Dynamics of the Atlantic salmon (Salmosalar L. ) population of the River Foyle, Ireland.
by
Patrick Boylan
A thesissubndtted for the degreeof Doctor of Philosophy.
Division of Environmental and Evolutionary Biology
Institute of Biomedical and Life Sciences
University of Glasgow
March 2004
OPatrick Boylan, 2004 Acknowledgments
I wish to acknowledgethe help and support of the Chief Executive, Mr. Derick
Anderson,the staff and the Board of the LoughsAgency who providedthe majority funding for this study. I also wish to recognisethe financial supportof the Marine
Institute and in particularDr. Niall O'Maoileidigh. Thanksare also due to the rest of
Agency's Scientific Advisors Dr. Walter Crozier, Dr. Philip McGinnity and Dr.
GershamKennedy for their manyhelpful comments. I wish also to acknowledgethe help and supportI receivedfrom manypeople at GlasgowUniversity and in particular my supervisorDr. Colin Adams. Thanks are also due to Prof. Felicity Huntingford
Midam Geurts,Sanne de Boer andthe staff at the University Field Station. Contents Page
Page no.
Summary i
Chapter 1. General Introduction 1
1.1 Atlantic salmonLife-cycle 1
1.2 ConservationStatus of Atlantic salmon 3
1.3 PhysicalDescription of Foyle Catchment 4
1.4 SalmonManagement in the Foyle System 4
1.5 PopulationRegulation 7
1.5.1 Density-DependentFactors 7
1.5.2 Density-IndependentFactors 8
1.6 PopulationRegulation in Salmonids 9
1.6.1 PopulationRegulation Processes in the JuvenileFreshwater 9 Phase
1.6.2 PopulationRegulation Processes in the Marine Phase 11
1.6.3 Population Regulation Processesin the Adult freshwater 13 Phase
1.6.4 MathematicalModels 14
1.7 Modelling of JuvenileSalmonids 17
Chapter 2. Competing modes of exploitation and their effects on 21 spawning successin Atlantic salmon (Salmo salar L. ) in the Foyle catchment, Ireland
2.1 Introduction 21
2.2 Materialsand Methods 23
2.2.1 StudyArea 23 2.2.2 Exploitation Catch Data 24
2.2.3 Salmon Population Size Estimates 26
2.2.4 Analysis 27
2.3 Results 27
2.4 Discussion 37
Chapter 3. Life-stage specific, stochastic environmental effects, 41 overlay density-dependent filial cohort strength effects in an Atlantic salmon (Salmo salar L. ) population from Ireland
3.1 Introduction 41
3.2 Materials and Methods 43
3.2.1 Study Area 43
3.2.2 Salmon Population Estimation 43
3.2.3 Population Structure 46
3.2.4 Data Analysis 47
3.3 Results 54
3.3.1 Density-Dependent Regulation 54
3.3.2 Life-Stage Specific Environmental Effects on Population 57 Size
3.4 Discussion 67
Chapter 4. The influence of broad scale climatic phenomena on long 72 term trends in Atlantic salmon population size: an example from the River Foyle, Ireland
4.1 Introduction 72
4.2 Materials and Methods 75
4.3 Results 77
4.4 Discussion 81 Chapter 5. Local instream and catchment spatial scale habitat 83 characteristics determine 0+ fry density of Atlantic salmon (Salmo salar L. ) in the River Foyle
5.1 Introduction 83
5.2 Materialsand Methods 84
5.2.1 CatchmentScale Characteristics 86
5.2.2 Analysis 89
5.3 Results 90
5.4 Discussion 94
Chapter 6. General Discussion 98
References 104 List of Figures
Figure no. Legend Page no.
1.1 Life-cycle of the Atlantic salmon 1
1.2 Location map of the Foyle catchment 4
1.3 Five year running averageof commercial salmon catches 6 in the Foyle 1993-2001
1.4 Hypothetical stock-recruitment model 14
1.5 Three common stock-recruitment curves: 0--Linear; 16 2=Beverton & Holt (1957); 3=Ricker (1954)
2.1 Drift, Draft nets,uncorrected Sport Angling catches& 29 uncorrectedredd counts in the Foyle area1952-2000
2.2 Regressionof total annualcatch of salmonin the Foyle 35 areaand corrected redd counts
2.3 Commercialcatch residuals regressed on correctedangling 37 catch
2.4 Total catchresiduals regressed on correctedredd counts 38
2.5 Total catch residualsregressed on correctedredd counts 38 excludingtwo outlying points
3.1 Salmonlife history stagesfor onelife-cycle 49
3.2a A linear model fitted to parental population size (egg 55 equivalent *100,000) and filial population size (egg equivalent*100,000)
3.2b A Beverton& Holt model fitted to parentalpopulation size 56 (egg equivalent *100,000) and filial population size (egg equivalent*100,000)
3.2c A Ricker model fitted to parental population size (egg 56 equivalent *100,000) and filial population size (egg equivalent*100,000)
3.3 Regressionsof environmentalvariables during life-stage 58 specific eventsand populationsize residualsderived from the Ricker parental-offspringcurves 3.4 Mean residual population size values derived from the 64 Ricker parental-offspring curves for years with high (upper 30 percentile); low (lower 30 percentile) and intermediate (remaining 40 percentile) values for environmental variables during the life-stage specific periods.
4.1 Annual commercial catch of salmon in the Foyle area 77 1875-2000
4.2 The five year averagecommercial catches of migrant 78 Atlantic salmonfrom the Foyle catchment1875 to 2001 & the five yearaverage NAOI: 1875- 2000
4.3 The relationship between the winter NAO and commercial 79 fishery catches of Atlantic salmon returning to the River Foyle over 122 year. Breakpoint analysis shows a uncoupling of a negative relationship about a NAOI of 0.151
4.4 The change in North Atlantic Oscillation predicted by 80 sevenclimate changemodels. Dotted line showsthe mean of the sevenmodels, solid lines showthe 2 extremesof the predictedNAO change
6.1 Foyle area annual commercial salmon catches 1962-2003 102 with regression line List of Tables
Table no. Legend Page no.
Stock-recruitmentequations for Beverton & Holt (1957) 16 andRicker (1954)models
2.1 Correlations of commercial fishing catches and corrected 32 sport angling catches
2.2 Correlationsof commercialfishing catches,corrected sport 34 angling catches, corrected total catch, corrected redd countsand corrected total population
3.1 Comparisonof predictive linear; Beverton & Holt and 54 Ricker models
3.2 Regressionof environmentalvariables during life-stage 60 specific eventsand populationsize residualsderived from the Ricker parental-offspringcurves
3.3 ANOVA results of environmentalvariables during life- 62 stagespecific eventsand populationsize residualsderived from the Ricker parental-offspringcurves
5.1 Local scale site-specific variables collected for each 85 samplingfor 350 sitesin the Foyle catchment
5.2 Catchmentscale characteristicscalculated from ordnance 86 survey and geological survey maps for each of the 350 samplingsites in the Foyle catchment
5.3 Stepwise regression of local site specific stream 90 characteristics on 0+ salmon numbers for a) all sites (including sites with no 0+ salmon); b) only sites where 0+ salmon were present
5.4 Stepwise regressionof catchmentcharacteristics on 0+ 91 salmonnumbers for a) all sites (including siteswith no 0+ salmon);b) only siteswhere 0+ salmonwere present
5.5 Stepwise regression of local scale and catchment 92 characteristicscombined on 0+ salmonnumbers for a) all sites (including sites with no 0+ salmon); b) only sites where0+ salmonwere present Summary
This study examines the dynamics of the Atlantic salmon (Salmo salar L. ) population of the Foyle catchment in Ireland, through the analysis of long-term datasets and extensive field data. In Chapter I the current conservation status of the salmon is discussed with particular reference to the Foyle. A number of methods used in studying the juvenile and adult life-stages are reviewed.
Chapter 2 considersthe interaction betweencommercial netting and recreationalsport angling and the effect of total combined exploitation on an independentmeasure of populationsize (redd counts)using a 49 year dataset. While recognisingthat commercial netting had a relatively small negativeimpact on recreationalsport angling, the evidence suggeststhat year class strength is the principal modulator of variation in commercial catches, sport angling catchesand spawning escapement.
Chapter 3 examines the role of density-dependencein the Foyle salmon population. A
Ricker density-dependent model showed that spawning adult population size significantly predicted variation in the resultant filial generation, however, a significant amount of variation (ca. 68%) remained unexplained. Environmental factors were significant in explaining some of the remaining variance and these influences were linked to specific life stages. This finding strongly suggestspopulation bottlenecks in the complex life cycle of this species, during which specific environmental factors may have an impact, that they do not have during other periods. It was concluded that these life stage specific environmental effects are likely to contribute to the stochastic variation in population size resulting from the application of traditional stock-recruitmentmodels and that the identification and quantification of these effects should allow improved model accuracy.
Chapter4 investigatesthe effect of marineclimatic conditionsin the North Atlantic on the abundanceof returningmigrant Atlantic salmon,using a 126 year datasetof commercial catchesand an index of climate in the northernhemisphere, the North Atlantic Oscillation
(NAOI). The NAOI when below 0.151 explaineda significant proportionof variation in five year running meancatches of migrant Atlantic salmonreturning to the River Foyle.
This indicatesthat a significant proportion of the variancein populationsize in the past was the result of variability in conditionsin the marineenvironment. However, when the
NAOI was above0.151, this relationshipuncoupled. Models of climate changeindicate that the NAOI is likely to increasesignificantly with time. If thesemodels are correct,this study would lead to the conclusionthat a decouplingof broad scaleclimate effects on salmon population size will become the norm. Data presentedhere suggest two consequencesof this: that the value of the NAOI as a predictivetool for forecastingadult salmonpopulation size will be limited; and that the medianpopulation size will become lower in the future.
Chapter 5 tests the capacity of local instream and broadscale catchment characteristics to predict 0+ salmon abundance within the Foyle area. Using a combination of local site specific data, from 350 sites chosen for annual electrofishing surveys of 0+ salmonids, and semi-quantitative morphometric information and broadscale catchment characteristics derived from Ordnance and Geological survey maps, two models were constructed. Model
I used data for all available sites i. e. including those where 0+ salmon were absent, while model 2 used only those sites where salmon were present. Both of these models were
ii significant predictorsof juvenile abundanceusing site-specificvariables. However,both models were improved significantly by the inclusion of wider broadscalecatchment characteristics(model I r2=30%;model 2 r2=43%). When testedagainst an independent data set using paired Mests neither model differed significantly from the actual catch of salmon,although both had relatively high standarderror rates (84%±13 and 56%±15.2 respectively). It was concludedthat site-specifichabitat characteristicswere significant predictors of juvenile abundance,but with the inclusion of broadscalecatchment characteristicsthe models predictive power was greatly increased. This chapter also highlighted the potential detrimental impact of increasingurbanisation on the salmon stocksof the Foyle area.
iii Chapter 1. General Introduction
1.1 Atlantic salmon (Salmo salar L.) Life-cycle
The Atlantic salmon (Salmo salar L. ) is an anadromous species of fish whose range extends on both sides of the North Atlantic ocean. Its life cycle is outlined in Figure 1.1 and has been described by various authors (Jones, 1959; Shearer, 1992; Hansen, L. P. &
Quinn, T. P., 1998; Klemetsen et al., 2003).
Figure 1.1 Life cycle of the Atlantic salmon (Salmo salar L. )
'A ter Diet: Diet: F\ edOva
AL 7.77t'Air
Alevin
(-3 C111)
Sand Eel
A Parr (4- 1 cn 0 12-1
I Copyright Central Fisheries Board, Dublin The adults spawn in gravel areasof streamsand rivers betweenNovember and January each year. The femaleslay eggs in shallow depressions(called redds) which are then fertilised by the male fish and covered by gravel. The eggs remain buried until
February/Marchwhen they hatchand the alevinsmigrate through the gravel into the river.
The juvenile fish will remain in the river for at least one year, but more usually two in
Ireland (Crozier & Kennedy, 1997)before going to seaas smolts. A numberof marine feeding areashave been identified, namely the Faroes,Iceland and the west coast of
Greenland. The fish will remain at sea for one or more winters. Those that return to freshwater after spending one winter at sea are commonly known as I seawinter salmon or grilse, but if they spendtwo or more winters at sea they are known as multi-seawinter salmon.
Adult fish may return to their natal streamat any time of year but most rivers have well defined periods of migration. For example,in the Foyle catchment,Ireland, the system which is the subjectof this study, the River Finn has a large numberof multi-seawinter salmon returning before the end of May supportinga recreationalspring fishery. This river also has an early I sea-winterrun of salmonwhich return at the end of May and into
Juneand July, with few fish generallyentering the river after mid July. In contrast,on the
River Mourne, a neighbouringcatchment, the majority of salmonenter the river during
June,July and August with salmonreturning, albeit in smaller numbers,until December
(LoughsAgency, unpublished data).
On their return migration from the sea salmon stop feeding and will live on their body energyreserves until spawning. After spawning,some salmon known as kelts maketheir way back downstreamand return to the marineenvironment. A small proportionof these
2 may return to spawn in subsequentyears. In Scottish Rivers previous spawners contributed between5-10% of spawning stock on average,with >95% of these being female(Shearer, 1992).
1.2 Conservation Status of Atlantic Salmon
In much of its rangethroughout the north Atlantic, the Atlantic salmonhas declined,both in termsof catchesand actual population size, since the 1960'sand early 1970's(Parrish et aL, 1998;Jonsson et aL, 2003).
In someparts of its range,such as the Bay of Fundy in Canada,and the Gulf of Maine in the USA, stockshave been listed as endangered(Chase, 2003). Commercialharvesting of salmonhas been stopped in manyareas of North America as a conservationmeasure, with the exceptionof somesubsistence catches (O'Connell et al., 1992;Dempson, et al., 2001b) and in some regions sport angling has been closed, or is subject to closure during the normal sportangling season(Dempson, et al., 2001a).
In a Europeancontext, commercialfisheries for salmon have been tightly regulatedfor many years. Despitethis however,southern stocks, in contrastto stocksin more northerly latitudes, still appearto be declining (Parrish et al., 1998). In order to protect current strongstocks of salmonthe EuropeanUnion (EU), underthe Directive on the conservation of natural and semi-naturalhabitats and of wild fauna andflora (92143EEC)(Habitats
Directive 1992)(Anon., 1992a), havedesignated many rivers throughoutEurope for extra protectionas SpecialAreas of Conservation(SAC). Within the Foyle systemthe Rivers
Finn, Mourneand Derg havebeen afforded such designation.
3 1.3 Physical Description of the Foyle System
The Foyle catchmentis situatedin the north west of the island of Ireland (Figure 1.2) and is approximately 4,5OOkM2in area. The catchment encompassestwo international jurisdictions; thoseof the Republic of Ireland and Northern Ireland, which is part of the United Kingdom.
Figure 1.2 LocationMap of the Foyle catchment
N FoyleCatchment A
R. Roo 2ý1 C,
R. FWin
OWO*kw P,
Ft S" 10
05 10 20 30 4
1.4 Salmon Management in the Foyle System
The Loughs Agency is a statutorycross-border body, which aims to provide sustainable social, economic and environmental benefits through the effective conservation, management,promotion and developmentof the fisheries and marine resourcesof the
Foyle and Carlingford Areas. Ile Loughs Agency and its predessorthe Foyle Fisheries
4 Commissionhave collated a long dataset of catchfigures and populationestimates. These data indicatethat salmonabundance peaked in the late 1960's and subsequentlydeclined until the early to mid 1970's.
As a result of thesedeclines, a report was commissioned(Elson & Tuomi, 1975)which examinedthe salmonresource of the catchmentand maderecommendations as to how best managethe stock. Following this report the Foyle FisheriesCommission set up a real time managementregime, which, with somerefinements, is still in operation. Managementof the fishery is predicatedon meeting in-seasonpopulation size targetsfor the returning migrant populationabove a control site on the River Mourne at Sion Mills measuredusing a Logie resistivity 2100C fish counterin-addition to a visual count of salmoncrossing the weir outwith the countingchannels. The first of thesetargets is the 3& June,that is, a little over two weeks into the commercialfishing season,which starts on the 15a'June.
This first target is setat 2,600returning migrant salmon(or in periodsof low water on the next flood). If the targetis not achieved,depending on the size of the flood, as measuredat a local Departmentof the Environment gauging station, both commercial fishing and recreationalangling will be curtailed by either 24 or 48 hours respectively. Similarly by the 10thJuly if a target of 4,200 returning migrant salmonhas not passedSion Mills, the fishing may againbe closedfor either 24 or 48 hours. However,the fishing may only be closed for a maximumof 48 hours in any one season. If by the 24thJuly the numberof salmonabove Sion Mills exceeds8,000, an extensionof 96 hoursfishing will be grantedto commercial netsmen. Finally, if by the 2& Septemberless than 7,000 salmon have crossedthe weir at Sion Mills recreationalangling will be curtailedby 10 days.
5 This end of season target of 8,000 salmon is based on the amount of nursery habitat
available upstream and its juvenile carrying capacity. Target egg deposition levels are set
-2 -2 for three juvenile habitat grades i.e. Grade I= 10 eggs in ; Grade 2=5 eggs M and
Grade 3=2.5 This habitat eggs M-2. combination of quality and quantity assessment
equates to approximately 6,500 salmon (male and female) spawning at the end of each
seasonwithin the catchment. Twenty five percent is added to this figure to allow for sport
angling, poaching and natural mortality bringing the final target population size to 8,000.
These targets are monitored using data from fish counters and adjusted for a visual
assessmentof the numbers of fish which cross the weir without passing through the
counting channels. The counting site isjust above the tidal limit.
Since this managementregime's introduction, stocks have increased and there is currently
a five-year running average commercial catch of salmon between 30,000-35,000 (Figure
1.3).
Figure 1.3 Five year running average of commercial salmon catches in the Foyle area 1993-2001.
50000
45000
35000 C 0 30000 E 2 25000
6 20000
15000
10000 5000 N 0 1997 1998 1999 2000 2001 Year 1.5 Population Regulation
1.5.1 General
As with all animals, mechanismsoperate to limit the size of individual populations.
Alisauskasand Arnold (1994)in a studyon Americancoots (Fulica americana)found that aI inearrelationship explained a high proportionof variance(12---0.54; p<0.00 1) betweenan estimateof populationsize (numbersshot eachyear) and the numberof temporaryponds availablefor breedingon the prairiesthe preceedingsummer. Models suchas theseallow predictionsof spatialpatterns of habitatuse and give an insight into mechanismsoperating on the population,thus enabling predictionsto be made which support population and habitat management.This is particularly importantin speciesof high conservationvalue or which are exploited.
Many studieshave attempted to identify thesemechanisms for Atlantic salmonbut most of these have been limited by the use of short-termdata sets or small sample size. One generalaim of this study is to uselong-term data sets on populationsize and exploitation to attempt to identify mechanismsand principal controlling factors influencing population size in the Foyle catchment.
Population regulatory mechanisms are frequently divided into two main types: density- dependentand density-independent factors.
1.5.2 Density-Dependent Factors
Density-dependentcauses of mortality are thosewhose mortality rate changeswith density e.g. competitionfor spaceand food (Elliott, 1985,1994;Haldane 1953; 1956;Kennedy &
7 Strange,1982; Kennedy& Strange,1986a; Kennedy & Strange,1986b). Both negative and positive effectsof populationdensity on survival havebeen shown for fish. Chapman
(1986) in a review of historical catch recordsof Pacific salmon species(Oncorhynchus tshawytscha0. kisutak 0. nerka 0. keta, 0. mykiss)from the ColumbiaRiver in North
America, identified reduced aboriginal fishing pressure,around 1800, as the cause of subsequentdecreasing runs. The author postulated that this was due to excessive escapementof salmonup river and density-dependentmortality occurringin the freshwater habitat. Heavy fishing pressureafter 1850probably increasedruns initially, but led to a sharp subsequentreduction as stocks were over exploited. Ross and Almeida (1986) showedthat silver hakes(Merluccius bilinearis) were subjectto density-dependentcontrol mechanisms,with lower growth ratesat high stockdensities.
Negative density-dependencemay also be known as concurrent (Solomon, 1949), compensatory(Neave, 1953)or regulatory(Nicholson, 1957)and simply meansthat as a population increases,the probability of individual survival decreases(Elliott, 2001).
However,population increases may not alwaysact negativelyon a species. For example, amonggoldfinches (Carduelis carduelis), birds in flocks of greaterthan eight consumedup to 2.3 times more seedsper unit time than singletons,thus increasingtheir survivorship potential(GRIck, 1986).
1.5.2 Density-independent Factors
Density-independentcauses of mortality are thoseexerting a similar effect on mortality, independentof the populationdensity. In general,density-independent factors are mainly thought to be environmentalin origin, with eventssuch as drought, temperature,storms
8 etc. being someof the main factorsidentified. Among Peregrinefalcons (Falco pergrinus) in Australia, for example,breeding successwas negatively related to rainfall with the effect that the total productionof the populationwas lowest in the wettestyears, mainly as a result of direct rainfall andrunoff flooding nestsites (Olsen & Olsen,1989).
In reality thereis no sharpdelineation between density-dependent and density-independent factors. For instance,deaths arising from biological causes,such as disease,parasitism, malnutrition and predation,which may be consideredto be initially density-independent will usually becomemore frequentas densityincreases. Density-dependent factors can act togetheror can compensatefor eachother. For exampleJenkins et al. (1963), in studying the effects of shootingon Red Grouse(Lagopus lagopus scoticus), found that lossesat
30% of the post-breedingnumbers were entirely compensatedfor by reducednatural loss, causingno depletionin breedingstocks (Jenkins et al., 1963).
1.6 Population Regulation In Salmonids
1.6.1 Population Regulation Processes In the Juvenile Freshwater Phase
There is evidencethat density-dependentmortality in the juvenile stagescan be a control of population size in salmonids(Ricker, 1954; Kennedy & Strange, 1982; Kennedy &
Strange1986a; Kennedy & Strange1986b; Egglishaw & Shackley,1985; Milner, et al,
2003; Ruggerone& Rogers,2003; Dumas & Prouzet,2003). Gee et aL (1978),in a study on the River Wye in Wales,found that maximumsmolt production,which they equatedto juvenile salmonsurviving to 2 yearsof age,was attainedat a fry densityof 0.75m'2on the l' of June. Above and below this stocking density, production decreasedfollowing a dome-shapedproduction curve. Gardinerand Shackley(1991), in a study of a six-year
9 dataset of salmonfry to their first and secondautumns on the ShelliganBum in Perthshire,
Scotland,found that productionfollowed a similar domeshaped Ricker model. However, in a review of the stock-recruitmentmodels of the time, Solomon(1985) suggestedthat, while Gee et al. (1978) werejustified in claiming that the curve was indeeddome-shaped, migration could explain differencesin the densitiesof fry in subsequentgenerations. In contrast,Gardiner and Shackley(1991) suggestedon the basis of work carried out by
Egglishawand Shackley(1977) that densityrelated losses in their study were most likely lossesof youngerparr dying in the stream. Elliott (1994,1989,1985),showed in his long- term studieson two small streams(Black Brows Beck and Wilfin Beck) in England,that density-dependentmortality is not necessarilyalways the most significant population regulatingeffect in salmonidpopulations. Black Brows Beck (holding a migratorybrown trout [Salmotrutta L. ] population)always had higherjuvenile densitiesthan Wilfin Beck
(containingresident brown trout) and was regulatedby density-dependentsurvival in the early stagesof the life cycle. There was no evidencefor this, however,in Wilfin Beck, wheresimple proportionatesurvival occurredwith fairly constantloss rates. Survival was reducedin both populationsby summerdroughts and also by spatesin Wilfin Beck. Black
Brows Beck trout were alwayslarger than thoseof Wilfin Beck of similar age,mainly due to fry size at the start of the growth season.Variations in water temperaturewere chiefly responsiblefor differencesin growth betweenyear classeswithin eachpopulation. Food intake was not a limiting factor except in the first winter of the life cycle and for adults over 3 years old in Wilfin Beck. Variation in individual size was inversely density- dependentin Black Brows Beck and decreasedwith age in Wilfin Beck, thesechanges being due to natural (stabilising)selection. There was also strongevidence for genotypic differencesbetween the populations.
10 Resultsfrom studiessuch as theseindicate a natural inter-systemvariation in the relative dominanceof density-dependentand density-independentcauses of mortality in population sizecontrol. This leadsto the conclusionthat, in orderto studythe dynamicsof individual systemsfor the purposesof populationmodelling, an assumptionof a genericrelationship appliedacross systems is likely to be inadequate.
1.6.2 Population Regulation Processes in the Marine Phase
Density-dependent mortality as a result of competition between Asian Pink salmon
(Oncorhynchus gorbuscha) and Alaskan sockeye salmon (a nerka) has been shown to occur at sea (Ruggerone et aL, 2003). It is not known if density-dependent mortality occurs with Atlantic salmon in the marine phase. As major differences exist between the
Pacific and Atlantic species it is impossible to apply conclusions based on Pacific salmon studies to Atlantic salmon. For example, Pacific salmon are the dominant fish species at sea while Atlantic salmon constitutes a very small percentageof the total marine biomass.
What is known, is that mortality in Atlantic salmon, once smolts leave the river, varies dependingboth on the river location and stock component(Anon., 2003c). Rod catches from the River Spey, Scotland indicate that migrating spring stocks of multi-seawinter salmon are falling (Anon., 1998a). However, numbersof grilse and summermigrating multi-seawintersalmon remain more stable. The mechanismsbehind Atlantic salmon survival at sea are not clearly understood. However, North American studies have observedgood correlationsbetween sea survival and oceantemperatures and the projected amountof winter habitatavailable to their stocksof fish at sea(Amiro, 1998).
11 In many Canadianrivers, juvenile productionis thoughtto be optimal. Despitereductions
in the commercialfishing effort and periodic increasesin the numberof salmonreturning to spawn,however, anticipated increases in subsequentrecruitment of spawningstock have
failed to materialise(Dempson et aL, 2001; Amiro, 1998;Ritter, 1997). This trend is also
shownby the decline in survival ratesof hatchery-releasedsmolts, which are independent
of the freshwaterhabitat. Amiro (1998) suggeststhat an increasein predatornumbers,
particularly harp seals(Phoca groenlandica), may explain the deficit betweenactual and
expectedreturns. In a Europeancontext, data from the River Bush in N. Ireland indicate
that salmonsurvival at seaoutwith direct anthropogenicinfluences has decreasedfrom an
averageof 30% pre 1998 (Crozier & Kennedy, 2001) to 10% in 2002 (Crozier et aL,
2003d). These declinesare also occurring at a time when smolt production from the freshwaterphase is decreasing(Kennedy & Crozier, 1991). Changesin the productivity
of Pacific salmon specieshave also been documentedand linked to climatic conditions
(Petermanet aL, 2003) and somestudies have developedpredictive models to track these
changes(Logerwell et aL, 2003). Chapter4 investigateslarge-scale climatic effects on
marine survival and the potential impact of global warming on future trends in salmon
abundancein the Foyle system.
With the decline in salmonabundance, attention has tendedto focus on the commercial netting sector and in particular on interceptorymixed-stock fisheries, such as those off
Greenland,the Faroes,the Northumberlandcoast of England, Northern Ireland and the
Republicof Ireland. In 2002 the Greenlandfishery was allocateda total allowablecatch of between20 and 55 tonneswhilst for the long line Faroesefishery no quota was set for
2002 or the 2003 season(Anon., 2003d). In Northern Ireland, within the Fisheries
ConservancyBoard area,a buy-out schemefor the commercialnetsmen was introducedin
12 2002. The drift net fishery limits off the Republic of Ireland's coasthave beenreduced from 12 to 6 miles and fishing is only allowedbetween 6am and 6pm Mondayto Thursday
(Anon., 1997a). In 1998the drift netscaught 58% of the total run of fish back to the Irish coast (O'Maoileidigh, pers. comm.). In 2001 carcasstagging was introducedon an all
Island of Ireland basis,whereby each individual salmon,whether caught commercially or by sport angling, must carry a tag with a unique identifying number.This has greatly reducedlarge scalepoaching and with the introductionof quotas(Republic of Ireland) to both commercialfisheries and sportangling, further restrictsexploitation on stocks.
Within the Foyle system,in line with theserestrictions, the commercialfishing seasonhas been reducedfrom March to Septemberin the 1960's and 1970's to a six-week period during Juneand July at present. In addition,there are only four days' fishing allowedeach week for both the drift and draft nets. The drift netshave also had the permissablefishing hours,reduced to 12 hoursper day.
1.6.3 2 Population Regulation Processes In the Adult Freshwater Phase
On return to fresh water, the main source of mortality in adult salmon is exploitation. It is estimated that estuarine or in-river draft nets catch approx. 28% to 30% of fish entering the
system on the River Eme. This figure is regarded as a national average rate of exploitation
for catchments where draft net exploitation occurs in Ireland (O'Maoileidigh, pers. comm.). Angling exploitationcan rangefrom 9.7% to 20.6%(Anon., 1992b)for grilse and
35-40% for stocks of Irish multi-seawinter salmon (O'Maoileidigh, pers. comm.).
Although, in the Foyle system, sport angling exploitation has reached50% for grilse
(Elson & Tuomi, 1975). As a responseto high exploitation levels, the Environment
13 Agency in Englandand Wales has introducedmandatory catch and releaseof salmonby anglers before 10 June (Anon., 1999). Many rivers in Scotland,while not having legislation to enforce similar controls, actively encourageanglers to return spring fish
(Atlantic SalmonTrust 1998). In the FisheriesConservancy Board areaof jurisdiction in
N Ireland,catch and release must be practicedon salmonprior to the I" June.
In this study (Chapter2) 1 attemptto examinethe role of competingexploitation methods on populationsize and also the interactionbetween these capture methods.
1.6.4 Mathematical Models
Severalmathematical models have been proposed to describedensity-dependence (Ricker,
1954;Beverton and Holt, 1957). When thesemodels are appliedto fish populations,they
are usually termedstock-recruitment models as they describethe relationshipbetween the
numberof recruits (R) to the populationand the parentstock (S) of fish. A hypothetical
stock-recruitmentrelationship is outlinedin Figure 1.4.
Figure 1.4 Hypotheticalstock-recruitment model.
Gn
I Teýgýgs) 14 Parents These models can be used to examinehow changesin, for example,fishing mortality,
affect yield. For thesemodels to be effective it is necessaryto know the productionof
parentstock over a wide rangeof recruit numbers. The parentstock can be expressedas
numbersof adults, biomassor egg production (Elliott, 1994; Chadwick, 1985; Elson &
Tuomi 1975).
A numberof stock-recruitmentmodels are shownin Figure 1.5 and their derivedequations
outlined in Table I. I. Model 0 shows a population that is increasing without any
restriction. While in model 1, as egg density increases,a dome-shapedcurve emerges.
Thus the initial effect of increasingpopulation size is density-dependenceoperating in a
positive manner. However,as the populationreaches maximum return it startsto decline
as negative density-dependenteffects become dominant. In model 2, the relationship
reachesan asymptoteas egg densityincreases (Elliott, 1994),but thereafterremains stable.
Figure 1.5 Three common stock-recruitmentcurves: 0--Linear; 1= Ricker (1954); 2=Beverton& Holt (1957).
12-
Cý% 10- 02 r: 8-
6- 0 1. 4- E 0 2- z 0-. 0 1111 20 40 60 80 100 120 140 Number of eggs (S m-2)
15 Table 1.1 Stock-recruitmentequations for Beverton& Holt (1957)and Ricker (1954) modelswhere R representsthe recruitsand S theparental stock. Proportionate survival Mortality rate [In (SIR)] Replacement Key (RIS) abundance fisheries Equation Function Relationship Function Relationship (S* for R =;S) references with S (all with S (all decreasing increasing
monotonic) monotonic) (I)R=aS exp(- alexp(bS) Concave bS-In a Linear (In a)lb Ricker bS) upward (1954) (2)R=aS1(l+bS) Concave In Convex (a-l)lb Beverton al(l+bS) upward (I+bS)-In upward & Holt a (1957) In most studies on salmonids to date, model 1, the Ricker (1954) stock-recruitment model,
hasbeen found to be the most applicable.Elliott (1994)found that this modelgave the best
fit to his datafor sea-trout(Salmo trutta L. ). Elson and Tuomi (1975), Gee et aL, (1978),
Gardinerand Shackley(1991), Kennedyand Crozier (1993) fitted Ricker curvesto their
data on Atlantic salmon and many North American studies of Pacific salmon have utilised
this model (Ricker, 1954,1989). Buck and Hay (1984), however,found that a Beverton
and Holt curve fitted their population data on Atlantic salmon, with survivor density
increasing to an asymptote (curve 2, Figure 1.5) rather than following a dome-shaped
curve.
All of thesepopulation models incorporatea number of inherentproblems. Regression
methods for estimating the parametersof these simple stock-recruitmentequations
generallylead to upwardbias in assessmentsof productivity at low spawningstock sizes,
and underestimatesof the spawningstock that would producemaximum averagesurplus
for harvest(Hilbom, 1997;Myers, et aL, 1995;Myers, et al., 1994).Bias is due to effors-
16 in-variables effects when spawning stock measurementhas been inaccurate(Elson &
Tuomi, 1975). It can also be difficult to generatedata points for an over-abundanceof parent stock or egg deposition(Kennedy & Crozier, 1991; Smith & Walters 1981),and thereforeit may be difficult to determineif the data set takesthe form of a dome-shaped
Ricker curve or a flat-toppedBeverton and Holt curve. There is also bias in time-series effects when harvestrates have been relatively stableso that recruitmentanomalies result in subsequentchange in spawningabundance (Walters & Ludwig 1981). There are no satisfactorycorrection methodsfor thesebiases, since they involve fundamentalloss of information about averageresponse, especially at low spawningabundance (Myers 1997;
Schnute& Kronlund 1996.). Beyond such bias problems,there is also uncertaintyas to how to measurethe reliability of parameterestimates and define policy prescriptions especiallyin circumstanceswhere the recruitmentrelationship may be non-stationarydue to persistent directionalecological or environmentalchanges (Amiro, 1998;Ritter, 1997).
As a result of these uncertainties,Walters (1981) recommendedthat regular probing experimentsbe carriedout on modelsand that in generalresults from thesetypes of model be treatedwith caution.
In this study,Chapter 3 attemptsto addressthe role of density-dependencein determining adult returnsand also the potentialfor environmentalfactors in influencingthese returns, with the useof long term dataavailable for the Foyle system.
1.7 Modelling of Juvenile Salmonids
Density-dependentmortality is known to occur in juvenile salmonids(Elliott, 1994,1989,
1985;Kennedy & Crozier 1991). Ibis may be as a result of competitionfor resourcesand
17 territoriality (Heggenes& Saltveit, 1990; Titus & Mosegaard,1992; Kennedy 1983;
Kennedy & Strange 1981; Marshall, 1995). The measurementand classification of instrearnhabitat areasis thereforecritical in estimatingthe potentialcarrying capacityof streamsand thereby deriving data for stock-recruitmentrelationships. In assessing instrearnhabitat, some studies have used wetted area of the river accessibleto fish (Elson
& Tuomi, 1975)while othershave quantified the availablenursery area (Inoue et al., 1997;
FHR, 1991). In theselatter studiesattempts were madeto quantify the river area,based on its morphology,and then relate the stock (as determinedfrom electrofishingsurveys) of each habitat reach or type to give an overall estimateof productivity. However, these modelsquantify actualproductivity at any given time and makeno attemptto predict fish numbers. In dynamicriverine environments,change, whether natural or man-induced,is always occurring, with concomitant effects on fish and other animal assemblages.
Therefore,predictive modelshave been developedin an attemptto monitor and explain these impacts (Wright et al., 1989; Whitehead, 1992; Milner et al. 1998). The River
InvertebratePrediction And ClassificationSystem (RIVPACS), developedby the Institute of FreshwaterEcology, is one suchexample (Wright et al., 1989; 1996). It encompassesa computermodel built from a databaseholding biological and physicaldata collected from over 8,000near pristine samplingsites in Englandand Wales. Siteswere selectedto cover as wide a range of environmental variables as possible, with the exclusion of unrepresentativepolluted sites. Invertebrateswere collectedfrom a wide variety of habitat types at each samplingsite and samplingwas repeated over different seasons.Samples were collected,sorted and identified to provide speciespresence/absence, abundance and diversity estimatesfor each site. The Biological Monitoring Working Party (BMWP)
scoreand AverageScore Per Taxa (ASPT) were computedfor each site as a measureof
the richness and diversity of the invertebrate community. Using the TWINSPAN
18 technique(Hill, 1979),RIVPACS invertebratedata were analysedto provide a systemfor the classificationof river siteson the basisof the distributionof invertebratefauna present.
In addition to the invertebratesamples collected, a large number of physico-chemical characteristicswere estimatedor measured,such as channelwidth, gradient,flow velocity and substratetype. On the basisof (a) theseenvironmental observations, (b) the observed invertebratepopulation data and (c) the TWINSPAN classifications,a predictivetechnique basedon multiple discriminantanalysis (MDA) was developed.This is usedto predict the probability of occurrenceof individual taxa, and BMWP and ASPT scoresfrom valuesof physico-chemicalvariables alone. At a river site of interest,RIVPACS can be used to predict targetvalues of BMWP, ASPT and the expectedcomposition of invertebratefauna on the basis of a limited number of observations/measurementsof physico-chemical parameters.Comparison of observedand predicted scoresusing results from the direct samplingof invertebratesat the site and RIVPACS predictions(based on physico-chemical parametersat the site), gives a measure of any loss of biological quality due to environmentalstress.
Another model which usesa similar approachto RIVPACS is HABSCORE(Milner et aL,
1993). This is a fisheriesmanagement model and is designedspecifically for salmonids.
Direct sampling by electrofishing is conducted at a number of sites, alongside measurementsand observations of physicalhabitat features, such as:
1) > Conductivity(pS cuf
> Cover
> Flow type
> Substrate
19 > Mean width
> Mean depth
Theseare combined with datafrom maps,including:
> Gradient(m km")
> Catchmentarea (km -2)
> Altitude (m OD)
> Distanceto mouth(km)
The model is calibratedon the basisof theseobserved data and can then generateestimates of fish populations from measurementsof physical habitat variables alone. As with
RIVPACS, a comparison of model predictions with population estimates made by electrofishingmay be usedto detectanomalies in observedecological data which may be attributableto impactingfactors. The modelswere producedfrom a data set of 224 sites on II catchmentsthroughout Wales. The siteswere screenedcarefully to avoid any which were subject to identifiable environmentalimpacts. Models were developedfor four categoriesof salmonid:trout 0+, trout >0+, salmon0+ and salmon>0+. The predicted densitiesrepresent an index of site expectation. This is not to be confusedwith carrying capacitywhich may be rarely reachedin sites receivingonly naturalrecruitment, because density-independenteffects keep the populationat lower levels (Milner et aL, 1998).The likelihood that many sitesused in habitatmodelling have populationdensities below their carryingcapacity is oneof the principal constraintson the performanceof habitatmodels.
Chapter5 attemptsto build upon existing models by using site specific and broadscale catchmentcharacteristics to investigatetheir relative importancein determiningAtlantic salmonfry numberswithin the Foyle system.
20 Chapter 2. Competing modes of exploitation and their effects on spawning success in Atlantic salmon (Salmo salar L.) in the Foyle Catchment, Ireland
Z1 Introduction
Conflict arising between resource user groups exists in many differing situations around the world today (Quizilbash, 2001). With regard to animal populations, this can be an argument over resource allocation or concerns over the conservation status of the exploited population (Claytor, 2000). Many of these have in the past been pushed to the edge of extinction (blue whale (Baleanoptera musculus) (Clark, 1973)) or over it (Great Auk
(Pinguinus (Alca) impennis); PassengerPigeon (Ectopistes migratorius) (Newton, 1998)) through over-exploitation. It is the balancing of these arguments by managers which science strives to inform.
Amongst the animal species deemed to be of high conservation value and named in statutory instruments requiring protection, the Atlantic salmon (Salmo salar L. ) is unique
in the degree to which it is exploited across its range in Europe. Although designated as a
species whose conservation requires the establishment of Special Areas of Conservation
(SACs) under the European Union Habitats and Species Directive (Anon., 1992), exploitation through commercial high seas netting, coastal and estuarine netting and sport
angling is high. For example, the global harvest of this species in 1999 was ca 2,200
tonnes, the vast majority of this catch being made in North Eastern Atlantic waters (ca
2,000t) (Anon., 2001).
21 There have beengrowing efforts to reducethe impact of interceptorycommercial fishing on stocksso that individual populationswhich may be over exploitedare protected(Anon.,
1996).As well as conflict betweenexploitation and conservationthere is also scopefor conflict betweenthe multiple usersof a resource.There is a generallyaccepted belief that
commercial interceptionnetting of migrant salmon returning to the natal streams,by
trapping,gill or seinenetting in coastalwaters, estuaries and rivers, reducesthe potential
for exploitation by riverine sport angling (Johnston,2002; Anon., 2003b). Basedon the
presumptionthat the greatesteconomic value is derivedfrom sport angling,there has been
considerablepressure to close, or at least reduce, commercial fisheries to allow sport
angling to flourish (Whitehead,2003; Anon., 2003). As a result, interceptionnet and
trapping fisheries have been reducedconsiderably over the last 20 years or so (Anon.
2001). However,despite this widely held belief, there is a paucity of data which shows
significant benefits deriving for recreationalangling from the closure of a commercial
fishery (Jensenet aL 1999)and what datado exist may not provide a clear cut picture.For
example,Bowker et aL (1998) found someevidence to suggestthat diminishedestuarine
netting exploitationhad a beneficial impact on salmonangling successon the River Usk,
England, while with the closure of the marine Norwegian drift net fishery Jensenet al
(1999) concludedthat this contributedto increasedcatches of grilse and smaller 2 sea-
winter salmon in freshwater in Norway. McKinnell & Karlstrom. (1999) and
Romakkaniemi et aL (2003) found some evidence that reduction in commercial
exploitationcontributed to increasedstocks in manyBaltic rivers. A buyoutof commercial
netting stationsin Iceland by angling interestswas shownto have positive effects on the
sport angling catchesin the local rivers (Einarsson& Gudbergsson,2003). In contrast
however,Shearer (1992) found that following net closures,rod catchesdid not necessarily
improveon the AberdeenshireDee in Scotland.
22 This chapterexamines the potentialconflict betweenthe requirementsof managementfor protection of this species and exploitation (commercial net and trap fisheries and recreationalangling for sport) in one population,the Foyle system. Within this catchment area there has been long-term monitoring of catches of salmon from all forms of exploitation, in addition to the collection of an exploitation-independentmeasure of populationsize.
Specifically two hypothesesare tested:
1. Commercialnetting and trapping of Atlantic salmon in the Foyle systemsignificantly
impactsupon sportangling, by depressingangling catches.
2. Total exploitation (from all sources)has a significant impact upon the ultimate size of
the spawningpopulation.
Z2 Materials and Methods
2.2.1 Study Area
The River Foyle system drains ca 4,500km2of the northwest of the island of Ireland
(Figure 1.1), dischargingnorthwards through the Lough Foyle estuary into the northern
Atlantic Ocean.
This systemsupports a significantpopulation of Atlantic salmonfor which there is a long
history of commercialnet and trap fishing and sport angling stretchingback over several
hundred years (Foyle Fisheries Commission, 1953; Elson and Tuomi, 1975; Foyle
23 FisheriesCommission, 1996). Total catch in the Foyle areain 1999was approximately60 tonnes,representing 11 % of the total catchfor the islandof Ireland(Anon. 2000).
The LoughsAgency of the Foyle, Carlingford and Irish Lights Commissionis responsible for the protection,conservation, improvement and developmentof the salmonidfisheries within the Foyle and Carlingford catchmentareas. As part of theseresponsibilities, the agencyand its predecessorthe Foyle FisheriesCommission, have collected data on catches and populationsize on an annualbasis since 1952. Thesedata are describedbelow.
2.2.2 Exploitation catch data
Drift nets
Drift nets are gill nets deployedclose to the water surfacein the sealough and up to 12
miles out from the north coast. At presenta maximum of 112 licencescan be issued
annually. In the past,the numbersissued ranged from 39 in 1952to a peakof 139in 1962.
The nets utilised at seaare 1500in long, with a depth of 45 meshsand a meshsize of 63
mm knot-to-knot. Those used in the Lough are 900 in long but otherwise of similar
dimensions. The drift-netting seasoncurrently commenceson 15th Juneand endson 3 1"
July and nets are fished for four days each week betweenthe hours of 6 am and 6 pm.
Licenceholders have been required to makeannual returns as a condition of the grant of a
licencesince 1952.
Draft Nets
Draft netsare seinenets which am deployedin-river in the main River Foyle and the River
Roe at defined netting stations.They are currently used from 15'h Juneto 31" July, for
four dayseach week over the 24-hourperiod. Tidal movementand water conditionslimit
24 their use. On average50 licenceswere issuedannually over the last 5 years(Wysner, pers. comm.) However,this hasfluctuated between a maximumof 524 in 1959to a minimum of
43 in 1999. There is a statutoryrequirement for the usersof thesenets to make catch retums.
Stake Nets
Stake netting was a commercialfishery run by the Foyle FisheriesCommission between
1952and 1988. A barrier net stretchedat right anglesto the shorelineled to a box trap
wherethe salmonwere caught. They werefished in the lower River Foyle and the estuary.
Thesewere temporary fishing stations,which were installedannually. The Foyle Fisheries
Commissioncollected annual catch records for the durationof this fishery.
Bag Nets
Bagnets were of a similarconstruction to thestake nets but wereleft in placepermanently.
Onelicence was issued annually to be fishedin theLough Foyle estuary, although id 1988
anadditional licence was issued. Data for thesenets are available from 1964to 1990when
operationsceased. The Foyle FisheriesCommission collected annual catch records for the
full periodof their use.
Catch Effort of Commercial Net and Trap Fisheries
The effort employedin all commercialfisheries in the Foyle systemhas changed over the
years.The seasonis considerablyshorter now than previously (currently 15 June to 31
July, cf. March to Septemberin 1952)and the numberof hours where fishing is allowed
each week has been reduced. There has also been a shift in emphasisbetween the
fisheries, with the closure of some methodsof exploitation and a shift in the relative
25 importance of draft and drift net use. Over the years the actual returns from the commercialnets have beengood and would appearto tie in with field staff observations
(R. Wysner,pers. comm.). For the purposesof analysispresented here no correctionhas beenmade for changesin fishing effort as it is the absolutenumber of fish removedthat is most likely to impactupon recreationalcatches and population size.
SportAngling ExploitationReturns
It is a statutory requirement that sport angling licence holders make an annual return of
their catch in the Foyle and Carlingford areas.The returns from recreational angling are, in
general, low (Small, 1991); this is certainly also true of the Foyle system. Thus, the sport
angling catch return cannot be regarded as an accurate measure of total exploitation by
recreational anglers. Therefore, to estimate the actual sport angling catches a correction
factor was applied to the data (Small, 1991). This correction allowed for the non-reporting
of fish caught when anglers made a return and is as follows ((0.3/(percentage return of
licences/100))+0.7) this produces a raising factor which is applied to the actual number of
fish declared.
2.2.3 Salmon population size estimates
Redd Counts
Since 1952the Foyle FisheriesCommission (and subsequentlythe Loughs Agency) field
staff have made counts of redds (salmon nests) during and immediately following the
salmonspawning period in approximately260 zoneswithin the catchmentannually. This
is only a partial count of all redds within the Foyle catchment.The accuracyof counting
can be subjectto environmentalconditions such as high water flows, which can obstruct
26 proper assessmentthrough poor visibility, or flattening of reddsmaking them difficult to see. However, as a method of long-term population monitoring, it is recognisedas a useful tool (Elson & Tuomi, 1975;Hay, 1984;Isaak, et aL, 2003).Highly accuratedata on spawning population size for one year, 1999 showed that counts at these 260 sites representedI I% of the total reddsbased on the percentageof spawningareas available to fish within the Foyle catchmentcalculated from instrearnhabitat surveys (Loughs Agency, unpublisheddata). Thus to estimatethe absolutenumber of reddseach year in the Foyle catchment,a correctionfactor of 9.1 was appliedto redd countsfrom all years.Each redd was takenas representinga singlefemale salmon (Hay, 1987). In order to thereforeget an overall population size an additional 40% was addedto incorporatethe male conponent
(Wray, pers.comm. ).
2.2.4 Analysis
Catchesand redd count estimateswere assembledby calender year. The data were analysedusing the statistical software packageSPSS 10.0. Pearson'scorrelations and linear regressionswere usedto look for trends in relationshipsbetween exploitation and subsequentspawning population size. To avoid Type I statisticalerrors, the acceptable probability when using multiple tests was correctedusing a Bonferroni correction.For clarity the correctedprobability equivalentsare presentedhere where appropriate.
2.3 Results
Analysis of the 49-yeardataset showed considerable variation in the yield from the various fonns of exploitation.
27 The 49-year meancatch in the drift nets was 26,753 individual salmonbut varied from
2,347 in 1952to 65,654 in 1983(Figure 2.1). Reporteddraft net catcheswere generally lower than thoseof drift net catchesafter the mid 1970's,but the draft nets havea higher overall mean catch of 30,928. The highest reported annual catch was also by draft net,being 82,106 in 1962 while the lowest was 5,434 in 1991. The Foyle Fisheries
Commissionstake nets operatedfrom 1952to 1988,during which period they comprised on average13% of the total commercialnetting and trappingcatches. The bag nets,which were operatedbetween 1964 and 1990,averaged 1% of the total commercialnet and trap catches.
The sport angling fishery reportedyield over the 49 year run averaged1,443 with a range of 379 (1984)to 5,100(1965).
28 a) C14 Sport Angling catches (no. of salmon) / Redds 0000 0 C) 0 LO 0 LO C) N C\j 0
C) C)
rj) 0) CM Q) 1. Nd 4a) If oa CL U) ...... r,- OD
CM 00 (D
N- N- 0) i-
>-
04 t- o)
r,-
IV
CM (0
U I ozz:::- .. 4
C) N U) CY)
C\j LO CY)
CD CD 000 C:)
00 r,- (0 U) CY) EI (uowjes jo-ou) saq31eo lau UejCV4! jC1 Actual redd (nest)counts between 1952 and 2000 averaged5,383 with the lowestcount in
1999 of 640 and the highest in 1966 of 22,606. Thus the estimatedcorrected mean, minimum and maximum redd numbersfor the whole catchmentwere 81,647,9,607 and
342,858respectively.
2.3.5 Hypothesis 1: The interceptory commercial fishery for returning migrant salmon impacts upon the recreational angling fishery
If one form of exploitation were negatively impacting upon another then clearly the exploitationmethods conducted at earlier points of interceptionwill impact on successive forms of exploitation. Thus, one would predict that drift netting, being the first of the interceptionfisheries exploiting the returning migrant population,would impact on draft netting and the combination of drift and draft netting would impact upon recreational angling. If there were an impact of initial exploitation methodson later ones then one would expecta negativecorrelation between the catches.
Correlationanalysis was used to test for a negativerelationship in catchesbetween drift and draft netting in the Foyle system. Catchesfrom the drift net fishery were not correlatedwith equivalentannual returns from the draft net fishery over the period 1952to
2000 (Table2.1).
Similarly to test for an effect of all the commercialnet and trap fishery catchescombined upon the sport fishery, the annual draft and drift net returns from 1952 to 2000 were combinedwith the catchesfrom other commercialnet sources(stake nets from 1952 to
30 1988and bag nets from 1964to 1990),to producea measureof the total catchesfrom all commercialfisheries over the 49 year period. Thesedid not predict sport fishery catches (rable 2.1).
When the commercial (drift, draft, stake and bag net fishery catches)were compared individually with the sport fishery catches, no significant correlation relationships with the sportfishery returnswere found (rable 2.1).
31 (1q Ln (» (D Ln r- f2 Co N 0 C) Ln [ý C) 0) M 1- t M " . qt 1- (0 CM le CI) (0 r- Ict Ln le lq 0 Co v- LO 0 Co cm 0 CM 0 0
*t o 0 0 o- e
(7) 9 8 r- Lf) 0 N Le) 8 CD P CY) e CY r- 0 cr) (» lit > li t .u uý cý ce cý r- C) Co 0 0 (D E E 0
CX3 ciý CMl e vLO - N ei s> N 8 .2 c\I r-.:cý cm 2 ei e
C» cy r- r_ 0 to 0 * N 0) Co 0 cr) to 0 cli 0 Cf) r_ m cý Ce CY 0 rli (0 0 0 . (0 cý (D , -k '0 (D CY 0 0 0 0 0 (D .0 iu 4ma) (D 4Z90) 0 0 0 (D 0) c) 0 (D C) (D 0 0) 0 = (i Q = ci CD 0 0 0 0 0 P) cý U) U) c\I 2 CM ci 0A (0 ce cý (0 Co a z9 a) a) 6 0tu a tu(D a Q)cu 6 w0 Z! CL U5 Z CL U5 Z CL U5 Z CLu5 Z CL u5 Z CL ý5 Z Cl cm . . . rn 0 ZU E E 0 < E-4 Co EL 0 c] m Coig 0 C) F- Co , 2.3.6 Hypothesis 2: Exploitation negatively Impacts upon spawning population size To test the hypothesis that exploitation is having a significant negative impact on that population, a measure of the annual exploitation was derived by combining the catch returns from all fisheries (commercial netting, trapping and sport angling). This was used to test for a relationship with an indicator of breeding population size that was independent of exploitation; the number of redds (nests). Annual corrected redd counts were significantly but positively correlated with the total annual catch returns of all forms of commercial exploitation combined (Table 2.2). The total annual catch (including the sport fishery) also significantly predicted corrected annual redd counts in a linear regression model (FI,47=43.85; p<0.001; r2=0.47 - Figure 2.2). Looking at each exploitation method for which there are a full 49 years data, salmon population size, as determined by corrected redd counts, was also significantly and positively correlated with drift netting, draft netting, stake nets, bag nets and total catch (Table 2.2). When corrected using Bonferroni drift netting and bag netting were not significant. One possible explanation for the significant positive relationships between different exploitation types, and between exploitation and population size, is that years with relatively large returning migrant populations are swamping any negative impacts. To explore this further, the same relationships between exploitation rates with population size were examined, at high and low population levels, separately. 33 It en 10 9 cc 'o ,02 0 0 cq 0 !ý 9) "' Ui § r" q§ -M§- (), Ici 2 k§-. g§t,- E E 0 N (7) ý 9§ § CO - It . : C4 rIt ID V) I I GOD - "0 I., ca e 251c'1 r c C r_ C c c r_ 0 0 0 0 0 .0 .2 .2 .0 . . . 0 0 i a cts C ca C c r_ C C (D c cm 4 4 rz 6 caa) d, cis0) Ch 0Cli ca Ch (1) ca 6 cc(D b (1)ma ca CLco z a. 05 z M Fn Z C'Li5z CL05 Z COL)Co z (L rb z CL.Fn Z ('L ih z 0 i2 cc, b E 0 t 8 8 "0 8.0 s 9 20 g0a I Co' . , - E-4 Figure 2.2 Regression of total annual catch of salmon in the Foyle area and corrected redd counts. 250000 200000 150000 100000 0 0 20,000 40,000 60,000 80,000 100,000 120,000 140,000 160,000 Total Commercial salmon catch (no. ) In years when the sport fishing catch returns were below the mean for the period 1952 to 2000, there was no significant correlation between these catches and any of the commercial catches or corrected redd counts. In years when sport-angling catches were high, i. e. exceeding the long-term mean, no significant correlations were found. When the corrected redd counts were less or greater than the mean of the 49-year period (81,647) there was no significant correlation with total commercial catch or sport angling or all catchescombined. 35 To derive a measure of commercial netting catches which was free from the effects of population size, commercial catch was linearly regressed on total population size and the residuals extracted. As auto-correlation artificially increased the significance of the regression model and adjusted its direction it is therefore impossible to use this to make any predictions. No interpretation is therefore put on the significance of this model, rather it is a means to develop a straight line through the data and derive the residuals for further analysis. These residuals significantly predicted variation in sport angling catches in a linear regression (FI,47--210.69; p<0.002; r2= 0.168 - Figure 2.3) with sport angling catch declining with increasing commercial catches independent of population size. Probability analysis showed, that when commercial netting catches were less than expected for the returning population size (i. e. below a0 residual), there was a 2.06 times chance that the long anglers' catch would exceed the term mean (4,321 fish) (Probability analysis - Relative RiskJRRJ = 2.06 with 95% confidence limits 1.24 to 3.41: X2 = 6.89; p_0.009 with a Yates Correction) (Lachin, 2000). A similar approachwas used to look in greaterdetail at the impactof total catch(including commercialand sportfishing) on the subsequentspawning population or redd counts. The total catch was regressedagainst total populationsize and the residualsextracted. These were thenplotted against the reddcounts. The relationshipwas not significant). 36 Figure 2.3 Commercial catch residuals regressedon corrected sport angling catch. 0c Q ro C C I -60000 . 50000 -40000 -30000 -20000 -10000 0 10000 20000 300DO 40000 Commercial catch. residuals 2.4 Discussion It is axiomatic that, were there no coastal and tidal water interceptory fisheries for migrant salmon returning to the Foyle catchment, there would be more salmon available for exploitation by sport anglers. As the commercial fisheries intercept the returning population before it becomes available to the sport anglers it is commonly perceived that such an impact does exist and results in depressedangling catches. The availability of high quality data from a very long-terrn monitoring programme has enabled a pragmatic approach to assessmentof the impact of one exploitation method on another. Thus, here the hypothesis was tested that commercial netting and trapping catches in the Foyle system have a significant negative impact upon recreational angling catches. Analysis of the data 37 would suggestthat this doesnot occur in the Foyle area. Data from the drift net fishery, the first interceptionexploitation of returning migrants, showsno evidenceof negative impact on the in-river draft net fishery or upon upstreamsport angling leading to the tentativeconclusion that it is populationsize that controlscatches by anglers. Significant positive relationshipsalso exist betweencommercial netting and trapping catches. The analysisof the residualswith sportangling catches allowed a more detailedexamination of the interactionbetween two competingforms of exploitation. The result of this was,when commercialcatches deviate from thosepredicted by populationsize, theseresiduals were significant predictorsof the sport fishery, explaining 19% of the observedvariance. In addition,there was a reasonableprobability in yearsof low commercialcatches when there were strong cohorts of returning adult migrants,that the sport fishery would benefit by higher catchesof salmon. Thus the hypothesisthat commercialinterception netting and trapping of Atlantic salmon reducessporting angling catchesin the Foyle catchmentis supportedby the data. However, this should be qualified, as the amount of variance explainedby the commercialnetting returns is relatively low leaving 83% of variance unexplained. The secondhypothesis that exploitation from all sources(sport angling and commercial netting and trapping) significantly impact upon the ult imate size of the spawning population,is not supportedby the data. As all exploitationof salmontakes place prior to the spawning period in any year, logically, exploitation rates in any one year may be expectedto depressthe spawningpopulation size. Strong positive correlationsbetween annual exploitation and ultimate spawning population size over a 49 year period superficially suggeststhat, at the levels of exploitation practiced in this catchment, exploitation does not have a significant impact on spawning population size. Further 38 analysis of the data did not highlight any significant effects. and the most likely explanationfor positivecorrelations between exploitation and spawningescapement is that yearclass strength is the determiningfactor. Other studies have observed that many animal populations can sustain very high levels of exploitation without any apparent adverse impact on overall population size, e.g. Laurian et aL (2000) in their work on moose populations in Quebec, Canada found that the species could sustain high levels of exploitation and had a certain adaptability which allowed it to maintain high productivity despite the intensive harvesting. Also, Bosch et aL (2000) who looked at the effects of culling on gulls and Frederiksen et aL (2001) who examined culling and its impact on cormorants both found that the populations could sustain very high levels of harvesting. With regard to Foyle salmon one explanationfor these findings is that years of high abundanceare maskingthe effects of exploitation in yearsof lower numbersof returning salmon. To test this yearsof low sport angling catch and spawningpopulation size (redd counts) were examined.No significant negative relationshipswere found, this would suggestthat even in yearsof lower abundance,the commercialnet fisherieswere still not having a negative effect on rod catches and that total exploitation (commercial and sporting)rates were not havinga negativeeffect on subsequentspawning escapement. Thus it is concludedthat in the Foyle catchment,although commercial fishing has been shown to have a significant effect on sport angling catchesthere still remains 83% of variance unexplained,therefore the year class strength of returning migrants is the principal modulatorof variation in the commercialnet catches,sport angling catchesand 39 escapement,and that this effect over-rides any potential conflicting impacts of one exploitation type on anotherand of total exploitation on spawningpopulation size. It is unclear how these findings might apply to populationselsewhere. The Foyle system supportsa highly abundantpopulation comparedwith other EuropeanAtlantic salmon rivers (five year running averagecommercial catch between30,000-35,000 salmon) so it may be able to support multiple exploitation methodsthat other systemsmay not. In addition the strongmanagement structure in the Foyle systemhas resultedin tight control of exploitationin yearsof low abundanceso protectingstocks during theseperiods. Thus, in the Foyle area,the questionregarding cessation of netting may be regardedlargely not as a conservationissue, but ratherone of resourceallocation. 40 Chapter 3. Life stage specific, stochastic environmental effects, overlay density-dependent filial cohort strength effects in an Atlantic salmon (Salmo salar L.) population from Ireland 3.1 Introduction All animal populationsare subject to controlling mechanisms,which limit their size. However the relative roles of density-dependentand density-independentfactors in determiningultimate population size have long been controversial(Elliott, 1985,1994; Sinclair, 1989; Newton, 1998).It is now widely acceptedthat for populationsto persist over time, at least one negativedensity dependentmechanism (where the probability of survivorshipdecreases as populationsize increases)must operate(Elliott, 2001).In nature, ultimate populationsize is unlikely to be determinedsolely by negativedensity dependent regulatorymechanisms but is most likely to be the result of complex interactionbetween density dependentand density independentfactors (reviews in Newton, 1998; Elliott 2001). Elliott (1994) in long-term studieson trout (Salmo trutta L. ) showedthat both density-dependentand density-independentfactors contributed to ultimate populationsize but their relative effects varied betweenhabitats. Studies such as those by Whittaker (1971) and Newton and Marquiss (1986) have demonstratedthat populations may be regulatedby density independentfactors in stableenvironments but under lessfavourable conditions,density independent factors can assumemuch greater importance. The evidencefrom a significant number of studiesover a broad range of speciesis that factorsmodulating animal population size do not operateat all times over the life-cycle but may act disproportionatelyat one or more ontogeneticor life-cycle events(Achord et aL, 2003; Langeland & Pedersen,2000). In red grouse (Lagopus lagopus) for example, 41 negativedensity dependance has been shown to occur during autumnwhen territoriality is at its height (Jenkins,1963). Similarly in oystercatchers(Haematopus ostralegus) density dependentregulation occurred during spring when densitiesincreased during the breeding period(Harris, 1970). In a review of the literature Sinclair (1989) showedevidence of "populationbottlenecks", i.e. life stagespecific, negativedensity-dependent effects, on populationsize in a number of populationsof insects,fish, birds and mammals. Identification of life-stage-specific density independenteffects on ultimate population size has receivedless attention than density dependentfactors, possibly becauseby their nature they are more difficult to detect. However, some studies have shown such effects. In a study of an Australian peregrine falcon (Falco peregrinus) population, Olsen & Olsen (1989) demonstrated environmental(flooding) density-independenteffects during nesting on egg mortality. In sea-trout(Salmo trutta) in Black Brows Beck, Elliott and co-workers(Elliott et al. 1997) showedsignificantly depressedpopulation size as the result of summerdroughts and the effect was life-stagedependent while a numberof studieshave highlighted the potentialfor exogenousenvironmental impacts on a number of salmonidspecies (Einuin et al., 2003; Hill et al., 2003;Azumaya & Ishida,2004). As a result of the existanceof distinct phaseshifts in the life cycle and periods during which abundancechanges, the Atlantic salmon(Salmo salar) is a specieswith the potential for populationbottlenecks (Elliott 1994;2001). Here for the Foyle system, Ireland, several long-term datasetson returning migrant Atlantic salmon population size are combined with environmentaldatasets to test two 42 hypothesesrelated to the control of ultimate population size. It is postulatedthat this population is primarily regulatedby density dependentfactors, but that environmentally inducedeffects operatingat specific life stages,which affect the magnitudeof changein populationsize during populationbottlenecks, overlie this effect. 3.2 Materials and Methods 3.2.1 Study Area The River Foyle systemdrains ca 4,500krnýof the northwestof the islandof Ireland (Fig. 1.1),discharging northwards through the Lough Foyle estuaryinto the northernAtlantic. This system supports a significant population of Atlantic salmon for which there is a long history of commercial net and trap fishing and sport angling extending back over several hundred years (Foyle Fisheries Commission, 1953; Elson and Tuomi, 1975; Foyle Fisheries Commission, 1996). Total catch in the Foyle area in 2002 was approximately I 10 tonnes, representingca. 11% of the total catch for the island of Ireland (Anon. 2002). The Loughs Agency of the Foyle, Carlingford and Irish Lights Commission and its predecessorthe Foyle Fisheries Commission, have collected data on catches and populationsize on an annualbasis since 1952. Thesedata are describedbelow. 3.2.2 Salmon Population Estimation As the Foyle salmon population is exploited, an estimate of the returning migrant population size has been maintained by the statutory body by combining data on commercialnet catcheswith sport anglingcatches and the fishery escapement(that portion 43 of the populationthat remainsfollowing exploitation).The origins of a numberof elements of thesedata are described below. Commercial Salmon Catches Four forms of commercialexploitation of returningmigrant Atlantic salmonhave operated since1952 on the Foyle system. Drift nets are gill nets deployedclose to the water surfacein the Foyle Estuary and in inshorewaters up to 12 miles out from the north coast. Drift nets have been in constant operationin the areasince 1952. Draft netsare in-river seinenets which are fished in the main River Foyle andRiver Roe at definedpoints. Draft netshave been in constantoperation since 1952. Thereis a statutory requirement for the users of both drift and draft nets to make catch returns. Stake nets are net traps run as a commercial fishery by the Foyle Fisheries Commission between 1952 and 1988. Annual catch records were collected by the Foyle Fisheries Commissionfor the durationof this fishery. Bag Nets are trap netswere of a similar constructionto the stakenets but were left in place permanently. Data are availablefor thesefrom 1964to 1990,the period spannedby the fishery. 44 Catch Effort of Commercial Net Fisheries The effort employed in these fisheries has changed over the years. The season is considerably shorter now (currently 15th June to 3l't July, cf March to September in 1952 the first year of the data set) (Loughs Agency, unpublished data) than previously and the number of hours fished each week has been reduced. There has also been a shift in emphasisbetween the fisheries,with the closure of some methodsof exploitation and a shift in the relative importanceof draft and drift net use. The drift nets particularly have become more efficient at catching salmon with increasing mechanisationand the availability of more effective netting materials. Over the yearsthe numberof returnsof catchdata from the commercialnets have been high and catchdata appear to broadlytie in with field staff observations(Wysner, pers. comm.). For the purposesof this paper no correction to catch data has been made for effort, as it is the absolutenumber of fish removedin the fisheriesthat reflect returningmigrant population size. SportAngling ExploitationReturns Despiteit beinga statutoryrequirement that sportangling licence holders make an annual catchreturn, the numberof returns,as a percentageof total licencessold was highly variablebetween 1952 and 2000. This is typicalof previouslypublished studies (Small, 1991). To determineannual catch rate by therecreational fishery a correctionfactor was usedwhich wascalculated using a techniquedescribed by Small(1991) to determinea realisticmeasure of anglingcatch. This correctionallowed for the non-reportingof fish caught when anglersmade a return and is as follows ((0.3/(percentagereturn of licences/100))+0.7) this producesa raisingfactor which is appliedto theactual number of fish declared. 45 Spawning Population Estimates - Redd Counts Since 1952the Foyle FisheriesCommission (and subsequentlythe Loughs Agency) field staff have made counts of redds (salmon nests) during and immediately following the salmonspawning period in approximately260 zoneswithin the catchmentannually. This is only a partial count of all reddswithin the Foyle catchment.The accuracyof counting can be subjectto environmentalconditions such as high flows, which can obstructproper counting through poor visibility, or flattening of redds making them difficult to see. However,as a methodof long-termpopulation monitoring, it is recognisedas a useful tool (Elson & Tuomi, 1975;Hay, 1984;Isaak, et aL, 2003). Highly accuratedata on spawning populationsize for one year (1999) showedthat countsat these260 sitesrepresented I I% of the total reddsbased on the percentageof spawningareas available to fish within the Foyle catchmentcalculated from instrearn habitat surveys (unpublisheddata, Loughs Agency).Thus to estimatethe absolutenumber of reddseach year in the Foyle catchment, a correctionfactor of 9.1 was appliedto redd countsfrom all years.Each redd wastaken as representinga singlefemale salmon (Hay, 1987). 3.2.3 Population Structure Scaleanalysis of 813 migrant adult salmonreturning to the Foyle from 1968,1969 and 1970 showed that the population comprised93.8% 1-sea-wintersalmon (fish that had remained at sea for only one winter) (Anon., 1969; 1970; 1971). The age of metamorphosisfrom the freshwaterto the seawaterphase (smolt age)was also determined from thesefish. On averageover theseyears 1% left the river at one year old, 92% at two 46 yearsof age and 7% after three yearsin fresh water. Scaleanalysis on a sampleof 81 salmon in 1998 (W. Crozier, pers. comm) found that 97.5% were I-sea-winter and 13% hada smolt ageof I+yrs; 84% hada smolt ageof 2+, and 3% of 3+. For subsequentanalysis, all fish were assumedto be I sea-winterfish but variation in smolt ageswere appliedto the data so that the recruitswere correctly apportionedto their respectiveparental year classby the following method. From adult scaleanalysis the ratio of smolts ageswere known for the years 1952to present(Loughs Agency, unpublished data). This information was used to apportion the recruits of spawning fish to the populationin their year of return i.e. fish spawnedin 1952would return as adults in 1955, 1956or 1957and wereapportioned accordingly. To comparelike-with-like all populationdata (commercial catch, recreational fisheries and redd counts) were convertedto an "egg number equivalene'. As fecundity is related to body size, a fecundity of 1,430 eggs kg-1 of fish (Shearer 1992) was used. The mean weight of salmon was taken from the commercial fisheries for each year used in the analysis. In addition, as the sex ratio is typically skewed in adult migrant salmon populations,a sex ratio for the population of 60:40 (female to male) (Loughs Agency, unpublisheddata) was usedto adjustegg depositionnumber for datafrom the commercial andrecreational fisheries. 3.2.4 Data Analysis To examine the hypothesesthat the Foyle Atlantic salmon population is regulatedby density-dependentfactors, the relationshipbetween parent and progeny population size, (total eggequivalent derived as above)was examined. 47 Linear regressionsand curvi-linear stock-recruitmentmodels (derived from the literature) were testedto find the model which bestexplained the variancein recruit populationsize. To test for additional environmentallyinfluenced modulatorsof population size, during key life cycle stages (potential population bottlenecks),a number of environmental parameterswere usedto predict variation in residualsextracted from the stock-recruitment model. Ufe-HistoryStages Eight specificlife history stageswere. recognised (Figure 3.1): 48 0 z C 0 421 (D tb to 0 E V)as ff) ýA4 Spawning and egg incubation The first stage in the salmon's life cycle examined was from November in year x to January year x+l, when adult fish are engaged in spawning and eggs are developing in the redds. Four variables were used to test for a relationship between environmental conditions during this period (Nov-Jan) and the size of the returning migrant cohort. These were: rainfall, water flow, air temperature, and the North Atlantic Oscillation Index (NAOI) (see below). LarvaelAlevin Emergence Subsequent to this, variation in environmental conditions during the period of alevin emergencefrom the redd; - February year x+1 to March year x+l, was tested for its effect on population size. Environmental variables used were: rainfall, air temperature and water flow, during this period. Fry establishment - Year 0+ Summer Following emergence, fry establish territories in the streams during their first summer (May year x+I to August year x+l). Three variables were tested for significant effects on population size during this period, these were rainfall, air temperature and water flow. Year 0+ Overwintering Overwinteringof fry was examinedby looking for significant effects during the period Novemberin year x+1 to Januaryyear x+2. The variablestested for effectson population sizeduring this periodwere rainfall, air temperature,water flow and the NAOI. 50 YearI+ Summersurvival The effectsof the environmenton one year old parr summersurvival, coveringthe period May year x+2 to August year x+2 were tested using three variables: rainfall, air temperatureand waterflow. Year 1+ Overwintering Overwintering parr during the period November year x+2 to January year x+3, were tested for effects on the overwintering juvenile salmon population size using the environmental variables rainfall, air temperature, water flow and the NAOI Smolt migration The smolt migration period May - June year x+3 was testedagainst the environmental variablesrainfall, air temperatureand waterflow. Marine Survival year x+4 The NAOI and sea-surface temperature anomalies north of Iceland were tested for significant survival effects during the period when the salmon were in their first winter at sea, November year x+3 to February x+4. 51 EnvironmentalVariables The environmentalvariables used are explained in greaterdetail below: - Air Temperature,Local Sea-surface Temperature and Precipitation Data This information was obtained from the Meteorological Office in Dublin. These environmental data were collected at Malin Head, Co. Donegal, which is at the entrance to Lough Foyle. Air temperature and precipitation data were available from May 1955 to present, while sea-surfacetemperature data were available from May 1958. The data used were an average for each month available. Waterflow Thesedata were derived from a hydrometricstation on the CamowenRiver at Omagh,Co. Tyrone, which is operatedby the Departmentof the Environmentfor N Ireland. For the analysisan averagemonthly flow rate was used. The data were availablefrom January 1975to December2000. The North Atlantic Oscillation Index (NAOI) The NAOI is an atmospheric phenomenon, which is measured as the difference in air pressure between the Azores and Iceland (Hurrell, 1995; Hurrell, 2003; Gillett et al., 2002). The winter index, used in this analysis, is calculated by taking the mean of the index between December and the following March. This Winter NAOI Index then referencesto the January of that year. 52 Winter sea-surface temperature anomaliesfOr Grimsey Island, Iceland The period November x+3 to January x+4 was used to examine possible environmental effects of the marine environment. These data were acquired from the British Atmospheric Data Centre (BADC) and are representativeof temperature change. Analysis Two statisticalapproaches were adoptedto examinethe variationin populationsize independentof densitydependent effects. Three parental-offspring population size models were tested(linear; Beverton & Holt; Ricker) using the softwarepackages SPSS and Statistica.The model which best fitted the data and provided the greatest predictive ability waschosen for furtheranalysis. Environmental variables were then used as independent predictorsof residualsderived from this modelin regressionanalysis. However, as it is highlypossible that any environmental effect on populationvariation may not be linearor may showsignificant threshold effects (see Elliott, 1997),population residuals in years with high for examplewater flow (upper30 percentile),low water flow (lower 30 percentile)and intermediate water flow (mid40 percentile)were compared using ANOVA with a posthoc Tukey test pair-wise group comparison. This techniquewas used with all of the environmentalvariables. To avoid Type I statisticalerrors, the acceptable probabilitywhen using multiple tests was correctedusing a Bonferronicorrection. For claritythe corrected probability equivalents are presented here where appropriate. 53 3.3 Results 3.3.1 Density Dependent Regulation Of the three models examined relating parental size to the number of offspring recruited into the population, linear regression explained the lowest amount of variance (17.1%), Beverton and Holt 27.2% and the Ricker model explained most variance at 31.2% (Table 3.1; Figure 3.2a; 3.2b; 3.2c). Table 3.1 Comparison of predictive linear; Beverton & Holt and Ricker models. Model r2 p n F Linear 0.171 0.003 43 9.88 Model r, n constant parameter Beverton & Holt 0.272 43 5.94 0.0005 Ricker 0.312 43 4.385 0.00017 This model suggests that maximum filial population size resulted from parental egg deposits of around 500,000,000 eggs. Egg deposition in excess of this resulted in a low rate of decline in filial recruitment to the population, deposition rates below optimal filial recruitment declined more rapidly with decreasingegg deposition. Thus, both positive and negative density dependent factors appear to modulate recruitment in this population. To derive a measure of survivorship, i.e. population size independent of parental population, residuals from the Ricker curve were derived. These were then used to examine the role of environmental factors controlling variation in survivorship. 54 Figure 3.2a A linear model fitted to parental population size (egg equivalent *100,000) and filial population size (egg equivalent* 100,000). 18,000 16,000 14,000 12,000 g 10, ODO 8,000 LL 6,000 4,000 2,000 01 ODO 2000 3000 4000 50DO 6000 7000 8000 9DOO Parental (eggs*100,000) 55 Figure 3.2b A Beverton & Holt model fitted to parental population size (egg equivalent 100,000) and filial population size (egg equivalent* 100,000). Model: v3=(constant*v2Y(1+(param*v2)) y=((5.9411483)*xy(1 +((0.0005041)lx)) 18000 16000 14000 12000 10000 8000 ÖJ3 6000 4000 .1 a'! 2000 11« 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 I 1 Parental (eggs*] 00,000) Figure 3.2c A Ricker model fitted to parental population size (egg equivalent *100,000) and filial population size (egg equivalent* 100,000). Model: v3=constant*v2*exp(-param*v2) y=(4.3854143)*x*exp(-(0.0001741)*x) 16000 . . 12000 . 8000 (A 4000 n 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 I I Parental (eggs* I 00, OW) 56 3.3.2 Life-stage specific - environmental effects on population size. Spawning and egg incubation Four variables,air temperature,water flow, rainfall and the NAOI, were usedto test for correlationsbetween survivorship and environment. Water flow (positively) (Figure 3.3a) and air temperature(negatively) significantly predicted population size residuals (Figure 3.3b). The two remaining variables did not significantly predict survivorship (Table3.2). Years with low air temperatures during the incubation period had significantly higher survivorship than years with high air temperatures but there was no significant difference with years of intermediate temperature and high or low temperature years in ANOVA analysis (Table 3.3; Figure 3.4a). Years of low water flow had significantly lower survivorship than years with high water flow during this period (Table 3.3; Figure 3.4b). Survivorship in intermediate water flow years did not differ significantly from that of either high or low water flow years. However, when corrected using Bonferroni these were no longer significant. To determinethe environmentalvariable most influencing spawningand egg incubation survivorship two variables water flow, air temperature were entered into a stepwise forward multiple regression. Water flow was the most importantdeterminant of variability in survivorship,however, air temperaturealso contributedsignificantly to explainingvariance, with thesetwo variables aloneexplaining 54.9% of total variance(F2,19 = 13.177:p<0.001). 57 Figure 3.3 Regressionsof environmentalvariables during life-stage specific eventsand populationsize residualsderived from the Ricker parental-offspringcurves. a) water flow during spawning and incubation, p<0.001; b) air temperatureduring spawning and incubation,p<0.009; c) rainfall during larval/alevinemergence, P<0.029; d) North Atlantic Oscillation during 0+ winter, p<0.022;e) water flow during 0+ winter, p<0.012;f) North Atlantic Oscillationduring I+ winter, p<0.011; g) North Atlantic Oscillationduring winter marine feeding, p<0.008; h) sea surfacetemperature anomalies north of Iceland during winter marinefeeding, p<0.002. 39 2000- 30 0- low - t 7 67.OS le 67 670 6723 i000- 100047 67.05 6671 -eliz-74.672 67ý 5 -2000 3D00 -m -4000 -m (cun»co) Wo" flow (eunwu) wäter now 10000 i om)- - 3b ww oow - 60oD 6=- 40W 4000 1 2W0 1 et 6 t ýo ;4 01 -ýI 45 :: 4W4 6 ep- 0,0 -10 - ý 0.1 o _ -400 Air tompendm (Fa Im hWt) NAO 10000 iooo- 8000 30 am- 39 6= *Wo 4W0 20W 0 ) 10 -20M1 -4000. -40W -ein -ein NA O io oo o-- 3h oow - 3d 8m. 6000- 6=- 0 40W 2 0 4000- .e16 - , 2000. 60 - ý* 1 t- 00-w -ýO jr4_wýOD; j loo 2w 300, -40M- iw 6000- ---- . - KAO BST(anonn394 58 Larvae/AlevinEmergence Of thevariables tested rainfall, air temperatureand water flow in theupper catchment, only rainfall was shownto be significantduring this period,however, following correction usingBonferroni this was no longersignificant (Table 3.2; Figure 3.3c). When population size residuals were compared in years with low, intermediate and high flow, the data hinted at an inverse relationshipbetween flow and survival (Table 3.2) althoughthe resultwas not significant. The regressionof rainfall on survivorshipexplained 9.3% of the total variance(FI, 39 5.11; p<0.029). Fry establishment- Year 0+ Summer Threevariables (summerrainfall; air temperatureand waterflow) were regressedon survivorship. None of thesewere found to be significantpredictors of varianceor correlatedto the populationresiduals, although water flow duringthis periodwas nearly significant(Table 3.2). ANOVA analysisof years of high, intermediateand low rainfall showedno significant differencesin survivorshipbetween these years (Table 3.3). 59 I s ON 0 Cf) a 0 0 -4 -4 -0 -. 4 0 - cf) N C14C14 N cn Nr 1* N IRT NT C4 't "I N NT IRT Itt Cq It IRT (14 C, 0 0 r. 4.42 "t N 00 N N %0 cn r- 0 - r- "0 rf) r- tn .. 4 N'o -4 ". 4 m Cf) C% 00 0 9 6 6 01 6 6 d d 6 6 6 al61 C26 06 . 8 r-: 8 1cc-1 C5 GA ce A ý A (7, 1ý0 en 00 -4 r, ýo 10 It Itt a ý cn C-4 cn Ci Cn C) C) q -4 . 0 cn . 6 6 c; C5 r.5 c? (3 o - Cý M r- B r- ýo 10 (14 en cq cf) , 8 0 W) oo ýo W) W) ýo t W4 0 W)4 IR cl (7N I r-: C,; , C; ci wi r-: 6 C-i6 "0 110 C14 Cý ON * & 0 0ý tn r- I en :t 00 ON ;s ON -4 r- ON q ":cn eq "" ,rý: 7, ý6 00 r- IRT 00 IRcq cn C5 N - C14 ýocn -, 0 C? r, i en T I I "I ZIO 0 0 CD. 0 CI. 0 cl 0 E 2d E co 4 0 4-4 0 r- ý rA 4.4 0 9) 4.4 1.4 b4 Qn cc Cd 1-4 Cd 1 l I cq en en en en en cq (1) W) IRT It Itt C14 It It C*4 C*4 RT cn W) jW ej 4r.1, Ll pw 14 14i4 0 a U 00 tf) i 0 ýoc f ) 1.0- 4 - 4 8 tn ýc00 1W) eq Ici " m r. l: %ro, d d (::ý 0 6 t4r) N 00 t- IT eq cl I., en -4 N -4 - en cf) -4 0 0 - 0 -. 4 c.4 00 %.o 00 00 ON tn C-1w! -1 C-1 0 0 W) dC'4 00 cq ". 4 r- C5 r-: C)I" "0 I I I I I I I I I I 0 rý 4R oo ý:t , 'n 8 A c 4 00 en 4 c 't C, ' 06 00 ýc1-4 C5cý 11 00 Cý0? 'T 00 -, 1-4 "Ci cl En 8E r. .19C = E 0 = 0 s P-.4 o .2 -S C) -< 4) ý 101. Cd 0 W 0 C14 0 I I 1 ON 'o 'o "D r- ýo C140 C" C', m Itt cn r- m r en M 00 -- '4 a 00 00 00 C#) ýo ýo 00 tlý ýo -4 ON 4 "t 0 t, ý.q t (7Ncn ýo m 0 N Ch V) en 00 "0 .2 -, 00 r- - Cý - 00N - 0 W'l Cý NT r- en en - - 1 C,"04 C',cn r- C,4 eq en ;S r- wl, 00 ýý ýo R - 00 0. 00 W) X? C "1*"-0 " 10 00-4 cf) ON 00 Cý = &W Cl. r. "0 1 I l I ON CD ""t --, -4 "' I wl o 110 eq 00 N 00 WI) ýo 00 W) 19t W ) r - r- CIO 0 C\ 6 r- c) 00 r- 4 C5 6 6 d 6 6 d d C) (=5 c o o 1- 00 -0 00 00 00 ON CN ON C., 00 CN Cý CD 0 V) Ci 't -t n n cf) cf) - cf) cf) -t - 't "t cli C14C14 N NI N C141clt I Clf cli I Clf CIT Cý N N r- 00 C\ 00 %,o 't en - en V*j 'r, W" cn C) - r r a 00 s tq (ý m - Cý (ON ý14 i cf) * ON00 W) 00 c 00 en - Cn 00 00 c cli Cý C5d d 6 Cý 6 6 ý6 cli d -0 cp o 0 7; E E E tz = E C) o 4" C) jwo 0 o P4 z 00" 0 r. 0 12 elý I N4 m cu \O 0 C14 00 WI) ýo ýc t- W, 00 en (7% ýo oo %.o C-A r- '9 ' ") 00 C,4 2ýý2 ,r- C14N . n. . 00ýo MC,4 ýo- enr- r-ýo en cnC) enIt W)I- C14 N It 10 - cc Go 10 l tn 00 It ýr tn"' oo;i 00 rn 0 in 00 0 en a ON 00 14 800 1"01 1* -. , a0 00 00 Ici ýc r- cq 't 00 r- en -. 4 tn - In cn ý9 Mo cn ýo 0 N N r- 6 C5C5 C5 6 6 6 C5 C;6 N cl C,4 C-4 C-4 cT Cf C4 cf Cq Cý "t cn - M r- r- en r- r- W) CN 00 -4 c, N 0 N 1 3 L) 2. E 0 a P- ýu lu 12 12 d) CEo 46In ý C) N .21ý I ro 0 . < 2 w . is .ý4 04 < z ý: P.,:ý ý: ý 04-V) "01-41 cl 1.4 CIO) 0 L_ L_-- L :3 Figure 3.4 Mean residual population size values derived from the Ricker parental- offspring curvesfor years with high (upper 30 percentile);low (lower 30 percentile)and intermediate(remaining 40 percentile)values for environmentalvariables during the life- stagespecific periods separately for: a) air temperatureduring spawningand incubation;b) water flow during spawning and incubation; c) North Atlantic Oscillation during 0+ winter; d) water flow during 0+ winter; e) North Atlantic Oscillation during 1+ winter; f) rainfall during smolt migration;g) North Atlantic Oscillationduring winter marinefeeding. All figures are significant at p<0.05 but columns with identical lettering are not significantly different in post-hoctesting. Figures 4a and 4b were not significant after Bonferronicorrection. 4a 4e 15w- 2000 1000- 1128 1500 1804 6w- 1000 0--a 500 c AC 500 «Am 0 -0 . 1000 - -610 -5w A , 1500-. AB -1130 -468 . . 181361 -1000 -2000 -1500' low Intermediate high low Inteffneclate hIgh Air temperature NAOI 4b 41 2,366 25W 0 U2174 M-746 9.500 1500 11,500 600 000 -1125 B BW AB -326 c0A ; 1600 2260000, 25W I low intermediate high low ktermeclate high wetwIflow Raintaft 4C 4g 2500 25M 2000 Ai 2000 1500 1500 1000 1000 500 RAB 5w Ba 0 0 E-1328 -W c -500 A '500 A -387 -1000 -1000 -W -1500 1500 low intermediate high low Intermediate Ngh NAOI NAOI 4d 0- 9 »500- -1000. -719 -1115 ABC -1999 ABC 20D0 2500 low Intermediate high waliff now 64 Year 0+ OverWntering Of the four variables examined (rainfall, air temperature, NAOI & water flow) only the NAOI and water flow were found to significantly predict survivorship. The NAOI had an inverse relationship (Figure 3.3d) while water flow was positively correlated with survivorship (Table 3.2; Figure 3.3e). Correcting for multiple tests using Bonferroni however meant that the NAOI was not a significant predictor. Years with low, intermediateor high values for each of theseenvironmental variables showedsignificant differencesin survivorshipfor the NAOI (Table 3.3; Figure 3.4c) but not quite significant for water flow, althoughthe data hinted at a positive relationshipto higherflow ratesin the uppercatchment (Table 3.3; Figure 3.4d). To determinethe environmentalvariable most influencing 0+ winter survivorship,NAOI and water flow were enteredinto a stepwiseforward multiple regression.Water flow was found to be the most importantdeterminant explaining 23% of the total variance(F1,21 7.562;p<0.012). NAOI did not add significantlyto the model. Year 1+ Summer survival Summerrainfall, air temperatureand water flow effectson survival were ascertainedusing regressionand ANOVA. None of thesevariables were found to havea significanteffect. 65 Year 1+ overwintering Of four environmentalvariables (rainfall, air temperature,NAOI & water flow) examined only the NAOI was found to be significantly inverselyrelated to survivorship(Table 3.2; Figure 3.3f). Comparing survivorship in years of low, intermediate and high values of the four environmentalvariables showed that only survivorship in high, intermediateand low NAOI differed significantly (Figure 3.4e). Post-hoctesting showedthat survivorship in low NAOI years was significantly higher than that of intermediate years, but not significantfrom yearsof high NAOI. The NAOI explained12.5% of the total variance(FI, 42 = 7.157;p<0.01 1). Smolt migration Rainfall, air temperature, water flow and the sea surface temperature at Malin Head did not significantly predict survivorship. However, comparing years with low, intermediateand high values for rainfall using ANOVA showedsignificant differencesin survivorship(Table 3.3 & Figure 3.4f). Post- hoc testing showedthat survivorshipwas higher in yearswith a high rainfall at this time comparedwith yearsof low or intermediaterainfall. 66 Marine Feeding- year x+4 The NAOI significantly predicted survivorship(Table 3.2; Figure 3.3g) (inverse relationship),as did the seasurface temperature anomalies north of Iceland (positive relationship)(Figure 3.3h). Post-hoc testing showed that low NAOI years had significantly higher survivorship than yearswith intermediateor low NAOI (Figure3.4g). To determinewhich environmentalvariables (NAOI & SST) best predictedsurvivorship they were enteredinto a stepwiseforward multiple regression. Sea surfacetemperature was the most importantdeterminant of variability in population size residualsexplaining 21.7% of total variance(FI, 34 = 10.718:p<0.002). NAOI did not add significantly to the model. 3.4 Discussion The evidencepresented here supports the hypothesisthat density-dependentmortality is an importantunderlying mechanism controlling Atlantic salmonpopulation size in the River Foyle catchment. Modelling the effect of parentalpopulation size on the filial population showeda significantrelationship, with the Ricker curve providing the best model fit to the data. The domednature of this curve suggeststhat negativedependence occurs at high densities.Elliott (1994) showedfor brown trout that this modelfits datafor other salmonid populationswhile other authorshave shownthis dome-shapedmodel fits Atlantic salmon (Chadwick, 1982; 1985a; 1985b; Kennedy & Crozier, 1993;). However, in the study presentedhere these density-dependent effects only accountedfor 31% of the total annual 67 variability in population size. Clearly other factors also influence the population size. Here, a life stage specific approachwas adopted to aid the identification of factors modulating the remaining variation in population size. Biologically important critical survivorshipperiods ("bottlenecks"sensu Elliott, 2002) were identified and broad scale environmentaldata usedas predictorsof survivorshipvariance to determinetheir relative influence on a comprehensivesuite of the critical life stage specific events faced by Atlantic salmonduring ontogeny. This approachshows clear evidenceof environmentally inducedpopulation regulating effects at a numberof critical life stages. During the spawningand incubationperiod, standardisedin this study betweenNovember and Januaryof the following year, it has been shownthat air temperatureand water flow are good predictorsin explainingresidual variation. Higher survivorshipis correlatedwith lower air temperaturesand higher water flows in the upper catchment. Approximately 54.9%of survivorshipvariation was explained at this life stage.It is unclearif theseresults are linked to the spawningperiod itself or the incubationperiod of the eggs. However,it is assumedthat water temperaturewill be a function of air temperaturealthough probably with some time delay. In previous studies it has been shown that a drop in water temperatureis requiredbefore spawningoccurs (Shearer,1992) and it is well recognised that egg incubationis controlleddirectly by water temperature,higher temperaturesleading to fasterdevelopment and emergence from the redd(Heggeberget, 1988). On emerging from the redd, the alevin must establish a territory and this period is recognisedas a populationbottleneck in other salmonidspecies (Elliott, 1994;Jones et aL, 2003). High water during this time, when fish are first exposedto full stream flow conditions,can be particularly important in determiningultimate population size (Elliott, 68 1994). Consistentwith this, it was shown that rainfall explained9.3% of survivorship variationat this life cycle stage. Although water flow was not significantat this stage,this could be as a result of low samplesize as low rainfall predictedhigher survivorship. Previous studies (Elliott, 1994; Elliott, 1997) have linked high periods of mortality to summerwhen water conditionscan havea severeimpact on productivity. This may occur through direct mortality or indirectly, for exampleby affecting growth ratesand therefore potential for survival at other crucial periodssuch as first entranceto the marine phaseof the salmonids'life cycle. In the study presentedhere, there was no direct evidencefor environmentalcontrol of survivorshipduring summerperiods, although survival in either year 1 (0+) or year2 (1+) was nearlypositively significantly relatedto waterflow. In contrast,the periodswhich appearto influencesurvivorship in the Foyle catchmentare more prominentin winter. Survival for both 0+ and 1+ winter periodsis stronglylinked to the negativephase of the NAOI and in the first winter to increasedwater flow in the upper catchment. Thus environmentalfactors accountedfor 23% and 12.5% of survivorship during first and secondwinter periodsrespectively. The final critical phase in the freshwater element of the salmons' life-cycle is that of its entrance to the marine environment as smolts. Here it has been shown that high rainfall greatly benefits survivorship during this period but interestingly the worst case scenario appears to be that of intermediate flows, very low flows having a much smaller negative effect. It has been shown previously in other studies that high water flows greatly facilitate movement of these fish downstream (Shearer, 1992; McCormick et aL, 1998; Byrne et aL, 2003) and that low flows may be detrimental to survival. However the impact of these 69 intermediateflows havenot been observedpreviously. One possibleexplanation for this finding is that in very low flows the fish are less likely to migrate actively downstream while in intermediatewater conditionsfish will attempt migration and thereforemay be moreprone to predation. Survival once the salmonreach the seais known to be affectedby marine conditionsand thesehave been linked previouslyto the NAOI and also seasurface conditions for North American stocks of salmon (Friedland et al., 1993; 2003a; 2003b). Very little similar information is available for Europeanpopulations although Martin and Mitchell (1985) linked sea surfacetemperatures north of Iceland to differing abundancesbetween single and multi-seawinter salmonand Friedlandet al (1998) linked a thermal habitat index of 10-13'C in the North Sea.to survival of Scottishand Norwegian stocks. The NAOI may influencesurvival in a numberof ways, suchas affectingcurrents, salinity profiles and seasurface temperatures. These conditions may impact on salmondirectly, for instance it is known that salmon cease feeding at temperaturesbelow 3'C (Shearer,1992), or indirectly by affecting prey or predator abundance. Sea surface temperaturenorth of Iceland appearsto be of particular importanceas this variable was found to explain 23% of variance. There may of coursebe other mechanismsor factors which are linked to theseconditions of which little is known at presentand which are affectingthe abundanceof the salmonduring this phaseof their life-cycle. The Ricker density-dependentmodel explaineda proportionof the variancein recruitment, however, this is not to state that the mode of operation of environmentalfactors is completelydensity-independent. For example,in looking at the effects of water flow on 70 0+ and 1+ over wintering, higher water flows lead to greaternumbers of fish surviving. This is possiblydue to the effect of greaterareas of habitatbeing available as a direct result of the increasedflow. If so, this implies that theseenvironmental factors are operatingin a density-dependentmanner as well as having somedensity-independent effects. This study highlights the importanceof density-dependentregulation as a controlling mechanismfor the salmonpopulation in the Foyle area. However,it has beenshown that environmentalfactors also play an important role in determining ultimate returning populationsize driving life-stageevents that may potentially act as populationbottlenecks. It is concludedthat the life-stage specific environmentaleffects which are shown to be operatingin this populationare likely to havesimilar effects in other salmonidpopulations, and will contributeto the apparentstochastic variation in populationsize resultingfrom the applicationof traditional stock-recruitmentmodels. The identification and quantification of these effects, from this long-term dataset,should improve the predictive ability of models by enabling the constructionof more sophisticatedmodels that combine parent populationsize and environmental factor variance. 71 Chapter 4. The influence of broad scaleclimatic phenomenaon long- term trends in Atlantic salmon population size: an example from the River Foyle, Ireland. - 4.1 Introduction Many animal populations are known to experience large-scale fluctuations in population size e.g. lemmings (Lemmus lemmus (L)) and voles (Clethrionomys rufocanus) in Northern Norway (Ekerholm et aL, 2001) and red grouse (Lagopus lagopus)throughout their range(Jenkins et aL, 1963;Newton, 1998). Many of these cycles are related to natural phenomena, operating independently of direct anthropogenicinfluence, for example,changes in the distribution and abundanceof the Pacific sardine(Sardinops caeruleus) and northern anchovy(Engraulis mordax) havebeen shown to be relatedto climate shifts in the north-eastPacific over a 70 year period (Rodrfguez-SAnchezet aL, 2002). With regard to Atlantic salmon Martin and Mitchell (1985) showed correlations betweensea surface temperatures north of Icelandand the abundanceof I -sea-winter and multi-sea-winterfish returningto the River Dee, Scotland. Increasingsea surface temperaturesin the sub-Arctic were associatedwith larger numbers of adults returning as multi-sea-wintersalmon and fewer returning as I-sea-winter salmon. They also showed that the averageweight of I-sea-winter salmon increasedwith populationsize of that group. Friedlandet aL (2000) also found linkagesbetween survival of salmonand sea-surfacetemperatures in the North seaarea. 72 Similar linkageshave been found for Pacific salmon(McFarlane et aL, 2000; Mueter et aL, 2002) and North American stocks of Atlantic salmon, where studies have correlatedmarine survival of Atlantic salmonto winter seasurface temperatures in the Labradorsea (Friedland, 1998; Drinkwater, 2000; Reddin et at., 2004). Previous studies have shown that the atmospheric phenomenon the North Atlantic Oscillation is also related to stock abundance in Atlantic salmon (Friedland et aL, 2003; Beaugrand & Reid, 2003). This index can be used as a proxy for climate variation on a large scale. The North Atlantic Oscillation in winter (NAOI) has been shown to provide a good index of the dominant mode of winter climate variability in the North Atlantic region ranging from central North America to Europe and into much of Northern Asia. It is calculated as a ratio of mean atmospheric pressure in the Azores to that of Iceland between December and the following March. Thus extreme high mean values result from intense low pressure centred over Iceland, corresponding with high pressure centred over the Azores; when this situation is reversed, the corresponding values of the NAOI are low (Hurrell et aL, 2003). A high value NAOI is likely to be indicative of higher frequency and more violent winter storms crossing the Atlantic Ocean on a more northerly track. This results in warmer and wetter winters in Europe and in colder and dryer winters in northern Canada and Greenland, while the eastern United States is more likely to experience milder and wetter winter conditions (Hurrell et aL, 2003). As a consequence,sea surface winter temperatures over much of the Arctic will generally be colder at a higher NAOI (Dickson & Turrell, 2000; Hurrell et aL, 2003). Low NAOI values are likely to indicate fewer and weaker winter storms bringing moist air into the Mediterranean and cold air to northern Europe. However, areas such as Greenland will have milder 73 winter temperaturesand as a result, Arctic seasurface temperatures will be generally higher (Huffell et aL, 2003). The NAOI has previously been shown to correlate with changesin behaviourand abundanceof a wide range of natural animal populations, from the timing of migration in birds (Forchhammeret al., 2002),plankton abundance (Reid & Planque, 2000),abundance of squid (Loligoforbesi) in Scottishwaters (Pierce & Boyle, 2003) to emergenceof sea-troutfry from spawningbeds (Elliott et aL, 2000). The spatialscale and geographicrange over which the NAOI appearsis indicativeof climate patterns and makes it a potentially powerful predictor of population fluctuations in European Atlantic salmon foraging at sea in winter. Determiningclimatic influenceson the dynamicsof wild populationsrequires long- term measuresof populationabundance; such datasets are very rare. Howeverwithin the Foyle catchment,Ireland, there is a long history of commercialfishing for Atlantic salmon extendingback severalhundred years (Anon. 1953; Elson & Tuomi, 1975; Anon., 1996). Commercialcatch records have been maintained since 1875with only a short break beforeand during the First World War (1909 to 1919). Theserecords show periods of abundanceand periods when the catcheswere substantiallylower. Many reasonsfor these changeshave been proposedsuch as Ulcerative Dermal Necrosis(UDN), drainageschemes, pollution and land use change(Elson & Tuomi, 1975; Shearer,1991; Magee et aL, 2003; Waring & Moore, 2004). However, here theserecords are usedas an index of population size, to examinelong-term change and the potential effects of climate. Specifically, the hypothesisis testedthat the 74 NAOI can be used to explain variance in the abundanceof salmon in the Foyle catchment.In addition, published data from climate change models predicting the effect of global warmingon the NAOI areused to makepredictions about the possible effectson salmonpopulations in the future. 4.2 Materials and Methods Catch data are not absolutemeasures of populationsize (Shearer,1992), but there is evidencethat thesedata do representa significant relative measureof the dynamic change in returning migrant population size for the Foyle system. Firstly, catch returns apparentlyaccurately reflect actual catchesfrom the commercial fisheries (Wysner,pers. comm. ). Secondly,over a 49-year period since 1952,there has been an independentmeasure of populationsize. The numberof redds(nests) have been monitored annually at 260 sites spreadthroughout the catchment.Over the period 1952 - 2000, commercialfishery catch was a very good predictor of redd number ?=0.45; (Linear regression:FI, 47= 39.9; p<0.001).These data strongly supportthe principal assumptionof the study presentedhere; namely that commercialnet catches reflect change in the relative population size of the returning migrant salmon populationto the River Foyle. Data from commercialcatch returnsare collatedby the LoughsAgency of the Foyle, Carlingford and Irish Lights Commission. Prior to 1999, its predecessorsthe Foyle FisheriesCommission (1952 - 1999)and the Foyle andBann SystemsLtd. (pre 1952) gatheredand collatedthis information. Overall commercialcatch returnsdating from 1875 are availableon a yearly basis. The commercialcatch returns arise from four 75 types of fisheries.Drift nets are gill nets deployedclose to the water surfacein the lough seawardsof Lough Foyle. Draft netsare seinenets, which are deployed in-river in the main River Foyle and the River Roe at defined netting stations.Stake netting was a commercialtrap fishery operatedby the Foyle FisheriesCommission between 1952and 1988.Bag netsemployed a similar constructionto stakenets but wereleft in place permanently.One licencewas issuedannually to be fished in the Lough Foyle estuary,although in 1988 an additional licence was issued. Data for thesenets are availablefrom 1964to 1990when operationsceased. Post 1952when more detailed records are available, the data show catchesof adult salmon peaking during the monthsJune to July. Fish countingstations operating at presentindicate that the main run of returning adult salmon occurs during these months with lower numbersof salmon returning in the spring and autumn/winter periods (Loughs Agency, unpublisheddata). The catchdata used here were not correctedfor effort, as prior to 1952no measureof effort wasrecorded. Data on the NAOI are availablefrom 1856 onwards(Hurrell, 1995; Hurrell, et aL, 2003). The winter index usedin this analysiswas calculatedby taking the meanof the index betweenDecember and the following March. This winter NAOI then referencesto the Januaryof that year. NAOI varies from year to year, but also exhibits a tendencyto remain in one phasefor intervalslasting severalyears. In order to smoothout inter-annualvariation and examineperiods of abundance,a five-year runningaverage of both the NAOI andcommercial catches was used in the analysis. 76 The datawere analysedusing the statisticalsoftware package SPSS 10.0. Linear and breakpointregression analyses were used to determinethe relationshipbetween the long-termcatch data and the NAOI. RelativeRisk Analysis was usedto determinethe probability of catchesexceeding a definedthreshold (Lachin, 2000) 4.3 Results The commercialcatches of Atlantic salmonin the Foyle systemvaried significantly between 1875 and 2001 from a maximum of 149,635 to a minimum of 12,500 salmon.Annual meanreported catch was 52,082+/- 2,541 (mean+/- S.E. ). Although there is considerableinter-annual variation evident, some underlying year-on-year trendsare apparent(Figure 4.1). A five year running meanshowed that catcheswere high at the end of the 1800s;in the mid 1930sto 1940s;in the 1960sand early 1970s and againin the 1980sand early 1990s(Figure 4.2). Figure 4.1 Annual commercialcatch of salmonin the Foyle area1875-2000. 160000 140000- 120000- 0 E 100000- 1 80000- is 60000- c; 40000 - 20000- 0i 1875 1900 1925 1950 1ý75 2000 Year 77 Figure 4.2 The five year average commercial catches of migrant Atlantic salmon from the Foyle catchment 1875 to 2001 & the five year average NAOI: 1875 - 2000. 140000 4 no. of salmon 120000-- --3 ...... NAOI 100000-- 2, ý 0 :% S 80000-- It ev Oi 60000-- --t 40000-- 20000-- ---2 0 -3 1875 1900 1925 1950 1975 2000 Year To examinethe effect of climatic changeon catches,the five-year running meanwas regressedon the five year running mean for the NAOI. This showed a highly significant negative relationship (Fj,jo6 = 79.07, r2 = 0.43, P<0.001, b= -0.654, intercept= 55,790,test of gradientt= -8.892, P<0.0001). However, there is some evidencethat the relationshipbetween NAOI and catch was not constantacross the full range of NAOI values. Breakpoint analysis was used to determine the point of any potential change of influence of the NAOL This analysis suggesteda substantial NAOI relationship change above a NAOI value of 0.151. Years when the was above this point showed no significant relationship between the NAOI and catches (FI, 50= 1; 1.08). However, for 1.16; P=0.286; b= -0.15; intercept = 45,66 test of gradient: t= - years with a NAOI below 0.15 1, there was a highly significant negative relationship b= intercept 41,005.8; (FI, 54 = 129.97; r2 = 0.701; P<0.001; -0.84; = test of 78 gradient: t=- 11.4; P<0.00 I) (Fig. 4.3). It was noted that the years of the lowest recorded NAOI (1960's) coincided with the highest catches (Fig. 4.1). Figure 4.3 The relationship between the winter NAO and commercial fishery catches of Atlantic salmon returning to the River Foyle over 122 year. Breakpoint analysis shows a uncoupling of a negative relationship about a NAOI of 0.15 1. 140000 "" 120000 "" w 100000 .' . 80000 " "" Z 0 60000 fj 40000 $ (0 20000 02 34 NAO Seven models of the effects of climate change on the NAOI resulting from a 1% per annum compound change in atmospheric carbon dioxide levels have been summarised by Gillett et al. (2003) (Figure 4.4). All models predicted an increasing NAOI with time. The mean of the seven models suggestsNAOI will increase approximately five fold over the next 75 years. Assuming that the NAOI has a similar responseto change over time as does the NAOI for the rest of the year, NAOI values of under 0.151 will become less frequent than in the past, the frequency of occurrence decreasing with time. There is some evidence that this may already be occurring as in only 2 of the last 10 years has the NAOI dropped below 0.151. As a result, data presented here would suggest that the linkage between the NAOI and population size would frequently remain uncoupled in the future. In addition, these data predict a change in 79 the frequencyof abundant,returning migrant populationsof salmon.Relative Risk Analysis showsthat the probability of achievinga catch of salmongreater than the median(47,800 between 1879 and 2001) is 2.34 times lower for a NAOI greaterthan 0.151 than for a NAOI of less than 0.151 (Relative Risk Analysis: RR=2.34; 95% confidence limits = 1.52-2.34; Yates corrected P<0.0005). Thus, assuming that climate changemodels do reflect likely patternsof changein the NAOI, then the long-termfuture medianpopulation size of returningmigrant salmon is likely to be lower thanthat of the past. Figure 4.4 The changein North Atlantic Oscillation predicted by seven climate changemodels. Dotted line showsthe meanof the sevenmodels, solid lines show the 2 extremesof the predictedNAO change. 12 10 "" "_"_4 ." _. -... . 4- . "" CL 0 4.4 4c z 2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 Year I-Predicted Max -Predicted Min ...... Average 80 4.4 Discussion The population size of adult Atlantic salmon in the Foyle system as indicated by commercialcatch returns has been highly variable over the years since 1875, but catchesare generallyregarded as large when comparedwith other systems(Shearer, 1992). Within this variation there are generaltrends of changingabundance linked with variation in the North Atlantic Oscillation. This feature may affect salmonid abundancein severalways. In the freshwaterphase, Elliott et aL (2000) found that the dateof fry emergenceover 30 yearsin a seatrout nurserystream, predicted by an individual basedmodel (Elliott & Hurley, 1998),correlated highly significantly with the NAOL Theseauthors also found a significant relationshipwith instreamwater temperatureand suggestedthat this may be the actualmechanism driving variation in the model. As a measureof the dominant mode of winter climate in the northern hemisphere,the NAOI has also beenlinked directly to seatemperatures which affect salmonwhen feeding at sea(Martin & Mitchell, 1985;Reddin et aL, 2000;Dickson & Turrell, 2000). Here, a negativerelationship is shown betweenthe NAOI and the abundanceof migrantsalmon returning to the River Foyle in Ireland for indicesbelow 0.151. Above 0.151 this relationshipbreaks down and at higher values there is no significant relationshipbetween NAOI and catch abundance.The proximateclimatic mechanismmodulating salmon abundanceis unclear, but there are a number of possibilities.Dickson and Turrell (2000) outlined the relationshipbetween ice flux and a positiveNAOL which increasedinto the 1990s.This potentiallyaffects salinity and watertemperature. The NAOI may also indicatechanging thermal habitat optima available to salmon through the cooling or warming of either migration routes or feeding areassuch as thosenorth of Iceland or off Greenland. Martin and Mitchell 81 (1985) found evidenceto suggestthat the abundanceof multi-sea-winter salmon returning to the River Dee in Scotland, was linked to an increasein sea-surface temperaturein the waters north of Iceland, which has in turn been linked to a decreasingNAOI (Hurrell et aL, 2003). This mechanismsuggests that a strongly negativeNAOI resultsin an increasedquantity or quality of thermalhabitat available to Atlantic salmon,resulting from increasingwater temperaturein theseareas. In the northwest Atlantic, a thermal habitat index has been used effectively to predict salmon abundance.This index shows a significant relationshipwith the NAOI and may thereforebe considereda useful tool in determiningsalmon abundance (Dickson & Turrell 2000; Beaugrand& Reid, 2003). The study presentedhere indicatesthat this may not be valid now or in the future. Although when hindcastingover the last 126 yearsthe NAOI is clearly a significant predictorof returningsalmon migrant numbers, it is clear that the value of the NAOI as a forecastingtool to predictpopulation size is highly suspect.Here it is shownthat when the NAOI exceeds 0.151, then predictivecapacity breaks down. Over the last decadethe NAOI has been greaterthan 0.151 in 8 out of the 10 years. Basedon current climate changemodels, this is a trend that is likely to becomemore extreme. Thus it is concludedthat it is unlikely the NAOI will be a useful tool to forecast Atlantic salmon population abundanceand that its inclusion in predictive models shouldbe viewed with caution. 82 Chapter 5. Local instrearn and catchment spatial scale habitat characteristics determine 0+ fry density of Atlantic salmon (Salmo salar L.) in the River Foyle. 5.1 Introduction Temporal variations in population size arise as a result of density-dependentor density-independentfactors and the interplay betweenthem (Elliott, 1994; Newton 1999; Harwood et aL, 2003). In contrast,spatial variation in density is primarily a function of the habitat in which animals live (Jenkins, 1964; Kennedy & Strange, 1980;Kennedy & Strange,1982; ElliotL 1994). For a numberof speciesenough is known about the relationshipbetween habitat use and population size to be able to construct predictive models to estimatecarrying capacity of a given habitat. For example Alisauskas and Arnold (1994) in a study on American coots (Fulica americana)found that a linear relationshipexplained a high proportion of variance (r2---0.54;p<0.001) between an estimateof populationsize (numbersshot eachyear) and the numbersof ponds on the prairies the preceedingsummer. Models such as these allow predictionsof spatial patternsof habitat use and give an insight into mechanismsoperating on the populationthus enablingpredictions to be madewhich supportpopulation and habitatmanagement. This is particularly importantin species of high conservationvalue or which are exploited. One such speciesis the Atlantic salmon (Salmo salar L. ). A number of habitat type, population regressionmodels linking local habitat type and density already exist for stream dwelling salmonids (B inns & Eiserman,1979; Milner et al, 1995,1998;Grant et al., 1998;Poff & Huryn, 1998). 83 As some elementsof habitat for stream dwelling salmonids, for example water quality, are partly a function of activity higher in the catchmemthis study attemptsto extend existing models by considering environmental effects at both local and catchmentscales. Specifically, it attemptsto model effectsof the uppercatchment on the deviationsin densityestimated by local habitattype. 5.2 Materials and Methods Five minutetimed electrofishingsurveys (Crozier & Kennedy,1994) were carried out using an Electracatchbackpack electrofisher at 307 and 279 sites, during July and Augustof 2001 and2002 respectively,throughout the River Foyle catchment,Ireland. This work is part of the LoughsAgency's ongoingdevelopment programme and was carriedout by field staff. The Agencyhas a maximumof 350 suitablesites identified, a target of 200 of theseare to be fished at a minimum each year. For each site all underyearling(0+) salmonfry and brown trout (Salmotrutta L. ) were counted.Fish lessthan 7 cm werecounted as 0+. Other fish speciessuch as stoneloach(Barbulata barbulatus),roach (Rutilus rutilus), eels (Anguilla anguilla), encounteredwere noted as present. Sites were chosenspecifically for their generalsuitability for juvenile salmonids,i. e. relatively shallow (10-30 cm deep)and relatively coarsesubstrate. Sixteen measures of in-streamand banksidehydrology and geomorphologywere gatheredconcurrently with the electrofishingdata and used in the analysis(Table 5.1). The widest site fished was 33 in with a maximum depth of 30 cm. The greatestmaximum depth fishedwas 120cm with a site width of 24 in. 84 Table 5.1 Local scalesite-specific variables collected for eachsampling for 350 sites in the Foyle catchment. Width - this was the average width at the site and was measured in metres. Minimum depth - minimum depth of water measured in centimetres. Average depth - average depth of water at the site measured in centimetres. Maximum depth - maximum depth of water measured in centimetres at the site. Flow (m-) -A net float was timed over a 10 m stretch and recorded in seconds. Bankside cover - right hand bank - this was recorded as a percentagefor both banks. Bankside cover - left hand bank Over hanging cover - right hand bank - this was recorded for presenceor absence. Over hanging cover - left hand bank Percentagebedrock Percentageboulder Percentagecobble Percentagegravel Percentagefines Percentagesand Percentagemud For the purposesof analysis,the data sets were combined, and every secondsite extracted. One set of datawas usedto constructthe model while the secondset was used to test its predictive power. To categorisethe broader scale catchment on salmon fry density, data for the catchmentupstream of each survey site were extractedfrom 1:50,000 Discoverer SeriesOS mapsand GeologicalSurvey of N Ireland maps(1: 50,000 Drift Edition and 1:250,000 Quarternary Edition). Catchmentdata collected included (Table 5.2): 85 Table 5.2 Catchment scale characteristicscalculated from ordnance survey and geologicalsurvey maps for eachof the 350 samplingsites in the Foyle catchment. Catchment area (krrý) Stream order Distance of the site from the stream source (km) Distance of the site from the streamsconfluence with the main River Foyle (km) Altitude (m) Slope at the site Presenceof lake upstream No. of housesper km' upstream of site Distance to nearesthouse (km) Distance to nearestfarm (km) Percentageof urban areasupstream Percentagearea of grassland upstream Percentage rea of woodland upstream Percentagearea of peat upstream Percentagearea of glacial alluvium upstreaiý Percentagearea of glacial lacustrine upstream Percentagearea of glacial calcareoustufa Percentagearea of glacial boulder clay Percentagearea of glacial sand and gravel Percentagearea of bed rock Percentagearea of gneiss Percentagearea of felsite Percentagearea of diorite Percentagearea of sandstone 5.2.1 Catchment Scale Characteristics Catchment area The catchment area above each sampling site was calculated by determining the river networkboundary above the samplingsite andcalculating catchmentarea in kM2. Streamorder Streamorder was derivedusing the methoddeveloped by Horton (1945). 86 Distance from source The distanceof the electrofishingsite from the upper-mostreaches of that arm of the river system. Distance from confluence Wth the River Foyle This is the distanceas measuredfrom the map of the survey site to the streams confluencewith the main-stemof the River Foyle. Altitude This was taken by interpolationfrom height contourson the map either side of the site. Stream Gradient The stream'sgradient was derivedby measuringthe horizontal distancebetween the nearesttwo contourlines and dividing this by the changein altitude. Presence or absence of a lake upstream Number of houses present This wascalculated by countingthe individual housesfrom the map. 87 Housing density A densityfor the catchmentwas then derivedby dividing the numberof housesby the catchmentarea (km2)above the samplingsite. Distance to nearest house The distanceto the nearesthouse from eachsite was measuredin metres. Land use Three categoriesof landuse:urban; grasslandand woodland were defined from the 1:50,000 series 'Discoverer Series' maps. The areaof eachwas measuredin km2for eachcatchment, above each sampling site. Thesewere then convertedto percentages of the sub-catchmentarea and arc-sine transformed before fuller analysis. Geology Seven geologicalcharacteristics of the catchmentoutlined below were also categorisedfor eachsite. These were obtained from 'Drift andQuaternary editions of GeologicalSurvey of N Ireland' maps. The areas,in kmý,covered by the various depositswere measured on a digitisingpad and converted to percentageswhich were thenarc-sine transformed for analysis. 88 The characteristicsused were: Drift 9 Peat * Glacial alluvium - Waterbornematter depositedby rivers 9 Glacial lacrustine alluvium - Similar to alluvium only deposited in a lacustrineenvironment " Glacial calcareoustufa " Glacial boulderclay " Glacial sandand gravel Solid " Bedrock " Gneiss " Felsite " Diorite " Sandstone 5.2.2 Analysis The data was modelled in SPSSV. 10 using single and stepwiseforward multiple- regressionshaving 0+ salmonas the dependentvariable. 89 5.3 Results Regressingthe catchof 0+ salmonfry from the 5 minute electrofishingsamples using all sites including thosewhich had no 0+ salmonpresent, on local site specific data (Table 5.1) in a forward stepwisemultiple regressionshowed that maximum water depth at the site was the best predictor of 0+ fry number,deeper water having lower 0+ abundance(Fi, 139=4.91; r2---0.027; p--0.028). The model was significantly improved however, by the addition of the width at the site (F2,138=4.65;r2=0.05; p---0.011)and the percentageof bedrock present(F3,137=4.317; r2---0.066; p<0.006) (Table5.3). Table 5.3 Stepwise regression of local site specific stream characteristics on G+ salmon numbers for a) all sites (including sites with no 0+ salmon); b) only sites whi-rP ftA.cnlmnn were nrerzent. a) F df rz p< Maximum water depth at site 4.91 1,139 0.027 0.03 Maximum water depth at site Width 4.65 2,138 0.050 0.01 Maximum water depth at site (-); Width (+); 4.32 3,137 0.066 0.006 Bedrock b) Bedrock 3.844 198 0.028 0.053 Bedrock(-); Average water depth 3.821 12,97 0.054 0.025 Increasingriver width had a positive effect on numbersof 0+ salmonpresent but the percentageof bedrockat the site was negativelyrelated to salmonnumbers. When only sites with 0+ salmonpresent were included it was found that the percentageof bedrockand averagedepth providedthe best model (F2,97=3.82;r2---0.054; p=0.025), both of thesewere negatively related to increasingsalmon numbers. Local instrearnhabitat characteristicswere highly significant predictorsof between site variation in salmonnumbers, however the percentageof variation explainedwas low (6.6% for all sites;5.4% for sites without 0+ salmon). To test if broaderscale 90 catchmentcharacteristics were affecting salmon abundance,wider catchmentscale variables(Table 5.2) were used as predictors against 0+ salmon numbers.Firstly, thesewere examinedagainst all of the available sites including those which did not have 0+ salmon. The percentageof glacial sandand gravel in the catchmentabove the site was found to be the best predictor (FI,73ý--11.83; rý--0.128; p<0.001), the addition of felsite (F2,72--9.63;r2=0.189; p<0.001) and altitude at the site greatly improvedits predictive ability (F3,71=8.421;r2=0.231; p<0.001) (Table 5.4a). When only sites where salmon were found were included the size of the catchmentarea abovethe site wasthe bestpredictor (FI, 48=10.426; r2---0.164; p<0.002), the addition of the percentageof sandstone(F2,47=7.846; r2--0.218; p<0.001) and the distanceof the site from the source of the river improved the predictive capacity of the model (F3,46=7.045;r2=0.270; p<0.001) (Table 5Ab). Table 5A Stepwiseregression of catchmentcharacteristics on 0+ salmonnumbers for a) all sites (including sites with no 0+ salmon);b) only sites where 0+ salmonwere present. a) F df r' P< Glacial sand & gravel 11.83 1,73 0.128 0.001 Glacial sand & gravel Felsite 9.63 2,72 0.189 0.001 Glacial sand & gravel Felsite Altitude 8.42 3,71 0.231 0.001 b) I I Catchment area 10.43 1,48 0.164 0.002 Catchment area Sandstone 7.85 2,47 0.218 0.001 Catchment area Sandstone (-); Distance from 7.05 3,46 0.27 0.001 mainstern Foyle I I I Both of thesespatial scaleswere then combinedin a single analysisto determineif catchmentfeatures were more importantthan site specific instrearncharacteristics. It was found that when all the sites(including thosewith salmonabsent) were included glacial sandand gravel,percentage of felsite, altitude, minimum depthof water at the site and stream width provided the best predictive model (F5,69=7.347;r2=0.300; 91 p<0.001) (Table 5.5a). Of these, altitude and the minimum depth of water were negatively related while the rest were positive. When the catchment scale characteristicsand instrearnvariables were included for only those sites with 0+ salmonpresent, the bestpredictors were found to be catchmentarea above the site, the maximumdepth of water found at the site, the numberof houseskrn-2 upstream, the amount of urban area upstream,the percentageof boulder clay upstreamand the percentageof sandstonein the catchment above the site (F6,43=7.190;r2=0.43 1; p<0.001)(Table 5.5b). Maximum depthof water, the amountof urban areaupstream and the percentageof sandstonewere negativelyrelated to 0+ numberswhile the rest of the variableswere positive. Table 5.5 Stepwiseregression of local scaleand catchmentcharacteristics combined on 0+ salmonnumbers for a) all sites (including sites with no 0+ salmon);b) only siteswhere 0+ salmonwere present. Equation 1: Y= (Glacial sand/gravel*267.777)+ (Felsite*810.421)+ (Altitude*- 0-0676)+ (Minimum waterdeDth at site*-0.300)+ (Width at site*0.543)-0.0439 a) F df r' P< Glacial sand & gravel 11.83 1,73 0.128 0.001 Glacial sand & gravel Felsite 9.63 2,72 0.189 0.001 Glacial sand & gravel Felsite Altitude 8.42 3,71 0.231 0.001 Glacial sand & gravel Felsite Altitude 7.60 4,70 0.263 I 0.001 Minimum water depth (-) I Glacial sand & gravel Felsite, Altitude 7.35 I 5,9 0.300 10.001 Minimum water deptli (-); Stream width (+) 92 Equation2: Y= (Catchmentarea*0.124) + (Maximum water depth at site*-0.331)+ No. houseskrd I *2.182) + (% urban area upstream*-1197.368)+ (Glacial boulder clav*252.548)+ (Sandstone*-2099.587)-5.801 b) F df r' p< Catchment area 10.43 1,48 0.161 0.002 Catchment area Maximum water depth 8.54 2,47 0.235 0.001 Catchment area Maximum water depth No. 7.31 3,46 0.279 0.001 2(+) houseskm 1 1 Catchment area Maximum water depth No. 6.92 4,45 0.326 0.001 houseskrný (+); Urban area upstream (-) Catchment area (+); Maximum water depth No. 7.04 5,44 0.381 0.001 houses krlf2 Urban area upstream (-); Glacial boulder clay Catchment area (+); Maximum water depth (-); No. 7.19 6,43 0.431 0.00 houses kM, 2 (+); Urban area upstream (-); Glacial boulder clay (+); Sandstone(-) In order to test the predictive ability of the combined local and catchmentwide characteristicsthe seconddata set of electrofishingsites was used. Salmonnumbers predictedfrom the sites not used in model development(equation 1: using all sites) significantlypredicted actual numbers caught, although with a low degreeof variance explained (FI,80=4.849; r2=0.045; p---0,03 1). Using equation 2 i.e. only those sites where 0+ salmon were present, also significantly predicted the actual numbers of fish caught,although again a very small amountof variancewas explained(Fl, go= 7.51; ? ---0.045;p=0.007). In order to investigate this further, paired Wests were used, it was found that using salmonnumbers predicted by equationI did not differ significantly from actual number (t=-0.609; 8ldf; p---0.544). The mean error, the differential betweenactual and predictedfish being 84% ±13S.E. Similarly, using the model developedfor only siteswhere salmon were detected(equation 2) actualfish numbers did not differ significantly from predictednumbers (t=0.354,91df; p=0.724). The meaneffor rate being 56%±15.23S. E. 93 5.4 Discussion Local habitat structure is known to influence the density of juvenile salmonids (Kennedy & Strange, 1982,1986a, 1986b; Elliott, 1994; Kocik & Ferreri, 1998). However,wider catchment,characteristics may be importantfactors in explainingboth temporaland spatialsalmonid variability (Milner, et aL 1998;Pess et aL, 2002). The model developedin this study proposessome novel additions to previous models, which are shownto be significantin explaining0+ salmonnumbers. Two variantsof the model were developed,variant I included local instreamand catchmentscale characteristicsand all of the available sites (including those where 0+ salmonwere absent) while the second variant (including local and broadscalecharacteristics) included only sites at which 0+ salmon were present. It was found that this combination of broader scale catchmentcharacteristics (glacial sand and gravel, felsite and altitude) and instreamvariables (minimum water depth and river width) explained30% of the observedvariation in variant 1. In this model catchmentwide features clearly explained more of the variation in salmon numbers than local instreamfeatures. In the secondmodel variant, differing broadscalecharacteristics were also found to be important(catchment area above the site, the numberof houses km-2upstream, the amount of urban area upstream,the percentageof boulder clay upstreamand the percentageof sandstonein the catchment)and also different site specificdata (maximum depth of waterfound at the site). This variant of the modelas might be expectedalso explained a higher proportion of the explainablevariance (r2---45%) The relationshipsin both of these models exposea strong relationship between broadscalecatchment wide featuresand 0+ abundanceand are consistentwith current 94 understandingof how theseunderlying physical attributescan influence fish habitat potential. Geology and geomorphicprocesses dictate the range of morphological characteristicsa streamreach can exhibit, thus partially determiningthe physicaland biological characteristicsof fish habitat.In model I which includedall sites,the area upstreamof glacial sand and gravel had a positive effect on salmon abundance. Juvenilesalmon generally inhabit areasof shallowfast flowing water with a substrate, which is gravel and cobble dominated4in addition, adult salmon require gravel depositsfor spawning(Jones, 1956; Shearer, 1992). Thereforeit is likely that greater areasof sandand gravel depositswill contribute to providing such suitablehabitat. The secondvariable, felsite, is an igneousrock. Hicks and Hall (2003) found that in the presenceof anotherigneous rock, basalt, production of juvenile salmonidswas positively relatedto its abundance,and they found that streamswhich flowed through areas of this rock type had generally steepergradients than those which flowed throughsedimentary sandstone rocks and that riffle habitatpredominated which is the favouredhabitat forjuvenile salmonids.Salmon abundance decreased with increasing altitude,higher reaches of streamsare generally regarded as being lessproductive as a result of lower temperatureswhich are known to adverselyaffect growth rates and densitiesof salmonids(Alabaster, and Lloyd, 1987;Elliott et aL, 1998). The first of the two local habitat features included was minimum water depth which had a negative effect on salmonid abundance. Symons and Heland (1979) found in laboratoryexperiments that 0+ salmonpreferred depths in the range6-9cm deep. It is postulatedthat, as minimum depth at the surveyedsite decreased,it was dropping below this preferredrange for 0+ fish, and thereforedensities decreased as a result. The secondand final local habitat feature was river width which was positively related to 0+ abundance.This variable is possibly related to fry dispersal post 95 hatchingasadults tend to spawnin the upper reacheswith fry dispersingdownstream (Shearer,1992). In model variant 2, broaderscale catchment characteristics were more importantthan site specific variables in explaining variation. Larger catchmentshad a positive influenceon salmonabundance, as wasfound by Milner et al. (1995). Interestinglyas the numberof housesincreased salmon numbersalso increasedalthough increasing urban areashad a negativeaffect. This suggeststhat up to a certain level increasing housenumbers may improve productivity, possibly through increaseof nutrientsbut above this threshold they may become detrimental to stocks, with increasing proportionsof urban areasin catchmentshaving a negativeeffect on fish abundance. Sandstone,being a sedimentaryrock is the corollary of felsite, and as notedby Hicks and Hall (2003) showed negative relationshipsbetween salmonid abundanceand increasingpresence of of this rock on rivers in north America. They found that rivers which flowed through sandstonewere generally low gradient with long reachesof deepslow flowing water,areas such as theseare not favourableto 0+ Atlantic salmon production. The only site specific variablewhich was included in the secondmodel variant was maximum water depth. Kennedy and Strange (1985) also found a negativerelationship between increasing water depthand salmonidabundance in their study on streamsin N Ireland, with preferred0+ salmonhabitat in water generally lessthan 20cm in depth. The results show that there is no significant statisticaldifference between the actual abundanceof salmonand those predictedby the models,but the averagedeviation betweenactual and predictedcatches was 84 and 56% respectively. Thus the second 96 variant of the model was the more robust of the two. This is to be expectedas only siteswith 0+ salmonwere included. The relatively low predictivecapacity and high error rate may in part be attributableto the choice of site. The sites chosenwere identified specifically as good quality sites for 0+ salmon and so are likely to be similar within a narrow range of attributes, thus eliminating large variation. The predictivecapacity would in all likelihood be improvedthrough the selectionof sites which arenot as suitedto 0+ salmonproduction. In summary, it is shown in this study that habitat on two scales, site specific and catchment wide, affects the juvenile density of salmon in the River Foyle catchment. Instream habitat characteristics are shown to be significant predictors of juvenile abundance, but with the inclusion of broader catchment scale characteristics the predictive power of the model is greatly increased. Increasing urbanisation is highlightedas a potentialfuture threatto the salmonof the Foyle area. 97 Chapter 6. General Discussion The Atlantic salmon is in decline in many parts of its range. However, the Foyle population, having declined from a period of high catchesin the 1960's and 1970's (Figure4.1), remainsrelatively stableat a lower level of productionwhich is still high in comparisonwith manyother systems(Anon., 2001). In periods when abundanceis high, conflicts between users (sport anglers and commercialharvesters) of the resourcemay diminish. However, during periodsof low abundancethis is liable to change. Theseconflicts generallyarise out of concernover resourceallocation and issuesof conservation. Chapter2 illustrated,using a long-time seriesof data the impactsof each of the user groups on each other and on successful spawningescapement. A 49 year time seriesdataset comprising commercial net and trap catches,recreational angling returnsand an independentmeasure of breedingpopulation size for the Atlantic salmonin the Foyle catchmentwas usedto investigatethe impactof competing methods of exploitation on each other and the effect of exploitation on populationsize. Positive correlationsbetween commercial netting and trapping catches over the 49 years of this study did not support the hypothesis that commercial interceptionfisheries impactednegatively on each other. In order to test the hypothesis that large returning migrant population sizes mask underlying negative relationships betweenexploitation, or betweenexploitation and spawningpopulation size, years with smaller than average returns were examined separately. No significant negative correlations between catchesfrom differing forms of exploitation nor between total exploitationand spawningpopulation size were detectedin this subsetof the data. With 98 regardto sport angling catchesit was shownthat a reductionin commercialexploitation may result in an increasein the sport fishery catch. However with only 19% variance explainedby commercialcatches, this effect was weak. Positivecorrelative relationships betweenall forms of exploitation combinedand an independentmeasure of spawning populationsize did not supportthe hypothesisthat exploitationhad a significantnegative impactupon ultimatespawning population size between 1952 and 2000. It was concludedthat the Atlantic salmon population supportedby this catchmentis sufficiently large to mask any potential for negative impacts of exploitation on sustainabilityof stocks.In addition, althoughthere is somesmall statisticallydetectable impact from commercial harvesting on the sport fishery, the different modes of commercial exploitation do not significantly impact detrimentally upon each other. Therefore, in the Foyle area, cessationof commercial harvesting should be largely regardedas an issueof resourceallocation rather than conservation. In Chapter3 using the samelong-term data set, the role of density dependentand life- stagespecific environmentalfactors in controlling population size was investigated.A Ricker density-dependentmodel showedthat spawningadult populationsize significantly explained variation in the resultantfilial generation,however a significant amount of variation (ca. 68%) remainedunexplained. It was shownthat environmentalfactors were significant in explaining someof the remainingvariance and that theseinfluences were linked to specific life stages. This finding strongly suggestedpopulation bottlenecks in the complex life cycle of this species,during which, specific environmentaleffects may have had an impact they did not have during other periods.It was concludedthat these life stage specific environmentaleffects were likely to contribute to the stochastic 99 variation in population size remaining after the application of traditional stock- recruitmentmodels and that the identification and quantificationof theseeffects should allow improvedmodel accuracy. Chapter3 highlightedthe importanceof variousenvironmental parameters on salmonand in Chapter4 this was investigatedfurther by the testing of the effect of marineclimatic conditionsin the North Atlantic on the abundanceof returning migrant Atlantic salmon, using a 126 year datasetof commercialcatches. ' Catches of salmon from commercial netting stations significantly predicted a measureof population size independentof catchesover a 49 year period; hence commercialnet catcheswere assumedto be an adequatemeasure of relative population size. The North Atlantic Oscillation index in winter (NAOI) provides a generalisedmeasure of climate variability for the northern hemisphere. Between 1875 and 2001, the NAOI explaineda significant proportion of variation in five year running meancatches of migrant Atlantic salmonreturning to the River Foyle. When the index was below 0.151, the NAOI correlatednegatively with salmon catches 70%) indicating that a significant proportion of the variance in population size in the past was the result of variability in conditions in the marine environment.However, when the NAOI was above 0.151, this relationshipuncoupled. The probability of catchesexceeding the long-term median, was 2.34 times lower in yearswhere the NAOI was abovethe 0.151breakpoint than for yearswhen it was below 0.151. The NAOI had exceededthe 0.151 thresholdon 8 out of the 10 years prior to 2001. Models of climatechange indicate that the NAOI is likely to increasesignificantly with time. If these models are correct, this study would lead to the conclusionthat a decouplingof broad scale climate effects on salmon population size will becomethe norm. Data presentedhere suggest two consequencesof this: that the value of the NAOI 100 as a predictivetool for forecastingadult salmonpopulation size will be limited; and that the median population size will become lower in the future. Chapter5 testedthe capacityof local instreamand broadscalecatchment characteristics to predict 0+ salmonabundance within the Foyle area.Data was collectedfrom 307 and 279 sites for annual electrofishing surveys from 2001 and 2002 respectivelyof 0+ salmonidsand semi-quantitativestream morphometric information. Using a combination of these local site-specificvariables and broadscalecatchment characteristics derived from Ordnanceand Geological survey maps, two models were constructed. Model I useddata for all availablesites i.e. including thosewhere 0+ salmonwere absent,while model 2 wasdeveloped using only thosesites where salmon where present. Both of these models were significant predictors of juvenile abundanceusing site-specific variables althoughthe varianceexplained was small. However, both were improved significantly ? by the inclusion of wider broadscalecatchment characteristics (model 1 =30%; model 2 x2=43%).When testedagainst an independentdata set using pairedt-tests, neither model differed significantly from the actualcatch of salmon,although both had relatively high standarderror rates (84%±13 and 56%±15.2respectively). It was concludedthat site- specifichabitat characteristics were significantpredictors of juvenile abundance,but with the inclusionof broadscalecatchment characteristics the models' predictivepowers were greatly increased. Although the Foyle population of salmonhas declined from the 1960's and 1970's in commonwith many other systemsaround the world, if the longer term catchesare taken into considerationthis period of lower level abundancemay just be part of a longer term 101 natural cycle. If the period from the mid 1960's is taken as an example (Figure 6.1) it may be seenthat the population appearsto be dramatically declining. Figure 6.1 Foyle area annual commercial salmon catches 1962-2003 with regression line. 160,000 - 140,000- y= -2578x + 115209 120,000- R2=0.6468 r- 100,000- 0 E 80,000 - C 60,000- 40,000- 20,WO - 0 -, IýIIIIIIýIýIIIIIIIIIIIýIIIIII-, IýIIII do ýý KI, Aý # -x'o ,pI FJK -lq Ile -lq Nq -IR,-IC', NOI -lq NOI-lq NOI-l0j ":b -1010 CNNoj ,q-, oj Year Most of the intensive research work on Atlantic salmon commenced during this period. However, if the longer term picture is looked at (Figure 4.1) it may be seen that the catches during the 1960's coincide with the highest recorded abundance in the long-term datasetand the period we are in now is almost like a background level of abundance.This poses the question if there were similar long-term datasets for other stocks of salmon would they show a similar picture? Fluctuations such as these have been observed in many different animal populations as previously seen with lemmings, voles (Ekerholm et al., 2001) and red grouse (Jenkins et aL, 1963; Newton, 1998). Many of these cycles can be related to natural phenomena, 102 which operateindependently of direct anthropogenicinfluence, as for examplechanges in the distribution and abundanceof the Pacific sardineand northernanchovy (Rodrfguez- SAnchezet aL, 2002). The resultsof this study have thereforewider implicationsthan just for the managementof salmonand it is hopedthat it will contributeto the debateon the managementof commerically exploited speciesand the relative roles of density driven populationmechanisms and environmental factors. Specifically with regard to the Foyle salmon this study has highlighted a number of issueswhich may impact on the return of the populationto the higher levels previously experienced.Of particularconcern are thosewhich may not have beenan issueduring previous periods of lower abundance,namely direct and indirect anthropogenic influences. Furtherwork is requiredto understandhow theseinfluences interact to affect the Atlantic salmon in particular further developmenton the combination of stock recruitmentand environmentalmodels where availabledata permit. It is hoped that this thesis will form the basis of future study, in particular the developmentof pre-fishery abundancemodels which can be linked by genetic analysisto specific stock structures within the Foyle catchment. 103 References Achord, S. Levin, P.S. & Zabel, R.W. (2003). Density-dependentmortality in Pacific salmon:the ghostof impactspast? Ecology Letters 6,335-342. Alabasterand Lloyd (1987).Water Quality Criteria for FreshwaterFish. London: Food and Agriculture Organisation. Alisauskas,R. T. and Arnold, T.W. (1994). American Coot. Migratory shoreand upland game bird managementin North America (T.C. Tacha & C.E. Bmun eds.). Kansas:Allen Press.pp. 127-143. Amiro, P.G., (1998a).Recruitment of the North Americanstock of Atlantic salmon(Salmo salar) relative to annual indices of smolt production and winter habitat in the northwest Atlantic. Canadian Stock Assessment Secretariat Research Document 98/45. Amiro, P.G., (1998b). The abundanceof Harp sealsin the North Atlantic and recruitment of the North American stock of Atlantic salmon (Salmo salar). Canadianstock AssessmentSecretariat Research Document 98/84. Anon. (1953). Foyle Fisheries Annual Report for 1953, Londonderry: Foyle Fishery Commission. 104 Anon. (1969). Foyle Fisheries Annual Report for 1953, Londonderry: Foyle Fishery Commission. Anon. (1970). Foyle Fisheries Annual Report for 1953, Londonderry: Foyle Fishery Commission. Anon. (1971). Foyle Fisheries Annual Report for 1953, Londonderry: Foyle Fishery Commission. Anon. (1992a). Directive on the conservationof naturaland semi-naturalhabitats and wild faunaand flora. HabitatsDirective 92/43EEC. Anon. (1992b). Salmon ResearchAgency of Ireland Annual Report 1992, Newport: SalmonResearch Institute of Ireland,Farran Laboratory. Anon. (1996a). Foyle Fisheries Annual Report for 1996, Londonderry: Foyle Fishery Commission. Anon. (1996b). Report to the Minister of the Salmon Task Force, Dublin: Government PublicationsOffice. Anon. (1997).Habitat SurveyClassification. Departmentof Agriculture for N. Ireland. Anon. (1998a). SalmonTask Force, Implementation Document. Dublin: Departmentof the Marine andNatural Resources. 105 Anon. (1998b). Yhe Spey District Fishery Board Management Report 1998 and Policy Review.Scotland. Spey District FisheryBoard. Anon. (1999). Glas-Y-Dorlan 25, Environment Agency Wales, Spring. Anon. (2001). Report of the working group on north Atlantic salmon, ICES CM 2001/ACFM:15 Part one,Aberdeen, Scotland, 2-11 April 2001. Anon. (2003a). 77zeFoyle Fisheries Commission Annual Report and Financial Statements, 1997-1999.Belfast: TSO. Anon. (2003b). Economic evaluation of sport angling versus commercial netting. Dublin: CentralFisheries Board. Anon. (2003c). Report of the 2dh Annual Meeting of the Council of NASCO. Edinburgh, Scotland. Azurnaya, T. & Ishida, Y. (2004). An evaluation of the potential influence of SST and currents on the oceanic migration of juvenile and immature chum salmon (Oncorhynchusketa) by a simulationmodel. FisheriesOceanography 13,10-23. Beaugrand, G. & Reid, P.C. (2003). Long-term changesin phytoplankton,zooplankton and salmonrelated to climate.Global ChangeBiology 9,1-17. 106 Beverton,R. J. H., and Holt, S.J. (1957). On the dynamicsof exploited fish populations. FisheryInvestigations, London Series 2 19,1-533. Bosch, M., Oro, D., Cantos,F. J. and Zabala,M. (2000). Short-termeffects of culling on the ecology and population dynamics of the yellow-legged gull. Joumal ofApplied Ecology37,369-385. Buck, R.J. and Hay, D.W., (1984). The relation between stock size and progeny of Altantic salmon, Salmo salar L., in a Scottish stream. Joumal of Fish Biology 24, 1-11. Byrne, C.J., Poole, R., Rogan, G., Dillane, M. & Whelan, K.F. (2003). Temporal and environmentalinfluences on the variation in Atlantic salmonsmolt migration in the B urrishoolesystem 1970-2000. Joumal of Fish Biology 63,1552-1564. Cass,A. J., andWood, C.C. (1994). Evaluationof the depensatoryfishing hypothesisas an explanationfor populationcycles in FraserRiver sockeyesalmon (Oncorhynchus nerka). CanadianJournal of Fisheriesand Aquatic Science51,1839-1854. Chadwick, E.M. P. (1982). Stock-recruitmentrelationships for Atlantic salmon (Salmo salar) in Newfoundland rivers. Canadian Journal of Fisheries and Aquatic Sciences39,1496-150 1. 107 Chadwick,E. M. P. (1985a). The influenceof spawningstock on productionand yield of Atlantic salmon in Canadianrivers. Aquaculture and Fisheries Management1, 111-119. Chadwick, E.M. P. (1985b). Fundamentalresearch problems in the managementof Atlantic salmon (Salmo salar) in Atlantic Canada. Joumal of Fish Biology 27 (Suppl. A), 9-25. Chapman,D. W., (1986). Salmonand Steelheadabundance in the Columbia River in the nineteenthcentury. Transactions of theAmerican Fisheries Society 115,662-670. Chase,S. (2003). Closing the North American mixed-stockcommercial fishery for wild Atlantic salmon(Salmo salar L. ). In Salmonat the Edge(Mills, D. ed), pp. 84-92, Oxford, Blackwell Publishing. Clark, C.W. (1973). The Economicsof Overexploitation.Science, 181,630-634. Claytor, R. R. (2000). Conflict resolution in fisheries management using decision rules:an exampleusing a mixed-stockAtlantic Canadianherring fishery. ICES Journal of Marine Science 57,1110-1127. Crozier, W.W. and Kennedy, G.J. A. (1994a). Application of semi-quantitative electrofishingto juvenile salmonidstock surveys.Joumal of Fish Biology 45,159- 164. 108 Crozier, W.W., KennedyG. J. A. (1994b). Marine exploitation of Atlantic salmon(Salmo salar L. ) from the River Bush,Northern Ireland. FisheriesResearch 19,141-155. Crozier, W.W., Kennedy,G. J. A., Boylan, P. & Kennedy,R. (2003). Summaryof salmon fisheriesand statusof stocksin Northern ireland for 2002. ICES Working Group Paper no. Dempson, J.B., Schwarz, C.J., Reddin, D.G., O'Connell, M. F., Mullins, C.C. and Bourgeois, C.E. (2001). Estimation of marine exploitation rates on Atlantic salmon (Salmosalar L. ) stocksin Newfoundland,Canada. ICES Journal ofMarine Science 58,331-341. Dickson, R.R. and Turrell, W.R. (2000). The NAO: The Dominant AtmosphericProcess Affecting OceanicVariability in Home, Middle and Distant Waters of European Atlantic Salmon, In 77zeOcean Life of Atlantic salmon: Environmental and biological factors influencing survival. (Mills, D., ed.), pp. 92-115. London: FishingNews Books. Drinkwater, K. F. (2000). Changes in Ocean Climate and its General Effect on Fisheries: Examplesfrom the North-west Atlantic, In The Ocean Life of Atlantic salmon: Environmentaland biologicalfactors influencingsurvival. (Mills, D., ed.), pp. 116- 136. London:Fishing NewsBooks. 109 Dumas, J. & Prouzet,P. (2003). Variability of demographicparameters and population dynamicsof Atlantic salmon(Salmo salar L. ) in a southwestFrench river. ICES Journal ofMarine Science60,356-370. Egglishaw,H. J., and Shackley,P. E. (1977). Growth, survival and productionof juvenile salmonin a Scottishstream, 1966-1975. Joumal of Fish Biology 11,647-672. Egglishaw,H. J., and Shackley,P. E. (1985). Factorsgoverning the productionof juvenile Atlantic salmon in Scottish streams. Joumal of Fish Biology 27,27-33. Ekerholm, P., Oksanen,L. & Oksanen,T. (2001). Long-term dynamics of voles and lemmingsat the timberline and above the willow limit as a test of hypothesison trophic interactions.Ecography 24,555-568. Einarsson, S.M., Gudbergsson,G. (2003). The effects of the net fishery closure on angling catch in the River Hvita, Iceland. Fisheries Management and Ecology 10,73-78. Einum, S., Fleming, I.A., Cote, I. M. & Reynolds,J. D. (2003). Population stability in salmon species:effects of population size and female reproductive allocation. Joumal ofAnimal Ecology 72,811-82 1. Elliott, J.M. (1985). Population dynamicsof migratory trout, (Salmo truua) in a Lake District stream,1966-83, and their implicationsfor fisheriesmanagement. Joumal of Fish Biology 27,3543. 110 Elliott, J.M. (1989). The natural regulation of numbers and growth in contrasting populationsof brown trout, Salmotrutta, in two Lake District streams.Freshwater Biology 21,7-19. Elliott, J.M. (1994).Quantitative Ecology and the Brovm Trout. Oxford Seriesin Ecology andEvolution, Oxford: Oxford University Press. Elliott, J.M., Hurley, M. A. & Elliott, J.A. (1997). Variable effects of droughts on the density of a sea-trout population over 30 years. Joumal of Applied Ecology 34, 129-138. Elliott, J.M. and Hurley, M. A. (1997). A functional model for maximum growth of Atlantic salmonparr, Salmosalar L., from two populations in northwestEngland. Functional Ecology 11,592-603. Elliott, J.M. & Hurley, M. A. (1998). An individual-basedmodel for predicting the emergence period of sea-trout fry in a Lake District stream. Joumal of Fish Biology 53,414-433. Elliott, J.M. (2001). The relative role of density in the stock-recruitment relationship of salmonids. In Stock, recruitment and reference points. assess7nentand managementof Atlantic salmon,( Prevost,E. & Chaput,G. ed.), pp. 25-66, Paris: III Elliott, J.M., Hurley, M.A. & Maberly, S.C. (2000). The emergenceof seatrout fry in a Lake District streamcorrelates with the North Atlantic Oscillation.Joumal of Fish Biology 56,208-210. Elliott, S.R., Coe, T.A., Helfield, J.M. & Naiman, R.J. (1998). Spatial variation in environmentalcharacteristics of Atlantic salmon(Salmo salar) rivers. Canadian Journal of Fisheriesand Aquatic Sciences55 (Suppl. 1), 267-280. Elson, P.F. & Tuomi, A. L. W. (1975). 77zeFoyle Fisheries, A New Basis for Rational Management.Londonderry: Foyle FisheriesCommission. Frederiksen,M., Lebreton,J. -D. and Bregnballe,T. (2001) The interplay betweenculling and density-dependencein the great cormorant:a modelling approach. Journal of Applied Ecology 38,617-627. Friedland, K.D. (1998). Oceanclimate influenceson critical Atlantic salmon (Salmo salar) life history events. Canadian Journal of Fisheries and Aquatic Sciences 55 (Suppl. 1), 119-130. Friedland, K. D., HansenL. P., Dunkley, D. A. & MacLean, J.C. (2000). Linkage between ocean climate, post-smolt growth, and survival of Atlantic salmon (Salmosalar L. ) in the North Sea area. ICES Journal of Marine Science57, 419-429. 112 Friedland,K. D., Reddin,D. G. & Castonguay,M. (2003a). Oceanthermal conditions in the post-smoltnursery of North American Atlantic salmon. ICES Joumal of Marine Science60,343-355. Friedland, K.D., Reddin, D. G., McMenemy, J.R. & Drinkwater, K.F. (2003b). Multidecadaltrends in North American Atlantic salmon(Salmo salar) stocksand climate trends relevant to juvenile survival. CanadianJournal of Fisheries and Aquatic Sciences 60,563-583. Forchhammer,M. C., Post,E. & Stenseth,N. C. (2002). North Atlantic Oscillation timing of long- and short-distance migration. Joumal ofAnimal Ecology71,1002-1014. Gardiner, R., and Shackley, (1991). Stock and recruitment and inversely density dependentgrowth of salmon,in a Scottishstream. Joumal of Fish Biology 38,69 1- 696. Gee,A. S., Milner, N.J. and Hernsworth,R. J. (1978). The effect of densityon mortality in juvenile Atlantic salmon (Salmo salar). Joumal ofAnimal Ecology 47,497-505. Gillett, N.P., Graf, H.F. & Osborne,T. J. (2003). Climate changeand the North Atlantic Oscillation, In 77zeNorth Atlantic Oscillation: Climate Significance and EnvironmentalImpact. GeophysicalMonograph Series (Hun-ell, J. W., Kushnir, Y., Ottersen,G. andVisbeck, M., eds.) 134,193-209. 113 Gillooly, J.F., O'Keefe,T. C., Newman,S. P. and Baylis, J.R. (2000). A long-termview of density-dependentrecruitment in smallmouthbass from Nebish Lake, Wisconsin. Joumal of Fish Biology. 56,542-55 1. Grant, J.W. A., Steingrimsson,S. O., Keeley,E. R. & Cunjak, R.A. (1998). Implicationsof territory size for the measurementand prediction of salmonid abundancein streams.Canadian Joumal of FisheriesAquatic Science55 (Suppl. 1), 181-190. Haldane,J. B. S. (1953). Animal populationsand their regulation.New Biology 15,9-24. Haldane,J. B. S. (1956). The relation betweendensity regulation and natural selection. Proceedings of the Royal Society B 145,306-308. Hansen,L. P. & Quinn, T.P. (1998). The marine phaseof the Atlantic salmon (Salmo salar) life cycle, with comparisons to Pacific salmon. Canadian Journal of Fisheries and Aquatic Sciences55 (Suppl. 1), 104-118. Harris, M.P. (1970). Territory limiting the size of the breeding population of the Oystercatcher (Haematopus ostralegus) -a removal experiment. Journal ofAnimal Ecology 39,707-713. Harwood, A. J., Annstrong, J.D., Metcalfe, N.B. & Griffiths, S.W. (2003). Does dominancestatus correlate with growth in wild stream-dwellingAtlantic salmon (Salmosalar)?. BehavioralEcology 14,902-908. 114 Hassell, M.P. & May, R.M. (1990). Population regulation and dynamics.Philosophical Transactions of the Royal Society of London 330,121-304. Hay, D.W. (1984). The relationshipbetween redd counts and the numbersof spawning salmonin the Gimock Bum (Scotland), InternationalCouncil for the Exploration of the Sea,C. W. 1984/M:22, Anadromousand CatadromousFish Committee. Hay, D.W. (1987). The relationshipbetween redd counts and the numbersof spawning salmonin the Gimock Bum, Scotland.Journal du Conseil43,146-148. Heggberget,T. G. (1988). Timing of spawning in Norwegian salmon (Salmo salar L. ). CanadianJoumal of Fisheriesand Aquatic Science45,845-849. Heggenes,J., and Saltveit, S.J. (1990). Seasonaland spatial microhabitat*selection and segregation in young Atlantic salmon and brown trout in a Norwegian river. Joumalof Fish Biology 36,707-720. Hicks, B.J. (2003).Rock type and channelgradient structure salmonid populations in the Oregon Coast Range. Transactions of the American Fisheries Society 132,468- 482. Hilborn, R. (1997). Comment:Recruitment paradigms for fish stocks. CanadianJournal of Fisheriesand Aquatic Sciences54,984-985. 115 Hill, M. O. (1979). TWINSPAN -A FORTRAN programfor arrangingmultivariate data in orderedtwo-way tablesby classificationo the individualsand attributes.Ecology and Systematics,New York: Cornell University. Hill, M. F., Botsford, L. W. & Hastings, A. (2003). The effects of spawning age distribution on salmon persistenceinfluctuating environments. 7he Journal ofAnimal Ecology 72,736-744. Hilton J., Welton J.S., ClarkeR. T. & Ladle M. (2001). An assessmentof the potentialfor the application of two simple models to Atlantic salmon, Salmo salar, stock managementin chalk rivers. Fisheries Management and Ecology, 8,189-205. Holm, M., Holst, J.C. & Hansen,L. P. (2000). Spatial and temporal distribution of post- smolts of Atlantic salmon (Salmosalar L. ) in the Norwegian Sea and adjacent areas. ICES Journal ofMarine Science 57,955-964. Horton, R.E. (1945). Erosionaldevelopment of streamsand their drainagebasins: a hydro- physical approach to quantitative morphology. Bulletin of the Geological Society ofAmerica 56,275-350. Hurrell, J.W., Kushnir, Y., Ottersen,G. & Visbeck, M. (2003).An overview of the North Atlantic Oscillation. 77teNorth Atlantic Oscillation: Climate Significance and Environmental Impact. Geophysical Monograph Series (eds. Hurrell, J.W., Kushnir, Y., Ottersen,G. andVisbeck, M. ) 134,1-35. 116 Hurrell, J.W. (1995). Decadal trends in the North Atlantic Oscillation: regional temperaturesand precipitation. Science 269,676-679. Hutchinson P. & Mills D. H. (2000). Executive Summary. In 71e Ocean Life of Atlantic salmon: Environmental and biological factors influencing survival. (D. Mills ed.) pp. 7-18, London:Fishing News Books. Isaak, D. J., Thurow, R.F., Rieman, B. E. & Dunham, J.B. (2003). Temporal variation in synchronyamong chinook salmon(Oncorhynchus tshawytscha) redd countsfrom a wildemess area in centml Idaho. Canadian Journal of Fisheries and Aquatic Science60,840-848. James,G. W. A. andDonald, K. L. (1990). Territory size as a predictorof the upper limit to population density of juvenile salmonids in streams. Canadian Journal of Fisheries and Aquatic Science47,1724-1737. Jenkins,D., Watson,A. & Miller, G.R. (1963).Population studies on Red GrouseLagopus lagopus scoticus (lath. ) in north-east Scotland. Joumal ofAnimal Ecology 32,317- 376. 117 Jensen,A. J., Zubchenko,A. V., Heggberget,T. G., Hvidsten, N. A., Johnsen,B. 0., Kuzmin, 0., Loenko,A. A., Lund, R. A., Martynov, V. G., Nxsje, T. F., Sharov,A. F., and Okland, F. (1999). Cessationof the Norwegiandrift net fishery: changes observedin Norwegianand Russianpopulations of Atlantic salmon.ICES Journal of Marine Science56,84-95. Jones,J. W. (1959).7he Salmon, London: Collins, St JamesPlace. Jones, M., Laurila, A., Peuhkuri, N., Piironen, J. & Seppa, T. (2002). Timing an ontogeneticniche shift: responsesof emerging salmon alevins to chemical cues from predatorsand competitors. Oikos 102,155-163. Johnston,P. (2002). Benefitsof nettingbuy-out. Trout & SalmonJune, 23-25. Jonsson, N., Jonsson,B. & Hansen, L. P. (2003). The marine survival and growth of wild and hatchery reared Atlantic salmon (Salmo salar L. ). Joumal ofApplied Ecology 40,900-911. KennedyG. J. A. (1983). Someobservations on salmonidecology in upland streams,Irish Fisheries Investigations Series A. 23. Kennedy, G.J. A. and Strange,C. D. (1982). The distribution of salmonids in upland streamsin relationto depthand gradient. Joumal of Fish Biology 20,579-591. 118 Kennedy, G.J. A. and Strange,C. D. (1986a). The effects of intra- and inter-specific competitionon the survival and growth of stockedjuvenile Atlantic salmon,Salmo salar L., and residenttrout, Salmotrutta L., in an upland stream. Joumal of Fish Biology 28,479-489. Kennedy, G.J. A. and Strange,C. D. (1986b). The effects of intra- and inter-specific competitionon the distributionof stockedjuvenile Atlantic salmon,Salmo salar L., in relation to depth and gradient, in an upland trout, SaInw trutta L., stream. Joumal of Fish Biology 29,199-214. Kennedy,G. J. A. & Crozier, W.W. (1993). Juvenile Atlantic salmon (Salmosalar L. ) - productionand prediction. In Productionof JuvenileAtlantic salmon(Salmo salar L) in natural waters. Canadian Special Publication of Fisheries and Aquatic Sciences118,179-187. (eds.Gibson, R. J. & Cutting, R.E. ). Kennedy, G.J. A. & Crozier, W.W. (1997). What is the value of a wild salmon smolt, (SalmosalarL. )? Fisheries Management and Ecology 4,103-110. Klemetsen,A., Amundsen,P. A., Dempson,J. B., Jonsson,B., Jonsson,N., O'Connell, M. F. & Mortensen,E. (2003). Atlantic salmon (Salmo salar L. ), brown trout (Salmotmtta L. ) and Arctic chaff (Salvelinusalpinus L. ): a review of aspectsof their life histories. Ecologyof FreshwaterFish 12,1-59. 119 Kocik, J.F. & Ferreri, C.P. (1998). Juvenileproduction variation in salmonids:population dynamics, habitat, and the role of spatial relationships. Canadian Journal of Fisheries and Aquatic Sciences55 (Suppl. 1), 191-200. Krebs, J.R. (1970). Regulation of numbers in the great tit (Aves: Passeriformes). Joumal of Zoology (U)ndon) 162,317-333. Lachin, J.M. (2000). Biostatistical Methods - The Assessmentof Relative Risks. New York: Wiley. Langeland, A. & Pedersen,T. (2000). A 27-year study of brown trout population dynamicsand exploitation in Lake Songsjoen,central Norway. Journal of Fish Biology 57,1227-1244. Laurian C.; Ouellet J-P.; Courtois R.; Breton L.; St-Onge S. (2000). Effects of intensive harvesting on moose reproduction. Joumal ofApplied Ecology, 37,515-53 1. Logerwell, E.A., Mantua, N., Lawson, P.W., Francis, R.C. & Agostini, V.N. (2003). Tracking environmental processesin the coastal zone for understandingand predicting Oregon coho (Oncorhynchus kisutch) marine survival. Fisheries Oceanography 12,554-568. Magee,J. A., Obedzinski,M., McConnick, S.D. & Kocik, JR (2003). Effects of episodic acidification on Atlantic salmon (Salmo salar) smolts. Canadian Journal of Fisheriesand Aquatic Sciences60,214-22 1. 120 Marschall,E. A. and Crowder,L. B. (1995). Density dependentsurvival as a function of size in juvenile salmonids in streams. Canadian Journal of Fisheries and Aquatic Science52,136-140. Martin, J.H. A. & Mitchell, K.A. (1985). Influence of seatemperature upon the numbers of grilse and multi-sea-winterAtlantic salmon(Salmo salar) caught in the vicinity of the River Dee (Aberdeenshire).Canadian Journal of Fisheries and Aquatic Science42,1513-152 1. Milner, N.J., Wyatt, R.J. and Broad, K. (1998). HABSCORE - applicationsand future developmentsof related habitat models. Aquatic Conservation: Marine and FreshwaterEcosystems 8,633-644. Milner, N.J., Elliott, J.M., Armstrong, J.D., Gardiner, R., Welton, J.S. and Ladle, M. (2003). The natural control of salmonand trout populationsin streams.Fisheries Research 62,111-125. McCain, M., Fuller, D., DeckerL., and OvertonK. (1990). StreamHabitat Classification and Inventory Proceduresfor Northern California. FHR Currents, R-5s Fish Habitat RelationshipsTechnical Bulletin, No. 1, U. S. Department of Agriculture, Forest Service, Pacific Southwest Region. 121 McCormick, S.D., Hansen, L.P., Quinn,T. P. & Saunders,R. L. (1998). Movement, migration and smolting of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences55 (Suppl. 1), 77-92. McFarlane, G.A., King, J.R. & Beamish, RJ. (2000). Have there been recent changes in climate? Ask the fish. Progress in Oceanography 47,147-169. McKinnell, S.M. & Karlstrom, 0. (1999). Spatial and temporal covariation in the recruitmentand abundanceof Atlantic salmonpopulations in the Baltic Sea. ICES Journal of Marine Science56,433-443. Mikio Inoue, Shigeru Nakano and Futoshi Nakamura (1997). Juvenile masu salmon (Oncorhynchusmasou) abundanceand stream habitat relationships in northern Japan.Canadian Journal of Fisheriesand Aquatic Science54,1331 - 1341. Milner, N.J., Wyatt, R.J., and Scott, M. D. (1993). The use of habitat models to interpret salmonid distribution and abundancein streams. In ProceedingsFSBI Annual Symposium,Conwy. Mueter, F.J., Peterman, R. M. & Pyper, B. J. (2002). Opposite effects of ocean temperature on survival ratesof 120 stocksof Pacific salmon(Oncorhynchus spp. ) in northern and southernareas. CanadianJournal of Fisheriesand Aquatic Sciences59,456- 463. 122 Myers, R.A., Barrowman,LA., Hutchings,J. A., and Rosenberg,A. A. (1995). Population dynamicsof exploitedfish stocksat low populationlevels. Science 269,1106-1108. Myers, R.A., Rosenberg,A. A., Mace,P. M., Barrowman,and Restrepo,V. R. (1994).ICES Journal ofMarine Science51,191-205. Newton,1. (1998). PopulationLimitation in Birds. London:Academic Press. O'Connell, M. F., Dempson, J.B. and Reddin, D. G. (1992). Evaluation of the impacts of major managementchanges in the Atlantic salmon (Salmo salar L. ) fisheries of Newfoundlandand Labrador,Canada, 1984-1988. ICES Journal of Marine Science 49,69-87 Olsen,P. D. and Olsen,J. (1989). Breedingof the Peregrinefalcon (Falco peregrinus):III. Weather, nest quality and breeding success.Emu 89,6-14. Parrish, D. L., Behnke, R. J., Gephard, S.R., McConnick, S.D. & Reeves, G.H. (1998). Why aren't there more Atlantic salmon (Salmo salar)? Canadian Journal of Fisheries and Aquatic Sciences55 (Suppl. 1), 281-287. Pess,G. R., Montgomery,D. R., Steel,E. A., Bilby, R.E., Feist, BE and Greenberg,H. M. (2002). Landscapecharacteristics, land use, and coho salmon (Onchorhynchus kisutch) abundance,Snohomish River, Washington,U. S.A.. CanadianJournal of Fisheriesand Aquatic Science59,613-623. 123 PetermanR. M., Pyper, B.J. & MacGregor,B. W. (2003). Use of the Kalman filter to reconstruct historical trends in productivity of Bristol Bay sockeye salmon (Oncorhynchusnerka). CanadianJournal of FisheriesScience 60,809-824. Pierce,G. J. & Boyle, P.R. (2003). Empirical modelling of interannualtrends in abundance of squid(Loligoforbesi) in Scottishwaters. FisheriesResearch 59,305-326. Poff, N.L. & Huryn, A.D. (1998). Multi-scale determinantsof secondaryproduction in Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences55 (Suppl. 1), 201-217. Punt,A. E., and Hilborn, R. (1997). Fisheriesstock assessmentand decisionanalysis: The Bayesianapproach. Reviews in Fish Biology and Fisheries7,35-63. QuizilbashM. (2001). SustainableDevelopment: Concepts and Rankings.7he Journal of Development Studies 37,134-16 1. Reddin D.G., Helbig, J., Thomas, A., Whitehouse, B.G. & Friedland, K. D. (2000). Survival of Atlantic Salmon(Salmo salar L. ) related to Marine Climate. In 7he Ocean Life of Atlantic salmon: Environmental and biological factors influencing survivaL(D. Mills ed.) pp. 88-91,London: Fishing NewsBooks. Reddin, D.J., Friedland, K.D., Downton, P., Dempson, B.J. & Mullins, C.C. (2004). Thermal habitat experiencedby Atlantic salmon (Salmosalar L. ) kelts in coastal Newfoundlandwaters. FisheriesOceanography 13,24-35. 124 Reid, P.C, & Planque, B. (2000). Long-term planktonic variations and the climate of the North Atlantic. In 7he Ocean Life of Atlantic salmon: Environmental and biologicalfactors influencingsurvivaL (D. Mills ed.) pp.153-169, London: Fishing News Books. Ricker, W.E. (1954). Stock and Recruitment.Journal FisheriesResearch Board Canada 11,559-623. Ritter, J.A. (1997). Atlantic Salmon Maritimes Region Overview. DFO ScienceStock Status Report, D3-14. Rodrfguez-SAnchez,R., Lluch-Belda, D., Villalobos, H. & Ortega-Garcia,S. (2002). Dynamic geographyof small pelagic fish populations in the California Current Systemon the regimetime scale(1931-1997). CanadianJournal of Fisheriesand Aquatic Science59,1980-1988. Romakkanierni, A., Perd, I., Karlsson, L., Jutila, E., Carlsson, U. & Pakarinen, T. (2003). Developmentof wild Atlantic salmonstocks in the rivers of the northernBaltic Sea in responseto managementmeasures. ICES Journal of Marine Science60,329- 342. Ross, M. R., and Almeida, F. P. (1986). Density-dependentgrowth of silver hakes. Transactionsof the AmericanFisheries Society 115,548-554. 125 Ruggerone,G. T. & Rogers,D. E. (2003). Multi-year effects of high densitiesof sockeye salmonspawners on juvenile salmongrowth and survival: a casestudy from Exxon Valdezoil spill. FisheriesResearch 63,379-392. Ruggerone,G. T., Zimmennann,M., Myers, K.W., Nielsen, J.L. & Rogers,D. E. (2003). Competitionbetween Asian pink salmon (Oncorhynchusgorbuscha) and Alaskan sockeyesalmon (0. nerka) in the North Pacific Ocean.Fisheries Oceanography 12, 209-219. Schnute,J. T., and Kronlund, A.R. (1996). A managementorientated approach to stock recruitmentanalysis. Canadian Joumal of Fisheriesand Aquatic Science53,128 1- 1293. Shearer,W. (1992). 7he Atlantic Salmon: Natural History, Exploitation and Future Management. London: Fishing News Books. Silveria, L., Rodrigues,F. H. G., Ja.A como, A. T.A. and Filho, J.A. F. D. (1999). Impact of wildfires on the megafaunaof Emas National Park, central Brazil. Oryx 33,108- 114. Sinclair, A. E.G. (1989). The regulation of animal populations. In Ecological concepts. British Ecological Society Symposium(Cherrett, M., ed.), pp. 197-241,Oxford: Blackwell ScientificPublications. 126 Small, 1. (1991).Exploring dataprovided by angling for salmonidsin the British Isles. In in Freshwater Fisheries Catch Effort Sampling Strategies - their Application Management(Cowx, I. G. ed.), Oxford: Blackwell Scientific Publishing. Smith, J.N. M., Taitt, M.J., Rogers,C. M., Arcese,P., Keller, L.F., Cassidy,A. L. E.V. and Hochachka,W. M. (1996). A metapopulationapproach to the populationbiology of the SongSparrow Melospiza melodia. This138,120-128. Smith, A. D.M., and Walters, C.J. (1981). Adaptive Managementof Stock-Recruitment Systems.Canadian Journal of Fisheriesand Aquatic Science38,690-703. Solomon,D. J. (1985). Salmonstock and recruitment,and stock enhancement.Joumal of Fish Biology 27 (Suppl. A), 45-57. Stenning,M. J., Harvey, P.H. and Campbell,B. (1988). Searchingfor density-dependent regulation in a population of pied flycatchers Ficeduld hypoleuca Pallas. Journal of Animal Ecology 57,307-317. Titus, R.G. and Mosegaard,H. (1992).Fluctuating recruitment and variable life history of migratory brown trout in a small unstable stream. Journal of Fish Biology 41,239- 255. Walters, C.J. (1981). Optimum Escapementsin the Face of Alternative Recruitment Hypotheses.Canadian Joumal of Fisheriesand Aquatic Science38,678-689. 127 Walters, C.J., and Ludwig, D. (1981). Effects of MeasurementErrors on the Assessment of Stock-RecruitmentRelationships. Canadian Journal of Fisheries and Aquatic Science38,704-7 10. Waring, C.P. & Moore, A. (2004). The effect of atrazine on Atlantic salmon (Salmo salar) smoltsin fresh waterand after seawater transfer.Aquatic Toxicology66,93-104. Whitehead,P. (1992). Examplesof recent models in environmentalimpact assessment. Journal of the Institution of Waterand EnvironmentalManagement 6,475-484. Whitehead, A. (2003) Progress in ending mixed-stock interceptory fisheries: United Kingdom. In Salmon at the Edge, (Mills, D. ed), pp.78-83, Oxford, Blackwell Publishing. Wright, JR, Armitage, P.D., Furse, M. T., and Moss, D. (1989). Prediction of invertebrate communities using stream measurements. Regulated Rivers. Research & Management 4,147-155. Wright, U., Blackburn, J.H., Gunn, R. J.M., Furse, M. T., Armitage, P.D., Winder, J.M., and Symes,K. L. (1996). Macroinvertebratefrequency data for the RIVPACSIII SITES in Great Britain and their use in conservation evaluation. Aquatic Conservation:Marine and FreshwaterEcosystems 6,141-167. 128