Using acoustic remote sensing and point samples to map and
monitor biota in the dynamic sediments of the Wash, UK.
Robert Foster-Smith* and Paul Gilliland**
* SeaMap, Centre for Coastal Management, Newcastle
University, Newcastle upon Tyne, NE1 7RU
** English Nature, Northminster House, Peterborough, PE1 1UA
Abstract
There is a two-way flow of information between remotely sensed data and point field
samples (such as grabs and video): On the one hand remote images must be interpreted
using the field data for ground truthing and assessing the predictive power of distribution
maps. On the other hand the spatial significance of point data can be fully appreciated
only when set in a broader context provided by remote sensing. Thus, both remote images
of the sea floor and point samples are necessary for understanding distribution patterns
and heterogeneity of biota across a range of spatial scales. Broad scale maps smooth
heterogeneity and give a general account of the biota over large areas of the sea floor
suitable for summarising baseline knowledge. However, heterogeneity at a fine scale
must be investigated in order to design sampling strategies for monitoring targeted
biotopes.
These issues are illustrated using the outputs from a series of surveys of the Wash (UK)
between 1997 and 2001. The first surveys were designed to give a broad coverage of the
range of sublittoral biotopes and successive surveys focussed effort on the key features of
conservation interest identified, particularly the reefs of Sabellaria spinulosa
(Polychaeta). The surveys used two complementary acoustic remote sensing techniques
(Acoustic Ground Discrimination Systems and sidescan sonar) in conjunction with grab
sampling and video. The advantages and limitations of these techniques for mapping and
monitoring are discussed. It is concluded that although the techniques can be used to
investigate distribution patterns of biota, ultimately there will always be a high level of
uncertainty that must be considered in the design of monitoring programmes to assess the
condition of biotopes found in dynamic sediment environments.
Introduction
As part of the development of the Management Scheme for the Wash and North Norfolk
cSAC it is necessary to establish the baselines and condition and compliance monitoring
programs for the interest features to determine whether conservation objectives have
been, or are in the process of being achieved. Biogenic sand reefs built by the polychaete
worm Sabellaria spinulosa are one of the features of interest as a ‘reef’ in its own right,
having recently been upgraded from being a ‘key component of subtidal mixed sediment
communities’. The Wash and its approaches are considered particularly important
because they are thought to harbour the only known well-developed, stable reefs created
by this species in the UK.
The research program for the study of these biogenic reefs has developed out of a series
of surveys carried out by SeaMap, English Nature and the Eastern Sea Fisheries Joint
Committee that started in 1997. The original surveys were designed to map the
distribution of a wide range of biotopes using acoustic remote sensing combined with
point sampling and Sabellaria spinulosa reefs were not specifically targeted. It was not
until video evidence of well developed reefs was collected from the northern margins of
the sand extraction area 107 (outside of the cSAC) that the existence of these reefs was
brought to the attention of English Nature. Although there was no direct observation of
reefs within the Wash cSAC itself (due to poor underwater visibility during the surveys),
comparison of the infaunal composition of grab samples taken from certain sites within
the Wash and the observed reefs in area 107 suggested that reefs might also occur in the
cSAC. Certainly, extremely dense populations of Sabellaria were found. However,
Sabellaria biotopes ranged from low density populations, through high density
communities but with poor reef development to well developed reefs. There was also
considerable overlap between the species composition of dense Sabellaria communities
and those with high species richness but low densities of Sabellaria. Analysis of available
data also suggested that Sabellaria is patchily distributed and/or temporally very variable
(see also George and Warwick 1985, Attrill et al. 1996 and Holt et al. 1997). This
remained to be tested but, if reefs are dynamic rather than stable, this should influence
both management objectives for maintaining this interest feature and the design of
monitoring surveys for compliance.
This paper outlines the strategy that evolved for using acoustic remote sensing techniques
to stratify sampling in areas likely to support Sabellaria and techniques to study the
patchiness of Sabellaria and associated communities.
Methods
Survey design
The survey design was based on stratified and nested sampling of sites selected on the
basis of a broad scale predictive map derived from a comprehensive but low resolution
acoustic survey. The sample sites were then intensively surveyed acoustically to
determine fine scale spatial structures. In summary the strategy was as follows:1. Broadscale mapping based on an interpretation of ground truth and acoustic data
used to highlight areas likely to support Sabellaria spinulosa.
2. Re-survey candidate sample areas (boxes with sides of 1km) using RoxAnn in
real-time to confirm their suitability or otherwise for sampling Sabellaria and the
selection of 7 boxes lying along a transect from the inner to outer Wash.
3. Sampling 10 randomly selected sites within each box using both video and grab.
4. Replicate sampling (5 grabs) at 5 selected sites
5. Intensive acoustic survey of areas using AGDS and sidescan to detect fine scale
features.
The primary acoustic technique for broad scale mapping was a RoxAnn AGDS operating
at 200 kHz. This system is based on a single beam echo sounder and its use in broad scale
mapping has been described by Foster-Smith and Sotheran (in press). It returns point
values of hardness and roughness of the sea floor for areas ensonified by the sounder as
the boat traverses the survey area. A complete coverage of an area is interpolated from
these track point data. Resolution is primarily determined by track spacing and this varied
from about 500m for the broadscale mapping phase to 150m for the intensive survey of
the box sample areas. Although AGDS are low resolution systems, they can be used to
discriminate sediment types and the digital data are readily handled by computer image
processing software. In addition, the box areas were surveyed using a Geoacoustic
sidescan operating at 400 kHz and the overlapping swaths mosaiced to give complete
coverage.
Broadscale mapping
The broadscale survey has been described fully by Foster-Smith et al, 2000. In brief, the
AGDS tracks for the Wash and neighbouring inshore areas were amalgamated from
several surveys taken over three years areas from 1996-9. The data were interpolated and
digital images of depth and the acoustic indices for roughness, hardness and acoustic
variability were imported into Idrisi™ image processing software for supervised
classification adopting techniques similar to those used for interpreting satellite and
airborne remotely sensed images. The ground truth data were point samples of video
and/or grab samples classified to biotope. The resulting distribution map showed the most
probable biotope for each pixel and served to guide selection of areas for intensive
sampling in the second phase of survey.
Re-survey candidate sample areas
The re-survey took place in 2001. Since long term stability could not be assumed,
stratified sampling based on the original distribution map had to be modified based on the
acoustic ground characteristics typical of Sabellaria reefs contemporary with the
sampling program. RoxAnn was used to prospect for suitable acoustic ground in realtime. The hardness and roughness values associated with dense Sabellaria taken from
previous surveys were used as guidance for assessing suitability of candidate sample
areas. A few video drops were carried out to confirm the sea floor characteristics where
these were in doubt. If a candidate site proved unsuitable, alternative sites were then
surveyed. In all, seven sampling areas were selected that lay along a deep water channel
that runs north east from the Wash.
Sampling in the selected sample boxes
Ten 0.1m2 Day grabs were taken at randomly selected stations, accurately located to
within 25m and these assessed visually for reef development and sediment granulometry.
The infauna were extracted by allutriation because of the generally coarse sediment and
the washings passed over a 0.1mm sieve. Analysis of all the preserved infauna was
undertaken by Dr Peter Garwood of Identichaete. Each of these grab sample sites was
also sampled with a drop down video which not only could assess the physical scale of
reef development, but also be used to gauge the patchiness of the biotopes at a broader
scale than the grab sample.
The infauna were identified to species where possible and numbers counted. The infauna
were analysed using Multi Dimensional Scaling (MDS) of Bray-Curtis similarity indices
derived from double square root transformed counts of all species identified (Clarke and
Warwick, 1994). Sabellaria numbers were also used to compare samples. For
comparative purposes, similarities between Sabellaria and non-Sabellaria biotopes were
derived from all grab sample data throughout the Wash from previous years to be used as
a ‘yardstick’ to assess the overall effect of stratification on reducing inter-sample
variability.
Spatial non-randomness was tested for the 10 samples within each box using Moran’s I.
The basis of such indices is to create two site/site matrices of (1) separation (lag) distance
and (2) similarity and then calculate the cross-product of corresponding cells in the
matrices. The value of the Moran’s index approaches -1 when the sites over a given lag
distance are more dissimilar than might be expected (negatively correlated) and +1 when
are more similar (positively correlated). The indices can be calculated for different lag
distances and this gives an indication of the way dispersion/aggregation changes with
increasing distance separating the sites. Moran’s indices have been calculated for
increments in the lag distance of 150 m up to just over 1 km. Also, Moran’s I was
calculated using Sabellaria numbers.
Replicate sampling
In addition to the above sampling program, the data from a study undertaken in 2000
were available. Five replicate grabs were taken (using the same procedures described
above) from three stations from Area 107 and two from Long Sands, all close to or within
the boxes selected in 2001. The similarity between replicates was measured using BrayCurtis similarity indices.
Intensive acoustic survey of areas
The AGDS and sidescan were run together at a track spacing of about 150m and
additional AGDS data were collected during sampling. The spatial correlation between
AGDS data was investigated using variogram analysis (Buroughs and McDonnell, 1998)
and the AGDS data were also used in conjunction with the sample data to produce
interpreted maps of the predicted distribution of Sabellaria and sediment classes.
Results
Broadscale mapping
The broad scale interpretation of pre-2001 AGDS data is shown in Figure 1. The map
does not show all the minor biotopes for the sake of simplicity. The major distribution of
Sabellaria lay along the edges of the deep channel running north east from the Wash and
more particularly in the outer Wash. The map from the broad scale survey was used to
select the approximate location of the sample boxes (super-quadrats). The boxes were
selected on the basis of maximum probability of the occurrence of Sabellaria at high
densities.
Two important points emerged from analysis of the pre-2001 data: (1) The majority of
the Sabellaria sites were between 65% and 80% similar to each other as compared to the
much wider range within the non-Sabellaria biotopes (0-75%); (2) many sites with lower
densities of Sabellaria were similar in species composition to the high density Sabellaria
sites.
Figure 1. The 7 sampling boxes superimposed on the predicted infaunal and epifaunal
biota. The former is shown by the background colour and the latter by the hatch pattern.
The inset gives the location of the Wash in the United Kingdom.
Prospecting for Sabellaria
The AGDS acoustic roughness and hardness values associated with dense Sabellaria
from the pre-2001 data were not well defined, although in general the hardness values
were lower than the corresponding roughness values than for most other biotopes. On the
basis of these parameters and video evidence one candidate area was rejected since the
habitat had apparently changed from gravel to coarse silt-free sand. The alternative site
(box 4) proved to have extensive areas of reef. The inner most site (box 6) was not
acoustically suitable but included to complete the transect and was chosen as the most
appropriate site in the locality. There was also observations of Sabellaria from this
general locality in recent years. An additional site (box 7) was also selected to include a
site without the licensed sand extraction area 107. No reefs were observed in box 1 in
2001 where well-developed reefs were observed from 1996-2000 and it was suggested
that dredging activity may have been a factor in the apparent change.
Community description and broad scale trends
The MDS plots for the 70 grab samples from 2001 are shown in Figures 2 a & b. The
plots have been overlain on a contoured plot of the overall similarity of each site to a
single reference yardstick ‘site’ derived from the average composition of all sites where
Sabellaria was present at densitites about 200/ 0.1m2. The purpose of the contour plot is
to show the relative similarity of the various samples. The samples themselves have been
coded to show box number (Figure 2a) and Sabellaria density (Figure 2b).
Figure 2. MDS plots of species dissimilarity for all samples coded to show Box (left)
and Sabellaria density. Contours are levels of similarity to a yardstick derived from
the average faunal composition of samples with dense Sabellaria .
Most of the samples from each box were tightly clustered and were equally similar to the
yardstick reference composition with considerable overlap between samples from
different boxes. There was even greater overlap in the distribution of Sabellaria amongst
the samples. Box 6 was clearly distinct both in terms of faunal composition and
Sabellaria density. Thus, the faunal composition of the samples within boxes 1-5 and 7
were very similar, despite the high variability of the Sabellaria densities (it must be
remembered that the numbers individuals have been double square root transformed
analysis and this will reduce the influence of large numbers of a few species on the
outcome of the MDS analysis).
Note that boxes 1, 2 and 5 have a single outlier apiece. These outliers, when plotted
geographically, also lie towards the outside of the boxes.
The proportions of the main community types represented in the boxes have been shown
in Figure 3 (one extra location has been included further into the Wash than box 6 to
extend the transect) and there would appear to be some major progressive changes in
faunal composition from the outer to the inner sample areas. Ampelisca (Amphipoda)
communities were more typical of offshore sites whilst Ensis americanus (Bivalvia: a
small, introduced species) and Scoloplos (Polychaeta) were more characteristic of inner
sites. Sabellaria was little found in the inner Wash, but abundant in boxes 7 and 1-4.
Given the broad biotope distribution patterns (Figure 2) it is hardly surprising that trends
exist over such large distances. However, the major change along the transect would
seem to occur between boxes 5 and 4. Box 6 is obviously different from the remaining
boxes and the inclusion of site 8 reinforces the distinction between inner and outer Wash
sites.
Figure 3. Trends in the proportion of major community types along the transect. Note that
an extra site has been included from 1999 samples in the inner Wash for the purpose of
extending the trends observed in the boxes. The Figure is purely illustrative and must be
viewed with caution when drawing conclusions about trends.
Variability of infauna and Sabellaria density between samples
The variability between samples within each box is summarised in Table 1. The
variability is high, especially in the densities of Sabellaria. The average infaunal
similarity was 64.9%
Table 1. Summary statistics for each box: (a) the average similarity between
pairs samples within each box and (b) the density of Sabellaria
Site
1
2
3
4
5
6
7
Similarity of infauna Sabellaria density
Mean
Standard
Mean
Standard
deviation
deviation
64.45
12.39
229.4
212.1
57.64
13.57
87.8
96.9
71.02
3.82
123.1
61.1
66.21
4.98
249.8
166.4
60.28
12.51
114.6
129.8
64.32
4.65
2.4
4.7
66.09
3.12
95.7
127
This high variability might be expected due to the separation between the samples.
However, variation within replicate samples (the average similarity for all sample was
61.8%) is also high (Table 2).
Table 2. Average similarity (between pairs of 5 replicate grab samples
taken at 5 stations at 2 sites in 2000).
Station
1
2
3
4
5
Site name
Longsands
Longsands
107 south
107middle
107 north
Average similarity
51.6
66.0
63.0
68.8
59.4
Standard deviation
10.0
5.6
3.2
2.0
3.3
The distribution of samples are shown coded according to similarity of infaunal
composition to a ‘yardstick’ derived from samples with high Sabellaria densitites (Figure
4a) and densities of Sabellaria (Figure 4b). There is little evidence of any strong spatial
distribution patterns (with the possible exception of box 1).
Figure 4a
Figure 4b
Figure 4. Grab samples plotted within each of the 7 boxes colour coded to show (a)
similarity of infauna to a ‘yardstick’ derived from the average composition of dense
Sabellaria samples and (b) Sabellaria numbers in each grab.
There was also lack of evidence for spatial pattern (either clustering or over dispersion)
from the analysis of Moran’s I statistic over different lag distances. Thus, all graphs of
infaunal similarity tend towards very slight negative values as lag distance increases
(Figure 5). The initial variability at small lag distances is due to the low number of pairs
of sites that occur with these small separations and are not greatly significant. This
indicates that dispersion is much what would be expected by chance. In other words,
there is no detectable pattern to the distribution of the infauna at the scale of the sampling
and the best description of the boxes is their mean and variance.
Dispersion of similarities
0.4
0.2
Moran's I
0
-0.2
-0.4
-0.6
-0.8
0
200
400
600
800
1000
Lag distance (m)
Figure 5. Moran’s I calculated for each of the 7 boxes for lag distances ranging from
150m to 1050m.
Moran’s I can be calculated using Sabellaria numbers and these have been shown for
each box separately in Figure 6. Once again, Moran’s I tends towards a slight negative
value at the larger lag distances and at small lags (where the significance of the indices is
low because of the smaller number of pairs in the calculation) I is very variable. A few
sites show a gradual decrease in I which might indicate some positive correlation at small
lags. But the highest values are not large (approximately 0.25) and it is doubtful if the
trend is significant.
Dispersion of Sabellaria
1
Moran's I
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0
200
400
600
800
1000
Lag distance (m)
Figure 5. Moran’s I calculated for each of the 7 boxes for lag distances ranging from
150m to 1050m.
Patterns in AGDS data
The pattern of association between the hardness and roughness values and density of
Sabellaria where track points lie within a 25m buffer radius of the sample points were
used to interpret the complete AGDS data set (Figure 6). The interpreted track data
accord fairly well with the ground truth records. However, the tracks were very variable
and few boxes showed clear spatial patterns.
Figure 6. Track data classified to show likely associated Sabellaria densities. Actual
densities from grab samples are superimposed.
AGDS data are more closely related to the physical habitat than biota and a clearer
pattern might emerge from an interpretation of sediment type. The AGDS data have been
interpolated and interpreted from grab and video data to show probable distribution of
sediment (Figure 7). Many of the patterns in the boxes relate to the general northeast/south-west trend of the deep water channel. However, box 3 was relatively
homogeneous whilst the sediments in box 4 were very patchily distributed.
Figure 7. Interpolated AGDS track data have been classified according to sediment
categories.
Patterns of spatial correlation (the variograms) show that in most boxes there was a high
level of variability at the minimum track point lag distance (the so-called ‘nugget’ effect)
and in most cases the variance rose sharply over lag distances of between 100m and
200m. The exact nature of the variogram was related to the interpreted pattern of
Sabellaria and sediment distribution. Example variograms are shown in Figure 8. Box 1
shows a strong pattern that crosses the trend that crosses north-east/south-west trend and
the variance continues to rise over all lag distances. Box 4 illustrates the effect of fine
scale variability throughout the sample area and box 3 is typical of uniform ground.
Variograms for roughness index
0.04
Variogram
0.035
0.03
Box 1
0.025
Box 3
0.02
Box 4
0.015
Box 5
0.01
Box 6
0.005
0
0
200
400
600
800
1000
Lag distance (m)
Figure 8. Example variograms for the AGDS roughness index showing the decline in
spatial correlation as lag distance between pairs of track points increases.
Sidescan and video data: Very fine scale patterns
Sidescan images can be of high resolution and have the potential to detect fine scale
features. The separate track images were mosaiced and boundaries digitised around
discernible fine scale features (Figure 9). Many of the boxes were featureless, consisting
of an even ‘graininess’ typical of gravel and sand or cobble and gravel. Dredge marks
were characteristic of the northeast section of box 4 with a clear cut-off coincident with
the boundary of the 107 licensed aggregate extraction area. Boxes 7 and 2 showed no
obvious features. Boxes 3, 4, 5 and 6 all had features ranging from small ribbons to small
waves aligned northeast/southwest along the line of the transect. The video recorded
waves of gravel alternating with cobble/gravel troughs. Box 4 contained some large
gravel waves aligned northwest/southeast and these were clearly visible on the video.
These waves appear to be encroaching on level cobble and gravel that supported
Sabellaria reefs.
Figure 9. Track data classified to show likely associated Sabellaria densities.
Conclusions
Ecological patterns and their underlying processes cross scales from the very fine to the
broad biogeographic. It is important to appreciate how the tools we have for observing
the benthic environment limit our perceptions of pattern not only from the viewpoint of
theoretical ecology, but also for the design of survey programs for monitoring significant
change in the environment (Figure 10).
This study has applied a range of techniques for remote sensing and direct observation
with varying powers of coverage, resolution and discrimination. The nature of the benthic
environment in the study area is challenging in that there are no distinct or static
boundaries between habitats. Instead, the habitat is patchy at very fine scales and subject
to hydrodynamic processes that create mobile seabed features. The way in which the
infauna interacts with the environment is also variable and this creates high levels of
uncertainty in interpreting the distribution patterns of biota. For example, Sabellaria
spinulosa has a wide tolerance of sediment types, although favouring silty cobble and
gravely sand. These habitats also support a wide range of other fauna (including dense
epifauna) and sites with both very high and low densities of Sabellaria often share similar
infaunal communities. Thus, grab samples taken in close proximity to each other showed
a relatively high level of variability and this could be interpreted as a consistent level of
patchiness rather than homogeneity. This high level of variability would hide spatial
patterns until larger distances separate samples (see also Morrisey et al., 1992; Kendall
and Widdicombe, 1999). Thus it is unlikely that grab sampling will be successful in
detecting fine scale spatial patterns in faunal composition given the inherently high levels
of variability at the maximum sampling resolution available in this survey. However,
acoustic remote sensing shed light on the heterogeneity of the physical habitat.
Physiographic features & Broadscale trends
Scale of
seabed
features
Topographic features
Sediment features
Sediment textures
Reef patches
Scale of
observation
AGDS
Sidescan
Video
Grab
Point data
not easily
mapped as
continuous
coverages
Complete coverage,
good resolution and
positional accuracy.
Repeatable.
Broadscale mapping
with poor resolution
and accuracy with
poor repeatability.
Scale of
change
Patch
dynamics
Local sediment change
Regional habitat
change
Scale
1m
Scale of
mapping
10m
100m
1,000m
10,000m
Figure 10. Summary of the correspondence of the scale of observation of the different
remote sensing and direct sampling techniques and the scale of environmental features.
AGDS has been used as a rapid tool for broad scale survey to focus down onto areas
likely to support Sabellaria. Because of the high levels of variability in the distribution of
Sabellaria and its associated communities, the predictive capability of the remote sensing
techniques is not precise. However, the techniques increased the chances of finding
Sabellaria within a sampling area of 1km. Patterns in the distribution of Sabellaria at a
finer scale were harder to detect. There were patterns in the AGDS data, the interpreted
sediment distribution and seabed features (as revealed by the sidescan sonar and video).
However, the wide sediment tolerance of Sabellaria and together with its inherent
variability has meant that associating fine scale distribution with these patterns is difficult
without high precision tools for locating grab samples within these seabed features.
Is it likely that these fine scale patterns are of significance for monitoring the overall
condition of the status of Sabellaria in the Wash and its environs? This question is based
on a much wider spatial context than the fine scale pattern. Statistical grab sampling of
large quadrats backed up by acoustic remote sensing may provide a robust way to
monitor broad scale change that is not subject to fine scale fluctuations in population.
Acknowledgements
The authors would like to thank the Eastern Sea Fisheries Joint Committee for their
continued support over the years, especially to the crew of Surveyor . They would also
like to thank the Crown Estate for financial support. Dr Peter Garwood provided and
excellent service in the identification of the fauna collected in the grab samples.
References
Attrill, M.J., Ramsay, P.M., Thomas, R.M. and Trett, M.W. (1996). An estuarine
biodiversity hot-spot. Journal of the Marine Biological Association of the United
Kingdom, 76, 161-176.
Burroughs, P.A. and McDonnell, R.A. (1998). Principles of Geographic Information
Systems. 333pp. Oxford: Oxford University Press.
Clarke, K.R. and Warwick, R.M. (1994). Change in marine communities: an approach to
statistical analysis and interpretation. 144pp. Natural Environmental Research Council,
UK.
Foster-Smith, R.L. and Sotheran, I.S. (In Press). Mapping marine benthic biotopes using
acoustic ground discrimination systems. International Journal of Remote Sensing.
George, C.L. and Warwick, R.M. (1985). Annual macrofauna production ian a
hardbottom reef community. Journal of the Marine Biological Association of the United
Kingdom, 65, 713-735.
Holt, T.J., Hartnoll, R.G. and Hawkins, S.J. (1997). Sensitivity and vulnerability to man
induced change of selected communities: intertidal brown algal shrubs, Zostera beds and
Sabellaria spinulosa reefs. English Nature: English Nature Research Report 234
Kendall, M.A. and Widdicombe, S. (1999). Small scale patterns in the structure of
macrofaunal assemblages of shallow soft sediments. Journal of Experimental Marine
Biology and Ecology, 237, 127-140.