Size Composition in Fish Communities

D4 - Marine Food Webs

D4.2 - Proportion of selected species at the top of food webs

The Typical Length indicator measures the size-structure of fish and elasmobranch communities and it decreases under high fishing pressure. Although low compared to the 1980s, Typical Length for the assessed demersal fish has been recovering across the OSPAR Maritime Area since 2010. Pelagic fish generally show fluctuations but no trend. Locally, there are deviations from these patterns.

Area Assessed

Printable Summary

Background

OSPAR’s strategic objective with respect to biodiversity and ecosystems is to halt and prevent further loss in biodiversity, protect and conserve ecosystems and to restore, where practicable, ecosystems, which have been adversely impacted by human activities.

The Typical Length indicator is one of three food-web indicators currently used by OSPAR. It represents the average length of fish (bony fish and elasmobranchs) and provides information on the size composition within communities of fish. The indicator is calculated using catch data from species sampled by scientific surveys and given in units of centimetres. Communities are represented by habitat-based feeding assemblages (groups of fish): namely, demersal assemblages (i.e. species living on or near the sea floor) and pelagic assemblages (i.e. species living in the water column).

Fishing mortality constrains the age structure of fish populations, reducing the proportion of larger individuals (Figure 1). A gradual, steady decline in Typical Length is expected in response to high fishing pressure. This is because the size structure of the fish assemblage integrates the impacts of fishing pressure over long periods of time. Model simulations demonstrate that in food webs where predator-prey interactions dominate over other interactions, large species at high trophic levels (the position of the species within the food web) are highly sensitive to loss of diversity at lower trophic levels.

Figure 1: A large-bodied Atlantic Wolffish (courtesy of Jim Ellis)

Fishing mortality constrains the age structure of fish communities, reducing the proportion of larger / older individuals. Fishing is also size-selective, preferentially removing larger / older fish, and therefore fundamentally affects fish community size composition. So far three indicators relating to fish size have been developed to assess impacts of fishing on healthy fish communities and the foodweb, considering similar but different parameters.

The distribution of biomass over body size (size spectra; Kerr and Dickie, 2001) is an emergent property of food webs, therefore size-based metrics that are sensitive and specific to pressures can be used as indicators of food-web structure. Jennings et al. (2007) found that body size was related to trophic level in fish in the North Sea at the community level (see also Reum et al., 2015). Barnes et al. (2010) demonstrated the relationship between fish size and trophic transfer efficiency. Riede et al. (2011) demonstrated that log-mean body size was significantly related to trophic level in marine invertebrates, and ectotherm and endotherm vertebrates using data on multiple ecosystems. Model simulations (Rossberg et al., 2008) have demonstrated that in food webs where trophic interactions dominate over other interactions, large species at high trophic levels are highly sensitive to loss of diversity at lower trophic levels (ICES, 2014a).

Fishing is a size-selective process therefore fish body size decreases during overexploitation (Boudreau and Dickie, 1992). A gradual, steady decline in Typical Length is expected in response to high fishing pressure because the size structure of the community integrates the impacts of fishing pressure over long periods of time (Rossberg, 2012; Fung et al., 2013). Processes related to rising sea temperature also serve to reduce body size of fish (Daufresne et al., 2009; Gibert and DeLong, 2014). The indicator can respond to pressures on the marine environment that impact individual fish directly (entrapment activities) or indirectly (through change in their seabed or pelagic habitat, primary production and food-web interactions). Although species are combined within the habitat-based feeding assemblages, it is possible to compute the indicator for each species individually.

The indicator is aggregated at the survey level within each region assessed and complemented by sub-divisional analyses at a scale appropriate to pressures and habitats that can be highly localised. Sub-divisional metrics are aggregated through a weighted average where those weights are given by the total surveyed biomass of relevant assemblage in each sub-division.

The Typical Length (TyL) is the weighted geometric mean length of fish, with weights given by the standardised catch rate of individuals in an area and defined as follows:

Formula a. Typical Length indicator

where Miis the body mass (standardised to kilogrammes per unit area fished) of the i-th fish with length Li (in units of centimetres) in a sample of N fish.

Data for this indicator come from scientific fisheries surveys, which ideally sample the entire fish community but in practice do not. The indicator requires that surveys are conducted at regular intervals (e.g. annually) in the same area with a standard fishing gear. Sufficiency of available sample sizes can be judged using re-sampling techniques (Shephard et al., 2012, Lynam and Rossberg, 2017). The absolute biomass of individuals in length classes present in the environment is not recorded directly by surveys, rather observations are made from samples with detection error (including many false negatives). The detection error is further complicated by differing catchabilities over length classes and species such that the relative abundance between species and length classes observed is survey specific. Where available, catchability estimates can be used to attempt to correct for this component of the systematic measurement error (e.g. Fraser et al., 2007). However, such estimates are sparse in the scientific literature and prone to great uncertainty. In future, the recent catchability corrections estimated by Walker et al. (2017) could be considered. Alternatively, model-based estimates of absolute species abundance can be used to rescale observed abundances, but here model uncertainty is also great (ICES, 2014b). For simplicity, Typical Length is defined with reference to a particular sampling design with a varying limitation to the size range sampled by fishing gear. For each survey, this indicator is calculated for sub-divisions that represent different habitats and communities, where possible.

The data are collected under the national programmes and the Data Collection Framework (EC, 665/2008). Currently, the most important data sources for Typical Length are those groundfish surveys that are conducted through the International Council for the Exploration of the Sea (ICES). The International Bottom Trawl Survey (IBTS) programme in the Greater North Sea, Celtic Seas, and Bay of Biscay and Iberian Coast is particularly important since the trawl is a general-purpose design aimed to catch both demersal and pelagic species. However, beam trawl surveys are more efficient at catching benthivorous species such as sole (Figure a) and acoustic surveys, supplemented with pelagic trawling, are more suitable for pelagic species such as mackerel (Figure b) and time series of Typical Length from such surveys may be preferable should sufficient length sampling of fish be made.

Figure a: Common sole (Solea solea) (courtesy of Hans Hillewaert)

Figure b: Atlantic Mackerel (Scomber scombrus) (courtesy of Jacek Lesinowski)

Data Used and Quality Assurance

The assessment draws on raw data from the ICES database of groundfish surveys (DATRAS, www.ices.dk/marine-data/data-portals/Pages/DATRAS.aspx). These data have been quality controlled by OSPAR as part of this assessment process to generate a data product for assessment purposes. Time series of Typical Length for fish and elasmobranchs are derived from each available groundfish survey, where the community is separated into demersal and pelagic habitat-based feeding assemblages.

Time series of Typical Length by assemblage were determined for 20 surveys carried out across four separate sub-regions: the Greater North Sea, Celtic Seas, Bay of Biscay and Iberian Coast, and Wider Atlantic (Table a). Ecological sub-divisions were determined for the Greater North Sea using a simplification of those strata proposed by the EU financed project Towards a Joint Monitoring Programme for the North Sea and Celtic Sea (JMP NS/CS) that took place in 2013, and building upon work in the EU VECTORS project (Vectors of Change in European Marine Ecosystems and their Environmental and Socio-Economic Impacts) that examined the significant changes taking place in European seas, their causes, and the impacts they will have on society. In other OSPAR regions, the strata from the survey design were considered appropriate to represent the ecological sub-divisions. The survey areas and sub-divisions are shown in Figure ctoFigure r. Details on the hauls (samples) for each sub-division in different years has been presented in Table b to Table x.

Standard data collected on these surveys consists of numbers of each species of fish sampled in each sample, measured to defined length categories (i.e. so a fish with a recorded length of 14 cm would be between 14.00 cm and 14.99 cm in length). By dividing the species catch numbers-at-length by the area swept by the trawl on each sampling occasion, these catch data are converted to standardised estimates of fish density-at-length, by species, at each sampling location. However, the indicator is based on biomass rather than abundance, so these abundance densities have to be converted to biomass density data by applying species weight (w) at length relationships (of the form w = aLb, where a and b are species-specific parameters). Density estimates per length category per species based on biomass (kg per km2) are referred to below as catch-per-unit-area (CPUA).

These trawl-sample density-at-length estimates are averaged retaining year, species and length category information across all trawl samples within each sampling stratum (i.e. survey specific strata following the survey design, which is a rectangular grid in the Greater North Sea and generally depth-based strata elsewhere).

Table a: Groundfish surveys, region in which they operate, and period over which they were undertaken
Sub-regionSurvey Accronym1Survey period
Greater North SeaGNSEngBT31990–2015
GNSFraOT41988–2015
GNSGerBT32002–2015
GNSIntOT11983–2016
GNSIntOT31998–2016
GNSNetBT31999–2015
Celtic SeasCSFraOT421997–2015
CSEngBT31993–2015
CSIreOT42003–2015
CSNIrOT11992–2015
CSNIrOT41992–2015
CSScoOT11985–2016
CSScoOT41995–2015
Bay of Biscay and Iberian CoastBBIC(n)SpaOT41990–2014
BBIC(s)SpaOT11993–2014
BBIC(s)SpaOT41997–2014
BBICPorOT42002–2014
CSBBFraOT421997–2015
Wider AtlanticWAScoOT31999–2015
WASpaOT32001–2014

1.Survey acronym convention: first two to four capitalised letters indicate the European Union Marine Strategy Framework Directive (MSFD) sub-region (BBIC: Bay of Biscay and Iberian Coast; CS: Celtic Seas; GNS: Greater North Sea; WA: Wider Atlantic). Next capitalised and lower case letters signify the country involved (Spa: Spain; Por: Portugal; Fra: France; Eng: England; Ire: Republic of Ireland; NIr: Northern Ireland; Sco: Scotland; Ger: Germany; Int: International; Net: The Netherlands).

International refers to the two international groundfish surveys carried out in the Greater North Sea under the auspices of ICES. In the Bay of Biscay and Iberian Coast sub-region, Spanish surveys are further delimited by (n) for surveys operating in the northern Iberian coast area and (s) for surveys operating in the southern Iberian coast area).

Next two capitalised letters indicate the type of survey (OT: otter trawl; BT: beam trawl). Final number indicates the season in which the survey is primarily undertaken (1: January to March; 3: July to September; 4: October to December).

2.This is a single survey that operates across both the Celtic Seas and the Bay of Biscay and Iberian Coast sub-regions, from the southern coast of the Republic of Ireland and down the western Atlantic coast of France. For assessment purposes this single survey was split into its two sub-regional components.

Data Treatment

Surveys with rectangular sampling grids (GNSIntOT1, GNSIntOT3, GNSNetBT3, GNSGerBT3, GNSFraOT4)

Catch per unit swept area (CPUA) data (kg / km2) from multiple hauls are averaged by species for each rectangular grid cell using the strata below; in the Greater North Sea these are ICES statistical rectangles, in the eastern English Channel a mini-grid (0.25° by 0.25°) is used by GNSFraOT4. The resulting rectangle-based CPUA estimates are multiplied by the area (km2) of their rectangles (using a Lambert equal area projection) to give species biomass-at-length (now measured in kg per rectangle). Sub-divisional strata level (not GNSFraOT4) estimates of biomass-at-length are given by the sum of the rectangle-based biomass-at-length estimates and corrected by a scaling factor = 1 / (proportion of the area of sub-division monitored in the survey year) (units are now tonnes per sub-division). The scaling factor correction ensures that the weighting of the strata relative to each other in each year is not altered by the sampling levels. Sub-divisional estimates of Typical Length are calculated at this point for investigating local responses of each assemblage.

Regional estimates of biomass-at-length are estimated from the sum of sub-divisions (or in the case of GNSFraOT4 by the rectangle-based estimates). Typical Length is calculated from these data to give a survey level assessment within each region.

Figure c: Greater North Sea surveys with rectangular grids and sub-divisions used (note only GNSIntOT1 and GNSIntOT3 cover every sub-division appropriately)

The colours differentiate the different spatial sub-divisions used.

Table b: Hauls (samples) from in the GNSIntOT1 survey generating data used in the assessment
YearCentral WestKattegat & SkaggerakNorth EastOrkney& ShetlandSouth EastSouth WestTotal
19836433623514042376
19846936825615951453
19858131946517961511
19867038916520157522
19878843946118758531
19886535875511445401
19896442903814743424
19906841854310138376
19916937994612347421
1992634277387939338
19935340754911040367
1994714379378146357
1995644476287941332
1996614571407037324
1997634280388843354
19986941764611351396
1999694273378347351
2000694180429847377
20017542744212564422
20028243734410763412
2003814273469965406
2004694271448952367
2005734672469352382
2006734173459349374
2007634267409046348
2008654268479943364
2009644276459844369
2010704171479848375
2011673968449849365
2012664268478846357
2013654266478449353
2014613943368145305
2015644266478248349
2016644463417943334
Table c: Hauls (samples) from in the GNSIntOT3 survey generating data used in the assessment
YearCentral WestKattegat & SkaggerakNorth EastOrkney & ShetlandSouth EastSouth westTotal
1998544054315831268
1999714277507735352
20006875468238309
2001584177417437328
2002604275477333330
2003604270427035319
2004664274487135336
2005544280437335327
2006604169437235320
2007554069466635311
2008544069397934315
2009544145356334272
2010584171466032308
2011584373446634318
2012564170386435304
2013564075465433304
2014584275426334314
2015604374466935327
2016664587516939357

Figure d: Spatial coverage by the Netherlands groundfish survey

The colours differentiate the different spatial sub-divisions used.

Table d: Hauls (samples) from in the GNSNetBT3 survey generating data used in the assessment
YearCentral WestNorth EastOrkney & ShetlandSouth EastSouth WestTotal
1999211129318145
2000211229219146
2001211248310130
2002221129319147
2003211039422150
2004211049719151
20052211410022159
2006221149116144
2007221049416146
2008211238214132
2009211238815139
201022636414109
2011211247315125
2012221149421152
2013221238614137
2014231136515117
2015221249117146

Figure e: Spatial coverage by the German groundfish survey

The colours differentiate the different spatial sub-divisions used.

Table e: Hauls (samples) from in the GNSGerBT3 survey generating data used in the assessment
YearNorth EastSouth EastSouth WestTotal
20021328344
20031328243
20041338253
2005538245
20071332247
20081329244
20091440256
20101339254
20111340255
20121238252
20131240254
2014226230
20151340255

Figure f: Spatial coverage by the French channel otter trawl survey

The colours differentiate the different spatial sub-divisions used.

Table f: GNSFraOT4 samples used in the assessment
YearTotalYearTotal
198866200288
198961200390
199069200483
1991752005101
199254200694
199359200785
199480200892
199579200990
199650201077
199780201193
199873201283
199989201384
200089201494
200197201572
Surveys with irregular depth banded strata (i.e. all surveys other that those with rectangular sampling grids above)

Catch per unit swept area (CPUA) data (kg per km2) from multiple hauls averaged by species for each survey strata. Sub-divisional estimates of biomass-at-length are subsequently given by CPUA multiplied by area of the survey strata (km2, using a Lambert equal area projection). Sub-divisional estimates of Typical Length are calculated at this point for investigating local responses of each assemblage.

Regional sea estimates of biomass-at-length are estimated from the sum of sub-divisional estimates. Typical Length is calculated here for regional sea assessment.

Figure g: Depth strata for GNSEngBT4

The colours differentiate the different sub-divisions used.

Table g: GNSEngBT4 samples used in the assessment
YearFrench coast <25 mMid-ChannelUK coast <25 mTotal
199025221966
199130211970
199228301775
199328251972
199429261974
199529262176
199630281876
199729242073
199830281977
199926232271
200030212172
200129293391
200228282884
200329242073
200426242070
200522171958
200626271770
200726232170
200824261969
200925252171
201023202063
201124251665
201222201860
201323241764
201421261865
201523221661

Figure h: Depth strata for Irish Sea surveys (CSEngBT3, CSNIrOT1 and CSNIrOT4) note only CSEng BT3 includes St George’s Channel

The colours differentiate the different sub-divisions used.

Table h: CSEngBT3 samples used in the assessment
YearIrish Coast, <50 mIsle of Man, 50–100 mSt George's Channel <100 mEastern Irish Sea, <50 mTotal
1993910194987
1994614153065
1995614153065
1996614143266
1997614163066
1998613153064
1999613153064
2000614132962
2001612153063
2002613163065
2003612143062
2004612163064
2005613152963
2006612153063
2007612153063
2008613122960
2009612153063
2010612163064
2011612162963
2012614162965
2013614162965
2014614142963
2015613153064
Table i: CSNIrOT1 samples used in the assessment
YearIrish Coast, < 50 mIsle of Man, 50–100 mEastern Irish Sea, <50 mTotal
199211141035
199318161045
19941913840
19951915842
19961811938
19971914740
19981916944
19991915943
200019161146
200119171046
200221161150
200319161149
200418151144
20051911838
200619141144
200719161146
200819161147
200919151147
201019151148
201118151147
201219151147
201319161149
201419161149
201519161149
Table j: CSNIrOT4 samples used in the assessment
YearIrish Coast, < 50 mIsle of Man, 50–100 mEastern Irish Sea, <50 mTotal
199318161044
19941916742
1995187934
19961916944
19971916944
19981917945
19991817944
20001911939
200119151148
200218161148
200319151148
200419161149
200518161148
200618161145
200719161147
200919161149
201019161149
201119141146
201219161149
201319161149
19922412844
201419161149
201519161049

Figure i: Depth strata for CSIreOT4 survey

The colours differentiate the different sub-divisions used.

Table k: CSIreOT4 samples used in the assessment
YearVIa_CoastVIa_DeepVIa_MediumVIIb_CoastVIIb_DeepVIIb_MediumVIIg_CoastVIIg_MediumVIIj_CoastVIIj_DeepVIIj_Medium
200311121381167223148
200414111299611244135
20051091241377215159
200619111279810264178
200714111269810304186
200814816610513263158
2009211212611812233128
20101413136138143441910
20111681587615335199
201215717129616323237
2013141017712817314226
2014151112513717334197
201515101666515263175
Table l: CSIreOT4 hauls which were not included in the assessment (as they were not in the Celtic Sea region)
YearVIa_SlopeVIIb_SlopeVIIj_Slope
200311
200412
20051137
20065169
20074149
200841710
20093187
20103124
201133
20121106
20132125
20142114
2015394

Figure j: Depth-based strata for CSScoOT1

The colours differentiate the different sub-divisions used.

Table m: CSScoOT1 samples used in the assessment (red1 was excluded from further analysis due to poor sampling and because it is outside of the assessed region)
Yearblue1_lamblue2_lamclyde_lamgreen1_lamgreen2_lamlightblue_lampink_lamred2_lamred3_lamwindsock_lam
1985311116162581
19862815142141
198721018142254
19882917152333
1989291642263
19902101752133
1991113111152243
19922815141361
19932101615234
199431117142151
199531017142161
19964916122261
1997381813215
19983717141161
19992122815226
200031417142261
20013717142261
20023927142261
200331218342252
200421017152262
20054718252372
200658111152262
2007410112252554
20083819272472
20094719262432
201057111252462
20113428144362
201256213273241
201348312266263
20145521153271
2015559344273
201655214253362

Figure k: Depth-based strata for CSScoOT4

The colours differentiate the different sub-divisions used.

Table n: CSScoOT4 samples used in the assessment (the years 1995 and 1996 are excluded from further analysis due to poor sampling)
Yearblue1_lamblue2_lamclyde_lamgray_lamgreen1_lamgreen2_lam
1995222
1996373
1997322364
1998222354
1999322365
2000422373
2001512373
20025123103
20035222115
20045223103
20055123115
2006522115
20076223154
20085123114
20095123114
20113244
201231395
2013451
2014212383
20153113102
Table o: CSScoOT4 samples also used in the assessment (1995 and 1996 excluded from further analysis due to poor sampling)
Yeargreen3_lamgreen4_lamlightblue_lamred1_lamred2_lamwindsock_lam
19953126
199631291
1997414382
1998214291
1999115182
20004164122
2001516893
20026176107
2003517594
20045176115
20056275114
20065127104
20075277155
20084347113
20097367143
20113248135
20128238153
20131714
201451710114
20156246115

Figure l: Northern Celtic Sea strata for CSFraOT4 survey (data located above 48°N only from larger survey area) The colours differentiate the different sub-divisions used.

The colours differentiate the different sub-divisions used.

Table p: CSFraOT4 samples used in the assessment
YearCc3eCc4eCc4wCc5Cn2Cn3Cs4Cs5
1997529322133
1998919324126
1999748234114
2000626124126
20016816335166
20024514444167
20036612347156
2004769245145
2005666348144
200656944493
20078611245164
20086810346134
2009269336134
20104111334125
20114117457135
2012258145104
2013467445175
20147412345126
20153212446126
Table q: CSFraOT4 samples not used in the assessment (Cc6, Cs6, Cc7 and Cs7 are excluded from further analysis because they are not in the assessed region and are all poorly sampled)
YearCs6Cs7Cn2eCc3wCc6Cc7
19972131
199823
1999221322
200022
2001222122
2002231322
2003412131
20041332
2005421141
2006121331
2007322222
200825322
2009222222
2010222332
2011432312
2012423222
201323132
2014353422
2015231521

Bay of Biscay and Iberian Coast

Figure m: Depth-based strata for CSBBFraOT4 (southern area, south of 48°N)

The colours differentiate the different sub-divisions used.

Table r: CSBBFraOT4 samples used in the assessment
YearGn1Gn2Gn3Gn4Gn57Gs1Gs2Gs3Gs4Gs5Gs67
199741115187354233
19981810233263332
1999139186253116
2000241819723335
2001618226223324
20023318187262334
20032314198343324
20042219197333225
20051615179153315
20062415167333225
20073317177344116
20082316198253333
20093317207243143
20102518187243126
20112315217343216
20122416167343144
20133513227433234
20143516205253234
20152617207244334

Figure n: Depth-based strata for BBICnSpaOT4

The colours differentiate the different sub-divisions used.

Table s: BBICnSpaOT4 samples used in the assessment (note only data for the eastern strata available in the data product)
YearABPA
1990614
199111
19922
199313
1994149
1995156
1996145
19974
1998157
1999159
2000137
2001146
2002137
2003136
2004711
2005149
20061411
20071410
20081312
20091311
20101411
2011149
2012149
20131511
20141413

Figure o: Depth-based strata for BBICPorOT4

The colours differentiate the different sub-divisions used.

Table t(i): BBICPorOT4 samples used in the assessment (17, 21, 37 and 42 and all data pre-2005 are excluded from further analysis due to poor sampling)
Year012358910111213141516171819
20022142223221222211
20052341224362233222
20062341324542133122
20073241336541233131
20084131224531233112
200941612357212322122
20103331225831123112
2011414135431232222
20132351328431132212
2014135125532133122
Table t(ii): BBICPorOT4 samples used in the assessment (17, 21, 37 and 42 and all data pre-2005 are excluded from further analysis due to poor sampling) (continuation)
Year20212223242526272930313233363738394042
200211332621432133
200511232552315113441
20062134541315211631
20072113335524431263
20082133344241621351
2009114226314522462
2010311232562413211411
201131122255234222231
2013321225524133112631
2014312273334211421

Figure p: Depth-based strata for BBICsSpaOT1 and BBICsSpaOT4

The colours differentiate the different sub-divisions used.

Table u: BBICsSpaOT1 samples used in the assessment (A and E are excluded from further analysis due to poor sampling)
YearABCDETotal
19932555724
19943814319
19952123320
19973862221
1998115420
199921078128
20001579132
2001978731
20021279634
200431255530
20051268632
200611177632
20074154831
200841156531
200931047630
20104114827
201119106632
20123855223
201341077533
201441067532
Table v: BBICsSpaOT4 samples used in the assessment (A and E are excluded from further analysis due to poor sampling)
YearABCDETotal
1997345517
199831043525
19991474227
2000355720
2001889429
20024986431
200341377233
200441367333
200512610533
20061279533
20071058326
20084957631
2009412511335
20101389535
20111089431
20123117627
201411910434

Wider Atlantic

Figure q: Depth-based strata for WAScoOT3

The colours differentiate the different sub-divisions used.

Table w: WAScoOT3 samples used in the assessment (the outer area ‘mylightblue_lam’ was excluded from further analysis due to poor sampling)
Yearmyblue_lammygreen_lammyred_lammylightblue_lamTotal
1999431641
2001335644
2002225229
2003449760
2005232438
2006127432
2007432642
200833437
2009434341
20118195537
20126184331
20138155230
201411214238
20159214539

Figure r: Depth-based strata for WASpaOT3

The colours differentiate the different sub-divisions used.

Table x: WASpaOT3samples used in the assessment (the outer area ‘mylightblue_lam’ was excluded from further analysis due to poor sampling)
Year1213222311a11bTotal
2001168182021074
2002185172061076
2003207151451172
2004164151151061
200518516155968
2006196151551070
200719517166972
200819516147768
200920514176971
201017917166570
2011198171251071
2012187151641171
201319513206972
201419714177872

 

Overall Assessment Basis

Where multiple surveys were available for assessment, key surveys were prioritised for assessment given the length of time series available and spatial coverage. If these measures were equal between surveys, then whichever surveyed the greatest biomass by assemblage was selected for indicator assessment. The following surveys were considered key:

Greater North Sea

GNSIntOT1 for both demersal and pelagic assemblages was selected as the key survey (preferred) for the Greater North Sea, given that it is the longest survey with the best spatial coverage. For the eastern English Channel, GNSEngBT3 was preferred for demersal assemblage given more consistent sampling here than GNSIntOT1 and GNSFraOT4. GNSFraOT4 was preferred for the pelagic assemblage in the eastern English Channel given the length of time series available.

Celtic Seas

CSScoOT1 for both demersal and pelagic assemblages was preferred over CSScoOT4 and CSIreOT4 due to length of time series. CSIreOT4 for both demersal and pelagic assemblages was preferred for sub-divisions to the west of Ireland and in the northern Celtic Sea, but not in the north where there was overlap with CSScoOT1. CSFraOT4 for both demersal and pelagic assemblages was preferred in sub-divisions of the Celtic Sea, except where overlap occurred with CSIreOT4.

CSEngBT3 for the demersal assemblage was preferred for the Irish Sea over CSNIrOT1 and CSNIrOT4 given its greater spatial coverage. CSNIrOT1 for the pelagic assemblage was preferred for the Irish Sea over CSNIrOT4 given relatively high biomass of the assemblage caught and identical coverage spatially and temporally.

Bay of Biscay and Iberian Coast

BBICsSpaOT1 for both demersal and pelagic assemblages was preferred over BBICsSpaOT4 given the length of the survey. CSBBFraOT4, BBICPorOT4 and BBICnSpaOT4 did not overlap with any other surveys.

 

Time-Series Assessment

In each case, the minimum value observed over the time series, prior to the last six years, was considered as a lower limit that should be avoided in future. The long-term trend in each time series (sub-division and survey level) was modelled through the application of a LOESS smoother (i.e. locally weighted scatterplot smoothing) with a simple ‘fixed span’ of one decade.

Breakpoint analyses uses data to define stable underlying periods (see Probst and Stelzenmüller, 2015). The method makes it possible to say whether there is a significant change in the time series state over time, namely whether the recent period is not significantly different from the historically observed period. The method avoids the arbitrary choice of reference periods for assessment (i.e. how many years to use to calculate an average) which can lead to subjective assessments. The shorter the period chosen, the more likely it is that noise in the data or natural fluctuations in the system are being compared against each other. However, too long a period and it could be that actual changes in state are averaged out. The minimum detectable period is defined in this analysis as three years. The analysis uses two statistical approaches: First applying the ‘supremum F test’ to establish whether a non-stationary time series or a constant period for the entire time series is more suitable. If the former, then breakpoint analysis is applied to find periods of at least three years duration.

Populations should have a size structure indicative of sustainable populations and should occur at levels that ensure long-term sustainability in line with prevailing conditions. There should be no significant adverse change in the function of different trophic assemblage levels due to human activities. Appropriate baselines for both demersal and pelagic assemblages are not currently available to determine assessment thresholds. The current assessment uses a time-series approach to identify long-term changes in state and further investigation is required to identify if reductions in the size structure of assemblages is due to human activities, food web interactions or prevailing climatic conditions.

Results

The results of this assessment (Figure 2) apply at the community level and do not identify particular species.

Greater North Sea

The assessed demersal fish assemblage is recovering at the scale of the Greater North Sea as a whole due to recent increases in typical length indicator in some sub-divisions: Orkney / Shetland, Kattegat / Skagerrak and the United Kingdom coast in the English Channel. However, the current level is low relative to observed size structure in the early 1980s. Areas of concern, with long-term declines to lowest observed levels remain in the south-eastern and central-western North Sea. The pelagic fish assemblage generally shows fluctuations without trend, with the exception of a long-term decrease to a minimum level in the south-eastern North Sea.

Celtic Seas

Although the surveys showed mixed signals within the Celtic Seas for the typical length of the demersal fish assemblage, surveys in the north suggest some recovery from previous low states with increases to the west of Scotland. However, decreases are also apparent for shelf edge waters to the west. Elsewhere the picture is similarly mixed with decreases near the Irish coast of the Irish Sea and in the Clyde area, but increases to the south of Ireland, Isle of Man, Sea of the Hebrides, and The Minch. The pelagic fish assemblage generally shows no long-term change at the sub-regional level. However, increases are seen to the south of Ireland and decreases in some northerly areas including the Sea of the Hebrides and in coastal areas in the Irish Sea and to the west of Ireland.

Bay of Biscay and Iberian Coast

The typical length of the demersal fish assemblage has increased in this region due to long-term increases in northerly sub-divisions in shelf waters to the west of France and in the coastal area of the Sea of Cadiz. Many sub-divisions to the west of Portugal have also shown increases, in contrast to decreases in some areas to the south. The pelagic fish assemblage generally showed no long-term change. However, decreases to a low state relative to previously observed size structure were identified in northerly sub-divisions in shelf waters to the west of France.

Wider Atlantic

The typical length of the demersal fish assemblage has increased at the Porcupine Bank and the Rockall Bank. While fluctuations without long-term change in size structure have been shown in the pelagic fish assemblages, in the recent period (last six years) a linear increase has been shown for the Porcupine Bank.

There is moderate / low confidence in the method for this assessment and high confidence for data availability

Figure 2: Spatial pattern of Typical Length indicator and time series for key surveys

Typical Length for fish and elasmobranchs is separated into demersal assemblages (left) and pelagic assemblages (right) for sub-divisions for key surveys, where data are available. The duration of the period for which long-term change is defined is dependent on the survey data available, all time periods considered are over ten years long.

Figure s to Figure ay show time series for each survey sub-division. The label for the different sub-divisions is above the plot. For each figure the plot labelled ‘sea’ is a time series of the aggregated survey data for demersal and pelagic assemblages. Each mini-heading shows the p value for the supremum F test which demonstrates whether a significant long-term change is evident (the changes are shown by red dashed lines if significant, or a grey dashed line is used to show a mean level for the whole time series). Annual estimates are shown by blue circles with a fitted LOESS smooth plot (black line) with an estimate of spread shown (± 1 standard deviation). The solid horizontal blue line shows the minimum observed data point prior to the most recent six data points and two horizontal thin black lines showing the average indicator value for the first and last six years.

Greater North Sea

Figure s: Time series of Typical Length for each sub-division of the GNSIntOT1 survey (Demersal fish species)

Figure t: Time series of Typical Length for each sub-division of the GNSIntOT1 survey (Pelagic fish)

Figure u: Time series of Typical Length for each sub-division of the GNSIntOT3 survey (Demersal fish)

Figure v: Time series of Typical Length for each sub-division of the GNSIntOT3 survey (Pelagic fish)

Figure w: Time series of Typical length for each sub-division of the GNSNetBT3 survey (Demersal fish)

Figure x: Time series of Typical Length for each sub-division of the GNSGerBT3 survey (Demersal fish)

Figure y: Time series of Typical Length for each sub-division of the GNSEngBT3 survey (Demersal fish)

Figure z: Time series of Typical Length for each sub-division of the GNSFraOT4 survey (Pelagic fish)

Celtic Seas

Figure aa: Time series of Typical Length for each sub-division of the CSEngBT3 survey (Demersal fish)

Figure ab: Time series of Typical Length for each sub-division of the CSIreOT4 survey (Demersal fish)

Figure ac: Time series of Typical Length for each sub-division of the CSIreOT4 survey (Pelagic fish)

Figure ad: Time series of Typical Length for each sub-division of the CSNIrOT1 survey (Demersal fish)

Figure ae: Time series of Typical Length for each sub-division of the CSNIrOT1 survey (Pelagic fish)

Figure af: Time series of Typical Length for each sub-division of the CSNIrOT4 survey (Demersal fish)

Figure ag: Time series of Typical Length for each sub-division of the CSNIrOT4 survey (Pelagic fish)

Figure ah: Time series of Typical Length for each sub-division of the CSScoOT1 survey (Demersal fish)

Figure ai: Time series of Typical Length for each sub-division of the CSScoOT1 survey (Pelagic fish)

Figure aj: Time series of Typical Length for each sub-division of the CSScoOT4 survey (Pelagic fish)

Figure ak: Time series of Typical Length for each sub-division of the CSFraOT4 survey (Demersal fish).

Figure al: Time series of Typical Length for each sub-division of the CSFraOT4 survey (Pelagic fish)

Bay of Biscay and the Iberian coast

Figure am: Time series of Typical Length for the CSBBFraOT4 survey (Demersal fish)

Figure an: Time series of Typical Length for each sub-division of the CSBBFraOT4 data (Pelagic fish)

Figure ao: Time series of Typical Length for each sub-division of the BBIC(n)SpaOT4 data (Demersal fish)

Figure ap: Time series of Typical Length for each sub-division of the BBIC(n)SpaOT4 survey (Pelagic fish)

Figure aq: Time series of Typical Length for each sub-division of the BBIC(s)SpaOT1 survey (Demersal fish)

Figure ar: Time series of Typical Length for each sub-division of the BBIC(s)SpaOT1 survey (Pelagic fish)

Figure as: Time series of Typical Length for each sub-division of the BBIC(s)SpaOT4 survey (Demersal fish)

Figure at: Time series of Typical Length for each sub-division of the BBIC(s)SpaOT4 survey (Pelagic fish)

Figure au: Time series of Typical Length for each sub-division of the BBICPorOT4 survey (Demersal fish)

Wider Atlantic

Figure av: Time series of Typical Length for each sub-division of the WAScoOT3 survey (Demersal fish)

Figure aw: Time series of Typical Length for each sub-division of the WAScoOT3 survey (Pelagic fish)

Figure ax: Time series of Typical Length for each sub-division of the WASpaOT3 survey (Demersal fish)

Figure ay: Time series of Typical Length for each sub-division of the WASpaOT3 survey (Pelagic fish)

Assessment of Confidence

The method has been developed specifically for this assessment. There is consensus within the scientific community regarding this methodology, however further methodological development is required therefore, it has been rated as moderate / low.

There are no significant data gaps and there is sufficient spatial coverage, confidence for data availability is rated as high.

Conclusion

Long-term decreases in Typical Length, between the 1980s and 2000s in the Greater North Sea and from the 1990s to 2005 in the Celtic Seas, imply that the size structure of fish communities deteriorated such that communities are now more dominated by small-bodied fish. In the Wider Atlantic and Bay of Biscay and Iberian Coast, an overall increase has been observed since 2010.

However, while the indicator in demersal fish assemblages is often still at a relatively low value, recovery since 2010 appears to be underway in the Typical Length of demersal fish and elasmobranchs in the Greater North Sea and Celtic Seas, overall or at least in some sub-divisions. The pelagic fish assemblage shows no long-term change in much of the OSPAR Maritime Area.

In the fish and elasmobranch community, Typical Length responds to changes in the dynamics of the size distribution across the full assemblage including both large and small fish, yet the indicator is still robust to outliers in the data. Typical Length can be directly compared across geographic regions and the indicator can be computed for pelagic or demersal species. The sub-divisional strata are a useful means to capture local patterns in the indicator for specific benthic and water column habitats and the often-local impacts of pressures.

Within the Greater North Sea there are clear sub-divisional differences, with demersal and pelagic assemblages in the northerly areas now recovering, while the southerly areas continue to decline. For the Celtic Seas, decreases in the demersal fish assemblage appear greatest at the shelf edge, while decreases in pelagic fish occur in coastal areas.

While fisheries may have contributed to this depletion, it is unclear whether rising sea temperatures have led to increases in small-bodied fish (i.e. young fish and / or small species). In the Bay of Biscay, where declines are seen in the size structure of pelagic fish, increases are evident in demersal fish.

In addition to the key surveys for each assessment region, additional survey information was assessed which generally confirmed the overall conclusions.

The increase in the typical length of demersal fish since 2010 in the Greater North Sea, evident in the International Bottom Trawl Survey (IBTS) quarter one (Q1) survey, was also shown in the quarter three (Q3) survey (GNSIntOT1 and GNSIntOT3), while the more spatially restricted groundfish surveys showed no significant change (GNSNetBT3, GNSGerBT3, GNSEngBT3). For pelagic fish, no change was evident in the two IBTS surveys in the Greater North Sea.

Within the Celtic Seas, increases in the typical length of demersal fish were evident to the west of Scotland in two surveys (CSScoOT1 since a low value in 2010 and CSScoOT4 since 2005), increases were evident in the Irish Sea in two surveys (CSNIrOT1 and CSNIrOT4 since 2010) with no change in a third (CSEngBT3), to the south and west of Ireland increases were evident in one survey (CSIreOT4 since 2010). The pelagic fish generally showed no significant changes in typical length, with the exception of a decrease in the Irish Sea in one survey (CSNIrOT4 in 1998).

For the Bay of Biscay and Iberian coast, the typical length of demersal fish increased in two of the five available surveys (CSBBFraOT4 since 2004 and BBICnSpaOT4 since 1998) and no overall changes in the typical length of pelagic fish were detected.

An overall increase since 2002 in the typical length of demersal fish in the Wider Atlantic were significant in both assessed surveys (WAScoOT3 and WASpaOT3). No change in the typical length of pelagic fish was evident in WAScoOT3 but an increase from 2012 was evident in the WASpaOT3 survey.

Knowledge Gaps

Further work is required to evaluate appropriate baselines and assessment values for this indicator. This is necessary because any historical baseline for the fish and elasmobranch community is likely to represent an impacted state. Assessment values should preferably be identified through multi-species modelling.

The setting of assessment values for this indicator should consider their relation to the European Union’s Common Fisheries Policy targets aiming at Maximum Sustainable Yield and in relation to other fish community indicators.

Until more comprehensive investigations are complete, the minimum observed typical length in the available time series can be considered as a precautionary limit for the indicator. If indicator scores are at a minimum observed state, a positive (increasing) trend should be evident to avoid falling below the limit.

While reductions in fishing pressure in recent years appear to be driving improvements in the size structure of the demersal fish community in some areas, it should not be forgotten that the OSPAR Maritime Area has also warmed significantly recently (IPCC, 2014). These prevailing conditions may mean species composition is changing. Since Lusitanian (warm-water southern) species tend to be smaller bodied than boreal (cold-water northern) species, the size-structure may require longer than expected to recover to its historic values, if possible.

Barnes et al. (2010) Global patterns in predator–prey size relationships reveal size dependency of trophic transfer efficiency Ecology, 91(1), 222–232

Boudreau P. R., L. M. Dickie (1992) Biomass Spectra of Aquatic Ecosystems in Relation to Fisheries Yield. Canadian Journal of Fisheries and Aquatic Sciences, 1992, 49:1528-1538, 10.1139/f92-169

Commission Regulation (EC) No 665/2008 of 14 July 2008 laying down detailed rules for the application of Council Regulation (EC) No 199/2008 concerning the establishment of a Community framework for the collection, management and use of data in the fisheries sector and support for scientific advice regarding the Common Fisheries Policy,http://data.europa.eu/eli/reg/2008/665/oj

Daufresne M., Lengfellner K. and Sommer U. (2009) Global warming benefits the small in aquatic ecosystems PNAS 106 (31) 12788-12793, doi:10.1073/pnas.0902080106

Fraser, H. M., Greenstreet, S. P. R., and Piet, G. J. 2007. Taking account of catchability in groundfish survey trawls: implications for estimating demersal fish biomass. ICES Journal of Marine Science, 64: 1800–1819

Fung, T., Farnsworth, K. D., Shephard, S., Reid, D. G., and Rossberg, A. G. 2013. Why the size structure of marine communities can require decades to recover from fishing. Marine Ecology Progress Series, 484, 155—171. doi:10.3354/meps10305.

Gibert JP, DeLong JP. 2014 Temperature alters food web body-size structure. Biol. Lett. 10: 20140473. http://dx.doi.org/10.1098/rsbl.2014.0473

ICES, 2014a - Report of the Working Group on the Ecosystem Effects of Fishing Activities (WGECO). ICES Document CM 2014/ACOM:26, Copenhagen, Section 3.4.3 (http://tinyurl.com/p8vwu7d)

ICES. 2014b. Interim Report of the Working Group on Multispecies Assessment Methods (WGSAM), 20–24 October 2014, London, UK. ICES CM 2014/SSGSUE:11

IPCC, 2014: Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp.

Jennings, S., Oliveira, J. A. A. D. and Warr, K. J. (2007), Measurement of body size and abundance in tests of macroecological and food web theory. Journal of Animal Ecology, 76: 72–82. doi:10.1111/j.1365-2656.2006.01180.x

Kerr, S. R. & Dickie, L. M. 2001. The biomass spectrum: a predator– prey theory of aquatic production. New York, NY:Columbia University Press.

Lynam C.P. and A.G. Rossberg. (2017) New univariate characterization of fish community size structure improves precision beyound the Large Fish Indicator. Available: at arXiv:1707.06569

Probst WN, Stelzenmüller V (2015) A benchmarking and assessment framework to operationalise ecological indicators based on time series analysis. Ecological Indicators 55: 94-106, doi:10.1016/j.ecolind.2015.02.035

Reum, J. C. P., Jennings, S. and Hunsicker, M. E. (2015), Implications of scaled δ15N fractionation for community predator–prey body mass ratio estimates in size-structured food webs. J Anim Ecol, 84: 1618–1627. doi:10.1111/1365-2656.12405

Riede, J. O., Brose, U., Ebenman, B., Jacob, U., Thompson, R., Townsend, C. R. and Jonsson, T. (2011), Stepping in Elton’s footprints: a general scaling model for body masses and trophic levels across ecosystems. Ecology Letters, 14: 169–178. doi:10.1111/j.1461-0248.2010.01568.x

Rossberg, A. G., Ishii, R., Amemiya, T. and Itoh, K. (2008). The top-down mechanism for body-mass–abundance scaling. Ecology, 89: 567–580. doi:10.1890/07-0124.1

Rossberg, A. G. (2012). A complete analytic theory for structure and dynamics of populations and communities spanning wide ranges in body size. Advances in Ecological Research, 46, 429-522

Shephard, S., Fung, T., Houle, J. E., Farnsworth, K. D., Reid, D. G., and Rossberg, A. G. 2012. Size-selective fishing drives species composition in the Celtic Sea. ICES Journal of Marine Science, 69: 223–234

Walker, N. D., Maxwell, D. L., Le Quesne, W. J. F., and Jennings, S. (2017) Estimating efficiency of survey and commercial trawl gears from comparisons of catch-ratios. ICES Journal of Marine Science, doi:10.1093/icesjms/fsw250