effects of climate and ocean conditions on the marine survival of irish salmon (salmo salar, l.)

I first explored the levels of synchrony in the marine survival of European salmon, in order to assemble evidence on the spatial scale of the processes controllingsurvival.. Finally, I e

EFFECTS OF CLIMATE AND OCEAN CONDITIONS ON THE MARINE

SURVIVAL OF IRISH SALMON (SALMO SALAR, L.)

A Dissertation Presented

byARNAUD J PEYRONNET

Submitted to the Graduate School of theUniversity of Massachusetts Amherst in partial fulfilment

of the requirements for the degree of

DOCTOR OF PHILOSOPHY

February 2006

Wildlife and Fisheries Conservation

UMI Number: 3206206

3206206 2006

Copyright 2006 by Peyronnet, Arnaud J.

UMI Microform Copyright

All rights reserved This microform edition is protected against unauthorized copying under Title 17, United States Code.

ProQuest Information and Learning Company

300 North Zeeb Road P.O Box 1346 Ann Arbor, MI 48106-1346 All rights reserved.

by ProQuest Information and Learning Company

© Copyright by Arnaud J Peyronnet 2006

All Rights Reserved

EFFECTS OF CLIMATE AND OCEAN CONDITIONS ON THE MARINE

SURVIVAL OF IRISH SALMON (SALMO SALAR, L.)

A Dissertation Presented

byARNAUD J PEYRONNET

Approved as to style and content by:

Francis Juanes, Member

Niall O’Maoileidigh, Member

_Matthew J Kelty, Department Head

Department of Natural Resources Conservation

This work was funded by a Marine Institute grant, under the Marine Institute andNOAA Cooperation programme I wish to express my gratitude to several peoplewho have been instrumental in helping me complete this research

I thank my advisors, Kevin Friedland and Niall Ó Maoiléidigh, for helping medesign this study and conducting this research, and for their patience in reviewing

I wish to thank our director in ACMS and President of NASCO, Ken Whelan, forhis help and support throughout this fellowship and for his infecting enthusiasmabout salmon angling Many thanks also to the ACMS station manager, RussellPoole, and his staff, for providing me with a tremendous administrative andscientific support during my time in Furnace Thanks to Ger Rogan, for his helptracking information on Burrishoole salmon, for his great sense of humour and hisconstant good spirit Thanks also to all my other colleagues from the MarineInstitute, particularly Kathleen Sweeney and the rest of the ACMS team; but alsoGlen Nolan in OSS for his support and advice on oceanographic data

I am very grateful to all the individuals who provided data for this study: Jan-EvenNilsen for the MLD Data; David Johns and Darren Stevens (both in SAHFOS) fortheir help with CPR data; Rowan Fealy from NUI Maynooth; Aidan Murphy in MetEirann; Walter Crozier, Lars-Petter Hansen, Gudni Gudbersson, Lars Karlsson,Julian McLean and Jacques Dumas for the time series of marine survival of

European salmon

I would like to thank my parents, family and friends for their encouragements.Finally my greatest gratitude goes to my wife Joanne, for her tremendous supportand wonderful patience during the last few years This work is dedicated to her

EFFECTS OF CLIMATE AND OCEAN CONDITIONS ON THE MARINE

SURVIVAL OF IRISH SALMON (SALMO SALAR, L.)

FEBRUARY 2006ARNAUD J PEYRONNET, Bsc., UNIVERSITE BORDEAUX I

Msc., UNIVERSITY COLLEGE CORKPh.D., UNIVERSITY OF MASSACHUSETTS AMHERST

Directed by: Dr Kevin D Friedland

This dissertation investigates the role of climate and ocean conditions on the marinesurvival of Atlantic salmon The overall objective was to identify the relevantenvironmental variables controlling salmon marine survival, in order to establishpredictive models of marine recruitment These models are required to improve themanagement process of the Irish and European salmon resources

I first explored the levels of synchrony in the marine survival of European salmon,

in order to assemble evidence on the spatial scale of the processes controllingsurvival I demonstrate that these levels of synchrony are low and I conclude thatlarge scale events are not directly exerting a control on the rates of salmon survival,perhaps indicating the presence of several intermediary processes

Using information from a scale analysis of a monitored Irish population, I thenexplore the hypothesis that marine survival is linked to marine growth I presentevidence that the level of marine recruitment of 1SW salmon is linked to growthduring the marine residency, and that decreasing growth over the last 30 yearsexplains the observed decrease in salmon recruitment

Finally, I explore the role of several environmental variables on salmon marinesurvival, by constructing semi-parametric models (GAMs) of marine survival forwild and hatchery Irish populations These models explain an important part of theinter-annual variability in survival and provide a capability to forecast survival.These models also help to identify the role of specific variables, more specificallythe North Atlantic Oscillation, sea surface temperatures, and the abundance ofzooplankton, to explain the variations in survival.

I conclude that the changes in climate in the northeast Atlantic have affected thesalmon via bottom-up effect, by affecting the abundance, distribution and

phenology of key zooplankton species in the northern North Sea and southernNorwegian Sea

Keywords: Atlantic salmon; Climate; GAM, Growth; Marine survival; NAO;Ocean; Synchrony; Zooplankton

TABLE OF CONTENTS Page

ACKNOWLEDGEMENTS……… iv

ABSTRACT……… ………vi

LIST OF TABLES……….ix

LIST OF FIGURES……… x

CHAPTER 1 INTRODUCTION……….……… ……1

2 IS THERE COHERENCE IN THE MARINE RECRUITMENT OF NORTHEAST ATLANTIC SALMON, SALMO SALAR, L.? IMPLICATIONS FOR THE SPATIAL SCALES OF THE PROCESSES CONTROLLING……… … 11

3 LINK BETWEEN MARINE GROWTH AND SURVIVAL OF IRISH ATLANTIC SALMON (SALMO SALAR, L.)……….…52

4 EFFECTS OF CLIMATE CHANGE ON SALMON (SALMO SALAR, L.) MARINE RECRUITMENT IN THE NORTHEAST ATLANTIC………….76

5 CONCLUSION………112

6 BIBLIOGRAPHY……… ……… …117

LIST OF TABLES

2.1 Summary of the marine survival time series, the coordinates

of the point of entrance to the Sea, the methods used to generate

the data and information on the data holders and their affiliation…….… 44

2.2 Testing hypotheses on the structure of synchrony

in Marine survival of European salmon stocks……… 452.3 Matrix of inter-stocks distances (km)……….….462.4 Summary of pairwise cross-correlation coefficients

between aggregated stocks………….……… 472.5 Mantel tests results for synchrony versus distance

and other hypothetical models of synchrony organization……… …… 513.1 Correlation between growth indices and salmon recruitment………… ….734.1 Final GAM for Wild salmon survival and nested models……… … 1014.2 Numerical output from the model (GAM) of wild salmon survival….……1024.3 Final GAM for Hatchery salmon survival and nested models……… 1064.4 Numerical output for the model (GAM) of hatchery marine survival…… 107

LIST OF FIGURES

2.1 Locations of the European salmon stocks for which time-series of

marine survival were used in a cross-correlation analysis.………432.2 Mean marine survival (and Standard Deviation), for all stocks of

hatchery (H) and wild (W) origins……….…482.3 Mean synchrony in marine survival for all European sub-groups (regions)…49

2.4 Cross-correlation versus inter-stocks distance for all pairs of stocks in the northeast Atlantic……….……….503.1 Indices of Irish salmon recruitment……… 693.2 Mean decadal curves of inter-circuli distances

for Burrishoole Salmon ………703.3 Total marine growth, represented by the cumulative inter-circuli

distances, measured between the last freshwater circuli and the edge

of the scale………….……… ….……….713.4 Total marine growth, represented by the total number of circuli

measured between the last freshwater circuli and the edge of the scale …723.5 Scatterplots of growth versus Burrishoole salmon recruitment………74

3.6 Relationships between the total number of circuli and the numbers of

circuli at first point of maximum growth, first point of minimum

growth and second point of maximum growth……… 754.1 Standardized mean marine survival for hatchery

4.2 Standard Continuous Plankton Recorder regions……… 1004.3 Diagnostic plots for the GAM (Marine survival wild Irish salmon),

showing no violation of the assumption of normality (no pattern for the fitted residuals) and no violation of the assumption of homogeneity

(QQ-plot).……….1034.4 Fitted (GAM) and observed values of wild marine survival……….1044.5 Scatterplots of dependent variables (explanatory) versus marine

survival for wild Irish salmon……… ………105

4.6 Fitted (GAM) and observed values of hatchery marine survival ………….1084.7 Scatterplots of dependent variables (explanatory) versus

marine survival for hatchery Irish salmon ……… ………… 1094.8 Diagnostic plots for the GAM (Marine survival hatchery Irish salmon),

showing no violation of the assumption of normality (no pattern for the fitted residuals) and no violation of the assumption of homogeneity

(QQ-plot)……….1104.9 Link between marine growth (circuli number) and marine survival

for Burrishoole Wild salmon……….……… 111

CHAPTER IINTRODUCTIONThe last century has seen an extraordinary increase in the rate of development of thewestern civilization, but it is the rates of industrial and economic expansion andgrowth in human population that make it a turning point in human history Bothdevelopment and demographic pressure have contributed to equally importantchanges in the marine fisheries, from a localised coastal and artisanal activity to thetechnologically advanced and globalised industry we know today In the early days,catches and yield increased according to the level of effort and the resources wereseemingly unlimited (Hjort 1914) We know today that all marine resources arefinite and can get rapidly depleted by fisheries activities Fishing is responsible forthe collapse of single-species stocks but can also result in other pervading effects atthe ecosystem level, such as the removal of large predatory species (Ward andMyers 2005) and the decrease in trophic levels (Pauly et al 1998) Extreme cases ofover exploitation can even result in species extinction (Casey and Myers 1998).Changing trends in fisheries science have followed the pattern of changes in theindustry This very active and broad scientific discipline regularly provides newmethods and techniques for stock-assessment and management purposes Perhapsmore importantly, it is the overall objectives and fundamental ideas that seem tohave evolved most Sustainability is now central to the long-term management plans

of most commercial stocks (Rosenberg et al 1993; Pauly et al 2002) In addition, ashift is taking place from single species consideration to an ecosystem approach(Gislason et al 2000) Another modern development is the acceptance of

uncertainty (Ludwig et al 1993) and its implications for fishing activities.

Uncertainty derives from the complex nature of the marine ecosystems and our lack

of understanding or ability to conduct investigations at large spatial and temporalscales Further uncertainty is associated with environmental variability The effects

of environmental variability on fish production seem to be receiving increasingattention because of the widespread pattern of overexploitation of marine resources.Indeed, the consequences of environmental variability are more serious for stocksthat are already depleted by fishing activities (Pauly et al 2002) In the case ofexploited populations, the global stock production is often over reliant on strongcohort recruitment, and this exacerbates the role of the environment When

combined together, the global pattern of over-exploitation and the severe climaticchanges we are witnessing provide a very ominous context for the long-term

sustainability of many stocks (IPCC 2001; Walther et al 2002) In this context ofover-exploitation of the majority of the commercial stocks, the variability due toenvironmental conditions therefore appears to have an increasing influence over theoverall levels of abundance The effects of environmental conditions on fish

abundance have been widely reported and continue to stimulate much interest(Klyashtorin 1998; McFarlane et al 2000; Finney et al 2000; Klyashtorin 2001) Inthe north Pacific, climate variations have been linked to trends in fish productionover the last 2200 years (Finney et al 2002) Cycles of climate conditions and theirlink to salmon production have also been identified over multi-decadal periods(Hare and Francis 1995) These patterns are often detected through the presence ofperiods of pronounced high/low production termed regimes and occurring

synchronously to similarly strong/weak climatic signals Often, the biologicaloutcome of the climate patterns is reflected in the levels of production of severalspecies (Aebisher et al 1990) The variation in the overall salmon abundance in thenorth Pacific is mainly influenced by the state of these climate/production regimes(Hare and Francis 1995; Beamish et al 1997, 1999), rather than by the patterns ofproduction in freshwater Similarly, the variations in the abundance of Atlanticsalmon are largely controlled by events taking place during the marine residency(Chadwick 1987; Friedland et al 2005) Over the last three decades, salmon

abundance in the north Atlantic experienced a strong decline, particularly the

southern populations (Jonsson and Jonsson 2004) This decline occurred in thecontext of stable or increasing freshwater recruitment and reductions in the rate ofmarine exploitation Despite increasingly favourable conditions, adult returnsdecreased significantly, indicating a global decrease in marine survival rates

(Parrish et al 1998) This decrease in marine survival is thought to be linked tochanges in the climate and ocean conditions, with effects on the marine ecosystems

of the north Atlantic (Beaugrand and Reid 2003; Jonsson and Jonsson 2004;

al 1998, 2000) In years of warmer Sea Surface Temperatures (SST’s), survival for

a Scottish and a Norwegian population, both bordering the North Sea, were found to

be better than during colder years The survival rates for the Scottish stock werealso linked to the growth rate of the postsmolts (Friedland et al 2000) Survival forBaltic populations is also positively correlated with SST’s in early summer (Kallio-Nyberg et al 2004) Jonsson and Jonsson (2004) suggested that increasing SST’scould increase both food consumption and growth rate, and hence result in highersurvival The patterns of decrease in salmon abundance in the 1980’s, and the factthat it took place in the context of a well reported global increase in temperatures(IPCC 2001), remains an important predicament to support this hypothesis.

Opposite relationships have been identified for north American populations, forwhich abundance is negatively correlated with SST’s and thermal habitat in June(Friedland et al 2003) Negative correlations between salmon abundance and

temperature have also been reported in the northeast Atlantic (Beaugrand and Reid2003) In this case, the increase in Northern Hemisphere Temperatures (NHT) is the

main explanation for the northward shift of the distribution of the copepod Calanus

finmarchicus, with effects on the pelagic ecosystem and on salmon Such changes in

copepod abundance and distribution can result in lower growth rates, which in turnhave control on the overall survival rates In Ireland, previous studies have indicatedthe absence of a link between growth and survival for both the river Bush (Crozierand Kennedy 1999) and the Burrishoole (McLoone 2000) However, links betweengrowth and recruitment have been reported for Pacific salmon (Beamish et al 2004)and for Atlantic salmon (Friedland et al 2000; Jonsson et al 2003), hence providing

an indication of the mechanisms controlling survival

Salmon fisheries taking place in the high sea are managed by the north AtlanticSalmon Conservation Organisation (NASCO), whereas homewater fisheries areunder national jurisdiction In recent years, the management process for the Irishsalmon resource has seen considerable new developments, with changes in scopeand purpose of the stock assessment, the methodology of this assessment, andfinally the quality and the relevance of the advice delivered Up to 2002, the

management regulations only affected the level of effort and were not based on thestatus of the resource A review of the management of the Irish salmon resource(Anon 1996) recommended that the management objective should be the

attainment of conservation limits (CLs) to ensure sufficient escapement of adultspawners, rather than conservation of the resource via effort control Total

Allowable Catch (TAC) was chosen as a technical instrument to ensure that theconservation limits were attained for each salmon stock (i.e fishery districts inIreland) Initially, these CLs were provided by pseudo stock-recruitment analyses,following the methodology used by Potter et al (1998) Both the pseudo-

recruitment and TAC setting-up approaches require information on the number offish returning to spawn This information can be obtained using catch numbers,which are modified to account for the levels of non-reported catches (local

expertise) and exploitation (derived from a Coded Wire Tagging programme,Browne 1982), providing an index of Pre-Fishery Abundance (PFA) Once theeffects of natural mortality are accounted for, the number of spawners can be found

by subtracting the catches from the PFA Run reconstruction models of PFA areavailable for Irish 1SW fish since 1970 (Anon 2003) In fisheries science,

Biological Points References (BRPs) are used to define safe levels of harvests.When spawners numbers are considered with the lagged values for the resultingrecruits, stock-recruitment analysis can be conducted and the BRPs can be derived.However, this approach has several limitations.

Firstly, the CLs are not river specific and correspond to district CLs The attainment

of district CLs will not guarantee that each river can reach its conservation limit interms of spawning escapement In order to overcome this problem, new methodshave been applied to determine CLs for each river There are only few monitoredrivers for which stock-recruitment analysis can be conducted This is because themonitoring of these rivers generally requires a substantial and costly effort

However, a new methodology has been implemented, where the BRPs can betransferred from monitored to non-monitored systems The BRPs for the non-monitored rivers are modified to account for differences in latitude of the river(with both climatic and ecological (Hutchings and Jones 1998) implications) andthe total wetted area available (Prevost et al 2003)

Secondly, under the present stock/recruitment and catch advice processes, theadvice is delivered using information from catches in previous years, and withoutconsideration for the conditions in the marine environment The PFA model used inthe northeast Atlantic is a run-reconstruction model using past information oncatches A pro-active approach requires forecasting PFA according to the number ofspawners in the previous years (or to an index of juvenile abundance), and to

marine conditions A similar forward running model has been established in thenorthwest Atlantic, based on the number of spawners and sea temperatures in the

spring, but the relationship with the thermal habitat did not hold (Crozier et al.2003) A forecasting model of PFA has several advantages over the present back-calculated model It allows one to predict returns, according to the conditions in themarine environment, and to provide advice before the fishing season, on the basis of

a realistic assessment of the status of the stock In addition, it can also help toformulate advice for the high sea fisheries of West Greenland and the Faroes.Finally, as more fisheries close and fishing effort declines, the quality of the back-calculated index of PFA (which relies on good catch information) is decreasing Apredictive model based on spawners numbers/juveniles and environmental

conditions would not be affected by these limitations The development of such aforecasting model for PFA has been considered, but so far our understanding of therelationships between climate and marine conditions has not been strong enough toconstruct such a model (Crozier et al 2003) This study was designed to exploremore explicitly the links between the marine conditions and the levels of

recruitment for Irish populations

The index of PFA is central to the delivery of the scientific advice for catches ofIrish salmon A forecasting model of PFA, based on levels of juvenile productionand environmental conditions, can improve the accuracy of the advice by making it

a pro-active rather than a reactive process Time series of marine survival are

available for several European rivers, usually as a result of trapping and/or taggingprogrammes The number of fish returning to freshwater is compared to the totalnumber of smolts migrating the previous year This proportion is then raised toaccount for the rates of exploitation by the coastal fishery and for the rate of non-

reported catch, hence providing a percentage of returns to the coast, or marinesurvival Since the early 1980’s, these indices display important inter-populationand inter-annual levels of variability However, a common declining trend is presentfor nearly all these populations This variability can be an indication of the diversity

of the processes controlling survival Alternatively, the global decline in

recruitment, also observed on the time-series of PFA, seems to suggest that largescale climate/ocean processes are involved These large-scale events can havedifferent outcomes, by demonstrating disparity in amplitude between

geographically distinct regions For a given wide-scale process, the effects onsalmon survival will be decided through interactions taking place in the specificcontext of the local ecosystem conditions These large-scale events are unlikely todirectly control salmon survival, but it is the number of intermediary processes, andtheir nature, that will dictate their relevance as a predictor for salmon recruitment

In order to limit the number of potential predictors, it would be useful to get aninsight into the spatial scale of these events/processes A similar approach has beenused to identify the factors controlling the marine recruitment of Pacific salmon

(Oncorhynchus spp.) (Pyper et al 2001, 2002; Mueter et al 2002, 2005) The levels

of coherence/synchrony in marine survival between distant populations were used

to infer the spatial scale of the factors controlling survival The idea being that ifsurvival is controlled by basin-wide (northeast Atlantic) processes, there should beimportant similarities in recruitment rates between populations, even at large

distances Alternatively, a lack of synchrony could be a reflection of differences inlife-history traits between populations (e.g proportion of multi sea-winter),

differences in the timing ( or the occurrence of distinct factors controlling thesurvival of the various populations (i.e presence of lags due to the number ofintermediary processes).

In the first chapter, I compare the inter-annual variation in the rates of marinesurvival of 18 European populations using cross-correlation analysis The patterns

of coherence in marine survival across Europe and their link with geographicdistances are further explored through the use of Mantel tests I demonstrate thatthere is little synchrony in marine survival, and that similar results also arise fromthe analysis of coherence of the longer time-series of PFA Several potential

interpretations are provided as well as their consequences for the study of thefactors controlling survival

In the second chapter, I investigate the nature of the link between marine growthand marine recruitment for a monitored Irish population over four decades Scalesfrom 1SW Burrishoole salmon were used to create indices of marine growth thatwere compared to rates of marine survival and return to the coast A strong linkbetween marine growth and marine survival is demonstrated and its importance tounderstand the mechanisms controlling recruitment is discussed

Finally, in the third chapter, I produce semi-parametric models of marine survivalfor Hatchery and Wild Irish salmon, based on the selection of both linear and non-linear predictors, using Generalized Additive Models (GAM) The most relevantand parsimonious GAM explain a large amount of the variability in survival of IrishHatchery and Wild populations These models provide an insight into the role ofparticular predictors, and their relevance is interpreted More specifically, the

variability in survival seems to result from the synergetic effects of climate change

on the pelagic ecosystem of the northeast Atlantic

CHAPTER 2

IS THERE COHERENCE IN THE MARINE RECRUITMENT OF NORTHEAST

ATLANTIC SALMON, SALMO SALAR L.? IMPLICATIONS FOR THE

SPATIAL SCALES OF THE PROCESSES CONTROLLING SURVIVAL

AbstractThe levels of inter-annual synchrony in marine recruitment of 18 Europeanpopulations of Atlantic salmon were examined, in order to determine the spatialscales of the processes regulating marine survival A series of hierarchicalhypotheses were formulated and tested to examine the patterns of synchrony acrossthe northeast Atlantic Time series of marine survival and pre-fishery abundanceboth indicated that, despite common declining trends, inter-annual synchronybetween populations is poor This result would suggest that processes operating atthe scale of the northeast Atlantic may have control over the survival rates, but theireffect on inter-annual variation is suppressed or modulated by other processes Norwas evidence found supporting the hypothesis of control at the coastal level.Instead, our results suggest the existence of a combination of factors that arespecific to each population and thus possibly reflecting the importance of eventstaking place before the stocks overlap at sea, and perhaps in freshwater

Keywords: Atlantic salmon; marine survival; Mantel tests, recruitment; synchrony;spatial scale

In the last thirty years, the total reported catch of salmon in the north Atlanticdecreased more than four fold, from approximately 12,000 tonnes in 1972 to 2,625tonnes in 2002 (Anon 2003) Management measures have contributed to reduce thecatches but the overall pattern points towards a decline in marine survival (Anon.2002; 2003)

Initially, the patterns of salmon recruitment were considered to be primarily driven

by density-dependent processes taking place during the juvenile freshwaterresidency (Chadwick 1985) More recent studies have revealed that the overallsalmon recruitment is in fact largely a result of the variability in survival during themarine rather than the freshwater part of the life cycle (Chadwick 1987; Bradford1995; Friedland et al 2003), because survival during the freshwater phase exhibitsless variation than the marine survival

This recent shift of interest from the freshwater to the marine environment hashighlighted the existence of important gaps in our knowledge of the ecology ofsalmon at sea From a management point of view, and in the context of dwindlingstocks, it is important to understand the processes responsible for the regulation ofsalmon populations in the marine environment This requires the understanding ofthe spatio-temporal scale of the processes controlling salmon survival, in order toestablish the role of potential predictor variables upon salmon recruitment, whichmight lead to more informed management of the exploitation

In the north Pacific, several studies have considered temporal synchrony (or

co-variation) in survival, for species of Oncorhynchus originating from different

regions, to be indicative of the spatial scale at which marine survival was regulated(Peterman et al 1998; Pyper et al 2001; Mueter et al 2002; Pyper et al 2002;2005) These studies have shown that coastal processes controlled survival for thesespecies (Peterman et al 1998; Pyper et al 2001; 2002; 2005) This contradictsprevious studies suggesting that marine survival was determined via basin-wideprocesses (Beamish and Bouillon 1993; Noakes et al 1998; Beamish et al 1999).Intuitively the amplitude of the pattern of decline in the recruitment of Atlanticsalmon (i.e the fact that it affects all the populations across the salmon range)suggests that the recruitment is likely to be controlled by events taking place in themarine environment, where and when the stocks can be found together Forinstance, the survival of stocks from Scotland and Norway has previously beenlinked to the extent of favourable thermal habitat in the North Sea (Friedland et al.

1998, 2000) The level of synchrony between populations could help us toconceptualise and test further hypotheses regarding the process controlling survival

at sea Synchrony in inter-annual recruitment of northeast Atlantic salmon has beenreported for two sets of geographically close pairs of populations, The North Esk(Scotland) and the Figgjo (Norway) (Friedland et al 1998), and the Teno and theNäätämöjoki (both from Finland) (Niemelä et al 2004) Comparisons betweenpopulations from the northeast Atlantic have also been conducted (Crozier et al.2003; Potter et al 2004), however, the implications for the spatial scales of theregulating process were not explored

To examine whether salmon populations display coherence in recruitment

variability on an inter-annual basis, the levels of synchrony in recruitment for

distant populations have been quantified This analysis uses information from 18European populations to test hypotheses regarding the spatial scales of the

processes controlling salmon recruitment at sea The conclusions from this studyprovide new information on selecting the appropriate scale on which to identify themechanisms that control survival

Methods

To investigate the levels of synchrony in marine recruitment of salmon populationsacross the northeast Atlantic, two sets of information on stock’s specific marinesurvival and regional levels of pre-fishery abundance were examined

Marine survival time-series

We first used time-series of marine survival from 13 rivers, from six Europeancountries, to provide the best possible geographical coverage of the region and ofthe likely regulating processes experienced by salmon populations across thenortheast Atlantic (Fig 2.1; Table 2.1) Information from stocks from the Baltic wasnot included in this analysis, as these populations have a marine distribution largelyrestricted to the Baltic itself (McKinnell and Karlstrom 1999; Kallio-Nyberg et al.2004) In some instances, several time-series of marine survival data were availablefor a single river This occurred when both wild and hatchery stocks weremonitored, and when several hatchery stocks were used (e.g Corrib) These time-series were initially standardized (zero mean and unit variance) and aggregated tocreate stock specific single survival time-series The combined catches for these sixcountries constitute approximately 80% of the total catch in the northeast Atlantic

(Anon 2004) There is little information on the marine distribution of Europeanpostsmolts Captures in high sea fisheries have shown that MSW populations fromsouthern Europe can be found west of Greenland, whereas northern European fishtend to be found mainly in the Norwegian Sea, close to the Faroes, along with 1SWfrom southern countries (Holm et al 2000) Recent surveys at sea have shown thatsouthern populations can be found along a narrow corridor between the Faroes andShetlands, few weeks after the smolt migration (Shelton et al 1997; Holm et al.2000) The distribution of the postsmolts in the Faroe-Shetland channel matches theboundaries defined by the slope current, the main current feature in this area Thedistribution of the postsmolts becomes more scattered as they leave the Faroe-Shetland channel and travel in the northeast direction along the main circulationpattern, and can eventually be found in most of the Norwegian Sea (Holm et al.2000) Information on age structure of the catches made north of the Faroes havealso revealed that southern populations, including fish from Ireland, are found there

in the late autumn and early winter Northern populations seem to be present in anarea slightly further north, and later in the winter, so that there is probably littleoverlapping with fish from southern Europe (Jacobsen et al 2000) The time-series

of marine survival for the monitored European stocks were established throughtagging and/or trapping programs and represent values of survival between thesmolt stage and returning one sea-winter fish (1SW) exclusively; i.e fish that havespent one winter at sea These indices do not include survival values for multi-seawinter (MSW) fish The rational behind the focus on 1SW salmon is based on thegeneral belief that the overall marine survival is determined during the first year in

the ocean (Doubleday et al 1979; Scarnecchia et al.1989; Potter and Crozier 1999;Friedland et al 2003) As a consequence of differences in stocks composition,maturity and migration schedules, the ICES working group on north Atlanticsalmon (Anonymous 2002) has separated European salmon stocks into two sub-groups The northern European group includes stocks from Iceland, Norway,Russia, and Sweden, whereas the southern group is composed of stocks fromEngland, France, Ireland, Scotland, and Spain In this study, we remain consistentwith this nomenclature but we also refer to the stocks from Norway and Sweden asthe “Scandinavian stocks” For the purpose of the marine survival analysis, thestocks from the river Bush (Northern Ireland) were included in the Irish sub-group.Whereas wild and hatchery fish can demonstrate difference in marine survival, theuse of a single index for each river allows an exploration of the spatial dimension ofthe variation in survival Each time-series was tested for normality and for theoccurrence of autocorrelation.

Cross-correlation, as a measure of the levels of synchrony between and amongstocks

To quantify the level of synchrony in marine survival between stocks,

cross-correlation coefficients (Box et al 1994; Ranta et al 1995; 1997; 1998; Bjornstad et

al 1999a; Botsford and Lawrence 2002) were computed on the standardized, andaggregated indices The cross-correlation coefficients measure the co-variabilitybetween two times series The cross-correlation coefficients were calculated as:(1) ρ = cov (i, j) / δi δj

wherecov (i, j) is the covariance of the populations i and j for a given year, and δi δj

is the product of the standard deviations of the populations i and j The mean

synchrony, or the average value of the cross-correlation coefficients (Bjornstad et

al 1999a), was calculated for a range of stock associations, within and betweenregions As the cross-correlation coefficients are not independent, Bootstrap

procedures were used to produce confidence intervals for the average synchronies(Sokal and Rohlf 1995; Manly 1997)

Autocorrelation

Traditional inference tests are based on the assumption of serial independence.However, biological time-series often contain significant amounts of temporal auto-correlation (Thompson and Page 1989; Legendre 1993; Bence 1995; Pyper andPeterman 1998; Lichstein et al 2002) To account for this, it is possible to eitherremove the auto-correlation by applying pre-whitening or first-differencingtechniques (Thompson and Page 1989; Quinn and Niebauer 1995; Pyper andPeterman 1998; Beaugrand and Reid 2003), or to compute new numbers of degrees

of freedom before conducting the tests of association (Garrett and Toulany 1981;Pyper and Peterman 1998; Friedland et al 2003) The latter method is sometimesrecommended because pre-whitening and first-differencing can remove much of thesignal under scrutiny, particularly where low frequency processes are an importantsource of co-variation between recruitment series (Pyper and Peterman 1998) Inthis study, the time series of marine survival with significant levels of auto-correlation were initially first-differenced (Thompson and Page 1989) beforecomputing cross-correlation coefficients between time-series To test if this

procedure produced correct Type-I error rates without removing importantsynchronous signal, cross-correlation coefficients were also computed using the realnumber of degrees of freedom, according to the level of autocorrelation present(Garrett and Toulany 1981) The difference between the levels of mean cross-correlation resulting from each approach was tested using t-tests.

Spatial analysis

The spatial scales at which the drivers of marine recruitment act were explored byperforming a spatial analysis on the results of inter-stock synchrony in marinesurvival Our analytical framework was designed to provide tests of the relativesupport and interaction of various scales of spatial synchrony This framework can

be best described as a hierarchy of hypotheses:

1) Marine survival is regulated at a large oceanic scale (i.e at the scale of thenortheast Atlantic basin)

2) Marine survival is a function of intermediate meso-scale factors (i.e.regional processes operating in the open ocean at the scale of 100 to 500km)

3) Marine survival is a function of coastal factors

Statistical analysis

A series of Mantel tests were used to compare the levels of synchrony in marinesurvival with distance among stocks originating from different regions (Mantel1967; Legendre and Fortin 1989; Leduc et al 1992; Fortin and Gurevitch 1993;Sokal and Rohlf 1995; Bjornstad et al 1999a; Koenig 1999) The Mantel proceduretests the correlation between two distance matrices The Null hypothesis is that the

two matrices are not associated As the values in a matrix are not independent, avalue of significance for the test is produced using a permutation procedure where anew distribution of the statistic is produced by reshuffling one of the matrices manytimes and the sample statistic is then compared to the new distribution (Bjornstad et

al 1999a; Koenig 1999) A hierarchical approach was used to define the structure

of synchrony in marine survival across Europe, from the broadest (northeast

Atlantic) to the finest (coastal) scale Within each of the three stated hypotheses anumber of models were tested representing different spatial scales and patterns ofsynchrony in marine survival (Table 2.2) The shortest distance between two rivermouths (great circle distances) obtained from their respective latitude and

longitude) were used as a measure of spatial distance (Table 2.3)

• To test the first hypothesis (i.e regulation of salmon marine survival by broadscale processes), a Mantel test of synchrony versus distance was used for the entirenortheast Atlantic (all stocks, test 1 in Table 2.2), the southern (test 2) and thenorthern (test 3) European regions separately If processes regulating marine

survival take place at these scales, there should be high synchrony between all thepopulations and it would be expected that the tests would not demonstrate

significant changes in synchrony with distance, because the various stocks shouldall be subject to similar conditions

• Mantel tests were carried out to test the second hypothesis and to explore thestructure of marine survival synchrony among European regions (Table 2.2, test 4-10) These tests were conducted by comparing the pair-wise matrix of cross-

correlations of marine survival with a series of model matrices each representing a

specific hypothesis regarding a pattern of similarities between stocks, or synchrony.Matrices containing hypothetical patterns of synchrony between regions werecreated using numerical weight coefficients The degree of association between thetwo matrices indicates the degree of fit of the model (Fortin and Gurevitch 1993,Legendre and Legendre 1998) Tests 4 and 5 test hypotheses of super-regionalpatterns of synchrony, whereas tests 6 to 10 explore the presence of regional

patterns (Table 2.2)

• Finally, to test the hypothesis (i.e regulation at a fine/coastal scale), a Mantel test(test 11, Table 2.2) of synchrony versus distance was conducted with the stocksfrom the Irish region because it provided information on several populations at aregional level If marine survival regulation occurs at a fine/coastal scale, synchronywould be expected to decrease with distance because neighbouring stocks are morelikely to experience similar environment than distant stocks

Time-series of Pre-Fishery Abundance

In addition to the survival information, we also analysed time series of maturing1SW pre-fishery abundance (PFA) to detect synchrony in marine recruitment acrossregions The PFA indices are produced by run-reconstruction models (Potter et al.1998) and provide an estimation of the number of fish available before exploitationtakes place The PFA is calculated from the number of fish caught, and estimations

of non-reported catches and natural mortality Whereas by definition, the indices ofmarine survival offer an accurate description of the variation in marine recruitment,the PFA indices are not restricted to this part of the life cycle and also includevariability in catches, which can be linked to freshwater production However,

because the variability in the marine recruitment is considerably more importantthan the variability in freshwater recruitment (Chaput et al 1998), the PFA indiceshave the potential to depict the signal of synchrony in marine recruitment acrossregions These indices are available for a longer period, and thus are likely tocontain more contrast than the indices of marine survival However, PFA valuesbefore 1980 must be considered with caution because of the important level ofestimation involved (Crozier et al 2003) In the northeast Atlantic, the PFA indicesprovide information on recruitment since the start of the 1970’s The same analysis,previously applied on marine survival indices, was used on PFA information.However, because PFA indices are produced for regional groupings rather thanindividual rivers, they do not allow the use of spatial statistics methods involvingpoint distances (e.g Mantel tests).

This framework, considered in the context of the overall levels of synchrony acrossthe northeast Atlantic, allows the evaluation of the relative importance of global(tests 1-3), super-regional (tests 4-5), regional (tests 6-10), and local processes (test11), allowing us an explicit analysis of the relative importance of different scalesand identifying possible multiple-scales of synchrony

Results

Marine Survival

The raw time series used in this analysis generally started after 1983 (Table 1; range10-20 years, mean 14.88 years) Salmon marine survival was highly variable fromyear to year but also from one river to another (Fig 2.2) The average marine

survival over the period of study ranged from 0.98 ± 0.72 percent (Midfjardarawild, Iceland) to 27.24 ± 7.38 percent (Bush wild, Northern Ireland) The two riversfrom the north coast of Iceland (Midfjardara and Vesturdalsa) had the lowest marinesurvival, followed by the Drammen (south-east Norway) and then the hatcherystocks from the Corrib and the Screebe rivers (both in Ireland) The survival ofhatchery stocks was significantly lower than the survival of wild stocks (t test;

p ≤ 0.01), with the exception of the wild stocks from the northern coast of Iceland.The highest mean survival was recorded for the wild stock from the river Bush, thetwo stocks (wild and hatchery) from the Burrishoole (Ireland) and the wild stocksfrom the North Esk (Scotland) and the Corrib (Fig 2.2)

Synchrony in Marine Survival between and among regions

The results of the cross-correlation analysis are reported in Table 2.4, in the form of

a pairwise cross-correlation matrix Among 78 pairwise cross-correlations carriedout, 59 (75.6%) were positive and seven were found to be significant (p ≤ 0.05).The overall synchrony in marine survival across Europe was approximately 20%(cross-correlation coefficient r = 0.195, [0.134; 0.249]) The highest synchronybetween stocks was observed between the Shannon (Ireland) and the Nivelle(France), r = 0.724 and p ≤ 0.05 The synchrony between wild stocks (r = 0.199)was inferior to the synchrony between hatchery stocks (r = 0.327) and also to thesynchrony between wild and hatchery populations (r = 0.207)

The patterns of mean synchrony for each region, and between regions are reported

in Figure 2.3 At the regional level, the lowest synchrony was observed for northernEurope (Scandinavia and Iceland), r = 0.082 (-0119; 0.238) The highest synchrony

was recorded for the Irish region, r = 0.323 (0.253; 0.397) The inter-stockssynchrony for the southern region was low but mostly positive Cross-correlationbetween regions was generally poor, as illustrated by the synchrony betweenIceland and Scandinavia The survival of Icelandic stocks was poorly correlatedwith the survival of the other European regions (Fig 2.3) Synchrony levelsobtained after correcting the number of degrees of freedom were similar to thelevels obtained after first-differencing.

Synchrony in Pre-Fishery Abundance

The level of synchrony in PFA across European regions was poor The overallsynchrony in PFA for all regions (northern and southern) was less than 22%.Similar results were found for the synchrony in the northern region (r < 0.120) andthe synchrony between the northern and southern region (r < 0.140) The synchronyinside the southern region was more pronounced This is also the only region forwhich the method of first-differencing was found to significantly (t-test, p = 0.031)underestimate the level of cross-correlation (r1 = 0.243 after first-differencing and

r2 = 0.509 after correcting the number of degrees of freedom)

Results of the Mantel tests for the northeast Atlantic

The plot of synchrony versus distance (Fig 2.4) indicates that synchrony at shortdistance is low but remains largely positive, and that negative synchrony tends tooccur at distances greater than 1000 km The results of the various Mantel tests arereported in Table 2.5 The Mantel test for all European stocks (test 1) produced aMantel coefficient r = -0.237 (p = 0.069) The negative coefficient indicates that,

overall, there was a negative relationship between synchrony and distance for theEuropean stocks The Mantel tests involving the southern (test 2; r = -0.186,

p = 0.118) or the northern (test 3) European regions were not significant, indicatingthat there was no relationship between synchrony and distance at these regionallevels

Results of the Mantel tests at the regional and coastal levels

Test 4 verified the absence of synchrony in marine survival between the Icelandicand European regions, already observed via the cross-correlation coefficients Thispattern was further explored by test 5 which revealed that this lack of synchronycould be linked to the lack of synchrony between northern Iceland and the Europeanstocks Tests 9 and 10 confirmed that marine survival for south Iceland was moresynchronous with marine survival for Scandinavia and Europe respectively, ratherthan survival with northern Iceland The test for the Irish region only (test 11),produced a positive test coefficient (r = 0.327; p = 0.055), therefore indicating thatsynchrony increased with distance between stocks

Taken together the results from the levels of cross-correlation and the Mantel testspresent a wide-ranging analysis of the scales and patterns of the synchrony ofsalmon stocks in the north Atlantic Firstly, there is no evidence of global basin-scale drivers of synchrony (tests 1, 2 and 3) The Mantel test for the hypothesis ofregulation at the scale of the northeast Atlantic shows a decrease of synchrony withdistance, which would indicate that regulation occurs at a smaller spatial scale TheMantel tests at the scale of the southern and northern European region are notsignificant, potentially indicating regulation at these specific scales, but they occur

in the context of very poor overall of synchrony between these stocks, which

suggests that the regulating factors are not acting at these scales Secondly, there issome evidence of a difference in the dynamics of salmon marine survival betweennorthern Iceland and the rest of the northeast Atlantic, including southern Iceland.Thirdly, there is no evidence of regional differences in the dynamics of marinesurvival between northern (without Iceland) and southern European stocks Finally,there is evidence against coastal drivers of synchrony

DiscussionHigh levels of synchrony between distant populations are often interpreted asresulting from those populations experiencing a correlated environment (Ranta et al.1998; Ripa 2000); this is usually referred to as the Moran effect (Royama 1992).Moran (1953) argued that two populations with the same density-dependent

structure could fluctuate in synchrony under the effects of correlated

density-independent factors such as climate conditions (Post and Forchhammer 2002) Thiscould be particularly relevant for a species like Atlantic salmon, for which there arestrong indications of density-independent mortality (Jonsson and Jonsson 2004).Inaddition, synchrony can also be the result of dispersal or migration (Hanski andWoiwod 1993; Holmes et al 1994; Molafsky 1994; Ranta et al 1995; 1998;

Bascompte and Sole 1998) or predation by nomadic predators (Ims and Steen 1990;Neubert et al 1995; Gurney et al 1998; de Roos et al 1998) Synchrony in marinesurvival values has been used to infer the patterns of temporal and spatial

distribution of salmon populations in the north Pacific and to identify the relevant

environmental variables influencing the overall recruitment of the stocks (Peterman

et al 1998; Pyper et al 2001, 2002, 2005; Mueter et al 2002)

Salmon undertake extensive migration, during which they are exposed to changingenvironmental conditions Populations originating from distant rivers can

geographically overlap in the sea and therefore can be regulated by conditionsoccurring at fine spatial scales Thus, it is not unreasonable to suggest that a processoccurring at fine spatial scale (e.g 100 km) might have consequences detectable onpopulations originating from very distant areas Therefore, and in contrast to otherspecies, high synchrony detected across distant populations of Atlantic salmon isnot necessarily an indication of large-scale processes regulating these populations.Salmon populations are also reproductively isolated (Saunders 1967, Stabell 1984,Saglio 1994, Youngson et al 2003) Straying rates are negligible and wild

populations return with great accuracy to their river of origin to spawn (Jonsson et

al 2003) Thus, despite the extent of the migration undertaken, the population’sdispersal does not result in increased contributions of individuals to non-natalpopulations Here, dispersal becomes a potential contributor to population

synchrony because populations are able to congregate and thus be exposed tosimilar environmental conditions It is suggested therefore that, dispersal acts as apromoting agent for the Moran effect

The pronounced declines of catches and marine survival in the northeast Atlantic,suggest that the levels of mortality experienced by the various populations couldhave a common origin (i.e single process acting at a very broad scale) In trying toexplain the recent and unprecedented low salmon returns, one has to account for the

global scale of this decline The similarities in the declining trajectories over the lasttwo or three decades have suggested the role of broad scale rather than more

localized or stock specific processes One of the main objectives of this study was

to test the validity of this assumption of coherence in decline

Testing of the hypothesis of stock regulation at the basin scale

The results of this study show that the marine survival of European stocks is poorlycorrelated among European regions on an inter-annual basis The synchrony acrossall stocks examined was less than 20% These levels correspond to the lower range

of the patterns of synchrony in marine survival observed for Pacific salmon Forexample, the synchrony for salmon stocks, from 14 regions of the northeast Pacific,has been found to range between 26% and 71% (Pyper et al 2001) McKinnell andKarlstrom (1999) have also reported higher levels of synchrony in recapture rates oftagged salmon populations in the Baltic They found a mean correlation of r = 0.43between recapture rates of tagged salmon from the East coast of Sweden and thenorth coast of Finland The level of synchrony reported in this study also

corresponds to the proportion of significant correlations previously found for

European populations (Crozier et al 2003) There was a concern that the use ofhatchery fish, for the purpose of comparing survival patterns, might weaken thesynchronous signal due to the differences in ecology between wild and hatcheryfish However, the fact that the synchrony between wild stocks was not superior toboth the synchrony between hatchery stocks, and the synchrony between hatcheryand wild populations, seems to support the results obtained with these time series

The cross-correlation analysis performed on the PFA indices supports the resultsfrom the marine survival analysis The overall synchrony in PFA indices is poor andsimilar to the levels of synchrony in marine survival The longer time series of PFA(1971 to 2003) were used to overcome the potential shortcomings in the time series

of marine survival (i.e lack of contrast due to a shorter coverage, 1980-1999) Theonly noticeable incidence of a significant positive synchrony in PFA was recordedfor the southern region A closer analysis reveals that the pattern of synchrony inPFA in southern Europe is largely reflecting the level of synchrony between

populations that are geographically close (i.e Northern Ireland, England, Scotlandand Wales) This is consistent with the pattern of common captures for these stocks

in the northeast Atlantic (Holm et al 2000, 2003; Jacobsen et al 2001) As PFAindices are a reflection of the general productivity of a stock over its entire lifecycle, it is not possible to exclude that the synchrony in PFA indices for thesepopulations (UK region) is also a reflection of the patterns of freshwater rather thanmarine recruitment This would corroborate with the absence of coherent signal inmarine survival for this region

If the stocks were regulated by events occurring at the scale of the northeast

Atlantic, there should be more pronounced evidence of important synchrony inmarine survival and PFA between populations from different regions, let alone thesame region The low levels of synchrony in marine survival and PFA indicate thatsurvival and recruitment are unlikely to be directly controlled by a unique large-scale process under a Moran effect scenario Low synchrony could indicate thatseveral processes taking place in different areas and/or at different times control the