Progress in Oceanography 117 (2013) 106–117

Contents lists available at ScienceDirect

Progress in Oceanography

journal homepage: www.elsevier .com/locate /pocean

Fish behaviour in response to tidal variability and internal wavesover a shelf sea bank

0079-6611/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.pocean.2013.06.013

⇑ Corresponding authors. Address: School of Biological Sciences, Institute ofBiological and Environmental Sciences University of Aberdeen, Tillydrone Avenue,Aberdeen AB24 2TZ, UK. Tel.: +44 (0)7977 064939; fax: +44 (0)1224 272396 (C.B.Embling), tel.: +44 (0)1224 273257; fax: +44 (0)1224 272396 (B.E. Scott).

E-mail addresses: clare.embling@plymouth.ac.uk (C.B. Embling), jons@liverpool.ac.uk(J. Sharples), e.armstrong@marlab.ac.uk (E. Armstrong), rolm@noc.ac.uk (M.R.Palmer), b.e.scott@abdn.ac.uk (B.E. Scott).

C.B. Embling a,b,c,⇑, J. Sharples d,f, E. Armstrong e, M.R. Palmer e, B.E. Scott a,⇑a School of Biological Sciences, Institute of Biological and Environmental Sciences, University of Aberdeen, Tillydrone Avenue, Aberdeen AB24 2TZ, UKb Marine Institute, University of Plymouth, Plymouth PL4 8AA, UKc Centre for Ecology and Conservation, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UKd School of Environmental Sciences, University of Liverpool, Liverpool L69 3GP, UKe Marine Scotland Science (MSS), Marine Laboratory, PO Box 101, Victoria Road, Aberdeen AB11 9DB, UKf National Oceanography Centre, Liverpool L3 5DA, UK

a r t i c l e i n f o a b s t r a c t

Article history:Available online 28 June 2013

The availability of fish to top mobile predators such as seabirds and fishermen and to fisheries surveysdepends on their schooling behaviour. In temperate coastal ecosystems, tide is the main driver of ecosys-tem dynamics. To investigate the mechanisms linking fish schooling behaviour to bio-physical oceanogra-phy in a tidal ecosystem, we combined novel fine scale survey data of fish behaviour using fisheriesacoustics with co-located oceanographic data from contrasting topographical locations. The schoolingbehaviour of pelagic fish was examined in relation to bio-physical oceanography by carrying out repeatedsurveys of the same locations over the daily tidal cycle at spring and neap tides over a small bank (JonesBank) and a nearby flat region. Different fish species were found over the bank and in the flat region14 km to the south-east, and fish schooling behaviour varied similarly. Results showed that the bank playedan important role in influencing fish distribution and behaviour, with shallow pelagic schools concentratedover the bank and closer to the surface at times of high internal wave activity. Behaviour of schools iden-tified close to the bottom was partially influenced by oceanographic variables that drove internal wavesover the bank slopes, but were most strongly influenced by the spring–neap cycle. These deep schools werelarger, closer to the bottom and less dense at neap tides. Identifying some the of the bio-physical variableswhich drive changes in fish distribution and behaviour will help quantify fish catchability and assist in thedesign and interpretation of better targeted surveys for top predators and prey fish species.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The availability of fish to predators and to fisheries surveys isdependent on many aspects of fish behaviour (Fréon et al., 1993;Perry et al., 2000). Fish behaviour can change diurnally (Benoîtand Swain, 2003; Johnsen and Godø, 2007), tidally (Michalsenet al., 1996; Embling et al., 2012), or correlated with environmentalvariables such as chlorophyll concentration (Robinson, 2004) andoxycline depth (Bertrand et al., 2010, 2011; Grados et al., 2012).Fishermen adapt their fishing techniques to take advantage ofthe behaviour of their target prey, for instance by fishing with tidal

currents to maximise catches (Perry et al., 2000). Changes in fishbehaviour have also been shown to have a significant effect onthe foraging behaviour of other top predators such as seabirds(Embling et al., 2012). For example, sandeel schooling behaviourwas shown to change with tidal currents making them more acces-sible to surface feeding kittiwakes, and it was hypothesised thatthis was topographically driven (Embling et al., 2012). Regions ofsteep seabed topography, such as seamounts and the shelf edge,can provide a habitat that attracts pelagic fish and other predatorsby producing aggregations of plankton prey species (e.g. Genin,2004) or by supporting a suitable food resource (e.g. Sharpleset al., 2009). On smaller spatial scales, internal waves over theedges of banks in shelf seas can enhance surface aggregation ofprey species resulting in spatially focused foraging by top preda-tors (Stevick et al., 2008).

Jones Bank is a relatively small (20 km � 5 km) bank within theCeltic Sea over which internal waves have been observed (Sharpleset al., this issue-b), and so provides an ideal location from which tostudy the effect of topography and internal waves on the schoolingbehaviour of pelagic fish, and hence the implications for top

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 107

predators such as fishermen and seabirds. The Celtic Sea is animportant fishing ground for vessels from France, Ireland, UK,Spain and Belgium (Marchal and Horwood, 1996). It is a relativelynew fishery; having only developed since the Second World War,yet despite this it has already shown changes in the trophic leveland size of both surveyed and landed fish due mainly to fishingactivities (Pinnegar et al., 2002; Blanchard et al., 2005). The mainfish species landed have changed from being predominated by hightrophic level fish such as cod and hake (e.g. Merluccius merluccius)to lower trophic level zooplanktivores such as horse mackerelTrachurus trachurus, and blue whiting Micromesistius poutassou(Pinnegar et al., 2002). Jones Bank within the Celtic Sea (Fig. 1)has high fishing effort in the summer months as shown from VesselMonitoring System (VMS) data (Sharples et al., 2013a). Fisherieslandings data show that the areas around Jones Bank support afishery of mainly benthic fish such as anglerfish Lophius spp., andmegrim Lepidorhumbus whiffiagonis, as well as demersal fish suchas hake M. merluccius (Sharples et al., this issue-b; Martinezet al., this issue).

Understanding the mechanistic links between oceanography,fish distribution and changes in fish behaviour over and around ashallow bank is key to understanding top predator foraging distri-butions, fishery data and fish survey data. It is therefore vital formany aspects of ecosystem conservation, from the ecosystemapproach to fisheries management (EAFM), to marine spatialplanning including the effects of placement of renewable energydevices, to the potential effects of climate change (Inger et al.,2009; Simpson et al., 2011). The aim of this study was thereforeto investigate how fish schooling behaviour contrasted between abank site and a flat site, including any changes driven byspring–neap and semi-diurnal tidal cycles. By combining novel

Fig. 1. Locations of the bank and flat circuits in the Celtic Se

survey data using fisheries acoustics with co-located oceano-graphic information we are able to investigate important spatialand tidal contrasts in fish schooling behaviour, and suggest the rolethat the physics may play in determining fish behaviour.

2. Materials and methods

2.1. Data collection

The surveys were conducted between 2 and 26 July 2008 fromthe RRS James Cook in the Celtic Sea over and adjacent to JonesBank (49.8–50.0�N, 7.9–8.1�W; Fig. 1). This paper concentrateson the repeated survey tracks carried out in two regions: overthe eastern edge of Jones Bank, and in a contrasting flat region14 km to the south-east of Jones Bank (Fig. 1). Surveys were carriedout at a vessel speed of approximately 8 knots, repeatedly runningcircuits around a closed loop track 15 km long and 2 km wide(Fig. 1). The speed and loop circumference allowed the same loca-tion to be sampled roughly every sixth of the tidal cycle (2 h 5 min)corresponding with previous survey methods (Embling et al.,2012). The bank circuit was surveyed continuously for 25 h at bothspring and neap tides: spring tides on 8 July 00:00–9 July 01:00(C1), and neap tides on 15 July 15:00–16 July 16:00 (C3), thus eachsurvey covering two tidal cycles and one diurnal cycle (Fig. 2). Thecontrasting flat circuit was continuously sampled during daylighthours only at both spring and neap tides: neap tides on 13 July06:30–19:30 (C2), and spring tides on 20 July 04:00–19:30 (C4),so that each survey covered at least the time over one tidal cycle(Fig. 2). All times can be assumed to be Greenwich Mean Time(GMT). During each survey, synoptic monitoring of the water col-umn and pelagic fish was carried out. Oceanographic sampling

a and overlaid on the measured underlying bathymetry.

Fig. 2. Tidal range during the survey, with the times of the circuits in relation to the tidal cycle highlighted in grey (from left to right: C1 – bank at spring tide, C2 – flat at neaptide, C3 – bank at neap tide and C4 – flat at spring tide).

108 C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117

was carried out using an MVP (Moving Vessel Profiler: C1) or aScanFish (C2–C4) undulating instrument packages, providingvertical (1–2 m resolution) and horizontal (�0.5 km resolution)information on temperature, salinity, density and chlorophyllfluorescence. The MVP could only be used in one direction on July8–9th bank circuit due to the effects of storm winds with peakspeeds of 13.4 ms�1. Current data was provided from bottom-mounted 300 kHz ADCP moorings on the top of the bank (for C1and C3) and on the control flat site (for C2 and C4) (Fig. 1). Forthe collection of zooplankton a set of 200 lm and 95 lm bongonets, were deployed using vertical tows. Two tows were taken ateach location. The locations of the stations were above the fourlocations where ADCPs were deployed such that samples rangedfrom on top the bank, down the edge of the bank and over the flatsite (see Palmer et al., 2013 for map). Spring and neap periods wereequally sampled. Each sample was stored in 4% formalin solutionand species identification and abundance estimates were pre-formed by Sea Fisheries Institute, Gdynia, Poland. Fisheries acous-tic data were collected using a keel-mounted SIMRAD EK60scientific echosounder (Bodholt et al., 1989), operating at fivefrequencies: 18, 38, 70, 120 and 200 kHz. The echosounder wasconfigured to ping at each frequency simultaneously, every 1 s,with pulse duration of 1 ms for each frequency. Data were loggedfrom the echosounder to a personal computer with Myriax’s Echo-log software (Myriax Ltd., Australia).

Fish samples were collected during the survey by commercialfishing vessels between 22 and 26 July 2008 in daylight hours atspring tides (Martinez et al., this issue). Both pelagic and benthictrawls were carried out over the bank and over the control flat site14 km to the south-east of the bank. The predominant pelagic spe-cies on the bank were the zooplanktivores horse mackerel T. tra-churus, and boarfish Capros aper. The predominant pelagic specieson the control flat site were the zooplanktivore blue whiting M.poutassou and the piscivore hake M. merluccius. It is therefore likelythat the pelagic schools detected using the fisheries acoustics com-prise these main species.

2.2. Oceanographic data processing

The data from the towed instrumentation were summarisedinto 5-min bins representing approximately 1.24 km of distancealong the survey track. This time period was selected to coincidewith the 5-min bird data counts (Scott et al., 2013). Using thetemperature, density and chlorophyll fluorescence profiles 5-min averaged parameters were defined and are presented in Ta-ble 1. Echoview was used to extract the water depth during thesurvey, this is a hydroacoustic data processing tool for visualis-ing, processing and characterising data from echosounder andsonar systems (Myriax Ltd.). The extracted depth was summa-

rised into 5-min bins of mean depth Depth (m) and variationin depth Depthvar (m2). Variation in depth is a proxy for the sea-bed slope.

2.3. Fisheries acoustics data processing

The 38, 120 and 200 kHz echosounders were calibrated using a38.1 mm, tungsten–carbide, standard target (Foote, 1987) near thestart of the survey (Foote and MacLennan, 1984). The transducerswere keel mounted at a depth of 5 m below the sea surface. Datawere analysed between 15 m below the sea surface (10 m belowthe transducers) and 0.5 m above the seabed, to exclude the nearfield effect of the transducers and bottom pixels respectively. Echo-view (Myriax Ltd.) was used to analyse the acoustics data to sepa-rate fish from plankton. This involved a linear addition of allfrequencies’ volume backscattering strength (Sv) followed by athreshold to exclude the majority of plankton ‘noise’. This tech-nique may not exclude very dense patches of zooplankton, how-ever visual checks suggested that <5% of all schools identifiedwere dense plankton patches. The Echoview school detection algo-rithm (Barange, 1994) was used on the plankton-removed data toautomatically identify and select fish schools. Parameters were se-lected such that small schools were included as well as the largerschools, with the aim of identifying the small fish echoes risingoff the seabed (see Figs. 3 and 4 inset). Multi-frequency techniquesfor species identification (e.g. Knudsen et al., 2006; Jurvelius et al.,2008; Fernandes, 2009; Korneliussen et al., 2009) were attemptedfor distinguishing between fish species. However, due to the lack ofknowledge on the frequency responses of the species found duringthe survey, and the presence of mixed species schools, we were un-able to divide schools by species. The only differentiation betweenfish schools was made by depth since there was a clear depth divi-sion between fish in the upper water column (upper 60–75 m) andthose found associated with the bottom. Therefore fish schoolswere divided either into ‘shallow’ fish schools associated withthe upper water column, or ‘deep’ fish schools associated withthe bottom depending on where the clustering occurred (C1:63 m; C2: 70 m; C3: 75 m; C4: 60 m). School characteristics (schoolheight and width, school perimeter and area, school depth, com-pactness, and Sv) were exported and averaged in 5 min bins to cor-respond to the time intervals of the oceanographic data. Sv is thevolume backscattering strength for each school at 38 kHz and isan indicator of the fish density inside each school. Compactnessis a measure of how close to a circle the school resembles: it is cal-culated as the ratio between the radius of a circle with the sameperimeter as the school to the radius of a circle with the same areaas the school (so a circle has compactness of 1). Only daylightschools were used in further analyses so that all circuits could becompared (since no night data was collected at the flat site).

Table 1Names and definitions of all the explanatorily variables used in the analysis.

Variable name Description of variable

PhysicalBankFlat Bank or flat is a factor depicting whether the data is from the Bank or Flat siteDepthmax (m) Maximum depth is the maximum depth (m) within the 5 min bin from the Scanfish or MVP instrumentsDepthvar (m2) Variation of depth is the variation in depth with the 5 min bin from the Scanfish or MVP instrumentsDThermmax (�C m�1) Mean maximum thermocline gradient defined as the mean of the 5 min of continuous data for the maximum dT/dz from the Scanfish or MVP

instrumentsDepthDThermmax (m) Mean depth of maximum gradient of the thermocline within the 5 min binSpU (m s�1) Mean current speed in the surface mixed layer (SML) above the depth of maximum thermocline gradient from either Main Station 1 (MS1) if

for the Bank Site or Main Station 4 (MS4 for the Flat site.SpL (m s�1) Mean current speed in the bottom mixed layer (BML) below the depth of maximum thermocline gradient with location same as Sp UDirU (� degrees) Mean current direction in the surface mixed layer (SML) above the depth of maximum thermocline gradient with locations same as for current

speedsDirL (� degrees) Mean current direction in the bottom mixed layer (BML below the depth of maximum thermocline gradient with locations same as for current

speedsDifSp. (m s�1) Difference between the speeds above and below the depth of maximum thermocline gradient. SpU–SpL (note this does not take account of the

direction of the current)Shear Bulk Shear. Calculated as the absolute difference in velocity between upper and lower layers

p(du/dz2 + dv/dz2). Units are s�1 but the values

are not absolute as the dz values are arbitrary as no account has been made for the difference in depth between layers.Windsp (m s�1) Mean Wind speed per 5 min bin from the HMS James CookWindir (� degrees) Mean Wind direction per 5 min bin from the HMS James Cook

BiologicalCHLmax (mg m�3) Maximum chlorophyll biomass defined as the highest concentration of chlorophyll within each 5 min bin of observations, regardless of where

it occurred vertically in the water column. Biomass values were converted from the Scanfish samples using the following equation Chl = Chl(Scanfish voltage) � 1.612–0.0758. As there were few chlorophyll samples take during the limited use of the MVP chlorophyll biomass valueswere calculated by deriving a conversion factor from the relationship for chl biomass from the Scanfish data. The conversion factor was theaverage of the maximum MVP/maximum Scanfish converted value + minimum MVP/minimum converted Scanfish value

DepthCHLmax (m) Mean depth of maximum chlorophyll biomass (m) per 5 min bin

TemporalTime (decimal) Time of day converted 0:00 to 24:00 to a decimal from 0 to 1SpringNeap Spring or Neap (a factor) The Spring period is defined as the 7 days within a 14 day period with maximum tide speeds. Date with maximum

speeds is normal 3 days after either full or new moon such that dates of spring tides will cover the dates 3 days before and after thatmaximum. For neap tides the dates straddle the date of minimum tidal speeds normally 3 days after half moons each 14 day period

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 109

2.4. Statistical data analysis

Principal Components Analysis (PCA) was used to reduce thenumber of fish school characteristics to a smaller number of syn-thetic variables that represented the dominant behaviours for bothdeep and shallow schools (based on: Boon et al., 2007; Dingemanseet al., 2007; Martin and Réale, 2008). This technique is widespreadwithin animal behaviour literature, and used to reduce the numberof behaviours recorded in a wide range of animals from fish(Dingemanse et al., 2007) to squirrels (Boon et al., 2007) and birds(Patrick and Browning, 2011) in order to relate behaviour to lifehistory (Boon et al., 2007), populations (Dingemanse et al., 2007)or environment (Minderman et al., 2009). Similar to Dingemanseet al. (2007) we used the PCA to reduce the behavioural variablesto one or two ‘summary’ behaviour variables that were then usedas response variables in a mixed model. The school characteristicsof all schools in all circuits were used in the PCAs, however PCAsfor each individual circuit produced very similar components vali-dating the combining of all circuits in the analyses. The dominantschool behaviours were the grouped combination of school charac-teristics based on the principal component (PC) score. Only the firsttwo components were retained to limit the number of responsevariables, and only one PC was retained if it explained >70% ofthe variance (Jolliffe, 2002).

The two dominant fish schooling behaviours for both deep andshallow schools were used as response variables in GeneralisedLinear Models (GLMs) to determine which of the physical, biolog-ical or temporal variables presented in Table 1, influences fishschooling behaviour. Since the data were collected from repeatedsurveys of the same locations over time, the data were highlytemporally and spatially autocorrelated, so the autocorrelationstructure was modelled using Generalised Estimating Equations(GEEs) using an ar1 autocorrelation structure. A Poissondistribution was assumed for the distribution data (number of

schools per 5 min bin), and a Gaussian distribution was assumedfor the school behaviour response variables. Forward and back-ward step-wise model selection was used to select variables: add-ing variables stepwise to the null model (of no predictor variables)if they were significant (p < 0.05) and maximised the Wald statistic(forward), or removing variables that were not significant thatminimised the Wald statistic (backward). The best model includedonly the variables that were significant in both forward and back-ward step-wise selection models.

The same methods were used to model the distribution of fishschools (number of fish schools per 5 min bin) with the predictorvariables. Models were built separately for each location (bank orflat), and for each fish school depth division (schools in the upperwater column i.e. ‘shallow’ schools and schools associated withthe bottom, i.e. ‘deep’ schools). Therefore there were 12 modelsbuilt (4 groups of models of distribution, 4 for the first behaviouralPC, and 4 for the second behavioural PC). All analysis was carriedout in R version 2.12.0 (The R Statistical Foundation 2010) usingthe ‘geepack’ library for carrying out the GLM-GEE models(Halekoh et al., 2006).

Where current direction was significant in a model, and if auto-correlation was low, the variable was modelled within a General-ised Additive Model (GAM) using the ‘mgcv’ library in R (Wood,2006) and a cyclic smoothing function (ts = ‘‘cc’’).

3. Results

3.1. General comparisons between circuits

At the start of the survey (the first 5 days) there was a stormwith winds exceeding 10 ms�1 that had a significant effect on themixing of the upper water column and hence on lower trophiclevels (Palmer et al., 2013; Davidson et al., 2013). This storm-driven

Fig. 3. EK60 echograms for 38 kHz at 70 dB for the bank slope region at (a) spring tides on 8 July 2008 between 10:35 and 11:10 (GMT); (b) neap tides on 16 July 2008between 10:10 and 10:45 (GMT) (note that this neap circuit was carried out in the opposite direction to the spring tide circuit). Higher echo returns from fish schools areshown in green–yellow–red, plankton or lower density fish schools are shown in dark blue. The thermocline is found at the base of the dense dark blue plankton layer around40 m. The inset figure shows the schools as identified by the school detection algorithm.

110 C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117

mixing resulted in the top of the thermocline being eroded but didnot affect the depth of the base of the thermocline. This resulted in aweaker SCM (lower CHLmax) at the beginning of the survey thanlater in the survey when conditions improved (Davidson et al.,2013). The differences that such high winds have had on the overallresults are addressed in the discussion.

There were generally more fish schools at neap than at springtides, with most of the shallow fish schools concentrated in thevicinity of the bank (Tables 2 and 3, Figs. 3 and 4). Based on visualinspection of the echograms, there also appeared to be very highplankton density above and within the thermocline at spring tidesat both sites (Figs. 3a and 4a). Apparent differences in plankton dis-tribution through the water column could also be observed at theflat site, with seemingly high plankton throughout the water col-umn at neap tides, but much lower (other than in the thermocline)at spring tides (Figs. 3 and 4). The hydroacoustics also suggested acontrast in the distribution of zooplankton from the shelf edge tothe bank region with high levels of zooplankton mixed throughoutthe water column over the top of the bank, a feature which was notapparent away from the bank (Fig. 5). However the vertical towsdata showed no significant differences in mesozooplankton num-bers associated with the bank and off-bank sites suggesting distri-butional rather than abundance changes over spring neap cycles.As an example, Table 4 summarises the mean and standard devia-tions of total copepods. Comparison of either all stations on thebank and all stations off the bank, or all stations on the bank andstations downstream of the bank in the mean flow, show no signif-icant contrast in either copepod numbers or in numbers of copepod

eggs. A similar lack of significant contrasts was seen for all otherzooplankton taxa.

Fish schooling behaviour varied both on and off the bank andwith the spring–neap cycle (Table 3). Deep and shallow schoolsshowed opposite trends with the spring–neap cycle: deep schoolswere closer to the bottom, of larger size and denser at neap tides,whereas shallow schools were closer to the surface and of smallersize at neap tides (Table 3). The main difference in school behav-iour between the bank and flat site was exhibited by the shallowschools which were of larger size in the bank circuits than inthe flat circuits (Table 3). These differences can be seen visibly inthe EK60 echograms, with numerous small fish schools close tothe bottom on the flat site at springs (Fig. 4a), but dense aggrega-tions at >100 m depth at the flat site at neap tides (Fig. 4b).

The EK60 also allows for clear visualisation of the progression ofinternal wave generation over the bank edge at spring tides (Fig. 6).As the bottom current speeds increase towards the south (off-bankin Fig. 6), there is a gradual depression over and beyond the bankcrest followed by a rapid increase in thermocline height on theflanks of the bank. This is due to hydraulic control of the flowand is dependent on high flow speeds and water depth (describedin detail in Palmer et al., 2013). The thermocline depression pro-gresses into high-amplitude internal waves (40–50 m high, equiv-alent to half the water depth). During this period the lower watercolumn current velocity is at a maximum and that in the upperwater column current is close to minimum resulting in a strongshear layer in the thermocline. As the lower layer flow decreasesthere is a reduction of the amplitude of oscillation of the thermo-

Fig. 4. EK60 echograms for 38 kHz at 70 dB for the flat region at (a) spring tides on 20 July 2008 between 11:15 and 11:45 (GMT); and (b) neap tides on 13 July 2008 between09:45 and 10:05 (GMT). Higher echo returns from fish schools are shown in green–yellow–red, plankton or lower density fish schools are shown in dark blue. The thermoclineis at the base of the dense blue plankton layer around (a) 40–50 m, and (b) 30–40 m depth. The inset figure shows the schools as identified by the school detection algorithm.

Table 2Summaries of 5-min bin data included in the analysis (median and interquartile range).

C1 (Bank springs) C2 (Flat neaps) C3 (Bank neaps) C4 (Flat springs)

Deep Schools/bin 2.2 32.7 17.2 4.8Shallow Schools/bin 2.3 0.5 2.4 0.1Topography

Mean Depth (m) 112 (106–114) 123 (122–124) 111 (98–115) 123 (122–125)Depthvar (m2) 0.6 (0.3–1.4) 0.1 (0.1–0.3) 0.9 (0.3–2.1) 0.3 (0.2–0.5)

Number of 5 min bins 53 114 144 168

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 111

cline (Fig. 6) and thermocline shear decreases in intensity. Thistransition can occur suddenly in the form of a hydraulic jump asthe strong lower water column flow is released from hydrauliccontrol (Palmer et al., 2013).

3.2. Principal Component Analysis of fish schooling behaviour

The first two principal components for both deep and shallowschools, explaining 63.4% and 71.6% of the variance respectively,were retained from the PCA (Table 3). The first component in bothcases describes school morphology (school width, height, perime-ter and area) (Table 3). However, the polarities were opposite for

deep and shallow schools: thus an increasing first componentcorresponds to a decreasing deep school size but an increasingshallow school size. The second component in both school typeswas bipolar with Sv38 (school density) opposing school depth (Ta-ble 3): thus an increasing second component corresponds to an in-crease in school density but a decrease in school depth (seeTable 5).

3.3. Mixed models of fish distribution with bio-physical characteristics

In all models of fish school distribution, except for shallowschools in the bank circuits, the spring–neap cycle was the most

Fig. 5. EK60 at 120 kHz and 75 dB over the entire span of Jones Bank travelling from the direction of the shelf edge to the southwest to the northeast end of the bank on 26July 2008 05:30–17:30 (see inset map showing the cruise track overlaid on bathymetry). Bottom echo shown as deep red, echoes from higher density organisms (such aszooplankton) within the water column are shown in green–yellow–red.

Table 3Summary of fish school characteristics by circuit (median and interquartile range).

C1 (Bank springs) C2 (Flat neaps) C3 (Bank neaps) C4 (Flat springs)

Deep SchoolsSv38 (dB) �54 (�56 to �51) �57 (�60 to �55) �56 (�58 to �55) �53 (�54 to �51)Depth (m) 81 (75–89) 114 (107–120) 108 (101–113) 109 (99–114)Length (m) 13 (9–15) 17 (12–28) 17 (12–33) 12 (10–16)Thickness (m) 1.5 (1.2–1.9) 1.5 (1.3–2.7) 1.7 (1.3–3.6) 1.5 (1.2–2.1)Perimeter (m) 32 (22–46) 47 (34–88) 50 (29–120) 34 (26–45)Area (m2) 10 (8–15) 13 (8–29) 13 (8–38) 10 (7–16)Compactness 9 (�1 to 21) 23 (9–43) 26 (11–51) 6 (�1 to 19)N 115 3723 2461 802

Shallow SchoolsSv38 (dB) �47 (�50 to �45) �47 (�51 to �43) �47 (�50 to �45) �48 (�50 to �45)Depth (m) 39 (32–46) 29 (20–35) 35 (29–48) 39 (37–50)Length (m) 14 (9–20) 8 (4–12) 12 (8–21) 8 (8–11)Thickness (m) 5.4 (2.7–8.8) 2.3 (1.5–4.2) 4.8 (1.9–8.6) 3.1 (2.1–4.7)Perimeter (m) 53 (28–101) 22 (12–36) 43 (20–103) 30 (20–40)Area (m2) 47 (17–104) 12 (7–30) 36 (10–102) 17 (13–47)Compactness 6.0 (2.3–12.6) 3.0 (1.5–4.9) 5.4 (2.4–10.5) 2.7 (�1 to 4.9)N 120 62 346 18

Table 4Means and standard deviations for copepods and copepod eggs from whole watercolumn bongo net hauls (95 lm mesh). ‘‘Bank’’ stations were at MS1 and MS2, ‘‘OffBank’’ stations at MS3, MS4 and MS5. Numbers for ‘‘Downstream’’ are for samplescollected at MS4 only.

Bank Off Bank Downstream

Copepods (�104) m�2 11.9 ± 4.8 (n = 9) 11.4 ± 5.3 (n = 8) 11.4 ± 5.8 (n = 4)Copepod eggs (�104)

m�21.3 ± 0.9 (n = 6) 0.9 ± 1.0 (n = 5) 1.2 ± 1.2 (n = 3)

112 C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117

important predictor of fish schools: there were fewer fish schoolsat spring than at neap tides (Tables 2 and 6). The most significantpredictor of shallow fish schools in the bank circuit was waterdepth, with the number of schools increasing as depth decreased

(Table 6). This implies that the highest number of shallow schoolswas found over the top of Jones Bank rather than over the bankslopes (the top of the bank was around 65 m depth). Also CHLmax

and DifSp were significant in explaining the distribution of shallowschools, with the number of schools reducing as CHLmax increased,and with higher numbers of schools where the bottom current wasfaster than the surface current (Table 6). DifSp does not account forcurrent direction, so it is not equivalent to Shear. When examiningthe relationship between DifSp and Shear, it is clear that DifSp ispredominantly negative during the high Shear, i.e. during hydraulicjump events (see Palmer et al., 2013) of C1. Thus for the bank anal-ysis DifSp emphasises the high Shear events occurring during thesprings from C1 reducing the significance of lower Shear and lackof hydraulic jump events at neaps in C3. The depth of thethermocline (DepthDThermmax) was significant in explaining the

Fig. 6. Internal wave progression over time over the south-east slope of Jones Bank at spring tides on 8 July 2008 from the EK60 200 kHz echosounder at 78 dB: (a) 06:40(GMT) with easterly (60–65�) current flows of 0.46 m s�1 and shear 0.063; (b) 08:40 (GMT) showing a depression of the thermocline over the bank slope with east-southeasterly current flows (87–121� upper/lower water column) of 0.331–0.506 m s�1 (upper/lower) and shear 0.275; (c) 10:45 (GMT) showing clear high amplitudeinternal waves over the bank slope with southerly (off-bank) current flows (169–191� lower/upper) of 0.134–0.513 m s�1 (upper/lower) and shear 0.392; and (d) 12:50 (GMT)with south-westerly current flows (228–233�) of 0.242–0.391 m s�1 (upper/lower) and shear 0.159. The thermocline is clearly defined, with the surface water coloured inblue, and bottom mixed layer in green/yellow/red. The bottom echo is shown by the deep red line.

Table 5PCA loadings for fish schooling behaviours for deep pelagic schools (>60 m) andshallow pelagic schools (<60 m). Coefficients larger than 0.4 which are therefore asign of a meaningful relationship between variables are shown in bold.

Deep Schools (n = 8515) Shallow Schools (n = 736)

PC1 PC2 PC1 PC2

Sv38 �0.066 0.716 0.025 0.745Depth �0.073 �0.695 0.178 �0.595Width �0.539 �0.009 0.468 0.055Height �0.366 0.049 0.465 0.135Perimeter �0.529 �0.006 0.494 �0.017Area �0.517 0.002 0.495 0.116Compactness �0.135 �0.036 0.211 �0.236

SD 1.80 1.11 1.93 1.14% Total variance 46.2 17.2 53.2 18.4% cum Variance 46.2 63.4 53.2 71.6

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 113

number of deep schools in both bank and flat circuits (Table 6).There were also fewer deep schools in the flat site with increasingcurrents in the lower water column (Table 6).

Table 6Summaries of models of fish school distribution for deep and shallow pelagic schools at thresponse and fitted with Generalised Linear Models with Generalised Estimating Equationorder autoregressive (ar1) time structure. The correlation coefficient (a) and the samplen = 282 for the deep schools at the flat site; (iii) a = 0.207, n = 197 for the shallow schools

BANK

Variable Estimate s.e. Wald

DEEP schools Intercept �6.03 1.16 27.1SpringNeap �2.27 0.44 27.0Depth 0.09 0.01 69.5DepthDThermmax �0.02 0.01 8.8

SHALLOW schools Intercept 4.48 0.56 65.2CHLmax �0.63 0.24 6.7Depth �0.02 0.01 9.7DifSp �1.54 0.71 4.7

3.4. Mixed models of fish behaviour with bio-physical characteristics

Forward and backward model selection resulted in the samemodels except for PC2 for shallow schools in the bank circuits,where forward model selection resulted in a model withDThermmax, and backward model selection resulted in a modelwith CHLmax (with DThermmax being the last variable to beremoved, and hence explaining the highest amount of variance).School morphology of deep pelagic schools was most stronglyrelated to the spring–neap cycle: PC1 was higher at spring thanat neap tides translating to smaller schools at spring tides (Table 7).Also deep schools in the bank circuit changed their morphologyaround the slope region, with smaller schools (higher PC1) at stee-per slopes (higher Depthvar) (Table 7). Current variables influenceddeep schooling morphology at both sites (Table 7). On the bankdeep school size decreased as lower layer current speed increased,and also as the difference between the upper and lower currentspeed decreased (Table 7). Over the flat site, deep school sizedecreased with increasing shear (Table 7). Shallow schoolmorphology was only weakly influenced by the environment.Schools in the bank circuit were smaller with increasingDepthDThermmax and increasing CHLmax (though both had low

e bank and flat sites, modelling the number of schools per 5 min bin with a Poissons (GLM–GEEs) with repeat number as the correlation grouping factor (id) with a first-size (n) are (i) a = 0.744, n = 197 for the deep schools at the bank site; (ii) a = 0.345,

at the bank site; and (iv) a = 0.154, n = 282 for the shallow schools at the flat site.

FLAT

p Variable Estimate s.e. Wald p

<0.001 Intercept 3.15 0.14 531.2 <0.001<0.001 SpringNeap �1.56 0.13 146.5 <0.001<0.001 DepthDThermmax 0.01 0.00 12.7 <0.001<0.01 SpL �1.95 0.67 8.4 <0.01

<0.001 Intercept �0.60 0.07 69.8 <0.001<0.01 SpringNeap �1.61 0.43 14.3 <0.001<0.01<0.05

Table 7Summaries of models of fish school morphology for deep and shallow pelagic schools at the bank and flat sites, modelling PC1 for all schools with a Gaussian response and fittedwith Generalised Linear Models with Generalised Estimating Equations (GLM–GEEs) with repeat number as the correlation grouping factor (id) with a first-order autoregressive(ar1) time structure. The correlation coefficient (a) and the sample size (n) are (i) a = 0.01, n = 2576 for the deep schools at the bank site; (ii) a = �0.011, n = 4525 for the deepschools at the flat site; (iii) a = 0.028, n = 466 for the shallow schools at the bank site; and (iv) n = 80 for the shallow schools at the flat site (no variables significant).

BANK FLAT

Variable Estimate s.e. Wald P Variable Estimate s.e. Wald p

DEEP schools Intercept �0.66 0.08 78.3 <0.001 Intercept �0.95 0.15 39.3 <0.001SpringNeap 0.43 0.09 23.6 <0.001 SpringNeap 0.41 0.05 67.7 <0.001Depthvar 0.01 0.00 87.6 <0.001 Shear 3.07 0.49 39.9 <0.001SpL 1.99 0.32 38.1 <0.001 Time 0.38 0.15 6.3 <0.05

SHALLOW schools Intercept 3.17 0.85 13.8 <0.001 NSDepthDThermmax �0.03 0.01 7.1 <0.01CHLmax �0.67 0.34 3.9 <0.05

114 C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117

Wald test statistics: 7.1 and 3.9 respectively). No variables weresignificant in explaining shallow school morphology on the flat site(Table 7).

School behaviour on the second component axis of the PCA(explaining school Sv and depth) for deep pelagic schools was bestexplained by the spring–neap cycle, with schools being denser andshallower at spring tides than at neap tides for both sites (Table 8).The behaviour of deep schools on the bank site was significantlydependent on a high number of environmental variables otherthan the spring–neap cycle. The next most important predictorwas the direction of the lower water current DirL (Table 8). Sincecurrent direction is a circular variable and the autocorrelation fac-tor showed low levels of autocorrelation (a = 0.0142), this variablewas modelled within a Generalised Additive Model (GAM) frame-work to determine which directions resulted in changes of fishschooling behaviour (Fig. 7. Without allowing for changes in thespring–neap cycle, DirL on its own showed high PC2 when currentswere running perpendicular to the main axis of Jones Bank (SE orNW currents). Allowing for changes in the spring–neap cycle, PC2was only higher when currents were flowing in a southwesterlydirection, i.e. onto the bank during an ebb tide (Fig. 7). Thus therewere denser and shallower schools when the currents were flow-ing onto the bank.

The main predictor of PC2 for shallow pelagic schools on thebank was variation in depth Depthvar, with schools becoming lessdense and deeper at the slope. These schools were also less denseand deeper with deeper thermoclines (Table 8). Only time of daywas a significant predictor in explaining shallow school depthand density on the flat site, with shallower, denser schools as theday progressed from dawn to dusk (Table 8).

4. Discussion

Pelagic fish distribution and schooling behaviour was mainlyinfluenced by three main factors: the presence of the bank andits role in generating internal waves via the hydraulic jump; thespring–neap cycle; and the storm event.

4.1. Role of the bank to fish in the upper water column (shallowschools)

Jones Bank played an important role in influencing both the dis-tribution and behaviour of fish schools during our survey. Fish inthe upper water column (shallow schools) concentrated over thebank regardless of the spring–neap cycle (e.g. Table 3). Based onthe sampling carried out during the survey, these shallow fishare most likely to comprise the zooplanktivores horse mackerelT. trachurus and boarfish C. aper (Martinez et al., this issue). Accu-mulation of fish and other top predators over banks is a wellknown occurrence (Coyle et al., 1992; Holland et al., 1999; Stevicket al., 2008). The mechanisms explaining the attraction of fish and

top predators to banks and seamounts are still not clear, but arelikely to be most strongly related to aggregations of prey on oraround the bank region driven by oceanographic processes suchas internal waves (Stevick et al., 2008). There was no clear differ-ence in the abundance of zooplankton between the bank and flatsite, but future quantitative analysis of the zooplankton componentof the hydrographic data (e.g. using methods detailed in Ballónet al., 2011) may reveal differences in the distribution of zooplank-ton through the water column that may make the bank a moreprofitable foraging ground if the zooplankton is more concentratedthere.

The presence of internal waves may increase the profitability offoraging for both fish and seabirds at the edges of the bank, by con-centrating prey into denser aggregations and forcing it towards thesurface (Stevick et al., 2008). Jones Bank showed a high level ofinternal wave activity during our survey, with the highest ampli-tude internal waves occurring during spring tides, and loweramplitude internal waves observed at both spring and neap tides(Fig. 4, Palmer et al., 2013). Internal waves have been shown tobe important in concentrating plankton into patches and soexplaining zooplanktivorous fish spatial distributions in multi-scale studies off Peru (Bertrand et al., 2008; Grados et al., 2012).In our study, the role of internal waves is supported by the bio-physical variables related to internal wave generation that areimportant in describing shallow school distribution and behaviourover the bank. The large amplitude internal waves observed duringspring tides (Fig. 6) were associated with hydraulic jump activity(Palmer et al., 2013). Shear was relatively high over the bank dur-ing both spring and neap tides, but only associated with the largeamplitude internal waves during spring tides (Palmer et al., 2013).Shear was not a significant variable in explaining either shallowschool distribution or schooling behaviour, however the differencebetween the upper and lower water column currents DifSp was:with higher numbers of schools associated with a high negativeDifSp. DifSp is similar to shear in that it is related to the differencebetween surface and bottom water speeds, however as DifSp doesnot account for current direction it can be low when shear is high(e.g. when both surface and bottom current speeds are high buttravelling in opposite directions). During the bank circuits (C1and C3), DifSp was mainly highly negative (i.e. bottom currentmuch faster than surface current) when shear was high duringspring tides. During the spring tide over the bank this strong neg-ative DifSp was associated with hydraulic jumps that drive thelarge amplitude internal waves. These results suggest that therewere higher numbers of shallow fish schools over the bank at timesof high internal wave activity.

4.2. Role of the bank to fish associated with the bottom (deep schools)

Jones bank also had a significant effect on the fish associatedwith the bottom (deep schools), though in a different way to the

Table 8Summaries of models of fish school depth and density behaviour for deep and shallow pelagic schools at the bank and flat sites, modelling PC2 for all schools with a Gaussianresponse and fitted with Generalised Linear Models with Generalised Estimating Equations (GLM–GEEs) with repeat number as the correlation grouping factor (id) with a first-order autoregressive (ar1) time structure. The correlation coefficient (a) and the sample size (n) are (i) a = 0.0142, n = 2576 for the deep schools at the bank site; (ii) a = 0.985,n = 4525 for the deep schools at the flat site; (iii) a = 0.096, n = 466 for the shallow schools at the bank site; and (iv) a = 0.925, n = 80 for the shallow schools at the flat site.

BANK FLAT

Variable Estimate s.e. Wald p Variable Estimate s.e. Wald p

DEEP schools Intercept 3.62 1.15 9.9 <0.01 Intercept �0.42 0.16 6.8 <0.001SpringNeap 1.83 0.14 178.5 <0.001 SpringNeap 1.54 0.35 19.8 <0.001DirL 0.001 0.00 23.9 <0.001DepthDThermmax �0.02 0.01 11.6 <0.001Depth �0.03 0.01 6.4 <0.05Time 0.44 0.20 4.9 <0.05

SHALLOW schools Intercept 0.92 0.20 20.4 <0.001 Intercept �0.36 0.13 7.2 <0.01Depthvar �0.01 0.00 230.3 <0.001 Time 0.93 0.27 11.9 <0.001DThermmax

* �1.80 0.37 23.3 <0.001

* By backward selection this variable is replaced by CHLmax, with DThermmax being the last variable to be removed from the model. Wald statistics were quite a lot lower, sothe model selected by forward selection is shown in the table.

Fig. 7. Circular smooths of DirL explaining deep fish school behaviour (PC2) on the bank, with (left) no other variables in the model; and (right) factor (SpringNeaps) includedin the model. The dashed line indicates the 95% confidence intervals around the smoother.

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 115

shallow schools. The majority of deep schools were found off thebank, however, there were still small schools of deep fish overthe bank slopes (Fig. 3b) with very few observed over the top ofthe bank.

The bank slopes and tidal currents interacting with the bank, re-sulted in different deep fish schooling behaviour around the bankthan in the flat region. Around the bank, deep schools were smallerover the slope region (higher Depthvar, Table 7). This suggests thatthe deep fish may be foraging over the bank slopes since smallerschools tend to be associated with foraging activity in schoolingfish, a behaviour that reduces the competition for food (Hoareet al., 2004). The bank is also associated with an intrusion of chlo-rophyll into the bottom mixed layer (Palmer et al., 2013), and amovement of the bulk of the diatom population into sub-euphoticdepths (Davidson et al., 2013). If zooplankton are also pushed dee-per (as is suggested at neap tides in Figs. 4b and 5), there would bea resulting increased supply of food to deep zooplanktivore schoolsover the edge of the bank. The bank also interacted in a more phys-ical way with deep fish schooling behaviour: where the bottomcurrent flowed onto the bank schools were pushed towards thebank slope becoming denser and shallower; as the current flowedoff-bank the schools become deeper and more diffuse. Changes invertical fish distribution have been similarly recorded in otherstudies of tidal currents interacting with bank slopes (e.g. Cottéand Simard, 2005).

The flat site, over which the contrasting circuits were con-ducted, was positioned 14 km to the southeast of Jones Bank. Since

the prevailing wind is from the west (Inall et al., 2013), productionon the bank was shown to be swept to the south east, in the direc-tion of the flat site, affecting plankton in the upper water columndownstream from the bank and within a 25 km range of the bank(Davidson et al., 2). The flat site had the highest abundance of deepfish schools (those well below the thermocline) especially at neaptides, so may benefit from the transport of production from thebank. Low amplitude internal waves were observed at the flat siteat both spring and neap tides (Fig. 3) resulting in high shear valuesat times of higher mixing. Thus the smaller deep schools observedat the flat site associated with higher shear, may indicate higherforaging activity with this increased mixing (Mackinson et al.,1999; Hoare et al., 2004; Robinson et al., 2004).

4.3. The role of the spring–neap cycle

The spring–neap cycle was the most significant variable influ-encing both deep and shallow fish school distribution and behav-iour, except for shallow schools found over the bank. Littleresearch has been carried out looking at the effect of the spring–neap cycle in non-estuarine habitats especially on higher trophiclevels such as fish. Generally benthopelagic fish such as cod Gadusmorhua will avoid the higher currents found during spring tides,staying close to the bottom to save energy, and only emerge whenthe currents are slower during the neap tides (Nichol andSomerton, 2002; Yousif and Aglen, 1999). At higher trophic levels,predators such as seabirds will forage in different locations depen-

116 C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117

dent on the spring–neap cycle (Anderson et al., 2004; Irons, 1998),presumably due to different prey behaviours during the differenttides. Our study found higher numbers of fish schools, deep andshallow, during neap tides when the current speeds were lower,so perhaps it is energetically more profitable for the fish to feedduring the weaker neap tides.

Deep fish also showed a change in schooling behaviour with thespring–neap cycle, with larger, more diffuse (less dense) deeperschools at neap tides. This may suggest that the deep fish areexhibiting foraging behaviour, since schooling fish tend to formsmaller schools or form more diffuse aggregations during foragingactivity to reduce the strength for food competition (Hoare et al.,2004; Robinson et al., 2004). Though we must stress that it mayalso reflect a change in fish species between spring and neap tidessince the school structure differs substantially between spring andneap tides (Fig. 3) and fish sampling only took place at the end ofthe study during spring tides. Future advances in understandingthe frequency responses of different species of fish (and mixed spe-cies schools) will enable species identification from the echograms(e.g. Korneliussen and Ona, 2002; Knudsen et al., 2006; Fernandes,2009; Korneliussen et al., 2009). In our study there were insuffi-cient data on the frequency responses of the different species,and sufficient evidence that schools comprised of mixed speciesmaking identification with acoustics difficult.

4.4. The role of the storm event

The strong winds from the storm during the initial 10 days ofthe study period did appear to influence fish distribution andbehaviour, although to a lesser extent than topography and thespring–neap cycle. Shallow schools may have been more stronglyinfluenced by the storm event than the deep schools, perhapsunsurprisingly since the wind mainly affects the surface mixedlayer. There were higher numbers of shallow schools associatedwith lower CHLmax; smaller schools with deeper DepthDThermmax;and more diffuse (less dense) deeper schools with lower CHLmax.All of these conditions are created with the increased mixing bywinds caused by the storm event but also by increased mixingvia internal wave activity. Therefore it is likely that the relationshipbetween shallow schools and low CHLmax levels was an indirectone, reflecting the level of mixing of the water column rather thana preference for lower CHLmax per se. The change in behaviour tomore diffuse schools suggests that these conditions may increasethe opportunities for foraging behaviour, since fish tend to spreadout when foraging (Hoare et al., 2004; Robinson et al., 2004). Fromthese results it is clear that vertical mixing has an effect on fishbehaviour and that both storms and tidal mixing can create similareffects on fish school behaviour.

The significance of thermocline depth on deep fish schools overthe flat area is also likely to be a result of the storm event. Duringthe event, there were more deep fish schools with a deeper ther-mocline: thus more fish schools with a more mixed water column.The increased mixing by the storm winds forced much of the dia-tom population to deeper waters (Davidson et al., 2013), and islikely therefore to have increased bottom mixed layer zooplanktonpopulations. There is therefore some indication from our resultsthat the storm event does not just affect the fish schools in theupper water column, but to a lesser extent fish schools close tothe bottom.

4.5. Conclusions

Our study has defined some of the mechanisms responsible forchanges in pelagic fish availability to predators and has highlightedthe importance of shallow banks in shelf waters as well as thespring–neap tidal cycle to a range of pelagic fish species. Fish spe-

cies within the upper water column such as horse mackerel, mack-erel and boar fish appear to aggregate over bank features mostlikely due to predictable aggregations of zooplankton prey speciescaused by the internal waves generated over the bank slopes. Dee-per fish species also utilise the combination of topographical andtidal interactions and seem to favour neap tides for feedingopportunities.

Understanding the mechanistic importance of banks, the sur-round area that they influence, and how that affects pelagic fishbehaviour is essential to successful marine spatial planning, andfor the designation of Marine Protected Areas (MPAs). Shallowbanks and seamounts are already well known locations for highbiodiversity (Grantham et al., 2011) so protection of these areaswould benefit not only benthic biodiversity but the pelagic fishthat aggregate over them.

The importance of the spring–neap cycle on fish schoolingbehaviour will obviously have strong influence on predator forag-ing behaviour, such as seabirds (Scott et al., 2013), and fishermen(Sharples et al., 2013a) as well as possibly biasing fisheries surveyresults. Fishermen are known to alter their fishing behaviour withthe spring–neap cycle (e.g. Bell et al., 2008), and fish catchabilityhas been shown to vary with tidal currents (Yousif and Aglen,1999), however the mechanisms driving these changes werepoorly understood. The understanding this study has provided inindentifying the bio-physical variables which drive changes in fishdistributions and behaviour will help to quantify fish catchabilityand assist in the design of better targeted surveys for top predatorsand prey fish species.

Acknowledgements

This project was funded by the NERC ‘Sustainable Marine Biore-sources’ funding, with the research cruise funded by the NERCOceans2025 research programme. Special thanks to Sophie Field-ing for the EK60 calibration, to the crew of the RSS James Cook,and for the technical and engineering support provided by theUK National Marine Facilities. Two anonymous reviewers provideduseful comments on previous drafts of this manuscript.

References

Anderson, C.D., Roby, D.D., Collis, K., 2004. Foraging patterns of male and femaledouble-crested cormorants nesting in the Columbia River estuary. CanadianJournal of Zoology 82, 541–554.

Ballón, M., Bertrand, A., Lebourges-Dhaussy, A., Gutiérrez, M., Ayón, P., Grados, D.,Gerlotto, F., 2011. Is there enough zooplankton to feed forage fish populationsoff Peru? An acoustic (positive) answer. Progress in Oceanography 91 (4), 360–381.

Barange, M., 1994. Acoustic identification, classification and structure of biologicalpatchiness on the edge of the Agulhas bank and its relation to frontal features.South African Journal of Marine Science 14, 333–347.

Bell, M.C., Elson, J.M., Addison, J.T., Revill, A.S., Bevan, D., 2008. Trawl catchcomposition in relation to Norway lobster (Nephrops norvegicus L.) abundanceon the Farn Deeps grounds. NE England. Fisheries Research 90, 128–137.

Benoît, H.P., Swain, D.P., 2003. Accounting for length- and depth-dependent dielvariation in catchability of fish and invertebrates in an annual bottom-trawlsurvey. ICES Journal of Marine Science 60, 1298–1317.

Bertrand, A., Ballón, M., Chaigneau, A., 2010. Acoustic observation of livingorganisms reveals the upper limit of the Oxygen Minimum Zone. PLoS ONE 5(4), e10330. http://dx.doi.org/10.1371/journal.pone.0010330.

Bertrand, A., Chaigneau, A., Peraltilla, S., Ledesma, J., Graco, M., et al., 2011. Oxygen:A fundamental property regulating pelagic ecosystem structure in the coastalSoutheastern Tropical Pacific. PLoS ONE 6 (12), e29558. http://dx.doi.org/10.1371/journal.pone.0029558.

Bertrand, A., Gerlotto, F., Bertrand, S., Gutiérrez, M., Alza, L., Chipollini, A., Díaz, E.,Espinoza, P., Ledesma, J., Quesquén, R., Peraltilla, S., Chavez, F., 2008. Schoolingbehaviour and environmental forcing in relation to anchoveta distribution: ananalysis across multiple spatial scales. Progress in Oceanography 79, 264–277.

Blanchard, J.L., Dulvy, N.K., Jennings, S., Ellis, J.R., Pinnegar, J.K., Tidd, A., Kell, L.T.,2005. Do climate and fishing influence size-based indicators of Celtic Sea fishcommunity structure? ICES Journal of Marine Science 62, 405–411.

Bodholt, H., Ness, H., Solli, H., 1989. A new echo-sounder system. Proceedings of theInstitute of Acoustics 11 (3), 123–130.

C.B. Embling et al. / Progress in Oceanography 117 (2013) 106–117 117

Boon, A.K., Réale, D., Boutin, S., 2007. The interaction between personality, offspringfitness and food abundance in North American red squirrels. Ecology Letters 10,1094–1104.

Cotté, C., Simard, Y., 2005. Formation of dense krill patches under tidal forcing atwhale feeding hot spots in the St. Lawrence Estuary. Marine Ecology ProgressSeries 288, 199–210.

Coyle, K.O., Hunt, G.L., Decker, M.B., Weingartner, T.J., 1992. Murre foraging,epibenthic sound scattering and tidal advection over a shoal near St. GeorgeIsland Bering Sea. Marine Ecology Progress Series 83, 1–14.

Davidson, K., Gilpin, L. C., Pete, R., Brennan, D., McNeill, S., Moschonas, G., Sharples,J., 2013. Phytoplankton and bacterial distribution and productivity on andaround Jones Bank in the Celtic Sea. Progress in Oceanography 117, 48–63.

Dingemanse, N.J., Wright, J., Kazem, A.J.N., Thomas, D.K., Hickling, R., Dawnay, N.,2007. Behavioural syndromes differ predictably between 12 populations ofthree-spined stickleback. Journal of Animal Ecology 76, 1128–1138.

Embling, C.B., Armstrong, E., Illian, J., van der Kooij, J., Sharples, J., Camphuysen, C.J.,Scott, B.E., 2012. Investigating fine scale spatio-temporal predator-prey patternsin dynamic marine ecosystems: a functional data approach. Journal of AppliedEcology 49, 481–492.

Fernandes, P.G., 2009. Classification trees for species identification of fish-schoolechotraces. ICES Journal of Marine Science 66, 1073–1080.

Foote, K.G., 1987. Fish target strengths for use in echo integrator surveys. Journal ofthe Acoustic Society of America 88, 1543–1546.

Foote, K.G., MacLennan, D.N., 1984. Comparison of copper and tungsten carbidecalibration spheres. Journal of the Acoustic Society of America 75 (2), 612–616.

Fréon, P., Gerlotto, F., Misund, O.A., 1993. Consequences of fish behaviour for stockassessment. ICES Marine Science Symposium 196, 190–195.

Genin, A., 2004. Bio-physical coupling in the formation of zooplankton and fishaggregations over abrupt topographies. Journal of Marine Systems 50, 3–20.

Grados, D., Fablet, R., Ballón, M., Bez, N., Castillo, R., Lezama-Ochoa, A., Bertrand, A.,2012. Multiscale characterization of spatial relationships among oxyclinedepth, macrozooplankton, and forage fish off Peru using geostatistics,principal coordinates of neighbour matrices (PCNMs), and wavelets. CanadianJournal of Fisheries and Aquatic Sciences 69, 740–754.

Grantham, H.S., Game, E.T., Lombard, A.T., Hobday, A.J., Richardson, A.J., Beckley,L.E., Pressey, R.L., Huggett, J.A., Coetzee, J.C., van der Lingen, C.D., Petersen, S.L.,Merkle, D., Possingham, H.P., 2011. Accommodating dynamic oceanographicprocesses and pelagic biodiversity in marine conservation planning. PLoS ONE 6(2), e16552.

Halekoh, U., Højsgaard, S., Yan, J., 2006. The package geepack for generalisedestimating equations. Journal of Statistical Software 15, 2.

Hoare, D.J., Couzin, I.D., Godin, J.-G.J., Krause, J., 2004. Context-dependent group sizechoice in fish. Animal Behaviour 67, 155–164.

Holland, K.N., Kleiber, P., Kajiura, S.M., 1999. Different resident times of yellowfintuna, Thunnus albacores, and bigeye tuna, T. obesus, found in mixed aggregationsover a seamount. Fisheries Bulletin 97, 392–395.

Inall, M., Aleynik, D., Neil, C., 2013. Horizontal advection and dispersion in astratified shelf sea: the role of inertial osciallations. Progress in Oceanography117, 25–36.

Inger, R., Attrill, M.J., Bearhop, S., Broderick, A.C., James, G.W., Hodgson, D.J., Mills, C.,Sheehan, E., Votier, S.C., Witt, M.J., Godley, B.J., 2009. Marine renewable energy:potential benefits to biodiversity? An urgent call for research. Journal of AppliedEcology 46 (6), 1145–1153.

Irons, D.B., 1998. Foraging area fidelity of individual seabirds in relation to tidalcycles and flock feeding. Ecology 79 (2), 647–655.

Johnsen, E., Godø, O.R., 2007. Diel variations in acoustic recordings of blue whiting(Micromesistius poutassou). ICES Journal of Marine Science 64, 1202–1209.

Jolliffe, I.T., 2002. Principal Component Analysis. Springer-Verlag, New York, pp.112–114.

Jurvelius, J., Knudsen, F.R., Balk, H., Marjomäki, T.J., Peltonen, H., Taskinen, J.,Tuomaala, A., Viljanen, M., 2008. Echo-sounding can discriminate between fishand macroinvertebrates in fresh water. Freshwater Biology 53, 912–923.

Knudsen, F.R., Larsson, P., Jakobsen, P.J., 2006. Acoustic scattering from a larvalinsect (Chaoborus flavicans) at six echosounder frequencies: implications foracoustic estimates of fish abundance. Fisheries Research 79, 84–89.

Korneliussen, R.J., Ona, E., 2002. An operational system for processing andvisualizing multi-frequency acoustic data. ICES Journal of Marine Science 59,293–313.

Korneliussen, R.J., Heggelund, Y., Eliassen, I.K., Johansen, G.O., 2009. Acoustic speciesidentification of schooling fish. ICES Journal of Marine Science 66, 1111–1118.

Mackinson, S., Nøttestad, L., Guénette, S., Pitcher, T., Misund, O.A., Fernö, A., 1999.Cross-scale observations on distribution and behavioural dynamics of oceanfeeding Norwegian spring-spawning herring (Clupea harengus L.). ICES Journalof Marine Science 56, 613–626.

Marchal, P., Horwood, J., 1996. Long-term targets for the Celtic Sea mixed-speciesmulti-métiers fisheries. Aquatic Living Resources 9, 81–94.

Martin, J.G.A., Réale, D., 2008. Temperament, risk assessment and habituation tonovelty in eastern chipmunks, Tamias striatus. Animal Behaviour 75, 309–318.

Martinez, I., et al., this issue. Small scale changes in fish assemblages on the JonesBank (Celtic Sea). Progress in Oceanography Special Issue.

Michalsen, K., Godø, O.R., Anders, F., 1996. Diel variation in the catchability ofgadoids and its influence on the reliability of abundance indices. ICES Journal ofMarine Science 53, 389–395.

Minderman, J., Reid, J.M., Evans, P.G.H., Whittingham, M.J., 2009. Personality traitsin wild starlings: exploration behaviour and environmental sensitivity.Behavioural Ecology 20 (4), 830–837.

Nichol, D.G., Somerton, D.A., 2002. Diurnal vertical migration of the Atka mackerelPleurogrammus monopterygius as shown by archival tags. Marine EcologyProgress Series 239, 193–207.

Palmer, M.R., Inall, M.E., Sharples, J., 2013. The physical oceanography of JonesBank: A mixing hotspot in the Celtic Sea. Progress in Oceanography 117, 9–24.

Patrick, S.C., Browning, L.E., 2011. Exploration behaviour is not associated withchick provisioning in Great Tits. PLoS ONE 6 (10), e26383. http://dx.doi.org/10.1371/journal.pone.0026383.

Perry, R.I., Boutillier, J.A., Foreman, M.G.G., 2000. Environmental influences on theavailability of smooth pink shrimp, Pandalus jordani, to commercial fishing gearoff Vancouver Island, Canada. Fisheries Oceanography 9, 50–61.

Pinnegar, J.K., Jennings, S., O’Brien, C.M., Polunin, N.V.C., 2002. Long-term changes inthe trophic level of the Celtic Sea fish community and fish market pricedistribution. Journal of Applied Ecology 39, 377–390.

Robinson, C.J., 2004. Responses of the northern anchovy to the dynamics of thepelagic environment: identification of fish behaviours that may leave thepopulation under risk of overexploitation. Journal of Fish Biology 64, 1072–1087.

Robinson, C.J., Anislado, U., Ramos, P., 2004. Shoaling fish and red crab behaviourrelated to tidal variations in Bahia Magdalena, Mexico. Deep-Sea Research II 51,767–783.

Scott, B.E., Webb, A., Palmer, M.R., Embling, C.B., Sharples, J., 2013. Fine scale bio-physical oceanographic characteristics predict the foraging occurrence ofcontrasting seabird species; Gannet (Morus bassanus) & storm petrel(Hydrobates pelagicus). Progress in Oceanography 117, 118–129.

Sharples, J., Ellis, J.R., Nolan, G., Scott, B.E., 2013a. Fishing and the oceanography of astratified shelf sea. Progress in Oceanography 117, 130–139.

Sharples, Y.J., Moore, C.M., Hickman, A.E., Holligan, P.M., Tweddle, J.F., Palmer, M.R.,Simpson, J.H., 2009. Internal tidal mixing as a control on continental marginecosystems. Geophysical Research Letters 36, L23603. http://dx.doi.org/10.1029/2009GL040683.

Sharples, J., Scott, B.E., Inall, M.E., Ellis, J.R., Nolan, G.D., this issue-b. Introduction:the physics and biology of a temperate shelf sea in summer. Progress inOceanography Special Issue.

Simpson, S.D., Jennings, S., Johnson, M.P., Blanchard, J.L., Schön, P., Sims, D.W.,Genner, M.J., 2011. Continental shelf-wide response of a fish assemblage torapid warming of the sea. Current Biology 21 (18), 1565–1570.

Stevick, P.T., Incze, L.S., Kraus, S.D., Rosen, S., Wolff, N., Baukus, A., 2008. Trophicrelationships and oceanography on and around a small offshore bank. MarineEcology Progress Series 363, 15–28.

Wood, S.N., 2006. Generalized Additive Models – An Introduction with R. Chapman& Hall/CRC.

Yousif, A., Aglen, A., 1999. Availability of pelagic distributed cod (Gadus morhua L.)to bottom trawls in the Barents Sea. Fisheries Research 44, 47–57.