Acoustics 08 Paris

Estimation of macrophytes using single-beam and

multibeam echosounding for environmental monitoring
of arctic fjords (Kongsfjord, West Svalbard Island)
A. Krussa , P. Blondelb , J. Tegowskia , J. Wiktora and A. Tatareka
a
Institute of Oceanology Polish Academy of Sciences, Powstancow Warszawy 55, 81-712
Sopot, Poland
b
University of Bath, Department of Physics, Claverton Down, BA2 7AY Bath, UK
[email protected]

1743
Acoustics 08 Paris

Abstract: This paper presents results of a study on the spatial distribution and biomass of macrophytobentos in a
fjord of Arctic Svalbard. Kongsfjord represents a periglacial environment with intense morphodynamic
processes and rapidly progressing changes in the biotic environment, making it one of the most promising areas
to research climate impact on ecosystems. The main objective was to provide an acoustic tool for the evaluation
of benthic habitats. The 2007 field survey included systematic, co-registered, single-beam and multibeam
echosounder measurements. Acoustic observations were verified by biological samplings and observations for
the classification algorithm development and verification. Analyses of acoustic signals scattered on bottoms
covered by algae providing a map of phytobenthos distribution and biomass estimation in Kongsfjord. The
algorithms designed and tested for processing single- and multibeam data allow extracting the morphological
forms of the bottom and determine the areas covered by algae. This survey was the first instance where a
multibeam imager was used to map macrophytes in an Arctic environment, in a wide variety of depths and
ranges.

restricted algae investigations to the inner part of

Kongsfjord. The main point of interest was the euphotic
1 Introduction zone down to 25-m deep and along about 42Nm of shore
line. All acoustical equipment was mounted on a small,
aluminium-hull boat (Buster-type). Research transects were
Single-beam (SBES), down-looking echosounders have taken perpendicular to the coast and placed every 30-60m.
long been the tool of choice for underwater habitat During 11 days of effective survey time (because of the
mapping, because they are simple to use and widespread on weather conditions), we investigated more than 20 Nm
nearly all vessels. The shape of the echo carries a lot of along the shore, covering the entire southern part of the
information about bottom hardness and roughness. New or fjord. In addition, biological samples were taken in 4
developed techniques of acoustic signal analysis allow to different locations, thanks to a group of German divers,
differentiate all collected data and distinguish benthic confirming our acoustic recognition of macroalgae height
habitats using various processing methods [1-4,9]. Single- and density.
beam echosounders are good acoustic tools but have also
disadvantages, the most important for this research being a Kongsfjord, as many Arctic fjords, is surrounded by
rather small swath width, so it was reasonable to use melting glaciers and replete with underwater rocks and
multibeam echosounder for complementary measurements. icebergs which makes research conditions difficult and hard
to predict. Water state is very variable; transparent areas
Multibeam sonar systems (MBES) are the most advanced become cloudy very fast because of suspensions from
acoustical tools for remote observations and glacier rivers (a strongly limiting factor for other types of
characterizations of the seafloor. They are capable of remote sensing, such as aerial photography). The fjord’s
mapping backscattering strength coinciding with fine bottom is partly sandy and flat but can be very steep or
bathymetry, improving its usefulness to discriminate rocky in some places. This difficult environment requires
different types of seafloor habitats [5-7]. The small size of special methods of acoustic data analysis, extracting the
modern shallow-water multibeam transducers enables most significant information for habitat mapping
small-boat deployment in difficult, Arctic conditions. independently of water condition [3]. Our measurements of
All data were collected in Kongsfjord (79˚N) which is a macroalgae spatial distribution in particular showed that
part of Spitsbergen Island and is an excellent example of algae could be found even on the fjord’s bottom very close
the Arctic environment. The presence of the international to the glacier, in very poor water visibility and with dense
polar station in Ny-Ålesund makes research easier and sediments suspensions, proving the definite advantage of
safer. More importantly, it means there is good, long term acoustic techniques in these conditions.
documentation of the changing environment. Our acoustic
habitat mapping project has added an innovative approach
and interesting results to this pool of knowledge, providing 3 Data acquisition
much useful and new information about algae distribution
and biomass changes.
Until now, new and improved algorithms have been tested 3.1 Single-beam echosounder
using wavelet transformation and statistical echo
parameters as base for fuzzy logic and genetic algorithms The primary tool for this study was a Biosonics DTX
classification [7]. Efficient and precise distinction of echo single-beam echosounder (SBES). It operates at 420 kHz
signals coming from vegetation-covered and bare bottoms with a narrow 3-dB beamwidth of 5.2˚. At a range of 0.5-
allowed us to create accurate maps of the spatial 30 m and with a pulse length of 0.1ms, it performs with a
distribution of macrophytobentos in the research area. resolution of ~0.3–0.9 m (according to depth) x 0.08m,
making it an efficient device for algae investigation. The
transducer was mounted on the port side of the Buster boat
2 Methods and study area and connected to a surface unit. A GPS antenna was placed
exactly over the transducer and the signal NMEA was sent
The survey presented here was conducted in summer 2007 to the Biosonics Visual Acquisition software, which
with logistical support from the Alfred-Wegener Institute managed all single-beam data acquisition. Research
and Paul-Emile Victor (AWIPEV) Polar Station in Ny- transects were taken perpendicular to the coast, depending
Ålesund. The short time available for the field survey on the shores’ shape. Especially near the glaciers, the

1744
 Acoustics 08 Paris

bottom could be very steep, reaching 30-m depths a few one event), this did not prove a problem to calculate the
meters from the shore. In other places, rocks go down bio-areas in individual snapshots (ping by ping) and to
straight to the sea and algae grow on vertical underwater deduce the corresponding bio-volumes. The MBES was
walls, hard to investigate by SBES. successfully used in water depths as small as 0.5 m, with
As shown on the typical SBES echogram in Fig.1, algae are excellent correlation with SBES measurements and optical
acoustically visible as weaker backscattering areas on the checks in clear water, validating its shallowest deployment
hard bottom. The horizontal scale of this echogram is the ever.
pulse number, i.e. along-track position as moving at This MBES principally acquires bathymetry measurements
constant speed. The colour bar (greyscale in the printed for each beam, but backscatter strengths can also be derived
version of this article) shows the amplitude of the from the individual measurements, and the returns from the
backscattered signal in dB. The individual echoes above the centre beam(s) can create an echosounder-like profile (Fig.
seabed are in red (darker), and the smaller acoustic 2). This figure shows a typical case in very shallow water
intensities are identified as macrophytes, in green (lighter). (2.2 m). The range of the sonar can be dynamically altered
This was confirmed by visual inspection, made possible by (here to 5 m) to be optimised for the current depth. Strong
the high clarity of the water in this particular area. Results reflections from the flat seabed are clearly visible as red
of measurements conducted by the echosounder show that dots (smeared as the imaging angle increases, for the outer
the average height of macrophytes is varying here from 1 to beams). Because of the high pulse-repetition rate, 5 shots
1.5 m. Areas with bare seabed are also clearly visible. are averaged for the current display. Additional echoes in
the water column are attributed to Laminaria algae, and
confirmed with visual checks from the survey vessel.
Echoes below the seabed are in fact secondary reflections
from previous pings: smaller and more variable from ping
to ping, they can be safely disregarded.

Fig.1 Typical echogram taken with the single-beam

echosounder BioSonics DTX. Left: bottom profiles along-
track, imaged ping by ping with backscattering intensity.
Right: ping analysis (in this case ping no.1) is also done
sample by sample. Macroalgae (framed) are green (lighter) Fig.2 Typical MBES display during acquisition. The
on the hard bottom, orange (darker). number of windows is usually reduced to allow a better
visibility of the raw multibeam bathymetry (background
image). Clockwise from top, the superposed images
3.2 Multi-beam echosounder correspond to GPS navigation, centre-beam returns, and
sidescan-like processing.

The Imagenex 837 Delta-T multibeam sonar is operating

with 120 beams and working at a frequency of 260kHz. Multibeam bathymetry has already been used with success
Beamwidths are 1° across-track and 20˚ along-track. to map the coverage and bio-characteristics of seagrass, for
Innovative digital signal processing is used to optimise data example on the coast of Japan (e.g. [10]). Although their
usage from all channels to achieve the best possible study was limited to specific seagrass types, rather different
resolution at every point in the field of view. The range- from the Laminaria kelps studied here, their raw multibeam
resolution is 0.2% of the range selected by the user (i.e. 1 measurements showed strong similarities. The raw
cm at 5 m, 16 cm at 80 m). Multibeam profiles are acquired multibeam imagery acquired in the present survey was
at rates higher than 20 per second, giving an excellent similar but with a higher resolution. Fig. 3 shows an
coverage of the seabed at all depths and allowing interpreted profile, where one can note the high resolution
comparisons of series of pings (for example to detect of individual macrophytes even at far-range. The nominal
transient signals or shoals of fish). The acquisition gain was vertical resolution of the MBES at this range is 1 cm,
kept fixed at 20 dB. The MBES was pole-mounted (on the enough to resolve the broad, flat leaves typical of the
starboard side of the boat) and connected to a GPS antenna Laminaria witnessed below the boat. Interestingly enough,
mounted directly on top of the pole for accurate the thick stalks cannot be resolved (although they could be
positioning, enabling accurate co-registration with the seen on the Biosonics echosounder measurements,
SBES measurements. No Motion-Reference Unit was used, demonstrating the complementarities of the two
for logistical reasons, but as the survey was conducted only approaches). At far range, the flat leaves are still visible,
in very flat sea (Beaufort Sea States 0-2, and Sea State 3 in although their apparent tilt is more likely increased by the

1745
Acoustics 08 Paris

angle of ensonification. The out-of-range echoes are not The first step of the discrimination procedure was to divide
used during data acquisition, but their range (twice the all echoes into bare-bottom areas and areas with algae. It
depth of the macrophytes) shows clearly they are secondary was then possible to look for different height or species
reflections from the sea surface, arriving 1 ping later on the distribution of the macrophytobentos present.
multibeam receiver. Post-processing could look at their
significance as “pseudo-bistatic” echoes, similar to those
studied by [11] with echosounder data.
160 samples
1st bottom return

Samples

Fig.4. Example of the return echo from one ping of the

Fig.3 MBES echoes from Laminaria and gravelly bottom. BioSonics DTX SBES. Backscattering strengths of each
Note the resolution of individual leaves even at far range. sample value are colour-coded. The 160 samples cut out
around the first bottom return are marked by arrows.
It is possible to approximate sidescan sonar imagery by
adequate processing of multibeam data. Biological
constructs on the seabed can be accurately delineated on
sidescan imagery alone (e.g. [3, 4, 9]) and it was worth
checking what could be achieved by processing the
multibeam data to mimic sidescan sonar imagery. The
seabed looks rather homogeneous at this high frequency
(although iceberg scours or large algal clumps can be seen
in places), and macrophytes generally have very subtle
expressions, unless the range selection means they are
visible in the water column. The combinations of range and
gain best suited to detect the macrophytes in the bathymetry
were usually not ideal for the sidescan-like processing.
Fig.5. Preliminary map of algae distribution and height
The returns from the centre beam(s) can also be processed variability in Kongsfjord, based on BioSonics DTX
to give SBES-like profiles (Fig.2). Measured depths match measurements. The colourbar intensity is related to the
exactly what was observed with the other instruments, height [m] of algae, which reaches 1.5m in some places.
namely the Biosonics echosounder and the survey vessel’s
own echosounder. Individual returns correctly identify the Using theSonar5 Balk&Lindem software, a preliminary
macrophytes observed with the SBES and areas of bare map of algae distribution was computed (Fig.5).
seabed (including mud, correctly identified when also Presence/absence of algae and heights were calculated from
looking at wider ranges across-track). Current processing the backscattering strength. “First bottom” is found from
efforts are focusing on the 3-D identification of algae from maximum backscatter amplitudes and then a threshold for
ping to ping, and their matching with what is known from the tops of algae is selected after visual study of the
centre-beam returns, using traditional image processing. echograms.
These method provides good and fast results for a first
analysis of the entire, large dataset. However, parts of the
4 Data processing and classification signal from deeper places, below 20m, were often
recognized as algae although they correspond in fact to a
The single-beam echosounder writes backscattered signals muddy bottom. Preliminary results (Fig. 5) show estimated
in BioSonics DT4 format using 1728 samples per ping. algal coverage of 44% on the area investigated, with a mean
These were corrected for TVG (Time-Varying Gain) and height of around 1m.
converted into ASCII format before further processing. Taking this into consideration, more advanced classification
Each echo signal contains information from several bottom techniques were then used to find seafloor properties
returns, due to repeated reflections of the signal between significant for habitat discrimination. Referring to known
the shallow bottom and the water surface. To characterize methods [e.g. 1,3-5], we chose cluster analysis based on
benthic habitats, only the first part of the return echo is various features.
needed, so the most useful first 160 samples were cut out
from the echogram [8]. Fig.4 shows an example of the Two sets of parameters were calculated based on single
envelope of a return ping and its limitation to 160 samples. beam echo data. The first one contained statistical features
The highest peak, about –23 dB, represents part of the first [4] of each cut ping, corresponding to morphological and
echo returned from the bottom. physical properties of benthic habitats. The second set

1746
 Acoustics 08 Paris

consisted in continuous wavelet-transformation coefficients

(with a Coiflet mother wavelet of the 4th order), showing
the frequency components of signal intensity and their
spatial position [7]. All parameters were normalized and put
into a Matlab-based classification system based on fuzzy
logic and cluster analysis. The diagram of both methods is
shown in Fig.6.
So far, statistical parameters mostly used for single-beam Pings No
echosounder echo classification [1,3-5,6] were, for each
ping, kurtosis, maximum, mean value, second-order
Fig.7. Portion of an echogram with classification line. The
moment from origin, skewness, standard deviation and
dark part is a hard, rocky bottom, lighter rough shapes are
minimum, calculated and normalized. This set of features
algae. The line beneath the echogram contains 3 colours for
was divided into 3 classes during fuzzy-logic cluster
each class of data: black – bare bottom, green – high algae
analysis. Each data point belongs to a cluster to some
(over 0.6m), yellow – medium-high algae (up to 0.6m).
degree specified by a membership grade; after many
iterations, the optimal centers of the clusters are obtained
and the data set is classified for the assigned number of
classes.

STATISTICAL PREPROCESSING

kurtosis, maximum, mean,

second-order moment from origin,
skewness, standard, deviation,
minimum FUZZY CLUSTERING
Cut Out Echo

Signal Input
FEATURES EXTRACTION

CONTINUOUS WAVELET ANALYSIS

Fig.8. Portion of an echogram with classification line based
Wavelet Coefficient (Ci) = Signal*Mother Wavelet (scaled (i) and shifted)
on statistical parameters. The dark part is a hard, rocky
W l tE F t (i) Σ C2 bottom, lighter rough shapes are algae. The line beneath the
echogram contains 3 colours for each class of data: black –
Fig.6. Classification algorithm. bare bottom, green – high algae (over 0.6m), yellow –
medium-high algae (up to 0.6m).
Fig.7 shows classification result based on wavelet energy
features. A portion of an echogram is shown with a line Work on further analysis is in progress, but even now, these
beneath representing 3 classes: bare bottom, medium-high new algorithms give promising results and suggest SBES
algae (up to 0.6 m) and very high algae (over 0.6 m). These and MBES systems can be effectively used for mapping
results were validated with visual biological inspection. benthic habitats and support biological and ecological
This kind of parameters allows avoiding misclassification studies.
of muddy bottoms as algae-covered, recognizing mud
instead as a bare bottom.
Similar results are obtained from habitats discrimination 5 Discussion – Conclusion
characterized by statistical features (Fig. 8). The same 3
classes were used as previously and classification efficiency The main goal of this project was to research the range of
is similar, the difference being about 7%. These existence and biomass assessment of macrophytobentos
misclassifications are mostly between bare bottom and low using acoustical methods. Pre-processing and signal
algae, especially at higher depth. analysis allow to distinguish echoes from areas with and
Simultaneous biological samples were taken from 4 points. without algae what resulted in preliminary map of algae
They confirm the acoustical characterizations of the height and spatial distribution. Biomass estimation of algae
seafloor and gave input values for biomass estimation in could be done after finishing work on efficient algorithm
selected areas. For now, the biomass of algae can be based only on acoustical data and needs more biological
calculated from their spatial distribution, obtained from samples from areas of interest.
acoustical data, and the biological quantitative information Using earlier experiences we chose wavelet and envelope
from samples. The next step is to create methods for parameters as input arguments for fuzzy logic cluster
completely acoustically-based estimations. analysis classification. Genetic algorithms [5] were also
used for clusters center optimization using mentioned
parameters but long operating time and huge amount of
data forced us to live this method for farther work, although
first results are very promising. More SBES and MBES
signal processing and algorithms verification is needed to
create comparable results.

1747
Acoustics 08 Paris

Acknowledgments
We would like to thanks ARCFAC-026129-70 project for
financial support, German Divers Team: Max Schwanitz
(leader), Anita Flohr, Mark Olischläger, Peter Leopold for
samples collecting, and the Ny-Alesund staff for their
support, In-kind support from the European distributor of
Imagenex Inc., Hydro-Product UK, Balk&Lindem Sonar5
Software for Echosounder Signal Analysis and Biosonics
and Zbigniew Burczyński for Biosonics DTX, are all
gratefully acknowledged.

References
[1] N.G. Pace, H. Gao, "Swathe Seabed Classification",
IEEE Journal of Oceanic Engineering, 13(2), 83-90,
(1988)
[2] Preston J.M., Christney A.C., Beran L.S., and Collins
W.,T., “Statistical Seabed Segmentation-from images
and echoes to objectiv clustering”, In Proc. of the 7th
European Conference on Underwater Acoustics,
ECUA 2004, Delft, The Netherlands, pp. 813- 818
(July 2004)
[3] J. Tegowski, A. Kruss, “Parametrical Analysis of
Acoustic Echoes from Macrophytes”, In Proc. of the
7th European Conference on Underwater Acoustics,
ECUA 2006, Carvoeiro, Portugal (June 2006)
[4] J.Tegowski, N. Gorska, Z. Klusek, “Statistical Analysis
of acoustic echoes from underwater meadows in the
eutrophick Puck Bay (Southern Baltic Sea)”, Aquatic
Living Resources, 16, 215-221 (2003)
[5] S. Silva, Y.T. Tseng, “Classification of Seafloor
Habitats using Genetic Programming”, GECCO’05,
Late-Breaking Papers, Washington (June 2005)
[6] J. Simmonds, D. MacLenan, “Fisheries Acoustics”,
Blackwell Publishing, Oxford (2005)
[7] J. Tegowski, “Acoustical Classification of Bottom
Sediments”, Institute of Oceanology PAS, Sopot,
Poland(2006)
[8] I.M. Parnum, P.J.W. Siwabessy, A.N. Gavrilov,
“Identification of Seafloor Habitats in Coastal Shelf
Waters Using a Multibeam Echosounder”, Proceedings
of Acoustics 2004, Gold Coast, Australia(2004)
[9] Ph. Blondel, B.J. Murton, "Handbook of Seafloor
Sonar Imagery", PRAXIS-Wiley & Sons, 314 pp.
(1997)
[10] Komatsu T., C. Igarashi, K. Tatsukawa, S. Sultana,
Y. Matsuoka and S. Harada, Use of multi-beam
sonar to map seagrass beds in Otsuchi Bay on the
Sanriku Coast of Japan, Aquatic Living Resources,
16 (2003), 223–230
[11] Heald G.J., N.G. Pace: An analysis of 1st and 2nd
backscatter for seabed classification. Proc. ECUA,
649-654 (1996), Heraklion (1996)

1748