In sublittoral environments, hard substrates like cobbles, boulders, blocks and bedrock provide essential ecosystem functions [1
]. Especially single stones (cobble, boulders and blocks) in areas where soft substrates dominate, are hotspots for marine biodiversity [3
]. For fishes and marine mammals, those areas are important breeding and feeding places [4
The localization, mapping and documentation of stones in the North Sea (and elsewhere) gains increasing importance since the European Union (EU) implemented the protection of stony areas in the European seas in 1992 [5
]. Detailed seafloor information is also a relevant precondition for resource assessments, coastal management and further protection measures. In the German Bight, existing area wide sediment distribution maps [6
] were interpolated from grab-sample data, which lack resolution and tend to under-represent “coarse sediment” (gravel and stones). Especially stones can hardly be representatively sampled with grab samplers.
Currently, there are no clear definition and consistent demarcation criteria in the international literature for stony areas. Descriptions like “high concentration of stones” [8
] or “reef like structures” [9
] are very vague and insufficient for habitat classification and modelling according to e.g., EUNIS (European Nature Information System) or HELCOM HUB (HELCOM Underwater Biotope and Habitat System) standards. Standardized demarcation criteria such as the distance between stones and the number of objects on a certain area do not exist. Both approaches require at least the localization of the single stones in adequate resolution. In this context, remote sampling techniques such as trawling, dredging, grab sampling and video inspection are inappropriate.
Sidescan sonar (SSS) imaging is a method to detect differences in material and texture of the seabed as well as to identify objects on the seafloor [10
]. In general, coarser sediments (rougher surface) will backscatter more energy than finer sediments (smoother surface) [12
]. Elevated objects affect the sonar pulse by causing high backscatter at the front and an acoustic shadow (no backscatter) behind the object. The shadow length is strongly related to the location of the object relative to the sound source (here: the SSS tow fish) and the object height [10
]. The geometric relationship between these parameters enables the calculation of object heights. Objects close to the tow fish cast shorter shadows when compared to objects of the same size located farther away. As a result, objects close to the sonar produced a shadow below the sonar data resolution and makes them undetectable. Further limitation in the detection of objects is given by benthos living on it. The biomass may absorb and scatter part of the acoustic signal so that the typical high backscatter at the front of the object does not occur [10
]. Therefore, automatic detection of objects like stones in sidescan sonar backscatter data is difficult. It basically works out for objects (e.g., mines) lying on homogeneous backscatter grounds [14
]. In principle, manual detection is possible but it requires time and high resolution of backscatter to detect small objects. Moreover, the smaller the objects to be identified, the higher resolved data are necessary and the more time consuming is the data acquisition as the survey speed needs to be reduced in order to fit more pings on a given survey line. If the data resolution is insufficient to show the acoustic shadow, a differentiation from the surrounding sediment is hardly possible.
The use of swath systems like multibeam echo sounder (MBES) for object identification is only known for mine detection [15
]. The implementation for area-wide stone detection is not practicable due to the very narrow swath in shallow waters and relative low resolution compared to SSS.
Acoustic ground discrimination systems (AGDS) can be instrumental in the effort to demarcate stony areas [16
]. Aside from the standard single-beam technique that does not allow areawide measurements, the disadvantage of such systems is the relatively large opening angle causing a large foot print (e.g., ~5.2 m at 30 m water depth). A smaller foot print is given by parametric sediment echo sounders (pSES) [19
]. The nonlinear acoustic propagation generated by pSES provides high vertical resolution (decimeter) data with small footprints (e.g., ~1.0 m at 30 m water depth) at low frequencies (e.g., 12 kHz). This technique has been successfully used for the detection of anthropogenic objects lying on or embedded in the seabed like wrecks or pipelines [19
]. Generally, hard objects are indicated in the pSES as hyperbolic curves.
The motivation of this paper is to develop an innovative and novel approach to efficiently and reproducibly identify and demarcate areas of exposed stones (erratic cobble, boulders and blocks) by combining SSS and pSES data. In particular, the goals for this new method are to:
Validation measurements were carried out in the area of the paleo Elbe valley (PEV) and the protected area “Sylt Outer Reef” (SOR) in the German Bight. The working areas include homogeneously distributed sands with single scattered stones on the seafloor and areas with heterogeneous sediment distribution where glacial lag deposits (mixed sediments including stones), gravelly substrate and Holocene fine sands change on small scales.
The area-wide, low-resolution SSS backscatter mosaic in combination with the ground-truth data suggest to differentiate the working area into two sedimentologically different areas as shown in Figure 2
The PEV reveals low backscatter data representing muddy fine sands (mean 100 µm). The SOR located east of the valley shows small-scale changes of high, medium and low backscatter. Low-backscatter data represent fine sands (mean 190 µm) and medium backscatter rippled medium sands (mean 320 µm). High backscatter represents two different coarse sediments (Figure 3
A differentiation of gravelly substrate and lag deposits by means of the backscatter on a larger scale is hardly possible. There are no differences in the backscatter values and typical ripple marks or stone signatures are hardly visible at a spatial resolution of 1 m (Figure 3
). In the case study area with a spatial resolution of 0.25 m such signatures are apparent. The amount of stones identified manually from the two SSS mosaics differ significantly (844:11,079) (Figure 4
a,b). Generally, the stones were identified on high backscatter data in the central part of the case study area. Almost no stones are present in the northeastern high backscatter area.
The AGDS data suggest a sediment distribution pattern similar to the low resolution SSS backscatter (Figure 5
). The E1/E2-scatter plot indicates two clusters: (1) smoother and softer values which corresponds with low SSS-backscatter values (sand); and (2) rougher and harder values which corresponds with high backscatter data (gravelly substrate and lag sediments). Moreover, lag sediments with stones do not produce significantly different AGDS-values compared to rippled sandy gravel to gravel areas (Figure 5
In areas with high SSS backscatter, the pSES data sometimes reveal diffracted hyperbolas at the sediment surface (Figure 6
). They become best-defined with the following pSES settings: high frequency, small pulse length and a minimum window length (Figure 6
c). The position of the diffraction hyperbolas is plotted on top of the SSS backscatter mosaic in Figure 2
b and Figure 4
c,d. Every hyperbola always represents multiple acoustic pings. Extensive ground truthing data, including high-quality underwater videos, show that these hyperbolas are associated with accumulations of pebbles, cobbles, boulders and blocks occurring together with fine sands to gravel (lag deposits). Locations of the hyperbolas clearly correlate with locations of stone deposits that were identified manually in the backscatter mosaics of the case study area (Figure 4
Area-wide mapping of SSS backscatter data showed that the differentiation of lag sediments (containing stones) from gravelly substrate (without stones) is challenging in the area of the SOR, because backscatter values are too similar and typical textures and bedforms were unidentifiable in the 1 m resolution data set. To figure out ecologically important stony areas, further information or high resolution data is needed.
The case study showed that it is principally possible to identify stony areas using SSS data with adequate resolution. However, the effort to manually “click” objects visible to the eye on a computer screen for demarcation purposes is high. It is only feasible for small areas or special investigations such as marine constructional ground investigations and environmental impact assessments. Extrapolating observations from smaller case studies to larger areas is generally impossible [35
]. Even studies using a large dataset of sediment sample information often underestimate size classes of gravel and stones [36
]. Moreover, earlier studies reported that grab samplers deployed in stony areas mainly catch finer sediment in between stones and are therefore not representative [37
]. This has been confirmed during surveys for this study.
Automated object identification for demarcation purposes was not applicable for this data set because stones are always present in areas of mixed sediments which generate a diffused backscatter. So far, quick and automated object identification is only successful when the objects are sitting on homogeneous, low backscatter substrates (e.g., fine sand) [14
]. Object identification becomes difficult and even impossible close to the nadir and also when the backscatter is unclear (e.g., stones on mixed sediment). Moreover, noises in the record created by rough sea conditions or by fishes in the water column can produce spots on the record that may be misinterpreted. Even the data acquisition time is almost sixfold for data with a resolution of 0.25 m when compared to one-meter resolution.
The AGDS RoxAnn, originally designed for the fishing industry [41
], to distinguish rocks from sand (to save bottom trawls from being destroyed) was indeed able to differentiate between fine and coarse sediments at the SOR. Fine sands with only small ripples have a much smoother surface (lower backscatter) than the large gravelly ripples and the lag deposits with stones (higher backscatter). However, the acoustic properties of the two coarse sediment types are very similar and cannot be differentiated by RoxAnn with certainty. Similar findings were reported by Humborstad et al. [42
] who related this deficiency to the relation of size between footprint and hard bottom patches and to the effect of sessile biota on the acoustic characteristics. The substrate patches at the SOR are mostly larger than the footprint due to relative low water depth. That sessile biota affects the acoustic properties of objects such as stones is reasonable by all means. Michaelis et al. [43
] have shown that more than 85% of stones in the area of the SOR are colonized to a varying degree by sessile organisms. As a result, the biotic cover may mask stones and make them undetectable by AGDS. Further, RoxAnn cannot clearly distinguish areas with and without stones as the stones of mostly cobble and boulder size are usually much smaller than the footprint (3.5–8 m).
Areas of stone occurrence are clearly apparent in the echogram of the pSES. The elevated surfaces of stones on the seafloor caused the hyperbolas in the pSES data. The incident wave is diffracted at the edges of the stones similar as it is known from manmade structures like pipelines at the sediment surface as described by several authors e.g., [19
]. Müller and Wunderlich [44
] have shown that even objects (pipeline) of 40 cm diameter can be successfully detected due to the optimized signal to noise ratio of pSES. Dybedal and Bøe [46
] reported that one single pass over or along the object, in that case a pipeline, is sufficient for accurate detection and positioning. According to a formula for vertical resolution given by Wunderlich et al. [19
], objects of even 5 cm elevation can be detected if the pulse length is low and both the secondary frequency and the ping rate are high.
Aside from stones, the number of features that would generate hyperbolas is small in the study area. Pipelines, wrecks, moorings or rebuilt foundations are the only features that tower over the seafloor. These are usually easy to identify as manmade structures because they are normally registered in navigation charts. Geogenic features like eroded substrate (e.g., marine clay horizons or peat) could possibly generate comparable hyperbolas but so far none of these structures have been observed during ground truthing within the study area. Features that do not exist in the German North Sea, but that are potentially able to produce similar signals, include reef building fauna (e.g., Sabellaria) or manganese nodules. Lo Iacono et al. [47
] have observed impedance anomalies close to the seafloor that were likely produced by fish schools and seagrass meadows. However, these hyperbolas differ in intensity and geometry from those generated by stones. Generally, the hyperbolas of stones indicate much stronger impedance than those produced by organic material (e.g., seagrass meadows). Unknown is the effect of sessile assemblages attached on stones. Possibly, they attenuate the acoustic wave and cause less clear or intense hyperbolas. According to the high proportion of colonized stones in the study area [43
] it has to be expected, however, that most of the hyperbolas are altered. This is not the case, though. Only view diffuse hyperbolas have been observed. The authors assume that the degree of colonization plays a role. For future research it is interesting to investigate the effect of sessile fauna on the hyperbola characteristics and whether the SES approach can also be used to differentiate colonized from barren stones.
The distribution pattern of hyperbola position and manually marked stones reveals a good to perfect match (Figure 4
). However, the single beam information by oneself is insufficient to trace each single stone and to discriminate larger areas of stone coverage in high detail, especially in areas with heterogeneous sediment distribution. The information can be enhanced through a narrow grid of profiles at the cost of more ship time. Best demarcation of stony areas will be achieved when simultaneously recorded pSES and SSS will be combined. SSS backscatter data of lower resolution (e.g., 1 m, i.e., greater survey speed, wider transect distances) proved to be sufficient and can reduce time and costs.
In this study, the combination of SSS backscatter and pSES hyperbola occurrence provides a more detailed description of stone occurrence east of the PEV. Here, several authors e.g., [25
] described frequent occurrences of lag deposits and single stones. However, intensive mapping using the SSS-pSES approach reveals that the area east of the PEV can be divided in two areas: SOR I is characterized by frequent occurrences of lag deposits and stones whereas in SOR II rippled sandy gravel to gravel deposits largely without stones dominate. As yet, such a discrimination exclusively based on low resolution backscatter data was not possible.
Stones located at the seafloor surface provide important ecosystem functions and thus need to be precisely identified and mapped. Hydroacoustic mapping has proven to be an adequate cost-effective way to produce gapless maps but it lacks the ability to distinguish between stony areas and coarse sediments without stones (e.g., gravel). In this context, our study has provided a time efficient and reproducible method to demarcate stony areas in marine environments.
AGDS turned out to be impracticable to demarcate stony areas because the footprint of the linear acoustic propagating system is too large. To meet the optimum cost-effective ratio, we recommend to simultaneously record wide swath, low resolution SSS backscatter data and pSES data. As a result, SSS data provide regional precise maps of coarse sediments (including areas with and without stones) while pSES data provide information on the occurrence of exposed stones along the nadir. The pSES information about stones is not area-wide but it is dense enough to allow for extrapolation from one transect line to the next. The combination of SSS and pSES compensate for individual limitations of the respective instruments, allow a survey speed of 6 kn (or more) and prevent excessive survey and data acquisition time as well as time consuming data processing and manual stone identification. The extraction of stone signals from pSES data is easy and fast and will be automated for future surveys.