ICES SGBICEPS REPORT 2010

SCICOM STEERING GROUP ON ECOSYSTEMS FUNCTION

ICES CM 2010/SSGEF:03

REF. WGNAS, WGRECORDS, SSGEF SCICOM, ACOM

Report of the Study Group on Biological Characteristics as Predictors of Salmon Abundance (SGBICEPS)

24–26 November 2009 ICES Headquarters, Copenhagen, Denmark

International Council for the Exploration of the Sea Conseil International pour l’Exploration de la Mer

H. C. Andersens Boulevard 44–46 DK‐1553 Copenhagen V Denmark Telephone (+45) 33 38 67 00 Telefax (+45) 33 93 42 15 www.ices.dk [email protected]

Recommended format for purposes of citation:

ICES. 2010. Report of the Study Group on Biological Characteristics as Predictors of Salmon Abundance (SGBICEPS), 24–26 November 2009, ICES Headquarters, Copen‐ hagen, Denmark. ICES CM 2010/SSGEF:03. 158 pp.

For permission to reproduce material from this publication, please apply to the Gen‐ eral Secretary.

The document is a report of an Expert Group under the auspices of the International Council for the Exploration of the Sea and does not necessarily represent the views of the Council.

© 2010 International Council for the Exploration of the Sea

ICES SGBICEPS REPORT 2010 | i

Contents

Executive summary ...... 3

1 Introduction ...... 4 1.1 Main tasks ...... 4 1.2 Participants ...... 4 1.3 Background ...... 5

2 Summary of literature ...... 6 2.1 Salmon life history strategies ...... 6 2.2 Salmon in the Sea ...... 8 2.3 Climatic/oceanic factors ...... 18 2.4 Salmon in Freshwater ...... 22

3 Data sets ...... 32 3.1 Data sources and requirements ...... 32 3.2 Data on biological characteristics – update on developments ...... 33 3.3 Data quality issues – caveats and limitations ...... 35 3.4 Environmental data sets ...... 39

4 Case Studies ...... 40 4.1 Long‐term changes in biological characteristics of smolts on the River Bush, N. Ireland and associations with environmental parameters ...... 40 4.2 Update on biological characteristics of salmon from the River Dee, North Wales and other monitored rivers in UK (England & Wales) and associations with environmental parameters ...... 44 4.3 Biological characteristics of salmon from the River Frome – UK (England & Wales) ...... 50 4.4 Biological characteristics of salmon from the River Test – UK (England & Wales) ...... 50 4.5 Evidence for later age at maturity in Norwegian salmon stocks in recent years ...... 51 4.6 Baltic Sea – changes in post‐smolt survival and the factors affecting it ...... 52 4.7 Baltic Sea – review of Swedish tagging experiments and implications for estimating post‐smolt survival ...... 62 4.8 Fecundity of Penobscot River broodstock ...... 62 4.9 Smolt summary data for two monitored river sites in UK(Scotland) ...... 62 4.10 Burrishoole wild salmon census programme ...... 65

5 Exploratory analyses ...... 69 5.1 Analyses of long‐term variation in condition factor in relation to ocean climate ...... 70

ii | ICES SGBICEPS REPORT 2010

5.2 Biological characteristics data sets – temporal trends ...... 71 5.3 Biological characteristics data sets – spatial patterns...... 73 5.4 Overview of preliminary analyses of temporal and spatial trends ...... 81 5.5 Exploration of two‐way relationships ...... 88 5.6 Long‐term variation and changes in age at smoltification in four major Scottish salmon rivers ...... 100 5.7 Among‐river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 2SW “summer” salmon (returning May‐August/September) ...... 107 5.8 Among‐river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 1SW grilse (returning April–August/September) ...... 113 5.9 Analysis of the time‐series of the emigrant smolt run, River North Esk (1964–2008) ...... 119 5.10 Correlations between the length of returning 1SW salmon and the PFA of maturing (1SW) salmon ...... 123 5.11 Is there evidence of a change in overall (egg‐adult) mortality over time? ...... 126

6 Overview and recommendations ...... 137

Annex 1: List of participants ...... 139

Annex 2: List of working documents/presentations and data sets ...... 141

Annex 3: References ...... 143

ICES SGBICEPS REPORT 2010 | 3

Executive summary

A second meeting of the Study Group on the identification of biological characteristics for use as predictors of salmon abundance [SGBICEPS] was agreed by ICES at the 2009 Annual Science Conference (C. Res. 2009/2/SSGEF03) and met at ICES HQ, Co‐ penhagen from 24 to 26 November, 2009. The meeting was chaired by Ian Russell, UK and attended by 11 people from five European countries. Data were made available for stocks throughout the geographic range of Atlantic salmon ‐ Canada, USA, Iceland, Russia, Finland, Norway, Sweden, UK (Scotland), UK (England & Wales), UK (N. Ireland) and France. The main objectives of the meeting were to identify and compile time series of data on biological characteristics of Atlantic salmon and conduct preliminary analyses on these data as a basis for developing, and where possible testing, hypotheses relating any observed changes in these characteristics to trends in freshwater/marine mortal‐ ity and/or abundance of Atlantic salmon stocks and/or environmental changes. The Study Group report, in addressing the ToR, provides: (1) an updated summary of the available literature; (2) a description of the available data sets and how these were compiled; (3) a number of case studies; (4) details of exploratory analyses of the full data set; and (5) an overview and recommendations. In brief:  Literature ‐ the review summarises the life history strategies of salmon and changes in biological characteristics of different life stages across the geo‐ graphic range of the species in relation to key environmental variables. A number of existing, and sometimes conflicting, hypotheses relating to factors regulating the mortality of salmon are considered.  Data sets – additional data sets providing time series of various biological characteristics were made available to the Study Group. Information was also compiled in relation to the various sampling programmes from which data were derived, and methodological differences noted.  Case studies ‐ information from a number of new river or area‐specific inves‐ tigations were presented and reviewed and other case studies updated. A number of hypotheses were investigated.  Exploratory analyses ‐ the extended data sets of stock‐specific biological characteristics were re‐examined for possible temporal trends and to explore changes in biological characteristics over broader spatial scales. A number of hypotheses were also explored. Results included a clear relationship between the size of 1SW salmon returning to Europe and their abundance.  Recommendations – the Study Group highlighted the importance of monitor‐ ing programmes for collecting information on biological characteristics of salmon and recommended better utilisation of such data sets, recognising that the value of analysing a number of data sets together far exceeds their value when analysed separately. While further work was warranted, the Study Group felt that it had probably gone as far as it could in addressing the ToR and that further progress might best be made by other groups (e.g. in‐ volving modellers, biological oceanographers), potentially in response to specific management requirements.

4 | ICES SGBICEPS REPORT 2010

1 Introduction

1.1 Main tasks In June 2008, NASCO asked ICES to ‘continue the work already initiated to investi‐ gate associations between changes in biological characteristics of all life stages of At‐ lantic salmon, environmental changes and variations in marine survival with a view to identifying predictors of abundance’. WGNAS had begun work on this question, but had been unable to make significant progress due to other work pressures. A need was therefore identified for a separate expert group to take on this task if sig‐ nificant progress was to be made with addressing NASCO’s request for this advice. At the 2008 Annual Science Conference, ICES made a resolution (C. Res. 2008/2/DFC02) that a Study Group on the identification of biological characteristics for use as predictors of salmon abundance [SGBICEPS] (Chair: Ian Russell, UK) should be set up. This met for the first time in Lowestoft, England, from 3 to 5 March, 2009. In June 2009, NASCO repeated its request to ICES (as above) and it was subse‐ quently recommended at the 2009 Annual Science Conference (C. Res. 2009/2/SSGEF03) that the Study Group would meet again at ICES HQ, Copenhagen from 24 to 26 November 2009 to continue its work. By addressing this topic within a Study Group, ICES hoped it would be possible to provide the opportunity for scien‐ tists working on both Baltic and Atlantic salmon to contribute to the work. The terms of reference given by ICES were as follows: a ) identify data sources and compile time series of data on marine mortality of salmon, salmon abundance, biological characteristics of salmon and re‐ lated environmental information; b ) consider hypotheses relating mortality (freshwater and marine) and/or abundance trends for Atlantic salmon stocks with changes in biological characteristics of all life stages and environmental changes; c ) conduct preliminary analyses to explore the available datasets and test the hypotheses. The second Study Group meeting was attended by 11 people and these are listed in Section 1.2; the full address list of the participants is provided at Annex 1. The Study Group received a number of presentations and considered various working docu‐ ments; these are listed at Annex 2. In addition, the Study Group further updated and examined data relating to biological characteristics of salmon from a wide range of stocks around the North Atlantic and Baltic. Data were available for analysis from salmon stocks in Canada, USA, Iceland, Russia, Finland, Norway, Sweden, UK (Scot‐ land), UK (England & Wales), UK (N. Ireland) and France. A full list of the data sources and the contributors is also provided at Annex 2.

1.2 Participants Ian Russell (Chair) UK (England & Wales) Miran Aprahamian UK (England & Wales) Ian Davidson UK (England & Wales) Peder Fiske Norway Anton Ibbotson UK (England & Wales) Richard Kennedy UK (N. Ireland) Julian MacLean UK (Scotland) Stig Pedersen Denmark

ICES SGBICEPS REPORT 2010 | 5

Ted Potter UK (England & Wales) Ger Rogan Ireland Chris Todd UK (Scotland) Two people who attended the first Study Group meeting, but were unable to attend the second meeting, also made contributions to this second report of the Study Group. Subsequent to the completion of the second meeting, Erik Petersson (Sweden) and Jon Barry (UK (England & Wales)) completed further analyses of the updated data sets. Their valuable contribution to the work of the Study Group is therefore ac‐ knowledged.

1.3 Background Over the past 20–30 years there has been a marked decline in the abundance of Atlan‐ tic salmon across the species’ distributional range (Figure 1.3.1). Wild Atlantic salmon populations are declining across most of their home range and, in some cases, disap‐ pearing (ICES, 2008a). Generally, populations on the southern edge of the distribu‐ tion seem to have suffered the greatest decline (Parrish et al., 1998; Jonsson & Jonsson, 2009; Vøllestad et al., 2009). This may be linked to climatic factors. The decline in salmon abundance has coincided with a variety of environmental changes linked to an increase in greenhouse gases and a corresponding increase in temperatures (IPCC, 2001), which is most likely to have manifest effects at the edge of the species range. However, these areas are often also the ones with higher human population density and therefore, typically, where potential impacts on the freshwater environment may also be greater. This has potential implications for the survival of juvenile salmon and their resulting fitness when they migrate to sea as smolts (e.g. Fairchild et al., 2002). In addition to changes in climate and potential freshwater issues, various other factors have been postulated as possibly contributing to the decline in stock abundance, in‐ cluding predation, aquaculture impacts and the effects of fisheries.

Figure 1.3.1. Approximate oceanic distribution area of Atlantic salmon.

6 | ICES SGBICEPS REPORT 2010

Atlantic salmon occupy three aquatic habitats during their life‐cycle: freshwater, es‐ tuarine and marine. Similar factors contribute to mortality in each of these habitats – competition, predation and environmental factors ‐ but despite occurring in different habitats these are not independent. Conditions experienced within the freshwater environment can affect the survival of emigrating smolts and marine conditions may subsequently modify the spawning success of fish in freshwater. It should be noted that the decline in salmon populations has occurred despite sig‐ nificant reductions in exploitation, although this does not preclude possible fishery effects. An underlying cause has been a marked increase in the natural mortality of salmon at sea – the proportion of fish surviving between the smolts’ seaward migra‐ tion and their return to freshwater as adult fish (e.g. Peyronnet et al., 2008). The proc‐ esses controlling marine survival are relatively poorly understood (Friedland, 1998), although there is growing support for the hypotheses that survival and recruitment is mediated by growth during the post‐smolt year, for European stocks at least (Fried‐ land et al., 2009). In addition to the declines in abundance, changes in salmon life histories are also widely reported throughout their geographic range, affecting factors such as sea‐age composition, size at age, age at maturity, condition, sex ratio and growth rates (e.g. Nicieza & Braña, 1993; Hutchings & Jones, 1998; Niemelä et al., 2006; Peyronnet et al., 2007; Aprahamian et al., 2008; Todd et al., 2008). Changes are also manifest in fresh‐ water stages, affecting factors such as the size and growth of parr and the age of smolting (e.g. Davidson & Hazelwood, 2005; Jutila et al., 2006). In addressing the terms of reference posed by NASCO, the Study Group have been asked to consider hypotheses relating marine mortality and/or abundance trends for Atlantic salmon stocks with changes in biological characteristics of all life stages and environmental changes. The purpose is to determine whether declines in marine sur‐ vival and abundance coincide with changes in the biological characteristics of juve‐ niles in fresh water or are modifying characteristics of adult fish (e.g. size at age, age at maturity, condition, sex ratio, growth rates, etc.), and whether these changes are linked with environmental change. As a foundation for addressing these questions, Section 2 attempts to summarise available information on the life history strategies of salmon and changes in their bio‐ logical characteristics in relation to some of the key environmental variables, as a ba‐ sis for developing hypotheses as to what the possible underlying mechanisms might be. This review of the literature was compiled and included as part of the first Study Group report (ICES, 2009a). However, brief updates have been included here where new information has come to light.

2 Summary of literature

2.1 Salmon life history strategies Atlantic salmon have highly diverse and plastic life‐history traits and occupy a di‐ verse array of environments (Elliott et al., 1998). This diversity has been suggested as the mechanism that enables small populations to persist (Saunders & Schom, 1985). Atlantic salmon populations vary from fully freshwater resident to anadromous forms. Freshwater resident populations occur throughout the range of the species in North America, but are relatively uncommon in Europe (MacCrimmon & Gots, 1979). Non anadromous populations exist in water courses that are isolated from the sea but also in sympatry with anadromous populations. In contrast to Arctic charr, there is

ICES SGBICEPS REPORT 2010 | 7

no correlation between the prevalence of anadromous forms and latitude (Klemetsen et al., 2003). Atlantic salmon also exhibit variability, both within and among popula‐ tions, in factors such as freshwater habitat use, length of freshwater residence and age at maturity (Klemetsen et al., 2003). Most Atlantic salmon populations are anadromous, with smolts migrating to sea to exploit the more abundant marine food resources and attain a large size at maturity before returning to freshwater to breed. The majority of these fish undergo extensive oceanic migrations (Hansen & Quinn, 1998). However, stocks in the Baltic Sea (Karls‐ son & Karlström, 1994) and Inner Bay of Fundy (Amiro, 1998) tend to stay within the confines of these respective areas, although there is evidence that the latter may not be a successful strategy and may be changing (Hubley et al., 2008). Some stocks at the northern extremity of the North American distribution remain close to the river, ma‐ ture after just a few months at sea and are known as ‘estuarine’ salmon. There is some evidence that the incidence of these fish has become more prevalent in recent years in certain rivers (Downton et al., 2001). Fish typically return to their natal rivers to spawn, resulting in a certain degree of reproductive isolation between different populations, although some straying into other rivers does usually occur (Marshall et al., 1998). Salmon normally spawn in the same season that they return to freshwater, but this does not always apply (Webb & Campbell, 2000). Unlike most Pacific salmon, Atlantic salmon can spawn repeatedly (i.e. they are iteroparous), although there is wide variability between populations. Salmon mature at various sea‐ages, typically maturing for the first time as 1 to 3 sea‐winter (SW) fish, but also sometimes at older sea‐ages. The biological characteristics (e.g. size at age, sex ratio, smolt age, fecundity, etc.) of these sea‐age groups vary widely among stocks and with geographic location. For example, maiden 5SW salmon occur in the River Tana (Teno) in northern Europe (Erkinaro et al., 1997), while stocks in New‐ foundland consist almost entirely of salmon which mature as 1SW fish (Dempson et al., 1986). Life‐history characteristics are further complicated by the fact that parr can also be‐ come sexually mature. This is typically restricted to males, although there are isolated reports of mature female parr (e.g. Power, 1969; Moore & Riley, 1992). Sexually ma‐ ture male parr successfully mate with mature adult females both in the presence and absence of adult males (Myers & Hutchings, 1987) and are thought to play an impor‐ tant role in maintaining small populations (L’Abée‐Lund, 1989). Reduction of the ef‐ fective population size (Ne) can radically alter the rate of loss of genetic heterozygosity in any population and this may be especially pertinent to small river stocks. Martinez et al. (2000), for example, assessed the frequency of successful fertili‐ zations by precocious male parr in three threatened river stocks at the southern dis‐ tributional limit of salmon in Europe. They found frequent multiple paternity of eggs within redds, and that precocious parr were sufficiently successful to have significant effects in increasing Ne. Life‐history strategies are a means to successful reproduction and flexibility of these strategies is a characteristic of salmonid species (Thorpe, 1990). Stocks of salmonids are often defined by a propensity to migrate and mature at particular ages, but if transplanted to a non‐native environment they perform differently (Thorpe, 1990). Hence, the environment influences their genetic predisposition to follow a particular life‐history strategy. However, the direction of those decisions may depend on the current metabolic performance of the individual. Therefore, such developmental de‐

8 | ICES SGBICEPS REPORT 2010

cisions may be viewed within an abiotic‐biotic regulatory continuum, depending on the abiotic environment for their initiation and the biotic environment for their com‐ pletion (Thorpe, 1990). The high degree of variability in life‐history characteristics and phenotypic plasticity needs to be borne in mind and provides a necessary backdrop for reviewing and as‐ sessing recent changes in biological characteristics of the fish. It is also important to bear in mind that many of these biological characteristics are inextricably linked – e.g. growth, maturation and run timing – and thus impossible to consider in isolation.

2.2 Salmon in the Sea

Marine survival Marine survival of salmon is typically expressed as the proportion of emigrating smolts that return to homewaters (to the coast) or to their river of origin as 1SW or 2SW adults. In reality, these ratios are return rates rather than survival rates (Crozier et al., 2003) since they reflect the effects of both mortality and maturation. Changes in the age at maturation may affect the relative proportions of a smolt cohort that return as 1SW or 2SW fish, but this can also result from changes in natural mortality in dif‐ ferent areas of the ocean. Nevertheless, these return rates can be considered as con‐ venient indicators of survival (Crozier et al., 2003). Numerous factors are thought to affect the survival of salmon in the sea, both biotic and abiotic, although their relative impact and the interaction between them are poorly understood. Marine mortality of salmon is considered to be density‐ independent since salmon abundance is not constrained by the carrying capacity of the NE Atlantic (Jonsson & Jonsson, 2004). Instead, density‐independent processes are believed to regulate marine mortality either directly (physiologically) or indirectly by controlling a fish’s ability to feed (find high densities of prey), migrate or escape predators. Sources of marine mortality, in general, are poorly understood due to a lack of basic knowledge about post‐smolt distributions and habits (Friedland, 1998; Dadswell et al., 2010), although information on post‐smolt distribution is improving (e.g. Holm et al., 2000). However, it is generally accepted that the main marine mortality events take place during the first year of sea life when survival, maturation, and migration trajectories are being defined (Hansen & Quinn, 1998; Potter & Crozier, 2000; Fried‐ land et al., 2005, 2009). The key factors influencing the mortality of salmon in the sea are believed to be:  Environment – climatic variations play a key role in shaping the marine environment, affecting currents, gyres and sea surface temperatures (SST). Such factors can impact upon salmon directly (e.g. migration routes) or in‐ directly (e.g. effects on the abundance and distribution of prey species or predators). The broad scale declines in salmon abundance and the more pronounced declines for MSW salmon point to changes in the marine envi‐ ronment affecting the survival of salmon at sea. Friedland & Reddin (1993) demonstrated correlations between the area of potential post‐smolt habitat in the sea, defined as the area combining their optimal temperature and full salinity, and catches of salmon from that area. More recent studies have confirmed such links for European stocks occupying the eastern North Atlantic, though the pattern for North American stocks and the western North Atlantic (e.g. Gulf of St Lawrence) are rather more complex (Friedland et al., 2003a, b, 2009).

ICES SGBICEPS REPORT 2010 | 9

 Food – growth and survival are likely to be affected by the abundance and distribution of suitable prey, particularly during the period of initial ma‐ rine residence. When smolts enter saltwater their energy expenditure in‐ creases and scarce food resources at this time may result in increased mortality. A lack of food would also reduce growth and increase the likeli‐ hood of predation. The diet of salmon has been shown to change over time (Andreassen et al., 2001; Haugland et al., 2006) and with sea‐age (Jacobsen & Hansen, 2001), but is still poorly understood. For example, although stomach contents analysis permits an appraisal and comparison of prey items taken in differing locations and for salmon of differing size and sea‐ age, no detailed assessment of active prey choice is yet possible because of a lack of comprehensive data on prey availability.  Growth – salmon exhibit higher rates of growth and mortality than other pelagic species (Cairns, 2003). This strategy is speculated to represent a trade‐off between the two and highlights the possible importance of ma‐ rine growth in controlling marine survival and recruitment. It is generally accepted that mortality at sea is growth mediated (Friedland et al., 1996,) and significant relationships have been demonstrated between SST, sur‐ vival and growth in the first sea year (Friedland et al., 2000), and particu‐ larly in the first summer (Friedland et al., 2009). Distinguishing between environmental and genetic influences on growth is difficult. Differences in growth may reflect variable food supply or more general changes in oceanographic processes, which could affect some populations more than others depending on their marine distribution. Differences in growth rates may also be influenced by natural selection processes in individual rivers.  Predation – this may be the most important source of salmon mortality, al‐ though quantitative information on predation in the sea is scarce. Preda‐ tion is believed to be most severe on smolts and post‐smolts ‐ small fish are vulnerable to a larger range of predators than large fish and more preda‐ tors occur on the continental shelf than in oceanic areas. Mortality is gener‐ ally lower for larger smolts (Lundqvist et al., 1994). Size clearly is an important factor determining the survival of hatchery‐reared smolts (e.g. Kallio‐Nyberg et al., 2004; Lacroix & Knox, 2005) and the typically larger size of hatchery‐reared smolts may, to some extent, compensate for their origin in comparison to wild smolts (Jutila et al., 2006; Kallio‐Nyberg et al., 2004, 2006). But the role of smolt size in influencing the performance of wild stocks has been poorly studied and remains unclear. For example, for wild fish from the River North Esk (Scotland), smolt size clearly had no in‐ fluence on resultant survival rate to adulthood, whereas for the River Fig‐ gjo (Norway), adult survival rate was correlated with size at sea entry (Friedland et al., 2009).  Competition – while intra‐specific competition may be unlikely in the North Atlantic, there is some evidence to suggest that inter‐specific compe‐ tition could occur. For example, negative relationships have been observed between herring abundance in the Norwegian Sea and salmon catch and between herring abundance and marine survival of smolts from the River Figgjo (Crozier et al., 2003). For salmon in the Baltic Sea, survival indices have been correlated with both the total production of wild and hatchery‐ reared smolts in the Baltic and herring recruitment (ICES, 2008b).

10 | ICES SGBICEPS REPORT 2010

A number of these possible regulatory factors are considered in greater detail below in relation to different salmon stock characteristics.

Age at Maturation Age at maturation is believed to be a key life‐history trait, as fitness is reported to be more sensitive to changes in this trait than to changes in many other life‐history traits (Stearns, 1992). Early maturation reduces the generation time and increases the chance of surviving to breed, but early maturing individuals are smaller and produce fewer or smaller progeny. Hence, an optimal trade‐off will depend on age‐specific growth and mortality rates (Stearns, 1992). In salmon, any such trade‐offs are also likely to be influenced by the levels of repeat spawning. Life‐history theory suggests that a trait will change in relation to changes in age‐specific mortality, growth and fecundity to ensure fitness is maximized (Roff, 1992). Thorpe (2007) cautions that age at first maturity or size at first maturity in salmon can be misleading measures, since they appear to suggest that steps toward reproductive ripeness have not taken place until a specific age or size is reached at which spawn‐ ing can occur. In Atlantic salmon, maturation begins in the egg soon after fertilization and continues intermittently until the individual is capable of spawning. Thorpe (2007) suggests that the fish’s developmental decisions are likely to be based on proximate cues, both internal and external, largely independent of size and age. In Atlantic salmon, fish mature at various ages and this affects the patterns of return ‐ run timing ‐ of adult fish to freshwater.

Run Timing and Sea-age Run timing in adult Atlantic salmon is highly variable. Different sea‐age classes of salmon have different patterns of run timing and these vary on a geographic scale, but also between stocks in a region and within stocks over time. A change in the pat‐ tern of run timing could therefore result from a change in the balance between the various sea‐age classes, a change in run timing within sea‐age classes or both. There is widespread evidence of change in the sea‐age composition of salmon throughout their geographic range (e.g. Gough et al., 1992; Anon., 1994; Summers, 1995; Gudjons‐ son et al., 1995; Welton et al., 1999; Juanes et al., 2005; Quinn et al., 2006; Aprahamian et al., 2008). In the UK, 1SW salmon mainly enter rivers from June to August, though some rivers have strong autumn runs, 2SW fish enter throughout the year, but sometimes with spring, summer or later peaks, while 3SW fish generally enter rivers early in the year, with few entering after about May. Fish spawning for a second time tend to adopt similar run timing to that of their first migration. In Norway, most salmon enter riv‐ ers from May to October, with MSW salmon tending to enter earlier than 1SW fish (Jonsson et al., 1990a,b). Broadly similar patterns apply in eastern Canada, although some stocks are characterized as ‘early’ or ‘late’ running stocks (Klemetsen et al., 2003). In Scotland, some fish have been reported to enter rivers over a year before they will spawn (Webb & Campbell, 2000). An analysis of long‐term data sets for 12 salmon stocks in the UK (Anon., 1994), indi‐ cated similar changes in the monthly pattern of catches and in the contribution of dif‐ ferent sea‐age classes. The spring component of the catches increased both numerically and as a proportion of total catch from 1910 to about 1930, remained generally stable until the early 1950s, but subsequently showed a steady decline to the current low levels. It was concluded that the dominant process in these shifts in

ICES SGBICEPS REPORT 2010 | 11

timing of runs and catches was a change in sea‐age composition. While there was some evidence of a shift in run timing within sea‐age classes, this was evidently not the main mechanism of change. Extensive variability in the sea‐age composition of stocks over the past 100 years or more has been demonstrated in other studies (George, 1984, 1991; Martin & Mitchell, 1985; Summers, 1995; Heddell‐Cowie, 2005), with evidence of 1SW salmon being pre‐ dominant for some periods and MSW salmon at others. These changes have occurred at broadly similar times among rivers, suggesting that common factors operating in the marine environment have been the main cause of change in age at maturity (Aprahamian et al., 2008). Over recent decades, many stocks around the North Atlan‐ tic have experienced long‐term declines in the MSW ‘spring’ component, which ap‐ pears to be driven primarily by an increase in marine mortality (ICES, 2008a). The sea‐age composition of stocks and adult run timing are influenced by various factors. Maturation rate is a function of both stock genetics (e.g. Stewart et al., 2002) and environment, although the relative influence of these factors is not clear (Thorpe, 1994, 2007; Friedland & Haas, 1996; Friedland, 1998). There is evidence for an inher‐ ited link with maturation (Hansen & Jonsson, 1991) – typically, rapidly developing parents tend to produce early maturing offspring. Environment also plays a role in determining the sea‐age, maturation and run timing of salmon, operating in both the marine environment and freshwater (Scarnecchia et al., 1989, 1991; Jonsson et al., 1991a; Nicieza & Braña, 1993; Anon., 1994; Gudjonsson et al., 1995; Frieldland & Haas, 1996; Friedland, 1998; Friedland et al., 2003a,b; L’Abee‐Lund et al., 2004; Juanes et al., 2005; Peyronnet et al., 2008). There is, however, no mechanistic framework to explain how seasonal growth and ocean environment combine to produce annual variability in maturation. Stocks have very different maturity schedules, often associated with latitude, and age at maturity can change over time (Anon., 1994; Summers, 1995). An examination of the variation in sea‐age at maturity in 158 Norwegian rivers over large spatial and temporal scales (L’Abee‐Lund et al., 2004) found no general tempo‐ ral trend in the proportion of 1SW salmon in the catches: the proportion decreased significantly in 10 stocks and increased significantly in 11 stocks. There was some evidence of coherence in temporal patterns at a regional level, and river‐specific fac‐ tors (river discharge, topographic gradient and the presence of lakes) explained a large percentage of the spatial variation in the proportions of 1SW fish. This propor‐ tion increased with decreasing river size (measured as water discharge), where most of the discharge occurred during the summer period (the main migration period), and for rivers located closer to the open ocean, probably reflecting the effect of early feeding on growth and maturation. There were large regional differences in the 1SW proportion, not obviously associated with latitude. However, 1SW proportions were generally higher in the northern part of Norway than in the south, possibly reflecting large‐scale differences such as oce‐ anic migration routes, although there was no evidence of a general effect of the North Atlantic Oscillation Index (NAOI) on temporal trends in 1SW fish proportions. How‐ ever, Jonsson & Jonsson (2004) indicated the existence of a correlation between NAOI and age at maturation for one Norwegian river, with a positive winter NAOI (indica‐ tive of a mild and stormy winter) correlated with a decreased age at maturity. It was suggested that salmon grow quicker and therefore mature earlier in mild winters, and that climatic conditions at the time of sea entry may be important for the later performance (growth) of the spawners.

12 | ICES SGBICEPS REPORT 2010

Selective exploitation of older sea‐age classes can also affect the sea‐age composition of stocks (Gee & Milner, 1980; Moore et al., 1995). Rod fisheries can be responsible for much higher levels of exploitation on MSW spring fish than of 1SW fish from the same stock. The high seas fisheries at West Greenland and Faroes have also mainly exploited potential 2SW and older returnees. A correlation between run timing of salmon and water temperature has been demon‐ strated in the Baltic region (Dahl et al., 2004), with a strong correlation between the migration peak and mean monthly sea and river temperatures in spring: fish ran ear‐ lier in years when the water temperature was higher. It was speculated that this may reflect earlier gonad maturation at higher temperatures; a condition‐dependent initia‐ tion of migration; or a means to reduce total migration energy costs since the cost of migration increases with temperature. Further, river conditions can also directly af‐ fect run timing to a river, with low flows potentially causing considerable delays. It is not entirely clear what role the freshwater environment plays in determining the sea‐age of salmon through its influence on the growth rate of parr, and the evidence is inconclusive and sometimes contradictory (Anon., 1994). For some stocks, an in‐ verse ratio hypothesis was proposed that suggested slow‐growing parr which be‐ come smolts at older river age tended to return as younger sea‐age fish. However, in other stocks it is evident that fish which develop quickly in freshwater continue to do so once they migrate (Anon., 1994). It has been concluded that there is no fundamen‐ tal causal mechanism linking freshwater growth and sea‐age (Gardner, 1976; Bielak & Power, 1986; Anon., 1994). Evidence from long‐term monitoring of marine processes indicates the existence of major fluctuations in parameters such as temperature, salinity and plankton produc‐ tivity, occurring over a similar timescale to that in the observed shifts in salmon run timing and in the declines in many stocks (Dickson & Turrell, 1999; Beaugrand & Reid, 2003). Direct causal links have not been identified and Beaugrand & Reid (2003) conclude that results are not necessarily indicative of a trophic cascade or bottom‐up control of salmon abundance. However, it seems clear that fluctuations of such fun‐ damental biological significance in the marine environment are likely to be key proc‐ esses in regulating the sea‐age composition and run timing of salmon, as well as the overall abundance of stocks. While the mechanisms by which changing marine conditions may influence the sea‐ age composition of stocks are not clear, there are a number of possibilities. These in‐ clude: direct physiological effects on the growth and maturation processes of indi‐ vidual fish; an increase in total natural mortality throughout the period of marine residence, thus reducing the proportion of older sea‐age fish as well as the overall numbers; differential mortality leading to a shift in genetic tendency to a particular sea‐age; and differential mortality within each smolt year class of fish destined to re‐ turn as 1SW or MSW salmon, since sea‐age groups are known to inhabit different ar‐ eas at different times during their marine migrations. Friedland (1998) reported that the abundance of 2SW spawners in North America is directly scaled to the size of over‐wintering thermal habitat in the NW Atlantic, sug‐ gesting a link between maturation and environment. This is consistent with the mi‐ gration‐maturation hypothesis described by Friedland et al. (1998b). This hypothesis is specified for southern N. American stocks, but is intended as a general hypothesis applicable to all stocks and speculates that post‐smolts that migrate to more northerly areas are affected differently by over‐wintering conditions and, at the end of the win‐ ter, find themselves in areas where they fail to receive cues related to sensing their

ICES SGBICEPS REPORT 2010 | 13

home rivers. As a result, the fish feed and grow, join feeding migrations and their maturation state regresses. Those fish that have a more southerly post‐smolt migra‐ tion experience different over‐winter conditions and are closer to their home rivers after winter. Thus, they are more likely to sense cues, develop sexually and invoke behaviours to navigate home. There are, however, alternative hypotheses relating to the marine migration of Atlan‐ tic salmon. Dadswell et al. (2010) argue that a gyre model for ocean migration is highly plausible and have termed this scenario, originally formulated by Reddin et al. (1984), as the ‘Merry‐Go‐Round’ hypothesis. This proposes that salmon undertake trans‐Atlantic migrations using surface currents of the North Atlantic sub‐polar gyre (NASpG). European and North American stocks enter the NASpG on their respective sides of the ocean and migrate counter‐clockwise around the North Atlantic feeding and growing until they mature after one, two or more years, at which point they de‐ part the NASpG at the appropriate location and return to their native river. They ar‐ gue that it is unlikely that post‐smolts would have the necessary stored navigational information to make straight line movements to specific areas such as West Greenland. Rather, they suggest that the interplay of physiology and environmental cues probably maintain salmon in the appropriate areas of the NASpG for year‐round maximum growth and survival.

Growth and Age at Maturity Growth in salmonids is regulated by temperature and typically increases linearly with water temperature up to an optimum rate given an adequate food supply (Brett, 1979). However, it has been reported that growth in Atlantic salmon post‐smolts has a non‐linear response to temperature, with optimum growth occurring at 13°C (Han‐ deland et al., 2003). Temperature can act on growth directly by affecting physiological processes or indirectly by modifying ecosystems (e.g. food availability or changes in other aspects of the rearing environment). Higher temperatures may increase meta‐ bolic demand beyond available food resources and inhibit growth. Some fish on low food rations have been shown to have a lower temperature preference than fish on high rations (Despatie et al., 2001), suggesting that when food is limiting growth will be optimised at lower temperatures. However, other factors also influence fish growth and the size that fish attain at maturity. Large body size confers certain advantages for spawning fish, but can be balanced by the greater probability of mortality associated with spending more time at sea and by potential difficulties in accessing spawning habitats, particularly in smaller rivers. Such factors are thought to drive the adaptation of locally adapted age and size dis‐ tributions for fish (e.g. Jónasson et al., 1997). Mean body size has been shown to vary consistently among populations with some rivers supporting large MSW fish, while others support smaller 1SW salmon (Schaffer & Elson, 1975). River discharge volume may be the main factor determining the among‐population variation in fish size for rivers where mean annual water dis‐ charge is <40 cumecs (Schaffer & Elson, 1975; Scarnecchia, 1983; Jonsson et al., 1991a). Mean size and age at maturity increase with river discharge for these smaller rivers, although there is no such relationship for larger rivers (Jonsson et al., 1991a). Within populations, temperature may be a key factor influencing temporal differences in growth rate at sea through its influence on food resources and the growth potential of the fish (Friedland et al., 1998a; Friedland et al., 2000; Jonsson et al., 2001).

14 | ICES SGBICEPS REPORT 2010

There is conflicting evidence as to whether Atlantic salmon smolt size influences sub‐ sequent post‐smolt growth. Both negative relationships (Skilbrei, 1989; Nicieza & Braña, 1993; Jonsson & Jonsson, 2007) and positive relationships (Lundquist et al., 1988; Salminen, 1997) have been reported, and Friedland et al. (2006) indicated that marine growth of post‐smolts in the Gulf of St. Lawrence from August to October was independent of freshwater growth history. Various hypotheses have been pro‐ posed to explain the slower growth at sea of larger smolts (Jonsson & Jonsson, 2007). This has been attributed to growth of the gill surface area, with oxygen consumption becoming gradually more limiting for growth as the fish get larger (Pauly, 1981). Wootton (1998) suggested that the surface area for absorbing food may limit growth in larger fish, and Einum et al. (2002) proposed that fish growing more slowly in freshwater, and which are small as smolts, may be better adapted to the marine envi‐ ronment. It has also been demonstrated that the age at sexual maturity decreases with decreas‐ ing first year growth at sea (Jonsson et al., 1991a; Jonsson & Jonsson, 2007). Evidence of a linkage between higher growth rates during the first winter at sea and higher age at maturity has also been demonstrated for populations in northern Spain (Nicieza & Braña, 1993). This contrasts evidence from the Baltic (Salminen, 1997) where sea‐age at first spawning was inversely related to the marine growth of sea‐ranched salmon. Jonsson & Jonsson (2007) suggest this discrepancy may be related to the expected fit‐ ness gain of fish exhibiting different growth trajectories resulting in different reaction norms for growth (Stearns, 1992). Fish that grow fast initially may either attain sexual maturity relatively early, when later growth is environmentally constrained, or later when the high growth rate is maintained (Jonsson & Jonsson, 1993, 2004). Thus salmon and a number of other fishes appear to delay maturity when the growth rate is consistently high throughout life, but mature early if growth rate starts to level off (Jonsson & Jonsson, 1993). Exploitation has also been shown to affect the size and age of returning fish. For ex‐ ample, the cessation of drift net fishing in Norway has been shown to affect the struc‐ ture of the spawning run in a number of Norwegian and Russian salmon populations (Jensen et al., 1999). In the Miramichi River, Canada ‐ increases in mean length at age and the abundance of repeat spawners have been attributed to reductions in exploita‐ tion (Moore et al., 1995). Long‐term declines in the size of 2SW salmon and a reduced proportion of previous spawners were also attributed to the selective effects of fish‐ ing for a stock in Quebec (Bielak & Power, 1986).

Sex Ratio There is a tendency for male fish to mature and return to freshwater at a younger age than females. In many stocks the 1SW component of the run is predominantly male while 2SW fish are predominantly female (Anon., 1994). Female fish have predomi‐ nated in the catch at West Greenland, where most fish are approaching their second sea winter when caught and where fish from a wide range of stocks are exploited. Some studies suggest that the sex ratio in 3SW salmon is more even, while others in‐ dicate a bias towards females in some stocks and males in others (Anon., 1994). Since many stocks have experienced a decline in the proportion of older MSW salmon and a relative increase in the predominance of 1SW fish this could have a marked effect on the sex ratio of the spawning stock and the overall level of egg deposition.

ICES SGBICEPS REPORT 2010 | 15

Previous/Repeat Spawners The proportion of salmon surviving to spawn varies markedly among and within Atlantic salmon populations. In some stocks very few fish survive after spawning, while in other populations fish may return repeatedly and fish with a wide range of life histories can contribute to egg deposition in any year, even within populations with an apparently simple sea‐age structure. For example, although most stocks in Newfoundland consist almost entirely of salmon which mature as 1SW fish, signifi‐ cant numbers of these fish can migrate back to sea after spawning to return and spawn repeatedly (Dempson et al., 1986). Some fish return to spawn in a number of consecutive years (consecutive spawners) while others return every other year (alter‐ nate spawners), although fish have also been known to switch between these two strategies (Klemetsen et al., 2003). Levels of repeat spawning are clearly influenced by the overall level of exploitation and also the possible size selectivity of fishing gears. Repeat spawning is affected by the size of the fish, with the proportion of repeat spawners decreasing with size. This is possibly related to energy expenditure during spawning: 1SW salmon may allocate 50% of their energy for spawning (Jonsson et al., 1991b) compared with 70% in older fish (Jonsson et al., 1997). A study based on stocks in 18 Norwegian rivers (Jonsson et al., 1991a) indicated that in small rivers (flow rates <40 m3 sec‐1) salmon tended to ma‐ ture at smaller size and age, but post‐spawning survival was high with fish tending to spawn annually (consecutively). Large rivers were characterized by larger matur‐ ing salmon, with low post‐spawning survival and the repeat‐spawning fish mainly returning to spawn in alternate years. Within populations the proportion of alternate spawners increased with the size at first maturity. A study conducted in a sub‐arctic region at the border between Finland and Norway showed that the proportion of repeat spawners has increased in the last 10 years (Niemelä et al., 2006), more so in females than males. For the Miramichi stock, there has been a significant recent increase in the rate of salmon returning to spawn for a second time as consecutive spawners, but not for the alternate spawning life history strategy, for both 1SW and 2SW maiden salmon (ICES, 2008a). This has been associ‐ ated with years of a high biomass index of small fish in the southern Gulf of St. Law‐ rence, a change attributed to reduced predation pressure resulting from the collapse of the previously dominant groundfish stocks in this area (cod, skate, flatfish species) (Benoit & Swain, 2008). It has been suggested that this may reflect bottom‐up effects of prey availability on adult fish abundance as prey abundance is an important factor in post‐spawner survival in Atlantic salmon.

Changes in Size at Maturity Previous investigations have demonstrated that the marine growth of salmon has decreased over the last 20 years (Crozier & Kennedy, 1999; Jonsson et al., 2003). There have also been recent widespread reports of unusually small 1SW fish returning to rivers in many parts of Europe with the mean weight of fish in a number of stocks being the lowest in the time series (ICES, 2007a, 2008a; Todd et al., 2008). However, these changes are not manifest in all populations (Davidson & Hazelwood, 2005; ICES, 2008a). The decrease in growth in recent years has been linked to indirect ef‐ fects of anomalous warming in areas where salmon are located at sea (Todd et al., 2008). The mean standardised weights of 1SW fish from 20 Norwegian rivers have also correlated positively with the estimated pre‐fishery abundance (PFA) of the cor‐ responding sea year class, and the annual mean weight of small salmon (< 3 kg) from

16 | ICES SGBICEPS REPORT 2010

the River Drammen has correlated positively with the estimated survival of hatchery reared smolts released in the same river (ICES, 2008a). Todd et al. (2008) have indicated almost identical temporal patterns in growth condi‐ tion variation for two Scottish data sets – a single river stock and a mixed stock fish‐ ery ‐ with an overall decrease of 11–14% over the past decade. Growth condition has fallen as SST has risen, and for each year class, a negative correlation was identified between the midwinter (January) SST anomalies in the Norwegian Sea and the final condition of the fish on return during the subsequent summer. The study explicitly drew no connection with stock abundances, but did also demonstrate that under‐ weight individuals had disproportionately low reserves of stored lipids, which are crucial for successful spawning of individuals. It was felt that the investigation was consistent with other analyses in providing evidence of major, recent climate‐driven changes in northeast Atlantic pelagic ecosystems, and the likely importance of bot‐ tom‐up control mechanisms manifest as changes in the quality and/or quantity of available prey. ICES (2008a) cautioned that the growth of salmon during the first year at sea, as as‐ sessed from returning 1SW salmon, provides an indirect measure of growth rate. If the conditions that smolts experience during the first weeks or months at sea and growth during this period are crucial for size‐selective mortality, measurements of circuli on scales during this period (Beamish et al., 2004; McCarthy et al., 2008) may be better correlated with survival than growth over the whole period. Furthermore, if mortality is size‐selective, in years with harsh conditions, only the largest fish are likely to survive and this is likely to compromise potential correlations between sur‐ vival and the size of returning fish. Peyronnet et al. (2007) noted that the length of 1SW fish returning to the River Burrishoole had varied little over a 40‐year period and that there was no significant correlation between average length and marine recruitment. It was suggested that length at return was unlikely to be a good indicator of the limiting effects of the envi‐ ronment on survival, since such effects would probably manifest mainly when fish are small and most vulnerable. At larger size, salmon would have potential to feed and grow intensively at different times. Thus, growth in the first months at sea may have a strong influence on recruitment, but may have a relatively small effect on length at return compared with growth experienced after the first winter. This Burrishoole study looked at various growth measurements based on inter‐circuli distances and found that the number of scale circuli deposited during the marine phase was highly correlated to both rates of marine survival and to the time series of PFA. However, the only significant relationship was found with the distances over the first ten scale circuli. This was a negative relationship, suggesting that high growth during early marine residency results in lower overall marine survival. Davidson & Hazelwood (2005) found only weak correlations between weight‐at‐ return and post‐smolt growth (first‐year marine) increments, indicating that the for‐ mer is not heavily influenced by initial growth at sea. Cairns (2003) suggests that salmon aim to achieve a target weight prior to return and one of the consequences of this is that fish must compensate for poor initial growth by increased foraging activ‐ ity but at the cost of greater susceptibility to predation. Relatively stable return weights for 1SW salmon for stocks in the UK (Rivers Wye and Dee) over the last 40 years are consistent with this target weight hypothesis. Conversely, however, the weight of 2SW salmon appears to have been increasing in recent years (Rivers Severn, Wye and Dee). Annual variations in adult weight‐at‐return data for these rivers were

ICES SGBICEPS REPORT 2010 | 17

also highly synchronous, especially within sea‐age groups, but showed no strong as‐ sociations with SST variables or the NAO index. Long‐term changes have been observed in the mean size of salmon caught in the fish‐ ery at West Greenland, with the mean size of European origin fish declining more markedly than that of American fish (ICES, 2008a). This may suggest that ocean habi‐ tat has been more limiting for European fish, although appears to be at odds with the apparent increase in size of 2SW salmon noted for rivers in England and Wales.

Growth and Marine Survival A number of authors have provided evidence that growth during the first year at sea has a critical influence over marine survival and that recruitment is strongly linked to growth (Friedland & Haas, 1996; Friedland et al., 1998a; Friedland et al., 2000; Jonsson et al., 2003; Friedland et al., 2005; McCarthy et al., 2008; Peyronnet et al., 2007; Peyron‐ net et al., 2008). Beamish & Mahnken (2001) have proposed a ‘critical size and period hypothesis’ to explain the recruitment of coho salmon. This proposes that the fish must reach a threshold size (possibly varying with year) by the end of the first sum‐ mer and autumn at sea to be able to cope with the metabolic demands of winter. However, while Friedland et al. (2009) identified a critical period of Atlantic salmon post‐smolt growth and survival, their results did not suggest that a critical size mechanism was controlling adult abundance. However, Crozier & Kennedy (1999) reported that survival of cohorts to both the coast (pre fishery) and freshwater for salmon from the River Bush were unrelated to variation in growth from smolt migration to the end of the first winter in the sea. Fur‐ ther, variability in marine growth was much less than the variation in natural sur‐ vival at sea, suggesting that factors instead of, or in addition to, growth influence natural marine survival. Davidson & Hazelwood (2005) identified common patterns of post‐smolt (first‐year marine) growth among salmon stocks around the UK and Ireland, suggesting these were likely to be influenced by a mixture of environmental processes operating throughout the post‐smolt period and possibly indicative of shared trends in sea survival. However, no data were available in this study to ex‐ plore links between smolt survival and post‐smolt growth rate. Observations that marine survival trends are correlated among some geographically separated rivers in both the NE Atlantic (Crozier & Kennedy, 1993; Friedland et al., 1998a) and NW Atlantic (Friedland et al., 1993) imply that major regulating factors operate when stocks mix and utilise a common shared habitat in the first autumn at sea. However, analysis of long‐term catch data from Norway and Scotland has dem‐ onstrated high levels of homogeneity at small, local scales, but marked heterogeneity at larger scales (Vøllestad et al., 2009). This anlysis suggested a trend of healthy popu‐ lations in the north and decline, as evidenced by falling catches, in the more southerly stocks. It was speculated that northern Norwegian populations were closer to cool oceanic regions, which are likely to be highly productive. In contrast, southern popu‐ lations must undergo much more extensive migrations through larger areas of the ocean to reach productive northern feeding areas, leaving them more vulnerable to mortality factors. However, these studies do not necessarily support the hypothesis that growth influences natural survival (Crozier & Kennedy, 1999). Further, while there appear to be critical periods for marine mortality, differences have been indi‐ cated between European and North American stocks of Atlantic salmon. The survival rates of 1SW and 2SW salmon from two European stocks ‐ the Figgjo (Norway) and River North Esk (Scotland), both of which discharge into the North Sea

18 | ICES SGBICEPS REPORT 2010

– were found to be correlated both within‐ and between‐stocks (Friedland et al., 1998a). This coherence in recruitment pattern from non‐neighbouring stocks suggests that survival effects act on the broad spatial scale or when stocks are mixed. Further, survival was found to be positively correlated with the area of 8‐10oC water in May. An analysis of SST distributions indicated that when cool surface waters dominate the Norwegian coast and North Sea at this time salmon survival was poor, but when the 8oC isotherm extended along the Norwegian coast during May, survival was good. Post‐smolt growth increments for returning 1SW fish also showed that en‐ hanced growth was associated with years when temperatures were favourable, in turn resulting in higher survival rates (Friedland et al., 2000). The evident link be‐ tween growth and survival suggests that growth‐mediated survival mechanisms (e.g. predation) are the dominant source of recruitment variability and recent work (Fried‐ land et al., 2009) has reinforced this view. Further, similar fluctuations in survival for both 1SW and 2SW salmon suggest that the possible contribution of variable matura‐ tion can be discounted. For North American stocks, correlations have also been demonstrated between sur‐ vival and thermal conditions, with similar trends in return rates observed for North American rivers over a broad geographical range, consistent with factors acting on fish when they are mixed and utilising a shared habitat. Associations have also been demonstrated between the increased summer growth of 2SW fish and increased re‐ turns of 1SW fish and a higher 1SW fraction (Friedland & Haas, 1996). Friedland et al. (1993) showed that the distribution of the winter habitat in the Labrador Sea and Denmark Strait was critical for post‐smolt survival. However, it was recognised that the mechanisms for this correlation were somewhat obscure since the observations conflicted with conventional thinking that the spring period is more important – i.e. when the smolts first migrate to sea. Subsequent investigations (Friedland et al., 2003b) compared thermal conditions in potential post‐smolt nursery areas with the pre‐fishery abundance of North American stocks and found that stock size was nega‐ tively correlated with mean SST during June. The results suggest that post‐smolt sur‐ vival is negatively affected by the early arrival of warm ocean conditions in the smolt nursery area and indicates that June conditions (the first month at sea for most stocks in the region) are pivotal to survival.

Hatchery-reared salmon Many salmon stocks are supplemented by salmon released from hatcheries. In gen‐ eral, wild fish are reported to survive significantly better than hatchery‐reared fish (e.g. Jonsson et al., 2003). Recent studies have also indicated that hatchery fish seem to be subject to other mortality events, above those experienced by wild fish (Peyronnet et al., 2007). Modeling studies (Peyronnet et al., 2008) indicated that survival of hatch‐ ery fish was primarily explained by coastal SSTs one year before the migration of smolts suggesting that warmer conditions during freshwater rearing appear to affect fitness at migration.

2.3 Climatic/oceanic factors The following section reviews certain environmental factors that have been linked with salmon recruitment; additional information in relation to environmental data sets is provided in Section 3.5.

ICES SGBICEPS REPORT 2010 | 19

North Atlantic Oscillation (NAO) The NAO is the dominant atmospheric process in the North Atlantic throughout the year. It accounts for more than one third of the total variance in sea‐level pressure and represents the large‐scale shift in atmospheric mass between a ‘high‐index’ pat‐ tern, characterised by an intense Iceland Low with a strong Azores ridge to its south, and a ‘low‐index’ pattern in which the pattern is reversed. The pressure difference between these two areas is the conventional index of NAO activity (Dickson & Tur‐ rell, 1999). The NAO has exhibited considerable long‐term variability, which appears to be am‐ plifying with time, although the index has been weak and variable over the past dec‐ ade. The 1960s exhibited the most protracted and extreme negative phase of the Index; the late 1980s – early 1990s experienced the most prolonged and extreme posi‐ tive phase. Changes in the NAO Index produce a wide range of physical and biologi‐ cal responses in the North Atlantic, including effects on wind speed, evaporation and precipitation, SST, ocean circulation, storminess, and the production of zooplankton and fish recruitment (Dickson & Turrell, 1999). There is thought to be a link between salmon marine survival and the NAO. There are various mechanisms by which this might occur through: (a) a positive NAO is linked to lower transport through the Faroes‐Shetland Channel with important effects on mixing processes (Parsons & Lear, 2001) and recruitment of other fish species such as cod; (b) a positive NAO is correlated with high SST and reduced mixing and zoo‐ plankton abundance in the NE Atlantic (Beaugrand & Reid, 2003) and salmon growth has been negatively correlated with SST (Friedland et al., 2005); (c) a positive phase of the NAO has also been reported to have had a negative effect on catches from the River Foyle (Boylan & Adams, 2006). Peyronnet et al. (2008) directly link the NAO signal to salmon survival (rather than catches). Davidson & Hazlewood (2005) have demonstrated a strong correlation between the NAO index and patterns of post‐smolt growth for various 1SW salmon stocks in the UK. Peyronnet et al. (2008) have also explored the influence of climate and oceanic conditions on the marine survival of 1SW Irish salmon using Generalised Additive Models and have demonstrated a link between a positive phase of the NAO, along with low plankton abundance and high SST in the NE Atlantic with a decrease in salmon survival. The NAO is also known to affect freshwater ecosystems via effects on temperature, rainfall and wind speed, highlighting the over‐riding influence of the NAO as the dominant atmospheric process in the North Atlantic and one which appears to serve as a general surrogate for a number of climatic effects operating over land and sea (Dickson and Turrell, 1999). For example, Jonsson et al. (2005) reported a positive cor‐ relation between NAOI and water temperature & discharge in the River Imsa in Norway during winter 1976–2002, indicating a significant oceanic influence on winter river conditions. Thus, while the global decline in abundance of Atlantic salmon is usually assumed to stem mainly from changes in the marine environment (Friedland, 1998), the potential for large scale processes, such as the NAO, to affect both freshwa‐ ter and marine environments together may mean that the influence of freshwater fac‐ tors in this decline have been understated (Crozier & Kennedy, 2003). Davidson & Hazlewood (2005) have demonstrated that year‐to‐year variations in both river and sea surface temperatures around England and Wales, and the NAO Index were highly synchronous. Further, Peyronnet et al. (2008) have shown that the NAO in the winter, prior to smolt migration, explained 70% of the deviance in marine

20 | ICES SGBICEPS REPORT 2010

survival of wild 1SW salmon from Ireland. This suggests that the NAO may be affect‐ ing pre‐smolts in freshwater with knock‐on effects for the fitness of fish during their early marine migration, although various mechanisms might apply. Warmer condi‐ tions during the juvenile rearing period may be detrimental to the future survival at sea by affecting overall fitness, or the match between the migration time and the op‐ timum migration period. Long‐term variation in catches for two rivers within the northernmost distribution area of the species has been related to mean SSTs in the Kola section of the Barents Sea in July (Niemelä et al., 2004). However, no link was identified for these stocks be‐ tween NAO and catches. In contrast to most of the North Atlantic area, catches in these rivers demonstrate no consistent recent trend of declining abundance. The Atlantic Multidecadal Oscillation (AMO) provides an alternative measure of At‐ lantic climate variability to the NAO. Information is provided in Section 3.5.

Gulf Stream North Wall (GSNW) The latitudinal position of the Gulf Stream is generally defined as the landward edge of the Gulf Stream (east coast of USA), and is referred to as the North Wall. The posi‐ tion is indicated by isotherm gradient data and is derived from satellite, aircraft and surface observations. This has also been used as an exploratory variable in relation to the changing abundance of Atlantic salmon (Crozier et al., 2003). However, while a variety of relationships between marine conditions (GSNW, NAO & SST) and sur‐ vival and abundance of stocks were indicated, both at single river and wider levels, these were not always consistent or intuitively correct, and the number of significant relationships was little greater than expected by chance.

Mixed Layer Depth (MLD) The MLD describes a thin homogeneous surface layer mixed by the wind which has consistent temperature & salinity. The thickness of the MLD varies and reflects the level of oceanic stratification. Warmer spring temperatures result in a shallower MLD and promote production since nutrients become trapped and phytoplankton can de‐ velop. This link between MLD and plankton productivity has also been related to the marine survival of coho salmon (Hobday & Boehlert, 2001). Peyronnet et al. (2008) found that the depth of the MLD in June had a significant con‐ tribution to models of marine survival for Irish salmon, with June MLD < 25m being generally positively associated with survival. Peyronnet et al. (2008) further noted that the timing of the formation of the MLD varied between years and suggested that delays in the formation of a shallow mixed layer could affect salmon survival. This might occur through reduced primary productivity, or a mismatch between the tim‐ ing of plankton abundance and the presence of post‐smolts. It was also suggested that since post‐smolts are distributed close to the surface, a shallow mixed layer might contribute to better survival through keeping their prey close to the surface.

Temperature Temperature has been identified as an important factor influencing growth and maturation of Atlantic salmon (Scarnecchia, 1983; Scarnecchia et al., 1991). However, it has also been suggested that thermal conditions can influence maturation inde‐ pendently of growth by influencing migration patterns (Friedland et al., 1998b). For example, temperature may directly affect the physiology of post‐smolts through in‐ creased swimming requirements as a response to migration cues or to avoid unfa‐

ICES SGBICEPS REPORT 2010 | 21

vourable thermal conditions (Salminen, 1994). This assumes post‐smolts have very specific thermal preferences and would seek to locate optimum temperatures during the first months at sea. A number of authors have noted a negative correlation between the growth of salmon and SST, affecting fish in both the NE and NW Atlantic (Peyronnet et al., 2007; Friedland et al., 2003b) and the Pacific (Wells et al., 2008). Because growth in sal‐ monids typically increases linearly with water temperature up to an optimum rate given an adequate food supply (Brett, 1979), and there is an evident link between size and survival, this presents something of a dilemma in making sense of the observa‐ tion that warmer conditions are associated with poorer survival, at least for fish which are experiencing temperatures which are at or below the optimum for growth. This possibly suggests a shortage of food ‐ under warmer conditions this could result in higher mortality rates, possibly due to an increasingly fragmented distribution of prey exacerbated by the effects of higher temperatures on fish’s metabolism. Fried‐ land et al. (2003b) explored four hypotheses: how ocean conditions may affect growth of post‐smolts, whether variations in climate have altered predation pressure or prey abundance, and how migration might be affected by variations in climate. Having reviewed the evidence, and given that post‐smolts represent a relatively minor com‐ ponent of the food web, they suggested that post‐smolts may not be affected by fluc‐ tuations in predator and prey levels as much as they are controlled by the individual demands of swimming and migration. Strong migrational behaviours that select for particular preferred temperature ranges could impose significant energetic costs on fish as they seek optimal temperatures. Davidson & Hazlewood (2005) have reported increasing trends in temperature for rivers and coastal waters in England over the last 20–40 years. These are consistent with warming trends across the north Atlantic (Hughes & Turrell, 2003) and globally (IPCC, 2001). Temperature fluctuations among river and sea sites were highly syn‐ chronous, suggesting large‐scale climatic processes are influencing both freshwater and marine environments simultaneously. However, for both rivers and coastal wa‐ ters there were geographical differences in warming rates with southerly sites warm‐ ing faster than northerly sites. Rises in temperature have been linked to marked declines in the status of salmon stocks in parts of southern England (Solomon & Lightfoot, 2007). For example, in some instances ‐ during very warm summers ‐ high estuarial temperatures appear to deter adult salmon from entering freshwater, or may even prove lethal (Solomon & Lightfoot, 2007). Given these common patterns in freshwater and marine temperature change, and the importance of temperature to the biology of fishes, Davidson & Hazlewood (2005) suggested that coherent trends in the age and growth of pre‐ and post‐smolt salmon might be expected. However, SST data were generally poorly associated with post‐ smolt growth patterns. Similarly, more direct measures of sea survival and adult abundance have shown a poor association with SSTs (Crozier et al., 2003). This does not mean that environmental temperature is unimportant for salmon in the sea, rather that other (perhaps more complex) factors have been the main cause of change and/or the temperature variables examined have poorly represented the true influ‐ ence of temperature on salmonid biology/ecology. A correlation between SST and survival has been observed for salmon stocks in the NE Atlantic suggesting temperature either directly affects growth or modifies post‐ smolt behaviour (Friedland et al., 2000). Beaugrand & Reid (2003) demonstrated highly significant relationships between SST in the NE Atlantic, Northern Hemi‐

22 | ICES SGBICEPS REPORT 2010

sphere temperatures, the NAO and long‐term changes in 3 trophic levels (phyto‐ plankton, zooplankton and salmon). A stepwise shift was identified that started after a pronounced increase in Northern Hemisphere temperature anomalies at the end of the 1970s. The biological variables showed a pronounced shift after ~1982 for euphausiids (decline), 1984 for total abundance of small copepods (increase), 1986 for phytoplankton biomass (increase) and the copepod Calanus finmarchius (decrease) and 1988 for salmon (decrease). Beaugrand & Reid (2003) concluded that regional temperature increases were an im‐ portant parameter currently governing the dynamic equilibrium of NE Atlantic pe‐ lagic ecosystems with possible consequences for fisheries. However, the results did not necessarily indicate a trophic cascade or bottom‐up control of salmon abundance. Nonetheless, it was suggested that the findings possibly open the way to the devel‐ opment of predictive tools, based on physical and plankton indicators, which might be used to assess future changes in the abundance and distribution of salmon. Scenarios for future climate change (UKCIP, 2002) suggest that warming trends are likely to be far more severe in the coming decades than those experienced in the past – even under the best case ‘low emissions’ scenario. Davidson & Hazlewood (2005) used temperature profiles forecast to the end of this century (UKCIP, 2002) to predict the growth rates of salmon in freshwater based on the model of Elliott & Hurley (1997). This indicated that for rivers in the south‐west and north of England, freshwater growth rates could generally improve under the ‘low emissions’ scenario but may fall below current levels under the ‘high‐emissions’ scenario as temperatures exceed optimum levels in the latter half of the century. On rivers in the south‐east of England (represented by the Thames) ‐ where warming is expected to be greatest ‐ declining growth rates may result, with adverse conse‐ quences for abundance and survival. Warming is only one aspect of climate change that might prove detrimental at ex‐ treme levels; expected increases in the frequency of summer droughts and winter floods could also adversely affect survival and abundance (UKCIP, 2002). Jonsson & Jonsson (2009) provide a recent review of the effects of climate change on Atlantic salmon and brown trout, with particular reference to water temperature and flow. The authors suggest that there is reason to expect a northward shift in the thermal niche of anadromous salmonids, with decreased production and population extinc‐ tion in the southern part of the distribution areas. Effects on stocks are expected to include earlier migrations, later spawning, younger ages at smolting and attainment of sexual maturity, and increased mortality. The effects of global warming on salmon in the sea are more speculative given our poorer knowledge of ocean processes and of the marine life of Atlantic salmon (Hughes & Turrell, 2003). However, given that temperature increases over land are expected to exceed those over the surface of the oceans, the changing state of condi‐ tions in freshwater may be the more important factor controlling future distribution and viability of the species, while change in the marine environment may be the main factor regulating stock productivity (Friedland et al., 2009).

2.4 Salmon in Freshwater Both density‐dependent and density‐independent factors regulate the abundance of Atlantic salmon in freshwater (Elliott, 2001). The carrying capacity for salmon juve‐ niles is limited and the response of juvenile abundance to intra‐specific competition is strongly compensatory (Elliott, 2001). Aside from the effects that density may have on

ICES SGBICEPS REPORT 2010 | 23

freshwater survival, other factors such as disease, predation and displacement by ad‐ verse environmental conditions may vary with the abundance of juveniles. There are many estimates of mortality rates for the juvenile stages of salmon because the freshwater life stages are readily accessible and are not subject to fisheries (Crozier et al., 2003). Thus, freshwater mortality is relatively well characterised and tends to be less variable than marine mortality (Friedland, 1998). Estimating mortality during the marine phase is more difficult because the returning adults have commonly been exposed to both natural and fishing mortality at sea. Fac‐ tors which act in freshwater, but may be manifest as variations in sea survival (e.g. Peyronnet et al., 2008) also need to be borne in mind in the context of marine survival. In addition, conditions experienced within the marine environment (e.g. prey avail‐ ability, growth rate, climatic conditions), clearly have an impact on the subsequent freshwater stages of the salmon through modifications to run timing and spawning. High densities of juveniles, among other factors, can affect juvenile growth and ulti‐ mately the size of migrating smolts (Jutila et al., 2006). As there is evidence of size‐ dependent survival at sea, the conditions experienced in freshwater are likely to di‐ rectly affect survival in the ocean. Environmental conditions in freshwater, inde‐ pendent of abundance, can also affect the physiology and energy reserves of salmon as they migrate to sea. Recent research has demonstrated very clearly that the conditions experienced within the freshwater environment are critical to the survival of salmon in the marine envi‐ ronment. For instance exposure of juvenile salmon to diffuse pollution in freshwater (e.g. pesticides) modifies run timing, development of the parr‐smolt transformation and the subsequent survival and behaviour of post‐smolts during the freshwa‐ ter/marine transition and once they enter the marine environment (Fairchild et al., 2002; Moore et al., 2003, 2007, 2008; Lower & Moore, 2007). However, poor water quality is simply one factor within the freshwater environment that may regulate and modify juvenile salmonid production. Habitat availability, feeding opportunities, and modifications to the thermal and flow regimes of rivers and streams may all interact to influence the “quality” of the smolt, which may compromise its ability to survive in the marine environment and successfully complete its extensive marine migra‐ tions.

Spawning Salmon spawn in river systems in the autumn and winter wherever suitable spawn‐ ing and nursery habitat areas exist; spawning habitat requirements are described by Armstrong et al. (2003). Female fish exhibit a degree of parental care burying their eggs in redds (Marshall et al., 1998). Fish typically spawn earlier in more northerly areas and in the upper reaches of rivers where water temperatures are lower and later in more southerly areas and in the warmer, lower reaches of rivers. Although the spawning season may last only a few weeks, fish may enter the river at any time and some fish can spend over a year in freshwater prior to spawning (Webb & Campbell, 2000). It is presumed that entering the river well in advance of spawning confers some benefit in terms of reproductive success. This might be due to: better survival in freshwater; easier upstream migration past falls and rapids prior to full gonad devel‐ opment; optimising opportunities for encountering specific, favourable river condi‐ tions for upstream migration; and avoiding migration through estuaries and the lower reaches of rivers during warmer periods when flows, temperature and water quality may be less suitable for migration (Anon., 1994).

24 | ICES SGBICEPS REPORT 2010

In many river systems, the earliest running fish in each sea‐age class tend to penetrate furthest into the river system. It is not clear whether this is because the fish happened to enter first or whether they specifically did so because they had farthest to go within the system. A genetic component to run timing has also been indicated (e.g. Stewart et al., 2002). Large body size enables female fish to produce more and larger eggs, compete more effectively for spawning sites and dig deeper redds. Large males are better able to compete for access to females and are preferred by females as mates (Quinn, 2005). However, the benefits of large size are likely to be balanced by the greater probability of mortality associated with spending more time at sea and possibly by difficulties in accessing spawning sites in smaller streams. In each river, these factors are expected to drive local adaptations in age and size distribution (e.g. Jónasson et al., 1997). There is also evidence that females arrive on the spawning ground earlier than males (Dahl et al., 2004). This could be due to the fact that females are commonly older, and older individuals often migrate earlier. Another possibility is that females try to mo‐ nopolize preferred spawning sites, while males normally compete only for access to a specific female. It is also known, however, that pheromones released by female salmon attract male fish and help to synchronise spawning events (Moore & Waring, 1996). Spawning date seems to have evolved to allow progeny to emerge at a locally appro‐ priate time in the spring relative to the average temperature regime that occurs dur‐ ing incubation (Quinn, 2005). However, there can be significant variation in the time of spawning among salmon within a river system (Webb & McLay, 1996), and there is some evidence to suggest a recent shift in the timing of spawning, in some rivers at least (W. Riley, pers comm). The temperatures experienced by female salmon in the months before spawning can have a major effect on egg quality and subsequent survival. Elevated temperatures and temperature ‘spikes’ can have a negative effect on egg size and fertility rates (King et al., 2003, 2007), affect physiological pathways (Watts et al., 2004) and ovula‐ tion (Taranger & Hansen, 1993).

Fecundity Fecundity is another trait that is highly variable both within and among stocks. Lar‐ ger fish typically produce more and larger eggs (Thorpe et al., 1984; Jonsson et al., 1996). Thus, absolute fecundity varies greatly among individual fish relative to their size at maturity. Relative fecundity (eggs per kg total egg mass) varies much less, typically by a factor of 1.5 to 2.0 within a population (Klemetsen et al., 2003). Eggs are large and incubation relatively long, with embryos relying on endogenous food reserves for a lengthy period before first feeding (Marshall et al., 1998). There is a reported trade‐off between egg size and fecundity (Jonsson et al., 1996) ‐ fish may ei‐ ther spawn large and few eggs or small and many eggs. The faster that fish grow in freshwater before smolting, the smaller their relative egg size on reaching maturity, This has been explained as a phenotypic response to the potential growth opportuni‐ ties in the natal river and assumes that the freshwater feeding conditions experienced by the parents as juveniles is a good predictor of what their offspring would experi‐ ence.

ICES SGBICEPS REPORT 2010 | 25

Sexual maturity in parr The sexual maturity of salmon parr is common for males. Development is thought to have both heritable and environmental components. Therefore, the proportion of ma‐ ture male parr can vary considerably between rivers, between river stretches and be‐ tween years depending on stock‐specific characteristics and growing conditions (Dalley et al., 1983; Bagliniere & Maisse, 1985; Hutchings & Jones, 1998), but usually increases with parr age (Dalley et al., 1983). O’Connell & Ash (1993) also reported a higher incidence of precocious parr in lacustrine habitats compared with fluvial habi‐ tats in the same river catchment. Possible environmental triggers for precocious male parr are fast early growth in fa‐ vourable conditions (e.g. Thorpe, 1990), with high seasonal fat content possibly the most important precondition for maturation (Rowe & Thorpe, 1990; Rowe et al., 1991; Simpson, 1992). Prevost et al. (1992) found that large size after the first year and short winters favoured maturation of 1+ male parr; and that the incidence of maturing 1+ males was greater where high densities of parr were recorded in the second season of growth. In northern (sub‐arctic) Atlantic salmon rivers the reported maturation rates of male parr are low (0–25%, typically ~10%) (Elo et al., 1995; Heinimaa & Erkinaro, 2004; Hutchings & Jones, 1998) compared with mean reported proportions among 1+ parr elsewhere – 57% in Newfoundland, 53% in the Maritime Provinces and 22% in Que‐ bec (Hutchings & Jones, 1998); 28–52% in east coast rivers in the U.S.A. (Whalen & Parrish, 1999); 0–100% in France (Bagliniere & Maisse, 1985). The short growing sea‐ son, poor seasonal growth and long winters consuming body lipid reserves may ex‐ plain this finding (Heinimaa & Erkinaro, 2004). In New England, U.S.A, where maturation of male 0+ parr is typically rare (~5%) Letcher & Terrick (1998) reported a high incidence of 0+ parr maturation (74%) fol‐ lowing a massive, localised flood that appeared to be the result of increased growth of parr following the flood. They suggested that such strong environmental distur‐ bances can alter the direction and timing of salmon life‐histories by influencing community structure and growth opportunity. It has also been suggested that maturity in male parr will be evolutionary stable only if adult mortality is high (Myers, 1984), and that over‐fishing could eventually elimi‐ nate anadromy in male members of a population (Myers & Hutchings, 1987). If true, male precocity should be expected to be less common in lightly exploited stocks. Ri‐ ley & Power (1987) carried out a study to test this hypothesis using data from land‐ locked and anadromous stocks in Canada. They concluded that high post‐smolt mortality in anadromous stocks is conducive to male parr maturity at an early age, and suggested the lower proportion of maturing male parr in landlocked stocks was related to competition among males for mates and the smaller size of the spawning adult landlocked salmon. The incidence of mature female parr is rare (Fleming, 1996), with only a few recorded incidences in anadromous Atlantic salmon (e.g. Moore & Riley, 1992). There is no evidence to suggest any change.

Egg and alevin development Salmon spawning substrates vary in composition (Armstrong et al., 2003). Spawning redds generally cover areas of 1–11 m2 (Bardonnet & Bagliniere, 2000) and consist of a ‘pot’ or ‘pit’ and a ‘tail’ (Crisp & Carling, 1989). The pot is the deepest excavated part at the upstream end of the redd, where most eggs are deposited. The depth to which

26 | ICES SGBICEPS REPORT 2010

eggs are buried in spawning gravels is related to female fish length (e.g. Crisp & Car‐ ling, 1989) and typically in the range 15–25 cm. The eggs are buried to protect them from light, predators and high water flow, which can result in mortality due to me‐ chanical shock (Bardonnet & Bagliniere, 2000). Egg burial depth also has some influ‐ ence on rate of development (temperature) and on the likelihood of washout, asphyxiation or exposure during low flows (Crisp, 1996). Egg development is influenced strongly by temperature and this provides the best predictor of hatching time (Crisp, 1996). However, a number of other factors can in‐ fluence hatching time, including incident light, dissolved oxygen concentration, sub‐ lethal mechanical shock and low temperatures at the time of hatching (Crisp, 1996). A number of factors also affect the survival of the intragravel stages (eggs and alevins) of salmon (e.g. Crisp, 1996; Youngson et al., 2004; Lapointe et al., 2004). Tem‐ perature has a direct effect on the survival of eggs and can also influence the size of alevins at hatching through regulating the relative proportions of the yolk sac used for metabolism and tissue growth. Oxygen supply rate is also critical. Oxygen re‐ quirements vary at different stages of development and are further influenced by fac‐ tors such as egg size, temperature, the spatial arrangement of eggs within the redd and the velocity of intragravel water flow (Crisp, 1996; Youngson et al., 2004). Other factors affecting egg and alevin survival include the gravel composition, stream bed conformation and hydraulics, patterns of discharge and mechanical shock (Crisp, 1996). A range of factors thus influence the survival of salmonid eggs and alevins, and these can interact in a complex manner (Crisp, 1996).

Fry emergence and dispersal The period of emergence and establishment of feeding territories is a time when mor‐ tality can be very high, and during which the strength of a cohort may be established, and can be regarded as a critical period (Armstrong et al., 2003). The timing of fry emergence in salmon is influenced by environmental conditions during egg devel‐ opment, most notably water temperature (e.g. Elliott & Hurley, 1998; Garcia de Leaniz et al., 2000). Thus the combination of spawning date and the temperature‐ dependent rate of egg and alevin development effectively determine when this will occur (e.g. Jensen et al., 1989). It is generally accepted that spawning dates are adapted to current thermal and flow conditions such that juvenile emergence timing is optimized as a result of selection pressures (e.g. Heggberget, 1988; Jensen et al., 1991). Marked changes in temperature or flow during early development may there‐ fore create a mismatch between emergence and environmental conditions, resulting in increased levels of early juvenile mortality (Jensen et al., 1991). Emergence from the redd is generally associated with the initiation of exogenous feeding (Dill, 1977), although fry may also disperse with varying amounts of unab‐ sorbed yolk sac, particularly during the early part of the dispersal period (Garcia de Leaniz et al., 2000). Emergence is typically clumped, with the majority of fry from a redd emerging over a short period (e.g. 80% over a 2–3 day period reported by Gustafson‐Marjanen & Dowse, 1983), and occurs at night (e.g. Gustafson‐Marjanen & Dowse, 1983). Clumped emergence at night is considered to be a predator avoidance tactic through avoidance of visual detection by predators and reducing the individual risk of predation through being in a large group (Gustafson‐Marjanen & Dowse, 1983). In Chinook salmon (Oncorhynchus tshawytscha), Field‐Dodgson (1988) noted that flow rate had a strong effect on emergence timing, while the effects of tempera‐ ture and light level were less.

ICES SGBICEPS REPORT 2010 | 27

After emergence from the gravel, some 0+ young‐of‐the‐year fry drift, while others, so called resident fry, remain close to the redd area and seek to establish and defend territories (Bujold et al., 2004). It has been suggested that those fry that fail to gain ter‐ ritories close to the redd will be displaced downstream and die (Elliott, 1984), but others suggest dispersal of viable fry may be an important means of maximizing use of available habitat, particularly where redds are widely spaced (Crisp, 1995), or of avoiding densely‐populated areas and obtaining individual territories elsewhere (Mikheev et al., 1994). As with emergence, dispersal occurs mostly at night (Kennedy & Strange, 1986), and is typically downstream of the redd, although upstream dispersal can also occur (Egglishaw & Shackley, 1973; Crisp, 1995). The most intensive dispersal of 0+ fry oc‐ curs during the two months after emergence (Beall et al., 1994), and Marty & Beall (1989) reported two distinct waves of dispersal, one soon after emergence and an‐ other 10 to 20 days later. The majority of fry typically remain close to the redd site ‐ less than 1 km downstream, with a high proportion within 50m (e.g. Egglishaw & Shackley, 1973; Kennedy & Strange, 1986; Marty & Beall, 1989; Crisp, 1995; Webb et al., 2001). Dispersal distances also appear to be affected by rainfall and thus flow rates (Crisp, 1995). Once settled, salmon appear to die in situ rather than move during the summer (Egglishaw & Shackley, 1980; Armstrong et al., 1994), although a further dis‐ persal may occur in autumn (Rimmer et al., 1983; Rimmer et al., 1984). Garcia de Leaniz et al. (2000) report marked variability in the developmental stage of fry at dispersal ‐ early in the dispersal period all fry had remnants of a yolk sac, but fry dispersing later in the season had no yolk sack left. Bujold et al. (2004) report that larger (but not older) fry tended to stay near the redd while smaller fry tended to drift/disperse downstream, probably as a result of intraspecific competition for food and space. Differences in the dispersal behaviour of emerged fry may also be related to gender; in masu salmon (Nagata & Irvine, 1997) found that male fry were more likely to remain near the redd and female fry to move downstream. After dispersal, Garcia de Leaniz et al., (2000) report that individual growth of fry is highly variable. Crisp (1995) found that the weight of 0+ fry in the autumn following dispersal was directly related to the distance dispersed and inversely related to popu‐ lation density. Two hypotheses were proposed to explain this observation: (a) that the fry that dispersed the longest distances tended to be of above average size at the time of dispersal, and (b) that the fish which dispersed farthest were of average or below average size at the time of dispersal, but, once dispersed, encountered less competi‐ tion and were able to grow more rapidly. It was noted that hypothesis (a) was con‐ trary to the view that displaced fish are typically smaller than the population average. Metcalfe & Thorpe (1992) demonstrated that earlier‐feeding fry were dominant over their later‐feeding siblings, despite not being any larger, but soon established and maintained a size advantage. This led to an increased probability of early‐feeding fish migrating to sea as one‐year‐old smolts (rather than 2 or more) (Metcalfe & Thorpe, 1992), indicating the link between growth rates and future life‐history traits (Thorpe, 1989). It was subsequently demonstrated (Metcalfe et al., 1995) that the correlation between date of first feeding and dominance was a consequence of early‐feeding fry having relatively higher metabolic rates (Metcalfe et al., 1995). However, environ‐ mental factors such as temperature or flow can also have an important influence on fry survival. Good et al. (2001) reported that size‐selective mortality of salmon fry was relatively weak and directed towards the smaller fry in the population during a drought year (1995), but that during a subsequent wet year (1996), selective mortality

28 | ICES SGBICEPS REPORT 2010

was relatively strong and directed towards the larger fry from the same population. They concluded that hydoclimatic events can select against either small or large fish and is a key determinant in the mean size of juvenile salmon fry at the end of their first summer.

Growth in freshwater Aprahamian et al. (2008) reported relatively stable freshwater growth of salmon parr on the River Dee, Wales between the late 1930s and mid‐1980s (using mean lengths back calculated from adult scales). However, by the end of the 1990s they reported that juvenile salmon were, by the end of their first and second year, respectively, ~60% and ~ 19% larger, on average, than between the late 1930s and mid‐1980s. Pos‐ sible explanations proposed included: a reduction in density dependent processes; an increase in river temperature affecting both parr growth rate and timing and size of alevins at emergence; and changes in agricultural practices in the late 1970s resulting in enhanced invertebrate production so that food was no longer limiting growth. Rivot et al. (unpublished data) found a significant increase in the length of one‐year old juveniles between 1973 and 2002 in Brittany and Lower Normandy. Stable isotope analysis on archived scales of juveniles showed that the C13 ratio in scales also in‐ creased. They suggested that the changes in juvenile growth were in response to warming (climate change) and/ or a change in primary production. Jonsson et al. (2005) reported a positive correlation between growth rate of parr in their first year and river temperature, flow and NAOI during the winter of egg incu‐ bation, but only NAOI was significant when cross‐correlating the two series using a time difference of 1 year. Jensen (2003) analysed the effects of altered water temperature on parr growth, due to hydropower regulation of the River Alta, northern Norway. After hydropower development, water temperature decreased 1–2°C during June, July and the first half of August and parr growth rates decreased; later in the season when water tempera‐ ture increased by up to 3°C parr growth increased. However, only minor overall changes in annual growth rates were observed. In contrast, mean annual size‐at‐age of juvenile salmon decreased in the Northwest Miramichi and Southwest Miramichi Rivers in the Maritime Provinces of eastern Canada between 1971 and 1999 (Swansburg et al., 2002). Fork lengths of parr were significantly and negatively associated with spring air and water temperatures. Therefore, increases in air and water temperatures as predicted from climate change models may adversely affect growth of juvenile salmon parr, reducing overall pro‐ ductivity. Davidson & Hazelwood (2005) used the model of Elliott and Hurley (1997) to predict pre‐smolt (freshwater) growth rates for the Rivers Wye and Dee, Wales. These sug‐ gested that growth rates had improved on these rivers in response to increasing tem‐ peratures, but not significantly so. These relatively stable growth predictions contrast a marked decline in mean smolt age observed on some rivers (Severn, Wye and Dee), which began around the early 1980s but has proceeded at different rates thereafter. It was suggested, therefore, that factors other than, or in addition to, temperature were promoting faster growth in pre‐smolt salmon on these rivers and, as a result, a de‐ cline in the mean smolt age. Gurney et al. (2008) also reported changes in age at smolting (and growth) consistent with changes in temperature regime, but indicated that changes in early density‐dependent mortality, rather than the physiological ef‐ fects of temperature, explained the large attenuation between a dramatic fall in

ICES SGBICEPS REPORT 2010 | 29

spawner numbers and a relatively minor diminution in total smolt production in the Girnock Burn, Scotland. In general, it is anticipated that parr would be expected to grow faster (and hence probably smolt at a younger age) with increased temperature, up to the optimal tem‐ perature for growth (16–19°C) (numerous authors). With further increases in tem‐ perature, growth would be reduced (resulting in probable increased age at smoltification). However, at temperatures ranging from 22–24°C juveniles seek ref‐ uge from thermal stress (Cunjak et al., 1993), at 28°C they die in 7 days and at 33°C they die in 10 minutes (Elliot, 1991).

Parr movement/ migration Downstream movement of salmon parr during the autumn has been recorded in populations in both North America and Europe (Youngson et al., 1983; Cunjak & Chadwick, 1989; Riley et al., 2002). Studies of the migration patterns of Atlantic salmon parr tagged with passive integrated transponder (PIT) tags on the River Frome, a chalk stream in southern England, have demonstrated that a substantial proportion (25% of the spring smolt run in absolute terms) of the population can mi‐ grate downstream during the autumn with the peak movement occurring during Oc‐ tober and November (Pinder et al., 2007). The ecological drivers for autumn migrations of salmon are unknown (Riley et al., 2008), although a number of mechanisms have been proposed. These include dis‐ placement of subordinates by dominant fish (Bjornn, 1971; Mason, 1976), the re‐ quirement for juvenile salmon to migrate to more suitable freshwater habitats (Riddell & Leggett, 1981; Huntingford et al., 1992; Riley et al., 2008) or the requirement for mature male parr to locate mature female adult salmon to maximise reproductive success (W. Riley unpubl. data for the Kielder Burn, England). In some cases the au‐ tumn migrations of juvenile parr have been associated with elevated stream dis‐ charge (Youngson et al., 1983; A. Pinder, personal communication, for the River Frome, England; W. Riley, unpubl. data for the River Ceiriog, Wales). Such move‐ ments have also been shown to comprise predominantly precocious male parr (Buck & Youngson, 1982; W. Riley, unpubl. data for the Kielder Burn, England), but can also comprise fish of both sexes (Riley et al., 2008). Recent studies on the River Frome have indicated that the autumn migrants, includ‐ ing those that subsequently move to and reside within the tidal reaches of the river during the winter months, are not physiologically adapted to permit permanent, or early, entry into the marine environment (Riley et al., 2008). Frequency histograms of seasonal downstream movements of juvenile Atlantic salmon in the UK suggest a dual peak in the autumn/winter migration, the first oc‐ curring in early autumn, the second during the spawning season for the river system in question (Pinder et al., 2007; Riley et al., 2002; Riley, 2007; W. Riley, unpubl. data for the River Ceiriog, Wales). Although there is often no information from these studies on the sex composition of the migrants, it is speculated that the later migration may involve mature male salmon parr and be related to reproductive activity. These fish are also older than those migrating during the autumn (Riley, 2007). The extent to which the timing and relative magnitude of these migrations might vary between rivers or over time is unclear. Whilst adult returns have previously been reported from parr leaving highland tribu‐ tary spawning streams during the autumn (Youngson et al., 1994), Riley et al. (2009) recently reported adult returns from parr that were known to have migrated down‐

30 | ICES SGBICEPS REPORT 2010

stream to the lower river/ tidal reaches thus confirming the importance of these areas as habitat utilised by juvenile salmon (Cunjak & Chadwick, 1989; Riley et al., 2008) before they undergo the parr‐smolt transformation.

Smolt Age The age of smoltification of Atlantic salmon juveniles is closely associated with growth rates: fast growing populations smoltify at younger ages (Swansburg et al., 2002). Over the geographic range there is a significant negative correlation between the age of smoltification and an index of growth potential based on degree days and photoperiod length (Metcalfe & Thorpe, 1990). Aprahamian et al. (2008) reported that juvenile salmon on the Welsh Dee were larger at the end of the 1990s than between the late 1930s and mid‐1980s. This was reflected in a change in the age composition of smolts, with the mean smolt age declining from ~ 2 years prior to the 1980s to ~1.6 years in the late 1990s. Similarly, Cragg‐Hine et al. (2006) reported a marked increase in the proportion of 1+ smolts on the Welsh Dee (from ~5% in the period to 1949, to almost 40% in the decade to 2002). They con‐ cluded that as the Welsh Dee had become colder during the period April to October (post‐1964 due to flow regulation schemes) factors other than climate change were likely to be involved. A decline in mean smolt age has been similarly observed in other British and French populations of salmon (Bagliniere et al., 2004). Rivot et al. (unpubl.) also reported a significant reduction in the mean age at smoltification in Brittany and Lower Nor‐ mandy and suggested this reduction may be due to changes in juvenile growth in response to warming (climate change) and/ or to a change in primary production, yet acknowledged that selective fishing on late maturing fish could also be responsible for a decrease in the mean age at sexual maturity. Davidson & Hazlewood (2005) re‐ ported an overall decline in the mean smolt age of 1SW and 2SW salmon returning to the Dee and Wye (Wales) and Severn (England) since the 1960s. However, the onset of this decline did not appear until the 1980s, and was notably less marked on the Severn. On the River North Esk (Scotland) there has been a slow but steady decline in mean smolt age over the period 1971–2004 (see Figure 4.9.1 and discussion below) and on the R. Frome (England) the decline in smolt age has resulted in the near elimination of 2‐year old smolts from the population (Section 4.3). Englund et al. (1999) reported considerable inter‐annual and tributary specific differ‐ ences in the age structure of smolts in the River Teno, Finland between 1989 and 1995. Jonsson et al. (2005) reported a positive correlation between river temperature and the proportion of salmon cohorts smolting and migrating to sea at age‐1 and suggested that long‐term effects of climate (e.g. NAOI) during early development may be more important than generally recognised.

Smolt size Smolt size can vary widely among populations. Within populations, smolt size is flexibly dependent on growth rate – smolt size typically increases with age. Klemet‐ sen et al. (2003) noted that large smolts (mean lengths typically >20 cm) tend to occur in rivers that enter a cold ocean, such as rivers in parts of Quebec, while small smolts (averaging 12–13 cm) can occur in cold rivers flowing into a relatively warm ocean, such as the glacier‐fed rivers along the west coast of Norway. However, individual river, or even tributary, characteristics, and year to year fluctuations, may play as large a part in determining smolt size as any such broad geographical trends.

ICES SGBICEPS REPORT 2010 | 31

Jutila et al. (2006) reported a significant negative regression between the annual mean smolt size and the density of wild >1 year parr in the previous autumn in a northern Baltic river, but not between the annual mean smolt size and age. They hypothesised that the increased density of wild >1 year parr may have contributed to the decreased smolt size since the 1990s, and that the reduced size of wild smolts could result in reduced post‐smolt survival.

Smolt run timing It is generally accepted that there are two distinct processes involved in controlling the downstream movement of salmon smolts in fresh water. First, the physiological development of migratory readiness ‐ whereby the juvenile salmon undergo the physiological and morphological changes associated with smoltification (migration disposition), and second, the environmental signals (external releasing factors) which stimulate downstream movement once smoltification has been completed (Bagger‐ man, 1960; Solomon, 1978). Changing day length has been suggested as the ultimate trigger to smolt migration, with river conditions such as temperature, discharge, turbidity and nightfall serving as proximate triggers controlling the smolt run in different rivers (Jutila et al., 2005; McCormick et al., 1998; Riley, 2007; and many others). Thus the timing of smolt mi‐ gration varies with latitude, with southern populations moving out to sea earlier than northerly ones. Migration is correlated also to body size, with larger smolts typically migrating earlier, and also appears to have a genetic component (Stewart et al., 2006). Jutila et al. (2005) found that the onset of the smolt run was positively correlated with river temperature during 1972–2002 in a river flowing into the northern Baltic Sea; a rise in water temperature above 10°C being the main proximate environmental trig‐ ger. In addition, they found that the duration of the main run was shorter in the years when the onset of the smolt run was delayed. However, Zydlewski et al. (2005) found that temperature experience over time is more relevant to initiation and termination of downstream movement than a temperature threshold. There are concerns that variations in climate might modify the run timing of smolts. As freshwater conditions are more responsive to air temperatures than the sea, and smolt migration is cued by water temperatures, mismatches relative to marine condi‐ tions might result and affect return rates. For example, Staurnes et al. (2001) indicated that smolts transferred to seawater in which the temperature gradient was more than 4–6°C had poorer seawater challenge performance. Evidence from some rivers in Newfoundland (Crozier et al., 2003) indicated a relationship between run timing and subsequent survival at sea, indicating a possible optimum window of opportunity for smolts entering the marine environment. However, this relationship was not consis‐ tent among rivers in the area. Unusually early runs of smolts were observed on the River Bush in the late 1990s and a significant linear relationship determined between smolt migration date and marine survival, suggesting that smolt cohorts which migrated earlier had poorer survival (Crozier et al., 2003). Average SST taken from a coastal station 25 km north west of the river mouth indicated a temperature difference of 2.4oC with that of the river mouth during the years of poor survival, whereas the differential was only 0.3oC during ear‐ lier years when marine survival was higher (Section 4.1).

32 | ICES SGBICEPS REPORT 2010

Role of lacustrine habitat Klemetsen et al. (2003) have reviewed the importance of lacustrine habitat for juvenile salmon. It has been suggested that juveniles occupying such sites are displaced from preferred stream habitats. However, it appears that growth and survival is often bet‐ ter in lacustrine sites and that use of such habitats is of adaptive significance. In New‐ foundland, a positive relationship has been identified between mean smolt size and the proportion of lacustrine habitat, with some evidence also of better survival to adult for the larger smolts. Overall generation times have also been shown to be lower in lacustrine than fluvial systems. Although mean smolt age was similar in the two systems, there was a pre‐ ponderance of 1SW adults arising from lake origin fish, while fluvial systems pro‐ duced a significant component maturing at older sea‐ages. This was consistent with the view that given environmental conditions conducive to rapid growth, fish will mature as soon as developmentally possible and may explain why many Newfound‐ land rivers are dominated by 1SW fish. Studies in other countries have also indicated that faster‐growing juveniles and larger smolts favoured earlier age at first maturity (Nicieza & Braña, 1993 – northern Spain; Erkinaro et al., 1997 – Finland). Provisional results for the Frome and Dee suggest that the sex of smolts may be important to the strength of relationship between smolt size and age at maturity (Section 5.7).

3 Data sets

3.1 Data sources and requirements The exploration of biological characteristics for indicator salmon stocks and the pos‐ sible relationship of such metrics with environmental parameters and abun‐ dance/survival measures requires data from three distinct areas. First, time series of standard biological measures pertaining to particular salmon stocks are needed, ide‐ ally covering an adequate period to account for natural variability and to facilitate trend analysis. This process was originally initiated through WGNAS (ICES, 2008a) and was facilitated, at that time, by the production of a data entry spreadsheet for pertinent biological variables (e.g. mean annual length, weight, sea‐age of returning adults) to ensure the collection of time series information in a standardised format. Further substantial progress was made with this compliation of data at the first SGBICEPS meeting (ICES, 2009a) and at the most recent meeting (Sections 3.2–3.3). A second requirement is the identification and compilation of appropriate environ‐ mental datasets. The Study Group reviewed the types of information that would be available in taking forward exploratory analyses at its first meeting (ICES, 2009a), in particular relating to the marine environment. The Study Group recognised that the lack of a clear understanding of the distribution of salmon at sea remained a con‐ straint in this regard, and further recognised that specific requirements for environ‐ mental data or efforts to link these with changes in biological characteristics would need to be refined once clear hypotheses could be developed, for example in relation to specific stocks or stock complexes. Few further efforts were made to review envi‐ ronmental data sources at the second meeting (see Section 3.4) and no oceanographer was able to attend this meeting. However, a number of case studies presented at the second meeting examined changes in biological characteristics in relation to specific environmenmtal changes. Further details are provided in Sections 4 and 5. Finally, information on the relative performance of each available indicator stock, or stock complex, is required in terms of survival, mortality and abundance. The Study

ICES SGBICEPS REPORT 2010 | 33

Group examined the available data relating to these stock status variables at its first meeting (ICES, 2009a). Some further discussion of the stock status variables for spe‐ cific stocks is provided in Section 3.3 and analyses exploring relationships between biological characteristics and abundance are provided in Section 5.

3.2 Data on biological characteristics – update on developments The Study Group previously noted that there is a vast amount of biological data relat‐ ing to salmon stocks potentially available in pursuit of its objectives. Initial efforts to collate such data in a standardised from were made by WGNAS (ICES, 2008a) using an Excel spreadsheet developed by Tim Sheehan (USA and Chair of WGNAS at the time). Additional data sets that might be informative in addressing the Study Group’s terms of reference were collected in the run up to the first SGBICEPS meeting and preliminary analyses were completed by the Group (ICES, 2009a). The data entry spreadsheet requires various background information for each stock, as well as data for a range of biological characteristics. In summary, the fields are: Field Description Year Return year River Country Stock Complex e.g. NAC, northern & southern NEAC Latitude Latitude of river mouth Origin Origin of majority of spawning fish – hatch‐ ery/wild Size Approximated categorical description of the mean annual size of the spawning popula‐ tion Status Some standard estimate of stock status (e.g. count, trap catch) or index of abundance Status type Description of ʹstatusʹ variable Median run date Median date of the spawning run Mean run date Mean date of the spawning run Mean river age Mean river age of returning fish Mean sea‐age Mean sea‐age of returning fish Mean total age Mean total age of returning fish Proportion of run by sea‐age Proportion of the run by size/age grouping or by sea‐age Mean fork length Mean fork length (cm) of returning fish by size/age grouping or by sea‐age Mean whole weight Mean whole weight (kg) of returning fish by size/age grouping or by sea‐age Proportion female Proportion of returning fish that are female by size/age grouping or by sea‐age Proportion maiden spawners Proportion of returning fish that are maiden spawners by size/age grouping or by sea‐age Where data were requested by size/age grouping or by sea‐age, the preference was for information to be split between each sea‐age class – i.e. 1SW (1 sea‐winter), 2SW, 3SW, 4SW, 5SW and previous spawners (PS). Where such data were not available, information was provided for small/1SW and large/MSW (multi sea‐winter) catego‐ ries.

34 | ICES SGBICEPS REPORT 2010

A number of additional data sets were made available to the Study Group at, and after, the second meeting, and a number of the existing data sets were updated. This helped to extend the spatial and temporal coverage of the information available, for example, with the inclusion of some data for Baltic stocks. However, for a number of the new data sets, particularly those from Norway, a shorter time series was avail‐ able, typically only extending back as far as 1989. This had implications for the sub‐ sequent time series analyses. However, the Study Group previously indicated that data sets with less than 15 year time series should be included in the analyses where these provide greater spatial scale or allow comparison with other available datasets (ICES, 2009a). Table 3.2.1 summarises the latest information available to the Study Group. A full list of all those who provided data is provided at Annex 2. These data were further analysed by the Study Group with the additional assistance of Erik Pe‐ tersson (Sweden) and Jon Barry (UK (England & Wales)), who participated in the first meeting of the Study Group, but were unable to attend the second meeting. At the first meeting, the Study Group recognised that freshwater data had been largely ignored in their initial deliberations and recommended that the collection and collation of data sets should be extended to include freshwater stages (e.g. smolt age, smolt size, growth at age, etc.), not least since there was increasing evidence that freshwater influences might be instrumental to what subsequently happens in the sea. Such data, other than mean river age (derived from returning adult scales), have not been included in the data sets reported in Table 3.2.1. However, a number of case studies and analyses included in Sections 4 and 5 have incorporated analysis of juve‐ nile stages.

ICES SGBICEPS REPORT 2010 | 35

Table 3.2.1. Biological characteristics data provided for each stock (Y denotes data available for some or all of the time series) class

age ‐ sea

by

age ‐ age ‐ sea

spawners

sea by weight

date

length by Factor date

mean

run mean

fork whole run complex maiden Run female status

age series

age

Stock Country Stock Hatchery/Wild Time Prop. Prop. Mean Mean Condition Prop. Mean Latitude Stock Median River Sea NAC (N) Canada Western Arm Brook W 1971‐06 51.2YYYYYYYYYY Y Canada De la Trinité W 1980‐0949.4Y YYYYY Canada Saint‐Jean W 1981‐0948.8Y YYYYY Canada Middle Brook W 1975‐0548.8YYYYYYYYYY Y Canada Conne River W 1986‐0647.9YYYYYYYYYY Y Canada Miramichi W 1971‐0747.0Y YYYYYYY Y NAC (S) Canada Nashwaak W 1972‐0846.0YYYYYYY Y Y Canada St John (Mactaquac) W 1978‐0845.3YYYYYYYYYY Y Canada St John (Mactaquac) H 1978‐0845.3YYYYYYYYYY Y Canada La Have W 1970‐0844.4YYYYYYYYYY Y Canada La Have H 1972‐0844.4YYYYYYYYYY Y USA Penobscot H 1978‐0844.5YYYYYYYYYY Y N NEAC Finland/Norway Teno W 1972‐0770.8Y YYYYYYY Finland/Norway Näätämöjoki W 1975‐0669.7Y YYYYYYY Russia Tuloma W 1983‐0868.9YYYYYYYYYY Y Norway Vestre Jakobselv W 1989‐0870.1Y YYYYYY Norway Årgårdsvassdraget W 1992‐0864.3Y YYYYYY Norway Nausta W 1989‐0861.3Y YYYYYY Norway Gaula Sogn og Fjordane W 1989‐0861.2Y YYYYYY Norway Etneelva W 1989‐0859.4Y YYYYYY Norway Skienselva W 1989‐0859.1Y YYYYYY Norway Numedalslågen W 1989‐0859.1Y YYYYYY Norway Enningdalselva W 1990‐0858.6Y YYYYYY Norway Gaula W 1989‐0863.3Y YYYYYY Iceland (N&E) Laxa I Adaldalur W 1974‐08 65.6 Y Y Y Y Y Iceland (N&E) Hofsa W 1971‐08 65.4 Y Y Y Y Y S NEAC Iceland (S&W) Nordura W 1968‐08 64.6 Y Y Y Y Y Iceland (S&W) Ellidaar W 1949‐08 64.1 Y Y Y Y Y UK (Scot) N. Esk W 1981‐0856.7Y YYYY UK (NI) Bush W 1973‐0755.1YYYYYYYYYY Y UK (E&W) Lune W 1987‐0854.0Y YYYYYY UK (E&W) Dee W 1937‐0853.4Y YYYYYYY UK (E&W) Wye W 1910‐0751.6Y YYYYYYY UK (E&W) Frome W 1968‐08 50.7 Y Y Y Y France Bresle W 1984‐08 50.1 Y Y Y Baltic Sweden Kalix W 1980‐08 65.8 Y Y Finland/Sweden Tornionjoki W 1980‐09 65.5 Y Y Y Y Y Y Sweden Ume/Vindel W 1987‐08 63.8 Y Y Y Y Sweden Ume/Vindel H 1974‐0863.8YYYYYY Y Y

3.3 Data quality issues – caveats and limitations At the first meeting, the Study Group noted the importance of various factors that might affect data quality, and that these needed to be considered carefully before reaching conclusions about the relevance of any apparent links between different bio‐ logical variables and measures of salmon abundance or environmental variables (ICES, 2009a). The Study Group therefore recommended that data for all stocks should be accompanied by a full description of data sources and of the methodology used to record each variable to aid interpretation. Further, an email address of an ap‐ propriate contact person, able to provide a more detailed description of each data set, should be appended to the spreadsheet. In the run up to the second meeting, such information was therefore sought from all those who contributed data sets (see An‐ nex 2) and the responses collated and included in this report. Table 3.3.1 summarises

36 | ICES SGBICEPS REPORT 2010

the information provided and the following short section provides a brief description of the key points:

The stocks The majority of stocks for which data have been provided are regarded as wild popu‐ lations ‐ i.e. not subject to supplementary stocking from hatcheries as part of restora‐ tion, enhancement or mitigation schemes. However, a small number of the stocks are subject to stocking (Table 3.3.1). In the case of two rivers, the St John (Mactaquac) and La Have Rivers in Canada, entirely separate hatchery and wild data sets have been generated, and for the purposes of subsequent analyses these have been treated as separate stocks. For certain other stocks, varying levels of hatchery contribution have applied. In these cases, the data presented typically represent a composite of the wild and hatchery fish in the population.

Stock status The data sets provided for each country include a “status” column, which provides some measure of abundance. A wide range of stock sampling methods have been used to derive information on stock status, and these are variously listed as: adult traps, counting fences, catches (by rods, nets and both combined), counters, index trap nets and snorkel surveys (Table 3.3.1). For stocks in North America and a num‐ ber in the Southern NEAC area, the stock status variables are adjusted to take account of in‐river fishery catches to provide an estimated return to the river. However, for the majority of Northern NEAC stocks the status variable typically comprises the catch in freshwater by rods, or by rods and nets combined. For many rivers these data are compiled over fixed time intervals, but are not adjusted for effort. For the majority of stocks, no further adjustment was made to estimate returns to the coast (i.e. before coastal homewater fisheries), either because no such fisheries operated or due to the lack of robust information to make such estimates (e.g. stock‐specific ex‐ ploitation rates), but this applied in some cases. This highlights the widely differing approaches used for assessing the various stocks and highlights that changes in fishery regulations and trends in fishery catches will have influenced the robustness of the various abundance estimates. The Study Group noted these potential disparities between the measures available for different rivers and recognised that caution was therefore needed in their interpretation and use.

Sampling periods The Study Group previously noted that the extent to which the mean values included in the spreadsheets (e.g. run dates, weights, lengths, ages and sexes) truly reflect a particular stock would be affected by the period over which the data were collected in any year (ICES, 2009a). As indicated above, a wide range of sampling methods has been used. Some of these methods, typically traps and counting facilities, have the advantage that data can be collected throughout the entire year, although even these can be affected by issues such as poor trap efficiency during flood events or counter ‘downtime’. In contrast, fishery derived information (e.g. samples from net or rod fisheries) is restricted to the fishing season, and may have been subject to change over time. Table 3.3.1 indicates that just under half of the stocks for which data were provided reported that sampling proceeded throughout the year and could be considered rep‐ resentative of the entire run. Such sampling included the majority of the stocks in the NAC area and many in the NEAC Sothern area; information on run timing is almost

ICES SGBICEPS REPORT 2010 | 37

exclusively confined to these stocks. In the remainder of cases, data were typically sourced from fishery‐related sampling and were thus constrained by fishing seasons and potentially also influenced by temporal changes in the methods used and in the timing and level of fishing effort employed. However, for most such data sets, efforts have been made to provide data for fixed periods (e.g. June to August inclusive in respect of the Norwegian data sets) and/or to use logbook schemes or particular groups of fishermen to ensure consistency. Accordingly, the accuracy of the estimates of median and mean spawning run dates, average weights, lengths, ages and sexes will vary depending upon the source of the data and will need to be accounted for in any analysis.

Fish weight With respect to fish weight, the Study Group previously noted (ICES, 2009a) that data derived from net fisheries or sampling facilities close to the coast were most likely to represent the mass of fresh run fish while those derived from rod fisheries (and pos‐ sibly in some cases trap data) will be derived from a mix of fresh run and earlier run fish, which may have already lost some body weight compared to when they origi‐ nally entered freshwater. Table 3.3.1 indicates a reasonably even split between sam‐ pling in coastal and more upstream facilities, including fish captured by anglers. Thus, there are potential issues in comparing both weights and condition factors among stocks where the weight variable has been recorded from different sources. Further, there may be problems in interpreting weight trends within a stock using data derived from a rod fishery as the proportion of fresh to non fresh‐run salmon in the sample may vary throughout the time series.

Age data For the majority of rivers, age data for both adults and juveniles have been derived from scale reading, with scales mainly collected as part of targeted sampling pro‐ grammes. In some instances, scale samples have also been provided by anglers. Sam‐ ples have typically been taken throughout the run (see sampling periods above), but in some cases may reflect only part of the run – e.g. in Norway the scale samples were collected by anglers between the start of June and the end of August. The Study Group previously recognised that scale reading can be subject to potential differences in methodology and interpretation (ICES, 2009a), which might lead to possible incon‐ sistencies between stocks. However, the use of scale reading guidance, periodic workshops and interpretation by experienced ‘readers’ is generally considered to minimise such concerns. In the case of Iceland and one stock in the Baltic, sea‐age composition has been as‐ sessed from length and/or weight distribution data. Studies in Iceland have indicated little overlap in size between 1SW and 2SW fish (confirmed by earlier scale reading), with very few previous spawners or fish of older than 2SW occurring (Scarnecchia, 1983). This approach is thus considered to provide a robust alternative to analysis of individual scales. In respect of the river age variable, this has been determined in all but one case (where smolt scales were read) by the examination of scales from returning adults. The Study Group previously noted that such data might not accurately reflect the age composition of smolt cohorts, for example if different age/size classes of smolts are subject to differential rates of mortality in the sea (ICES, 2009a).

38 | ICES SGBICEPS REPORT 2010

Table 3.3.1. Summary of stock sampling procedures for the different biological characteristics data sets (see Key for abbreviations).

Stock complex NAC NEAC (N) NEAC (S) Baltic

River name Western Arm Brook Trinité la De Saint-Jean Brook Middle RiverConne Miramichi Nashwaak (Mactaquac) St John La Have Penobscot Teno Näätämöjoki Tuloma Jakobselv Vestre Årgårdsvassdraget Nausta SognGaula og Fjordane Etneelva Skienselva Numedalslågen Enningdalselva Gaula Laxa I Adaldalur Hofsa Nordura Ellidaar N. Esk Bush Lune Dee Wye Frome Bresle Kalix Tornionjoki Ume/Vindel Wild only (W) or combined wild & hatchery (W/H)? WWWWWWWW/HW/HW/HWWWWWWWWWWWWWWWWWW/HW W W WWWWW/H

NC/ Stock sampling procedure RC/ Co/ NC/ CFATSnCFCFIETCFCFATATFCFCATRCRCRCRCRCRCRCRCRCRCRCRCRCCoAT AT AT RC Co AT Co RC AT Stock status variable takes in-river fishery catches into account Y Y Y Y Y Y Y Y Y Y n/a n/a N n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a Y Y Y Y n/a N N N n/a n/a Returning Stock Estimate (RSE) takes coastal catches into account (Y/N)? NNNNNNN N N N NNNNNNNNNNNNNNNNN Y N N N NNNN N Does biological sampling of stock cover entire run? YYNYYYY Y Y Y NNYNNNNNNNNNNNNNY Y Y Y N YYYN Y Weight data - where collected? CUUCCCn/aC U C UUCUUUUUUUUUUUUUC CC/UCC/UUCn/an/aC Length/weight data - how Nt/R collected? SSSSSSS S S SNt/RNt/R SRRRRRRRRRRRRRSS/R/S Nt/S R/Nt R S Co n/a S Sea-age - how assessed? Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc O O O O Sc Sc Sc Sc Sc Sc Sc O Sc Sc River-age - how assessed? Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc Sc ScJ n/a Sc Sc Sex ratio - how assessed? I I I I I I/E E E E E I I I n/a n/a n/a n/a n/a n/a n/a n/a n/a E/I E/I E/I E/I E E E E n/a E/I E n/a I E Major changes in data collection procedures over time (Y/N)? NNNNNNN N N N NNNNNNNNNNNNNNNNN N N N N NNNN N

Other comments? sex in Change determination methods Changes in hatchery practices Sexvalidation from broodstock.of hatchery age in Change smolts. L/W data from experienced fishermen fisheryNo trap below fishNew ladder fishery coastal in Changes fishery coastal in Changes Changes in coastal fishery; farm escapees Changes in coastal fishery; farm escapees Changes in coastal fishery; acidification fishery coastal in Changes fishery coastal in Changes Reductions in coastal net fishery Sexratio data only in last 2 years Reductions in estuarine net fishery Reductions in estuarine net fishery also changes fishery Some 1991. in fishway with dam of Construction

Key:

Stock sampling procedure: CF = counting fence with adult trap; Sn = snorkel survey; IET = index estuary trap net; AT = adult trap; FC = Freshwater catch (nets & rods); RC = rod catch; NC = net catch; Co = counter. Weight data ‐ where collected: C = close to river mouth (recent entry); U = up river (possibly in freshwater for some time). Length weight data ‐ how collected: S = scientific sampling; R = rod fishery; Nt = net fishery; Co = counter. Sea age data: Sc = derived from scale reading; O = other (e.g. use of length/frequency data). River age data: Sc = derived from scale reading (adults); ScJ = derived from scales (juveniles); O = other (e.g. use of length/frequency data). Sex ratio: I = internal examination; E = external appearance; I/E = both internal & external methods used.

Sex ratios Table 3.3.1 indicates that, where available, sex ratios have been determined by both internal and external morphometric examination of returning fish. For some rivers, a combination of methods has been applied and, where possible, assessments based on external characteristics have been validated from broodstock (e.g. Penobscot). On some rivers, the methodology has also changed over time (e.g. Miramichi) or only been introduced recently (e.g. Lune). It is recognised that the more reliable data will be that obtained from fish dissection since this provides unequivocal confirmation of a fish’s sex. The allocation of fish to particular sexes by examination of external mor‐

ICES SGBICEPS REPORT 2010 | 39

phometric features will inevitably vary at different times of the season and among different observers. So, for some catchments sex determination has been restricted to more mature fish, presumably sampled later in the run. Such uncertainties clearly carry the risk of introducing some bias into any derived estimates of sex ratios. For a number of rivers, no effort has been made to assess the sex of returning fish, particularly where sampling of other biological parameters (e.g. length and weight) has been provided by anglers or other volunteers, and where are a large number of ‘samplers’ might be involved. This would clearly introduce substantial additional uncertainty.

Other data quality issues The original request for information indicated that data should only be included where this had been derived from sample sizes of at least 10 individual fish. In prac‐ tice, it is not clear that this has been applied in all cases. Provision of a “sample size” variable would allow a better appreciation of the likely error around the mean values presented for each of the variables in the spreadsheet. Some concerns were also ex‐ pressed about the use of annual means and that access to raw data might be of par‐ ticular added value in some instances. The Study Group previously recognised that requests for additional data sets, or for the raw data behind the annual mean values, should ideally be made in response to specific lines of enquiry and where worked examples/case studies suggested useful lines to pursue (ICES, 2009a).

Summary The foregoing comments highlight the extensive variability that exists in the ap‐ proaches used for assessing stocks and deriving time series of data on various bio‐ logical characteristics. The Study Group therefore recognised the need for cautious interpretation, particularly where comparisons were made between stocks. However, the Study Group noted that comparisons over time for a particular stock were likely to be more robust, provided data had been collected in a consistent manner over time.

3.4 Environmental data sets The Study Group reviewed the types of environmental information that could be employed to develop exploratory analyses at its first meeting (ICES, 2009a), with par‐ ticular emphasis on marine environmental data. The availability of the latest ICES Report on Ocean Climate (IROC) produced annually by the ICES Working Group on Oceanic Hydrography (WGOH; Holliday et al., 2009) was noted. The Study Group previously noted the current lack of a clear understanding of the distribution of salmon at sea and the multiple possible migratory destinations of in‐ dividuals within the same maturity grouping (e.g. southern European MSW fish mi‐ grating either to the Norwegian Sea or to Greenland). This remained a constraint, and it was recognised that clarifying specific environmental data requirements or making efforts to link these with changes in biological characteristics would need to be refined once clear hypotheses could be developed in relation to, for example, ob‐ served changes in specific stocks or stock complexes.

40 | ICES SGBICEPS REPORT 2010

4 Case Studies

The Study Group reviewed the latest information from the Baltic Salmon Working Group (WGBAST) and considered updated/new information from monitored stocks in the North Atlantic that have time series of data on biological characteristics of salmon or on changes in salmon abundance relative to environmental variables. These have loosely been grouped into a series of case studies and are described be‐ low. Hypotheses that were considered within these case studies include:  Marine survival is independent of the time of smolt emigration [Section 4.1];  Marine survival is independent of the temperature difference experienced by smolts when moving from fresh to salt water [Sections 4.1, 4.2];  Recruitment success (e.g. recruits (parr or smolt) produced per spawner) is independent of river conditions [Section 4.2];  Survival of Norwegian salmon in their second year at sea has not increased relative to their first year in the sea [Section 4.5];  The proportion of males among 2SW fish has not increased (indicating no evidence of later maturation of male fish) [Section 4.5];  In addition, case studies from the Baltic were reported by WGBAST which looked at hypotheses relating to the effects of competition, food availabil‐ ity, sea predation and smolt quality on at‐sea survival [Section 4.6].

4.1 Long-term changes in biological characteristics of smolts on the River Bush, N. Ireland and associations with environmental parameters The River Bush salmon project represents a long‐term study of the biological charac‐ teristics and stock recruitment dynamics of a typical salmon stock from UK( N. Ire‐ land). The project is based around annual whole river trapping census of returning adults and emigrating smolts at Bushmills in Co. Antrim. A long‐term data series, inclusive of a range of biological characteristics, measured on both juvenile and adult Atlantic salmon, is available for the stock.

Biological characteristics of R. Bush salmon over time The basic trends in biological characteristics from the River Bush datasets were exam‐ ined at the first meeting of the Study Group and are included in the earlier report. These data indicated that 2+ smolts were the most commonly encountered age class across the time series. The annual age structure has been quite variable and unstable temporally (Figure 4.1.1). The mean weights and lengths of Bush smolts across the entire time series have illustrated fair levels of variability with little evidence of a dis‐ crete directional temporal trend.

ICES SGBICEPS REPORT 2010 | 41

100%

80%

60% % 3+ % 2+ 40% % 1+ proportion

20%

0%

0 8 4 0 8 02 982 990 996 004 1978 198 1 1984 1986 198 1 1992 199 1 1998 200 20 2 2006 200 year

Figure 4.1.1. Age composition of smolts in the annual wild smolt runs on the River Bush, 1978– 2009.

Thermal mis-match and smolt survival Previous work on the timing of the River Bush smolt run has indicated that the mi‐ gration has shifted with increasing numbers of fish moving earlier in the year (O’Maoileidigh et al., 2003) with potential implications for subsequent survival driven by increased thermal discrepancy between river water and sea water. The biological characteristics of wild smolts from the River Bush were examined over a thirty year period (1978–2008). Changes in the timing of the smolt run were evident, with progressively earlier emigration periods detected across the time series (Figure 4.1.2). The rate of change has been around 3.6–4.8 days per 10 years, with the onset of the run (measured as the date of 25% total annual emigration) around 10–11 days earlier in recent years than at the start of the time series. The timing of smolt emigra‐ tion has been linked to ambient river temperature patterns. Distinct seasonal patterns have also been evident for biological characteristics of River Bush smolts with mean age and length decreasing throughout the emigration period.

42 | ICES SGBICEPS REPORT 2010

160

150

140

130

Ordinal Days 120

110

100

8 0 2 0 2 4 0 2 4 98 98 99 99 99 00 00 00 197 1 1 1984 1986 1988 1 1 1 1996 1998 2 2 2 2006 2008 Year

Figure 4.1.2. Smolt emigration times (ordinal date) on the River Bush between 1978–2008. The plots indicate the ordinal date by which standard percentages of the total annual smolt run had emigrated. Boxes indicate dates at which 25%, 50% and 75% of run had occurred; whiskers indi‐ cate dates at which 5% and 95% of run had occurred.

Marine survival patterns in 1SW salmon from the River Bush have been strongly in‐ fluenced by the run timing of the preceding smolt year, such that later emigrating cohorts demonstrated increased survival (Figure 4.1.3). A possible mechanism for this relationship may be the consequence of local climatic variation and potential thermal mismatch between freshwater and marine environments. The temperature contrast between fresh and salt water experienced by River Bush smolts is exacerbated by the absence of an estuary on the river. The River Bush discharges directly into the North Atlantic Ocean across a relatively steep and narrow stony beach, where the river mouth fans out to form a shallow braided channel. The physiological stresses associ‐ ated with osmotic regulation upon marine entry may be compounded by an in‐ creased requirement for thermal acclimation over a very limited transitional zone. Smolts are particularly vulnerable during the transition to salt water and the relative risk of predation has been demonstrated to increase during the acclimation of smolts to seawater (Handeland et al., 1996). These processes may have become more pro‐ nounced in recent years on the River Bush due to an observed increasing thermal disparity between the two discrete aquatic environments (Figure 4.1.4) (Kennedy & Crozier, 2010).

ICES SGBICEPS REPORT 2010 | 43

40

35

30 (% )

25

20 Survival

15

Marine 10

5

0

y y y y y pr pr a a a a A A - - -M -M 23-Apr 25 27 29-Apr 01 03 05-M 07-Ma 09-M Date of 25% smolt emigration

Figure 4.1.3. Relationship between the date of the onset of the smolt run (denoted by the data on which 25% of smolt emigration had occurred) and subsequent marine survival of 1SW River Bush salmon.

12 River Temp 11.5 SST C

o 11

10.5

10

9.5

9 Mean Daily Temperature Temperature Daily Mean

8.5

8

8 87 90 93 96 99 02 97 984 9 9 9 9 9 0 1 1981 1 1 1 1 1 1 2 Year

Figure 4.1.4. Mean daily water temperature (oC) for River Bush (Bushmills) and the sea surface (Malin Head) recorded on the date of 25% total Bush smolt emigration.

44 | ICES SGBICEPS REPORT 2010

4.2 Update on biological characteristics of salmon from the River Dee, North Wales and other monitored rivers in UK (England & Wales) and associa- tions with environmental parameters It was noted at the first meeting of the Study Group (ICES, 2009a) that there are eight monitored rivers (Dee (Wales), Test, Itchen, Frome, Tamar, Fowey, Lune and Kent) in UK (England & Wales) which report annual Returning Stock Estimates (RSEs) for salmon. Four of these rivers (Dee [Wales], Tyne, Tamar and Lune) are also identified as monitored ‘index’ rivers and are distinguished by having established or develop‐ ing (partial) adult trapping programmes to collect biological information (e.g. on age, size, sex, etc.) (Davidson, 2008). These rivers are spread around UK (England and Wales) and are broadly representative of the main salmon and sea trout producing regions. The group includes some of the leading ‘chalk’ rivers of southern England (e.g. Test, Itchen, Frome) as well as the more usual ‘rain‐fed’ rivers of the west and north. While the monitoring programmes for individual rivers may differ in their de‐ tail, they share a number of common objectives in terms of the information collected and its application to management (Davidson, 2008). Information on changes in biological characteristics of salmon from the River Dee (UK, England & Wales) was reported at the first meeting of the Study Group (ICES, 2009a). The following provides some new analyses in respect of the River Dee and also an initial examination of spring river and coastal temperatures to assess whether there is any evidence of substantive changes over time.

Long-term changes in spring temperatures – any evidence for a thermal ‘mis-match’ between rivers and coastal areas? Further to the investigations on the River Bush (Section 4.1), where a river/sea tem‐ perature mis‐match at the time of smolt migration has been linked to marine survival, initial investigations were completed for a number of sites in UK (England & Wales) to see if similar patterns might be evident. Time‐series of river and sea surface tem‐ perature data were collated/derived for four rivers spread from south to northwest around the coast ‐ namely: Hampshire Avon; Tamar; Dee and Lune ‐ to explore spa‐ tial and temporal patterns in these data. In all cases, river temperatures were collected at main stem sites at or close to the head‐of‐tide. Sea surface temperature data were extracted from records collated and published by Cefas (http://www.cefas.co.uk/data.aspx) and obtained at various coastal stations. The proximity of these stations to the selected rivers was variable and usually amounted to a distance of several km. There were a number of gaps in the data sets, particularly for the river temperature time‐series, and to fill these monthly mean temperatures were predicted from simple linear regression relationships with parallel temperature series. In the case of sea temperatures, these parallel data series were from the nearest adjacent sampling sta‐ tion to the required data sets. In the case of river temperatures, local air temperature time‐series (more widely available than river temperature time‐series) were used as predictors of river temperature. For example, Figure 4.2.1 shows the regression rela‐ tionship between monthly mean observed air temperatures at Slapton Ley (extracted from the UK Meteorological Office MIDAS Land Surface Station Data set http://badc.nerc.ac.uk/home/) and water temperatures on the River Tamar.

ICES SGBICEPS REPORT 2010 | 45

25.00

20.00

15.00

10.00

5.00 Tamar mean monthly water temp C

0.00 0.00 5.00 10.00 15.00 20.00 Slapton mean monthly air temp C

Figure 4.2.1. Regression relationship between the monthly mean water temperature for the River Tamar and the air temperature recorded at Slapton (in the same region).

Figure 4.2.2 shows mean April temperatures (usually calculated from daily records) for the four rivers along with mean April sea surface temperatures for adjacent coastal areas. The difference between the two data sets is presented in Figure 4.2.3, together with a plot of the 3‐year mean difference. The value of collating and analysing river and sea surface temperature data in this way is that it provides some insight into spatial and temporal variations in these fac‐ tors around England and Wales and their possible role in influencing changes in salmon biology and abundance. However, the data have a number of shortcomings which need to be considered; for example:  Predicting missing sea and river temperature data points from parallel temperature series is less than ideal. While regression relationships be‐ tween, for example, air and water temperature time‐series are usually highly significant, predictions made well beyond the period of tempera‐ ture observations are likely to be particularly uncertain. That said, this type of approach is often necessary if we want to examine past changes ‐ par‐ ticularly so in the case of river temperatures where long time‐series of data are scarce.  Stations where sea surface temperature data have been collected are often remote from the rivers of interest and may not reflect temperature condi‐ tions in the immediate vicinity of those rivers, or indeed the estuaries themselves ‐ where the effects of mixing on temperatures experienced by smolts or adult salmon may be significant, particularly in physically com‐ plex estuaries.  Use of monthly mean temperature data to explore relationships with fish survival or abundance is a relatively crude approach when periods of key interest (e.g. the main smolt migration period) may be confined to a few days. Therefore exploration of temperature data at a finer resolution is likely to be preferable in most circumstances (although data sets may not be readily available in this format).

46 | ICES SGBICEPS REPORT 2010

Hants Avon: April 14 12 10 8 6 4 Mean temp C Mean temp River temp 2 Sea temp 0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Year

Tamar: April 14 12 10 8 6 4 Mean temp C Mean temp River temp 2 Sea temp 0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Year

Dee: April 14 12 10 8 6 4 Mean C temp River temp 2 Sea temp 0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Year

Lune: April 14.0 12.0 10.0 8.0 6.0 4.0 Mean temp C temp Mean River temp 2.0 Sea temp 0.0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Year

Figure 4.2.2. April mean water temperatures for four rivers in UK (England and Wales) compared with sea surface temperatures in adjacent coastal areas. [Predicted rather than observed monthly data points are indicated by partially filled symbols.]

ICES SGBICEPS REPORT 2010 | 47

Hants Avon: April 5.0 Temp difference 3.0 3-year mean difference 1.0

-1.0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

-3.0

-5.0 Hants Avon river temp C temp Avon river Hants

Diff Bournemouth sea minus minus sea Bournemouth Diff -7.0

Tamar: April 5.0 4.0 Temp difference 3.0 3-year mean difference 2.0 1.0 0.0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 -1.0 -2.0 -3.0 Tamarriver temp C -4.0 Diff Plymouth sea minus Plymouth Diff -5.0

Dee: April 5.0 4.0 Temp difference 3.0 3-year mean difference 2.0 1.0 0.0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 -1.0 -2.0

Dee river temp C temp Dee river -3.0 -4.0

Diff Wylfa/Moelfre sea minus sea minus Wylfa/Moelfre Diff -5.0

Lune: April 5.0 4.0 3.0 2.0 1.0 0.0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 -1.0 -2.0

Lune river temp C temp river Lune -3.0 Temp difference -4.0 Diff Heysham sea minus Diff 3-year mean difference -5.0

Figure 4.2.3. The difference in mean April water temperatures for four rivers in UK (England and Wales) with that of adjacent coastal areas (sea surface temperatures) and the 3‐year mean differ‐ ence (solid line).

The mean monthly April temperatures (Figure 4.2.2) suggest modest warming in riv‐ ers and coastal waters in the the last 10–20 years ‐ particularly evident for the two most southerly systems: Hampshire Avon and Tamar. However, there is little evi‐ dence on any of these rivers over the last 40+ years of any strong pattern in terms of thermal mis‐match (i.e. the difference between river and sea temperatures). This is also the case for May (data not shown). The Lune ‐ the most northerly of the four riv‐ ers ‐ shows the greatest inter‐annual variability in thermal mis‐match. Thus, while spring river temperatures on the River Bush (Section 4.1) have been increasing faster than sea temperatures over recent years, based on this preliminary investigation, there appears to be no similar trend in rivers around England and Wales.

River Dee – recruitment and environmental factors Studies on the River Test reported at the first meeting (ICES, 2009a) had indicated a strong negative correlation between recruitment success – expressed as the numbers of smolts produced per spawner – and mean flow during the egg incubation period

48 | ICES SGBICEPS REPORT 2010

(Figure 4.2.4). More specifically, it was suggested that during high flow periods sediment mobilisation increases (Figure 4.2.5) with adverse consequences for devel‐ oping eggs and pre‐emergent fry as spawning gravels become clogged with sediment and intra‐gravel flow and oxygen delivery are reduced.

Figure 4.2.4. Relationship between river discharge during egg incubation and spawning success (smolts per spawner) on the River Test, southern England.

Figure 4.2.5. Relationship between suspended solids load and spawning success (smolts per spawner) on the River Test, southern England.

Prompted by these findings, a similar analysis was carried out on salmon stock and recruitment data from the River Dee, N. Wales to investigate whether any relation‐ ship between recruitment success and flow was evident. In the case of the Dee, no census data for smolt output were available, but indices of 0+ juvenile recruitment, derived from the results of extensive autumn electrofishing surveys, were obtained. Measures of recruitment success (‘recuits per spawner’ or R/S) generated from these data sets were plotted against spawner numbers and a linear regression fitted to the data points (Figure 4.2.6). Residuals from the regression line were, in turn, correlated with mean flows calculated for various periods in the life‐cycle (‘egg incubation’, ‘emergence’ and ‘first summer’). These flow variables are summarised in Table 4.2.1.

ICES SGBICEPS REPORT 2010 | 49

River Dee: Recruits as autumn (0+) fry 0.000 0 2000 4000 6000 8000 10000 -0.200 R2 = 0.26 -0.400 P = 0.101

-0.600

-0.800

-1.000

-1.200

-1.400

Loge (Recruits per spawner) per Loge (Recruits -1.600

-1.800 Stock as Spawners

Figure 4.2.6. Relationship between recruitment success (‘recuits per spawner’) and spawner num‐ bers for the River Dee.

Correlation coefficients and significance levels were examined for all combinations of R/S residuals and flow variables. Scatter plots for these comparisons are shown in Figure 4.2.7 and significant correlations (P<0.05) indicated with a # symbol.

Table 4.2.1. Summary of flow variables for various periods in the life‐cycle (‘egg incubation’, ‘emergence’ and ‘first summer’) for the River Dee, as used in exploratory analysis of recruitment success.

Life stage Variable as monthly (geometric) mean flow (cumecs) for period indicated: Incubation Dec Jan Feb Mar Apr Mean (Dec‐ Mean (Mar‐ (‘IFlow’) Mar) Apr) Emergence Apr May Jun Mean (Apr‐ Mean (‘EFlow) May) (Apr‐Jun) First summer Jul Aug Sep Mean (Jul‐ Mean (‘0+_SFlow’) Aug) (Jul‐Sep)

Like the Test example, this analysis suggests that high flows during the winter pe‐ riod, principally December to January, may have an adverse effect on recruitment success of Dee salmon at the egg incubation phase. However the possible mecha‐ nisms at work (e.g. siltation and/or scour of spawning gravels) are not clear at this stage and further work is needed. There is little evidence that flows examined during the other phases of the life‐cycle are similarly strongly associated with recruitment success. Unlike the Test example, however, the scatterplots exploring the influence of ‘incuba‐ tion’ flow do not suggest that a flow ‘threshold’ is operating above which recruitment success falls off sharply. This may relate to the different nature of the Test and Dee catchments. While both studies refer to juvenile recruitment success at similar stages (the Dee from spawner to autumn 0+ the Test from spawner to smolt – where the great majority of smolts emigrate as S1s), and are both whole catchment studies, the Dee is a larger and more complex catchment and a greater proportion of juvenile production is likely to come from sub‐catchments – i.e. away from the main river where the flow measurements used in these analyses have been taken.

50 | ICES SGBICEPS REPORT 2010

Scatterplot of R/S residual vs IFlow_Dec, IFlow_Jan, IFlow_Feb, ...

IFlow_Dec IFlow_Jan IFlow_Feb IFlow_Mar IFlow_Apr 0.5

0.0

-0.5 30 60 900 40 80 0 50 100 20 40 60 15 30 45 IFlow_M_Dec_Mar IFlow_M_Mar_Apr EFlow_Apr EFlow_May EFlow_Jun 0.5

0.0

-0.5 20 40 60 10 30 50 15 30 4510 20 30 10 15 20 Eflow_M_Apr_May Eflow_M_Apr_Jun 0+_SFlow_Jul 0+_SFlow_Aug 0+_SFlow_Sep 0.5 R/S residuals 0.0

-0.5 10 20 3010 15 20 20 40 60 10 20 30 10 20 30 0+_SFlow_M_Jul_Aug 0+_SFlow_M_Jul_Sep 0.5

0.0

-0.5 10 20 30 8 16 24

Figure 4.2.7. Scatter plots of recruitment success (R/S) residuals and flow variables at various stages of the life‐cycle; significant correlations (P<0.05) are indicated with a # symbol.

Accordingly, data from the River Dee should be explored at a sub‐catchment level to examine relationships between recruitment and flow at a finer resolution and investi‐ gate spatial differences in the flow response. [Note: In a related analysis of River Dee data to look at recruitment success from adult spawner to adult spawner, flows during the egg incubation period did not ap‐ pear influential, i.e. any effect was probably lost among a host of other factors acting over this longer period of the life‐cycle.]

4.3 Biological characteristics of salmon from the River Frome – UK (England & Wales) Information on changes in biological characteristics of salmon from the River Frome (UK, England & Wales) was made available at the first meeting of the Study Group (ICES, 2009a). There were no new data or developments to report at the current meeting.

4.4 Biological characteristics of salmon from the River Test – UK (England & Wales) Information on changes in biological characteristics of salmon from the River Test (UK, England & Wales) was made available at the first meeting of the Study Group (ICES, 2009a). There were no new data or developments to report at the current meeting.

ICES SGBICEPS REPORT 2010 | 51

4.5 Evidence for later age at maturity in Norwegian salmon stocks in recent years The Study Group previously examined information from Norway indicating recent changes in the age at maturity of returning salmon. This analysis was updated at the latest meeting to incorporate the most recently available data. Analysis of the pre‐fishery‐abundance (PFA) of Norwegian salmon stocks indicates a significant positive relationship between the number of 1SW salmon in one year and the number of 2SW salmon in the following year (Hansen et al., 2007; Hansen et al., unpublished). The regression between PFA for 1SW salmon in year n and PFA of 2SW salmon in year n+1 shows positive residuals in the more recent years in three regions in Norway (Figure 4.5.1), suggesting that more salmon return as 2SW fish than expected from the number of 1SW fish the previous year. The most recent data remain consistent with this observation. The 1SW component is typically male‐dominated, and if there is a tendency for later maturity one would expect an increase in the male proportion among 2SW salmon. The apparent later age at maturity might be explained both by more salmon delaying age at maturity, or by survival in the second year at sea having increased relative to survival in the first year at sea. If the first hypothesis is true one might expect an in‐ crease in the proportion of male salmon among the returning 2SW fish, whereas the sex ratio among 2SW fish is more likely to be unchanged if the second hypothesis is more valid.

Southern Norway Central Norway

Northern Norway Unstandardized resituals

Year

Figure 4.5.1. Unstandardised residuals from the regressions of PFA of 1SW salmon in year n and 2SW salmon in year n+1 in three Norwegian regions.

52 | ICES SGBICEPS REPORT 2010

4.6 Baltic Sea – changes in post-smolt survival and the factors affecting it The Study Group once again reviewed information on post‐smolt survival arising from the work of the Baltic Salmon and Trout Assessment Working Group (WGBAST). WGBAST initiated preliminary analyses during their 2008 meeting to evaluate the possible reasons for the low at‐sea survival of salmon stocks in the Baltic Sea; investigations continued in 2009. Further details are available in the WGBAST reports (ICES, 2008b; ICES, 2009b), and work is expected to continue during the WGBAST meeting in 2010.

Background The post‐smolt survival of salmon in the Baltic Sea has decreased in recent years, both for wild and hatchery‐reared smolts (ICES, 2008b). According to estimates generated from the assessment model, this decline started in the mid 1990s; the reasons behind the decline are unclear. It is possible that changes in the Baltic Sea ecosystem have negatively affected salmon post‐smolt survival rates by, for example, changing the abundance of prey species or though increased competition or predation from other species. The Baltic Sea has undergone pronounced changes in the last two decades, characterised by several regime shifts in species composition (ICES, 2007b). The main drivers of this process are believed to have been climate changes affecting salinity, oxygen levels and temperature, but fisheries and eutrophication are also believed to have had an impact. It has also been hypothesized that increased mortality among hatchery‐reared post‐smolts may be linked to changed hatchery practices. These have been subject to continual improvement – for example, the use of feeds with higher fat and energy content. Favourable river temperatures, especially in autumn, have also resulted in improved growth rates and increasingly large smolts. There are concerns that the large size of reared smolts may have negative fitness consequences in the wild.

Preliminary tests of hypotheses using multivariate analyses WGBAST initially agreed on the following specific hypotheses to test with multivari‐ ate analyses (ICES, 2008b):  Competition hypothesis: the aim was to test if intra‐specific competition for food is important in explaining variation in post‐smolt survival. In this case, the survival of smolts would decrease as the smolt abundance in‐ creases.  Food availability hypothesis: Young herring are considered important prey for young salmon in the Baltic Sea. Therefore, increased recruitment of 0+ herring in the smolt year should lead to higher survival if herring abun‐ dance is a limiting factor. This hypothesis is directly linked to the competi‐ tion hypothesis above. If herring recruitment directly affects post‐smolt survival then, by definition, at‐sea survival is density‐dependent and should be affected by smolt production in rivers and hatcheries.  Seal predation hypothesis: If salmon smolts are subject to substantial seal predation, then an increase in the number of seals along the migration path of post‐smolts should coincide with lower survival. If such a pattern is re‐ vealed, the predation hypothesis remains possible, but is not necessarily more likely than other hypotheses that might lead to similar predictions.  Smolt quality hypothesis: This relates only to hatchery‐reared smolts and hypothesises that increased mortality is due to changed hatchery practices.

ICES SGBICEPS REPORT 2010 | 53

Pre‐release factors suggested to be important explanatory variables include size, condition factor and level of fin damage. In preliminary analyses, the model which included only the seal counts was assigned the highest probability (>0.4) as a predictor. Models including estimated smolt abun‐ dance were assigned low probability compared to other combinations.

Ongoing analyses during 2009 During the WGBAST meeting in 2009 (ICES, 2009b), three data sets describing salmon at‐sea survival were evaluated: 1) post‐smolt survival estimates (Mps) for wild salmon from the Baltic assessment model; 2) return rates to the river of reared salmon from the River Umeälven (Ume), and 3) tag‐recapture rates of salmon reared in the River Dalälven. The data for potential explanatory variables characterising the Baltic Sea ecosystem and the smolt release hatcheries were updated and some more de‐ tailed time series were added.

Statistical procedures From initial analyses of individual predictor variables using a log‐linear regression model (ICES, 2008b), a number of variables were selected for further statistical evaluation. Predictor variables that were considered important included seal abun‐ dance, herring recruitment and abundance, sprat abundance, smolt production esti‐ mates (both wild and hatchery‐reared smolt production) and trawl effort. Temperature was not included, as previous studies have shown that sea surface tem‐ perature may only have an indirect effect on salmon survival through its positive ef‐ fect on herring recruitment (e.g. Kallio‐Nyberg et al., 2006). The selected predictor variables were used to test the hypotheses formulated during the previous WG meet‐ ing (see above). The Mps and Ume return rate survival indices were used to test the seal predation and food abundance hypotheses using two slightly different Bayesian models, whereas tag‐recapture data from the River Dalälven were used to test the smolt quality hypothesis. Post‐smolt survival estimates ‐ Mps estimates were extracted from the assessment model by calculating the median of the posterior distribution of the instantaneous mortality rate, and by calculating the correlation matrix of the joint posterior. Then the median values were used as observations and the correlation matrix was used as a model for correlated measurements. The mortality was assumed to vary randomly between years with an annual expected value depending on explanatory variables according to following equation:

E(My) = qSEALy + exp(α – β1HSD30y+1 – β2HSD31y+1 – γ2HCBy – γ2SCBy) Where q is the catchability coefficient of seals and α, β1, β2, γ1 and γ2 are regression coefficients. All these parameters are assumed to be positive, which means that the increase in seal abundance is assumed to increase the mortality, while the variables reflecting availability of food are assumed to decrease the mortality. The explanatory variables are described in the following table. Note that the relative recruitment of 0+ herring was approximated in the model by shifting the estimates of 1+ recruitment backwards by one year. SD refers to ICES subdivision.

54 | ICES SGBICEPS REPORT 2010

VARIABLE DESCRIPTION SEAL Total Baltic seal counts in Swedish surveys HSD30 SD30 1+ herring recruitment / (SD30+SD31 smolts) HSD31 SD31 1+ herring recruitment / (SD31 smolts) HCB Herring stock biomass in central Baltic SCB Sprat stock biomass in central Baltic

The five explanatory variables can be combined in 31 different ways. The relative credibility of these alternative models was estimated by calculating posterior prob‐ ability for each of the models. Such a Bayesian Model Averaging approach weights the models based on the goodness of fit and typically induces a penalty for complex‐ ity. For models that included the number of seals as a predictor, it was possible to calcu‐ late an estimate of the number of smolts eaten by an average seal. These values could then be used to assess whether the results of the model were plausible compared to independent knowledge about the diet of seals. Return rate to River Umeälven ‐ The assumed model structure for the return rate analysis was very similar to the model used for the estimated post smolt mortality rates. The main difference was that offshore and coastal fishing effort were included as predictors in each combination of explanatory variables. Furthermore, instead of using herring data from SD(ICES subdivision)31, both 0+ and 1+ recruitment of SD30 were included as explanatory variables since it has been suggested that larger reared post‐smolts may feed on both one year old and young of the year herring (Salminen et al., 2001). The mortality rates were assumed to vary randomly around the annual mean which depends on the explanatory variables:

Zy = Fy + My

Fy = q(pCOASTy + (1 ‐ p)OFFSHy)

My = dSEALy + exp(α – β1H1y – β2H0y – γ1CBHy – γ2CBSy) The mortality rate was then converted to expected return rate as:

Ry = exp(‐My). In order to compare the results of the modelling of the return rate to the estimated post‐smolt survival, the expected natural marine survival of released River Ume smolts was calculated as:

Ry = exp(‐My).

VARIABLE DESCRIPTION SEAL Total Baltic seal counts in Swedish surveys H0y = H1y+1 SD30 0+ herring recruitment / (SD30+SD31 smolts) H1 SD30 1+ herring recruitment / (SD31+SD31 smolts) HBC Herring stock biomass in central Baltic CBS Sprat stock biomass in central Baltic COAST Mean coastal effort during the next two years after the release year, all gear combined OFFSH Mean offshore effort during the next two years after the release year, all gear combined

ICES SGBICEPS REPORT 2010 | 55

Tag‐recapture rates from River Dalälven ‐ Tag‐recapture data is dependent on fish‐ ing effort and the willingness of fishermen to report tags, and is therefore more prob‐ lematic to use for analyses of trends in survival over time if these problems are not dealt with carefully. This dataset was therefore used only to evaluate the smolt qual‐ ity hypothesis described above, more specifically the effects of smolt length and hatchery injuries on subsequent survival following release. The fish were divided into four categories: healthy (no injuries), dorsal fin injured, other fins injured, and dorsal fin and other fins injured. Release year, release date (Julian date) and length of fish at tagging were included as continuous variables. The dataset was first analysed with PROC LOGISTIC (SAS statistical software) in order to investigate the overall effect of the independent variables. Thereafter, the data was analysed with PROC GENMOD (SAS statistical software), assuming binary response of recapture, in order to calculate Least‐square means for the different categories.

2009 results The work on salmon post‐smolt mortality in the Baltic Sea is still at an initial stage, and these results should be viewed as preliminary. It is important to highlight the fact that a correlative approach has been applied, which means that observed associations between survival estimates and biological predictors are not necessarily due to causal relationships. The independent dataset of return rates of reared salmon to the river mouth of the River Umeälven was in close agreement with the Mps model estimates (Figure 4.6.1.a). The decrease in salmon fishing effort was not taken into account in examin‐ ing the Ume dataset. Despite that, return rate seems to have decreased over the last 8 to 10 years, which supports the model estimates indicating a negative trend in post‐ smolt survival over the last decade. The natural marine survival (Sy) estimated from the River Umeälven return rates are lower than the post‐smolt survival estimated using the stock assessment model (Figure 4.6.1.b). The independent survival esti‐ mates show similar annual variation and the decreasing trend, which gives strong support for the conclusion that post‐smolt survival has been decreasing.

56 | ICES SGBICEPS REPORT 2010

0.4 0. 08 a) 0. 07

0.3 0. 06

0. 05

0.2 0. 04

0. 03 Return rate

Survival rate (Mps) 0.1 0. 02

0. 01 Mps wild Mps reared 0 0. 00 Ume return rate

8 1 5 8 1 4 8 9 94 9 97 9 00 0 0 9 9 9 9 0 0 0 1987 1 1989 1990 1 1992 1993 1 19 1996 1 19 1999 2 2 2002 2003 2 2005 2006 Smolting year

0.2 0.3

0.18 b) 0.25 0.16

0.14 0.2 0.12 River Ume model 0.1 0.15 Ass ess ment model 0.08 0.1 0.06 Post smolt survival Natural marine survival 0.04 0.05 0.02

0 0

7 1 5 9 3 4 8 9 92 9 96 9 00 0 0 9 9 9 9 9 9 0 0 0 1 1988198919901 1 199319941 1 199719981 2 200120022 2 2005

Figure 4.6.1. Changes in: a) post‐smolt survival of wild and hatchery‐reared salmon estimated by the assessment model and the return rate of reared salmon to the River Umeälven, and b) post‐ smolt survival of hatchery‐reared salmon estimated by the assessment model and estimated natu‐ ral marine survival of hatchery‐reared salmon from the River Umeälven, for 1987–2007 smolt year classes.

The predation and food availability hypotheses Initial analyses of predictor variables indicated that survival of wild post‐smolts was negatively correlated with seal abundance (Figure 4.6.2) and total smolt production in the Gulf of Bothnia (Figure 4.6.3), and positively correlated with herring recruitment in both the Bothnian Bay (Figure 4.6.4) and Bothnian Sea. No relationships were ob‐ served between survival and total trawl effort in the Bothnian Bay or pelagic trawl effort in the Bothnian Sea.

ICES SGBICEPS REPORT 2010 | 57

Figure 4.6.2. a) Changes in survival of wild post‐smolts and seal abundance along the Swedish coast, and, b) indications of a negative correlation between these variables.

58 | ICES SGBICEPS REPORT 2010

0.4 6000 a) 0.35 5500

0.3 5000

0.25 4500

Mps wild 0.2 4000 Smolt production BoS Mps

0.15 3500

0.1 3000

0.05 2500 Total smolt smolt Total production BoB &

0 2000

7 8 0 1 2 4 5 6 7 9 0 1 3 4 5 7 9 0 0 98 99 99 00 1 198 1989 19 9 1 1 1993 199 199 199 19 19 98 199 2 20 2002 200 20 0 20 20 06 200 Smolting year

0.4 b)

0. 35 y = -6E-05x + 0.4844 R2 = 0.4492

0.3

0. 25 Mps 0.2

0. 15

0.1

0. 05 3000 3500 4000 4500 5000 5500 6000 Total smolt production BoB & BoS

Figure 4.6.3. a) Changes in survival of wild post‐smolts and the production of wild and hatchery‐ reared smolts in the Bothnian Sea and Bothnian Bay, and, b) indications of a negative correlation between these variables.

ICES SGBICEPS REPORT 2010 | 59

0.4 1000000 a) 900000 0. 35 800000 )

0.3 3 700000 0. 25 600000

0.2 500000 Mps 400000 0. 15 300000 0.1 No 0+ xherring (10 200000 0. 05 100000 Mps wild 0 0 Herring recr BoB 6 7 8 93 94 95 987 989 9 9 9 99 99 001 002 003 1 1988 1 1990 1991 1992 1 1 1 1 199 1 1999 2000 2 2 2 2004 2005 Smolting year

0.4 b)

0.35

0.3

0.25

Survival wild Survival 0.2 y = 1E-07x + 0.199 R2 = 0.2344 0.15

0.1 0 100000 200000 300000 400000 500000 600000 700000 800000 900000 1000000 Herring recruitment BoB

Figure 4.6.4. a) Changes in survival of wild post‐smolts and the recruitment of herring in Both‐ nian Bay, and, b) indications of a positive correlation between these variables.

From the 31 models fitted to post‐smolt survival estimates, three models stood out clearly in terms of the posterior probability (Figure 4.6.5). Each of these include seal abundance and a combination of the SD31 and SD30 0+ herring available per smolt. The model with seal abundance and both SD31 and SD30 as predictors had the high‐ est degree of determination. Fitting models with only seal abundance and only her‐ ring availability indicates that the overall trend in post‐smolt survival can be explained by the increased number of seals, and the annual variation in survival co‐ incides with the variation in availability of 0+ herring per smolt.

60 | ICES SGBICEPS REPORT 2010

Probability 0 0.05 0.1 0.15 0.2 0.25 0.3

Seal+HSD30

Seal+HSD31+HSD30

Seal+HSD31

Seal+HSD30+HCB

HSD30+HCB

Model HSD30

Seal+HSD31+HCB

Seal+HSD31+HSD32+HCB

Seal

HCB

Figure 4.6.5. Posterior probabilities of the 10 most likely predictor variable combinations to ex‐ plain the variation in the survival of wild post‐smolts.

Results from the analysis of the River Umeälven return rate produced a similar pic‐ ture (Figure 4.6.6), although the importance of seal abundance as a predictor was less pronounced. In general, seal numbers seem to explain the trend in mortality for a longer time span, while herring abundance in SD 30 is an important predictor for an‐ nual variations. According to modelling of post‐smolt survival with the assumption of seal predation, an average seal would have to consume approximately 50–100 smolts per year in or‐ der to generate the estimated increase in mortality. This uncertainty arises mainly from the uncertainty about the true size of the seal population.

Probability 0 0.05 0.1 0.15 0.2 0.25

H1+H0 H0 Seal+H1+H0

H1 Seal+H0

Model Seal+H1 H0+H1+CBH

Seal Seal+H1+H0+CBH Seal+H0+CBH

Figure 4.6.6. Posterior probabilities of the 10 most likely predictor variable combinations to ex‐ plain the variation in the return rate of reared salmon from River Umeälven.

ICES SGBICEPS REPORT 2010 | 61

The smolt quality hypothesis There was an overall difference in recapture rate between the four categories of inju‐ ries recorded for the hatchery‐reared smolts. Length at release, and day and year of release also influenced recapture rate, while smolt length also had a positive effect on tag recapture rate. In respect of fin injuries, fish with only dorsal fin injuries had the same recapture rate as healthy fish. However, fish classified as having ‘other’ fins injured had a recapture rate that was only 60% of that in healthy individuals indicat‐ ing that injuries on fins than the dorsal fin have much more severe effects on survival following release to the wild. A more detailed description of these analyses was in‐ cluded in the first SGBICEPS report (ICES, 2009a).

Conclusions and future studies The results suggest that herring recruitment in the Gulf of Bothnia is able to explain a large proportion of the annual variation in post‐smolt survival, but is not able to ex‐ plain the negative trend in survival over the last decade. The addition of seal abun‐ dance to the models captured this decline. However, it is important to note that because a correlative approach was used, we cannot conclude that the observed pat‐ terns are due to causal relationships. The estimated number of salmon that an average seal has to consume per year in or‐ der to explain the estimated reduction in survival is within realistic levels. However, the available information on seal diet composition is limited, and future studies should focus on collecting such data in order to evaluate if the present seal popula‐ tion is actually able to have a negative effect on salmon survival, indicating a possible causal relationship. Information about seal diet would be particularly important for the summer when migrating smolts are passing the densest seal populations. As re‐ gards herring, previous studies have indicated that the recruitment of this prey spe‐ cies may be one of the most important factors regulating salmon abundance in the Baltic Sea, and our results support these earlier findings. The next step will be to con‐ sider a model which takes into account detailed information on salmon size, prey size distribution and predation pressure from seals during the post‐smolt migration to the southern Main Basin. Even though it is not clear whether the estimated effects of seal abundance and her‐ ring recruitment represent a true causality, the strong co‐variation can potentially be utilised when predicting the post‐smolt survival. Smolt production and seal abun‐ dance are both likely to change slowly, whereas the recruitment of herring is proba‐ bly less predictable. However, herring recruitment has been found to be correlated with the water temperature, which might be predictable to some extent. Since the number of seals and smolts are not expected to drop in the near future, the observed co‐variation suggests that post‐smolt survival is expected to remain low for the next couple of years, at least. Annual variation is more difficult to predict. It is also evident from this study that fin damage in hatchery‐reared smolts can affect survival following release to the wild, and while fin damage is certainly not the only important factor (as noted above) it may, in part at least, explain the comparably low survival of many hatchery‐reared stocks. Clearly, the way hatcheries are managed will have an impact on how reared salmon are able to cope with the wild environ‐ ment.

62 | ICES SGBICEPS REPORT 2010

4.7 Baltic Sea – review of Swedish tagging experiments and implications for estimating post-smolt survival Analyses of Swedish tagging data were reviewed at the first meeting of the Study Group (ICES, 2009a). There were no new data or developments to report at the cur‐ rent meeting, other than those included in Section 4.6.

4.8 Fecundity of Penobscot River broodstock Information on long‐term changes in broodstock fecundity for the Penobscot River (USA) was made available at the first meeting of the Study Group (ICES, 2009a). There were no new data or developments to report at the latest meeting.

4.9 Smolt summary data for two monitored river sites in UK(Scotland) The Study Group reviewed new information from two monitored river programmes in UK(Scotland). Data on the biological characteristics of emigrating smolts are avail‐ able for two monitored sites in Scotland: the North Esk and the Girnock Burn.

River North Esk The River North Esk drains into the North Sea on the east coast of Scotland. A sample of the total smolt run from the river is sampled on an annual basis by use of a trap that is situated on a mill lade that runs off the river some 7km from the river mouth and rejoins the river at the head of tide. As flow in the lade is maintained at a con‐ stant level, but river flow fluctuates naturally, the proportion of the smolt run sam‐ pled will vary both, throughout the annual smolt run, and among years. Thus, the sample from which we have derived summary biological characteristic parameters is biased. Bearing this caveat in mind, annual summary statistics have been provided, for inclusion in the SGBICEPS master file, for annual smolt age composition and mean fork lengths and weights by smolt age group. The annual smolt age composition for the River North Esk for the period 1964 to 2008 is shown in Figure 4.9.1.

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008

1 2 3 4

Figure 4.9.1. Annual smolt age composition in the River North Esk, Scotland, 1964–2008.

ICES SGBICEPS REPORT 2010 | 63

Smolts emigrate from the North Esk after spending between one and four years in freshwater. Two year old smolts (S2) comprise the majority of the run throughout the series, followed by S3, then S1, with S4 comprising a very small component. Throughout the time period there has been a tendency for an increase in the propor‐ tion of younger smolts (S1 and S2) and a decrease in the proportion of older smolts (S3 and S4). Figure 4.9.2 shows the mean fork lengths (cm), by smolt age, of emigrating smolts over the period 1964 to 2008. Generally, in any given year, older smolts have a greater mean fork length. For all four age groups there is a pattern of decreasing fork length until the early 1980s with lengths increasing thereafter.

14.5

14.0

13.5

13.0

12.5

12.0

11.5

11.0

10.5

10.0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

1 2 3 4

Figure 4.9.2. Mean fork length (cm), by smolt age, in the River North Esk, 1964–2008.

Figure 4.9.3 shows the mean weight (g), by smolt age, of emigrating smolts over the period 1975 to 2008. The weight data provide a similar picture to that observed for the length data. Generally, in any given year, older smolts have a greater mean weight. For all four age groups there is a pattern of decreasing mean weight until the early 1980s with weights increasing thereafter.

64 | ICES SGBICEPS REPORT 2010

32.0

28.0

24.0

20.0

16.0

12.0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

1 2 3 4

Figure 4.9.3. Mean weight (g), by smolt age, in the River North Esk, 1975–2008.

In general, there is a close correspondence between the annual directional changes in mean fork lengths and mean weights. The exception is in the earlier period (1975– 1980). Having considered the field conditions under which both size parameters were collected, it is likely that the fork length parameter is the most accurate indicator of size.

Girnock Burn The Girnock Burn is a tributary of the river Dee, one of Scotland’s major east coast salmon rivers. A total trap captures all emigrating smolts and a representative sam‐ ple of the fish has a sample of scales removed for aging purposes. The annual smolt age composition has been provided, for inclusion in the SGBICEPS master file. The annual smolt age composition for the Girnock Burn for the period 1990 to 2008 is shown in Figure 4.9.4.

ICES SGBICEPS REPORT 2010 | 65

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

5 1 7 991 997 003 1990 1 1992 1993 1994 199 1996 1 1998 1999 2000 200 2002 2 2004 2005 2006 200 2008

1 2 3 4

Figure 4.9.4. Annual smolt age composition in the Girnock Burn, 1990–2008.

Two‐ and three‐year old smolts comprise the majority of the run throughout the Gir‐ nock Burn series with only small percentages of S1 or S4 smolts being recorded. Throughout the time period there has been a tendency for an increase in the propor‐ tion of S2s and a decrease in the proportion of S3s. In conclusion, the summary biological characteristics of the North Esk and Girnock Burn smolts presented here provide a valuable addition to the SGBICEPS master file and for use in investigating possible links between the trends observed and both en‐ vironmental drivers and abundance indicators.

4.10 Burrishoole wild salmon census programme The Study Group received new information in relation to the Burrishoole monitored river programme in Ireland.

The Burrishoole system (9o 55ʹW 53o 55ʹN) is located in the Nephin Beg mountain range in the west of Ireland. The freshwater catchment has a total area of 8949 ha and the main land uses in the area are subsistence agriculture and forestry. There are three main lakes, Lough Furnace (brackish lake ‐ 141 ha), and two freshwater lakes, Lough Feeagh (410 ha) and Bunaveela L. (46 ha). Research into stock dynamics of salmon, sea trout and eels commenced in 1955 and during the 1960s a partial fish cen‐ sus programme commenced with the construction of upstream and downstream trapping facilities on the Mill Race. Full census commenced in 1970 with the comple‐ tion of the Salmon Leap trap. Census data is used to determine annual spawning es‐ capement, survival between generations, survival from ova to smolt, smolt to adult, adult to kelt and kelt to second spawner. The trapping facilities also enable the collec‐ tion of biological data such as length, weight and sex of juveniles and adults as well as biological samples for ageing and genetic studies. In conjunction with the wild census programme a salmon ranching programme, us‐ ing smolts originally derived from wild Burrishoole 1SW salmon has been in place since 1964. Since the 1980s all ranched smolts have been microtagged prior to release and data from the ranching programme has been incorporated into the Irish national

66 | ICES SGBICEPS REPORT 2010

microtagging programme (O’Maoiléidigh et al., 1995). Information from this pro‐ gramme, used in conjunction with information from the trapping facilities and rod fishery at Burrishoole, provide additional information on a broad range of population characteristics associated with the Burrishoole stock: adult survival rates to the coast, exploitation rates by commercial and recreational fisheries and spawning escape‐ ment. In addition to the census programme, the Burrishoole catchment has a well estab‐ lished environmental monitoring programme with instrumented platforms in place in two lakes (Lough Feeagh and Lough Furnace) to continuously monitor a range of climate and aquatic variables http://lakes.gleon.org/lakes/. Physical and chemical pa‐ rameters in three of the main river sub‐catchments are also monitored using high fre‐ quency automatic monitoring stations. Automated and manual weather stations, in combination with a network of water level recorders and rain gauges, monitor cli‐ matic and hydrological changes in the catchment. A set of biological indices are also routinely measured, including lake chlorophyll a, zooplankton and river macroinver‐ tebrates. Quantification of juvenile salmonid populations, yellow eel stocks and silver eel escapements are conducted annually throughout the catchment.

Available time-series of data on biological characteristics of Burrishoole wild salmon: Smolts Length and weight 1991 onwards Sex 1991 onwards Run timing 1970 onwards Survival to freshwater 1970 onwards Adults Length and weight 1980 onwards Sex 1970 onwards Run timing 1970 onwards

Adult Returns Data from the census programme show that the number of wild grilse returning to Burrishoole has fluctuated considerably over the time period particularly in the 1970s (Figure 4.10.1). Overall returns to freshwater have ranged from a maximum of 1777 recorded in 1973 to a minimum of 252 recorded in 1990. The conservation limit (CL) for the Burrishoole river system has been calculated at 615 spawning salmon (Na‐ tional Salmon Fisheries Review, 2008) and although this was met consistently from 1969 to 1976, it has only been achieved in four of the subsequent years since 1976. One of the factors known to have impacted negatively on the numbers of fish return‐ ing to freshwater in Burrishoole has been the level of coastal exploitation in home waters. Following the cessation of mixed stock drift netting in 2007 the number of grilse recorded at Burrishoole exceeded the CL. However, in both 2008 and 2009 the spawning stock again dropped below its CL. This recent low return rate to freshwater has occurred despite the absence of a coastal fishery and with a catch and release rod fishery.

ICES SGBICEPS REPORT 2010 | 67

2000 14.0 wild grilse 1800 % return to freshwater

m 12.0 1600 r 1400 10.0

1200 8.0 1000 6.0 800

600 4.0 % return to freshwate 400

Number of grilse counted grilseupstrea counted of Number 2.0 200

0 0.0 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007

Year

Figure 4.10.1. Numbers and percentage of wild 1SW salmon (grilse) returns to Burrishoole, 1971– 2008.

Adult salmon scale samples from Burrishoole and other river systems suggest that marine growth influences survival, particularly during the late summer and early winter of the first year at sea and that the rate of marine survival for a cohort could be predicted early in the year from the pattern of circuli numbers of first returning 1SW (Peyronnet et al. 2007).

Adult run timing The timing of return to freshwater for 1SW salmon has shown major fluctuations in recent years. The early component of the run (May to June) has ranged from 60.1% in 2002 to 1.4% in 2006. In years of low early summer rainfall, fresh fish arriving from the sea into Lough Furnace can be delayed in their upstream migration through the traps into Lough Feeagh. Thus, their arrival in the upstream traps does not necessar‐ ily indicate arrival time in the Burrishoole system. However, the recent low 1SW re‐ turn rates during early summer have occurred in years of high summer rainfall and in the absence of a coastal fishery, thereby implying that fish are arriving later, rather than reflecting a delay in upstream migration. There has been considerable fluctuation in the late season run (October to December) from 23% in the early 1970s to 13.7% in the 1990s. In recent years the late arrival in freshwater has not resulted in a corresponding increase in the proportion of late‐ running fish, which has been less than 4% of the total run.

Smolt output The variation in freshwater adult returns is also reflected in the wild smolt output. The maximum number was 16 136 recorded in 1976 and the minimum was 3794 re‐ corded in 1991.The main period of downstream smolt migration through the traps generally occurs during April and May. Byrne et al. (2003), identified two groups of environmental variables which had a significant influence on the daily salmon smolt output. One group of variables, dominated by photoperiod and temperature, oper‐ ated prior to the smolt run and were considered as variables regulating the develop‐ ment of smoltification. The second group of variables, dominated by light and water

68 | ICES SGBICEPS REPORT 2010

level, operated within the smolt run and were considered as factors involved in con‐ trolling daily smolt migration. Examination of the time series of smolt migration since 1970 suggests that the salmon smolt run is tending to be earlier with a significant trend towards an earlier commencement date (5% of the run) (Figure 4.10.2).

5% 50% 95% Linear (5%) Linear (50%) Linear (95%) 170

160

150

140

130

120 Days from 1stJanuary

110

100 1970 1975 1980 1985 1990 1995 2000 2005 2010 Year

Figure 4.10.2. Annual smolt run timing data for the Burrishoole system 1971–2008. The y‐axis de‐ notes the number of days, from the first of January, taken for 5%, 50% and 95% of the smolts to migrate downstream through the traps. The solid line is the trend for 5% of the run, the dashed line for 50% and the dot‐dash line for 95%.

The survival of wild salmon ova to smolt at Burrishoole has varied between 0.32% and 1.22% with no significant trend over the time period (Figure 4.10.3). Examination of decadal periods from 1970 to 2008 indicates that survival rates increased from an average of 0.5% in the 1970s to 0.7% in the following decades. The absence of signifi‐ cant differences in the mean values of smolt/egg between any periods examined, suggested that whatever changes occurred to the overall relationship over time were essentially quite small, (Crozier et al., 2003). The survivals shown in Figure 4.10.3 assume a 0% contribution from ova of ranched stock.

ICES SGBICEPS REPORT 2010 | 69

Burrishoole wild egg to smolt survival (assuming 0% survival reared egg to smolt)

1.4 1.2 1 0.8 0.6

% survival 0.4

0.2 R2 = 0.0927 0 1970 1980 1990 2000 ye ar

Figure 4.10.3. Burrishoole wild salmon egg to smolt survival, 1971–2008.

5 Exploratory analyses

The Study Group completed a number of exploratory analyses using the biological characteristics data sets described in Section 3.2. While the very variable nature and quality of the data for different stocks made it difficult to test hypotheses using the full data set, it was possible to continue to explore relationships within the data that provide pointers to the mechanisms controlling salmon survival at sea. Specific hy‐ potheses that are considered in the following analyses include:  Changes in salmon condition factors are independent of oceanic conditions [Section 5.1];  There are no temporal trends or spatial patterns in the biological character‐ istics data sets [Sections 5.2 and 5.3];  There are no relationships between mean river‐age, sea‐age, length, weight, condition or run size observed in monitored stocks over time [Sec‐ tion 5.5];  There have been no changes in the mean age of emigrating smolts or mean river age of returning adults (by sea age) among rivers and over time [Sec‐ tions 5.6, 5.7, 5.8, 5.9];  There is no relationship between the size of returning adult salmon and es‐ timates of pre‐fishery abundance [Section 5.10];  There is no evidence of a change in overall (egg‐adult) mortality over time [Section 5.11]. Not all of the analysis related to these hypotheses could be completed and some work is on‐going.

70 | ICES SGBICEPS REPORT 2010

5.1 Analyses of long-term variation in condition factor in relation to ocean climate The Study Group reviewed information on salmon condition factor during the first meeting of the SGBICEPS group and recommended further investigation (ICES, 2009a). Initial analyses (1993–2006) of condition factor variation in relation to ocean climate change (Todd et al., 2008) involved interrogation of Marine Scotland time‐ series for salmon from the mixed stock fishery at Strathy Point (N. Scotland) and the in‐river net fishery on the River North Esk (E Scotland). Those analyses included only 1SW fish and were constrained by the netting season to the period late June–end Au‐ gust. Analyses of sea surface temperature (SST) variations focused on the Norwegian Sea (67.5°N 4.5°E) and utilised the NOAA OISSTv2 SST database, which is based upon a combination of satellite telemetry and in situ recordings. This time‐series dates from December 1980 and is available on a monthly basis, and with a gridded resolution of one degree latitude/longitude. Analyses of condition factor variation and ocean climate (SST) are ongoing and pres‐ ently include both 1SW and 2SW fish and multiple rivers in Scotland (Tweed, Tay, Spey, North Esk) and Wales (Dee) as well as three rivers from southern, central and northern Norway. These data will permit analysis of salmon time‐series from NEAC southern and northern populations, some of which date back to the early 1960s. With the incorporation of data for 2SW fish, SST analyses have had to be extended to the western North Atlantic in order to include the foraging areas of coastal Greenland and are not yet complete. One complication of this analytical extension of condition factor lies in the choice of SST database. OISSTv2 is not available prior to 1980 and alternatives must be sought. The ERSST, Kaplan and HadSST time‐series have been investigated and, although they share common data, they differ in how the data have been processed and inter‐ polated to provide gridded predictions for data‐poor areas of ocean, or periods of missing values. As a generalisation, all these databases show closely similar trends in SST variation over time, but they differ in the monthly and spatial detail. SST time‐series present analytical difficulties because of the problem of autocorrela‐ tion – that is, if it is anomalously warm in one month it probably will be also in the preceding and succeeding months. The monthly SST anomaly values are not, there‐ fore, independent. The outcomes of our analyses (such as whether or not there is a significant correlation between ocean climate change and salmon condition) rely ul‐ timately on the correlations derived not for the raw SST anomalies, but for their re‐ siduals (i.e. departures of actual monthly observations from the predicted, or smoothed, trend values). Because these databases often show slightly different SST values for a given grid box in a given month, the residuals derived by de‐trending in order to remove autocorrelation are not the same. As a consequence, the different SST databases ultimately yield different residuals and final analytical outcomes, when correlated with the residuals from de‐trending (smoothing) the salmon condition fac‐ tor time‐series. That is to say, depending upon the selected SST database, the final analytical outcome (e.g. a correlation between ocean warming and salmon condition factor) will differ.

ICES SGBICEPS REPORT 2010 | 71

OISST 250 km kernel (x axis) versus HadSST2 250 km OISST 250 km kernel (x axis) versus ERSST 250 km kernel (y axis) RESIDUALS kernel (y axis) RESIDUALS

1 0.25

0.8 0.2 0.15 0.6 0.1 0.4 0.05 0.2 0 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 0 -0.05 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 -0.1 -0.2 -0.15 y = 0.1716x - 7E-05 -0.4 y = 1.0281x - 2E-05 2 -0.2 R = 0.1214 R2 = 0.6376 -0.6 -0.25 -0.8 -0.3

OISST 250 km kernel (x axis) versus Kaplan 250 km kernel (y axis) RESIDUALS

0.8

0.6

0.4

0.2

0 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 -0.2

y = 0.519x - 7E-05 -0.4 R2 = 0.3387

-0.6

Figure 5.1.1. Correlations of residuals for OISSTv2 versus HadSST2, ERSST and Kaplan Sea Sur‐ face Temperature databases for the eastern North Atlantic. Monthly values (December 1980 – March 2008) are temperature anomalies derived from de‐trending (smoothing) the average tem‐ perature recorded for a spatial kernel of standard deviation 250 km, and centred on the grid box at 67.5°N 4.5°E. The residuals are the observed monthly departures from the smoothed (de‐trended) time‐series.

This problem has been investigated in detail by Hughes et al. (2009). They affirmed the differences inherent in the data comprising various time‐series. For example, they noted that particularly prominent monthly SST anomalies in the North Sea could be up to 6 months apart in different time‐series. Discrepancies in the residuals of the magnitude illustrated in Figure 5.1.1, and in such timing of unusually anomalous events (Hughes et al., 2009), present considerable analytical and interpretational chal‐ lenges. For analyses post‐December 1980, the OISST series probably remains the da‐ tabase of choice. Hughes et al. (2009) noted generally good concordance of HadSST2 to OISSTv2 and in situ observations, though HadSST2 did commonly show marked discrepancies at high latitude in the Arctic, perhaps due to sea ice. One solution to this seemingly intractable problem might be to undertake an ensemble analysis with the salmon condition factor time‐series that incorporates all the varying SST data‐ bases. This is subject to ongoing investigation.

5.2 Biological characteristics data sets – temporal trends At the first Study Group meeting, preliminary examination of the various stock‐ specific data sets indicated substantial evidence of variability in different biological characteristics both over time and among stocks (ICES, 2009a). These data sets were subsequently analysed using the non‐parametric Mann‐Kendall statistic (Mann, 1945; Kendall, 1975) and the statistical programming environment R (R Development Core Team, 2007) to investigate possible time trends. One of the reasons this method was selected was that it does not assume any particular functional form for the trend (e.g. linear, exponential). These initial analyses were performed over a standardised time period starting from 1984 and extending to the most recently available data, typically 2007 (ICES, 2009a).

72 | ICES SGBICEPS REPORT 2010

At, and subsequent to, the second meeting, a number of additional data sets were made available; a number of which had slightly shorter time series (see details in Sec‐ tion 3.2). The Mann‐Kendall statistic was once again used to investigate the time trends in these extended data sets, but over a standardised time period starting from 1989 and extending to the most recently available data (mainly 2008). A number of the data sets were also updated at the second meeting with the addition of data from the most recent year(s). Essentially, the Mann‐Kendall statistic assesses each data point and counts the num‐ ber of data points in future years that exceed it (assigned +1) and also the number that are less (assigned ‐1). These values are summed over all data points. Thus, if there are

n observations in the time series Yk (k=1,...,n) then the statistic is:

n n M  I(Yk ,Y j ) k11jk

where I(,) is an indicator variable defined by the sign of D  Yk  Y j . If D is positive then I = 1, if D is negative then I = ‐1, if D = 0 then I = 0. The null distribution of the statistic M (i.e. assuming no trend) was calculated by Monte‐Carlo simulation. The observations were re‐ordered at random so that they were potentially assigned to different years and the value of M calculated. This was repeated 1000 times. These values of M under the null hypothesis were compared against the observed value of M to estimate the p‐value (Manly, 2001). The null hypothesis was that there is no trend. It was recognised that a potential problem with this approach is that it does not taken account of any correlation be‐ tween successive or neighbouring years. The presence of such correlations might af‐ fect the p‐values obtained and may require further consideration in future. The results are presented in Table 5.2.1 Missing values indicate no time series avail‐ able; ‘o’ indicates a non statistically significant trend (P>0.05); ‘‐‘ is a negative trend (p<0.05); and ‘+’ is a positive trend (p<0.05). It is evident that there are significant trends over time for many of the variables explored.

ICES SGBICEPS REPORT 2010 | 73

Table 5.2.1. Trends in biological characteristics over time: ‘o’ means not enough evidence at the 5% level to detect a trend. ‘+’ is a positive trend (p>0.05), ‘‐‘ is a negative trend (p<0.05). 2SW 1SW 2SW 1SW

PS

run run in in in in

in

date

run in in

date

mean

in

run mean

run female PS 2SW 1SW female status complex

age

female series maiden maiden length weight condition length weight condition

age

condition weight length

River Prop Sea Mean Prop. Prop. Prop. Prop. Median Prop. PS PS 1SW 1SW 1SW 2SW 2SW 2SW PS Country Stock Hatchery/Wild Time Latitude Stock Prop Prop Stock NAC (N) Canada Western Arm Brook W 1971‐06 51.2 + ‐ 0 ‐ +000+++‐ 0+++000000 Canada De la Trinité W 1980‐09 49.4 ‐‐0+++0++0+00++0 Canada Saint‐Jean W 1981‐09 48.8 0 + ‐ 0++‐ +00+000+0 Canada Middle Brook W 1975‐05 48.8 0 0 ‐ 00000++0 ++0‐ 00 0 Canada Conne River W 1986‐06 47.9 ‐ 000+‐ ++++00 000000‐ + Canada Miramichi W 1971‐07 47.0 ‐‐+00++0‐ +00++0‐ ++00 NAC (S) Canada Nashwaak W 1972‐08 46.0 0 0 0 ‐‐+ ‐ 0+ 00+‐ Canada St John (Mactaquac) W 1978‐08 45.3 ‐‐‐‐‐+ ‐ 0000‐‐ 00 +‐ Canada St John (Mactaquac) H 1978‐08 45.3 ‐‐‐‐‐+ ‐ 000‐‐‐ 00 +‐ Canada La Have W 1972‐08 44.4 0 ‐ 0+00000++0++00+0++00 Canada La Have H 1972‐08 44.4 ‐ 00‐ 00000++0++0+00+000 USA Penobscot H 1978‐08 44.5 ‐‐‐0 ‐ + ‐ 00++00+000‐ +0+‐ N NEAC Finland/Norway Teno W 1972‐07 70.8 0 + ‐ 00+00000000000‐ Finland/Norway Näätämöjoki W 1975‐06 69.7 0 + ‐ +00+0000000‐ 0 ‐ 0 Russia Tuloma W 1983‐08 68.9 0 0 + ‐ 0000000‐‐0000000+0 Norway Vestre Jakobselv W 1989‐08 70.1 + ‐ 00+000‐‐‐‐ Norway Årgårdsvassdraget W 1992‐08 64.3 ‐ 00‐ +00000‐ 0000 Norway Nausta W 1989‐08 61.3 0 0 + ‐ +000‐ 000 Norway Gaula Sogn og Fjordane W 1989‐08 61.2 0 0 + ‐ +00‐ 00‐‐ Norway Etneelva W 1989‐08 59.4 ‐‐0 ‐ 0000‐‐‐0 Norway Skienselva W 1989‐08 59.1 0 0 + ‐ ++00000‐ . Norway Numedalslågen W 1989‐0859.10 0000000‐‐‐‐ Norway Enningdalselva W 1990‐08 58.6 0 ‐ + ‐ +00‐ 0000 Norway Gaula W 1989‐08 63.3 0 ‐ 000+000‐ 00000 Iceland (N&E) Laxa I Adaldalur W 1974‐08 65.6 ‐‐+ ‐ 00‐ Iceland (N&E) Hofsa W 1971‐08 65.4 ‐‐+ ‐‐‐ 0+ S NEAC Iceland (S&W) Nordura W 1968‐08 64.6 ‐‐+ ‐‐‐ 0+ Iceland (S&W) Ellidaar W 1949‐08 64.1 ‐‐+ ‐‐‐ +0 UK (Scot) N. Esk W 1981‐08 56.7 0 0 0 ‐‐‐0 ‐ 0 UK (NI) Bush W 1973‐07 55.1 ‐ +0000‐ 0 ‐‐+000000000+ ‐ UK (E&W) Lune W 1987‐0854.00 0000 0000‐ 0 UK (E&W) Dee W 1937‐08 53.4 ‐‐‐+ ‐ 00+++++‐‐‐0+ 0 UK (E&W) Wye W 1910‐07 51.6 0 + ‐‐++‐ ++0000‐‐0 UK (E&W) Frome W 1968‐08 50.7 ‐‐ 00 France Bresle W 1984‐0850.1 00000 0 Baltic Sweden Kalix W 1980‐08 65.8 + Finland/Sweden Tornionjoki W 1980‐09 65.5 0 ‐ +0 00 00 ‐ Sweden Ume/Vindel W 1987‐08 63.8 + + 0 Sweden Ume/Vindel H 1974‐08 63.8 0 0 0 0 0 0 0 + + 0 + Summary ‐ all areas No. of stocks for which data available 33131432323636273132283031261717172322151313 No. of stocks with significant declining trend125418129111277713622232116 No. of stocks with significant increasing trend4223712968116555541210261 No. of stocks with no significant trend 176 811131516202114151813151011141810126 6 % of stocks with significant declining trend 363829563825314 62225234223121212139 7 846 % of stocks with significant increasing trend 12 15 14 9 22 33 25 22 26 34 21 17 16 19 29 24 6 9 45 13 46 8 % of stocks with no significant trend 52465734414244746844546042585965827845804646

5.3 Biological characteristics data sets – spatial patterns At the first meeting, the Study Group examined two approaches for assessing pat‐ terns in the changes in biological characteristics over broader spatial scales (ICES, 2009a). The first approach used a standardised (z‐score) analysis to examine the trend in mean smolt age, by way of an example, for different stock complexes. The second approach used meta‐analysis. The latter technique was applied to all the biological characteristics variables. The z‐score and meta analysis approaches provided broadly consistent results in respect of smolt age. Subsequent to the second meeting, the meta analysis was repeated using the extended data set available at that time. The meta‐analysis approach statistically combines the results of several studies (in this case different rivers) to address a shared research hypothesis. Just as individual studies summarise data collected from many participants in order to answer a spe‐ cific research question (i.e. each participant is a separate data‐point in the analysis), a meta‐analysis summarises data from individual studies that concern a specific re‐ search question (i.e. each study is a separate data‐point in the analysis). This is normally done by identification of a common measure of effect size. An effect size is a statistical measure portraying the degree to which a given event is present in a sample (Cohen, 1969). The type of measure (e.g. standardized mean difference) is called the effect, and its magnitude is considered an effect size. Different measures of effect size are calculated for different types of primary data, commonly modelled us‐ ing a form of meta‐regression. In this instance, all effect sizes were calculated from

74 | ICES SGBICEPS REPORT 2010

Pearson correlation coefficients between a given variable and year. Resulting overall averages when controlling for study characteristics can be considered meta‐effect sizes, which are more powerful estimates of the true effect size than those derived in a single study under a given single set of assumptions and conditions. Fail‐Safe Tests were also applied in these analyses. These tests are statistical methods for estimating the magnitude of the publication bias known as the file‐drawer problem (Rosenthal, 1979). These techniques calculate the number of non‐significant, unpub‐ lished studies (having effect size = 0) that need to be added to a summary analysis in order to change the results from significant to non‐significant. A large fail‐safe num‐ ber indicates that many unpublished studies are required to change the statistical re‐ sults, and thus one may be more confident in the results from the summary analysis. Thus, for example, in the case of mean river age (Figure 5.3.1) we need 551.9 rivers showing no correlation (r = 0) in order to say that there is no overall change in mean river age. The following stock complex groupings were used in the meta analysis [N.B. the standard WGNAS abbreviations NAC (North American Commission) and NEAC (North East Atlantic Commission) have been applied]:  Baltic – rivers draining into the Baltic Sea.  Northern NAC ‐ Rivers Western Arm Brook, Middle Brook, Conne and Miramichi. [N.B. The data for the De La Trinité and Saint‐Jean rivers (in‐ cluded in other analyses) were not available at the time that the meta analysis was undertaken and have been excluded here.]  Southern NAC ‐ Rivers Nashwaak, St John (Mactaquac), La Have and Pe‐ nobscot.  Northern NEAC north – rivers in Russia, Norway, Finland and Iceland (N&E).  Southern NEAC south ‐ rivers in UK, Ireland, France and Iceland (S&W). This approach was used to explore relationships for most of the available biological characteristics and a selection of results appears in Figure 5.3.1. These provide the central tendencies (means) for five stock groups and the total effect – if the error bar crosses the vertical zero‐line the effect is non significant at the p = 0.05‐level. These analyses were performed over a standardised time period starting from 1989 and ex‐ tending to the most recently available data, typically 2008. All available data were included, although the completeness of the data sets varied considerably for the dif‐ ferent biological characteristics, as indicated in Table 3.2.1. Median and mean run dates were converted to day of the year prior to analysis. All the results arising from the meta analysis are summarised in Table 5.3.1 (where ‘o’ denotes a non‐significant relationship, ‘+’ indicates a significant increase relative to the mean and ‘‐‘ denotes a significant decrease). These analyses indicated a number of significant trends over time for certain variables at the stock complex level (Table 5.3.2).

ICES SGBICEPS REPORT 2010 | 75

Figure 5.3.1. Example meta‐analysis plots for selected biological characteristics – for stock group‐ ings see text.

Mean river age Mean sea‐age

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=551.9 Rosenthal's=0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 -0.8 -0.4 0.0 0.4 0.8 Decreasing Increasing Decreasing Increasing river age river age sea age sea age

Mean total age Mean (adult) run date

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=94.6

-1.2 -0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 Decreasing Increasing Decreasing Increasing total age total age mean run date mean run date

76 | ICES SGBICEPS REPORT 2010

Proportion of 1SW fish in run Proportion of 2SW fish in run

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=0

-0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 0.8 Decreasing Increasing Decreasing Increasing proportion proportion proportion proportion

Proportion of previous spawners in run Mean length – 1SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=23.9 Rosenthal's=58.4

-0.6 -0.4 -0.2 0.0 0.2 0.4 -0.8 -0.4 0.0 0.4 0.8 Decreasing Increasing Decreasing Increasing proportion proportion mean length mean length

Mean weight – 1SW Mean condition factor – 1SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=28.2 Rosenthal's=0.0

-0.8 -0.4 0.0 0.4 0.8 1.2 -0.8 -0.4 0.0 0.4 0.8 1.2 Decreasing Increasing Decreasing Increasing mean weight mean weight condition factor condition factor

ICES SGBICEPS REPORT 2010 | 77

Mean length – 2SW Mean weight – 2SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=245.8 Rosenthal's=254.4

-0.8 -0.4 0.0 0.4 0.8 1.2 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Decreasing Increasing Decreasing Increasing mean length mean length mean weight mean weight

Mean condition factor – 2SW Mean length – previous spawners

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0.0 Rosenthal's=0

-0.8 -0.4 0.0 0.4 0.8 1.2 -0.8 -0.4 0.0 0.4 0.8 Decreasing Increasing Decreasing Increasing condition factor condition factor mean length mean length

Mean weight – previous spawners Mean condition factor – PS

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=0.7

-0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 0.8 1.2 Decreasing Increasing Decreasing Increasing mean weight mean weight condition factor condition factor

78 | ICES SGBICEPS REPORT 2010

Prop. female – 1SW Prop. female – 2SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=0

-0.6 -0.4 -0.2 0.0 0.2 0.4 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Decreasing Increasing Decreasing Increasing proportion proportion proportion proportion

Prop. female – previous spawners Prop. maiden spawners – 1SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=0

-0.4 -0.2 0.0 0.2 0.4 0.6 -0.4 -0.2 0.0 0.2 0.4 Decreasing Increasing Decreasing Increasing proportion proportion proportion proportion

Prop. maiden spawners – 2SW Stock size – 1SW

Total Total

Baltic Baltic

N NAC N NAC

S NAC S NAC

N NEAC N NEAC

S NEAC S NEAC Rosenthal's=0 Rosenthal's=97.5

-0.4 -0.2 0.0 0.2 0.4 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 Decreasing Increasing Decreasing Increasing proportion proportion stock size stock size

ICES SGBICEPS REPORT 2010 | 79

Stock size – 2SW

Total

Baltic

N NAC

S NAC

N NEAC

S NEAC Rosenthal's=0

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 Decreasing Increasing stock size stock size

80 | ICES SGBICEPS REPORT 2010

Table 5.3.1. Results of meta analysis at the stock complex level ‐ indicating significant increase (+) or decrease (‐) relative to the mean (o denotes non‐significant relationship). 2SW 1SW ‐ ‐ PS 2SW 1SW

1SW 2SW PS 1SW 2SW PS spawners spawners run run

in in in ‐ ‐ ‐

‐ ‐ ‐

date 1SW 2SW PS age run age 1SW 2SW in in

date age ‐ ‐ ‐

‐ ‐

in

run

length weight length weight length weight run total sea river complex size size PS 2SW 1SW maiden maiden female female female

series

Mean Mean Condition Mean Mean Condition Mean Mean Condition Prop. Prop. Prop. Prop. Median Mean Prop. Mean Prop. Mean Prop. Stock H/W Time Latitude Stock Stock Mean Prop. All 1989‐08 ‐ o ‐ oo‐‐oo+‐‐o ‐‐ooo+ooooo NAC (N) W 1989‐08 47.0 ‐ 51.2 o ‐ oo+‐‐ooo+oooooo+o‐ o ‐ oo NAC (S) H/W 1989‐08 44.4 ‐ 46.0 ‐‐ooo‐‐oo‐‐oooooooooo+oo N NEAC W 1989‐08 65.4 ‐ 70.8 o + ‐ +o ‐ ++‐‐‐‐‐ooooooo S NEAC W 1989‐08 50.1 ‐ 64.6 o o ‐‐‐ +ooo ‐‐o ‐‐oo‐ ooo+

Baltic H/W 1989‐08 63.8 ‐ 65.8 o + o o ‐ +oo

ICES SGBICEPS REPORT 2010 | 81

Table 5.3.2. Summary of significant trends in biological characteristics at the stock complex level derived from the meta analysis (PS = previous spawners).

Stock complex Increasing trend Decreasing trend NAC (N) Mean total age Stock status ‐ 2SW Mean length ‐ 1SW Mean run date (earlier) Mean weight ‐ PS Median run date (earlier) Proportion female in 1SW Proportion female in PS NAC (S) Proportion female in PS Stock status ‐ 1SW Stock status ‐ 2SW Mean run date (earlier) Median run date (earlier) Proportion PS in run Mean length ‐ 1SW N NEAC Stock status ‐ 2SW Mean river age Mean sea age Proportion 1SW in run Proportion 2SW in run Mean length ‐ 1SW Proportion PS in run Mean weight ‐ 1SW Mean Condition Factor ‐ 1SW Mean length ‐ 2SW Mean weight ‐ 2SW S NEAC Mean run date (later) Mean river age Proportion female in PS Mean sea age Mean total age Mean length ‐ 1SW Mean weight ‐ 1SW Mean length ‐ 2SW Mean weight ‐ 2SW Mean weight ‐ PS Baltic Mean total age Proportion 1SW in run Proportion 2SW in run

5.4 Overview of preliminary analyses of temporal and spatial trends Preliminary analyses of the biological characteristics data sets available to the Study Group focused mainly on the 1SW and 2SW sea‐age classes. Some data in respect of older sea‐age classes were available to the Study Group, but these were restricted to relatively few stocks and sample sizes for these stock components were believed to be small in some cases. In light of these caveats, and given the time available, the charac‐ teristics of these older stock components have not, as yet, been examined. However, some provisional analyses of fish categorised as previous spawners (PS) have been included in the following summary. For the purposes of this summary, the separate hatchery and wild data sets for the St John (Mactaquac) and La Have Rivers in Can‐ ada have been treated as separate stocks.

82 | ICES SGBICEPS REPORT 2010

Stock status Overall, 36% of the individual salmon stock data sets (Table 5.2.1) indicate a signifi‐ cant declining trend in the stock status variable (all sea‐age groups combined) over the time series analysed (1989–2008). Stocks with declining trends were more preva‐ lent in North America and the NEAC Southern area; only 2 stocks (17%) in the NEAC Northern area indicated a negative trend. Overall, only two North Atlantic stocks, the Western Arm Brook in Canada and the Vestre Jakobselv in Norway, showed a sig‐ nificant positive trend over the period. These are among the most northerly stocks for which data are available. Two of the three Baltic stocks for which data were available also showed a significant positive trend over the period. Meta analysis at the stock complex level and split by sea‐age (Table 5.3.1) indicates a significant decrease in the stock status variable for both 1SW and 2SW salmon in the southern part of the NAC and for 2SW salmon in the northern part of the NAC area, but a significant increase in stock status for 2SW salmon in the NEAC Northern area.

Mean river age The majority of the data sets (56%) indicate a significant decline in mean river (smolt) age over the standard time series (1989–2008) (Table 5.2.1). Relatively few stocks (9%) indicate a significant positive trend over the period and these are restricted to the NAC Southern area (only the La Have River ‐ wild) and the NEAC Northern area (the two most northerly stocks for which data were available). The data are summa‐ rised in Figure 5.4.1; limited data not included for the Baltic.

100 90 River age 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.1. Percentage of rivers in each stock complex indicating significant trends in mean river age.

Despite the evidence of declining smolt age in all areas of the North Atlantic, the meta analysis only indicates a significant decrease in mean river age for the NEAC Northern and Southern stock complexes (Table 5.3.1). The decline in mean smolt age may be the consequence of an increase in growth rate as the faster growing parr migrate to sea earlier (Metcalfe et al., 1989; Økland et al., 1993). The increase in growth rate may relate to an increase in temperature (Elliott et al. 2000; Jonsson et al., 2005), and/or an increase in growth as a result of density‐ dependent processes (Gibson, 1993; Jenkins et al., 1999; Imre et al., 2005; Lobón‐ Cerviá, 2005), and/or increased freshwater production. A possible consequence of the

ICES SGBICEPS REPORT 2010 | 83

increase in growth rate and smolts migrating at an earlier age is to dampen the im‐ pact of an increase in marine mortality. This assumes that the higher survival rate to smolt for a one‐year‐old smolt (S1) is not outweighed by their higher marine mortal‐ ity. A decline in smolt age may affect reproductive success as egg size is smaller for S1 as opposed to S2 smolts of the same sea‐age and early survival (egg to swim‐up) may also be lower (Moffett et al., 2006).

Mean sea-age The majority of stocks demonstrate significant trends in mean sea‐age since 1989 (Ta‐ ble 5.2.1), although patterns are variable. In total, 38% of the data sets indicate a sig‐ nificant decline in mean sea‐age, 22% a significant increase and 41% no apparent trend. There appear to be marked differences between the NAC Northern area, where 3 of the 4 stocks demonstrate significant increases in mean sea‐age and there are no declining trends, and the NAC Southern area where 4 of the 6 stocks show significant decreases and there are no increasing trends (Figure 5.4.2). The NEAC Southern area has a similar pattern to the NAC Southern area, while in the NEAC Northern area some stocks have decreasing trends, some increasing trends and some indicate no apparent trend.

100 90 Sea age 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.2. Percentage of rivers in each stock complex indicating significant trends in mean sea age.

Meta analysis at the stock complex level indicates a significant increase in mean sea‐ age in the NEAC Northern area and a significant decrease in the NEAC Southern area (Table 5.3.1).

Mean/Median run date Data on run timing are relatively sparse for stocks from the NEAC area. However, the available data suggest marked differences in the trend in run timing (data for all sea‐age classes combined) between the NAC and NEAC areas (Table 5.2.1). Around half the stocks in the NAC area demonstrate significant negative trends (i.e. towards earlier running) in adult run date, and no stocks demonstrate trends towards later running. In contrast, two of the three stocks in the NEAC area, for which such data are available, indicate significant positive trends (i.e. later running).

84 | ICES SGBICEPS REPORT 2010

Meta analysis at the stock complex level indicates a significant decrease in mean run date in the NAC area (both Northern and Southern) and a significant increase in mean run data in the NEAC Southern area (Table 5.3.1).

Proportion of different sea-age classes in the run The proportion of 1SW salmon in returning adult runs shows a significant increase in many stocks (33% of all stocks) (Table 5.2.1). A high proportion of the stocks in the NAC and NEAC Southern areas display similar positive trends (Figure 5.4.3). How‐ ever, there is a more variable pattern in the NAC and NEAC Northern areas, with more stocks having a significant negative trend than a positive one. Meta analysis at the stock complex level indicates a significant decrease in the proportion of 1SW salmon in the returning stocks for the NEAC Northern area and the Baltic (Table 5.3.1).

100 90 Prop. 1SW 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.3. Percentage of rivers in each stock complex indicating significant trends in the pro‐ portion of 1SW fish in the run.

The proportion of 2SW salmon in returning adult runs is increasing in 25% of stocks, decreasing in 31%, and shows no significant trend over the time period in 44% of stocks (Table 5.2.1). It should be noted, of course, that for many stocks 1SW and 2SW fish make up the majority of the run, and so a decrease in the proportion of one sea‐ age component will be reflected in an increase in the other. Consistent with this, the meta analysis indicates a significant increase in the proportion of 2SW salmon in the returning stocks in the NEAC Northern area and Baltic (Table 5.3.1). The analysis of the proportion of previous spawners (PS) in returning stocks indicates that this is increasing over time for 22% of the available data sets examined; only one stock (4%) shows a significant declining trend (Table 5.2.1). Increasing trends only occur in the NAC and NEAC Northern areas (Figure 5.4.4). There are no significant trends in the NAC Southern area and the only significant negative trend occurs in the NEAC Southern area. Meta analysis indicates a significant increase in the proportion of PS in the returning stocks in the NEAC Northern area, a significant decrease in the NAC Southern area, and no significant relationships elsewhere (Table 5.3.1).

ICES SGBICEPS REPORT 2010 | 85

100 90 Prop. PS 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.4. Percentage of rivers in each stock complex indicating significant trends in the pro‐ portion of 1SW fish in the run.

Mean length, weight and condition – 1SW salmon The available data sets indicate differing patterns in the mean size of returning 1SW salmon both among stocks and over time (Table 5.2.1). Significant increasing size trends are apparent for all of the stocks in the NAC Northern (Figure 5.4.5), whereas none of the stocks in the NAC Southern area show significant increasing trends in mean length, although over half do in respect of mean weight and condition factor; there are no negative trends in either length or weight in the NAC area.

100 90 1SW length 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.5. Percentage of rivers in each stock complex indicating significant trends in the length of returning 1SW fish.

There is little evidence for trends over time in the NEAC Northern area, although there are significant increases in the mean length of 1SW salmon returning to one river (8%). There is greater variability in the NEAC Southern area, with some stocks showing significant decreases in fish size, while others in relatively close geographic proximity show significant increases (Figure 5.4.5). Meta analysis (Table 5.3.1) indicates a significant increase in the mean length of 1SW salmon in the NAC Northern area, but a significant decrease in mean 1SW length in

86 | ICES SGBICEPS REPORT 2010

the NAC Southern area; there are no significant trends in weight or condition factor. There is a significant decrease in mean 1SW length, weight and condition factor for the NEAC Northern area, and a significant decrease in mean 1SW length and weight for the NEAC Southern area.

Mean length, weight and condition – 2SW salmon As with the 1SW fish, the available data sets indicate widely differing patterns in the mean size of returning 2SW salmon both among stocks and over time (Table 5.2.1). Both significant increases and significant decreases in the size of 2SW fish are appar‐ ent for different stocks in the NAC area, but with no clear geographical pattern (Fig‐ ure 5.4.6). In the NEAC Northern area, 5 stocks (42%) indicate a significant decline in the mean length of returning 2SW salmon over the period, with no stocks showing an increase. In contrast, in the NEAC Southern area no stocks indicate a significant de‐ cline in 2SW length (Figure 5.4.6), although there are significant increases in mean weight of 2SW fish for 4 stocks (57%) (Table 5.2.1).

100 90 2SW length 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.6. Percentage of rivers in each stock complex indicating significant trends in the length of returning 2SW fish.

Meta analysis indicates a significant decrease in mean length and weight of 2SW salmon in both the NEAC Northern and Southern areas, but no significant relation‐ ships in the NAC areas (Table 5.3.1). There are also no statistically significant trends in the condition factor of 2SW salmon at the stock complex level.

Mean length, weight and condition – Previous Spawners (PS) The length and weight of previous spawning (PS) fish is increasing in around half the stocks in the NAC area and there are no negative trends in length, weight or condi‐ tion factor (Table 5.2.1). The pattern in trends in the NAC area appears to be consis‐ tent with that seen in the 1SW salmon, the dominant sea‐age group in many of these stocks. In the NEAC Northern area, 2 of the stocks (25%) have declining trends in the length, weight and condition of PS, but no trends in the other stocks. Size data for PS are only available for three stocks in the NEAC Southern area and these indicate no apparent trend in one instance, but decreasing trends in length and weight for the other two stocks (Table 5.2.1). Meta analysis indicates a significant increase in the mean weight of PS in the NAC Northern area, and a significant decrease in the mean weight of PS in the NEAC

ICES SGBICEPS REPORT 2010 | 87

Southern area. There are no statistically significant trends in the condition factor of PS at the stock complex level (Table 5.3.1).

Proportion of female fish The available data sets indicate differing patterns in the proportion of female fish in the returning 1SW, 2SW and PS fish over the period from 1989 (Table 5.2.1). For 1SW salmon, the majority of stocks indicate no detectable trend in the proportion of fe‐ males (78%). There was evidence of a significant decline in the proportion of females in the NAC area, with only one stock (17%) indicated a significant positive trend. There were no significant trends in the NEAC Northern area, while in the NEAC Southern area the proportion of females increased for one stock (Figure 5.4.7). In contrast to the 1SW fish, around half the stocks in the NAC Northern and Southern areas and NEAC Southern area indicated a significant increase in the proportion of females among the 2SW salmon and there were no negative trends (Table 5.2.1; Fig‐ ure 5.4.8). Trends were more variable in the NEAC Northern area, with two stocks indicating a significant decrease in the proportion of female 2SW salmon, while one showed an increase.

100 90 Female ‐ 1SW 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.7. Percentage of rivers in each stock complex indicating significant trends in the pro‐ portion of returning 1SW fish that are female.

88 | ICES SGBICEPS REPORT 2010

100 90 Female ‐ 2SW 80 70 60 % 50 No sig. trend 40 30 Sig. increasing trend 20 Sig. declining trend 10 0 NAC(S) NAC(N) NEAC(S) NEAC(N)

Figure 5.4.8. Percentage of rivers in each stock complex indicating significant trends in the pro‐ portion of returning 2SW fish that are female.

Relatively few data sets are available in respect of PS. The majority of stocks (80%) indicate no detectable trend in the proportion of females (Table 5.2.1). There were increasing trends in the proportion of females for two stocks in the NAC area, and decreasing trends in one stock in the NEAC Northern area. Meta analysis indicates a significant decrease in the proportion of females in return‐ ing 1SW and PS salmon in the NAC Northern area and a significant increase in the PS females in both the NAC and NEAC Southern areas (Table 5.3.1).

Proportion of different sea-age groups among maiden spawners Information on the proportion of different sea‐age class fish among the maiden spawners (fish returning to spawn for the first time) was largely confined to North American stocks, and broadly reflects the proportions of different sea‐age fish in the run (see above). In respect of the proportion of 1SW salmon among the maiden fish, positive trends were evident for 4 (67%) of the stocks in the NAC Southern area, while in the NAC Northern area one stock shows a significant negative trend and there are no stocks with positive trends (Table 5.2.1). Where data were available for the NEAC area (one in the Southern area and one in the Northern area), these both indicated a positive trend in the proportion of 1SW salmon among the maiden spawners. Since 1SW and 2SW fish make up the majority of the run for many stocks, a decrease in the proportion of one sea‐age component will likely be reflected in an increase in the other. Thus, there were significant negative trends for 4 of the stocks in the NAC Southern area and a significant positive trend for one stock in the NAC Northern area. No significant trends in the proportions of maiden spawners were detected at the stock complex level using meta analysis (Table 5.3.1).

5.5 Exploration of two-way relationships At the first meeting, the Study Group completed some preliminary analyses to inves‐ tigate potential inter‐relationships between selected stock characteristics for each river, over a standardised time period from 1984 (ICES, 2009a). These analyses were repeated with the larger data set of rivers available at the second meeting, but using the period from 1989 as the standardised time period (as noted above). Simple linear

ICES SGBICEPS REPORT 2010 | 89

regression models were used to test each relationship since the analyses were essen‐ tially exploratory in nature. The relationships investigated were:  1SW weight v 2SW weight (fish sampled in the same year)  1SW weight (in year ‘n’) v 2SW weight (in year ‘n + 1’)  1SW condition factor v 2SW condition factor (fish sampled in the same year)  1SW condition factor (in year ‘n’) v 2SW condition factor (in year ‘n + 1’)  Mean river age v mean sea‐age  1SW length v 1SW weight  2SW length v 2SW weight  Mean river age v 1SW weight  Mean river age v 2SW weight  1SW weight v total run size  2SW weight v total run size  1SW weight v 1SW run size  2SW weight v 2 SW run size A series of plots were produced for each of the above relationships for each stock for which appropriate data were available (Figure 5.5.1 provides two examples). Each of the individual river plots also includes the p‐value to assess the statistical significance of the slope of the linear regression model. The Study Group noted that even if this p‐ value is not small it does not necessarily mean that there is no relationship – only that we are unable to detect it with these data. Table 5.5.1 summarises the results of all the linear regression models for each of the above relationships ‐ ‘o’ means that there was not enough evidence at the 5% level to detect a trend; ‘+’ denotes a positive trend (p>0.05), and ‘‐‘ denotes a negative trend (p<0.05). The results are discussed briefly below.

90 | ICES SGBICEPS REPORT 2010

Figure 5.5.1. Example two‐way plots for each individual stock of: (a) mean weight of 1SW salmon against the mean weight of 2SW salmon returning in the same year (i.e. from an earlier smolt cohort), and (b) mean weight of 1SW salmon against the mean weight of 2SW salmon returning in the following year (i.e. from the same smolt cohort). For stocks with both wild (W) and hatchery (H) stock components the data are plotted separately. The p‐values provide the statistical signifi‐ cance of the slope of each linear regression model. [N.B. the plots are presented by river name in alphabetical order.]

(a)

Argard (p= 0.05 ) Bush (p= 0.82 ) De la Trinite (p= 0 ) 1.4 1.8 2.2 1.9 2.2 2.5 1.0 1.2 1.4 Mean Weight 1SW Mean Weight 1SW Mean Weight 1SW Mean Weight

2.5 3.0 3.5 3.0 3.5 4.0 4.5 4.0 4.5 5.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Dee (p= 0 ) Ellidaar (p= 0.12 ) Enningdalselva (p= 0.31 ) 2.6 3.0 3.4 1.4 1.8 2.2 Mean Weight 1SW Mean Weight 1SW Mean Weight 1.9 2.2 2.5 1SW Mean Weight 4.5 5.5 6.5 4.0 4.5 5.0 5.5 4.04.55.05.5

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Etneelva (p= 0.13 ) Gaula (p= 0.1 ) Gaula Sogn (p= 0 ) 1.6 2.0 2.4 Mean Weight 1SW Mean Weight 1.41.61.82.0 1SW Mean Weight 1SW Mean Weight 1.2 1.5 1.8

3.5 4.0 4.5 5.0 5.5 5.0 5.5 6.0 3.5 4.0 4.5 5.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

ICES SGBICEPS REPORT 2010 | 91

Hofsa (p= 0.68 ) LaHaveW (p= 0 ) LaHaveH (p= 0 ) 1.6 2.0 1.5 1.8 2.1 Mean Weight 1SW Weight Mean 1.8 2.1 2.4 1SW Weight Mean 1SW Weight Mean

4.5 5.5 6.5 3.5 4.0 4.5 5.0 4.0 4.5 5.0 5.5

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Laxa i Adaldalur (p= 0.3 ) Lune (p= 0.49 ) Miramichi (p= 0.55 ) 1.5 1.7 2.3 2.6 2.9 Mean Weight 1SW Mean Weight 1SW Mean Weight 1SW Mean Weight 2.4 2.8 3.2 5.2 5.6 6.0 6.4 4.5 5.5 6.5 7.5 4.0 4.2 4.4 4.6 4.8

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Naatamojoki (p= 0.03 ) Nausta (p= 0.8 ) Nordura (p= 0.02 ) Mean Weight 1SW Mean Weight 1SW Mean Weight 1.1 1.3 1.5 1.7 1SW Mean Weight 2.0 2.2 2.4 1.4 1.8 3.8 4.2 4.6 5.0 3.8 4.0 4.2 4.4 4.6 4.2 4.6 5.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

92 | ICES SGBICEPS REPORT 2010

North Esk (p= 0 ) Numedalslagen (p= 0 ) Penobscot (p= 0.14 ) 1.5 2.0 2.5 1.6 2.0 2.4 2.8 1.6 2.0 2.4 Mean Weight 1SW Mean Weight 1SW Mean Weight 1SW Mean Weight

4.0 4.5 5.0 5.5 4.5 5.0 5.5 6.0 4.0 5.0 6.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Saint-Jean (p= 0.04 ) Saint John W (p= 0.75 ) Saint John H (p= 0.31 ) 2.0 2.4 2.8 2.0 2.6 3.2 1.50 1.65 Mean Weight 1SW Weight Mean 1SW Weight Mean 1SW Weight Mean 4.0 4.5 5.0 5.5 4.8 5.2 5.6 4.8 5.2 5.6 6.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Skienselva (p= 0.98 ) Teno (p= 0.35 ) Tuloma (p= 0.49 ) 1.3 1.5 1.7 Mean Weight 1SW Weight Mean 1SW Weight Mean 1SW Weight Mean 1.5 1.7 1.9 1.4 1.6 1.8 2.0 3.2 3.6 4.0 4.4 3.0 3.5 4.0 4.5 5.0 5.5 2.5 3.5 4.5

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

ICES SGBICEPS REPORT 2010 | 93

Ume_VindelH (p= 0.19 ) Vestre Jakobselv (p= 0.97 ) Western Arm Brook (p= 0.87 1.6 2.0 1.6 1.8 2.0 Mean Weight 1SW Weight Mean 1.2 1.6 2.0 1SW Weight Mean 1SW Weight Mean

4.0 4.5 5.0 5.5 6.0 4.0 5.0 6.0 3.4 3.6 3.8 4.0

Mean Weight 2SW Mean Weight 2SW Mean Weight 2SW

Wye (p= 0.7 ) Mean Weight 1SW Weight Mean 2.0 2.6 3.2 4.65.05.45.8

Mean Weight 2SW

94 | ICES SGBICEPS REPORT 2010

(b)

Argard (p= 0.74 ) Bush (p= 0.52 ) De la Trinite (p= 0 ) 1.4 1.8 2.2 1.9 2.2 2.5 1.0 1.2 1.4 Mean 1SW Weight Mean 1SW Weight Mean 1SW Weight

2.5 3.0 3.5 3.0 3.5 4.0 4.5 4.0 4.5 5.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Dee (p= 0 ) Ellidaar (p= 0.08 ) Enningdalselva (p= 0.03 ) 2.6 3.0 3.4 1.4 1.8 2.2 Mean Weight 1SW Weight Mean 1SW Weight Mean 1.9 2.2 2.5 1SW Weight Mean 4.5 5.5 6.5 4.0 4.5 5.0 5.5 4.04.55.05.5

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Etneelva (p= 0.06 ) Gaula (p= 0 ) Gaula Sogn (p= 0 ) 1.6 2.0 2.4 Mean Weight 1SW Weight Mean 1.4 1.6 1.8 2.0 1SW Weight Mean 1SW Weight Mean 1.2 1.5 1.8

3.5 4.0 4.5 5.0 5.5 5.0 5.5 6.0 3.5 4.0 4.5 5.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

ICES SGBICEPS REPORT 2010 | 95

Hofsa (p= 0.08 ) LaHaveW (p= 0 ) LaHaveH (p= 0.15 ) 1.6 2.0 1.5 1.8 2.1 Mean Weight 1SW Mean Weight 1.8 2.1 2.4 1SW Mean Weight 1SW Mean Weight

4.5 5.5 6.5 3.54.04.55.0 4.0 4.5 5.0 5.5

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Laxa i Adaldalur (p= 0.11 ) Lune (p= 0.55 ) Miramichi (p= 0.36 ) 1.5 1.7 2.3 2.6 2.9 Mean Weight 1SW Weight Mean 1SW Weight Mean 1SW Weight Mean 2.4 2.8 3.2 5.2 5.6 6.0 6.4 4.5 5.5 6.5 7.5 4.0 4.2 4.4 4.6 4.8

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Naatamojoki (p= 0 ) Nausta (p= 0.2 ) Nordura (p= 0 ) Mean Weight 1SW Mean Weight 1SW Mean Weight 1.1 1.3 1.5 1.7 1SW Mean Weight 2.0 2.2 2.4 1.4 1.8 3.8 4.2 4.6 5.0 3.8 4.0 4.2 4.4 4.6 4.2 4.6 5.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

96 | ICES SGBICEPS REPORT 2010

North Esk (p= 0 ) Numedalslagen (p= 0.22 ) Penobscot (p= 0.93 ) 1.5 2.0 2.5 1.6 2.0 2.4 2.8 1.6 2.0 2.4 Mean Weight 1SW Weight Mean 1SW Weight Mean 1SW Weight Mean

4.0 4.5 5.0 5.5 4.5 5.0 5.5 6.0 4.0 5.0 6.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Saint-Jean (p= 0.45 ) Saint John W (p= 0.04 ) Saint John H (p= 0.58 ) 2.0 2.4 2.8 2.0 2.6 3.2 1.50 1.65 Mean Weight 1SW Mean Weight 1SW Mean Weight 1SW Mean Weight 4.0 4.5 5.0 5.5 4.8 5.2 5.6 4.8 5.2 5.6 6.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Skienselva (p= 0.45 ) Teno (p= 0.15 ) Tuloma (p= 0.26 ) 1.3 1.5 1.7 Mean Weight 1SW Mean Weight 1SW Mean Weight 1SW Mean Weight 1.5 1.7 1.9 1.41.61.82.0 3.2 3.6 4.0 4.4 3.0 3.5 4.0 4.5 5.0 5.5 2.5 3.5 4.5

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

ICES SGBICEPS REPORT 2010 | 97

Ume_VindelH (p= 0 ) Vestre Jakobselv (p= 0.07 ) Western Arm Brook (p= 0.99 1.6 2.0 1.6 1.8 2.0 Mean Weight 1SW Weight Mean 1.2 1.6 2.0 1SW Weight Mean 1SW Weight Mean

4.04.55.05.56.0 4.0 5.0 6.0 3.4 3.6 3.8 4.0

Mean Weight 2SW (Next) Mean Weight 2SW (Next) Mean Weight 2SW (Next)

Wye (p= 0.83 ) Mean Weight 1SW Mean Weight 2.0 2.6 3.2 4.6 5.0 5.4 5.8

Mean Weight 2SW (Next)

98 | ICES SGBICEPS REPORT 2010

Table 5.5.1. Results of analysis of two‐way relationships based on simple linear regression mod‐ els: ‘o’ means that there was not enough evidence at the 5% level to detect a trend; ‘+’ denotes a positive trend (p>0.05), ‘‐‘ denotes a negative trend (p<0.05).

v v v v

v v

n+1) n) n) n) n+1) n) n) n) Sea 1SW

2SW

2SW Total 1SW Total v v

v

(yr (yr (yr (yr v v v v (yr (yr

(yr (yr

complex age age

age series Length CF Wt Length CF Wt CF CF wt wt wt wt wt wt wt wt

River age 1SW 2SW 1SW 2SW 1SW 2SW River 2SW Stock Wt River 1SW 2SW 1SW 2SW Country Stock H/W Time Latitude 1SW Run Run 2SW Run 2SW Run 1SW Wt 1SW NAC (N) Canada Western Arm Brook W 1984‐06 51.2 oooo‐ +o‐ o++oo Canada De la Trinité W 1980‐09 49.4 ++oo ++ ‐‐‐‐ Canada Saint‐Jean W 1981‐09 48.8 +o+o ++ ‐ ooo Canada Middle Brook W 1984‐05 48.8 ++ o oo Canada Conne River W 1984‐06 47.9 o++o oo Canada Miramichi W 1984‐07 47.0 ooo+ o++oooooo NAC (S) Canada Nashwaak W 1984‐07 46.0 + Canada St John (Mactaquac) W 1984‐07 45.3 o+ o+ o++oooooo Canada St John (Mactaquac) H 1984‐07 45.3 ooooo++ooooo+ Canada La Have W 1984‐07 44.4 ++++o++++oo‐‐ Canada La Have H 1984‐07 44.4 +o++‐ ++oo‐‐oo USA Penobscot H 1984‐07 44.5 ooooooooooooo N NEAC Finland/Norway Teno W 1984‐07 70.8 oo+o‐ ++ooo+oo Finland/Norway Näätämöjoki W 1984‐06 69.7 ++ooo++oo+oo+ Russia Tuloma W 1984‐08 68.9 ooo+ o++oooooo Norway Vestre Jakobselv W 1989‐08 70.1 ooo+o++o+oo‐ o Norway Årgårdsvassdraget W 1992‐08 64.3 + oooo++ooooo+ Norway Nausta W 1989‐08 61.3 ooooo++o+++o‐ Norway Gaula Sogn og Fjorda W 1989‐08 61.2 +++oo++ooo+oo Norway Etneelva W 1989‐08 59.4 ooooo++oooooo Norway Skienselva W 1989‐08 59.1 oo+ oo++oooooo Norway Numedalslågen W 1989‐08 59.1 + o+ oo++oooooo Norway Enningdalselva W 1990‐08 58.6 o+ o+ o++oooooo Norway Gaula W 1989‐07 63.3 o+ ooo++oooooo Iceland (N&E) Laxa I Adaldalur W 1984‐07 65.6 oo + ‐ o Iceland (N&E) Hofsa W 1984‐07 65.4 oo o o+ S NEAC Iceland (S&W) Nordura W 1984‐07 64.6 ++ o +o Iceland (S&W) Ellidaar W 1984‐07 64.1 oo + ‐ o UK (Scot) N. Esk W 1984‐07 56.7 + + + + ++ oooo UK (NI) Bush W 1984‐07 55.1 ooooo++oooooo UK (E&W) Lune W 1987‐07 54.0 oooooo+oooo‐ o UK (E&W) Dee W 1984‐07 53.4 ++oo+++‐‐oo‐‐ UK (E&W) Wye W 1984‐07 51.6 oo+o+++‐ ooooo UK (E&W) Frome W 1984‐08 50.7 France Bresle W 1984‐08 50.1 Baltic Sweden Kalix W 1980‐08 65.8 Finland/Sweden Tornionjoki W 1980‐09 65.5 Sweden Ume/Vindel W 1987‐08 63.8 Sweden Ume/Vindel H 1974‐08 63.8 o+ Summary ‐ all areas No. of stocks for which data available 31 31 26 26 30 28 27 30 28 28 28 26 26 % of stocks with significant declining trend 0 0 0 0100 01741171915 % of stocks with significant increasing trend 35 35 35 31 20 93 93 7 14 11 14 0 12 % of stocks with no significant trend 65 65 65 69 70 7 7 77 82 79 79 81 73

The size and condition of 1SW and 2SW fish returning in the same year It is evident that for a number of stocks (35%) there is a significant positive relation‐ ship between the weight of 1SW and 2SW salmon returning in the same year (i.e. de‐ rived from different smolt cohorts) (Table 5.5.1). In no instances are significant negative trends identified, and the majority of the plots in Figure 5.5.1(a) suggest positive trends even if the relationships are not significant. Significant positive rela‐ tionships are evident in all areas of the North Atlantic. A very similar pattern is evi‐ dent for mean condition factor. These common patterns of larger (or smaller) 1SW fish coinciding with larger (or smaller) 2SW fish in returning stocks in any year are consistent with common factors operating on the fish from the two sea‐group groups during their return migration. However, it should be noted that there was also evi‐ dence of a correlation between size and condition of 1SW fish in year ‘n’ and 2SW fish returning in the following year (‘n+1’) (Figure 5.5.1.b) suggesting common influences

ICES SGBICEPS REPORT 2010 | 99

in the early part of the marine phase (see below). The Study Group recognised that further analysis of such data would be valuable to explore temporal and spatial pat‐ terns.

The size and condition of 1SW in year ‘n’ and 2SW fish returning in the following year (‘n+1’) Similar positive relationships are also evident for many stocks (35%) between the weight of 1SW returning in one year and the weight of 2SW salmon returning in the following year (i.e. derived from the same smolt cohort). As with fish returning in the same year, there is no evidence of any negative relationships (Table 5.5.1 and Figure 5.5.1.b), and similar patterns are evident for mean condition factor. Significant posi‐ tive relationships are evident in all areas of the North Atlantic and for one hatchery stock in the Baltic. These concurrent patterns for particular smolt cohorts suggest that common factors operating in the first period at sea may have a large influence on growth and size at maturity of returning adult fish. Again, the Study Group recog‐ nised that further analysis of such data would be valuable. For some stocks, significant positive relationships are apparent between the weight/condition factor of returning 1SW salmon and the weight/condition factor of returning 2SW salmon in both the same and the following year. For other stocks, sig‐ nificant relationships were only evident for one of these two‐way comparisons. There is no apparent spatial pattern to these differences.

Mean river age v mean sea-age Analyses indicate very different relationships between river age and sea‐age for dif‐ ferent stocks (Table 5.5.1). In some rivers, significant positive relationships apply (i.e. older smolts tend to produce older adults). However, significant negative relation‐ ships are evident for some other stocks. Viewed spatially, there appears to be more evidence of variability in these relationships in the NAC and Northern NEAC areas, where both significant positive and negative relationships are evident. Only signifi‐ cant positive relationships (n = 3, 50% of the available data sets) are evident in the Southern NEAC area.

Length v Weight Unsurprisingly, the length and weight of returning 1SW and 2SW fish are positively correlated. In most, but not all, instances these relationships are significant (Table 5.5.1).

Mean river age v size of returning adults The age of smolts (ages determined from adult scales) and the size of returning 1SW and 2SW adults was compared to see whether there were any consistent patterns be‐ tween the age, and by inference the size of smolts, and the subsequent size (weight) of the returning adults. A number of significant relationships were apparent (Table 5.5.1). For 17% of the stocks (n = 5) there was a significant negative relationship be‐ tween river age and 1SW weight – i.e. older/larger smolts tended to result in smaller returning adults, while for 7% of stocks (n = 2) the opposite was true. The former is consistent with other evidence that smolt size can influence the subsequent growth rate of salmon, with larger smolts showing slower growth at sea (Jonsson & Jonsson, 2007). However, when comparing river age with the weight of returning 2SW sal‐ mon, more stocks (14%) had significant positive than negative relationships (4%), indicating that older/ larger smolts tended to result in larger returning adults. Such

100 | ICES SGBICEPS REPORT 2010

observations would appear to be at odds with the hypothesis that larger smolts have slower growth at sea, although the impact of smolt age/size on subsequent matura‐ tion (1SW or 2SW) may also be a factor.

Size of returning fish v stock status variable Provisional analyses were also completed to investigate possible links between the size (weight) of returning adult 1SW and 2SW fish in any year and the stock‐specific status variable for that year (Table 5.5.1). For these purposes, two‐way relationships were explored between the mean size of returning fish and both the total stock status (i.e. all sea‐age groups) and the age‐specific stock status (i.e. 1SW or 2SW), simply derived from the total returning stock figure and the estimated proportion of each sea‐class in the stock. The latter is expected to be more informative. The Study Group recognised that these analyses should be treated with some caution given the con‐ cerns noted previously (Section 3.3), perhaps in particular with regard to the robust‐ ness of the stock status variable. Considering only the comparisons between fish size and the age‐specific stock status variables, a small number of significant relationships are apparent (Table 5.5.1); these are both positive (i.e. fish of larger mean size associated with better stock status) and negative (i.e. fish of smaller mean size associated with better stock status). There is no apparent spatial pattern to these differences. However, the relationship between mean length and stock size is explored further, at the stock complex rather than indi‐ vidual stock level, in Section 5.10.

5.6 Long-term variation and changes in age at smoltification in four major Scottish salmon rivers Several recent publications have highlighted recent changes in age at smoltification (so‐called ‘river age’), as well as run‐timing and sea‐age composition of returning adult salmon populations in the British Isles (e.g. Cragg‐Hine et al., 2006; Quinn et al., 2006; Aprahamian et al., 2008). One working hypothesis is that a general warming of terrestrial/freshwater climate in recent decades might explain increases in length of the growing season, and hence smoltification of juveniles at a younger river age (e.g. Aprahamian et al., 2008). Aprahamian et al. (2008) showed that mean river tempera‐ tures of the Welsh Dee rose by ~0.02°C y‐1 between 1965 and 1999, and this may un‐ derlie an increased juvenile growth rate and observed switch to an earlier average age at smolting. A clear pattern of a marked and persistent increase in the proportion of S1 fish (i.e. fish that migrated at river age one) – and a concomitant fall in the proportion of S2 fish (Figure 5.6.1) – can be seen for the Welsh Dee from 1980 onwards (as determined from scale analysis in returning adults). However, Cragg‐Hine et al. (2006) reported that the Dee has become colder during the April‐October period since the mid‐1970s as a result of flow regulation. It remains rather unclear, therefore, whether or not freshwater climate warming alone accounts for this obvious multidecadal shift from S2 to S1 smolts in the Welsh Dee. Undoubtedly, it is important to obtain a wider geographic perspective on this aspect of salmon biology, and putative responses to a changing climate in both the freshwater and marine environments.

ICES SGBICEPS REPORT 2010 | 101

WELSH DEE 1SW 2SW Mean river age (1967-2008; Weeks 1-52) 2.40

2.20

2.00

1.80

1.60

1.40 Mean river age (years) river Mean 1SW adults 1.20 2SW adults

1.00 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Adult return year

Figure 5.6.1. River Dee, Wales. Mean river age derived from all available adult 1SW and 2SW salmon sampled between 1967 and 2008. The data for 1967–1989 relate to fish sampled in the es‐ tuarine net fishery (February‐September), whilst from 1991 onwards fish were monitored year‐ round at the Chester Weir trap (located close to the head of tide). Trap data indicate that ~80% of the returning adult runs were sampled by the nets.

The Marine Scotland time‐series of salmon age data for the rivers Tweed, Tay, Spey and North Esk are multidecadal in extent and comprise large sample sizes. Moreover, these data have been collated from commercial net fisheries in a consistent and com‐ parable manner throughout, and a detailed analysis was therefore considered expe‐ dient. These four Scottish east coast rivers comprise the focus for the present report, but comparative analyses are being undertaken which will include the Welsh Dee, Severn and Wye. The River North Esk is the most comprehensively monitored salmon river in Scot‐ land. Since 1970, downstream‐migrating smolts have been sampled during the spring smolt run by means of a purpose‐built trap in the diversionary lade at Kinnaber Mill, just upstream of the head of tide. Further details are provided in Section 4.9. In addi‐ tion, complementary data (1964–1969) for the age‐structure of the emigrant smolt population were obtained by sampling smolts caught in the water filter at the water‐ works situated on Kinnaber lade. Summary data are presented in Section 4.9, but de‐ tailed analyses are confined to the period 1975–2008 because of concerns over the comparability of data obtained prior to 1975. Since 1963, returning adults (1SW and MSW) have been routinely sampled on the North Esk throughout the season (16 February–31 August) from the commercial net and coble fishery in the river estuary. Between 2000 and 2004 the commencement of netting was delayed voluntarily by netsmen until 1 April for conservation reasons. Since 2005 the opening has been further delayed until 1 May. MSW salmon in the River North Esk are primarily 2SW adults. Historically there have been catch records for 3SW and 4SW fish but these are relatively scarce. Over the period 1963–1999, only 4.4% of the fish sampled were 3SW or 4SW return adults. For the present purposes, therefore, analyses have been confined to the 1SW and 2SW maturity groupings; to‐ gether these comprise approximately 95% of the overall historical catch. The 2SW fish sampled from the North Esk nets (Figure 5.6.2) show a significant re‐ duction in mean river age over the period 1963–2008. However, the summary statis‐

102 | ICES SGBICEPS REPORT 2010

tic – mean river age (that is, age at emigration of the smolt) – hides much ecologically important variation, and even statistically significant patterns in the data may be an artefact of the sampling regime. Importantly, mean river age may show no trend over time and yet the age‐structure of the population could have changed radically. It could also be argued that river age should be considered a categoric variable, rather than a continuous variable (and therefore not averaged within years). Because smoltification is an annual “all or none” decision, if a juvenile parr fails to smolt in one spring then it remains in the river for one or more further full years. There is therefore no such thing as an emigrant smolt of, say, 1.68 years river age.

NORTH ESK 2SW Mean River Age (1963-2008; February- August )

2.7

2.6

2.5

2.4

2.3

Mean River Age 2.2

2.1

2 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.2. River North Esk, Scotland. Mean river age (± se) for all available returning adult 2SW salmon sampled between February and August, 1963–2008. The downward linear trend is significant (r = 0.521; P <0.001). [The data for 1982 and 1983 were derived from Shearer (1990), his summary Tables 2 and 3.]

Figure 5.6.3 indicates that the recent, relatively low mean river age (Figure 5.6.2) is attributable to increases in the proportions of 2SW adults that migrated as S2 and, to a lesser extent, S1 smolts, and is reflected as a decline in the S3 component. However, it is important to stress that these are proportional data – a decline in one river age proportion will be reflected by an increase in one or more other components and the data are therefore not independent. Given the ecological improbability of a short‐ term downshift from S3 to S1, the likelihood is that these trends simply reflect a ten‐ dency to shift down from S3 to S2. Similarly, the most cautious interpretation of the Welsh Dee data (Aprahamian et al., 2008; Figure 2) is that a greater proportion of ju‐ veniles that previously would have smolted at S2 have tended to smolt at S1 since the 1980s. In the knowledge that the delays in the commencement of the netting season on the River North Esk since 1999 have led to a seasonal bias in the data time‐series, is the recent decline in mean river age of North Esk 2SW adults (Figures 5.6.2 and 5.6.3) real, or is it an artefact of the data?

ICES SGBICEPS REPORT 2010 | 103

North Esk 2SW ALL DATA (1963-99 February-August; 2000-04 April-August; 2005-08 May-August) 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 0.3

Proiportion River Age River Proiportion 0.2

0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.3. River North Esk, Scotland. Proportions of returning 2SW adults of varying river age (S1–S3; river ages 1 to 3 years, respectively). S4 and S5 fish are relatively rare in the data set and for clarity have not been plotted here. [The data for 1982 and 1983 were derived from Shearer (1990), his summary Tables 2 and 3.]

Figure 5.6.4 shows the fluctuations in mean river age for the 2SW fish sampled throughout the netting season between February and August over the period 1963– 1999. This time‐series therefore comprises seasonally comparable catches in every year. Despite the among‐year fluctuations, there is an apparent overall downward trend, but this is not statistically significant for a linear decrease (P >0.05). Figure 5.6.5 includes the separation of these 2SW adult data into their proportional river ages. Despite the inevitable between‐year variations – and although there is some suggestion of an increase in S1 and S2, in concert with a relative decrease in the S3 component – it is apparent that there is overall stability in the river age‐structure of the February–August 2SW samples over the available time‐series.

104 | ICES SGBICEPS REPORT 2010

NORTH ESK 2SW (1963-1999; February-August inclusive)

2.7

2.6

2.5

2.4

2.3

Mean river age 2.2

2.1

2 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.4 River North Esk, Scotland. Mean river age (± se) for all 2SW fish sampled annually February‐August inclusive; 1963‐1999. Sampling commenced on 1 April between 2000 and 2004, and on 1 May from 2005 onwards and hence these years are not included. [Data for 1982 and 1983 were derived from Shearer (1990), his summary Tables 2 and 3.]

NORTH ESK 2SW 1963-1999 (February-August; proportions of river ages, S1-S3) 0.9

0.8

0.7 0.6 S1 0.5 S2 0.4 S3

Proportion 0.3 0.2 0.1

0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.5. River North Esk, Scotland. Proportions of returning 2SW adults of varying river age (S1‐S3) over the period 1963–1999, when netting was monitored between February and August each year. Fish of S4 and S5 river age were scarce and have not been plotted. [Data for 1982 and 1983 were derived from Shearer (1990), his summary Tables 2 and 3.]

The indications from Figure 5.6.5 are that, for 2SW salmon, the S1 component may have increased linearly over time (r = 0.359; P <0.05; though the proportions typically are very low and the test therefore weak); but there is no statistically significant change either in S2 or S3 over that period. The marked reductions in overall river age apparent for the most recent decade (Figure 5.6.4) may, therefore, be primarily ac‐

ICES SGBICEPS REPORT 2010 | 105

countable as an artefact in the data because the start of sampling has been seasonally delayed since 2000. Figures 5.6.2 to 5.6.5 include both early‐running “spring” salmon and summer‐ running salmon. Notwithstanding our inability to categorically distinguish “spring” and “summer” salmon, pooling all the available data (February‐August) may there‐ fore occlude differences or contrasts in performance of the early and late‐running cohorts. The river age composition of early‐running 2SW spring salmon (Figure 5.6.6) typi‐ cally comprises adults of somewhat higher river ages compared to the later‐running 2SW salmon (included in Figure 5.6.3), with proportionally more S3 fish and S1 barely represented among the early‐running (February/March) 2SW adults. (N.B. one cannot infer that S2 are scarcer amongst these early‐running fish compared to the complete data in Figure 5.6.2, because a decrease in one proportion will lead to an increase in one or more others. The likelihood is that the critical difference here is a higher relative abundance of S3 among the early‐running spring salmon.) In inter‐ preting Figure 5.6.5 it is important, therefore, to ascertain the proportions of early (e.g. February/March) fish sampled in any one year relative to the later summer fish. As might be expected, the relative sample sizes of 2SW salmon taken in February‐ March versus April onwards has fluctuated considerably over the 1963–1999 time‐ series (Figure 5.6.7). The samples of early‐running 2SW spring salmon were consis‐ tently relatively large throughout the late 1960s and 1970s and this further com‐ pounds the difficulties in interpreting the long‐term shifts in river ages for 2SW fish. For these particular samples, the relative scarcity of early‐running 2SW salmon fol‐ lowing the mid‐1980s will result in a bias in the data towards rather younger river ages. Resolution of this problem will require more detailed modelling, but the likeli‐ hood is that over the period 1963–1999 there has been stability in the river age‐ structure of returnimg adult 2SW salmon (both in spring and in summer) for the river North Esk. The lack of data for the early‐running 2SW fish from 2000 onwards is re‐ grettable and precludes further analysis of changes apparent for the most recent years.

106 | ICES SGBICEPS REPORT 2010

North Esk 2SW (February + March only; 1963-99)

0.9 0.8 0.7 0.6 0.5 0.4

0.3 S1

Proportion River Age River Proportion 0.2 S2 S3 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.6 River North Esk, Scotland. Proportions of S1, S2 and S3 among 2SW “spring” salmon captured in February and March only [data from 1982 and 1983 are lacking in Marine Scotland records and are not derivable for February and March from the summary tabulations in Shearer (1990)]. The linear trend lines for S2 and S3 are shown, but these are not statistically significant (P >0.05).

NORTH ESK 2SW adults sampled February-August inclusive, 1963-1999

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Proportion of Feb+March fish in sample fish Feb+March of Proportion 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.6.7. River North Esk, Scotland. Proportion of annual samples of 2SW salmon that were taken in February/ March versus February‐August inclusive [data from 1982 and 1983 are lacking in Marine Scotland records and are not derivable for February and March from the summary tabulations in Shearer (1990)].

There is currently no objective and reliable means of separating early‐running “spring” salmon from “summer” salmon, or even from late‐running “autumn” MSW fish. In the knowledge that the early‐running 2SW fish tend to be of a higher overall river age (Figure 5.6.6) than those returning later in the season, and that these stock components may well respond differently to any climate‐related (or other environ‐ mental) drivers compared to later‐running 2SW salmon, it would appear inappropri‐ ate to further interrogate the River North Esk MSW time‐series for evidence of

ICES SGBICEPS REPORT 2010 | 107

climate‐related shifts in river age‐structure for MSW salmon. Certainly, the data for the early‐running stock component give no indication of any marked, persistent or statistically significant trend between 1963 and 1999 – either for mean river age of returning adults (Figure 5.6.4), or for the proportional representation of the various river ages (Figure 5.6.5). By contrast, the 1SW stock component is clearly identifiable. The timing of the netting season ensures that in every year the beginning of the 1SW return migration is effectively sampled, though the data are truncated by the close of the netting season before all 1SW fish have returned. Notwithstanding the latter qualification, the possibilities of detecting, with confidence, any evidence of demo‐ graphic shifts in time‐series for the 1SW component are clear and are discussed in Section 5.8.

5.7 Among-river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 2SW “summer” salmon (returning May-August/September) Time‐series data (from in‐river net fisheries) are available for the rivers North Esk, Tweed, Tay and Spey, with some extending back to the early 1960s. Historically, all these fisheries opened in February (Tay 6th, Spey 11th, Tweed 15th and North Esk 16th) though the opening has been voluntarily delayed by netsmen in recent years as a conservation measure. The close season either falls in August (Tay 20th, Spey 26th, North Esk 31st) or in September (Tweed 14th). On the strength of the foregoing analy‐ ses, and in recognition of the distinction between early‐running (“spring”) and late‐ running (“summer”) 2SW salmon, the following comparative analyses for 2SW fish are confined to years when sample data were available from early May onwards. All available data after the beginning of May within years have been retained and ana‐ lyzed so, for example, the Tweed analyses include September fish caught after the remaining three river fisheries had closed. Years in which sampling ceased >10 days prior to the seasonal close were deleted. The time‐series for the rivers Spey and Tay terminated in 1993 and 1996 respectively when the net fisheries were closed for con‐ servation reasons. The available time‐series for the rivers Tay, Tweed and Spey are not as complete as for River North Esk (Figure 5.7.1). Samples for the Tweed are available up to 2008, but the variations in the commencement and cessation dates of sampling within the netting season resulted in recent years of incomplete and non‐comparable data. The conservative approach was adopted throughout here in deleting from analysis all such years.

108 | ICES SGBICEPS REPORT 2010

TWEED 2SW Mean River Age (May onwards)

2.5

2.3

2.1

1.9

Mean River Age 1.7

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

TAY 2SW Mean river age (May onwards)

2.5

2.3

2.1

1.9

Mean River Age 1.7

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

NORTH ESK 2SW Mean river age (May onwards)

2.5

2.3

2.1

1.9

Mean River Age 1.7

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

SPEY 2SW Mean river age (May onwards)

2.5

2.3

2.1

1.9

1.7 Mean RiverAge

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.7.1. Rivers Tweed, Tay, Spey and North Esk, Scotland. Mean river age (at smoltification) (± se) for returning 2SW “summer” salmon sampled between May and the end of the netting sea‐ son (Spey, Tay, North Esk – August; Tweed – September).

As a generalization, the 2SW summer fish in all four rivers are dominated either by S2 or by S2 and S3 river ages. The Tweed, however, apparently differs fundamen‐ tally from the remaining three rivers in typically comprising an overall younger age‐ structure, although this varies considerably from year to year (see further discussion below).

ICES SGBICEPS REPORT 2010 | 109

Table 5.7.1. Correlation matrix for mean river age for 2SW “summer” salmon sampled on the four Scottish rivers (from data in Figure 5.7.1). Correlations in bold type are significant at the P = 0.05 level. The correlation in bold italics remains significant following Bonferroni Correction.

______SPEY r 0.186 P 0.490 TAY r 0.558 0.688 P 0.025 0.005 NORTH ESK r 0.221 0.521 0.329 P 0.349 0.027 0.182

TWEED SPEY TAY ______Analysis of the mean river ages illustrated in Figure 5.7.1 shows three significant cor‐ relations (Table 5.7.1), but only one (Spey–Tay) remains significant following Bon‐ ferroni Correction for multiple comparisons (critical value P = 0.008). The overall outcome here is that there is no strong evidence that these four Scottish rivers are re‐ sponding similarly – in terms of long‐term variation in mean river age of 2SW “summer” salmon – over the period from the late 1960s to the late 1990s. But as dis‐ cussed above, mean river age as a summary statistic occludes a considerable amount of ecologically relevant variation in the detailed age‐structure of return cohorts or year‐classes. Pair‐wise comparisons of, for example, the correlations between S2 and S3 fish within years are confounded by the lack of independence of proportional data, so in order to progress this a hierarchical analytical approach was adopted within a framework of generalized additive models (GAMs). River age for these four catchments was analyzed with a general model structure of: Model<‐gam Proportion river age~ s(Other proportional river ages) s(Year) s(Year, by=River) River Family=quasibinomial Gamma=1.4 Year and River (Tweed, Tay, North Esk, Spey) were treated as factors, and the quasi‐ binomial error structure was used due to the limitations of the response variable: the Proportion of sample population in the age‐class is a decimal that can range only be‐ tween 0 and 1. This error structure also controls for the over‐dispersion in the pro‐ portion data. Gamma was increased to 1.4 to control for over‐fitting within the model. Models were simplified to find the most parsimonious structure, with the validity and utility of the model being determined by the General Cross Validation (GCV) procedure. Model Forms: 1 ) ProportionS1 vs ProportionS2 + ProportionS3 + ProportionS4 2 ) ProportionS2 vs ProportionS3 + ProportionS4

110 | ICES SGBICEPS REPORT 2010

3 ) ProportionS3 vs ProportionS4 A separate model was used to test the relationship between each river age proportion and those above it (i.e. older). In each model, the Proportion of fish in the river age‐ class was compared with the (grouping of) older age class(es). The factors Year and River, and the Year*River interaction, were modelled as smooth functions. In the summary Table 5.7.2, the effects of River are assessed by comparing the River North Esk to the remaining three rivers. River North Esk is the comparator simply because it is first in the alphabetical list.

Table 5.7.2. Rivers Tweed, Tay, North Esk, Spey, Scotland. Summary of significant F ratios and probabilities from generalized additive model analysis of the variation in proportions of river age classes over the period 1963–2008 (data in Figure 5.7.2) ns = not significant. ______

GAM outputs for 2SW return adults (River comparisons Tweed, Spey, Tay against North Esk) S1 vs S2+S3+S4 Year River River*Year F P 2042 <<0.001 ns ns ns

S2 vs S3+S4 F P F P F P 82.46 <<0.001 2.49 0.024 ns 19.78<<0.001 (Tweed)

S3 vs S4 F P F P F P F P 9.95 0.002 4.10 0.001 ‐3.56 0.001 4.68<0.001 (Tweed) (Tweed) ‐2.79 0.006 (Spey) ______

ICES SGBICEPS REPORT 2010 | 111

TWEED 2SW proportional river ages (May-September fish) 1 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

TAY 2SW proportional river ages (May-August fish)

1 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

NORTH ESK 2SW proportional river ages (May-August fish)

1 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

SPEY 2SW proportional river ages (May-August fish)

1 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.7.2. Rivers Tweed, Tay, North Esk and Spey, Scotland. Proportions of “summer” 2SW adults of varying river age (S1‐S3) sampled between May and August/September each year. S4 fish were rare and have not been included.

112 | ICES SGBICEPS REPORT 2010

Figure 5.7.3. Rivers Tweed, Tay, North Esk, Spey, Scotland. Proportions of river ages S1‐S4 amongst return adult 2SW “summer” salmon.

Figure 5.7.3 indicates the marked contrasts in river age composition of “summer” 2SW salmon among the group of four rivers. Whilst the rivers Tay, Spey and North Esk show considerable similarity in overall age‐structure and temporal variation, the Tweed is clearly distinct: the major between‐river contrasts lie in the relative propor‐ tions of S1 and S3 fish. For the proportions of all the separate river ages (S1, S2, S3) there were consistent and significant effects of the proportions of the older river age(s) pooled. Indeed, for S1 fish this was the only significant outcome (Table 5.7.2), and the data for the River Tweed are especially notable in their variation over time. Given the scarcity of S4 fish, the comparison S2 vs S3+S4 essentially is a contrast between S2 and S3. The pro‐ portion of S2 vs S3+S4 showed a significant Year effect and, for the River Tweed, a significant interaction with Year. For S3, there was a significant Year effect (the pro‐ portion of S3 fish increased over time relative to the proportion of S4 [NOTE: for clar‐ ity, S4 data are not plotted in Figure 5.7.2]) and the Rivers Tweed and Spey showed declines in proportions of S3 (relative to S4) in comparison to the River North Esk. As a generalization, for the Tweed the demographic changes over time have largely been attributable to downward shifts from S2 to S1 smolts, whereas for the River North Esk the shift over the past decade has been from S3 to S2. The lack of compara‐ ble data for the other three rivers over the most recent decade does, however, restrict this observation to that one river. Whether or not this change is typical of Scottish rivers remains unclear. The higher prevalence of S1 smolts in the river Tweed pre‐ sumably is a reflection of this being essentially a lowland catchment with relatively

ICES SGBICEPS REPORT 2010 | 113

high productivity. By contrast, the catchments of the rivers Tay, North Esk and Spey include large proportions of highland terrain of relatively low productivity and typi‐ cally are characterised by colder waters. There presumably are severe ecological con‐ straints on highland parr growing sufficiently rapidly to smoltify only one year after hatching, and this is reflected in their comparative scarcity relative to the River Tweed. However, a downward shift from S3 to S2 in response perhaps to warmer river temperatures, increased growing seasons, or increased prey availability is per‐ haps much more possible for the rivers Tay, Spey and North Esk. For the Tweed, increases in the proportion of S1 smolts are very likely in response to more benign environmental conditions and it is noteworthy that the S1 component there can show such a wide range of fluctuations even over relatively short periods of time (e.g. 0% in 1988 and 69% in 1993; Figure 5.7.3). The overall outcome for “summer” 2SW salmon is that for the rivers Tay, Spey and North Esk over the period from the late 1960s to the early 1990s there are broadly similar river age‐structures that are relatively stable. Notwithstanding the lack of data for the most recent years, the indications are that these three rivers have shown a close concordance in their age‐class structure and its stability over time. However, the Tweed is quite distinct in showing a generally younger river age‐distribution and strongly variable annual dynamics. There is no evidence of concordance amongst these four rivers in shifts in mean age structure, or in the proportions of given river ages. The patterns of change in mean river age and in the proportionality of S1 and S2 do not match that reported by Apra‐ hamian et al. (2008) for the Welsh Dee (Figure 5.6.1). The short‐ and long‐term varia‐ tions in river age‐structure for these respective summer 2SW cohorts should therefore be considered in the context of a catchment‐specific perspective. The likelihood is that any short‐ or long‐term changes in river ages of return adult populations to these rivers are not driven by factors influential in the common (oceanic) environment, but by differences among river catchments and years which are attributable to temporal and geographic variation in the freshwater habitat.

5.8 Among-river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 1SW grilse (returning April–August/September) The adult 1SW stock component typically commences returning to these four Scottish rivers during April, May or June (with the Tweed consistently later than the remain‐ der); the start of the grilse run is comprehensively covered by all the available time‐ series. The netting season does, however, close before the completion of the grilse run and the available data are therefore inevitably curtailed. Comparisons of mean river ages, and proportional representation of the S1–S5 river ages, are therefore likely to be influenced by seasonal differences in migratory returns among rivers, and the possibility of different river ages returning at differing times of the year. Shearer (1990) reported that fish of younger river ages tend to return later in the calendar year, but he made no clear distinction in this regard between 1SW and MSW adults.

114 | ICES SGBICEPS REPORT 2010

TWEED 1SW Mean River Age (April-September)

2.5

2.3 )

2.1

1.9

1.7 Mean River Age (se Age River Mean

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

TAY 1SW Mean River Age (April-August)

2.5

2.3 )

2.1

1.9

1.7 Mean River Age (se Age River Mean

1.5

1.3 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

NORTH ESK 1SW River Age (April-August)

2.5

2.3 )

2.1

1.9

1.7 Mean River Age (se Age River Mean

1.5

1.3

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 SPEY 1SW River Age (April-August)

2.5

2.3 )

2.1

1.9

1.7 Mean River Age (se Age River Mean

1.5

1.3

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

Figure 5.8.1. Rivers Tweed, Tay, North Esk and Spey, Scotland. Mean river age (at smoltification) (± se) for adult 1SW salmon sampled between April and the end of the netting season (Spey, Tay, North Esk – August; Tweed – September).

The 1SW data for the four rivers (Figure 5.8.1) all include the beginning of the 1SW run, but the cessation of the netting season differs among rivers. The Tweed, for ex‐ ample, closes later, but the start of the grilse run is typically also later here than for the other three rivers. Over the complete 46‐year period, the mean river age of 1SW fish on the River North Esk has remained remarkably stable, and notably so over the past decade. Although variation was more marked in the earlier years, the overall

ICES SGBICEPS REPORT 2010 | 115

pattern there is of a mean river age of approximately 2.2 years. This is in marked con‐ trast to the Tweed, which shows an overall pattern of declining mean river age, albeit including a period of considerable among‐year fluctuations during the mid‐late 1980s (Figure 5.8.1). That period of high inter‐annual variation for the Tweed is certainly plausible if the “decision” to smolt at river age 1 (or to remain in‐river for a further year) is close, as a result of parr typically growing rapidly and well in their first one or two years: moreover, a high proportion of parr smolting at 1 year of age by defini‐ tion results in fewer available individuals to smolt at two years of age in the subse‐ quent year. Analyses for these four rivers (Table 5.8.1) reveal significant positive correlations in the changes in mean river age, but with only the Spey–North Esk comparison remain‐ ing significant following Bonferroni Correction. As before, however, consideration must be given to the detail of the age‐structures in order to ascertain whether or not these apparent positive correlations are real for particular river age groups.

Table 5.8.1. Rivers Tweed, Tay, North Esk and Spey, Scotland. Correlation matrix for mean river age of 1SW salmon sampled in the four Scottish rivers (see data in Figure 5.8.1). Correlations in bold face are significant at the P = 0.05 level. The correlation in bold italics remains significant following Bonferroni Correction.

______SPEY r 0.535 P 0.027 TAY r 0.503 0.625 P 0.028 0.010 NORTH ESK r 0.286 0.605 0.442 P 0.209 0.005 0.051 TWEED SPEY TAY ______

116 | ICES SGBICEPS REPORT 2010

TWEED 1SW proportional river ages (April-September)

1 0.9 0.8 0.7

0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

TAY 1SW river ages

1 0.9 0.8 0.7 0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 NORTH ESK 1SW river ages

1 0.9 0.8 0.7

0.6 S1 0.5 S2 0.4 S3 Proportion 0.3 0.2 0.1 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

SPEY 1SW river ages

1 0.9 0.8 0.7 0.6 S1 S2 0.5 S3 0.4 Proportion 0.3 0.2 0.1 0

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005

Figure 5.8.2. Rivers Tweed, Tay, North Esk and Spey, Scotland. Proportions of 1SW adults of varying river age (S1–S3; river ages 1 to 3 years, respectively) sampled between May and Au‐ gust/September each year. The proportions of S4 and S5 fish (which were scarce or absent) are not plotted [River North Esk data for 1982 and 1983 were derived from Shearer (1990), his sum‐ mary Tables 2 and 3].

ICES SGBICEPS REPORT 2010 | 117

Table 5.8.2. Rivers Tweed, Tay, North Esk, Spey, Scotland. Summary of significant F ratios and probabilities from generalized additive model analysis of the variation in proportions of river age classes over the period 1963‐2008 (data in Figure 5.8.1). ns = not significant.

______1SW return adults (River comparisons: Tweed, Spey, Tay against North Esk) S1 versus S2+S3+S4 Year River River*Year F P 9.39 <<0.001 ns ns ns S2 versus S3+S4 F P F P F P 36.08 <<0.001 2.42 0.027 ns 16.06<<0.001 (Tweed) S3 versus S4 F P F P F P 20.77 <<0.001 ns ‐2.27 0.025 8.60 0.004 (Spey) (North Esk) ‐2.32 0.022 3.54 0.002 (Tay) (Tweed) ‐9.13 <<0.001 (Tweed) ______As for “summer” 2SW fish (Table 5.7.2), each river age grouping for 1SW salmon showed a significant and positive effect of the proportion of the older river age‐ class(es) (Table 5.8.2). For the proportion of S2 there was a significant, positive effect of Year, indicating a general increase in the proportion of S2 compared to S3+S4 over the time series: that is, the dominant river age‐class shows a general increase with time. For S3 vs S4, all rivers showed a decrease in S3 fish and there were significant interactions among Years for the Rivers North Esk and Tweed. The overall outcome for 1SW adults is that there is a clear, general concordance in the time‐series of mean river age for the rivers Spey, Tay and North Esk (Table 5.8.1), with the Tweed quite distinct. Separation of the annual data into the three primary river‐age cohorts (Figure 5.8.2) also affirms the similarity of the rivers Tay, North Esk and Spey in contrast to the highly variable Tweed. Whilst Spey, Tay and North Esk 1SW salmon are characterised by river age structures that are broadly similar (dominated by S2 and S3), and are relatively stable over time, the Tweed is clearly distinct in showing a highly variable mean river age and occa‐ sionally very high proportions of S1 smolts (Figure 5.8.3). The similarity of the Tay data to the North Esk and Spey is especially striking because the Tay nets were sta‐ tioned in the estuary downstream of Perth and effectively comprised a mixed‐stock fishery: fish sampled at that juncture are variously destined for the rivers Tay, Earn, Almond, Isla, Ericht, Tummel and Garry. That this effective pooling of distinct popu‐ lations still renders an overall result closely comparable to the rivers North Esk and Spey is indicative of the generality of the causal environmental factors and the demo‐ graphic responses of the fish. The Tweed 1SW (Figure 5.8.2) and “summer” 2SW (Figure 5.7.2) adult cohorts show very similar fluctuations in river age‐structure, with the S1 component of the 1SW year‐classes ranging from 2% (in 1969) to 60% (in 1993). In view of the dominance of S2s amongst the returning adult 1SW fish it is likely that the changes for the Tweed

118 | ICES SGBICEPS REPORT 2010

are typified by a tendency for earlier smoltification and a downshift from S2 to S1. For the remaining three rivers there is a striking consistency of overall river age throughout the time series. Assuming that the river age‐structures of returning adult 1SW and 2SW adult salmon analyzed here are a reflection of the original annual emi‐ grant smolt cohorts, this indicates that changes in river age‐structure of the returning adult populations are a feature of the specific river catchment (the freshwater envi‐ ronment) and not the shared oceanic environment.

Figure 5.8.3. Rivers Tweed, Tay, North Esk, Spey, Scotland. Proportions of river ages S1‐S4 amongst returning adult 1SW salmon.

For the River North Esk, the trend for both the 1SW and “summer” 2SW cohorts in the most recent decade has been a slight but progressive shift towards younger over‐ all river age and this is reflected in a downshift from S3 to S2. It has to be reiterated that a decline in one proportion must lead to a compensatory increase in one or more other proportion(s), and that the present data cannot unequivocally identify if these changes relate to either or both of the S2 and S3 categories. However, considering the known relationship between growth rate and age at smoltification (e.g. Økland et al., 1993), the most cautious explanation is of an increasing recent tendency for River North Esk parr to undergo smoltification at two, rather than three years of age. Whether this recent pattern over the last decade is confined to the River North Esk, or is more widespread among these Scottish rivers, is not testable due to the lack of suitably comprehensive sampling on the remaining three rivers over that time period.

ICES SGBICEPS REPORT 2010 | 119

5.9 Analysis of the time-series of the emigrant smolt run, River North Esk (1964–2008) The comprehensive smolt trap and net samples for the River North Esk present the only opportunity of comparing the age‐structure of the annual emigrant smolt popu‐ lation with the river age‐structure of returning adults. Smolts have been sampled on the River North Esk since 1964. Data between 1964 and 1969 were derived from smolts caught in the filter screens of the waterworks at Kinnaber Mill Lade in the lower river. From 1970 onwards the data were derived from smolts sampled in a purpose‐built trap at Kinnaber Mill. Figure 5.9.1 indicates the age composition of the emigrating smolts for the two time‐series (1964–1969, 1970–2008).

NORTH ESK, emigrant smolts 1964-2008

1.0

0.9

0.8

0.7

0.6

0.5

0.4 Proportion 0.3

0.2

0.1

0.0 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008

S1 S2 S3 S4

Figure 5.9.1. River North Esk, Scotland. Proportional composition of the river ages of emigrant smolts, 1964–2008.

There has been a slight, but significant, increase in the proportion of S1 smolts (P = 0.037), no change in the proportion of S2 smolts (P = 0.068) and marked and signifi‐ cant decreases in both the S3 (P = 0.004) and S4 (P = 0.001) smolts. Given the scarcity of S4 juveniles throughout the time series, it appears that the major population‐level change in proportional age‐structure has been the decrease in the relative abundance of S3 smolts, reflected by the slight proportional increase in S1s and decrease in S4s. An overall slight downward shift in mean river age of the total emigrant smolt popu‐ lation (which includes fish which will subsequently mature as a mixture of 1SW and MSW adults) over the 45 years is therefore apparent. Such a trend may well be expli‐ cable by recent warming of the freshwater environment. For comparisons between the age‐structure of the returning adult cohorts and the relevant year‐classes of emigrant smolts, only the data between 1975 and 1999 are analyzed. These data include all years that adult salmon samples (1SW & 2SW) are available between February and August inclusive, and can be compared with the emigrant smolt data for just the Kinnaber trap. The latter excludes data between 1970 and 1974 (owing to concerns about the reliability of the data in those years) so that a period of inter‐annual comparability is assured. Although there is no netting be‐ tween 1 September and February 15, it is apparent (from the automated fish counter at Craigo in the lower River North Esk) that the bulk of both the 1SW and MSW stock components are effectively ‘sampled’ by the nets. 1SW fish continue to enter the river through September and October and MSW fish return in all months of the year, but

120 | ICES SGBICEPS REPORT 2010

numbers outwith the netting season are relatively small. Pooling of the 1SW and 2SW adult data (with appropriate 1 and 2 year lags) thus permits ‘reconstruction’ of smolt year‐classes of returning adults and matching to their year of emigration from freshwater. Throughout these analyses, the net catch data (taken over 6.5 months of the year) are assumed to have been representative of the entire annual return migra‐ tions of adult salmon to the River North Esk.

NORTH ESK, 1SW return adults, 1976-1999

1 0.9 0.8 0.7 0.6 0.5 S1 0.4 S2 0.3 S3 0.2 S4 0.1 Proportion in return adult 1SW adult Proportion in return 0 0 0.2 0.4 0.6 0.8 1 Proportion in emigrant smolt run (year - 1)

Figure 5.9.2. River North Esk, Scotland. Proportion of S1–S4 river ages among returning adult 1SW fish (1976–1999) in relation to the proportion of emigrant S1–S4 smolts in the previous year. The line is not a fitted regression, but a plot of the one‐to‐one relationship between the two vari‐ ates.

The proportional river age‐structure of returning 1SW fish and that of the emigrant smolts leaving the river the previous spring (Figure 5.9.2) shows a close concordance. As a generalisation, the overall river age‐structure of returning adult 1SW fish reflects that of the emigrant population in the previous year, although adults that migrated as S2 smolts are slightly over‐represented whereas adults that migrated as S1s and S3s are slightly under‐represented. A broadly similar pattern is apparent for 2SW returning adults compared with the smolt run two years previously (Figure 5.9.3), although in this case the S3 river age is slightly over‐represented.

ICES SGBICEPS REPORT 2010 | 121

NORTH ESK, 2SW return adults, 1977-1999

1 0.9 0.8 0.7 0.6 0.5 0.4 S1 0.3 S2 0.2 S3 0.1 S4 Proportion in return adult 2SW adult in return Proportion 0 0 0.2 0.4 0.6 0.8 1 Proportion in emigrant smolt run (year - 2)

Figure 5.9.3. River North Esk, Scotland. Proportion of S1–S4 river ages among returning adult 2SW fish (1977–1999) in relation to the proportion of emigrant S1–S4 smolts two years earlier. The line shows the one‐to‐one relationship between the two variates.

When the 1SW and 2SW adult cohorts in consecutive years are combined and com‐ pared with the relevant emigrant smolt runs (e.g. the age‐structure of 2SW adults in 1980 is combined with the 1SW adults of 1979, and the proportionalities of that pooled adult sample are plotted against the proportions of S1 to S4 fish in the 1978 emigrant smolt run), the data converge on the line of unity (Figure 5.9.4).

NORTH ESK, 1SW + 2SW return adults, 1977-1999

1 0.9 0.8 0.7 0.6 0.5

2SW S1 0.4 S2 0.3 S3 0.2 S4 0.1

Proportion in return adult 1SW + 1SW adult Proportionin return 0 0 0.2 0.4 0.6 0.8 1 Proportion in emigrant smolt run

Figure 5.9.4. River North Esk, Scotland. Proportion of S1–S4 river ages among returning adult 1SW and 2SW fish (1977–1999) in relation to the proportion of S1–S4 in the relevant emigrant smolt run. The line shows the one‐to‐one relationship between the two variates.

This convergence is summarised in Table 5.9.1, which provides the mean residuals (proportions) for the four river ages S1–S4 for 1SW and 2SW adults compared against the relevant emigrant smolt age‐structures. Larger mean positive and negative re‐ siduals respectively indicate over‐ and under‐representation of that river age‐group in the returning adult cohorts. The average residual (proportion) from the line of

122 | ICES SGBICEPS REPORT 2010

unity for the respective S2, S3 and S4 river ages in the melded year‐classes was 0.011 or less.

Table 5.9.1. Mean residuals (proportions) for the four river ages S1–S4 for 1SW adults 1976–1999 and 2SW adults 1977–1999 compared against the relevant emigrant smolt age‐structures for 1975– 1998. Note the convergence of the residuals for the melded population both for S2 and S3.

______S1 S2 S3 S4 1SW adults ‐0.013 0.038 ‐0.023 ‐0.002 2SW adults ‐0.027 0.024 0.043 0.008 1SW+2SW (meld) ‐0.019 0.011 0.004 0.004 ______For the period 1975–1999, there is a clear concordance between the age‐structure of the River North Esk emigrant smolt year‐classes and the age‐structure of the subse‐ quent returning adult cohorts one (1SW) and two (2SW) years later. This result may have important implications for our understanding of mortality of salmon at sea and in informing management strategies. If generally applicable to other river catch‐ ments, this implies that monitoring of the river‐age structure of emigrant smolts has predictive value for subsequent returning adults. Similarly, by ascertaining annual variation in the river age‐distribution from adult scale archives, the hindcasting of changes in age at smoltification can be undertaken with confidence. Perhaps the most important outcome from the data in Figure 5.9.4 is the clear evi‐ dence, for both the 1SW and 2SW stock components, of an overall lack of systematic, river age‐related variation in survival at sea, and an absence of any ocean‐related in‐ fluence on returning adult age‐structures. The River North Esk data show that river age at smoltification has no influence on ultimate adult survivorship probabilities and different aged smolts appear to have an effectively equal probability of returning to reproduce. Thus, for example, that S2 and S3 smolts appear to show no clear in‐ crease or decrease in their ultimate survivorship (Figure 5.9.4) indicates that early mortality at sea (which is generally perceived to be critical to adult recruitment) is not river‐age dependent. Since there is a clear tendency for older smolts to be slightly larger at first entry to seawater (Figures 4.9.2 & 4.9.3) and to migrate earlier in the spring smolt run (J.C. MacLean; unpublished data for River North Esk), this suggests that early mortality at sea is also independent of smolt size and timing of sea entry. Any specific statistical test for the extremes of river age is rendered weak for the River North Esk because of the relative scarcity of S1 and S4 fish. In this context, emi‐ grant smolt data for other rivers, such as the River Tweed, where S1 fish comprise a larger part of the smolt run, would have been invaluable. However, for the majority of River North Esk fish it is clear that S2 and S3 smolts have comparable rates of sur‐ vivorship at sea and probabilities of returning to spawn, either as 1SW or 2SW adults. These data also can further inform analyses of the possibilities of temperature / food / environment match‐mismatch hypotheses affecting salmon growth and survivorship early in their marine migration. If these data for the River North Esk are characteristic of other UK and southern NEAC populations, the possibility is that environmental match‐mismatch is not a driver of short‐ and long‐term variations in the age‐structure of returning adults. That is, at least for the period 1975–1999, earlier‐migrating (older and typically larger) smolts do not appear to have been subject to disproportionately high (or low) survivorship compared to younger (possibly later‐emigrating and typi‐ cally smaller) smolts. It has to be emphasized here that the present analyses focus on

ICES SGBICEPS REPORT 2010 | 123

relative frequencies of the river age‐classes, and not on abundances of the distinct maturity groupings and year‐classes. Environmental match‐mismatch may well in‐ fluence overall year‐class success in terms of adult recruitment abundances, but it appears unlikely to influence the river age‐structure of successful return adults. The combined River North Esk time‐series for emigrant smolt age‐structure, 1964– 2008 (Figure 5.9.1), indicates an overall trend of reducing mean river age and de‐ creases in the proportions of S3 and S4 smolts. This long‐term pattern is detectable also in the age‐structures of returning summer 2SW and 1SW adults (Figures 5.8.2. & 5.7.2), but objective and comprehensive testing of this is compromised by the lack of data for early spring 2SW fish from 2000 onwards. For the River North Esk, the last decade appears to have been characterised by progressive changes in the age‐ structures of both emigrant smolts and returning 1SW and 2SW adults. Unfortu‐ nately, the lack of comprehensive data for the remaining three monitored Scottish rivers (due to temporally inadequate sampling within years, voluntary deferment of the commencement of netting for conservation purposes, and the complete closure of two of the net fisheries) precludes among‐river comparisons for the most recent years.

5.10 Correlations between the length of returning 1SW salmon and the PFA of maturing (1SW) salmon The extended data sets of length at return made available to the Study Group enabled new analyses to be completed for NEAC stocks. In order to compare the size of 1SW salmon across a number of rivers in both the Northern and Southern NEAC areas, the size of maiden 1SW salmon was standardised using the mean and standard deviation for individual populations over the period 1989–2007. This period was chosen as it enabled information from as many rivers as possible to be included; data for 12 rivers in the Northern NEAC area and 6 rivers from Southern NEAC were examined. The z‐ score for 2008 was also calculated using the 1989–2007 period to standardise, though data for 2008 were only available for six rivers in Northern NEAC and three rivers in the Southern NEAC area. The mean length of 1SW salmon from Northern and Southern NEAC was found to be significantly correlated over the period 1989–2008 (r = 0.60, n = 20, p = 0.006, Figure 5.10.1 and Figure 5.10.2), suggesting that 1SW salmon across a large area experience similar growth conditions during their first year at sea. Furthermore, z‐scores of 1SW length were also significantly correlated with the pre‐fishery‐abundance of maturing 1SW salmon for both the Northern and Southern NEAC areas (N NEAC: r = 0.70, n = 20, p = 0.001, Figure 5.10.3; S NEAC: r = 0.49, n = 20, p = 0.027, Figure 5.10.4). This suggests that reduced growth leads to reduced survival or delayed maturation in the smolt cohorts experiencing lower growth, and/or that better growth leads to in‐ creased survival or earlier maturation.

124 | ICES SGBICEPS REPORT 2010

Figure 5.10.1. Z‐scores of 1SW salmon length computed for Northern and Southern NEAC areas for the period 1989–2008.

Figure 5.10.2. Z‐scores of 1SW salmon length computed for Northern and Southern NEAC areas for the period 1989–2008, plotted against each other.

ICES SGBICEPS REPORT 2010 | 125

Figure 5.10.3. PFA of maturing 1SW salmon plotted against the z‐score of 1SW length for North‐ ern NEAC populations.

Figure 5.10.4. PFA of maturing 1SW salmon plotted against the z‐score of 1SW length for South‐ ern NEAC populations.

126 | ICES SGBICEPS REPORT 2010

5.11 Is there evidence of a change in overall (egg-adult) mortality over time? This was explored using a simple life‐history model (Aprahamian et al., 2008). For salmon, Aprahamian et al. (2008) suggested that age at maturity was optimised to maximise life time fecundity, defined as the average number of eggs per female sur‐ viving to spawn. The optimal age at maturity is a function of mortality, growth (af‐ fecting smolt age) and fecundity. The model assumes a steady state, in that the number of eggs spawned in a particular year equals the number of eggs deposited by that cohort over its lifetime. The model requires the following input parameters: mean smolt age, proportion of fish maturing by sea age, fecundity by sea age and sex ratio of the returning adults. To validate the outputs of the model (number of spawning adults by sea age) the model was run on data from five Southern NEAC index rivers: Welsh Dee, Burrishoole, Bush, Frome and River North Esk, for various, ~ 5 year time periods. For each of the rivers and time periods, river specific estimates of smolt age and sea age were available; for fecundity, river specific values were available for all the rivers except the Frome and Bush where default values were used (Table 5.11.1). For all riv‐ ers the sex ratio for all sea age classes was assumed to be 1:1. To estimate the optimal age at maturity which maximises lifetime fecundity, the model was re‐run using the same life‐history parameters (smolt age, sex ratio, fecun‐ dity), the mortality estimate from the life history model, but with no constraint on the maturation schedule. (For ease of differentiating between the two models this model will be referred to as the Optimal Age model).

Mortality A comparison between the total number of salmon estimated to spawn (Observed) and that estimated from the life‐history model (Estimated) was within ±10% for the Frome, Dee, Burrishoole and for the Bush with the exception of the sample 1975–1979 (Table 5.11.2). For the River North Esk, the life‐history model consistently overesti‐ mated the number of spawning salmon by between 12 and 28%. The overestimation may arise if the fecundity estimate is too low and/or if the proportion of females is higher than 50%. For the rivers North Esk, Bush and Burrishoole where the time period extended pre and post‐1989, the year when the increase in marine mortality was perceived to have occurred, there does not appear to have been a systematic increase in the overall life‐ time mortality (Table 5.11.2) for these three stocks.

Sea age at maturity A comparison between the observed sea age structure of adult salmon spawning and that estimated from the model for a given egg deposition was compared using chi‐ square. There was no significant difference between the proportion of fish maturing at 1SW and 2SW between that observed and that estimated from the model for the rivers NorthEsk, Frome, Bush and Burrishoole (Table 5.11.2). However, for the Dee the life‐history model over estimated the proportion of MSW salmon, with the mean sea age being significantly higher by between 0.02 and 0.03 years when compared with the observed sea age structure. The fact that the difference in the case of the Dee was small (<5%) and that there was no significant difference between the observed and expected proportions for the other index rivers suggests that the life‐history model provides a reasonable estimate of the maturation schedule.

ICES SGBICEPS REPORT 2010 | 127

Table 5.11.1. Input data to estimate overall egg to adult mortality

River Time period Egg deposition Mean smolt age Fecundity Fecundity Fecundity Proportion Proportion Proportion 1SW 2 SW 3 SW 1SW 2SW 3SW 1981‐84 29 220 000 2.26 4861 8025 0.65 0.35 1985‐89 31 862 000 2.27 4861 8025 0.67 0.33 N. Esk 1990‐94 30 556 000 2.16 4861 8025 0.66 0.34 1995‐99 31 655 040 2.16 4861 8025 0.67 0.33 2000‐04 40 364 000 2.08 4861 8025 0.67 0.33 Frome 2000‐08 1 905 000 1.0 4310 7998 0.84 0.16 1992‐94 17 477 917 1.61 4632 7406 9114 0.69 0.27 0.01 Dee 1995‐99 13 938 870 1.63 4502 7622 9186 0.70 0.28 0.01 2000‐04 14 199 291 1.65 4485 7215 9529 0.73 0.24 0.01 1975‐79 1 354 000 1.66 3988 5438 0.77 0.23 1980‐84 1 382 000 1.87 4048 5712 0.85 0.16 1985‐89 3 332 000 1.88 3898 5712 0.76 0.24 Bush 1990‐94 2 646 000 1.80 3579 5730 0.85 0.16 1995‐99 1 654 000 1.76 3384 5730 0.82 0.18 2000‐04 1 122 800 1.84 3636 5898 0.91 0.09 1970‐74 2 185 560 2.0 4151 7081 0.98 0.02 1975‐79 1 540 520 2.0 4151 7081 0.95 0.05 1980‐84 1 012 920 2.0 4151 7081 0.92 0.08 Burrishoole 1985‐89 996 040 2.0 4151 7081 0.95 0.05 1990‐94 805 680 2.0 4151 7081 0.95 0.05 1995‐99 964 320 2.0 4151 7081 0.97 0.03 2000‐04 1 107 440 2.0 4151 7081 0.98 0.02

128 | ICES SGBICEPS REPORT 2010

Table 5.11.2. A comparison between the observed and the estimated number of salmon, by sea age, from the life history model, from a given egg deposition.

River Time Lifetime Number Number Number Mean sea‐ Number Number Number Number Mean sea‐ Number Significance period Mortality 1SW 2SW 3SW age (Obs.) 1SW (Est.) 2SW (Est.) 3SW (Est.) 4SW (Est.) age (Est.) Estimate / Chi‐square Z (Obs.) (Obs.) (Obs.) Observed 81‐84 1.665 5165 2721 1.35 6431 3387 1.34 1.24 P>0.05 85‐89 1.666 5688 2745 1.33 7289 3526 1.33 1.28 P>0.05 N. Esk 90‐94 1.700 5701 2906 1.34 6822 3483 1.34 1.20 P>0.05 95‐99 1.702 6367 3169 1.33 7154 3555 1.33 1.12 P>0.05 00‐04 1.730 7874 3938 1.33 9104 4545 1.33 1.16 P>0.05 Frome 00‐08 2.303 614 117 1 1.16 653 124 1 1.16 1.06 P>0.05 92‐94 1.815 4577 1736 91 1.30 4436 1743 136 20 1.33 1.02 P<0.05 Dee 95‐99 1.856 3768 1394 32 1.28 3593 1451 65 4 1.31 1.02 P<0.05 00‐04 1.843 4078 1343 35 1.26 3990 1347 74 6 1.28 1.02 P<0.05 75‐79 1.874 624 186 1.23 483 144 1.23 0.77 P>0.05 80‐84 1.830 530 97 1.16 542 100 1.16 1.02 P>0.05 85‐89 1.779 1292 399 1.24 1169 369 1.24 0.91 P>0.05 Bush 90‐94 1.834 1147 210 1.16 1133 210 1.16 0.99 P>0.05 95‐99 1.832 638 144 1.18 713 156 1.18 1.11 P>0.05 00‐04 1.865 518 52 1.09 532 53 1.09 1.03 P>0.05 70‐74 1.886 965 15 1.02 1018 21 1.02 1.06 P>0.05 75‐79 1.858 638 33 1.05 681 36 1.05 1.07 P>0.05 80‐84 1.835 398 33 1.08 425 37 1.08 1.07 P>0.05 Burrishoole 85‐89 1.858 414 20 1.05 440 23 1.05 1.07 P>0.05 90‐94 1.858 337 15 1.04 356 19 1.05 1.06 P>0.05 95‐99 1.876 416 12 1.03 441 14 1.03 1.06 P>0.05 00‐04 1.886 482 11 1.02 516 11 1.02 1.07 P>0.05

ICES SGBICEPS REPORT 2010 | 129

Optimal age at maturity The model was re‐run to predict the optimal sea age at maturity that maximised the number of eggs produced per female for the value of mortality estimated for the various periods from the five index rivers. The optimal age model used the estimate of lifetime mortality from the life‐history model, the same estimates of smolt age, growth and fecundity, but with no constraint on the sea age composition. There was no significant difference between the observed proportion of salmon maturing at dif‐ ferent sea ages and that estimated from the model for the rivers NorthEsk, Frome, Bush and Burrishoole (Table 5.11.3). This suggested that for these four rivers lifetime fecundity is maximised. However, for the Dee the model suggests a significantly higher proportion of MSW salmon than the proportion observed if fecundity is maximised. The difference in terms of the mean sea age is between 13 and 30% (Table 5.11.3). For the Welsh Dee, the difference between the observed and expected sea age at ma‐ turity may relate to smolt age. The optimal age model predicts that for any given level of mortality the tendency is for S1 smolts to mature at a higher sea age than S2 or older smolts (Figure 5.11.1). This is also evident from the data, where for female salmon there is a significantly higher proportion of S1 smolts maturing as MSW salmon (47.2%) when compared to S2 smolts (30.7%) (Table 5.11.4). For males the proportion is much similar, 16.4% and 14.0% for S1 and S2 smolts maturing as MSW salmon, respectively.

3.5 S1 3.0 S2 2.5

2.0 Mean sea age Mean 1.5

1.0 1.50 1.75 2.00 2.25 2.50 Egg-Adult instantanenous mortality

Figure 5.11.1. The optimal sea age at maturity for S1 and S2 smolts at different levels of mortality.

To explore the effect of gender differences, and as the model was based on optimising age at maturity for female salmon (with the assumption that males would respond in a similar manner), an analysis of just the female component was undertaken for the River Dee. The mean smolt age and sea age composition of female salmon from the Dee are shown in Table 5.11.5, together with the estimated life time mortality and the estimated sea age composition from the life‐history model. There was no significant difference between the observed and estimated sea age composition. The egg‐adult mortality estimated from the life‐history model and the mean smolt age of females were used in the optimal age model to estimate the sea age composition which

130 | ICES SGBICEPS REPORT 2010

maximises life time fecundity (Table 5.11.6). There was a significant difference be‐ tween the observed sea age composition and that estimated by the model; the optimal age model over‐estimated mean sea age by between 4.6 and 16.1% (Table 5.11.6). The model predicted a higher proportion of MSW females than observed. However, the difference between the observed and expected was less when the analysis focused solely on females than when both sexes were included in the analysis (Table 5.11.3).

ICES SGBICEPS REPORT 2010 | 131

Table 5.11.3. Optimal age at maturity for the N.Esk, Frome, Dee, Bush and Burrishoole for various time periods (MSA = mean sea age).

River Time period Lifetime Prop. Prop. Prop. Mean sea Prop. Prop.. Prop. Mean sea MSA Significance Mortality 1SW (Obs.) 2SW (Obs.) 3SW (Obs.) age 1SW 2SW 3SW age (Est.) Estimate / Chi‐square Z (Obs.) (Est.) (Est.) (Est.) Observed 1981‐84 1.665 0.655 0.345 1.35 0.657 0.343 1.34 0.998 P>0.05 1985‐89 1.666 0.674 0.326 1.33 0.674 0.326 1.33 1.000 P>0.05 N. Esk 1990‐94 1.700 0.662 0.338 1.34 0.661 0.339 1.34 1.001 P>0.05 1995‐99 1.702 0.668 0.332 1.33 0.669 0.331 1.33 0.999 P>0.05 2000‐04 1.730 0.667 0.333 1.33 0.667 0.333 1.33 1.000 P>0.05 Frome 2000‐08 2.303 0.839 0.16 1.16 0.858 0.142 1.14 0.983 P>0.05 1992‐94 1.815 0.71 0.27 0.01 1.30 0.311 0.689 0.000 1.69 1.30 P<0.05 Dee 1995‐99 1.856 0.73 0.27 0.01 1.28 0.561 0.439 0.000 1.44 1.13 P<0.05 2000‐04 1.843 0.75 0.25 0.01 1.26 0.542 0.458 0.000 1.46 1.16 P<0.05 1975‐79 1.874 0.77 0.23 1.23 0.79 0.21 1.21 0.984 P>0.05 1980‐84 1.830 0.85 0.16 1.16 0.88 0.12 1.12 0.972 P>0.05 1985‐89 1.779 0.76 0.24 1.24 0.78 0.22 1.22 0.986 P>0.05 Bush 1990‐94 1.834 0.85 0.16 1.16 0.83 0.17 1.17 1.012 P>0.05 1995‐99 1.832 0.82 0.18 1.18 0.79 0.21 1.21 1.021 P>0.05 2000‐04 1.865 0.91 0.09 1.09 0.91 0.09 1.09 0.998 P>0.05 1970‐74 1.886 0.98 0.02 1.02 0.98 0.02 1.02 1.004 P>0.05 1975‐79 1.858 0.95 0.05 1.05 0.95 0.05 1.05 0.998 P>0.05 1980‐84 1.835 0.92 0.08 1.08 0.92 0.08 1.08 1.002 P>0.05 Burrishoole 1985‐89 1.858 0.95 0.05 1.05 0.95 0.05 1.05 1.002 P>0.05 1990‐94 1.858 0.95 0.05 1.05 0.95 0.05 1.05 1.002 P>0.05 1995‐99 1.876 0.97 0.03 1.03 0.97 0.03 1.03 1.001 P>0.05 2000‐04 1.886 0.98 0.02 1.02 0.98 0.02 1.02 0.997 P>0.05

132 | ICES SGBICEPS REPORT 2010

Table 5.11.4. Proportion of S1 and S2 smolts from the River Dee maturing at different sea ages for female and male salmon, mean ± 95 CL for the period 1992–2008.

Sea age Smolt age Female Male 1SW 2SW 3SW 1SW 2SW 3SW S1 Smolt 0.528 ± 0.089 0.458 ± 0.087 0.015 ± 0.009 0.836 ± 0.040 0.157 ± 0.037 0.008 ± 0.003 S2 Smolt 0.693 ± 0.039 0.299 ± 0.039 0.008 ± 0.004 0.860 ± 0.046 0.134 ± 0.043 0.006 ± 0.003

Table 5.11.5. The mean smolt age and sea age composition of female salmon from the Dee and the estimated egg‐adult mortality and sea age composition from the life history model for three time periods.

Year Mean Prop. Prop. Prop. Mean sea Lifetime Prop. Prop. Prop. Mean sea Significance smolt age 1SW 2SW 3SW age Mortality 1SW 2SW 3SW age (Est.) Chi‐square (Obs.) (Obs.) (Obs.) (Obs) Z (Est.) (Est.) (Est.) 1992‐94 1.674 0.681 0.305 0.015 1.33 1.972 0.681 0.304 0.014 1.34 P>0.05 1995‐99 1.667 0.649 0.346 0.005 1.36 2.007 0.649 0.346 0.001 1.36 P>0.05 2000‐04 1.717 0.651 0.341 0.008 1.36 1.977 0.651 0.341 0.008 1.36 P>0.05

Table 5.11.6. The mean observed sea age composition and that estimated from the optimal age model which maximises life time of female from the River Dee for three time periods, together with the mean smolt age and life time mortality used to estimate the optimal sea age (MSA = Mean Sea Age)

Year Prop. Prop. Prop. Mean sea Lifetime Mean Prop. Prop. Prop. Mean sea Significance MSA 1SW 2SW 3SW age Mortality smolt 1SW 2SW 3SW age (Est.) Chi‐square Est. / Obs. (Obs.) (Obs.) (Obs.) (Obs) Z age (Est.) (Est.) (Est.) 1992‐94 0.681 0.305 0.015 1.33 1.972 1.674 0.452 0.548 0.000 1.55 1.161 P<0.05 1995‐99 0.649 0.346 0.005 1.36 2.007 1.667 0.581 0.419 0.000 1.42 1.046 P<0.05 2000‐04 0.651 0.341 0.008 1.36 1.977 1.717 0.569 0.431 0.000 1.43 1.055 P<0.05

ICES SGBICEPS REPORT 2010 | 133

Discussion The close agreement between the sea age at maturity estimated from the model and that observed suggests that fitness, defined as the number of eggs per female, is maximised indicating that the age at maturity is being optimised. This is supported from the findings from the rivers North Esk, Frome, Bush and Burrishoole and from females from the River Dee. There are significant differences between the observed and expected sea age composition and part of these differences may arise from the values of the parameters used in the model (sex ratio, fecundity). In addition, the analysis of the River Dee data indicates that for a given level of mortality and growth, males cannot automatically be assumed to respond in a similar manner as females and this is likely to be more evident when there are marked gender differences in life history parameters. In terms of whether biological characteristics, specifically sea age at maturity, can be used to predict changes in mortality seems unlikely. Changes in life history character‐ istics reflect changes in mortality and/or changes in growth and fecundity, as such, a change in mortality will only retrospectively manifest itself in terms of a change in age at maturity after an (unknown) number of generations. The findings presented here do not indicate that there has been any major change in total mortality, from egg to spawning adult, since the late 1970s – early 1980s in the rivers North Esk, Bush and Burrishoole. This is presumably because the decline in natural mortality, notably in marine survival around 1989, has been off‐set by an equivalent decline in fishing mortality and/or other anthropogenic mortalities. This would explain why there has been no major change in the age (sea) at maturity in these rivers over the time period. It is possible to retrospectively estimate mortality (lifetime mortality) using the life history model. For some river systems historic information are available on the smolt age and sea age composition and from these data together with default estimates for sex ratio and fecundity (as these are generally unknown) it is possible to estimate the lifetime mortality. This was undertaken for two river systems where a long time se‐ ries of data on biological characteristics were available: the Dee and Hampshire Avon, to investigate how the level of mortality had changed over a 60 year period (Table 5.11.7). The life history model estimates that the mortality on the Avon had been higher than on the Dee for all time periods, by between 5–13%. For both rivers it is estimated that mortality had increased, for the Dee mortality was estimated to have increased by 2.9% and the Avon by 6.4% between the late1930s and 1960s. However, between the late 1960s and 1990s there was estimated to be a larger increase in mor‐ tality on the Dee (22.5%) than on the Avon (13.6%). Over the entire 60 years the esti‐ mated increase in mortality on the Dee was slightly higher (26.0%) than on the Avon (20.9%). If the age at maturity is optimized then it is possible to use the optimal age model to examine the consequences of a change in mortality on the sea age composition for any given set of conditions (growth, smolt age, fecundity and sex ratio). A reduction in overall mortality has the effect of increasing the mean sea age (Figure 5.11.1) but will reduce the number of returning adults, an increase in mortality will have the op‐ posite effect. This was examined on the rivers Bush and North Esk where data were available on the number of salmon returning before exploitation in the net and rod fisheries. For these rivers it was assumed that the number of salmon returning to the river spawned (i.e. fishing mortality and all other anthropogenic mortality was zero).

134 | ICES SGBICEPS REPORT 2010

The instantaneous egg‐adult mortality in the absence of any mortality once the fish arrive in home waters was estimated as follows:

Z = (LnN0‐LnNt)/t equation 1 Where:

N0 = the number of eggs deposited

Nt = Number of salmon returning (prior to the fishery (Ntpf) or on the spawn‐ ing grounds (Nts)) t = total age. As the value of Z was not estimated in the same manner as in the life history model the value was adjusted as follows:

Zadj =(Zpf)/(Zs)* Zslh. Where:

Zpf = the estimate of mortality prior to the fishery using equation 1

Zs = the estimate of mortality at the time of spawning using equation 1

Zslh = the estimate of mortality at the time of spawning using the life history model. In the absence of any fishing mortality the optimal age model estimated a higher mean sea age indicating a higher proportion of MSW salmon than was observed (Ta‐ ble 5.11.8). If fishing mortality was reduced to zero, it is hypothesized that there would be an increase in the number of spawners until the carrying capacity was reached and then numbers of spawning adults would decrease as the age at maturity increased in order to maximise fecundity. It is appreciated that this is a simplistic ap‐ proach and that the change in sea age composition would not increase by the amount suggested in Table 5.11.8 if fishing mortality was reduced to zero. This is because the increase in the number of spawners would result in an increase in density dependent mortality resulting in a reduction in the age at maturity, assuming the catchment was at carrying capacity. Though the analysis must be viewed with caution it does illus‐ trate that it might be possible to increase the MSW component of the stock through a reduction in fishing (or other) mortality. It is thus important that such models inte‐ grate with population models. Certainly both the life history and optimal age models would benefit from estimating / using life stage specific values of mortality. This would then enable changes in mortality at specific life stages on the age at maturity to be evaluated. In terms of population models the fact that age at maturity appears to be optimized provides a feedback mechanism for such models, in that the overall mortality must be the same. Similar data as used in the life‐history model, namely the numbers of male and fe‐ male smolts emigrating and the numbers of adults returning by sex and sea age can be used to derive estimates of natural mortality during the marine phase for stocks which mature at two or more different sea ages (Ricker, 1976). One particular ap‐ proach referred to as “Murphy’s Method” (Ricker, 1975) was used to estimate the mortality of Icelandic ranched Atlantic salmon over their second year at sea (Jonasson et al. 1994). Chaput et al. (2003) described a variation of this method which allows estimation of mortality in both the first and second years in the sea. Application of this approach to the results described above (e.g. Table 5.11.5) might therefore allow partitioning of the life‐time mortality into separate life stages, namely freshwater and

ICES SGBICEPS REPORT 2010 | 135

the first and second years at sea. The Study Group recommended that this be inves‐ tigated further.

Table 5.11.7 Estimates of total egg‐adult mortality (Z) for the rivers Dee and Hampshire Avon, for three time periods using the life‐history model.

River 1937‐39 1967‐69 1997‐99

Dee 1.495 1.538 1.884 Hampshire Avon 1.635 1.739 1.976

136 | ICES SGBICEPS REPORT 2010

Table 5.11.8. Optimal age structure under zero fishing mortality.

Observed No net and rod fishery No rod fishery River Year Prop. Prop. Prop. Prop. Prop. Prop. Prop. Prop. Zslh MSA Zadj MSA Zadj MSA 1SW 2SW 1SW 2SW 3SW 1SW 2SW 3SW N.Esk 1981‐84 1.665 0.655 0.345 1.35 1.396 0.000 0.319 0.681 2.68 1.561 0.000 0.974 0.026 2.03 1985‐89 1.666 0.674 0.326 1.33 1.440 0.000 0.615 0.385 2.39 1.558 0.000 0.974 0.026 2.03 1990‐94 1.700 0.662 0.338 1.34 1.519 0.000 0.818 0.182 2.18 1.609 0.028 0.972 0.000 1.97 1995‐99 1.702 0.668 0.332 1.33 1.575 0.000 0.947 0.053 2.05 1.642 0.339 0.661 0.000 1.66 2000‐04 1.730 0.667 0.333 1.33 1.624 0.000 0.985 0.015 2.02 1.678 0.396 0.604 0.000 1.60

Bush 1975‐79 1.874 0.77 0.23 1.23 1.606 0.000 0.801 0.199 2.199 1980‐84 1.830 0.85 0.16 1.16 1.617 0.000 0.973 0.027 2.027 1985‐89 1.779 0.76 0.24 1.24 1.702 0.496 0.504 0.000 1.504 1990‐94 1.834 0.85 0.16 1.16 1.696 0.328 0.672 0.000 1.672 1995‐99 1.832 0.82 0.18 1.18 1.653 0.000 0.971 0.029 2.029 2000‐04 1.865 0.91 0.09 1.09 1.764 0.696 0.304 0.000 1.304

ICES SGBICEPS REPORT 2010 | 137

6 Overview and recommendations

Substantial further progress has been made, as detailed in this report, in addressing the three areas identified in the ToR, namely: (a) identifying data sources and compil‐ ing time series of data; (b) considering hypotheses relating to changes in biological characteristics of all life stages and relating these to mortality (freshwater and marine) and/or abundance trends; and (c) conducting preliminary analyses to explore the available datasets and test the hypotheses. A number of additional data sets were made available to the Study Group, and these further reinforced the evidence that the biological characteristics of Atlantic salmon stocks are changing markedly across the geographical range of the species (Sections 5.2–5.5). In addition, a number of hy‐ potheses were developed and explored through various case studies (Section 4) and other exploratory analyses (Section 5). The background to the original request from NASCO stemmed from an ‘interest in determining if declines in marine survival of Atlantic salmon coincide with changes in the biological characteristics of juveniles in freshwater or are modifying characteristics of adult fish (size at age, age at maturity, condition, sex ratio, growth rates, etc.) and with environ‐ mental change’. Evidence reported in the two Study Group reports clearly indicates marked changes in various biological characteristics coincident with a sharp decline in the marine survival for specific stocks. NASCO further asked ICES to ‘investigate associations between changes in biological characteristics of all life stages of Atlantic salmon, environmental changes and variations in marine survival with a view to identifying predic‐ tors of abundance’. The Study Group has identified links between the size of returning 1SW fish and their pre‐fishery abundance, for the NEAC area at least (Section 5.10), suggesting that reduced growth leads to reduced survival or delayed maturation in the smolt cohorts experiencing lower growth. However, while this suggests a direct link between growth/mortality and abundance, it seems unlikely that this could be used in any predictive capacity. In considering progress and possible next steps, the Study Group recognises that this is a vast subject area and that the time available did not allow a full appraisal of the topic, or even the available data. Further, in the absence of oceanographic expertise at the second meeting, or representation from the NW Atlantic at either meeting, the Study Group also recognises that its ability to explore changes in biological character‐ istics of salmon across the entire range of the species and relate these to environ‐ mental variables was inevitably somewhat constrained. The Study Group therefore recognises that further work to better explore temporal and spatial trends, investigate possible common patterns or regional groupings, and develop and test hypotheses would be of value. Members of the Group are involved with some ongoing investiga‐ tions and the production of a number of peer‐reviewed outputs from analyses initi‐ ated and reported here. However, it was felt that the Group had probably gone as far as it could in addressing the ToR and that further progress might best be made by other groups (e.g. involving modellers, biological oceanographers), potentially in re‐ sponse to more targeted questions or to address specific management requirements. The Study Group made the following recommendations: i ) The Study Group reiterated the importance of monitoring programmes to collect age, growth and other biological data from salmon stocks and the need for these to be continued. Such studies are important for exam‐ ining variations in marine recruitment and links with environmental conditions. Where possible, such studies should extend to the collection

138 | ICES SGBICEPS REPORT 2010

and collation of data sets on freshwater stages (e.g. smolt age, smolt size, growth at age, etc.), since there was increasing evidence that freshwater influences might be instrumental to the growth and survival of the fish in the sea. Changes in freshwater or marine growth may be among the first signs that climate is modifying salmonid populations and can be useful in understanding and explaining fluctuations in abundance and inform‐ ing fisheries management in a period of potentially unprecedented envi‐ ronmental change. ii ) The Study Group reiterated the view that such data sets needed to be bet‐ ter utilised and reported, and felt that (subject to funding being available) the development of an inventory of relevant data sets would be a useful initiative. Furthermore the value of a number of data sets when analysed together far exceeds their value when analysed separately. In this con‐ text, feedback from NASWG (ICES, 2010) had recommended that the data sets collated by the Study Group should be fully utilised and made available to NASWG in support of further analyses as appropriate. Pre‐ liminary discussions were held with the ICES Data Centre and further consideration will be given to archiving the data sets and providing suit‐ able supporting documentation. This could perhaps comprise part of an ICES Co‐operative Research Report in which the work of the Study Group could be consolidated. iii ) The Study Group noted that different members were developing peer‐ reviewed outputs related to investigations discussed at the meetings (and reported herein) and recommended that efforts continue to finalise these. Subjects included: smolt run‐timing and potential mis‐match with envi‐ ronmental variables; condition factor analyses of returning adults; life‐ cycle modelling and model validation; sea‐age/river‐age relationships; and an overview of spatial/temporal changes in salmon biological charac‐ teristics. iv ) The Study Group recommends that relationships between mortality (freshwater and marine) and growth, and resulting biological characteris‐ tics, might usefully be explored further through the application of ap‐ propriate life‐history models, including partitioning of the life‐time mortality into separate life stages. With the exception of the recommendation for a possible Co‐operative Research Re‐ port, which ICES would need to approve, these recommendations are addressed largely to the Study Group and to members of WGNAS and WGBAST.

ICES SGBICEPS REPORT 2010 | 139

Annex 1: List of participants

NAME ADDRESS PHONE/FAX EMAIL Miran Environment Agency TEL: +44 1925 542713 miran.aprahamian@environment‐ Aprahamian Richard Fairclough FAX: +44 1925 415961 agency.gov.uk House Knutsford Road Warrington WA4 1HG United Kingdom Ian Davidson Environment Agency TEL: +44 1244 894520 ian.davidson@environment‐ Chester Road FAX: +44 1244 550144 agency.gov.uk Buckley Flintshire CH7 3AJ United Kingdom Peder Fiske Norwegian Institute for TEL: +47 73 80 1522 [email protected] Nature Research FAX: +47 73 80 14 01 NO‐7485 Trondheim Norway Anton Ibbotson Game and Wildlife TEL: +44 1929 401873 [email protected] Conservation Trust, Burgate Manor, Fordingbridge, Hampshire SP6 1EF United Kingdom Richard River Bush Salmon TEL: +44 28 20731435 [email protected] Kennedy Station, 21 Church Street Bushmills, Co Antrim BT57 8QJ United Kingdom Julian C. Marine Scotland, FRS TEL: +44 1674 677070 [email protected] MacLean FL Field Station FAX: +44 1674 672604 Inchbraoch House South Quay, Ferryden Montrose, Angus DD10 9SL United Kingdom Stig Pedersen National Institute of TEL: +45 89213100 [email protected] Aquatic Resources, Department of Inland Fisheries, Vejlsøvej 39 ‐ 8600 Silkeborg, Den‐ mark Ted Potter Cefas Lowestoft TEL: +44 1502 524260 [email protected] Laboratory, Pakefield FAX: +44 1502 513865 Road, Lowestoft, Suffolk NR33 0HT United Kingdom

140 | ICES SGBICEPS REPORT 2010

Ger Rogan Aquaculture & Catch‐ TEL: +353 9842300 [email protected] ment Management Ser‐ FAX: +353 9842340 vices, Marine Institute, Furnace, Newport, Co. Mayo,

Ireland

Ian Russell Cefas Lowestoft TEL: +44 1502 524330 [email protected] (Chair) Laboratory, Pakefield FAX: +44 1502 513865 Road, Lowestoft, Suffolk NR33 0HT United Kingdom Chris Todd Scottish Oceans TEL: +44 1334 463454 cdt@st‐andrews.ac.uk Institute, FAX: +44 1334 463443 University of St Andrews School of Biology, St Andrews, Fife KY16 8LB United Kingdom

ICES SGBICEPS REPORT 2010 | 141

Annex 2: List of working documents/presentations and data sets

Aprahamian M. Is there evidence for optimal age at maturity in salmon ? Barry J. Re‐examination of updated biological characteristics datasets from different stocks – trend analysis. Barry J. Re‐examination of updated biological characteristics datasets from different stocks –linear regression analysis of two‐way relationships. Davidson I. Preliminary investigations into the effect of environmental factors, par‐ ticularly river flow, on recruitment success of salmon on the River Dee, UK (England & Wales). Davidson I. Investigation of long‐term trends in spring temperatures for selected monitored rivers in UK (England & Wales) and adjacent inshore coastal areas – any evidence of thermal mis‐match? Fiske P. Evidence of later age at maturity in Norwegian salmon stocks in recent years. Fiske P. Correlations between the length of returning 1 SW salmon and the PFA of 1 SW salmon. Kennedy R. Long term changes in biological characteristics of smolts on the River Bush, N. Ireland and associations with environmental parameters). MacLean J. Long term changes in the biological characteristics of smolts emigrating from two rivers in Scotland. Pedersen S. Update on work completed by the Baltic Salmon and Trout Assessment Working Group (WGBAST) ‐ changes in post‐smolt survival in the Baltic Sea and the factors affecting it. Petersson E. Re‐examination of updated biological characteristics datasets from dif‐ ferent stocks utilising Meta analysis. Rogan G. Report on the wild salmon census programme on the Burrishoole River, Ireland. Russell I. Update on salmon biological characteristics data sets available and review of earlier analyses and progress. Todd C. Analyses of long‐term variation in condition factor in relation to ocean cli‐ mate – update on progress. Todd C. & MacLean J. Long‐term variation and changes in age at smoltification in four major Scottish salmon rivers. Todd C. & MacLean J. Among‐river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 2SW “summer” salmon (returning May‐August/September). Todd C. & MacLean J. Among‐river comparisons of mean river age and proportional river age composition in four major Scottish salmon rivers: 1SW grilse (returning April‐August/September).

Data sets - as compiled at both meetings (and subsequently) Amiro P. & Gibson J. Salmon biological characteristics data for the La Have River (wild and hatchery‐origin) (1970‐08) ‐ Canada, Scotia Fundy Region.

142 | ICES SGBICEPS REPORT 2010

Chaput G. Salmon biological characteristics data for Miramichi River (1971‐07) ‐ Canada, Gulf Region. Dannewitz J. Salmon biological characteristics data for the Rivers Ume/Vindel (wild and hatchery‐origin) (1974‐08) and Kalix (1980‐08) – Baltic Sea (Sweden). Davidson I., Aprahamian M. & Russell I. Salmon biological characteristics data for the Rivers Wye (1910‐07), Dee (1937‐07) and Lune (1987‐07) – UK (England & Wales). Dionne M. Salmon biological characteristics data for the Rivers Saint‐Jean (1981‐09) and De La Trinité (1980‐09) ‐ Canada, Quebec Region. Erkinaro J. Salmon biological characteristics data for the Rivers Teno (1972‐07) and Naatamojoki (1975‐06) ‐ Finland/Norway. Euzenat G. Salmon biological characteristics data for the River Bresle (1984‐08) – France. Fiske P. Salmon biological characteristics data for the Rivers Gaula (1989‐08), Årgård (1992‐08), Enningdalselva (1990‐08), Etneelva (1989‐08), Gaula Sogn og Fjordane (1989‐08), Nausta (1989‐08), Numedalslågen (19989‐08), Skienselva (1989‐08) and Vestre Jakobselv (1989‐08) ‐ Norway. Gudbergsson G. Salmon biological characteristics data for Rivers Laxa i Aldaldalur (1974‐08), Hofsa (1971‐08), Ellidaar (1949‐08) and Nordura (1968‐08) ‐ Iceland. Ibbotson A. Salmon biological characteristics data for the River Frome (1968‐08) – UK (England & Wales). Jones R. Salmon biological characteristics data for the St John (Mactaquac) (wild and hatchery‐origin) (1978‐08); and Nashwaak Rivers (1972‐08) ‐ Canada, Scotia Fundy Region. Kennedy R. Salmon biological characteristics data for the River Bush (1973‐07) ‐ UK (N. Ireland). Maclean J. & Smith G. Salmon biological characteristics data for the River N. Esk (1981‐08) – UK (Scotland). Reddin D. Salmon biological characteristics data for Middle Brook (1975‐05) and Western Arm Brook (1971‐06) ‐ Canada, Newfoundland. Reddin D. & Dempson B. Salmon biological characteristics data for Conne River (1986‐06) ‐ Canada, Newfoundland/Labrador. Romakkaniemi A. Salmon biological characteristics data for the River Tornionjoki (1980‐09) – Baltic Sea (Finland/Sweden). Sheehan T. Salmon biological characteristics ‐ standardised spreadsheet design and set up. Trial J. Salmon biological characteristics data for the Penobscot River (1978‐08) ‐ USA. Zubchenko A. Salmon biological characteristics data for the River Tuloma (1983‐08) – Russia.

ICES SGBICEPS REPORT 2010 | 143

Annex 3: References

Amiro, P.G. (1998). The mystery of the missing inner Bay of Fundy salmon. Atlantic Salmon Journal 47: 50‐53.

Andreasson, P.M.R., Martinussen, M.B., Hvidsten, N.A. & Stefansson, S.O. (2001). Feeding and prey selection of wild Atlantic salmon post‐smolts. Journal of Fish Biology 58: 1667‐1679.

Anon. (1994). Run timing of salmon. Report of the Salmon Advisory Committee. London, MAFF: 55pp.

Aprahamian, M.W., Davidson, I.C. & Cove, R.J. (2008). Life history changes in Atlantic salmon from the river Dee, Wales. Hydrobiologia 602: 61‐78.

Armstrong, J.D., Shackley, P.E. & Gardiner, R. (1994). Redistribution of juvenile salmonid fishes after local catastrophic depletion. Journal of Fish Biology 45: 1027‐1039.

Armstrong, J.D., Kemp, P.S., Kennedy, G.J.A., Ladle, M. & Milner, N.J. (2003). Habitat require‐ ments of Atlantic salmon and brown trout in rivers and streams. Fisheries Research 62(2): 143‐170.

Atkins, C.F. (1873). On the salmon of Eastern North America and its culture. Report of the Commission of Fish and Fisheries. pp 226‐337.

Baggerman, B. (1960). Factors in the diadromous migrations of fish. Zoological Society Sympo‐ sium I: 33‐60.

Bagliniere, J.L., Denais, L., Rivot, E., Porcher, J.P., Prevost, E., Marchland, F. & Vauclin, V. (2004). Length and age structure modifications of the Atlantic salmon (Salmo salar) popu‐ lations of Brittany and Lower Normandy from 1972 to 2002. Rennes, Institut National de la Recherche Agronomique and Conseil Superieur de la Peche.

Bardonnet, A. & Bagliniere, J.L. (2000). Freshwater habitat of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 57: 497‐506.

Baum, E.T. & Meister A.L. (1971). Fecundity of Atlantic salmon (Salmo salar) from two Maine rivers. Journal of the Fisheries Research Board of Canada 28(5): 764‐767.

Baum, E.T. (1997). Maine Atlantic Salmon: A national treasure. Atlantic Salmon Unlimited. 224 pp.

Beall, E., Dumas, J., Claireaux, D., Barrière, L. & Marty, C. (1994). Dispersal patterns and sur‐ vival of Atlantic salmon (Salmo salar L.) juveniles in a nursery stream. ICES Journal of Ma‐ rine Science 51: 1‐9.

Beamish, R.J. & Mahnken, C. (2001). A critical size and period hypothesis to explain natural and regulation of salmon abundance and the linkage to climate and climate change. Pro‐ gress in Oceanography 49: 423‐437.

Beamish, R.J., Mahnken, C. & Neville, C.M. (2004). Evidence that reduced early marine growth is associated with lower marine survival of coho salmon. Transactions of the American Fisheries Society 133: 26‐33.

Beaugrand, G. & Reid, F.C. (2003). Long term changes in phytoplankton, zooplankton and salmon related to climate. Global Change Biology 9: 801‐817.

Benoit, H.P. & Swain, D.P. (2008). Impacts of environmental change and direct and indirect harvesting effects on the dynamics of marine fish community. Canadian Journal of Fisher‐ ies and Aquatic Sciences 65(10): 2088‐2104.

Bielak, A.T. & Power, G. (1986). Independence of sea‐age and river‐age in Atlantic salmon (Salmo salar) from Quebec North Shore Rivers. In: Meerburg D.J. (ed) 1986. Salmonid age at maturity. Canadian Special Publications in Fisheries and Aquatic Sciences 89: 70‐78.

144 | ICES SGBICEPS REPORT 2010

Bjornn, T.C. (1971). Trout and salmon movements in two Idaho streams as related to tempera‐ ture, stream flow, cover and population density. Transactions of the American Fisheries Society 100: 423‐438.

Boylan, P. & Adams, C.E. (2006). The influence of broad scale climatic phenomena on long term trends in Atlantic salmon population size: an example from the River Foyle, Ireland. Jour‐ nal of Fish Biology 68: 276‐283.

Brett, J.R. (1979). Environmental factors and growth. In: Fish Physiology, Vol VIII, pp 599‐675. Eds W.S. Hoar, D.J. Randall & J.R. Brett. Academic Press, London.

Buck, R.J.G. & Youngson, A.F. (1982). The downstream migration of precociously mature At‐ lantic salmon, Salmo salar L., parr in autumn; its relation to the spawning migration of mature adult fish. Journal of Fish Biology 20: 279‐288.

Bujold, V., Cunjak, R.A., Dietrich, J.P. & Courtemanche, D.A. (2004). Drifters versus residents: assessing size and age differences in Atlantic salmon (Salmo salar) fry. Canadian Journal of Fisheries and Aquatic Sciences 61: 273‐282.

Byrne, C.J., Poole R., Rogan, G., Dillane, M. & Whelan. K.F. (2003). Temporal and environ‐ mental influences on the variation in Atlantic salmon smolt migration in the Burrishoole system 1970‐2000. Journal of Fish Biology 63:6, 1552‐1564.

Cairns, D.K. (2003). Temperature mortality relations in marine phase Atlantic salmon, (Salmo salar L.): the search for underlying mechanisms. In: Marine mortality of Atlantic salmon: methods and measures (ed. E.C.E. Potter, N. O’Maoileidigh & G. Chaput). Fisheries and Oceans Canada; Canadian Science Advisory Secretariat Research Document 2003/101, 150‐ 160.

Cannaby, H. & Hüsrevoğlu, Y.S. (2009). The influence of low‐frequency variability and long‐ term trends in North Atlantic sea surface temperature on Irish waters. ICES Journal of Marine Science 10.1093/icesjms/fsp062.

Cefas & Environment Agency (2008). Annual assessment of salmon stocks and fisheries in Eng‐ land and Wales 2007: Preliminary assessment prepared for ICES, April 2008 (ISBN 978 0 907545 25 4).

Chaput, G., F. Caron, L. Marshall. (2003). Estimates of survival of Atlantic salmon in the first and second years at sea. In: E.C.E. Potter, N. Ó Maoiléidigh, and G. Chaput (Eds.), p.83‐ 109. Marine mortality of Atlantic salmon, Salmo salar L: methods and measures. DFO Can. Science Adv. Secr. Res. Doc. 2003/101.

Cohen, J. (1969). Statistical power analysis for the behavioural sciences. Academic Press, New York.

Cragg‐Hine, D., Bradley, D.C. & Hendry, K. (2006). Changes in salmon smolt ages in the Welsh River Dee over a 66 year period. Journal of Fish Biology 68: 1891‐1895.

Crisp, D.T. & Carling, P. A. (1989). Observations on siting, dimensions and structure of sal‐ monid redds. Journal of Fish Biology 34: 119‐134.

Crisp, D.T. (1995). Dispersal and growth rate of 0‐group salmon (Salmo salar L.) from point‐ stocking together with some information from scatter‐stocking. Ecology of Freshwater Fish 4: 1‐8.

Crisp, D.T. (1996). Environmental requirements of common riverine European salmonid fish species in fresh water with particular reference to physical and chemical aspects. Hydrobi‐ ologia 323: 201‐221.

Crozier, W.W. & Kennedy, G.J.A. (1993). Marine survival of wild and hatchery reared Atlantic salmon (Salmo salar L.) from the river Bush, Northern Ireland. In: D. Mills (ed.) Salmon in the Sea and New Enhancement Strategies. Oxford: Fishing News Books, Blackwell Scien‐ tific Publications, pp. 139‐162.

ICES SGBICEPS REPORT 2010 | 145

Crozier, W.W. & Kennedy, G.J.A. (1994a). Application of semi‐quantitative electrofishing to juvenile salmonid stock surveys. Journal of Fish Biology 45: 159‐164.

Crozier W.W. & Kennedy G.J.A. (1994b). Marine exploitation of Atlantic salmon (Salmo salar L) from the River Bush, N. Ireland. Fisheries Research 19: 141‐155.

Crozier, W.W. & Kennedy, G.J.A. (1999). Relationships between marine growth and marine survival of one sea winter Atlantic salmon, Salmo salar L., from the River Bush, Northern Ireland. Fisheries Management and Ecology 6: 89‐96.

Crozier, W.W. & Kennedy, G.J.A. (2003). Freshwater influences on marine survival of Atlantic salmon (Salmo salar L.): evidence from the River Bush, Northern Ireland. In: E.C.E. Potter, N. O’Maoileidigh, G. Chaput (eds.) Marine mortality of Atlantic salmon, Salmo salar L.: methods and measures. Research document, Canadian Science Advisory Secretariat 2003/101, pp124‐129.

Crozier, W.W., Potter, E.C.E., Prevost, E., Schon, P.J., & O’Maoiléidigh, N. (2003). A co‐ ordinated approach towards development of a scientific basis for management of wild At‐ lantic salmon in the North‐East Atlantic (SALMODEL). An EU Concerted Action ‐ Quality of Life and Management of Living Resources Key Action 5: Sustainable agriculture, fisher‐ ies and forestry, and integrated development of rural areas including mountain areas. Contract No. QLK5‐CT1999‐01546, 431 pp.

Cunjak, R.A. & Chadwick, E.M.P. (1989). Downstream movements and estuarine residence by Atlantic salmon parr (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 46: 1466‐1471.

Cunjak, R.A., Caissie, D., El‐Jabi, N., Hardie, P., Conlon, J.H., Pollock, T.L., Giberson, D.J. & Komadina‐Douthwright, S. (1993). The Catamaran Brook (New Brunswick) habitat re‐ search project: biological, physical and chemical conditions (1990‐1992). Canadian Techni‐ cal Report of Fisheries and Aquatic Science 1914.

Dadswell, M.J., Spares, A.D., Reader, J.M. & Stokesbury, M.J.W. (2010). The North Atlantic subpolar gyre and the marine migration of Atlantic salmon Salmo salar: the ‘Merry‐Go‐ Round’ hypothesis. Journal of Fish Biology 77: 435‐467.

Dahl, J., Dannewitz, J., Karlsson, L., Petersson, E., Löf, A. & Ragnarsson, B. (2004). The timing of spawning migration: implications of environmental variation, life history, and sex. Ca‐ nadian Journal of Zoology 82: 1864‐1870.

Dalley, E.L., Andrews, C.W. & Green, J.M. (1983). Precocious male Atlantic salmon parr (Salmon salar) in insular Newfoundland. Canadian Journal of Fisheries and Aquatic Sci‐ ences 40: 647‐652.

Davidson, I.C. & Hazelwood, M.S. (2005). Effects of climate change on salmon fisheries. Science Report W2‐047/SR, Environment Agency, Bristol, 52pp.

Davidson, I.C., Hazlewood, M.S. & Cove, R.J. (2006). Predicted growth of juvenile trout and salmon in four rivers in England and Wales based on past and possible future temperature regimes linked to climate change. Proceedings of Biology and Management of Sea Trout Conference, 6‐8 July 2004, Cardiff, Wales.

Davidson, I.C. (2008). Index river monitoring for salmon and sea trout: A progress report on the monitoring programmes on the Tyne, Tamar, Dee and Lune, July 2008. Environment Agency Wales. GEWA 108BOTB‐E‐E.

Dempson, J.B., Myers, R.A. & Reddin, D.G. (1986). Age at first maturity of Atlantic salmon (Salmo salar) ‐ influences of the marine environment. In: Meerburg D.J. (ed) 1986. Sal‐ monid age at maturity. Canadian Special Publication of Fisheries and Aquatic Sciences 89: 79‐89.

Despatie, S.P., Castonguay, M., Chabot, D. & Audet, C. (2001). Final thermal preferendum of Atlantic cod: effect of food ration. Transactions of the American Fisheries Society 130: 263‐ 275.

146 | ICES SGBICEPS REPORT 2010

Dickson, R.R. & Turrell, W.R. (1999). The NAO: the dominant atmospheric process affecting oceanic variability in home, middle and distant waters of European Atlantic salmon. In: The Ocean Life of Atlantic Salmon (ed. D. Mills). Fishing News Books, Oxford, pp. 92‐115.

Dill, P.A. (1977). Development of behaviour of alevins of Atlantic salmon, Salmo salar, and rainbow trout, Salmo gairdneri. Animal Behaviour 25: 116‐121.

Downton, P.R., Reddin, D.G. & Johnson, R.W. (2001). Status of Atlantic salmon (Salmo salar L.) in Campbellton River, Notre Dame Bay (SFA 4), Newfoundland in 2000. Canadian Science Advisory Secretariat Research Document: 73.

Egglishaw, H.J. & Shackley, P.E. (1973). An experiment on faster growth of salmon Salmo salar L. in a Scottish stream. Journal of Fish Biology 5: 197‐204.

Egglishaw, H.J. & Shackley, P.E. (1980). Survival and growth of salmon, Salmo salar, planted in a Scottish stream. Journal of Fish Biology 16: 565‐584.

Einum, S., Thorstad, E.B. & Naesje, T.F. (2002). Growth rate correlations across life stages in female Atlantic salmon. Journal of Fish Biology 60: 780‐784.

Elliott J.M. (1984). Numerical changes and population regulation in young migratory trout, Salmo trutta, in a Lake District stream, 1966‐1983. Journal of Animal Ecology 53: 327‐350.

Elliott J.M. (1991). Tolerance and resistance to thermal stress in juvenile Atlantic salmon, Salmo salar. Freshwater Biology 25: 61‐70.

Elliott, J.M., Hurley, M.A. & Fryer, R.J. (1995). A new, improved growth model for brown trout, Salmo trutta. Functional Ecology 9: 290‐298.

Elliott, J.M. & Hurley, M.A. (1997). A functional model for maximum growth of Atlantic salmon parr, Salmo salar, from two populations in northwest England. Functional Ecology 11: 592‐603.

Elliott, J.M. & Hurley, M.A. (1998). An individual‐based model for predicting the emergence period of sea‐trout fry in a Lake District stream. Journal of Fish Biology 53: 414‐433.

Elliott, J.M., Hurley, M.A. & Maberly, S.C. (2000). The emergence period of sea trout fry in a Lake District stream correlates with the North Atlantic Oscillation. Journal of Fish Biology 56: 208‐210.

Elliott, J.M. (2001). The relative role of density in the stock‐recruitment relationship of sal‐ monids. In: E. Prevost, Chaput, G. (eds.) Stock, recruitment and reference points ‐ Assess‐ ment and management of Atlantic salmon. Hydrobiologie et aquaculture, INRA, Paris, 25‐ 66.

Elo, K., Erkinaro, J., Vuorinen, J.A. & Niemela, E. (1995). Hybridization between Atlantic salmon (Salmo salar) and brown trout (S.trutta) in the Teno and Naatamo River systems, northernmost Europe. Nordic Journal of Freshwater Research 70: 56‐61.

Enfield, D.B., Mestas‐Nunez, A.M. & Trimble, P.J. (2001). The Atlantic multidecadal oscillation and its relationship to rainfall and river flows in the continental US. Geophysical Research Letters 28: 2077‐2080.

Englund, V., Niemela, E., Lansman, M. & Heino, M. (1999). Variations in Atlantic salmon, Salmo salar L., smolt age in tributaries of the River Teno, Finland. Fisheries Management and Ecology 6: 83‐86.

Erkinaro, J., Dempson, J.B., Julkunen, M. & Niemela, E. (1997). Importance of ontogenetic habi‐ tat shifts to juvenile output and life history of Atlantic salmon in a large subarctic river: an approach based on analysis of scale characteristics. Journal of Fish Biology 51: 1174‐1185.

Fairchild, W.L., Brown, S.B. & Moore, A. (2002). Effects of freshwater contaminants on marine survival in Atlantic salmon. NPAFC: 30‐32.

Field‐Dodgson, M.S. (1988). Size characteristics and diet of emergent Chinook salmon in a small, stable, New Zealand stream. Journal of Fish Biology 32: 27‐40.

ICES SGBICEPS REPORT 2010 | 147

Fleming, I.A. (1996). Reproductive strategies of Atlantic salmon: ecology and evolution. Re‐ views in Fish Biology and Fisheries 6: 379‐416.

Friedland, K.D. & Reddin, D.G. (1993). Marine survival of Atlantic salmon from indices of post‐ smolt growth and sea temperature. Salmon in the sea and new enhancement strategies. D. Mills (ed.). Oxford, Fishing News Books. Blackwell Scientific Publications: 119‐138.

Friedland, K.D., Reddin, D.G. & Kocik, J.F. (1993). Marine survival of North American and European Atlantic salmon: effects of growth and environment. ICES Journal of Marine Sci‐ ence 50: 481‐492.

Friedland, K.D. & Haas, R.E. (1996). Marine post‐smolt growth and age at maturity of Atlantic salmon. Journal of Fish Biology 48: 1‐15.

Friedland, K.D., Haas, R.E. & Sheehan, T.S. (1996). Post‐smolt growth, maturation, and survival of two stocks of Atlantic salmon. Fisheries Bulletin 94: 654‐663.

Friedland, K.D. (1998). Ocean climate influences on critical Atlantic salmon (Salmo salar) life history events. Canadian Journal of Fisheries and Aquatic Sciences 55 (supplement 1): 119‐ 130.

Friedland, K.D., Hansen, L.P. & Dunkley, D.A. (1998a). Marine temperature experienced by post‐smolts and the survival of Atlantic salmon, Salmo salar L. in the North Sea area. Fish‐ eries Oceanography 7(1): 22‐34.

Friedland, K.D., Reddin, D.G., Shimizu, N., Haas, R.E. & Youngson, A.F. (1998b). Stron‐ tium:calcium ratios in Atlantic salmon (Salmo salar) otoliths and observations on growth and maturation. Canadian Journal of Fisheries and Aquatic Sciences 55: 1158‐1168.

Friedland, K.D., Hansen, L.P., Dunkley, D.A. & MacLean, J.C. (2000). Linkage between ocean climate, post‐smolt growth and survival of Atlantic salmon (Salmo salar L.) in the North Sea area. ICES Journal of Marine Science 57: 419‐429.

Friedland, K.D., Reddin, D.G., McMenemy, J.R. & Drinkwater, K.F. (2003a). Multidecadal trends in North American Atlantic salmon (Salmo salar) stocks and climate trends relevant to juvenile survival. Canadian Journal of Fisheries and Aquatic Sciences 60: 563‐583.

Friedland, K.D., Reddin, D.G. & Castonguay, M. (2003b). Ocean thermal conditions in the post‐ smolt nursery of North American Atlantic salmon. ICES Journal of Marine Science 60: 343‐ 355.

Friedland, K.D., Chaput, G. & Maclean, J.C. (2005). The emerging role of climate in post‐smolt growth of Atlantic salmon. ICES Journal of Marine Science 62: 1338‐1349.

Friedland, K.D., Clarke, L.M., Dutil, J‐D. & Salminen, M. (2006). The relationship between smolt and postsmolt growth for Atlantic salmon (Salmo salar) in the Gulf of St. Lawrence. Fisheries Bulletin 104: 149‐155.

Friedland, K.D., MacLean, J.C., Hansen, L.P., Peyronnet, A.J., Karlsson, L., Reddin, D.G., Ó Maoiléidigh, N. & McCarthy, J.L. (2009). The recruitment of Atlantic salmon in Europe. ICES Journal of Marine Science 66: 289‐304.

Garcia de Leaniz, C., Fraser, N. & Huntingford, F.A. (2000). Variability in performance in wild Atlantic salmon, Salmo salar L., fry from a single redd. Fisheries Management and Ecology 7: 489‐502.

Gardner, M.L. (1976). A review of factors which may influence the sea‐age and maturation of Atlantic salmon (Salmo salar). Journal of Fish Biology 9: 289‐328.

Gee, A.S. & Milner, N.J. (1980). Analysis of 70‐year catch statistics for Atlantic salmon (Salmo salar) in the River Wye and implications for management of the stocks. Journal of Applied Ecology 17: 41‐57.

George, A.F. (1984). Scottish salmon and 1SW fish return‐migration variations over 200 years. Proceedings of the Institute of Fisheries Management 15th Annual Study Course, Stirling University. Institute of Fisheries Management, Nottingham. A. H. Holden (ed.): 23‐32.

148 | ICES SGBICEPS REPORT 2010

George, A.F. (1991). Climate and the Salmon. Salmon, Trout and Sea‐trout. December 1991.

Gibson, R.J. (1993). The Atlantic salmon in freshwater: spawning, rearing and production. Re‐ views in Fish Biology and Fisheries 3: 39‐73.

Gjerde, B. (1984). Response for individual selection for age at sexual maturity in Atlantic salmon. Aquaculture 38: 229‐240.

Good, S.P., Dodson, J.J., Meekan, M.G. & Ryan, D.A.J. (2001). Annual variation in size‐selective mortality of Atlantic salmon (Salmo salar) fry. Canadian Journal of Fisheries and Aquatic Sciences 58: 1187‐1195.

Gough, P.G., Winstone, A.J. & Hilder, P.G. (1992). Spring salmon. A review of factors affecting spring salmon from the River Wye and elsewhere, and proposals for stock maintenance and enhancement. Cardiff, National Rivers Authority: 57pp.

Gudjonsson, S., Einarsson, S.M., Antonsson, Th. & Gudbergsson, G. (1995). Relation of 1SW fish to salmon ratio to environment changes in several wild stocks of Atlantic salmon (Salmo salar) in Iceland. Canadian Journal of Fisheries and Aquatic Sciences 52: 1385‐1398.

Gurney, W.S.C., Bacon, P.J., Tyldesley, G. & Youngson, A.F. (2008). Process‐based modelling of decadal trends in growth, survival, and smolting of wild salmon (Salmo salar) parr in a Scottish upland stream. Canadian Journal of Fisheries and Aquatic Sciences 65: 2606–2622.

Gustafson‐Marjanen, K.I. & Dowse, H.B. (1983). Seasonal and diel patterns of emergence from the redd of Atlantic salmon (Salmo salar) fry. Canadian Journal of Fisheries and Aquatic Sciences 40(6): 813‐817.

Handeland, S., Järvi, T., Fernö, T. & Stefansson, S. (1996). Osmotic stress, antipredatory behav‐ iour, and mortality of Atlantic salmon (Salmo salar) smolts. Canadian Journal of Fisheries and Aquatic Science. 53(12), 2673–2680.

Handeland, S.O. Björnsson, B.T., Arnesen, A.M. & Stefansson S.O. (2003). Seawater adaptation and growth of post‐smolt Atlantic salmon (Salmo salar) of wild and farmed strains. Aqua‐ culture 220: 367‐384.

Hansen, L.P. & Jonsson B. (1991). Evidence for a genetic component in the seasonal return pat‐ tern of Atlantic salmon, Salmo salar L. Journal of Fish Biology 38: 251‐258.

Hansen, L.P. & Quinn, T.P. (1998). The marine phase of Atlantic salmon (salmo salar) life cycle, with comparison to Pacific salmon. Canadian Journal of Fisheries and Aquatic Sciences 55(supplement 1): 104‐118.

Hansen, L.P., Diserud, O., Fiske, P., Holm, M., Jensen, A.J., Melle, W., Mork, K.A., Friedland, K.D. & Reddin, D.G. (2007). Developments of models to predict marine survival and re‐ turn of salmon to Norway. ICES Working Paper, WP2007/12, 1‐11.

Hansen, M.J. & Nate, N.A. (2005). A method for correcting the relative weight (Wr) index for seasonal patterns in relative condition (Kn) with length as applied to walleye in Wisconsin. North American Journal of Fisheries Management 25: 1256‐1262.

Haugland, M., Holst, J.C., Holm, M. & Hansen, L.P. (2006). Feeding of Atlantic salmon (Salmo salar L.) post‐smolts in the Northeast Atlantic. ICES Journal of Marine Science 63: 1488‐ 1500.

Heddell‐Cowie, M. (2005). Importance of the River Teviot to Atlantic salmon, Salmo salar, rod catches in the River Tweed, Scotland. Fisheries Management and Ecology 12: 137‐142.

Heggberget, T.G. (1988). Timing of spawning in Norwegian Atlantic salmon (Salmo salar). Ca‐ nadian Journal of Fisheries and Aquatic Sciences 45: 845‐849.

Heinimaa, S. & Erkinaro, J. (2004). Characteristics of mature male parr in the northernmost Atlantic salmon populations. Journal of Fish Biology 64: 219‐226.

ICES SGBICEPS REPORT 2010 | 149

Hodbay, A.J. & Boehlert, G.W. (2001). The role of coastal ocean variation in spatial and tempo‐ ral patterns in survival and size of coho salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries and Aquatic Sciences 58: 2021‐2036.

Holliday, N. P., Hughes, S. L. & Beszczynska‐Möller, A. (Eds.) (2009). ICES Report on Ocean Climate 2008. ICES Cooperative Research Report No. 298. 66 pp.

Holm, M., Holst, J.C. & Hansen, L.P. (2000). Spatial and temporal distribution of post‐smolts of Atlantic salmon (Salmo salar L.) in the Norwegian Sea and adjacent areas. ICES Journal of Marine Science 57: 955–964.

Hubley, P.B., Amiro, P.G. & Gibson, A.J.F. (2008). Changes in scale circulus spacings of an en‐ dangered Atlantic salmon Salmo salar population: evidence of a shift in marine migration? Journal of Fish Biology 73: 2321‐2340.

Hughes, S. & Turrell, W.R. (2003). Prospects for improved oceanic conditions. In: Salmon at the edge. D. Mills (ed.), Oxford, Blackwell Science Ltd.: 255‐267.

Hughes, S.L., Holliday, N.P. & Beszczynska‐Möller, A. (eds.) (2008). ICES Report on Ocean Climate (2007). ICES Cooperative Research Report No. 291. 64 pp. ICES, Copenhagen, Denmark.

Hughes, S.L., Holliday, N.P., Colbourne, N.E., Ozhigin, V., Valdimarsson, H., Østerhus, S. & Wiltshire, K. (2009). Comparison of in situ time‐series of temperature with gridded sea surface temperature datasets in the North Atlantic. ICES Journal of Marine Science, 66: 1467‐1479.

Huntingford, F.A., Thorpe, J.E., Garcia de Leaniz, C. & Hay, D.W. (1992). Patterns of growth and smolting in autumn migrants from a Scottish population of Atlantic salmon, Salmo salar L. Journal of Fish Biology 41: 43‐51.

Hurrell, J.W. & Dickson, R.R. (2004). Climate variability over the North Atlantic. In Marine Ecosystems and Climate Variation, N.C. Stenseth et al. (eds). Oxford, U.K.: Oxford Univer‐ sity Press, 15–23.

Hurrell, J.W., (1995). Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science 269: 676‐679.

Hutchings, J.A. & Jones, M.E.B. (1998). Life history variation and growth rate thresholds for maturity in Atlantic salmon, Salmo salar. Canadian Journal of Fisheries and Aquatic Sci‐ ences 55 (supplement 1): 22‐47.

ICES (2006). ICES Report on Ocean Climate 2005. ICES Cooperative Research Report No. 280. 47 pp.

ICES (2007a). Report of the Working Group on North Atlantic salmon (WGNAS), ICES. Inter‐ national Council for the Exploration of the Sea. ICES CM 2007/ACFM:13. 253 pp.

ICES (2007b). Report of the Working Group on Integrated Assessments of the Baltic Sea. ICES CM 2007/BCC:04.

ICES (2008a). Report of the Working Group on North Atlantic salmon (WGNAS), ICES. Inter‐ national Council for the Exploration of the Sea. ICES CM 2008/ACOM:18. 235 pp.

ICES (2008b). Report of the Baltic Salmon and Trout Assessment Working Group (WGBAST). ICES CM 2008/ACOM:05, 267 pp.

ICES (2009a). Report of the Study Group on Biological Characteristics as Predictors of Salmon Abundance (SGBICEPS). ICES CM 2009/DFC:02, 119 pp.

ICES (2009b). Report of the Baltic Salmon and Trout Assessment Working Group 2009 (WGBAST), 24–31 March 2009, Oulu, Finland. ICES CM 2009/ACOM:05, 280 pp.

ICES (2010). Report of the Working Group on North Atlantic salmon (WGNAS), ICES. Interna‐ tional Council for the Exploration of the Sea. ICES CM 2010/ACOM:09. 294 pp.

150 | ICES SGBICEPS REPORT 2010

Imre, I., Grant, J.W.A. & Cunjak R.A. (2005). Density‐dependent growth of young‐of‐the‐year Atlantic salmon Salmo salar in Catamaran Brook, New Brunswick. Journal of Animal Ecology 74: 508‐516.

IPCC (2001) IPCC third assessment report: climate change 2001. Synthesis report (ed. R.T. Wat‐ son and the Core Writing Team). IPPC, Geneva, Switzerland , 184 pp.

Irish National Salmon Fisheries Review 2008. Report to the Minister of Communications, Ma‐ rine and Natural Resources. Prepared by the Marine Institute and Central Fisheries Board.

Jacobsen, J.A. & Hansen, L.P. (2001). Feeding habits of wild and escaped farmed Atlantic salmon, Salmo salar L., in the Northeast Atlantic. ICES Journal of Marine Science 59: 916‐ 933.

Jenkins, T.M.Jr., Diehl, S., Kratz, K.W. & Cooper, S.D. (1999). Effects of population density on individual growth of brown trout in streams. Ecology 80: 941‐956.

Jensen, A.J., Johnsen, B.O. & Saksgård, L. (1989). Temperature requirements in Atlantic salmon (Salmo salar), brown trout (Salmo trutta), and Arctic char (Salvelinus alpinus) from hatch‐ ing to initial feeding compared with geographic distribution. Canadian Journal of Fisher‐ ies and Aquatic Sciences 46: 786‐789.

Jensen, A.J., Johnsen, B.O. & Heggberget, T.G. (1991). Initial feeding time of Atlantic salmon (Salmo salar L.) alevins compared to river flow and water temperature in Norwegian streams. Environmental Biology of Fishes 30: 379‐385.

Jensen, A.J., Zubchenko, A.V., Heggberget, T.G., Hidsten, N.A., Johnsen, B.O., Kuzmin, O., Loenko, A.A., Lund, R.A., Martynov, V.G., Næsje, T.F., Sharov, A.F. & Økland, F. (1999). Cessation of the Norwegian drift net fishery: changes observed in Norwegian and Russian populations of Atlantic salmon. ICES Journal of Marine Science 56: 84‐95.

Jensen, A.J. (2003). Atlantic salmon (Salmo salar) in the regulated river Alta: effects of altered water temperature on parr growth. River research and Applications 19(7): 733‐747.

Jónasson, J., V. Johannsson, and S. Oskarsson. (1994). Ocean mortality of ranched Atlantic salmon during the second year in the sea. Nordic. J. Freshw. Res. 69: 79‐83.

Jónasson, J., Gjerde, B. & Gjedrem, T. (1997). Genetic parameters for return rate and body weight of sea‐ranched Atlantic salmon. Aquaculture 154: 219‐231.

Jones, R.E., Petrell, R.J. & Pauly, D. (1999). Using modified length‐weight relationships to as‐ sess the condition of fish. Aquaculture Management 20: 261‐276.

Jonsson, N., Jonsson, B. & Hansen, L.P. (1990a). Partial segregation in the timing of migration of Atlantic salmon of different ages. Animal Behaviour 40: 313‐321.

Jonsson, B., Jonsson, N. & Hansen, L.P. (1990b). Does juvenile experience affect migration and spawning of adult Atlantic salmon? Behavioural Ecology and Sociobiology 26: 225‐230.

Jonsson, N., Hansen, L.P. & Jonsson, B. (1991a). Variation in age, size and repeat spawning of adult Atlantic salmon in relation to river discharge. Journal of Animal Ecology 60: 937‐947.

Jonsson, N., Jonsson, B. & Hansen, L.P. (1991b). Energetic cost of repeat spawning in male and female Atlantic salmon (Salmo salar L.). Journal of Fish Biology 39: 739‐744.

Jonsson, B. & Jonsson, N. (1993). Partial migration: niche shift versus sexual maturation in fishes. Reviews in Fish Biology and Fisheries 3: 348‐365.

Jonsson, N., Jonsson, B. & Fleming, I.A. (1996). Does early growth cause a phenotypically plas‐ tic response in egg production of Atlantic salmon? Functional Ecology 10: 89‐96.

Jonsson, N., Jonsson, B. & Hansen, L.P. (1997). Changes in proximate composition and esti‐ mates of energetic costs during upstream migration and spawning in Atlantic salmon Salmo salar. Journal of Animal Ecology 66: 425‐436.

Jonsson, B., Forseth, T., Jensen, A.J. & Næsje, T.F. (2001). Thermal performance of juvenile At‐ lantic salmon, Salmo salar L. Functional Ecology 15: 701‐711.

ICES SGBICEPS REPORT 2010 | 151

Jonsson, N., Jonsson, B. & Hansen, L.P. (2003). The marine survival and growth of wild and hatchery‐reared Atlantic salmon. Journal of Applied Ecology 40: 900‐911.

Jonsson, N. & Jonsson, B. (2004). Size and age at maturity of Atlantic salmon correlate with the North Atlantic Oscillation Index (NAOI). Journal of Fish Biology 64: 241‐247.

Jonsson, N., Jonsson, B. & Hansen, L.P. (2005). Does climate during embryonic development influence parr growth and age of seaward migration in Atlantic salmon (Salmo salar)? Ca‐ nadian journal of Fisheries and Aquatic Sciences 62: 2502‐2508.

Jonsson, N. & Jonsson, B. (2007). Sea growth, smolt age at sexual maturation in Atlantic salmon. Journal of Fish Biology 71: 245‐252.

Jonsson, B. & Jonsson, N. (2009). A review of the likely effects of climate change on anadro‐ mous Atlantic salmon Salmo salar and brown trout Salmo trutta, with particular reference to water temperature and flow. Journal of Fish Biology 75: 2381‐2447.

Juanes, F., Gephard, S. & Beland, K.F. (2004). Long‐term changes in migration timing of adult Atlantic salmon (Salmo salar) at the southern edge of the species distribution. Canadian Journal of Fisheries and Aquatic Sciences 61: 2392‐2400.

Jutila, E., Jokikokko, E. & Julkunen, M. (2005). The smolt run and post‐smolt survival of Atlan‐ tic salmon, Salmo salar L., in relation to early summer water temperatures in the northern Baltic Sea. Ecology of Freshwater Fish 14: 69‐78.

Jutila, E., Jokikokko, E. & Julkunen, M. (2006). Long‐term changes in the smolt size and age of Atlantic salmon, Salmo salar L., in a northern Baltic river related to parr density, growth opportunity and post‐smolt survival. Ecology of Freshwater Fish 15: 321‐330.

Kallio‐Nyberg, I., Jutila, E., Saloniemi, I. & Jokikokko, E. (2004). Association between environ‐ mental factors, smolt size and the survival of wild and reared Atlantic salmon from the Simojoki River in the Baltic Sea. Journal of Fish Biology 65: 122–134.

Kallio‐Nyberg, I., Jutila, E. Jokikokko E. & Saloniemi, I. (2006). Survival of reared Atlantic salmon and sea trout in relation to marine conditions of smolt year in the Baltic Sea. Fish‐ eries Research 80: 295–304.

Karlsson, L. & Karlström, O. (1994). The Baltic salmon (Salmo salar L.): its history, present situation and future. Dana 10: 61‐85.

Kennedy, G.J.A. & Strange, C.D. (1986). The effects of intra‐specific competition on the distri‐ bution of stocked juvenile Atlantic salmon, Salmo salar L., in relation to depth and gradi‐ ents in an upland trout, Salmo trutta L., stream. Journal of Fish Biology 29: 199‐214.

Kennedy R.J. & Crozier W.W. (2010). Evidence of changing migratory patterns of wild Atlantic salmon (Salmo salar L.) smolts in the River Bush, Northern Ireland, and possible associa‐ tions with climate change. Journal of Fish Biology 76: 1786‐1805.

Kendall, M.G. (1975). Rank correlation methods. Fourth edition, Charles Griffith: London.

King, H.R., Pankhurst, N.W., Watts, M. & Pankhurst, P.M. (2003). Effect of elevated summer temperature on gonadal steroid production, vitellogenesis and egg quality in Tasmanian female Atlantic salmon. Journal of Fish Biology 63: 153‐167.

King, H.R., Pankhurst, N.W. & Watts, M. (2007). Reproductive sensitivity to elevated water temperatures in female Atlantic salmon is heightened at certain stages of vitellogenesis. Journal of Fish Biology 70: 190‐205.

Klemetsen, A., Amundsen, P‐A., Dempson, J.B., Jonsson, B., Jonsson, N., OʹConnell, M.F. & Mortensen, E. (2003). Atlantic salmon Salmo salar L., brown trout Salmo trutta L. and Arc‐ tic charr Salvelinus alpinus L.: a review of aspects of their life histories. Ecology of Fresh‐ water Fish 12: 1‐59.

LʹAbée‐Lund, J.H. (1989). Significance of mature male parr in a small population of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 48: 1015‐1021.

152 | ICES SGBICEPS REPORT 2010

LʹAbée‐Lund, J.H., Vøllestad, L.A. & Beldring, S. (2004). Spatial and temporal variation in the grilse proportion of Atlantic salmon in Norwegian rivers. Transactions of the American Fisheries Society 133: 743‐761.

Lacroix, G.L., & Knox, D. (2005). Distribution of Atlantic salmon (Salmo salar) postsmolts of different origin in the Bay of Fundy and Gulf of Maine and evaluation of factors affecting migration, growth, and survival. Canadian Journal of Fisheries and Aquatic Sciences 62: 1363–1376.

Lapointe, M.F., Bergeron, N.E., Bérubé, F., Pouliot, M.A. & Johnston, P. (2004). Interactive ef‐ fects of substrate sand and silt contents, redd‐scale hydraulic gradients, and interstitial ve‐ locities on egg‐to‐emergence survival of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 61: 2271‐2277.

Letcher, B.H. & Terrick, T.D. (1998). Maturation of male age‐0 Atlantic salmon following a massive, localized flood. Journal of Fish Biology 53(6): 1243‐1252.

Lobón‐Cerviá, J. (2005). Spatial and temporal variation in the influence of density dependence on growth of stream‐living brown trout (Salmo trutta). Canadian Journal of Fisheries and Aquatic Sciences 62: 1231‐1242

Lower, N. & Moore, A. (2007). The impact of a brominated flame retardant on smoltification and olfactory function in Atlantic salmon smolts. Marine and Freshwater Behaviour and Physiology 40: 267‐284.

Lundqvist, H., Clarke, W.C. & Johansson, H. (1988). The influence of sexual maturation on sur‐ vival to adulthood of river stocked Baltic salmon, Salmo salar, smolts. Holarctic Ecology 11: 60‐69.

Lundqvist, H., McKinnell, S., Fangstam, I. & Berglund, I. (1994). The effect of time, size and sex on recapture rates and yield after river releases of Salmo salar smolts. Aquaculture 121: 245‐257.

MacCrimmon, H.R. & Gots, B.L. (1979). World distribution of Atlantic salmon, Salmo salar. Journal of the Fisheries Research Board of Canada 36: 422‐457.

Manly B.F.J. (2001). Randomization, bootstrap and Monte Carlo methods in biology. Chapman and Hall/CRC, London, 399 pp.

Mann, H.B. (1945). Non‐parametric tests against trend. Econometrica 13: 245‐259.

Marshall, C.T., Needle, C.L., Yaragina, N.A., Ajiad, A.M. & Gusev, E. (2004). Deriving condi‐ tion indices from standard fisheries databases and evaluating their sensitivity to variation in stored energy reserves. Canadian Journal of Fisheries and Aquatic Sciences 61: 1900‐ 1917.

Marshall, E.A., Quinn, T.P., Roff, D.A., Hutchings, J.A., Metcalfe, N.B., Bakke, T.A., Saunders, R.L. & LeRoy Poff, N. (1998). A framework for understanding Atlantic salmon (Salmo salar) life history. Canadian Journal of Fisheries and Aquatic Sciences 55 (supplement 1): 48‐58.

Martin, J.H.A. & Mitchell, K.A. (1985). Influence of sea temperature upon the numbers of grilse and multi‐sea‐winter Atlantic salmon (Salmo salar) caught in the vicinity of the river Dee (Aberdeenshire). Canadian Journal of Fisheries and Aquatic Sciences 42: 1513‐1521.

Martinez J.L., Moran P., Perez J., De Gaudemar B., Beall E., & Garcia‐Vazquez E. (2000). Multi‐ ple paternity increases effective size of southern Atlantic salmon populations. Molecular Ecology 9: 293–298.

Marty, C. & Beall, E. (1989). Modalities spatio‐temporalles de la dispersion d’alevins de sau‐ mon atlantique (Salmo salar L.) a l’emergence. Revue des Sciences de l’eau 2 : 831‐846.

Mason, J.C. (1976). Response of underyearling coho salmon to supplemental feeding in a natu‐ ral stream. Journal of Wildlife Management 40: 775‐788.

ICES SGBICEPS REPORT 2010 | 153

McCarthy, J.L., Friedland, K.D. & Hansen, L.P. (2008). Monthly indices of the post‐smolt growth of Atlantic salmon from the Drammen river, Norway. Journal of Fish Biology 72: 1572‐1588.

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 Sci‐ ences 55 (supplement 1): 77‐92.

Metcalfe, N.B., Huntingford, F.A., Graham W.D. & Thorpe J.E. (1989). Early social status and the development of life‐history strategies in Atlantic salmon. Proceedings of the Royal So‐ ciety London B 236: 7‐19.

Metcalfe, N.B. & Thorpe, J.E. (1990). Determinants of geographic variation in the age of sea‐ ward‐migrating salmon, Salmo salar. Journal of Animal Ecology 59: 135‐145.

Metcalfe, N.B. & Thorpe, J.E. (1992). Early predictors of life‐history events: the link between first feeding date, dominance and seawater migration in Atlantic salmon, Salmo salar. Journal of Fish Biology 41(supplement B): 93‐99.

Metcalfe, N.B., Taylor, A.C. & Thorpe, J.E. (1995). Metabolic rate, social status and life‐history strategies in Atlantic salmon. Animal Behaviour 49: 431‐436.

Mikheev, V.N., Metcalfe, N.B., Huntingford, F.A. & Thorpe J.E. (1994). Size‐related differences in behaviour and spatial distribution of juvenile Atlantic salmon in a novel environment. Journal of Fish Biology 45: 379‐386..

Moffett, I.J.J., Allen, M., Flanagan, C.F., Crozier W.W. & Kennedy, G.J.A. (2006). Fecundity, egg size and early hatchery survival for wild Atlantic salmon, from the River Bush. Fisheries Management and Ecology 13: 73‐79.

Moore, A., Cotter, D., Rogan, G., Quayle, V., Lower, N. & Privitera, L. (2008). The impact of a pesticide on the physiology and behaviour of hatchery reared salmon smolts during the transition from the freshwater to marine environment. Fisheries Management and Ecology 15: 339‐345.

Moore, A., Lower, N., Mayer, I. & Greenwood, L. (2007). The impact of a pesticide on migra‐ tory activity and olfactory function in Atlantic salmon (Salmo salar L.) smolts. Aquacul‐ ture 273: 350‐359.

Moore, A. & Riley, W.D. (1992). A sexually mature female Atlantic salmon, Salmo salar L., smolt. Aquaculture and Fisheries Management 23: 273‐274.

Moore, A., Scott, A.P., Lower, N., Katsiadaki, I. & Greenwood, L. (2003). The effects of 4‐ nonylphenol and atrazine on Atlantic salmon (Salmo salar L.) smolts. Aquaculture 222: 253‐263.

Moore, A. & Waring, C.P. (1996). Electrophysiological and endocrinological evidence that F‐ series prostaglandins function as priming pheromones in mature male Atlantic salmon parr. Journal of Experimental Biology 199: 2307‐2316.

Moore, D.S., Chaput, G.L. & Pickard, P.R. (1995). The effect of fisheries on the biological char‐ acteristics and survival of mature Atlantic salmon (Salmo salar) from the Miramichi river. Water, Science and the Public: The Miramichi Ecosystem. E.M.P. Chadwick (ed.), Special Publications of Canadian Journal of Fisheries and Aquatic Sciences 123: 229‐247.

Myers, R.A. (1984). Demographic consequences of precocious maturation pf Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 41: 1349‐1353.

Myers, R.A. & Hutchings, J.A. (1987). Mating of anadromous Atlantic salmon, Salmo salar L., with mature male parr. Journal of Fish Biology 31: 143‐146.

Nagata, M. & Irvine, J.R. (1997). Differential dispersal patterns of male and female masu salmon fry. Journal of Fish Biology 51: 601‐606.

154 | ICES SGBICEPS REPORT 2010

Neff, B.D. & Cargnelli, L.M. (2004). Relationships between condition factors, parasite load and paternity in bluegill sunfish, Lepomis macrochirus. Environmental Biology of Fishes 71: 297‐304.

Nicieza, A.G. & Braña, F. (1993). Relationships among smolt size, marine growth, and sea‐age at maturity of Atlantic salmon (Salmo salar) in northern Spain. Canadian Journal of Fisher‐ ies and Aquatic Sciences 50: 1632‐1640.

Niemelä, E., Erkinaro, J., Dempson, J.B., Julkunen, M., Zubchenko, A., Prusov, S., Svenning, M.A., Ingvaldsen, R., Holm, M. & Hassinen, E. (2004). Temporal synchrony and variation in abundance of Atlantic salmon (Salmo salar) in two subarctic Barents Sea rivers: influ‐ ence of oceanic conditions. Canadian Journal of Fisheries and Aquatic Sciences 61: 2384‐ 2391.

Niemelä, E., Erkinaro, J., Julkunen, M., Hassinen, E., Länsman, M. & Brørs, S. (2006). Temporal variation in abundance, return rate and life histories of previously spawned Atlantic salmon in a large subatrctic river. Journal of Fish Biology 68: 1222‐1240.

O’Connell, M.F. & Ash, E.G.M. (1993). Smolt size in relation to age at first maturity of Atlantic salmon (Salmo salar): the role of lacustrine habitat. Journal of Fish Biology 42: 551‐569.

Økland, F., Jonsson, D., Jensen A.J. & Hansen L.P. (1993). Is there a threshold size regulating seaward migration of brown trout and Atlantic salmon? Journal of Fish Biology 42: 541‐ 550.

O’Maoiléidigh N., Potter, E.C.E., McGinnity, P., Crozier, W.W., Hansen, L.P., Gudbergsson,G., Prevost, E., Karlsson, L. and Maclean, J. (2003) Examination of marine survival datasets for Atlantic salmon (Salmo salar, L.) in the north‐east Atlantic area. In: Marine mortality of At‐ lantic salmon, Salmo salar, L: methods and measures (ed. E.C.E. Potter, N. O’Maoiléidigh and G. Chaput). Fisheries and Oceans Canada; Canadian Science Advisory Secretariat. Re‐ search Document 2003/101, pp

O’Maoiléidigh, N., Browne, J., McDermott, T., Cullen, A., Bond, N., O’Farrell, M. and Rogan, G. (1996). Marine survival and exploitation of Irish salmon stock. In “Irish Marine sci‐ ence 1995”. Eds – Keegan, B.F. and O’Connor R., Galway University Press Ltd.

Parrish, D.L., Behnke, R.J., Gephard, S.R., McCormick, S.D. & Reeves, G.H. (1998). Why arenʹt there more Atlantic salmon (Salmo salar)? Canadian Journal of Fisheries and Aquatic Sci‐ ences 55 (supplement 1): 281‐287.

Parsons, L.S. &, Lear, W.H. (2001). Climate variability and marine ecosystems impacts. A north Atlantic perspective. Progress in Oceanography 49: 167‐188.

Pauly, D. (1981). The relationship between gill surface area and growth performance in fish: a generalization of von Bertalanffyʹs theory of growth. Journal du Conseil‐Conseil perma‐ nent pour lʹExploration de la Mer 39: 251‐282.

Peyronnet, A., Friedland, K.D., O’Maoiléidigh, N., Manning, M. & Poole, W.R. (2007). Links between patterns of marine growth and survival of Atlantic salmon Salmo salar, L. Journal of Fish Biology 71(3): 684‐700.

Peyronnet, A., Friedland, K.D. & O’Maoiléidigh, N. (2008). Different ocean and climate factors control the marine survival of wild and hatchery Atlantic salmon Salmo salar in the north‐ east Atlantic Ocean. Journal of Fish biology 73: 945‐962.

Pinder, A.C., Riley, W.D., Ibbotson, A.T. & Beaumont, W.R.C. (2007). Evidence for an autumn seaward migration and the subsequent estuarine residence of 0+ juvenile Atlantic salmon Salmo salar L., in England. Journal of Fish Biology 71: 260‐264.

Potter, E.C.E. & Crozier, W.W. (2000). A perspective on the marine survival of Atlantic salmon. In: The Ocean life of Atlantic salmon: environmental and biological factors influencing survival. D. Mills (ed.), Oxford, Fishing News Books, Blackwell Science: 19‐36.

Power, G. (1969). The salmon of Ungava Bay. Institute of North America: 72.

ICES SGBICEPS REPORT 2010 | 155

Prevost, E., Chadwick, E.M.P. & Claytor, R.R. (1992). Influence of size, winter duration and density on sexual maturation of Atlantic salmon (Salmo salar) juveniles in Little Codroy River (southwest Newfoundland). Journal of Fish Biology 41: 1013‐1019.

Quinn, T.P. (2005). The behaviour and ecology of Pacific salmon and trout. University of Wash‐ ington Press, Seattle.

Quinn, T.P., McGinnity, P. & Cross, T.F. (2006). Long‐term declines in body size and shifts in run timing of Atlantic salmon in Ireland. Journal of Fish Biology 68: 1713‐1730.

R Development Core Team (2007). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3‐900051‐07‐0, URL http://www.R‐project.org.

Randall, R.G. (1989). Effect of sea‐age on the reproductive potential of Atlantic salmon (Salmo salar) in eastern Canada. Canadian Journal Fisheries and Aquatic Sciences 46: 221‐2218.

Rayner, N.A., Parker, D.E., Horton, E.B., Folland, C.K., Alexander, L.V., Rowell, D.P., Kent E.C. & Kaplan, A. (2003). Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late Nineteenth Century. Journal of Geophysical Research 108, (D14), 4407. doi:10.1029/2002JD002670

Reddin, D.G., Shearer, W.M. & Burfitt, R.F. (1984). Intercontinental migrations of Atlantic salmon (Salmo salar L.). ICES CM 1984/M11.

Ricker, W.E. (1975). Computation and Interpretation of Biological Statistics of Fish Populations. Bull. Fish. Res. Bd Can. 191.

Ricker, W.E. (1976). Review of the rate of growth and mortality of Pacific salmon in salt water, and noncatch mortality caused by fishing. J. Fish. Res. Bd Can. 33: 1483‐1524.

Riddell, B.E. & Leggett, W.C. (1981). Evidence of an adaptive basis for geographic variation of body morphology, and time of downstream migration of Juvenile Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences. 38: 308‐320.

Riley, S.C. & Power, G. (1987). Age at maturity in landlocked and anadromous Atlantic salmon parr from two Quebec rivers. Environmental Biology of Fishes 19: 223‐228.

Riley, W.D., Eagle, M.O. & Ives, S.J. (2002). The onset of downstream movement of juvenile Atlantic salmon, Salmo salar L., in a chalk stream. Fisheries Management and Ecology 9: 87‐94.

Riley, W.D. (2007). Seasonal downstream movements of juvenile Atlantic salmon, Salmo salar L., with evidence of solitary emigration of smolts. Aquaculture 273: 194‐199.

Riley, W.D., Ibbotson, A.T., Lower, N., Cook, A.C., Moore, A., Mizuno, S., Pinder, A.C., Beau‐ mont, W.R.C. & Privitera, L. (2008). Physiological seawater adaptation in juvenile Atlantic salmon, Salmo salar L. autumn migrants. Freshwater Biology 53: 745‐755.

Riley, W.D., Ibbotson, A.T. & Beaumont, W.R.C. (2009). Adult returns from Atlantic salmon, Salmo salar L., parr autumn migrants. Fisheries Management and Ecology 16: 75‐76.

Rimmer, D.M., Paim, U. & Saunders, R.L. (1983). Autumnal habitat shift of juvenile Atlantic salmon (Salmo salar) in a small river. Canadian Journal of Fisheries and Aquatic Sciences 40: 671‐680.

Rimmer, D.M., Paim, U. & Saunders, R.L. (1984). Changes in the selection of microhabitat by juvenile Atlantic salmon (Salmo salar) at the summer‐autumn transition in a small river. Canadian Journal of Fisheries and Aquatic Sciences 41: 469‐475.

Ritter, J.A. (1993). Changes in Atlantic salmon (Salmo salar) harvest and stock status in the North Atlantic. In: Salmon in the sea and new enhancement strategies, Mills, D.H. (ed.), Fishing News Books, Oxford, 3‐25.

156 | ICES SGBICEPS REPORT 2010

Rivot, E., Perrier, C., Jouanin, C., Roussel, J‐M., Riera, P. & Bagliniere, J‐L. (Poster) Climate change may affect growth and life history in Atlantic salmon juveniles over the last 30 years.

Roff, D.A. (1992). The evolution of life history. Chapman & Hall, New York & London.

Rogers, J.C. (1984). The Association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Monthly Weather Review 112: 1999–2015.

Rosenthal, R. (1979). The “file drawer problem” and tolerance for null results. Psychological Bulletin 86: 638‐641.

Rowe, D.K. & Thorpe, J.E. (1990). Suppression of maturation in male Atlantic salmon (Salmo salar L.) parr by reduction in feeding and growth during spring months. Aquaculture 86: 291‐313.

Rowe, D.K., Thorpe, J.E. & Shanks, A.M. (1991). Role of fat stores in maturation of male Atlan‐ tic salmon (Salmo salar) parr. Canadian Journal of Fisheries and Aquatic Sciences 48: 405‐ 413.

Russell, I.C., Ives, M.J., Potter, E.C.E., Buckley, A.A. & Duckett L. (1995) Salmon and migratory trout statistics for England and Wales, 1951‐1990. MAFF Directorate of Fisheries Research, Data Report No. 38, 252 pp.

Salminen, M. (1997). Relationships between smolt size and sea‐age at maturity in Atlantic salmon ranched in the Baltic Sea. Journal of Applied Ichthyology 13: 121‐130.

Salminen, M., Kuikka, S. & Erkamo, E. (1994). Divergence in the feeding migration of Baltic salmon (Salmo salar L.); the significance of smolt size. Nordic Journal of Freshwater Re‐ search 69: 32‐42.

Salminen, M, Kuikka, S. and Erkamo, E. (1995). Annual variability in survival of sea‐ranched Baltic salmon, Salmo salar L.: significance of smolt size and marine conditions. Fisheries Management and Ecology: 2, 171‐184.

Saunders, R.L. & Schom, C.B. (1985). Importance of the variation in life history parameters of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 42: 615‐ 618.

Scarnecchia, D.L. (1983). Age at sexual maturity in Icelandic stocks of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 40: 1456‐1468.

Scarnecchia, D.L., Isaksson, A. & White, S.E. (1989). Effects of oceanic variations and the West Greenland fishery on age at maturity of Icelandic west coast stocks of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 46(1): 16‐27.

Scarnecchia, D.L., Isaksson, A. & White, S.E. (1991). Effects of the Faroese long line fishery, other oceanic fisheries and oceanic variations on age at maturity of Icelandic north coast stocks of Atlantic salmon (Salmo salar). Fisheries Research 10: 207‐228.

Schaffer, W.M. & Elson, P.F. (1975). The adaptive significance of variation in life history among local populations of Atlantic salmon in North America. Ecology 56: 577‐590.

Shearer, W.M. (1990). The Atlantic salmon (Salmo salar L.) of the North Esk with particular reference to the relationship between both river and sea age and time of return to home waters. Fisheries Research, 10: 93‐103.

Simpson, A.L. (1992). Differences in body size and lipid reserves between maturing and non‐ maturing Atlantic salmon parr, Salmo salar L. Canadian Journal of Zoology 70: 1737‐1742.

Skilbrei, O.T. (1989). Relationship between smolt length and growth and maturation in the sea of individually tagged Atlantic salmon (Salmo salar). Aquaculture 83: 95‐108.

Solomon, D.J. (1978). Migration of smolts of Atlantic salmon (Salmo salar L.) and sea trout (Salmo trutta L.) in a chalkstream. Environmental Biology of Fishes 3: 223‐229.

ICES SGBICEPS REPORT 2010 | 157

Solomon, D.J. & Lightfoot, G. (2007). Climate change and chalkstream salmon. Environment Agency, South West Region report. South Wessex Area, Environment Agency. V 1.5: 23pp.

Staurnes, M., Sigholt, T., Asgard, T. & Baeverfjord, G. (2001). Effects of a temperature shift on seawater challenge test performance in Atlantic salmon (Salmo salar) smolts. Aquaculture 201: 153‐159.

Stearns, S.C. (1992). The evolution of life histories. Oxford, Oxford University Press.

Stewart, D.C., Smith, G.W. & Youngson, A.F. (2002). Tributary‐specific variation in timing of return of Atlantic salmon (Salmo salar) to fresh water has a genetic component. Canadian Journal of Fisheries and Aquatic Sciences 59: 276‐281.

Stewart, D.C, Middlemas, S.J. & Youngson A.F. (2006). Population structuring in Atlantic salmon (Salmo salar): evidence of genetic influence on the timing of smolt migration in sub‐catchment stocks. Ecology of Freshwater Fish 15: 552‐558.

Summers, D.W. (1995). Long‐term changes in the sea‐age at maturity and seasonal time of re‐ turn of salmon, Salmo salar L., to Scottish rivers. Fisheries Management & Ecology 2: 147‐ 156.

Sutton, S.G., Bult, T.P. & Haedrich, R.L. (2000). Relationships among fat weight, body weight, water weight, and condition factors in wild Atlantic salmon parr. Transactions of the American Fisheries Society 129: 527‐538.

Sutton, R.T. & Hodson, D.L.R. (2005). Atlantic forcing of North American and European sum‐ mer climate. Science 309: 115‐118.

Swansburg, E., Chaput, G., Moore, D., Caissie, D. & El‐Jabi, N. (2002). Size variability of juve‐ nile Atlantic salmon: links to environmental conditions. Journal of fish Biology 61: 661‐683.

Taranger, G.L. & Hansen, T. (1993). Ovulation and egg survival following exposure of Atlantic salmon, Salmo salar L., broodstock to different water temperatures. Aquaculture and Fish‐ eries Management 24: 151‐156.

Thorpe, J.E., Morgan, R.I.G., Talbot, C. & Miles, M.S. (1983). Inheritance of developmental rates in Atlantic salmon, Salmo salar L. Aquaculture 33:119‐128.

Thorpe, J.E., Miles, M.S. & Keay, D.S. (1984). Developmental rate, fecundity and egg size in Atlantic salmon, Salmo salar L. Aquaculture 43: 289‐305.

Thorpe, J.E. (1989). Developmental variation in salmonid populations. Journal of Fish Biology 35 (supplement A): 295‐303.

Thorpe, J.E. (1990). Variation in life‐history strategies in salmonids. Polskie Archiwum Hydro‐ biologii 37(1‐2): 3‐12.

Thorpe, J.E. (1994). Reproductive strategies in Atlantic salmon (Salmo salar L.). Aquaculture & Fisheries Management 25: 77‐87.

Thorpe, J.E. (2007). Maturation responses of salmonids to changing developmental opportuni‐ ties. Marine Ecology Progress Series 335: 285‐288.

Todd, C.D., Hughes, S.L., Marshall, C.T., MacLean, J.C., Lonergan, M.E. & Biuw, E.M. (2008). Detrimental effects of recent ocean surface warming on growth condition of Atlantic salmon. Global Change Biology 14: 958‐970.

Turrell, W.R. & Shelton, R.G.J. (1993). Climatic change in the north‐eastern Atlantic and its im‐ pacts on salmon stocks. In: Salmon in the sea and new enhancement strategies, Mills, D.H. (ed.), Fishing News Books, Oxford, 40‐78.

UK Climate Impacts Programme (2002). Climate change scenarios for the UK: The UKCIPO2 Briefing Report, April 2002.

Vøllestad, L.A., Hirst, D., LʹAbée‐Lund, J.H., Armstrong, J.D., MacLean, J.C., Youngson, A.F. & Stenseth, N.C. (2009). Divergent trends in anadromous salmonid populations in Norwe‐

158 | ICES SGBICEPS REPORT 2010

gian and Scottish rivers. Proceedings of the Royal Society B‐Biological Sciences, 276: 1021‐ 1027.

Watts, M., Pankhurst, N.W. & King, H.R. (2004). Maintenance of Atlantic salmon, (Salmo salar) at elevated temperatures inhibits cytochrome P450 aromatase activity in isolated ovarian follicles. General and Comparative Endocrinology 135: 381‐390.

Webb, J.H. & McLay, H.A. (1996). Variation in the time of spawning of Atlantic salmon (Salmo salar) and its relationship to temperature in the Aberdeenshire Dee, Scotland. Canadian Journal of Fisheries and Aquatic Sciences 53: 2739‐2744.

Webb, J.H. & Campbell, R.N.B. (2000). Patterns of run timing in adult Atlantic salmon return‐ ing to Scottish rivers ‐ some new perspectives and management implications. In: Manag‐ ing wild Atlantic salmon: New challenges ‐ New techniques. F.G.J. Whoriskey & Whelan, K.E. (eds.), The Atlantic Salmon Trust, 100‐138.

Webb, J.H., Fryer, R.J., Taggart, J.B., Thompson, C.E. & Youngson, A.F. (2001). Dispersion of Atlantic salmon (Salmo salar) fry from competing families as revealed by DNA profiling. Canadian Journal of Fisheries and Aquatic Sciences 58: 2386‐2395.

Wells, B.K., Grimes, C.B., Sneva, J.G., McPherson, S. & Waldvogel, J.B. (2008). Relationships between oceanic conditions and growth of Chinook salmon (Oncorhynchus tshawytscha) from California, Washington and Alaska, USA. Fisheries Oceanography 17(2): 101‐125.

Welton, J.S., Beaumont, R.C. & Ladle, M. (1999). Timing of migration and changes in age struc‐ ture of Atlantic salmon, Salmo salar L., in the River Frome, a Dorset chalk stream, over a 24‐year period. Fisheries Management and Ecology 6: 437‐458.

Whalen, K.G. & Parrish, D.L. (1999). Effect of maturation on parr growth and smolt recruitment of Atlantic salmon. Canadian Journal of Fisheries and Aquatic Sciences 56: 79‐86.

Wilkie, N.F. (2006). Phenotypic and genetic variation within and among seven populations (six endangered) of Maine Atlantic salmon (Salmo salar). University of Maine MS Thesis, 65pp.

Wootton, R.J. (1998). Ecology of Teleost Fish. Second edition, Kluwer, London.

Youngson, A.F., Buck, R.J.G., Simpson, T.H. & Hay, D.W. (1983). The autumn and spring emi‐ grations of juvenile Atlantic salmon, Salmo salar L., from the Girnock Burn, Aberdeen‐ shire, Scotland: environmental release of migration. Journal of Fish Biology 23: 625‐639.

Youngson, A.F., Jordan, W.C. & Hay, D.W. (1994). Homing of Atlantic salmon (Salmo salar L.) to a tributary spawning stream in a major river catchment. Aquaculture 121: 259‐267.

Youngson, A.F., Malcolm, I.A., Thorley, J.L., Bacon, P.J. & Soulsby, C. (2004). Long‐residence groundwater effects on incubating salmonid eggs: low hyporheic oxygen impairs embryo development. Canadian Journal of Fisheries and Aquatic Sciences 61: 2278‐2287.

Zydlewski, G.B., Haro, A. & McCormick, S.D. (2005). Evidence for cumulative temperature as an initiating and terminating factor in downstream migratory behavior of Atlantic salmon (Salmo salar) smolts. Canadian Journal of Fisheries and Aquatic Sciences 62: 68‐78.