Please noe ha his is an auhor-produced PDF of an aricle acceped for publicaion following peer review. The definiive publisher-auhenicaed version is available on he publisher Web sie ICES Journal of Marine Science DEC 2005; 62(8) : 1647-1664 doi:10.1016/j.icesjms.2005.06.009 2005 Inernaional Council for he Exploraion of he Sea Published by Elsevier Ld Archimer hp://www.ifremer.fr/docelec/ Archive Insiuionnelle de l Ifremer Combining indicaor rends o assess ongoing changes in exploied fish communiies: diagnosic of communiies off he coass of France Marie-Joëlle Roche a, *, Verena Trenkel a, Rober Bellail b, Franck Coppin c, Olivier Le Pape a, Jean- Claude Mahé b, Jocelyne Morin d, Jean-Charles Poulard a, Ivan Schlaich d, Arnauld Souple e, Yves Vérin c and Jacques Berrand a a IFREMER, Rue de l'ile d'yeu, B.P. 21105, 44311 Nanes Cedex 03, France b IFREMER, 8 rue François Toullec, 56000 Lorien, France c IFREMER, 150, quai Gambea, BP 699, 62321 Boulogne-sur-Mer Cedex, France d IFREMER, Avenue du Général de Gaulle, B.P. 32, 14520 Por-en-Bessin, France e IFREMER, Boulevard Jean Monne, BP 171, 34203 Sèe Cedex, France *: Correspondence o M-J. Roche: el: +33 2 40 37 41 21; fax: +33 2 40 37 40 75 ; mailo:mjroche@ifremer.fr. Absrac: We presen a mehod for combining individual indicaor resuls ino a comprehensive diagnosic of fishing impacs on fish populaions and communiies. A concepual framework for inerpreing combined rends in a se of simple indicaors is proposed, relying beforehand on qualiaive expecaions anchored in ecological heory. The iniial sae of he communiy is firs assessed using published informaion. Which combinaions of rends are accepable or undesirable is decided, depending on he iniial saus. The indicaors are hen calculaed from a ime-series and heir ime rends are esimaed as he slopes of linear models. Finally, he es resuls are combined wihin he predefined framework, providing a diagnosic on he dynamics of fishing impacs on populaions and communiies. The mehod is demonsraed for nine coasal and shelf-sea fish communiies moniored by French surveys. Mos communiies were persisenly or increasingly impaced by fishing. In addiion, climae change seems o have conribued o changes in Eas Alanic communiies. Résumé : Ce aricle propose une méhode uilisan des indicaeurs pour élaborer un diagnosic sur les effes de la pêche sur les populaions e les peuplemens de poissons. Un cadre concepuel perme d'inerpréer les endances conjoines d'indicaeurs à parir de la héorie écologique. L'éa iniial du peuplemen es d'abord évalué sur la base d'informaions publiées. En foncion de l'éa iniial e des objecifs de gesion, les combinaisons des endances son qualifiées d'indésirables ou saisfaisanes. Les indicaeurs son ensuie esimés à parir de données de campagnes de pêches scienifiques: abondance e longueur moyenne d'une sélecion de populaions, nombre, biomasse oale, poids moyen e longueur moyenne dans le peuplemen, e la pene du specre de aille mulispécifique. Les endances emporelles de ces indicaeurs son esées au moyen d'un modèle linéaire, e les résulas des ess son combinés en un diagnosic final. La méhode es mise en œuvre pour neuf peuplemens de poissons côiers e du plaeau coninenal, suivis par des campagnes françaises. Il en résule que la plupar de ces peuplemens son affecés par la pêche de manière saionnaire ou croissane. Par ailleurs, les changemens climaiques dans l'alanique Nord-Es conribuen aussi à des modificaions dans les peuplemens de poissons de cee région. Keywords: ecosysem approach o fisheries managemen; indicaors; mulispecies fisheries; scienific surveys 1
Inroducion There is increasing ineres in ecosysem-based fisheries managemen, from scieniss (Anon., 1999; ICES, 2000), fisheries managemen bodies, and oher sakeholders (FAO, 2001; Council of he European Union, 2002). Wihin his approach, i has been widely recognised ha managemen goals and aspiraions have o be ranslaed ino operaional objecives, which can be racked by indicaors (FAO, 2003; Degnbol and Jarre, 2004). Comparison of indicaors wih arge and limi reference poins provide decision crieria and also measures how well managemen performs. However, whereas reference poins are raher well defined in single-sock assessmens, an ecosysem approach complicaes maers for several reasons. Firs, several indicaors are needed o address he many dimensions of an exploied ecosysem. This makes an indicaor-based assessmen mulivariae in essence (Link, 2005). If every single indicaor could be associaed wih a reference poin, his would give many decision crieria ha would no necessarily be consisen and would have o be reconciled by managers. To address his, mulivariae reference spaces can be mapped in he indicaor space from he analysis of long-erm mulivariae daa series wih conrass among exploiaion periods (Link e al., 2002) or regions (Charve e al., 2000), or from simulaions (Picher and Preikhos, 2001). Second, even for a single objecive such as mainaining he size srucure of a fish communiy, several indicaors may be useful o help inerpre he informaion hey convey (Shin e al., 2005), because whereas indicaors ha are sensiive o fishing are numerous, none of hem is specific (Roche and Trenkel, 2003). Tha is, each of hem varies in response o a number of environmenal facors. Because indicaors migh differ in heir sensiiviy o differen facors, combining several of hem migh help disenangle he effecs of fishing from hose of oher facors. Third, reference poins migh be echnically and heoreically difficul o develop and jusify (Roche and Trenkel, 2003; Jennings and Dulvy, 2005) owing o difficulies in convering broad objecives ino specific limis and arges, insufficien daa-series a appropriae scales, and unavoidable bias inroduced by sampling gears. Insead, Jennings and Dulvy (2005) suggesed ha ime rends in indicaors could be compared wih reference direcions o judge managemen performance by evaluaing wheher ecosysem saus is improving or deerioraing. This is achievable because, for relevan indicaors of he impac of fishing on a communiy, reference direcions are well esablished: we know wheher fishing will increase or decrease he value of he indicaor (Roche and Trenkel, 2003). Fourh, he issue ha reference poins should incorporae medium o long-erm variabiliy in he environmen (Lassen, 1999), in life hisory (Rahikainen and Sephenson, 2004), or in prey availabiliy and in predaion moraliy (Collie and Gislason, 2001) has been raised a he single-sock level. This would be amplified in an ecosysem approach, due o he many dimensions o be considered, and he many ineracing facors affecing each dimension: mulivariae reference spaces would be condiional on he se of environmenal condiions ha prevailed during he period of he daa series, or on he assumpions used in he 2
simulaions. Moreover, a reference space would be moving in an unpredicable manner when environmenal condiions or he socieal background change. A a given ime, i seems more feasible o esablish wheher an ecosysem is moving in a desirable direcion, raher han o deermine how far i is from a broadly specified, moving mulidimensional arge (or limi). This saemen is sill paradoxical: a desirable direcion can only be defined in comparison wih a known arge. The soluion of his paradox lies boh in "a a given ime" and "broadly specified". Whereas i seems unreasonable o deermine a reference region in he space of indicaors relevan o boh managemen objecives and he curren siuaion, i migh be possible o assess wheher a siuaion was saisfacory or no some years laer, using addiional informaion, ha was no available or no inerpreable a he ime. We propose a mehod o combine simulaneous rends in several indicaors ino a diagnosic of he dynamics of a fish communiy. The firs sep consiss of assessing he iniial saus of he communiy relaive o is desirable sae. If a communiy was considered o sar in a non-impaced sae, he absence of ime rends (saionariy) is accepable. However, if a communiy was already impaced, no change is no good news. A rend-based assessmen will have o answer wo quesions, based on differen sources of informaion: i) wha was he saus of he communiy a he beginning of he assessmen period, and ii) has he condiion of he communiy improved or deerioraed since hen? To achieve his, a concepual framework for inerpreing combined rends in a se of simple indicaors of fishing impacs on fish populaions and communiies is proposed, relying beforehand on qualiaive expecaions anchored in ecological heory. Which combinaions are accepable or undesirable can hen be decided, depending on he iniial saus. These indicaors are hen calculaed from a daa imeseries, and heir ime rends are esimaed as he slopes of linear models. Finally, he es resuls are combined wihin he predefined framework, providing a diagnosic on he dynamics of fishing impacs on populaions and communiies. Moving from single-sock owards ecosysem-based managemen means aking noe ha exploiaion may no only modify arge populaions, bu whole communiies, i.e. all species ha direcly inerac wih he arge species hrough compeiion, predaion, oher biological processes, or ha are incidenally aken as by-cach (Hall, 1999). Inegraive communiy indicaors are no ye well developed and esed: i seems reasonable o complemen hem by monioring a wide selecion of populaions, including boh arge and non-arge species (Hall and Mainprize, 2004). This is he reason why our assessmen arges wo levels: populaions and he communiy as a whole. To address he impacs on boh commercial and by-cach species in a comparable way, hese indicaors are esimaed from scienific survey daa. Scienific rawl surveys have been conduced in he seas around France covering nine coasal and shelf sea communiies, wih he primary purpose of providing abundance or recruimen indices for sock assessmens. Recenly hese objecives have been broadened owards a more holisic approach o fisheries assessmen, wih an increasing proporion of axa being idenified, weighed and 3
measured. The primary purpose of his sudy is o implemen, on a large scale, an assessmen of he impac of fishing on hese fish communiies. Combining analyses across exploied ecosysems provides an empirical es of he approach we propose. The price o pay for his comparaive approach is ha he se of indicaors had o be resriced o hose ha could be esimaed by similar mehods across all ecosysems. Maerials and mehods We seleced six coninenal shelf communiies supporing mixed fisheries, which are moniored by boom-rawl surveys (Figure 1, Table 1). We also assessed hree esuarine communiies idenified as nursery areas for commercially imporan socks exploied elsewhere in mixed fisheries; hese communiies are sampled by beam-rawl surveys. The main purpose of he boom rawl surveys was iniially o provide abundance indices of commercial socks, whereas he coasal beam-rawl surveys were designed o provide recruimen indices. In all surveys, all fish are idenified and couned, and mos or all are measured. Hence hese surveys provide a picure of he oal fish communiy, a leas in he mos recen period. In all surveys, he sampling design is sraified according o deph and some oher crierion (e.g. Norh/Souh in he Bay of Biscay, boom subsraum in he Vilaine esuary). 4
-9-6 -3 0 3 6 9 54 IBTS Norh Sea 54 51 CGFS Channel 51 EVHOE Celic Seine Somme 48 Vilaine France 48 45 42 EVHOE Biscay MEDITS Lions MEDITS Corsica 45 42-9 -6-3 0 3 6 9 Figure 1: Map of communiies and survey areas (see deails in Table 1). Survey rawls do no sample all species equally well. Some species are rare or aggregaed in he survey area; some have par of heir range ouside he survey area; ohers escape from he gear for various reasons. This conribues o sampling biases and large sampling variances. Populaion indicaors were esimaed only for he fish species for which a reasonable precision could be achieved (Table 1). The crieria for selecing a species were a sufficien occurrence (proporion of hauls wih he species presen being > 5%), a sufficien densiy and/or commercial ineres, and availabiliy of lengh daa (e.g. in he MEDITS surveys, only a subse of species are measured). The species used for he esimaion of populaion and communiy indicaors are repored a hp://www.ifremer.fr/emh/publicaions.hm, and in Roche e al. (2005). 5
Table 1: Lis of he survey daa used in he analysis. Time series Number of Number of Toal area Deph Number of species Number of species included in Communiy Survey available Season sraa hauls per covered (km²) range (m) caugh populaion analysis year Seine Esuary Seine 1995-2002 Auumn 12 45 550 0-20 56 9 Somme Esuary Somme 1999-2002 Auumn 4 48-54 718 0-20 41 9 Vilaine Esuary Vilaine 1982-2002 Auumn 4 19-46 330 0-20 37 11 Eas Corsica MEDITS (1) 1995-2001 Spring 2 13-25 4 562 10-800 163 22 Gulf of Lions MEDITS (1) 1995-2002 Spring 2 64-76 13 860 10-800 179 22 Souhern Norh Sea IBTS (2) 1990-2000 Winer 80 143-210 252 813 15-100 106 13 Easern Channel CGFS (3) 1997-2002 Auumn 17 83-109 30 672 7-82 76 18 Celic Sea EVHOE (4) 1997-2002 Auumn 12 53-82 150 000 20-400 103 43 Bay of Biscay EVHOE (4) 1987-2002 Auumn 14 56-113 72 500 20-600 194 51 (1) Inernaional boom rawl surveys in he Medierranean (Anonymous, 1998). (2) Inernaional Boom Trawl Survey (ICES, 1996). (3) Channel Ground Fish Survey (Carpenier e al., 1989). (4) EValuaion des ressources Halieuiques de l'oues de l'europe par campagnes de chaluages programmés (ICES, 1991b). 6
Indicaors Seven indicaors for measuring fishing impacs on he populaion and communiy levels were seleced (Table 2). Two indicaors of populaion sae, log abundance, and average lengh in he populaion, were calculaed for a large se of seleced species (see previous secion). Boh are expeced o decrease in exploied socks (Beveron and Hol, 1957), so he direcion o wach ou for is a declining rend. The same inerpreaion of rends was assumed o apply o non-arge species, mos of which are aken as by-cach and hence should suffer similar impacs. Some non-arge species migh increase in abundance because of lower predaion by depleed arge predaor species, or in average lengh because of reduced compeiion (Hall, 1999). However, i was assumed ha for a majoriy of species, hese indirec effecs of fishing would be lower in magniude han he direc effecs. Table 2: Seleced populaion and communiy indicaors. For furher deails see Trenkel and Roche (2003). Level Indicaor Definiion Expeced effec of fishing Populaion ( N ˆ i, ) ln Ln-ransformed populaion abundance for species i i decrease L, Average lengh of populaion i decrease Bˆ Toal biomass decrease Nˆ Toal abundance in he communiy decrease W (1) Average weigh decrease Communiy L Average lengh decrease β (2) Size specrum slope decrease α (2) Size specrum inercep unknown (1) Average weigh was esimaed for he overall communiy as he raio of oal biomass o oal abundance, as individual weighs were no measured. (2) in N l ( ) = α exp ( β ( l l ), where N l is number a lengh l and l is medium lengh in he communiy, year, and he parameers are esimaed using a Generalized Linear Model wih lengh classes weighed by he number of hauls in which he lengh class was presen. Noe: β is generally a negaive number, hence decreasing β implies a seeper (more negaive) slope. 7
Similar direc effecs of fishing are expeced a he communiy level. Toal communiy biomass and number migh decrease under srong fishing pressure, and his decrease would impair he produciviy of he dependen fisheries (Roche and Trenkel, 2003). Average lengh and weigh (all species) decrease in exploied communiies (Jennings e al., 1999; ICES, 2001). The slope and he inercep of he size specrum were also esimaed. Fishing is generally expeced o decrease he slope of he size specrum (Pope and Knighs, 1982; Gislason and Rice, 1998). In order o consider only he descending limb of he specrum, size specra were calculaed for all sizes above he smalles lenghclass considered o be fully caugh by he sampling gear (15 cm for mos surveys). Size specrum slope and inercep were no esimaed for he coasal communiies, where surveys arge juvenile fish < 15 cm. In hese communiies, he variabiliy in size specra characerisics is mainly driven by flucuaions in he recruimen of he dominan species, which are influenced boh by iner-annual environmenal variabiliy and by survey iming, and migh no reflec populaion-level changes in abundance. Having esimaed an indicaor I for a given daa series, he parameers of a linear ime rend in he indicaor I ˆ = a + b were esimaed, and he null hypohesis ha b = 0 was esed (for deails, see Trenkel and Roche, 2003). Two-sided ess were preferred o one-sided ess wih he alernaive hypohesis ha he deeced change is due o fishing. This is because rends in he direcion opposie o fishing impacs also provide a signal, alhough wih a differen meaning (see below). Linear regression analysis assumes ha indicaors are normally disribued. This should be he case for all indicaors ha are mean values of some random variable as a resul of he cenral limi heorem. This migh also be expeced o be rue for oal abundance or biomass in he communiy. Finally, ln-ransformed abundance indices are commonly normally disribued. Hence he fundamenal condiions for simple linear regression can be expeced o be fulfilled for all indicaors used in his sudy. Insead of linear regression, non parameric ess such as he Mann-Kendall es could have been used. The advanage of he linear regression mehod is ha in addiion o providing a es for change, he slope esimaes allow ordering of species by inensiy of rend. From indicaors o communiy diagnosics We firs performed a quick assessmen of he saus of each communiy a he sar of he daa series. Signs of fishing impacs were gahered, based on sock assessmens by advisory bodies, published evidence, and any oher available informaion. We used he following crieria: communiies wih several overexploied commercial socks or bearing a high fishing effor or desrucive fishing mehods (discarding, habia damaging, ec.) were decided impaced; non-impaced communiies were hose wih none of hese characerisics. We defined a framework for inerpreing combinaions of pairs of indicaors a he populaion and communiy levels by seing up diagnosic ables (Table 3). For his firs applicaion of he 8
mehod, inerpreaions were based on common sense and basic ecological heory (Beveron and Hol, 1957; Har and Reynolds, 2002; Sibly e al., 2003). For example, if average lengh of fish in a populaion increased whereas is abundance remained saionary, his means ha here were boh more large fish and fewer small ones (Table 3a). This could be due eiher o reduced moraliy (fishing or naural moraliy, or boh) concomian wih decreased recruimen, or o faser growh, which migh in urn resul from changes in he environmen (food, emperaure, compeiion ), or from a geneic change in growh rae. Changes in recruimen could be due o changes in reproducion (lower reproducive capaciy of he sock, e.g. fishing-induced change in age-srucure, mauriy, fecundiy; disurbance of spawner aggregaions ), changes in larval survival (change in egg-size, in growh or moraliy raes, aribuable o changing environmenal condiions: food/predaion, or larval drif paerns alered by ocean circulaion). More or less fish in he survey area migh also indicae a shif in he disribuion area of he populaion, because of changes in environmenal condiions, in migraion paerns, or in he spaial disribuion of fishing effor. Each cell in able 3 conains a lis of poenial mechanisms, which could be furher disenangled by he use of addiional indicaors (see discussion). A similar framework was developed for communiy indicaors and combined in a wo-dimensional able o faciliae presenaion (Table 3b). Toal abundance and biomass of he communiy were seleced as he wo enries of he able, bu oher pairs could have been used as well. Which indicaor rend combinaions were accepable was decided he basis of inerpreaion ables and depended on he iniial sae assessmen. Based on a precauionary principle, we qualified 'deerioraing' (shaded cells in Table 4a) any combinaion of rends in populaion indicaors for which one of he poenial mechanisms was increased fishing moraliy, or he siuaion requires reducing fishing pressure o reverse he rends, even if no caused by fishing. When he communiy was considered impaced a he beginning of he ime-series, siuaions where none or only one indicaor was rending in he direcion opposie o expeced fishing impac were qualified as 'no improving' (hached cells in Table 5a). A formal es procedure was developed o assess rends in he wo indicaors over all populaions joinly. Probabiliies for each rend combinaion were esimaed, based on he null hypohesis ha populaions were sable and ha rends in log abundance and average lengh are independen. Under a neural model, a communiy is made up of independen individuals wih similar vial raes (Bell, 2001). Hence, populaions (groups of individuals) migh increase or decrease in abundance or average lengh independenly by chance. Under his null hypohesis, performing independen ess on he wo indicaors resuls in he probabiliies of combined rends summarised in Table 4a and 5a. For example, he probabiliy ha boh log-abundance and average lengh significanly increase in a given populaion is equal o he produc of he risk levels of each es, ha is, for wosided ess (α/2)², i.e. 0.025². A wo-sep procedure was used, depending on he iniial sae, o es 9
wheher overall here was evidence ha populaions were moving owards undesirable direcions or saying in undesirable saes (Table 4b and 5b). A he communiy level more han wo rends have o be combined: a sequenial procedure was used, which was summarised as a decision ree (Figure 2). A communiy was said o deeriorae as soon as one indicaor was poining in a deerioraing direcion. In addiion, combinaions of rends ha migh be inerpreed as fishing down he food web or decreasing sysem producion (Table 3b) were also labelled undesirable. By conras, an impaced communiy was said o recover only if boh oal abundance and biomass were increasing (Figure 2b). For example, saring from an impaced sae, he rend in average lengh is considered firs. If i is decreasing, he communiy is said o be deerioraing and he procedure is sopped. If no, he rend in oal biomass is examined in he same way, and so on. The indicaors esimaed wih he bes precision and giving clues abou he inerpreaion of rends (able 3b) enered he decision ree firs. Some indicaors canno be considered independen of ohers. For example, as average weigh was esimaed as he raio of oal biomass o oal abundance, a ime rend in mean weigh can be esed only if boh componens are saionary. The probabiliy of ending in a given combinaion of rends under he null model of a sable neural communiy can be calculaed as he produc of he probabiliy of each pah. Resuls a he populaion and communiy levels were finally combined ino a final diagnosic using a simple rule: as soon as one level was found deerioraing, so was he sysem. Conversely, improvemen a he wo levels was necessary o conclude ha he sysem was recovering. 10
Table 3: Inerpreing rends in indicaors joinly. a) Populaion indicaors: poenial mechanisms for each combinaion of rends in log populaion size and average lengh: direc effec of fishing, environmenal effec, a combinaion of fishing and environmen. b) Communiy indicaors: poenial mechanisms for each combinaion of rends in oal communiy abundance and biomass. In each cell, he firs line gives he mechanism, he bulles sugges poenial consisen signals, he las line an inerpreaion. a) ln ( N ˆ i, ) L i, Increase Saionary Decrease Increase More large fish: moraliy Faser growh decreases (F or M) Shif in spaial disribuion: more large fish More large fish (moraliy decreases or disribuion shif) and decreasing recruimen More fish and faser growh faser growh Saionary More small and large fish: good recruimen and low moraliy (F or M) Shif in spaial disribuion: more fish in survey area More old fish: moraliy decreases (F or M) and slower growh More small fish and faser growh Less small fish: decreasing recruimen or more undersized fish killed Shif in spaial disribuion: less small fish Less fish of any size and Less fish of all sizes: increased moraliy (F or M) and poor recruimen Shif in spaial disribuion: less fish in survey area Less old fish and faser growh Less small fish: decreasing recruimen and slower growh Decrease More fish and slower growh (densiy dependence) More small fish: increasing recruimen or improved seleciviy (undersized fish no killed) Shif in spaial disribuion: more small fish More small fish and increased moraliy (F or M) Slower growh Less large fish: increased moraliy (F or M) Shif in spaial disribuion: less large fish Less fish and slower growh 11
Table 3 (coninued) b) Toal number Nˆ Toal biomass Increase Saionary Decrease Bˆ Increase More animals Several populaions ln ˆ increase ( N i, ) Sysem produciviy increase Decreasing fishing impacs Bigger animals W increase Several populaions L i, increase Improved ransfer efficiency Less and much bigger animals W and L increase β increase Several populaions ln ˆ decrease ( ) N i, Saionary More and ligher animals W decreases Species replacemens Compensaions Many populaion s L i, increase Decreased inpus o he sysem (Primary producion / animal reproducion) Less and bigger animals W increase Decrease β decreases Several pop ln ( N ˆ i, ) increase Several populaions L i, decrease Fishing down he marine food web More and much ligher animals W and L decrease β decreases Several populaions ln ˆ increase ( ) N i, Many populaions L i, decrease Fishing down he marine food web Ligher animals W decreases Several pop L, decrease i Fishing down he marine food web β increases Several pop ln ( N ˆ i, ) decrease Several populaions L i, increase Decreased inpus o he sysem (Primary producion / animal reproducion) Less animals Several populaions ln ˆ decrease ( N i, ) Sysem produciviy decreases Fishing on oo small animals 12
Table 4: Diagnosic able when iniial populaion sae is saisfacory. (a) Expeced probabiliies for combinaions of rends in populaion indicaors under he null hypohesis of no change. Cells where one of he poenial driving mechanisms is increasing fishing impacs are shaded. Numbers are he expeced probabiliy of each cell when individual ess are performed independenly wih alpha = 0.05 (produc of marginal probabiliies). The expeced probabiliy of he shaded area (undesirable rends) is 0.04875. (b) Formal es procedure o decide wheher overall, populaions are moving in undesirable direcions. Tess are conduced sequenially and he procedure is sopped as soon as a conclusion regarding fishing impac is reached. (a) ln ( N ˆ i, ) L i, Increase Saionary Decrease Toal Increase 0.000625 0.02375 0.000625 0.025 Saionary 0.02375 0.9025 0.02375 0.95 Decrease 0.000625 0.02375 0.000625 0.025 Toal 0.025 0.95 0.025 1 (b) Sep Quesion asked H 0 Tes Conclusion 1 Did fishing No rend owards increasing Binomial model: If rejeced, impac appear? fishing impac (number of probabiliy of increasing evidence for populaions in red-shaded area fishing impac increasing consisen wih sable neural p = 0.04875, number of fishing impac. communiy). rials = number of populaions If acceped, go o 2. 2 Did he Saionary populaions G-es (log-likelihood If rejeced, populaions (number of populaions in raio es; Sokal and evidence for remain each cell of Table 4a Rohlf, 1995) comparison change saionary or consisen wih sable neural of expeced and observed probably no move in communiy). frequency disribuion. due o fishing. direcions differen from increasing fishing impacs? If acceped, populaions sable. 13
Table 5: Same as Table 4, when iniial populaion sae is srongly impaced by fishing. (a) Cells where one of he poenial driving mechanisms is increasing fishing impacs are shaded; cells wih suspicion of saionary fishing impacs are hached. The expeced probabiliy of he shaded area (undesirable rends) is 0.04875 and he hached area (no improvemen) 0.950625. (b) Formal es procedure. (b) Se p Quesion asked H 0 Tes Conclusion 1 Did he No change (number of Mulinomial model: cell If rejeced, go o 2. populaions remain saionary? populaions in he hree ypes of cell Table 5a consisen wih sable probabiliies from Table 5a, number of rials = number of populaions. If acceped, populaions remain impaced. communiy). 2 Did populaions No rend owards Binomial model: If rejeced, populaions show evidence of increasing fishing impac probabiliy of deeriorae. increasing fishing impacs, or he opposie? (number of populaions in shaded area consisen wih sable neural increasing fishing impac = p= 0.04875, number of rials = If acceped, populaions improve. communiy). number of populaions. 14
Figure 2: Diagnosic rees for he idenificaion of desirable/undesirable combinaions of rends in communiy indicaors when iniial sae a) bears no srong fishing impac b) is srongly impaced by fishing. One rend es is examined a each kno, and a branch is seleced depending on he es resul. Cells wihou a furher branch are endpoins. Cells are shaded if one of he poenial mechanisms for rends is increasing fishing impac, hached if here is suspicion of saionary fishing impacs. For case a) he probabiliy of ending in each endpoin cell under he null hypohesis of a sable neural communiy is given a he boom. For case b) hey can be calculaed as he produc of probabiliies associaed wih each branch of he pah leading up o he cell under consideraion. Resuls Iniial sae assessmens Mos of he communiies examined were already impaced by fishing a he beginning of he survey periods (Table 6). The wo excepions are he shelf o he eas of Corsica, where indusrial fishing effor is low, and he Vilaine esuary, where he desrucive shrimp fishery declined in he early 1980s (Fores, 1988). 15
Table 6: Assessmen of iniial sae and resuls of he combined rend diagnosics for nine fish communiies around France. Communiy Iniial year Iniial sae Descripion and sources Seine Esuary 1995 Impaced Fish habia loss and low suiabiliy (Riou e al., 2001) Desrucive shrimp and flafish fisheries (Bessineon e al., 1994) Somme Esuary Vilaine Esuary Trends in populaions Trends in communiy Overall diagnosic No improving No improving No improving 1999 Impaced Desrucive shrimp and flafish fisheries (Bessineon e al., 1994) No improving No improving No improving 1982 No srong impac Eas Corsica 1995 No srong impac Decline of he shrimp fishery in he early 80s (Fores, 1988) Moderae decline in diversiy of fish species of commercial ineres (Désaunay and Guéraul, 2003) Low fishing aciviy in he 80s (Lebeau, 1986) and no signal of exension since hen (Relini e al., 1999) Gulf of Lions 1995 Impaced Severe fishing impacs since he 1970s (Meurio e al., 1987; C.G.P.M., 1988) Major socks in a poor sae (Aldeber e al., 1993) Souhern Norh Sea Easern Channel 1990 Impaced Too high levels of exploiaion (ICES, 1991a) Norh Sea cod a risk of collapse (Cook e al., 1997) Long-erm fishing impac on communiies (Rijnsdorp e al., 1996; Jennings e al., 2002) High levels of discarding (Sraoudakis e al., 1998) 1991 Impaced Flafish and gadoid socks ouside safe biological limis (ICES, 1991a), due o high fishing moraliy raes and high levels of discarding (Mellon, 1998) Celic Sea 1997 Impaced A majoriy of socks ouside safe biological limis (ICES, 1998) Long-erm fishing impac on communiy (Pinnegar e al., 2002) High level of discarding (Roche e al., 2002) Bay of Biscay 1987 Impaced Increasing exploiaion level especially on young fish, oo small mesh sizes (ICES, 1991a) Deerioraing (P=0.0128) Saionary Deerioraing Saionary Saionary Saionary No improving No improving No improving Deerioraing (P=0.0026) No improving Deerioraing No improving No improving No improving Deerioraing (P=0.00016) Deerioraing (P<0.021) Deerioraing No improving No improving No improving 16
Coasal communiies In boh he Seine and Somme esuaries, no significan rends were found for any of he indicaors (Table 7, Figure 3a). These already impaced sysems remained impaced (saionary). For he Somme esuary however, his resul migh be due o he very shor ime series (four years only). Table 7: Combined rends in populaion indicaors. In each cell is he number of populaions wih he corresponding combinaion of rends. For shading and haching coding, see Tables 4 and 5. In he Vilaine esuary, hree populaions (pollack, Pollachius pollachius, plaice, Pleuroneces plaessa, and dab, Limanda limanda: P=10-4 o 0.04) decreased and wo (wedge sole, Dicologoglossa cuneaa; and grey gurnard, Eurigla gurnardus) increased among eleven species analysed. This change is aribued o fishing by he binomial es wih a P-level of 0.0128 (3/11 populaions in he shaded region in Table 7). On he oher hand, no sign of a deerioraion was deeced a he communiy level (Figure 3b). 17
Figure 3: Diagnosic rees of combined rends in communiy indicaors. a) Somme and Seine esuaries, Souhern Norh Sea, and Easern Channel; b) Bay of Vilaine and Eas Corsica; c) Gulf of Lions; d) Bay of Biscay; e) Celic Sea. Shading and haching coding as in Figure 2. Only he followed assessmen pahs are shown. 18
Medierranean communiies On he shelf o he eas of Corsica no significan rends in populaion abundances were found. Three rends in average lengh were significan, wih hake (Merluccius merluccius, P=0.0025) and horse mackerel (Trachurus rachurus, P=0.04) decreasing, and hornback ray (Raja clavaa) increasing (P=0.04, Table 7). Among he communiy indicaors, only he slope of he size specrum was found o increase, indicaing an increase in he proporion of large fish in he communiy (Figure 3b). By conras, he absence of any significan change in he Gulf of Lions is worrisome, as his communiy was considered already severely impaced a he beginning of he assessmen. The hornback ray populaion was found o decrease (P=0.003, Table 7), and no a single individual was caugh in 2002 nor in 2003. Two populaions were found o change significanly in average lengh, red mulle, Mullus barbaus (increase, P=0.05) and axillary seabream, Pagellus acarne (decrease, P=0.009). Like in eas Corsica, he only communiy indicaor wih a significan rend was he size specrum slope, i increased (Figure 3c). Norh Sea and English Channel shelf communiies For he Easern Channel and Souhern Norh Sea communiies, here were signs of worsening impacs of fishing on several populaions, whereas communiy-level indicaors remained sable (Tables 6 and 7, Figure 3a). In boh communiies he size specrum was curved and bumpy, so linear regressions were no fied as he inerpreaion of variaions in slopes and inerceps would be senseless. One populaion decreased in abundance in he Easern Channel, black bream (Spondyliosoma canharus, P=0.04), which had a srong year-class in 1997, he firs year of he daa series (Fores, 2001). Two populaions were found o decrease in average lengh (Table 7), herring (Clupea harengus), probably due o he srong 2001 and 2002 year classes (ICES, 2003), and whiing (Merlangius merlangus). Simply applying he assessmen rule of able 4b, he populaions are concluded o be overall deerioraing wih P=0.039. However, if he rends for black bream and herring are no aribued o fishing, he number of deerioraing populaions is one only, and he conclusion ha overall he populaions are no deerioraing canno be rejeced (P=0.37). In he Souhern Norh Sea, wo populaions were found o decrease in abundance, poor cod (Trisoperus minuus, P=0.002) and hornback ray (P=0.04), whereas he grey gurnard populaion was increasing (P=0.01). Two populaions had decreasing average lengh, plaice (P=0.001) and lemon sole (Microsomus ki, P=0.02). Alanic shelf communiies In he Celic Sea and Bay of Biscay, many populaions increased in abundance and many decreased in average lengh (Table 8). Boh observaions applied o imperial scaldfish, Arnoglossus imperialis, in 19
he Celic Sea; wedge sole, cuckoo ray, Leucoraja naevus, and smallspoed cashark, Scyliorhinus canicula, in he Bay of Biscay, a possible explanaion being an increase in recruimen. Many species changed in eiher lengh or abundance, including pelagic, benhic and demersal species. Some were commercial arge species like monkfish (Lophius piscaorius and L. budegassa) and dab, some were ypical discarded bycach like boarfish (Capros aper), and ohers had a low commercial value. Four species had similar posiive rends in boh ecosysems (imperial scaldfish, spoed dragone, Callionymus maculaus, conger eel, Conger conger, and hickback sole, Microchirus variegaus). The communiy indicaors were no compleely consisen wih he populaion indicaors (Table 7 and Figure 3d-e). In he Bay of Biscay, 40% of he populaions examined increased in abundance, bu oal abundance did no increase significanly, whereas communiy biomass did increase significanly (Figure 4), along wih he inercep of he size specrum (suggesing here were more animals of any size in he communiy). The inconsisency beween he significan exponenial increase in he abundance of 20 populaions versus saionary communiy abundance is due o he dominan species in he communiy no increasing, and he linear model no deecing he increase in communiy abundance, because of an oulier in 1994 and he S-shape of he ime-series (Figure 4). We conclude ha fishing impacs were no reduced in he Bay of Biscay, alhough he hypohesis of a sable communiy is rejeced (Figure 3d, Table 6). 20
Table 8 : Slope esimaes of rends in populaion indicaors for he Celic Sea and he Bay of Biscay. *: 0.01<P 0.05, **: 0.001<P 0.01, ***: P 0.001. Species Bay of Biscay, 1987-2002 Celic Sea, 1997-2002 Trend in ln ( N ˆ i, ) Trend in L i, Trend in ln ( N ˆ i, ) Trend in (y -1 ) (cm.y -1 ) (y -1 ) (cm.y -1 ) Argenina silus 0.48* -0.54* Argenina sphyrena -0.52* Arnoglossus imperialis 0.23*** 0.38* -0.37* Arnoglossus laerna 0.12* Buglossidium lueum 0.24** Callionymus lyra 0.23** Callionymus maculaus 0.18** 0.26* Capros aper 0.17* Cepola macrophhalma 0.15*** Chelidonichhys cuculus -0.38* Clupea harengus -0.52** Conger conger 0.17*** 0.20** Dicenrarchus labrax 0.10* Dicologoglossa cuneaa 0.17*** -0.26* Echiichhys vipera 0.20*** Enchelyopus cimbrius 0.18** Gadiculus argeneus -0.23* Galeus melasomus 0.19*** Helicolenus dacyloperus dacyloperus -0.31* Lepidorhombus whiffiagonis -0.46** Leucoraja naevus 0.09* -0.22* -1.60* Limanda limanda 0.37* Liza ramada 0.27** Lophius budegassa 0.14* Lophius piscaorius 0.23* Melanogrammus aeglefinus 0.56* Microchirus variegaus 0.17*** 0.17* Microsomus ki -0.30* Molva molva -11.37* Pleuroneces plaessa -1.14** Raja clavaa -0.89** Sardina pilchardus 0.10* Scomber scombrus 0.12* Scyliorhinus canicula 0.10** -0.34* Solea solea 0.12*** Spondyliosoma canharus 0.14** Spraus spraus 0.38* Trachinus draco -0.52*** Zeus faber 0.13* L i, 21
Toal biomass; P(no rend)= 0.019 Toal biomass; P(no rend)= 0.636 log(oal biomass) 19.0 19.5 20.0 20.5 log(oal biomass) 20.0 20.5 21.0 1990 1995 2000 1997 1998 1999 2000 2001 2002 Toal abundance; P(no rend)= 0.102 Toal abundance; P(no rend)= 0.761 log(oal abundance) 22.0 22.5 23.0 23.5 24.0 log(oal abundance) 22.6 23.0 23.4 23.8 1990 1995 2000 1997 1998 1999 2000 2001 2002 Mean weigh; P(no rend)= 0.242 Mean weigh; P(no rend)= 0.012 Mean weigh 0.02 0.04 0.06 0.08 Mean weigh 0.02 0.04 0.06 0.08 1990 1995 2000 1997 1998 1999 2000 2001 2002 Figure 4: Communiy indicaor rends in he Bay of Biscay (lef) and Celic Sea (righ): oal communiy biomass (kg, op); oal communiy abundance (middle); average weigh in he communiy (kg, boom). Error-bars: 95% confidence inervals. Line: fied linear rend, when significan. 22
In he Celic Sea, here were as many decreases in lengh as increases in abundance, and his resuled in a decrease in he average weigh in he communiy, alhough oal abundance did no decreased nor oal biomass increase, again because of he low power of he linear model for his shor ime-series (Figure 3e and 4). In addiion, he size specrum inercep decreased, suggesing a decrease in numbers a all sizes. Overall here is a srong signal of an increasing impac of fishing in he Celic Sea, on boh fish populaions and he communiy (Table 6). Discussion Fish communiies The overall picure of exploied fish communiies around France is ha impacs of fishing are seady or increasing in eigh of he nine communiies examined. Only wo communiies, he shelf eas of Corsica and he Vilaine esuary, were diagnosed as moderaely impaced a he beginning of he assessmen. This saisfacory sae persised only off Eas Corsica, bu we deeced a deerioraion in populaions of he Bay of Vilaine. In he oher seven ecosysems, already impaced communiies were no recovering. I is worh noing ha all hese communiies suppor, among ohers, ongoing rawl fisheries, whereas Eas Corsica suppors smaller scale fisheries using more selecive and less desrucive gears (Abbes, 1991). Boh Medierranean communiies were saionary, bu he final diagnosics conras wih he iniial saes. In he Gulf of Lions in he early 1970s, he demersal resources were sill considered under-exploied (Bonne, 1973). Firs diagnoses of overexploiaion here occurred afer he rapid developmen of a boom rawling flee in he mid-70s (Meurio e al., 1987). By he beginning of he MEDITS survey, severe impacs had already accumulaed (Table 6), and he presen resuls show ha hey have no lighened. By conras, he shelf eas of Corsica is narrow (9 km wide, surface area 1432 km², compared wih 74 km and 11262 km² for he Gulf of Lions) and has a rough boom, and he island marke has a limied demand for demersal fish. As a resul, a rawler flee did no develop and fishing aciviies in his area remained a a small scale, wih poenial impacs limied o he arge species (snappers, spiny lobser). Changes in boh he direcion of increasing fishing impacs and he converse were deeced for he hree Alanic communiies examined. Pars of hese changes migh be ascribed o he increase in ocean emperaure over he las 30 years (Kousikopoulos e al., 1998; Planque e al., 2003). There is indeed evidence of a srong influence of hydro-climaic condiions on fish communiies in his region, e.g., winer-spring freshwaer supply deermines spaial disribuion and abundance of cerain populaions (Le Pape e al., 2003). The hypohesis of climae-induced changes in species abundance 23
was horoughly examined for he Vilaine esuary and he Bay of Biscay, and could no be rejeced (Désaunay e al., In Press). Anoher relaed consequence of he warming rend is an increase in he number and abundance of ropical species as well as species ha are a heir norhern disribuional limi in he Bay of Biscay (Quéro e al., 1998; Poulard and Blanchard, In Press), ogeher wih a decline in boreal/emperae species (Poulard and Blanchard, In Press). In addiion, we also deeced ypical signs of increasing fishing impacs in he Celic Sea, which were consisen wih previous sudies showing ha in he Celic Sea undersize fish are caugh, mos of which are discarded (Roche e al., 2002; Trenkel and Roche, 2003), and ha he long-erm fishing-induced decline in rophic level of he communiy is ongoing (Pinnegar e al., 2002). According o commercial sock assessmens using age-srucured cach and effor models, mos socks in he Celic Sea are overexploied (ICES, 2002). Boh coasal and norhern shelf-sea communiies bear srong fishing impacs, and deerioraed over he sudy periods. The Norh Sea is well-known as a highly impaced ecosysem (see references in Table 6) and he siuaion is no improving (ICES, 2003). This impac exends o he adjacen English Channel, parly because some flees and socks occupying he wo areas. Species such as whiing and plaice, which have been diagnosed as overexploied in sandard sock assessmens (ICES, 2003), were also found here o be impaced by fishing. We also concluded ha species such as poor cod, black bream and hornback ray, which are no formally assessed and for which lile or no informaion oher han survey daa are available (Fores, 2001) were impaced by fishing. In addiion, he English Channel and adjacen coasal communiies suffer from polluion and oher human impacs (Rybarczyk, 1993; Desprez, 2000), especially Seine Bay (Abarnou e al., 2000). Impaced habias migh impair sensiive processes in he life cycle and limi he resilience of communiies o fishing impacs. One species, hornback ray, changed significanly in several of he communiies examined, eiher in abundance (decreasing in he Gulf of Lions and Norh Sea) or in average lengh (decreasing in he Bay of Biscay, increasing off Eas Corsica). This is no surprising because his species has long been known o be sensiive o fishing owing o is paricular life hisory characerisics (Walker and Hislop, 1998). In he Gulf of Lions a marked decline in mos commercial elasmobranch species has been noed since he 1980s (Aldeber, 1997). The Combined Indicaor Trends Mehod Our proposal for combining rends provides a picure of changes in he communiy as a whole, no jus commercially argeed species. As such, i is complemenary o he formal assessmen of arge socks, and migh give early signals of ongoing changes on differen scales. The mehod has he advanage of clearly separaing he esimaion and assessmen seps. For he esimaion mehod o give a reliable picure of communiy dynamics, i depends on i) an adequae inerpreaion of combined rends in a se 24
of relevan indicaors, ii) he availabiliy of sufficien daa, iii) an appropriae mehod o deec rends. Furher, he final oucome of he assessmen will depend on iv) an adequae assessmen of he iniial sae of he communiies and v) he decisions made abou accepable and undesirable combinaions of rends. The key sep of our approach is o provide a concepual framework for inerpreing combined rends, aiming a making he mos of he informaion enclosed in he indicaors. The framework developed here for a small number of indicaors should be refined. As he number of indicaors increases, he number of poenial combinaions does so in a facorial manner, bu he number of poenial inerpreaions in each cell decreases. The use of sequenial mehods wih end poins as soon as a firm inerpreaion can be reached should limi his problem. In his respec, he approach works like an idenificaion key: saring from broad characerisics and refining unil a final idenificaion can be made. We face he challenge of developping a axonomy of fishing impacs on populaions and communiies, and deermining appropriae sequences of indicaors. Which indicaors are o be examined firs depends on heir relevance o he managemen objecives, heir measurabiliy (availabiliy of daa and precision of indicaor esimaes) and he ime scale considered: he processes lised in Table 3a will respond o fishing and/or he environmen on differen ime frames. Unlikely combinaions (such as decreasing average weigh and increasing average lengh) should rigger furher invesigaions and a check of he field and esimaion mehods. Thinking boh backwards (saring from poenial combinaions and lising poenial mechanisms as done here) and forwards (saring from facors and predicing heir join effecs on indicaors) is necessary o increase he chances of esablishing a complee inerpreaion able. Populaion and communiy models would be useful for his sep. This firs aemp o combine rends was limied by he resriced se of indicaors used. Complemenary indicaors a he populaion level would help disinguish beween he various poenial explanaions of each combinaion of rends in log-abundance and mean lengh (Table 3a). For example, as exemplified by Channel herring, furher descripors of lengh disribuions, e.g. a low and a high percenile, would help o deermine wheher decreasing average lengh is due o more small fish or less large fish or boh. Similarly, properly combined indicaors of individual growh rae, moraliy, recruimen, and sock reproducive capaciy would allow us more firmly o ascribe deeced rends o fishing or o oher causes. Obviously, fishing pressure and environmenal indicaors such as waer emperaure or food abundance could bring useful addiional informaion. The idea is o eliminae poenial causes unil he mos plausible mechanisms of rends are idenified. Because of heir non-specificiy, indicaors canno be examined independenly for heir response o fishing, as done in many previous sudies. The quesion is no wheher an indicaor is relevan for assessing fishing impacs and he efficiency of managemen acions, bu wheher a suie of indicaors is. We found he indicaors seleced here o be generally consisen, excep he size specrum 25
descripors. In some insances we could no esimae he size specrum slope because of obvious nonlineariy. In oher cases, rends were found in he slope of he size specrum, whereas no oher sizebased indicaor was changing. This is probably due o he size specrum being esimaed only from he fully recruied size classes, conrary o he oher indicaors which include all fish caugh. In addiion, he generally acceped expecaion ha fishing should decrease size specra slopes migh be due o an oversimplificaion of he processes governing size specra dynamics (Benoî and Roche, 2004). We conclude ha size specra slopes migh no be very useful as shor o medium erm indicaors of ecosysem dynamics, because heir meaning is no obvious. Descripors oher han slope migh be more appropriae. This sudy also shows ha he appropriaeness of indicaors depends on he characerisics of he communiy being assessed and he expeced effecs of fishing on i. For example, in he coasal nurseries here is a coninuous inpu of small fish; fishing should no affec he size composiion of he communiy as much as abundances and species composiion. This is illusraed by he six populaions ha changed in average lengh in he Easern Channel and he Bay of Biscay, whereas hey did no in he adjacen esuaries. Thus he size-based indicaors do no convey much informaion abou nurseries. One very imporan poin for he implemenaion of he indicaor rends approach is o have a hand a long enough ime series of survey daa wih a consisen design. From he examples shown i is obvious ha 4 years is no enough. On he oher hand, he approach is mean for deecing curren and no long erm changes. However, our capabiliy o acually deec rends depends on he power of he rend ess, which in urn depends on he variabiliy of he indicaor esimaes. Unforunaely, he power of such ess is generally low. Nicholson and Jennings (2004) found ha around 14-16 years of IBTS (Inernaional Boom Trawl Survey) would be required o deec changes in mean lengh or weigh, given he rae a which hese were esimaed o change in he Norh Sea. Hence 10 o 15-years in a daa series seems adequae, alhough no easy o achieve wih a consan survey design. For he EVHOE survey, he scienific vessel changed in 1997. However, iner-calibraion experimens showed ha he difference beween vessels was lower han he uncerainy aribuable o spaial heerogeneiy and naural flucuaions (Pelleier, 1998). The fishing gear and mos oher characerisics of he survey design were kep consan, so he whole series was used for he Bay of Biscay. A he same ime, in he Celic Sea he spaial allocaion of hauls was changed and his led us o remove he earlier par of he series. All of he indicaors examined here, he abundance indices (a leas hose of he mos abundan species which are esimaed wih a sufficien precision), as well as all he size-based indicaors, are sensiive o changes in survey mehod (gear, design, ec.) (Trenkel e al., 2004). As soon as here are changes in he survey mehod, i becomes difficul o deermine wheher he rends in he indicaors are ascribable o he changes in he mehod or o oher impacs. Hence we recommend ha survey mehods be kep as consisen as possible. 26