novel image-processing methodology and spectral parameters of high-resolution DEM. The complete classification workflow developed herein provides a solution for the transparent generation of thematic maps of glacial landforms that may be reproducible and transferrable to various glacial regions worldwide.

Acoustic seafloor measurements with multibeam echosounders (MBESs) are currently often used for submarine habitat mapping, but the MBESs are usually not acoustically calibrated for backscattering strength (BBS) and cannot be used to infer absolute seafloor angular dependence. We present a study outlining the calibration and showing absolute backscattering strength values measured at a frequency of 150 kHz at around 10–20 m water depth. After recording bathymetry, the co-registered backscattering strength was corrected for true incidence and footprint reverberation area on a rough and tilted seafloor. Finally, absolute backscattering strength angular response curves (ARCs) for several seafloor types were constructed after applying sonar backscattering strength calibration and specific water column absorption for 150 kHz correction. Thus, we inferred specific 150 kHz angular backscattering responses that can discriminate among very fine sand, sandy gravel, and gravelly sand, as well as between bare boulders and boulders partially overgrown by red algae, which was validated by video ground-truthing. In addition, we provide backscatter mosaics using our algorithm (BBS-Coder) to correct the angle varying gain (AVG). The results of the work are compared and discussed with the published results of BBS measurements in the 100–400 kHz frequency range. The presented results are valuable in extending the very sparse angular response curves gathered so far and could contribute to a better understanding of the dependence of backscattering on the type of bottom habitat and improve their acoustic classification.

The rise of Artificial Neural Networks (ANNs) has revolutionized various fields of research, demonstrating their effectiveness in solving complex problems. However, there are still unexplored areas where the application of neural networks, particularly Convolutional Neural Network (CNN) models, has yet to be explored. One area where the application of Artificial Neural Networks is even expected is geomorphology. One of the tasks of geomorphology is the classification of landforms in a broad sense. Such classification requires a precise interpretation approach to create a homogeneous product, and this requires time and a uniform, consistent approach by the interpreter, which is not easy to achieve with manual operations. Classifications in geomorphology are mainly performed by manual or semi-automatic methods. The use of Artificial Neural Networks can complement and, in many areas, replace manual classification and reduce the time commitment of the interpreter, not least because of its repeatability and objectivity, which is a definite advantage in the case of geomorphological interpretation of vast areas. This study uses two popular CNN architectures, including VGG and Residual Neural Network (ResNet), to solve the problem of classifying glacial and fluvioglacial landforms based on a Digital Elevation Model (DEM). The results of this study show that CNNs can produce high accuracy scores (up to 87% overall accuracy) for a ground-based dataset and are a suitable method for identifying glacial and fluvioglacial landforms using DEM data as input. The presented method provides an objective, reproducible and fast tool for automatic geomorphological analysis of terrain imagery of vast areas.

High-resolution images of the seabed obtained with the use of hydroacoustic measurements allow a detailed identification of inaccessible seabed areas such as the Hans Glacier foreland in the Hornsund Fjord on Spitsbergen. Analyses presented in the paper were carried out on a Digital Elevation Model (DEM) of the bay’s seafloor exposed in the process of deglaciation, obtained from bathymetric data recorded by a multibeam echosounder. The main objective of this study was to show the relevance of the autocorrelation length parameter used to describe the roughness of the bottom surface based on the example of seafloor postglacial forms in the Hans Glacier foreland. The resulting parameter reflects the scale of the terrain roughness, which varies between geomorphologic forms. Maps of the autocorrelation length were derived from successive tiles of the data, overlapping by 90%. Based on this, the two-dimensional Fourier transform (2D FFT) was successively conducted, and the power spectral density and autocorrelation were calculated following the Wiener–Khinchin theorem. The thus obtained parameter describes the scale of the glacial bay seafloor roughness, which was assigned to the geomorphological features observed.