Testing hypotheses about temporary and persistent effects of otter trawling on infauna: changes in diversity rather than abundance

MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 385: 51–64, 2009doi: 10.3354/meps08056

Published June 18

INTRODUCTION

Fishing with towed bottom gear is a regional andglobal threat to the diversity and functioning of marinebenthic habitats (e.g. Collie et al. 1997, Kaiser et al.2002, 2006, Gray et al. 2006). The effects of trawlingand dredging range from very drastic impacts, suchas destruction of vulnerable habitats like cold-watercorals (e.g. Fosså et al. 2002, Costello et al. 2005,Wheeler et al. 2005) and sponges (e.g. Wassenberg etal. 2002), to more subtle effects like alterations in ben-thic production processes (Jennings et al. 2001). Themagnitude and persistence of fishing disturbancedepends on a large number of factors, such as the type

of fishery and the level of natural disturbance (e.g.Kaiser et al. 2006). Therefore, in order to provide guid-ance on levels of fishing that are sustainable in terms ofimpact to the environment, research is needed tounravel the magnitude, persistence and resilience oftrawl effects under different environmental conditionsand regimes (e.g. Gray et al. 2007, Løkkeborg 2007).

The most commonly used type of towed bottom fish-ing gear in the North Atlantic is the otter trawl (Rag-narsson & Steingrímsson 2003), of which the 2 compo-nents — otter-boards and the ground-rope — are pri-marily responsible for the sediment disturbance. Theotter-boards scour the seabed, generally resulting information of tracks of variable depth (e.g. Gilkinson et

© Inter-Research 2009 · www.int-res.com*Email: [email protected]

Testing hypotheses about temporary and persistenteffects of otter trawling on infauna: changes in

diversity rather than abundance

Stefán Áki Ragnarsson1,*, Mats Lindegarth2

1Marine Research Institute, Skúlagata 4, PO Box 1390, 121 Reykjavík, Iceland2Department of Marine Ecology — Tjärnö, University of Gothenburg, Tjärnö, 452 96 Strömstad, Sweden

ABSTRACT: A field experiment was conducted to examine the short- and long-term effects of ottertrawling on a macrobenthic infaunal community in shallow subtidal waters of Faxaflói Bay (SW Ice-land) that had never been trawled before. The experimental design consisted of 4 sites trawled 10times and 4 areas left undisturbed (controls). Sampling of fauna and sediments was carried out inJune 1997, immediately after trawling, and subsequently 2 and 7 mo later, in order to investigatelonger term impacts of trawling. No significant treatment effects could be detected on total abun-dance or on multivariate structure, and tests for individual species revealed only a single short-termeffect (for the bivalve Thyasira flexuosa). However, trawling affected several aspects of diversity withsignificant short-term reduction in species richness and persistent effects on the Shannon-Wienerindex. Power analysis revealed that larger changes were needed to detect changes in abundancecompared to measures of diversity. Analyses of qualitative changes revealed a decreasing trend inabundance of the majority of taxa (especially the rarer ones) immediately after trawling. This was themost likely cause for the reduced diversity and the augmented small-scale variability in species rich-ness. The size of initial impacts and the persistence of effects were largely consistent with those ofprevious studies in similar environments. Similar to other types of environmental impacts in aquaticenvironments, our analyses reveal that impacts of trawling on measures of diversity can be detectedat a high statistical power and that larger effects are necessary to detect impacts on abundances.

KEY WORDS: Otter trawl · Experimental fishing · Infauna · Variability · Analysis of variance ·Iceland · Diversity · Power analysis

Resale or republication not permitted without written consent of the publisher

Mar Ecol Prog Ser 385: 51–64, 2009

al. 1998). The area of seabed disturbed by the otter-boards is, however, very small relative to the areaswept by the ground-rope (e.g. Krost et al. 1990). Theground-rope, which is generally fitted with ‘rock hop-pers’, scours the upper sediment layers, thus changingthe sediment structure and reducing both the nutri-tional sediment quality and the amount of flocculatedmatter (Schwinghamer et al. 1998, Watling et al. 2001).

The majority of experimental studies on the effects ofotter-trawling in soft sediments have reported rela-tively minor effects in comparison to other types ofgear, such as beam-trawls (Kaiser et al. 2006). Conclu-sions about effects of trawling on assemblages in softsediments differ among studies in magnitude, persis-tence and in terms of the proportion of variablesaffected. This variability in response is partly driven bydifferences in the physical environment (e.g. hydrody-namic regime, depth, sediment characteristics andhabitat type) and in the species’ properties (e.g. life-history traits, lifestyle and feeding guild; Kaiser et al.2006). In sandy habitats, the observed impacts havegenerally been small and few. In muddy habitats, how-ever, otter-trawls appear to cause more complex im-pacts, with a short-term decrease in abundances ofindividual taxa (<1 wk), followed by increased abun-dances over longer periods of time, ranging fromweeks to months (Kaiser et al. 2006). Another type ofimpact that has been attributed to otter-trawling is anincrease in small-scale spatial and temporal variabilityof faunal assemblages (Lindegarth et al. 2000a). Thus,it appears that otter-trawling might affect benthic as-semblages selectively and in a number of ways.

Evidence of impacts of trawling comes from bothobservational and manipulative studies. Manipulativeexperiments offer unique opportunities to test hypo-theses about causal relationships and to estimate themagnitude and nature of effects (e.g. Underwood1997). Properly planned, conducted and interpreted,such experiments allow researchers to separate effectsof experimental treatments without being confoundedby variability caused by other factors that are not con-trolled by the experimenter. While such experimentsare ideally conducted at large spatial and temporalscales, in order to optimise relevance and realism(Thrush et al. 1998), practical constraints and costsoften limit their reliability (e.g. Lindegarth et al.2000b). In particular, the prevalence of spatial andtemporal variability in these systems (e.g. Currie &Parry 1996, Kaiser & Spencer 1996, Lindegarth et al.2000a,b) combined with a small number of indepen-dent replicate units may lead to insufficient statisticalpower (i.e. increased risk of Type 2 errors; e.g. Map-stone 1995) or confounding if experimental treatmentsare not replicated (i.e. Type 1 errors; e.g. Hurlbert1984, Underwood 1994, Lindegarth et al. 2000b).

In the present study we tested hypotheses abouteffects of otter trawling on abundance and diversity ofinfaunal communities in an area that had never beentrawled before. Using a multifactorial experimentaldesign, we simultaneously quantified and tested forthe existence of any temporary and persistent effectson means and on small-scale spatial variability of ben-thic assemblages in muddy areas off the coast of Ice-land following experimental trawling. In order to allowindependent estimates of the effects of trawling and ofspatial and temporal variability associated with factorsother than trawling, the experimental design involved4 areas that were experimentally trawled and 4 areasthat remained undisturbed. Furthermore, the results ofthe experiment were evaluated by assessing the mini-mum size of impacts that were necessary to be de-tected with an acceptable statistical power.

MATERIALS AND METHODS

Study site and experimental design. The study site(1 × 3.7 km; Fig. 1) is located north of the Reykjanespeninsula in Stakks Fjord (Faxaflói Bay), centred at63° 59’ N, 22° 28’ W. The whole of Faxaflói Bay (Fig. 1)has been closed to trawl fisheries since 1958 (Direc-torate of Fisheries, unpubl. data). Since 1991, Icelandicskippers are obliged to keep logbook records on the lo-cations of all hauls down to a spatial resolution of 1’ lat-itude and 1’ longitude. The analysis of logbook data(Marine Research Institute and the Directorate of Fish-eries database) showed that no trawling activity tookplace within Faxaflói Bay between 1991 and 2003.Furthermore, there is no evidence of demersal fishingprior to 1991 within Faxaflói Bay. The only towed mo-bile bottom gear allowed in Faxaflói Bay is the Danish

52

3.7 km

Faxaflói

T1

Keflavík

Stakks Fjord

C1 T2 C2 T3 C3 T4 C4

1.0 km

Fig. 1. Map of the study area in Stakks Fjord (location withinIceland shown in inserted maps) and the locations of the 4trawled (T1 to T4) and the 4 untrawled control (C1 to C4) plots

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

seine, and analysis of logbook data shows that fishingwith this gear has not taken place within the study area.

The depth within the study site ranged from 32 to35 m. The sediment can be categorised as mud ormuddy sand with the amount of silt varying from 46 to54% by weight (Bale & Kenny 2005). Storms are fre-quent and due to shallow depth at the study site,storm-induced suspension of sediments is likely to beconsiderable.

In June 1997, an experiment on the effects of ottertrawling on infauna was initiated. Eight plots, sepa-rated by about 400 m, were located within the studysite (Fig. 1). Four of these were trawled 10 times,whereas the remaining 4 were left undisturbed as con-trol plots. This intensity of trawling represents the up-per limits of yearly trawling in Icelandic waters (Rag-narsson & Steingrímsson 2003). Sampling of fauna andsediments was carried out immediately after trawling(June 1997) and subsequently 2 and 7 mo after trawl-ing (August 1997 and January 1998). Locations forgrab samples were assigned a priori at random alongthe midline of each plot. During sampling within thetrawled plots, care was taken that samples were col-lected more or less along the midline of the trawlswath. Within each plot, samples were collected from 8stations in August and January, but from 6 stations inJune. From each station, 2 replicate samples were col-lected with a Shipek grab (0.041 m2). Thus, a total of352 samples were planned for the study, but unfortu-nately 7 and 5 samples from trawled and untrawledplots, respectively (3.5% of the samples), were lostduring handling at sea and in the laboratory.

The trawling was carried out with an otter trawl fit-ted with 1100 kg polyvalent otter-boards. The lengthsof head-rope and bridles were 25 and 46 m, respec-tively, and the net was rigged with rockhoppers mea-suring 45 cm in diameter. The distances between thewing ends and the otter-boards during towing (mea-sured at a speed of 3.5 knots with SCANMAR acousticinstrumentation attached to the headline) were 13.3and 39.6 m, respectively. During towing, the vessel fol-lowed, as closely as possible, the midline of the firsttrawl path. The position of the vessel during towingwas recorded to an accuracy of ±10 m using dGPS (dif-ferential global positional system), relayed through thenavigational program package MaxSea (Radiomidun).In the MaxSea program, the distance between trawltracks could be measured with high accuracy. Basedon this information, the width of the swath within eachof the 4 trawled areas was estimated to range from 80to 120 m. Due to the shallow depth at the site, the off-set of the trawl relative to the vessel during towing wasconsidered to be minor.

Sample processing. Samples for faunal analysiswere sieved on board through a 0.5 mm mesh and

fixed in buffered 4% formaldehyde with seawater. Inthe laboratory, all animals were sorted, identified tothe lowest taxonomic level possible (usually species orgenus), counted and preserved in 70% alcohol. Themost abundant species in the samples was Myriocheleoculata, but this polychaete becomes easily fragmen-ted during sample processing. To obtain reliable abun-dance estimates of M. oculata, all tubes from a singlesample were dissected, revealing that 57% of all frag-ments of this polychaete contained heads. In subse-quent data analysis, all M. oculata fragments werecounted and multiplied with 0.57 to obtain correctedabundance estimates. Although this estimate wasbased on a single sample, dissecting more tubes to ob-tain more robust estimates would have been extremelytime consuming since the average number (±SD) oftube fragments per sample was 237 (±141.4).

Statistical analyses and hypotheses about impacts oftrawling. The linear model used for uni- and multivari-ate hypothesis testing contained 4 factors: ‘treatment’(Tr; trawled versus control plots; number of levels,a = 2) and ‘time’ (Ti; 0, 2 and 7 mo after trawling; num-ber of levels, b = 3) were fixed factors, while ‘plots’Pl(Tr) and ‘station’ (St[Tr, Ti, Pl]; 6 or 8 per plot andtime; number of levels, c = 6 or 8, each consisting of 2samples) were random, nested factors. The Pl(Tr) andSt(Tr, Ti, Pl) factors tested for significant differencesamong plots between treatments and for differencesamong stations within plots, respectively. Despite theapparent analytical complexity, relevant tests forephemeral and persistent effects of trawling are identi-fied a priori for both multi- and univariate analyses. Apersistent effect, i.e. difference between trawled andcontrol plots, which appears immediately after trawl-ing and does not change throughout the experiment,can be identified as a significant effect of Tr. A tempo-rary effect where the difference between trawled andcontrol plots differ among times is, however, detectedas a significant interaction between the experimentaltreatment and time, Tr × Ti.

Hypotheses about multivariate effects on abun-dance and composition of whole assemblages weretested using permutational multivariate ANOVA(PERMANOVA; Anderson 2001, McArdle & Anderson2001). Due to limitations with unbalanced data, sam-ples were randomly excluded to achieve equal samplesizes in all combinations of plots and times (n = 6) andconsequently the factor St(Tr, Ti, Pl) was removedfrom the model. These modifications may have someeffects on the precision of multivariate means withinplots at individual times, but, because they do notaffect the degrees of freedom for the important errorterms, the power and reliability of the test for effectsof trawling (Tr and Tr × Ti) will not be severely af-fected. Furthermore, the differences in the structure

53

Mar Ecol Prog Ser 385: 51–64, 2009

of assemblages in trawled and untrawled plots werevisualised using non-metric multidimensional scaling(nMDS; Plymouth Routines in Multivariate EcologicalResearch; Clarke & Warwick 1994). Ordinations ofsamples were done at individual times, and, to visu-alise changes among times, we plotted ordinationsusing centroids for individual plots at all times.PERMANOVA and nMDS ordinations were based onthe Bray-Curtis measure of similarity on square-root-transformed data.

Univariate analysis of variance (ANOVA) was usedto test hypotheses about persistent (Tr) and temporaryeffects (Tr × Ti) of trawling on averages of total num-ber of individuals and taxa, the 2 measures of speciesdiversity (Shannon-Wiener’s and Simpson’s indices)and on abundances of individual taxa. To achieve acomplete interpretation of temporary effects, theinteraction was complemented with a posteriori tests.We used Student-Newman-Keuls (SNK) procedures toevaluate at which sampling occasion there was a dif-ference between trawled and untrawled areas. Tomaximise the statistical power of all univariate tests,post hoc pooling was applied to insignificant terms(p > 0.25), mainly Pl(Tr), Ti × Pl(Tr) and St(Tr, Ti, Pl)(Winer et al. 1991, Underwood 1997). Due to theslightly unbalanced design (more samples collected inAugust and January compared to June), the Type IIIsums of squares were used (Henderson 1953). Prior toanalyses, data were log10(x + 1) transformed to nor-mality.

For univariate analyses, possible persistent effects oftrawling on small-scale variability (variability amongsamples) were examined using 2-tailed F-tests at α =0.025 (Underwood 1994). This was done by partition-ing the residual variability into that estimated fortrawled and control plots, respectively.

Power analyses. In order to evaluate the capacity ofthe experiment to detect any persistent or temporaryeffects of a given magnitude, we calculated the statis-tical power of a range of effect sizes. By defining adesired statistical power of 80% (which corresponds toa Type 2 error of 0.2), we estimated a minimum de-tectable effect size (MDES; Quinn & Keough 2002) forthe total number of individuals, 3 measures of diversityand for the abundance of the 3 most common and the 3least common taxa. Calculation of power requires thatalternatives (HA) to the null hypothesis (H0) aredefined. First, analyses of the power to detect persis-tent effects, i.e. the factor Tr, were modelled by assum-ing a proportional decrease (D = 0.02 to 0.5) in thetrawled locations compared to the controls such that:

HA,Tr:x–

Trawl × D < x–Control (1)

For each level of D, power was estimated by calculat-ing the non-centrality parameter λ from:

(2)

where MSerror is the appropriate error term of theF-test, x–i is the mean for the ith treatment and x is theoverall mean. Finally, the probability of detecting acertain effect was obtained from the non-centralF-distribution using the actual degrees of freedomused in the test (i.e. when possible, df for the pooledmean square were used).

Second, analyses of the power to detect temporaryeffects, i.e. the interaction Tr × Ti, were modelled byassuming that a proportional decrease in the trawledlocations would only occur at the first time of sampling,while at the other times there would be no differencesbetween trawled and control plots. Therefore, thealternative hypothesis of interest was:

HA,Tr × Ti: time 1, x–Trawl × D < x–Control; time 2, x–Trawl × D= x–Control and time 3, x–Trawl × D = x–Control

and λ was calculated as:

(3)

where x–ij is the mean of the ith treatment at the jthtime of sampling, x–i is the mean of the ith treatment, x–j

is the mean of the jth time and x– is the overall mean.

RESULTS

General observations

A total of 160 taxa representing 138 577 ind. wererecorded during the course of the present study. Twotaxa dominated in abundance, the tube-building poly-chaete Myriochele oculata and bivalves belonging tothe genus Abra, accounting for 38 and 27% of the totalabundance, respectively. The polychaetes Paraonis sp.,Cossura longocirrata, Scoloplos armiger, Pholoe mi-nuta, Sternaspis scutata and Eteone longa contributed25% to the total abundance. Polychaetes, bivalves,crustaceans and other groups comprised 69.6, 29.3, 0.7and 0.3% of the total abundance, respectively.

Effects on multivariate structure

Multivariate tests of hypotheses about effects of trawl-ing on the whole benthic community found no significantpersistent or temporary effects (Table 1). This suggeststhat there was no consistent difference between treat-ments at all times combined, nor were there any consis-tent differences in the way the 2 treatments changedover time. Consistent with this result, was the lack of

λ =− − +

==∑∑cn x x x xij i j

j

b

i

a

( )2

11

MSerror

λ =−

=∑bcn x xi

i

a

( )2

1

MSerror

54

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

clear separation among samples from trawled and un-trawled plots in nMDS ordinations based on data fromthe whole experiment or at individual times (Fig. 2). Nev-ertheless, there was a highly significant change in com-munity structure over time (Table 1, Fig. 2). Further-more, community structure differed significantlyamong plots, and the significant Ti × Pl(Tr) interactionshowed that trends in community structure over time

were not consistent among plots.These findings were supported by thenMDS ordination, which revealed thatthe spatial and temporal variationamong plots located in the eastern half(T3 to C4) was much greater comparedto that in the western half (T1 to C2) ofthe study area (Fig. 2).

Effects on total abundance anddiversity

The overall qualitative pattern of totalabundance and diversity is that all vari-

ables increased during the experiment (Fig. 3). Further-more, the mean values for all variables were smaller inthe trawled area compared with the controls immedi-ately after fishing, but, on subsequent sampling occa-sions, the rate of increase tended to be larger in trawledareas. In consistency with these observations, tests usingANOVA showed a range of different immediate (days)and long-term (months) effects on measures of diversity

55

Source df SS MS F p Permutable Unique units (n) values (n)

Treatment = Tr 1 679 679 0.342 1 8 915Time = Ti 2 10966 5483 8.356 0.001 24 868Plot(Tr) = Pl(Tr) 6 11930 1988 3.99 0.001 192 997Tr × Ti 2 1358 679 1.03 0.423 24 910Ti × Pl(Tr) 12 7874 656 1.32 0.004 192 1000Residual 168 83655 497

Total 191 116464

Table 1. Tests of fishing on the overall assemblage structure using permutationalmultivariate analysis of variance (PERMANOVA) on Bray-Curtis dissimilaritieson square-root-transformed non-standardised data. Probabilities were deter-mined by unrestricted permutation of raw data using correct permutable units.

Unique values: no. of unique values in permuted distribution

Stress: 0.15

a Whole period

3

3

33

3

2

2 2

211

11

1

2

3 1

2

4 44

44

4

T1 C1T2 C2T3 C3T4 C4

Stress: 0.25

d January 1998T1 C1T2 C2T3 C3T4 C4

c August 1997

Stress: 0.24

T1 C1T2 C2T3 C3T4 C4

b June 1997

Stress: 0.23

August

June

January

Fig. 2. Non-metric multidimensional scaling ordination of infauna (filled symbols: trawled plots; open symbols: controls) over: (a)the whole period (average abundance per plot, numbers identify plot number in each treatment) and on: (b to d) separate sampling

occasions (June, August and January, respectively). Plot abbreviations as in Fig. 1

Mar Ecol Prog Ser 385: 51–64, 2009

(Table 2). No significant effects were observed on the to-tal number of individuals (Table 2). The Shannon-Wiener index was significantly and persistently lower inthe trawled areas compared to the controls. Despite atendency for recovery during the experiment, the lack ofa significant interaction (Tr × Ti) indicated that the effectpersisted throughout the experiment (Fig. 3). Overall,the Shannon-Wiener index decreased by approximately3% in trawled areas (x–Control = 2.78 and x–Control = 2.69). Inall other community variables, the overall differences be-tween trawled and control areas were <2%.

Another observed effect that ap-peared to be persistent during the wholeexperiment was the significant increaseof the small-scale variability in speciesrichness at trawled locations (Table 2).This increase in variability was not onlysignificant but also substantial in magni-tude (67%). Finally, another observed ef-fect of trawling was a significant interac-tive effect, Tr × Ti, on the mean speciesrichness (Table 2). Inspection of themeans and SNK support the hypothesisthat this is a temporary negative ef-fect on species richness (Fig. 3; SNK ofTr × Ti; ‘June’: trawled < controls; ‘Au-gust’: trawled < controls and ‘January’:trawled = controls). Further inspectionsof the means show that the initial impactcaused the number of species to de-crease by 12%. Then, 2 mo later the dif-ference was 5.4% and after 7 mo the dif-ference was ~0.2%. The third index ofdiversity (Simpson) showed no signifi-cant effects of trawling.

Effects on abundances of individual taxa

In contrast to the measures of diversity, there wereno persistent long-term effects of trawling on abun-dances of individual taxa (Table 3). Furthermore, a sig-nificant short-term effect (i.e. Tr × Ti interaction) wasfound only for the bivalve Thyasira flexuosa, whichwas less abundant (70%) in trawled plots immediatelyafter trawling but more abundant in trawled plots atsubsequent sampling times (34 and 15%, respec-tively). Thus, significant long- or short-term effects on

56

2.3

2.35

2.4

2.45

2.5

2.55

2.6

2.65

June August January

1.26

1.28

1.3

1.32

1.34

1.36

1.38

1.4

1.42

1.44

1.46

June August January

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3

June August January

0.64

0.66

0.68

0.7

0.72

0.74

0.76

0.78

June August January

c

b

d

a

Lo

g (to

tal in

d. co

re–1)

Lo

g (n

o.

taxa c

ore

–1)

H’

Sim

pso

n’s

ind

ex

Fig. 3. Means (±SE) in: (a) total number of individuals, (b) Shannon-Wienerdiversity (H ’), (c) number of taxa and (d) Simpson’s reciprocal 1/D, in trawled

(black bars) and untrawled (grey bars) plots

Source df Total no. of individuals Total no. of taxa Shannon-Wiener Simpson’s Error termMS F p MS F p MS F p MS F p

Tr 1 0.02 0.22 0.659 0.04 2.06 0.20 0.65 4.75a+b+c 0.03 0.003 0.58a+c 0.45 Pl (Tr)Ti 2 0.31 14.14 0.001 0.09 48 0.0001 1.35 10.43 0.002 0.07 9.65 0.003 Ti × Pl(Tr)Tr × Ti 2 0.002 0.06b+c 0.937 0.01 6.00 0.02 0.13 0.96a+b+c 0.39 0.003 0.37 0.69 Ti × Pl(Tr)Pl(Tr) 6 0.07 2.21 0.045 0.02 4.67 0.002 0.09 0.64 0.69 0.001 0.18 0.98 St (Tr,Ti, Pl)Ti × Pl(Tr) 12 0.02 0.7 0.746 0.002 0.54 0.89 0.13 0.93 0.51 0.008 1.47 0.14 St (Tr, Ti, Pl)St(Tr, Ti, Pl) 152 0.03 1.45 0.010 0.005 1.24 0.09 0.14 1.2 0.12 0.005 1.31 0.04 ResidualResidual 163 0.02 0.004 0.12 0.004

Two-tailed F-tests of small-scale variabilityTrawled 79 0.02 1.048 0.418 0.005 1.67 0.01 0.09 1.39 0.07 0.004 1.25 0.16Controls 84 0.02 0.003 0.13 0.005

Table 2. ANOVA model for total number of individuals and species and for Shannon-Wiener and Simpson’s diversity measures.Significant results indicating impacts of trawling as persistent or temporary effects on the mean or effects on small-scale variabil-ity are shown in bold. F-tests were constructed using a priori error terms unless post hoc pooling was possible. Pooling terms are

indicated by superscript: (a) Pl(Tr), (b) Ti × Pl(Tr), (c) ST(Tr, Ti, Pl), (d) residual

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

average abundance were found for only 1 out of 32investigated taxa.

Similarly, the 2-tailed tests that examined changesin small-scale spatial variability showed few effects(Table 3). Interestingly, the abundance of Thyasiraflexuosa again appeared to be significantly affected,with the small-scale variability being ~35% smaller intrawled areas compared to the controls. For mytilid bi-valves, however, the small-scale variability in abun-dance was >2-fold greater in trawled areas, resultingin a significant 2-tailed test.

In contrast to the small number of significant effectsof trawling, abundances of all taxa varied significantlyin time and/or space (Table 3). Abundance of most taxadiffered significantly among plots within treatments,indicating that differences among plots were due tofactors other than trawling or that responses of benthicorganisms to fishing disturbances were not consistent

among plots. Relatively few taxa (Cossura longocir-rata, Scalibregma inflatum, Schistomeringos nigriden-tata, Goniada maculata, Ophelina acuminata, Hetero-mastus filiformis, Abra spp. and Arctica islandica)showed significant interactive variability (Ti × Pl(Tr)),i.e. plots within treatments changed similarly amongtimes. Further, there were few instances (Scoloplos ar-miger, Apistobranchus tullbergi, Terebellides stroemi,Sternaspis scutata, Spio sp., amphipods and Abra spp.)of significant variability among stations within plots(St[Tr, Ti, Pl]).

Power analyses

The power analyses of community variables re-vealed that smaller proportional changes are requiredto detect effects on any of the measures of diversity

57

Taxon Tr Ti Tr × Ti Pl(Tr) Ti × Pl(Tr) St(Tr, Ti, Pl) Two-tailed

PolychaetesMyriochele oculatab+c+d NS ** NS ** NS NS **Paraonis sp.b+c+d NS *** NS ** NS NS NSEteone longab+c NS NS* NS *** NS NS* NSScoloplos armigerc+d NS *** NS *** NS * NSCossura longocirrata NS *** NS *** *** NS NSPholoe minuta NS ** NS ** NS* NS NSSternaspis scutatab+c NS NS NS ** NS ** NSScalibregma inflatum NS *** NS *** ** NS* NSMaldane sarsib+c+d NS * NS *** NS NS NSBrada villosab+c NS *** NS ** NS NS NSPraxillella spp. NS *** NS *** NS NS NSSpio sp.b+c NS *** NS NS NS * NSSchistomeringos nigridentatac+d NS *** NS ** * NS NSChaetozone setosaa+c NS ** NS NS NS* NS NSSyllidaec+d NS *** NS ** NS* NS NSGoniada maculatac+d NS NS NS *** * NS NSLaonice cirratac+d NS *** NS NS NS* NS NSLumbrineris fragilisc+d NS NS NS *** NS NS NSOphelina acuminata NS ** NS *** *** NS NSAricidea sp.b+c+d NS ** NS NS* NS NS NSApistobranchus tullbergi NS *** NS *** NS * NSTerebellides stroemi NS *** NS ** NS ** NSHeteromastus filiformisc+d NS *** NS NS * NS NSCrustaceansIsopodaa+c+d NS* * NS NS NS NS NS Amphipodaa+c+d NS NS NS NS NS ** NSMolluscsAbra spp. NS *** NS *** *** ** NSMacoma calcareab+c NS * NS *** NS NS NSThracia spp. NS NS NS *** NS* NS NSArctica islandicac+d NS NS* NS *** *** NS NS Thyasira flexuosab+c NS *** ** ** NS NS *Mytilidaeb+c+d NS ** NS NS NS NS ** EchinodermsOphiura sp.a+b+c+d NS ** NS NS NS NS NS

Table 3. Results of ANOVA for the most abundant taxa. Significance was classified as *, ** and *** for p < 0.05, 0.01 and 0.001,respectively. Non-significant results were categorised into 2 groups: NS* and NS for 0.05 < p < 0.10 and p > 0.10, respectively.

Superscripts as in Table 2

Mar Ecol Prog Ser 385: 51–64, 2009

compared to the changes needed to detect effects ontotal abundance (Fig. 4). If there were persistent ef-fects on diversity these would have needed to be on theorder of 3 to 5% to be reliably detected using Shan-non-Wiener and Simpson’s indices. For species rich-ness and total abundance, the corresponding valueswere 15 and 25%, respectively. Note that the observeddifferences for these 2 indices were <2%.

The MDES values of short-term effects were gener-ally larger than those of any persistent effect (Fig. 4).To reliably detect changes in richness and Shannon-Wiener indices these need to be between 5 and 10%,but between 10 and 15% for the Simpson index, whilethe corresponding figure for total abundance is again~25%. Although there are slight differences in sensi-tivity among the different measures of diversity, theoverall impression is that the changes needed to detecteffects are smaller for these 3 measures of diversitythan for abundance.

The apparent difficulties associated with detectingeffects on abundance were highlighted by furtheranalyses of the MDES for abundance of the 3 mostabundant taxa (Myriochele oculata, Abra spp. andParaonis sp.) and the 3 least abundant taxa subjectedto detailed ANOVA (Ophiura sp., Lumbrineris fragilis

and Aricidea sp.). These analyses showed that to reli-ably detect impacts of trawling on abundant taxa,changes in abundance generally need to be >35%,both for tests of persistent and short-term effects(Fig. 4c,d). Interestingly, the MDES values of the 3rarer taxa were consistently smaller. Persistent effectscould be reliably detected at changes of 10 to 30%,while the corresponding figures for short-term effectswere 15 to 35%.

Qualitative changes of taxa

Even if mean and small-scale variability values in theabundance of very few individual taxa were signifi-cantly affected by trawling, it is possible that differ-ences between trawled and control areas are qualita-tively consistent. To test this proposition, simple sign(binomial) tests were performed. These analysesshowed that in June, 79% of the taxa had lower meanabundance in trawled plots compared to untrawledplots. This proportion is significantly greater than whatcan be expected by chance (p < 0.001), and these differ-ences were mainly driven by rarer taxa (Fig. 5). About50% of all taxa with average abundances of <1 individ-

58

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 5 10 15 20 25 30 35 40 45 50

Myriochele oculataAbra spp.Paraonis sp.Ophiura sp.Lumbrineris fragilis Aricidea sp.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 5 10 15 20 25 30 35 40 45 50

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 10 20 30 40 50

Po

wer

to d

etc

ta p

ers

iste

nt

ch

an

ge

Total abundanceRichnessShannon-WienerSimpson

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 10 20 30 40 50

Po

wer

to d

ete

ct

a s

ho

rt-t

erm

ch

an

ge

a

b

c

d

-

Proportional difference between TR and C

Fig. 4. Statistical power as a function of proportional difference between trawled and control areas: evaluated for persistent ef-fects (a,c) and short-term effects (b,d) and for total abundance, species richness, Shannon-Wiener index and Simpson’s index (a,b)and the 3 most abundant taxa (Myriochele oculata, Abra spp. and Paraonis sp.) and the 3 least common taxa (Ophiura sp., Lum-brineris fragilis and Aricidea sp.) that could be analysed statistically (c,d). The dotted lines refer to 80% statistical power

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

ual sample–1 declined by 30%, and 61.5% of all taxarepresented by 1 to 5 ind. on average declined by>20%. In August and January, the proportions of taxafound in lower abundance in trawled plots were 46 and50%, respectively. For the whole experiment the pro-portion was 50%. These differences were not signifi-cantly different from chance alone (p > 0.05; Fig. 5).

DISCUSSION

Type, size and duration of impacts

The aim of the present study was to test hypothesesabout temporary and persistent effects of otter trawlingin a previously undisturbed area at intensities thatare representative of those in heavily trawled areasaround Iceland (Ragnarsson & Steingrímsson 2003).Our results show that trawling caused short-term andpersistent effects on average and small-scale variabil-ity in diversity of infauna, while effects on abundancesof individual taxa were difficult to detect. Using appro-priate estimates of variability we also demonstratedthat effect sizes of >20% change are necessary to reli-ably detect short- and long-term effects on abundanceof individual taxa, while changes of <5% were oftensufficient for measures of diversity. Nevertheless,despite the difficulties in detecting the effects on abun-dance of individual taxa, we also demonstrated thatthere was a significant tendency for benthic taxa ingeneral to have lower abundances at fished comparedto unfished locations immediately after trawling.

The observed effects on diversity were consistentlylargest immediately after the experimental treatment.

Nevertheless, the fact that the effect on mean richnesswas detected as a significant interaction between timeand treatment, while that effect on the Shannon-Wiener measure was detected as a significant effect oftreatment, indicates that the persistence of observedeffects differed among measures of diversity. The spe-cies richness decreased by 12% immediately aftertrawling, whilst after 7 mo the difference was 0.2%.The corresponding figures for the Shannon-Wienermeasure were 5.4 and 0.9%. One reasonable interpre-tation of this result is that taxa that are initially re-moved from the area return relatively quickly, whilethe establishment of natural assemblages in terms ofrelative abundances and ecological associations takeslonger time. This interpretation is supported by the ob-servation of qualitative effects mainly involving de-creases in abundances of rare taxa immediately aftertrawling. Among these were the relatively large-bodied maldanid and terebellid taxa, which have pre-viously been reported to be sensitive to fishing (Kaiser& Spencer 1996, Tuck et al. 1998, Sanchez et al. 2000,Jennings et al. 2001, Duplisea et al. 2002, Hiddink etal. 2006). Although these differences were subtle, theymay nevertheless have accounted for the initial effectson richness. Similarly, Kaiser & Spencer (1996) andKenchington et al. (2001) reported abundance declinesin most taxa following trawling. These were generallymore pronounced for rarer taxa, but significant differ-ences for individual taxa were rarely attained. Hans-son et al. (2000) reported that, while 61% of all taxarecorded in their study were found at lower abundancein trawled plots, significance was only attained for 1 in-dividual taxon, and there was no obvious relationshipbetween the size of organisms and the size of impacts.

59

January

Mean abundance (log)

June August

All sampling months

Mean abundance (log)

Tr/C

Tr/C

00.20.40.60.8

11.21.41.6

0 0.5 1 1.5 2 2.5

0

0.5

1

1.5

2

2.5

0 0.5 1 1.5 2 2.5

0

0.5

1

1.5

2

0 0.5 1 1.5 2 2.5

0 0.5 1 1.5 2 2.50

0.5

1

1.5

2

Fig. 5. The ratio of mean abundance of individual taxa in trawled and untrawled plots (Tr/C) in different sampling periods as a function of overall mean abundance

Mar Ecol Prog Ser 385: 51–64, 2009

Impacts on spatial variability

The variability in species richness was augmented attrawled plots following trawling, but these effects didnot last long. Lindegarth et al. (2000a) reported in-creased temporal and spatial variability in the struc-ture of benthic communities at trawled sites. The rea-son for increased variance may have been due todisproportionate fishing impacts within the trawlswath. For soft bottoms at least, the otter-boards areconsidered to have more impact per unit area com-pared to the ground-rope, mainly because they pene-trate deeper into the seabed (e.g. Gilkinson et al.1998). In our study, the surface area swept by theboards in a single tow was small (0.8 to 1.5 m in width)compared to that of the ground-rope (~13 m). Thismeans that in a single tow, ~32% of the total tow width(distance between otter-boards; ~40 m) was swept withthe ground-rope and ~2 to 4% was swept with theotter-boards. The probability of retrieving a samplefrom a otter-board track in our study within a trawlswath of ~100 m that had been towed 10 times could besomewhere between 5 and 15%. If invertebrate densi-ties declined markedly only in those areas swept withthe otter-boards, it would be unlikely to detect anyoverall decline at the scale of the trawl swath, as thevast majority of samples would be obtained outside theotter-board tracks. Nevertheless, such disproportion-ate impact within the trawl swath may result in overallreduction in species richness and those diversity mea-sures that are influenced by the loss of the rarer taxa(like the Shannon-Wiener), while augmenting the spa-tial variability.

Consistency with previous studies

To further put our results into the context of previ-ous work on impacts of otter trawling in muddy areas,we compared the size and persistence of impacts tothose reported in the meta-analysis by Kaiser et al.(2006). To achieve this, we calculated the average dif-ference between trawled and control plots at individ-ual times and for all taxa used in ANOVA analyses(using the equation for response measure provided byKaiser et al. 2006), and compared it with meta-analysis outputs (see Figs. 2 & 3 in their paper) forstudies carried out in muddy habitats (Fig. 6). Ourstudy revealed that immediately after trawling, theabundances of individual taxa were on average 16%lower in trawled areas compared to undisturbedareas, while the corresponding figure for the meta-analysis was 28% (see Fig. 2 in Kaiser et al. 2006).The results in our study suggest that the differencescan be regarded as being significant as the confi-

dence limits do not include zero. Furthermore, thefact that the error bars in the present study overlaplargely with those of the meta-analysis suggests thatour results, in terms of initial impacts, are highly con-sistent with previous results using the same type ofgear in similar environments. On the other hand, ourresults are not entirely consistent with previous stud-ies with respect to patterns of recovery (see Fig. 3 inKaiser et al. 2006). Although detailed comparisons ofrecovery patterns for the 2 data sets are difficult dueto differences in times of sampling, there appears tobe consistency among studies in terms of initial (1 to2 mo) recovery of mean abundances. Beyond a fewmonths, however, it seems that most studies haveobserved increased abundances of individual taxa intrawled areas. Such a response could be attributed tocolonisation of opportunistic taxa benefiting from thechanged environment as a result of fishing distur-bance (de Juan et al. 2007). In our study, the averagechange in the abundance of individual taxa showedno signs of long-term increases in trawled areas.However, there was a tendency for increased totalabundances in trawled areas after 3 and 7 mo (Fig. 6),but, because our study can only provide a point esti-mate of change in total abundance, it is difficult toassess the significance of this result. Unfortunately,Kaiser et al. (2006) did not provide comparable dataon longer term effects of otter trawling in muddyhabitats with respect to total number of individuals or

60

–30

–25

–20

–15

–10

–5

0

5

10

15

20

25 75 125 175 225

Time from trawling (d)D

iffe

ren

ces b

etw

een

fis

hed

an

d u

nfish

ed

plo

ts (%

)

–25

Fig. 6. Responses (mean ± 2 SD) of benthic taxa (s), totalnumber of individuals (h) and species (n) to otter trawling

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

species richness (Fig. 6). Nevertheless, the initialimpacts on both total abundance and richness in ourstudy (Fig. 6) are in many respects similar to thosewhich have been observed in previous studies usingotter trawls in similar environments (e.g. Tuck et al.1998, Ball et al. 2000, Hansson et al. 2000, Lindegarthet al. 2000a, Sanchez et al. 2000, Drabsch et al. 2001,Kenchington et al. 2001, Kutti et al. 2005).

It is difficult to evaluate the extent to which the mag-nitude and persistence of initial impacts is representa-tive of a wider set of benthic assemblages and types offishing gear. Previous studies have highlighted fishinggear type, seabed characteristics, frequency of naturaldisturbances and water depth to be the most importantfactors (e.g. Kaiser et al. 2002). Other studies haveshown that trawling can modify the overall functioningof benthic communities, such as by altering the relativecomposition of functional groups (de Juan et al. 2007)and/or the biogeochemical environment (Allen &Clarke 2007). Hiddink et al. (2006) predicted, by usinga modelling approach, that the effects of trawling wereless strong in areas where shear stress is high and thebottom is of mud or muddy sands. The shallow depth atour study site does indicate that suspension and trans-port of sediments is likely to be considerable duringstorms, and the area can be regarded as a dynamic en-vironment (Churchill 1989). The infauna, dominatedby small-bodied (<1 cm), mobile, deposit-feeding taxa,is probably easily suspended during storms (e.g. Poseyet al. 1996), and is therefore likely to be adapted to thesediment disturbance caused by the experimentaltrawling in our study. This view of relatively resilientecological assemblages at the experimental site is alsosupported by the uni- and multivariate analyses thatshow substantial temporal variability, often across thewhole study area, but sometimes differently amongplots within the area.

The dynamic nature of the environment at the studysite indicates that recolonisation of animals (both prob-ably active and passive) is likely to be rapid. Our esti-mates on recovery rates are fully consistent with thoseobserved in other manipulative trawl impact studies(Kaiser et al. 2006). Recovery in our study, as for mostother small-scale trawl impact studies carried out onsoft sediments, is likely to be rapid due to the proxim-ity of undisturbed sea beds, where ~90% of the overallstudy area was undisturbed. However, at the scale ofthe commercial fishery (e.g. Thrush et al. 1998), themuch larger size of areas impacted by fishing impliesthat the distances to undisturbed areas are greaterand, thus, slower rates of recolonisation would beexpected. Similarly, recovery patterns in our study maydiffer from those in the real fishery, where the benthiccommunity can be subjected to fishing pressure overlonger periods of time (Kaiser et al. 2002).

Indicators of impacts

Findings from our study show that trawling causedsignificant negative impacts on several aspects ofdiversity of infauna, while significant effects on abun-dances of dominant taxa and on multivariate commu-nity structure were practically absent. Because none ofthese variables are strictly independent, this resultmay appear contradictory. Clearly, different taxa varyin their sensitivity to physical disturbance, but it is log-ically inconsistent to propose that species richness andShannon-Wiener’s index may be affected negativelyby trawling without any negative effects on individualtaxa. Nevertheless, we showed that taxa were in mostcases negatively affected by trawling, but it was gener-ally difficult to identify significant effects on particulartaxa. The most likely explanation of why significanteffects were detected for measures of diversity, but notfor abundance data, is that there were differences instatistical power between the 2 classes of variables(e.g. Osenberg et al. 1994, Mapstone 1995). These dif-ferences can, in turn, be explained either by differ-ences in effect sizes or in error variability (i.e. denomi-nator MS) in the tests critically evaluating hypothesesregarding the impacts of trawling. The analyses pre-sented here suggest that the main cause is that theerror variability was generally larger for measures ofabundance compared to those of diversity. The ratio-nale behind this conclusion is that the power analysesshowed that the minimum detectable effect size wasconsistently larger for measures of abundance, both forpersistent and short-term effects. Second, inspectionsof means in trawled and control plots showed thatthe observed effect sizes for measures of diversity werenot larger than those of abundances, and in fact thesewere often smaller. It is worth emphasising thatanother important issue to do with statistical power isthat the number of replicate units and error degrees offreedom did not differ between the 2 types of variablesand, therefore, cannot explain why measures of diver-sity appeared to be more powerful. Furthermore, itappears unlikely that alternative designs, involvingmore replicate plots, stations, or samples would haveimproved the statistical power substantially. This isbecause most tests were done using pooled error meansquares in such a way that many hypotheses about theeffects of trawling could be tested using >150 df (max-imum 333 df) in the denominator. By all reasonablestandards for experimental ecology this must be con-sidered a sufficient degree of replication (Underwood1997, Quinn & Keough 2002). Exceptions to this, how-ever, are the multivariate tests using PERMANOVA.For example, the tests for short-term effects were donewith only 12 df in the denominator mean square. Thismay explain the lack of significant multivariate effects

61

Mar Ecol Prog Ser 385: 51–64, 2009

that otherwise are considered particularly powerful forthe detection of environmental impacts (e.g. Gray et al.1990, Clarke 1993, Anderson & Thompson 2004).

In our study, measures of diversity appeared to bemore sensitive to environmental impacts than abun-dances of individual taxa in terms of the proportionaleffects that can be reliably detected. In general, stud-ies on soft-sediment macrofauna that have reportedimpacts of trawling on species diversity have alsoshown effects on abundances and/or community struc-ture (e.g. Kaiser & Spencer 1996, Tuck et al. 1998,Smith et al. 2000). Others have reported effects oftrawling on abundances and community structure, butnot on diversity (e.g. Hansson et al. 2000). Neverthe-less, in consistency with the observations presentedhere, previous analyses of statistical power of differentcommunity metrics in aquatic environments suggestthat it is usually possible to reliably detect smaller im-pacts on measures related to species diversity, such asrichness, evenness and environmental quality indices,compared to measures of abundance (e.g. Pratt & Bow-ers 1992, Johnson 1998, Sandin & Johnson 2000,Schratzberger et al. 2002). This does not necessarilymean that tests of hypotheses about effects on diversityare more powerful than those on abundance, butrather that diversity estimates may be less variable.The actual power is also determined by the strength ofmechanistic coupling between trawling and theresponse variable (i.e. the true effect size). Clearlysome taxa, such as epibenthic megafauna (e.g. Was-senberg et al. 2002), may be particularly vulnerable totrawling due to their specific morphologies and life-history characteristics. Due to the expected large effectsize, tests of impacts on such taxa are likely to be pow-erful despite large error variability. Such taxa were,however, largely missing or rare in our study. The taxathat showed signs of impacts (in small-scale variabilityand interaction between time and treatment), i.e. Myri-ochele oculata, Thyasira flexuosa and Mytilidae, doindeed live inside tubes and calcareous shells that maybe vulnerable to physical disturbance, but so are manyof the taxa that were not significantly affected. Thus, inthe assemblages present in our experimental plots,such life-history characteristics were not strong predic-tors of impact. Instead, one general lesson from thisstudy is that impacts on abundances of individual taxaare unlikely to be detected with a high probabilityunless the effect size is large.

Although the present study examined effects oftrawling with rigorous experimental design (4 repli-cate plots, 3 sampling occasions and a total samplingeffort of 340 samples), weak or no effects on abun-dance or on multivariate structure were detected.However, findings from our study indicate that theeffects of trawling in this particular environment were

subtle and mostly did not last long. They were detectedas changes in diversity, small-scale variability anddeclines in abundance of most taxa. Power analysessimilarly revealed that smaller statistical power wasneeded to detect changes in diversity than in abun-dance. Our findings indicate that these patterns are toa large extent driven by loss of the rarer taxa. This con-clusion is further supported by a power analysis show-ing that the statistical power to detect changes inmeans of the 3 rarest taxa (that could be statisticallyanalysed) was greater than the power to detectchanges in the means of the 3 dominant taxa. Similarlyto most other manipulative experiments on trawling,our study was designed to detect effects of a singleintense disturbance. The extent to which these impactsare representative of the size and persistence of con-tinuous, low-intensity disturbance represents an im-portant challenge to future studies.

Acknowledgements. S.A.R. is especially grateful to S. A. Stein-grímsson for his help in the planning of this project, commentson the earlier drafts of this manuscript and assistance duringfieldwork. The manuscript was improved by the comments ofE. Kenchington and E. Guijarro Garcia. S.A.R. also thanks H.Björnsson for preparing S-plus scripts and G. V. Helgason forreviewing the identification of polychaetes. Thanks are due tothe crews of the research vessels ‘Árni Fri7riksson’ and ‘Dröfn’for their assistance during sampling and to 3. Ásgeirsson andH. Skarphé7insdóttir for their hard fieldwork. S.A.R. alsothanks the staff at the Sandgerdi Marine Centre and R. Ólafs-dóttir for sample analysis. This project was partly supported byfunds from the Icelandic centre for research (RANNIS). Finan-cial support was provided to M.L. from the Swedish researchcouncil for environment, agricultural sciences and spatialplanning (FORMAS) through Contract 217-2006-357.

LITERATURE CITED

Allen JI, Clarke KR (2007) Effects of demersal trawling onecosystem functioning in the North Sea: a modellingstudy. Mar Ecol Prog Ser 336:63–75

Anderson MJ (2001) A new method for non-parametric multi-variate analysis of variance. Austral Ecol 26:32–46

Anderson MJ, Thompson AA (2004) Multivariate controlcharts for ecological and environmental monitoring. EcolAppl 14:1921–1935

Bale AJ, Kenny AJ (2005) Sediment analysis and seabed char-acterisation. In: Eleftheriou A, McIntyre A (eds) Methodsfor the study of marine benthos, 3rd edn. Blackwell Sci-ence, Oxford, p 43–87

Ball BJ, Fox G, Munday BW (2000) Long- and short-term con-sequences of a Nephrops trawl fishery on the benthos andenvironment of the Irish Sea. ICES J Mar Sci 57:1315–1320

Churchill JH (1989) The effects of commercial trawling onsediment resuspension and transport over the MiddleAtlantic Bight continental shelf. Cont Shelf Res 9:841–864

Clarke KR (1993) Non-parametric multivariate analyses ofchanges in community structure. Aust J Ecol 18:117–143

Clarke KR, Warwick RM (1994) Change in marine communi-ties: an approach to statistical analysis and interpretation.Plymouth Marine Laboratory, Plymouth

62

Ragnarsson & Lindegarth: Effects of otter trawling on infauna

Collie JS, Escanero GA, Valentine PC (1997) Effects of bottomfishing on the benthic megafauna of Georges Bank. MarEcol Prog Ser 155:159–172

Costello MJ, McCrea M, Freiwald A, Lundälv T and others(2005) Role of cold-water Lophelia pertusa coral reefs asfish habitat in the NE Atlantic. In: Freiwald A, Roberts JM(eds) Cold-water corals and ecosystems. Springer-Verlag,Berlin, p 771–805

Currie DR, Parry GD (1996) Effects of scallop dredging ona soft sediment community: a large-scale experimentalstudy. Mar Ecol Prog Ser 134:131–150

de Juan S, Thrush SF, Demestre M (2007) Functional changesas indicators of trawling disturbance on a benthic commu-nity located in a fishing ground (NW Mediterranean Sea).Mar Ecol Prog Ser 334:117–129

Drabsch SL, Tanner JE, Connell SD (2001) Limited infaunalresponse to experimental trawling in previously un-trawled areas. ICES J Mar Sci 58:1261–1271

Duplisea DE, Jennings S, Warr KJ, Dinmore TA (2002) Asize-based model of the impacts of bottom trawling onbenthic community structure. Can J Fish Aquat Sci 59:1785–1795

Fosså JH, Mortensen PB, Furevik DM (2002) The deep-watercoral Lophelia pertusa in Norwegian waters: distributionand fishery impacts. Hydrobiologia 471:1–12

Gilkinson K, Paulin M, Hurley S, Schwinghamer P (1998)Impacts of trawl door scouring on infaunal bivalves:results of a physical trawl door model/dense sand interac-tion. J Exp Mar Biol Ecol 224:291–312

Gray JS, Clarke KR, Warwick RM, Hobbs G (1990) Detectionof initial effects of pollution on marine benthos: an exam-ple from the Ekofisk and Eldfisk oilfields, North Sea. MarEcol Prog Ser 66:285–299

Gray JS, Dayton P, Thrush S, Kaiser MJ (2006) On effects oftrawling, benthos and sampling design. Mar Pollut Bull52:840–843

Gray JS, Dayton P, Thrush S, Kaiser MJ (2007) Fishing forfacts on the environmental effects of trawling and dredgefisheries: reply to Løkkeborg. Mar Pollut Bull 54:497–500

Hansson M, Lindegarth M, Valentinsson D, Ulmestrand M(2000) Effects of shrimp-trawling on abundance of benthicmacrofauna in Gullmarsfjorden, Sweden. Mar Ecol ProgSer 198:191–201

Henderson CR (1953) Estimation of variance and covariancecomponents. Biometrics 9:226–252

Hiddink JG, Jennings S, Kaiser MJ, Queirós AM, DupliseaDE, Piet GJ (2006) Cumulative impacts of seabed trawldisturbance on benthic biomass, production, and speciesrichness in different habitats. Can J Fish Aquat Sci 63:721–736

Hurlbert SH (1984) Pseudoreplication and the design of eco-logical field experiments. Ecol Monogr 54:187–211

Jennings S, Dinmore TA, Duplisea DE, Warr KJ, Lancaster JE(2001) Trawling disturbance can modify benthic produc-tion processes. J Anim Ecol 70:459–475

Johnson RK (1998) Spatiotemporal variability of temperatelake macroinvertebrate communities: detection of impact.Ecol Appl 8:61–70

Kaiser MJ, Spencer BE (1996) The effects of beam-trawl dis-turbance on infaunal communities in different habitats. JAnim Ecol 65:348–358

Kaiser MJ, Collie JS, Hall SJ, Jennings S, Poiner IR (2002)Modification of marine habitats by trawling activities:prognosis and solutions. Fish Fish 3:1–24

Kaiser MJ, Clarke KR, Hinz H, Austen MCV, Somerfield PJ,Karakassis I (2006) Global analysis of response and recov-ery of benthic biota to fishing. Mar Ecol Prog Ser 311:1–14

Kenchington ELR, Prena J, Gilkinson KD, Gordon Jr DC andothers (2001) Effects of experimental otter trawling on themacrofauna of a sandy bottom ecosystem on the GrandBanks of Newfoundland. Can J Fish Aquat Sci 58:1043–1057

Krost P, Bernhard M, Werner F, Hukriede W (1990) Ottertrawl tracks in Kiel Bay (western Baltic) mapped by side-scan sonar. Meeresforschung 32:344–353

Kutti T, Høisæter T, Rapp HT, Humborstad OB, Løkkeborg S,Nøttestad L (2005) Immediate effects of experimental ottertrawling on a sub-arctic benthic assemblage inside BearIslands fishery protection zone in the Barents Sea. In:Barnes PW, Thomas JP (eds) Benthic habitats and theeffects of fishing. Am Fish Soc Symp 41:519–528

Lindegarth M, Valentinsson D, Hansson M, Ulmestrand M(2000a) Effects of trawling disturbances on temporal andspatial structure of benthic soft-sediment assemblages inGullmarsfjorden, Sweden. ICES J Mar Sci 57:1369–1376

Lindegarth M, Valentinsson D, Hansson M, Ulmestrand M(2000b) Interpreting large-scale experiments on effects oftrawling on benthic fauna: an empirical test of the spatialconfounding in experiments without replicated controland trawled areas. J Exp Mar Biol Ecol 245:155–169

Løkkeborg S (2007) Insufficient understanding of benthicimpacts of trawling is due to methodological deficien-cies — a reply to Gray et al. (2006). Mar Pollut Bull 54:494–496

Mapstone BD (1995) Scalable decision rules for environmen-tal impact studies: effect size, type I and type II errors. EcolAppl 5:401–410

McArdle BH, Anderson MJ (2001) Fitting multivariate modelsto community data: a comment on distance-based redun-dancy analysis. Ecology 82:290–297

Osenberg CW, Schmitt RJ, Holbrook SJ, Abu-Saba KE, FlegalAR (1994) Detection of environmental impacts: naturalvariability, effect size, and power analysis. Ecol Appl 4:16–30

Posey M, Lindberg W, Alphin T, Vose F (1996) Influence ofstorm disturbance on an offshore benthic community. BullMar Sci 59:523–529

Pratt JR, Bowers NJ (1992) Variability of community metrics:detecting changes in structure and function. Environ Tox-icol Chem 11:451–457

Quinn GP, Keough MJ (2002) Experimental design and dataanalysis for biologists. Cambridge University Press, Cam-bridge

Ragnarsson SÁ, Steingrímsson SA (2003) Spatial distributionof otter trawl effort in Icelandic waters: comparison ofmeasures of effort and implications for benthic communityeffects of trawling activities. ICES J Mar Sci 60:1200–1215

Sanchez P, Demestre M, Ramon M, Kaiser MJ (2000) Theimpact of otter trawling on mud communities in the north-western Mediterranean. ICES J Mar Sci 57:1352–1358

Sandin L, Johnson RK (2000) The statistical power of selectedindicator metrics using macroinvertebrates for assessingacidification and eutrophication of running waters. Hydro-biologia 422/423:233–243

Schratzberger M, Dinmore TA, Jennings S (2002) Impacts oftrawling on the diversity, biomass and structure of meio-fauna assemblages. Mar Biol 140:83–93

Schwinghamer P, Gordon DC Jr, Rowell TW, Prena J, McKe-own DL, Sonnichsen G, Guigné JY (1998) Effects of exper-imental otter trawling on surficial sediment properties of asandy-bottom ecosystem on the Grand Banks of New-foundland. Conserv Biol 12:1215–1222

Smith CJ, Papadopoulou N, Diliberto S (2000) Impact of ottertrawling on an eastern Mediterranean commercial trawl

63

Mar Ecol Prog Ser 385: 51–64, 2009

fishing ground. ICES J Mar Sci 57:1340–1351 Thrush SF, Hewitt JE, Cummings VJ, Dayton PK and others

(1998) Disturbance of the marine benthic habitat by com-mercial fishing: impacts at the scale of the fishery. EcolAppl 8:866–879

Tuck ID, Hall SJ, Robertson MR, Armstrong E, Basford DJ(1998) Effects of physical trawling disturbance in a previ-ously unfished sheltered Scottish sea loch. Mar Ecol ProgSer 162:227–242

Underwood AJ (1994) On beyond BACI: sampling designsthat might reliably detect environmental disturbances.Ecol Appl 4:3–15

Underwood AJ (1997) Experiments in ecology: their logicaldesign and interpretation using analysis of variance. Cam-bridge University Press, Cambridge

Wassenberg TJ, Dews G, Cook SD (2002) The impact of fishtrawls on megabenthos (sponges) on the north-west shelfof Australia. Fish Res 58:141–151

Watling L, Findlay RH, Mayer LM, Schick DF (2001) Impact ofa scallop drag on the sediment chemistry, microbiota, andfaunal assemblages of a shallow subtidal marine commu-nity. J Sea Res 46:309–324

Wheeler AJ, Bett BJ, Billett DSM, Masson DG, Mayor D(2005) The impact of demersal trawling on northeastAtlantic deepwater coral habitats: the case of the DarwinMounds, United Kingdom. In: Barnes PW, Thomas JP (eds)Benthic habitats and the effects of fishing. Am Fish SocSymp 41:807–817

Winer BJ, Brown DR, Michels KM (1991) Statistical principlesin experimental design, 3rd edn. McGraw Hill, Boston

64

Editorial responsibility: Lisandro Benedetti-Cecchi,Pisa, Italy

Submitted: September 18, 2008; Accepted: April 09, 2009Proofs received from author(s): May 30, 2009