Polar Biology

, Volume 35, Issue 4, pp 593–609

Changes in Northwest Atlantic Arctic and Subarctic conditions and the growth response of Atlantic salmon

Authors

    • National Marine Fisheries Service
  • Christopher D. Todd
    • Scottish Oceans InstituteUniversity of St Andrews
Original Paper

DOI: 10.1007/s00300-011-1105-z

Cite this article as:
Friedland, K.D. & Todd, C.D. Polar Biol (2012) 35: 593. doi:10.1007/s00300-011-1105-z

Abstract

There has been a systematic change in the weight at age of Atlantic salmon (Salmo salar L.) in the Northwest Atlantic that is related to climate variability. This relationship emerged from analyses of broad-scale measures of ocean surface thermal habitat, which show that expansion of the area bounding 4–8°C is associated with greater growth. To further elucidate the effect of the environment on salmon growth, time series of sea surface temperature (SST), sea ice coverage, chlorophyll concentration, net primary production and zooplankton abundance were examined temporally and spatially in relation to changes in the weight of salmon. SST and zooplankton data were extracted from in situ analyses, whereas sea ice and chlorophyll-based measures of productivity were collected with satellite sensors. Salmon growth was found to be unrelated to productivity at the base of the food chain but highly associated with thermal regime during winter and spring. Warming conditions during specific segments of the salmon life cycle have been associated with poor adult recruitment; yet, warming during others is beneficial to salmon growth and is assumed to increase reproductive output of spawning fish. Despite these positive influences, climate change will continue to erode the viability of salmon populations while the negative effects of warming on survivorship outweigh the benefits of any increase in reproductive output related to growth.

Keywords

Atlantic salmonSea surface temperatureGrowthSea iceChlorophyll

Introduction

Climate variability in the North Atlantic area affects Atlantic salmon (Salmo salar L.) in a number of different ways (Graham and Harrod 2009; Jonsson and Jonsson 2009). Atlantic salmon may mature and return to freshwater to spawn either as one sea-winter (1SW) adults after 12–18 months at sea, or as much larger multi sea-winter (MSW) fish following 2–5 years of marine migration. The juveniles develop in freshwater nursery areas located in headwater streams and at 1–4 years of age they transform to the smolt stage, acquire the ability to osmoregulate in a hypersaline environment, and migrate to sea. In the ocean, now termed post-smolts, they are visual, generalist predators of zooplankton and nekton (Haugland et al. 2006) living mostly at the ocean surface; during this phase, they experience a period of high marine mortality and the recruitment pattern for return-migrant Atlantic salmon is associated with mortality during their first year at sea. Variation in the distribution of sea surface temperature (SST) has been associated with mortality both for European and North American salmon stock complexes. Thermal conditions impact the recruitment pattern of European Atlantic salmon through the indirect effect of food web changes on the growth of post-smolts (Friedland et al. 2009a). Shifts in food web composition have been associated with warming conditions in the Norwegian Sea resulting in poor growth and survival of salmon. The same climate forcing is hypothesized to control North American salmon recruitment, albeit governed by a different mechanistic linkage that appears to be related to predation rates during the early weeks after ocean entry (Friedland et al. 2003, 2009b). For North American stocks, growth during the post-smolt year does not appear to be critical to first-year survival, which is in contrast to the effect for European stocks (Friedland et al. 2009b; Hogan and Friedland 2010). However, the dynamics of salmon populations, though dominated by post-smolt survival, also are dependent on additional factors such as the overall growth and size of adults. Fecundity is a function of size (Jonsson et al. 1996) and will mediate recruitment potential by limiting the reproductive output of individuals (O’Connell et al. 2008). This has spurred an interest in the marine factors that control the growth and eventual size of returning spawners (Todd et al. 2008; Kuparinen et al. 2009).

Atlantic salmon utilize foraging areas in Arctic and Subarctic waters both in the western and eastern North Atlantic; these provide the conditions for an exceptional increase in size at age during the second post-winter season at sea. The Arctic regions of the North Atlantic undoubtedly are presently subject to increasing environmental change (Lindsay et al. 2009) and, for example, Wang and Overland (2009) include predictions of a sea ice-free Arctic in September by the year 2037. Such large-scale and pervasive changes in the physical environment will inevitably exert major influences at all levels of the pelagic food web in Arctic seas, including top predators such as Atlantic salmon. The exceptional size that MSW fish can achieve between their first and second sea-winter (Allen et al. 1972) is related to the availability of high-energy content prey. These food resources are exploitable by salmon because they have evolved the ability to migrate to distant Arctic habitats and return to their natal river to spawn (Dadswell et al. 2010). The nature of the migration control and specific routes taken by differing stocks is still under debate, but an anadromous life-history strategy clearly confers a fitness advantage to salmon because the remote feeding areas in the Arctic are typified by low species diversity but high seasonal productivity, which affords salmon feeding opportunity with little competitive challenge (Quinn 2005). It is therefore intuitive that variation in salmon growth may be a function of productivity at the base of the food chain in the Arctic, albeit perhaps with lag effects given that salmon prey upon larger zooplankton and forage fish. However, in Arctic waters, pelagic ecosystem productivity may be controlled more by the distribution of sea ice than the rates of primary productivity, because ice cover compromises water column light penetration and thus can delay the progression of areal production until the ice recedes (Zhang et al. 2010).

None the less, the growth of salmon is not simply a function of the energy content of the food ration, but also is dependent on metabolic factors controlled or influenced by their thermal regime. Under some circumstances, salmon growth may be independent of food quantity or quality if ocean migration is not constrained to surface waters of a restricted temperature range. Growth is temperature dependent in salmonids (Handeland et al. 2008), so if Atlantic salmon are unable to behaviorally modify their thermal regime, they may be subject to sub-optimal thermal experience, which itself would result in growth variation (Ottersen et al. 2010). Temperature may influence growth indirectly if the fish seek to thermo-regulate by swimming to regions to find desired thermal conditions. The migration range may consequently be increased, requiring a reallocation of resources from growth into swimming (Hubley et al. 2008), or perhaps even delays in the decision to mature after one or more years at sea (Martin and Mitchell 1985; Hutchings and Jones 1998; Jonsson and Jonsson 2004, 2007).

Given the foregoing, it is relevant that there has been systematic time-series change in the size of Atlantic salmon at their foraging areas in the Arctic; yet, the physical and biological forcing of this change in growth remains unknown. Here, we describe the change in size at age of salmon captured at West Greenland. We then examine the association between time-series changes in salmon growth and variability of several ocean climate thermal factors, in addition to measures of basal ecosystem productivity in the Northwest Atlantic.

Materials and methods

Thermal preferences of Atlantic salmon

Spatial variation in catch rates of Atlantic salmon from research vessel surveys provides a depiction of the specific thermal preferences. Reddin and Friedland (1993) reported catch rates of Atlantic salmon associated within a series of SST “bins” ranging from approximately 1–13°C, comprising data from multiple research vessel surveys in the Labrador Sea, Irminger Sea, and the Grand Banks during the period 1965–1991. These data are re-plotted here after transformation to metric units and application of a cubic B-spline smoother.

Size of Atlantic salmon at West Greenland

A long-standing port sampling program has monitored the size and continent of origin of the Atlantic salmon catch at West Greenland. Components of the catch are attributed to continent of origin using both a discriminant function analysis of scale growth pattern (Reddin and Friedland 1999) and genetic techniques (King et al. 2001; Sheehan et al. 2010). The size and age composition of the catch is raised from a stratified sampling design whereby fish age is determined by scale readings and length and weight are measured on whole and gutted fish. The weights of gutted fish are transformed to equivalent whole weights. The data for continent of origin and the size of 1SW salmon reflect the most recent assessments of the West Greenland fishery reported by the North Atlantic Salmon Working Group (ICES 2010). The length and weight time series extends from 1969 to 2009, albeit with missing data in 1977 and 1993–1994, and therefore provides 38 years of growth data for environmental analytical comparisons.

SST thermal habitat in the Northwest Atlantic

Trends in SST and the derived thermal habitat in the Northwest Atlantic were characterized using the extended reconstructed SST dataset (ERSST, version 3b; monthly data, 2° grid resolution). This time series is based on the SST compilation of the International Comprehensive Ocean–Atmosphere Data Set (ICOADS) and represents interpolation procedures that reconstruct SST fields in regions with sparse data (Smith et al. 2008). Thermal habitat was determined for the key temperature range of 4–8°C, as identified in previous climate studies focused on Atlantic salmon (Friedland et al. 1993). The extent of monthly thermal habitat was determined by summing the area of grid locations within that temperature range and across the region bounded by 65–41°W and 45–75°N for the years 1969–2009. Monthly SST data also were extracted for the same grid locations and over the same time period. Patterns of coherency in the thermal habitat time series were examined with principal component analysis. Thermal habitat and SST, together with other environmental variables, were analyzed for correlation with the weight of salmon at West Greenland; these correlations were corrected for time-series autocorrelation by computing an effective degrees of freedom, using the method of Pyper and Peterman (1998) with corrected probabilities reported as P* and uncorrected probabilities denoted as P.

Sea ice in the Northwest Atlantic

Sea ice concentration was extracted from the daily optimum interpolation sea surface temperature (OISSTv2) analysis database derived from the advanced very high-resolution radiometer (AVHRR) infrared satellite (Reynolds et al. 2007). The data include a large-scale adjustment of satellite biases with respect to in situ data from ships and buoys. Monthly sea ice concentration, expressed as a percentage within a grid box, was extracted for the years 1982–2009 and averaged over 5.0° grid locations within the region bounded by 65–40°W and 45–75°N.

Chlorophyll concentration in the Northwest Atlantic

The distribution of chlorophyll concentration, and the timing and dimensions of the spring phytoplankton bloom in the Northwest Atlantic were calculated from chlorophyll a concentrations derived from the Sea-viewing Wide Field of View Sensor (SeaWiFS) and the Moderate Resolution Imaging Spectroradiometer on the Aqua satellite (MODIS-Aqua) sensors. We used the level-3 processed data at temporal resolutions of 8 days and monthly, averaged over 5.0° grid locations, within the region bounded by 65–40°W and 45–75°N; this was the same grid used for the sea ice data. The monthly chlorophyll data, from the period 1998–2009, were used to characterize the distribution of phytoplankton biomass. The 8-day data permitted the identification of the start and duration of the spring bloom and, by use of sequential t-test analysis, the identification of transition points in chlorophyll time series for each grid location (Rodionov 2006). This procedure is illustrated in Fig. 1a for the composite chlorophyll climatology of the study area; the start date of the bloom is the first day of the first 8-day block included in the bloom period and the bloom duration is the number of days represented by the bloom period. In addition to the start date and duration, two bloom indices were computed. Bloom intensity was the average chlorophyll concentration during the bloom period, and the magnitude of the spring bloom was defined as the sum of chlorophyll concentrations for the 8-day blocks throughout the bloom period. Although the start and end of the spring bloom could be estimated for most years at a given grid location, there were some exceptions. The most common was a failure to identify the end of the spring bloom, in which case the climatological bloom duration for that area was used to estimate the end of the bloom. Other exceptions were missing data, which were filled using linear interpolation. The bloom indices allowed analysis of monthly spatial correlations for a time series of 12 years (1998–2009) duration.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig1_HTML.gif
Fig. 1

Chlorophyll concentration bloom cycle for the Northwest Atlantic, black line, with transition of bloom start and end determined by STARS algorithm, gray line, a. Net primary production (NPP) for the Northwest Atlantic b

Primary production in the Northwest Atlantic

We used net primary production (NPP) data reflecting two computational extensions of the Vertically Generalized Production Model (VGPM) as described by Behrenfeld and Falkowski (1997). The VGPM model is “chlorophyll based” and estimates NPP using a temperature-dependent description of chlorophyll-specific photosynthetic efficiency. The Eppley-VGPM extension employs an exponential function of temperature-dependent photosynthetic efficiencies (Morel 1991). The carbon-based Production Model extension (CbPM) models NPP as a product of carbon biomass and growth rate, rather than the traditional product of chlorophyll and photosynthetic efficiencies (Behrenfeld et al. 2005). We used monthly NPP data based on chlorophyll concentrations from SeaWiFS and MODIS sensors averaged over 5.0° grid locations within the region bounded by 65–40°W and 45–75°N, the same grid used for the sea ice and chlorophyll data (see above). Because NPP over the course of the year is not as highly patterned as chlorophyll concentration (Fig. 1b), no timing or duration statistics were calculated for NPP. The NPP data were available for the years 1998–2009, and thus, monthly spatial correlation analyses were undertaken for time series that were 12 years in duration.

Zooplankton in the Northwest Atlantic

The abundance of plankton organisms that may provide forage for Atlantic salmon (or forage fish exploited by salmon) and that may indicate changes in the food web supporting salmon were extracted from the Continuous Plankton Recorder (CPR) database hosted by the Sir Alister Hardy Foundation for Ocean Science (SAHFOS). We considered the abundances of four plankton categories: total small copepods (<2 mm, traverse stage in analysis); total large copepods (>2 mm, eyecount stage of analysis); total amphipods; total euphausiids. The input data were monthly abundances for the period 1991–2006 from six of the CPR standard areas (Fig. 13). The monthly means were averaged to represent two time periods; winter is represented by the mean of February–April inclusive and summer by May–July. The winter and summer abundances then were compared by correlation analysis with the North America 1SW salmon whole weight time series at West Greenland. The zooplankton data were available for the years 1991–2009 in most cases, and thus monthly spatial correlation analyses were undertaken for time series that were 19 years in duration.

Results

Thermal preferences and distribution of Atlantic salmon

Atlantic salmon utilize the Northwest Atlantic both as a nursery area for post-smolts and as foraging habitat for 1SW fish destined to mature as MSW adults. Although much of the coastal Northwest Atlantic Ocean appears to serve as part of the summer nursery for Atlantic salmon of North American origin (Dutil and Coutu 1988), by the fall of the year research surveys have demonstrated (Reddin and Short 1991) that most post-smolts appear to be concentrated in the southern portion of the Labrador Sea (Fig. 2). They move to the south and east to an over-wintering area that is at least in part associated with the Grand Banks (Reddin 1985). This overwintering area is associated with the dynamic distribution of temperature in the Northwest Atlantic, as evidenced from composite data summarizing the general relationship between catch rate and SST (Fig. 3); the 4–8°C thermal band should in most years intersect bathymetric features including the Grand Banks (45°N, 50°W) and Flemish Cap (47°N, 45°W). After their first sea-winter, Atlantic salmon follow receding ice cover and associated isotherms to re-enter the Labrador Sea on a feeding migration (Fig. 2). Salmon can be found throughout the Labrador Sea, but are concentrated along the coasts of Newfoundland, Labrador, and Greenland in response to the higher concentrations of prey associated with coastal habitats (Reddin and Shearer 1987). It is important to note, however, that suitable temperatures for salmon at West Greenland are restricted to the summer and early autumn months (Dadswell et al. 2010) and this geographic area is essentially unavailable to salmon for most of the year. The early post-smolt nursery for European Atlantic salmon is in the Norwegian Sea (Holm et al. 2000), so it must be assumed that the over-wintering area for European fish that ultimately will mature as MSW adults, and which enter the Labrador Sea the year following first migration from freshwater, is associated with the Irminger Sea and East Greenland (Jensen and Lear 1980).
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig2_HTML.gif
Fig. 2

Map of the Northwest Atlantic and Labrador Sea; gray region denotes area known to serve as post-smolt nursery habitat during the fall of the year. Dashed line arrow denotes general emigrational movement of post-smolts to over-wintering areas to the south and west. Solid line arrows denote general emigrational movements of one sea-winter salmon from over-wintering areas to feeding ground concentration off the Newfoundland, Labrador, and Greenland coasts

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig3_HTML.gif
Fig. 3

Catch rate versus sea surface temperature (°C) as an indicator of thermal preference of Atlantic salmon in ocean waters. These data are redrawn from Reddin and Friedland (1993) and comprise a compilation of data for multiple years, locations, and seasons to emphasize the overall pattern

Size of Atlantic salmon at West Greenland

The size of salmon captured at West Greenland has undergone systematic time-series change over the past three decades (Fig. 4a, b). Throughout, the sea age distribution has been dominated by 1SW salmon, averaging >95% of the catch by number, but the length and weight of European-origin 1SW fish was greater than North American-origin salmon from the beginning of the time series into the early 1990s (Fig. 4a, b). Whole weight of European salmon was >3.5 kg at the beginning of the time series but ≤3.0 kg for North American fish (Fig. 4a). The weight of both continental groups declined to an inflection point minimum around 1993, after which fish both of European and North American origin showed an increase. The weight disparity between the continental groups decreased to the extent that differences by year presently are negligible. A similar pattern is evident for length, in that European fish exceeded 68 cm before declining to a time series minimum of ~63 cm around 1993; North American fish were approaching 65 cm at the beginning of the time series but had declined to ~61 cm at the inflection point (Fig. 4b). North American fish accounted for ~50% of the West Greenland catch prior to 1993 and this has increased to >90% in recent years (Fig. 5). The implication of this dominance shift in continent of origin is that the size of European fish may not be well estimated in recent years and for that reason the ensuing analyses comparing environmental parameters with salmon growth response are focused exclusively on North American fish.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig4_HTML.gif
Fig. 4

Whole weight (a) and fork length (b) of 1SW Atlantic salmon captured at West Greenland by year. Salmon are disaggregated by continent of origin based on stock identification. Smoothing line is 5-point adjacent averaging

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig5_HTML.gif
Fig. 5

Percent composition of the salmon catch of North American origin taken at West Greenland by year. Smoothing line is 5-point adjacent averaging

Thermal habitat in the Northwest Atlantic

Thermal habitat associated with the distribution of Atlantic salmon varies seasonally both in area and with the extent of inter-annual variation. The area of 4–8°C thermal habitat in the Northwest Atlantic was minimal in March (<500,000 km2) and increased to a June maximum of ~1,887,000 km2 (Table 1). Winter habitat area has varied from ~300,000 to 900,000 km2, and this is reflected by the coefficients of variation (CV 0.25–0.30) for the winter months (February–April). Thermal habitat was less variable during the summer months (May–July), as indicated by declining CVs that were at a minimum of 0.11 in July. The time-series patterns of winter thermal habitats were, however, coherent (Fig. 6a) and distinct from the patterns observed for the summer months (Fig. 7). The winter thermal habitat time series share similarities with the time series of weight at age (Fig. 4a), the most obvious of which being the inflection points around 1993. This similarity is confirmed by the correlation between weight of 1SW salmon of North American origin and the extent of thermal habitat in February, March and April (Table 1). This coherence is lost by the summer months (Figs. 6b, 7), and weight at age is not correlated with thermal habitat either in May, June or July.
Table 1

The amount of 4–8°C thermal habitat by month in the Northwest Atlantic Ocean, mean (km2) over the years 1969–2009, standard deviation of the mean, SD, and coefficient of variation, CV

Month

Mean

SD

CV

r

P*

February

522,165

145,760

0.28

0.54

0.01

March

464,976

139,021

0.30

0.54

0.02

April

535,839

131,430

0.25

0.56

0.03

May

1,053,504

255,180

0.24

0.39

0.21

June

1,886,943

319,794

0.17

0.26

0.12

July

1,614,635

183,238

0.11

−0.18

0.29

Correlation between North America 1SW whole weight at West Greenland and thermal habitat, r, and time-series autocorrelation corrected probability, *P

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig6_HTML.gif
Fig. 6

Normalized extent of 4–8°C sea surface temperature thermal habitat area in the Northwest Atlantic by year for a the months of February, March and April and b May, June and July. Smoothing line is 5-point adjacent averaging

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig7_HTML.gif
Fig. 7

Principal components of monthly thermal habitat, factor coordinates for components 1 and 2

Sea surface temperature and salmon growth

The foregoing indicates that recent temporal changes in mean size of Atlantic salmon are related to ocean climate changes in the distribution of winter-spring SST in the Northwest Atlantic, but during February most of the Labrador Sea is only marginally within the thermal preference range of Atlantic salmon (Fig. 8a). For two regions within the study area, SST correlated positively with salmon growth in bands centered on 55°N, 47°W, and 63°N, 59°W (Fig. 8g). The more southerly band is associated with 4–8°C water (i.e. the preferred temperature for Atlantic salmon), whereas the northerly band is associated with temperature <2°C, which would place these areas at the low extreme of salmon thermal preference. This pattern and correlations persisted through March to May, with a slight translation of correlation distribution to the northwest of the area (Fig. 8b–d, h–j). By June, a significant portion of the preferred thermal range of salmon, as delimited by the 4°C isotherm, invaded the Labrador Sea and was associated with a large continuous region of high positive correlation with growth (Fig. 8e, k). The fisheries in the southern Labrador Sea historically were underway by July, which would accord with the distribution of 4–8°C SST and a pervasive pattern of positive correlation between growth and SST throughout the area (Fig. 8f, l).
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig8_HTML.gif
Fig. 8

Sea surface temperature (°C) distribution in the Northwest Atlantic and the Labrador Sea during February–July (af); shading added to enhance value contrast. Correlation between North America one sea-winter whole weight at West Greenland and sea surface temperature for February–July (gl). Light gray shading marks approximate regions with uncorrected correlations significant at P = 0.05, dark gray marks approximate regions significant at P = 0.01. Asterisks mark discrete locations where autocorrelation-corrected correlations are significant at *P = 0.05 and *P = 0.01, small and large asterisks, respectively

Sea ice and salmon growth

The Labrador Sea is dominated by sea ice through much of the spring, and its potential effect on growth of North American salmon may be restricted to a portion of the continental shelf region northeast of Newfoundland. Sea ice cover in the Labrador Sea ranged from 10 to 90% during February through March (Fig. 9a–c). A significant negative correlation between sea ice and growth emerged for an area centered around 53°N, 53°W (Fig. 9g–i), which corresponds to average SST in the area of <1°C (Fig. 8a–c). As the percent sea ice receded to <10% between May and July (Fig. 9d–f), the correlative pattern was lost (Fig. 9j–l).
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig9_HTML.gif
Fig. 9

Percent sea ice distribution in the Northwest Atlantic and the Labrador Sea for February–July (af); shading added to enhance value contrast. Correlation between North America one sea-winter whole weight at West Greenland and percent ice distribution for February–July (gl). Shading and P value details as for Fig. 8

Chlorophyll concentration and bloom dynamics impacts on salmon growth

The most prominent feature of the chlorophyll concentration fields of the Northwest Atlantic is the spring bloom that develops along the coast of West Greenland. Chlorophyll concentrations are understandably low during February and March because much of the region still shows significant ice cover (Fig. 10a, b). The beginning of the spring bloom was evident in the April field (Fig. 10c) off West Greenland and south of Newfoundland in an area associated with the Scotian Shelf (45°N, 55°W). The West Greenland bloom continued to build into May, whereas the Scotian Shelf bloom dissipated (Fig. 10d); summer chlorophyll concentrations were elevated and focused in the central portion of the Labrador Sea (Fig. 10e, f). The correlation fields associating salmon growth with chlorophyll concentration were weak and unstructured during February through April (Fig. 10g–i). However, some correlative density is detectable in the May and June fields associated with the West Greenland bloom in May (Fig. 10j) and the elevated concentration of the central Labrador Sea (52°N, 47°W) in June (Fig. 10k). In both cases, the correlations between salmon growth and chlorophyll concentration were negative and relatively weak considering the limited number of locations showing significant effect. The correlation pattern dissipated by July (Fig. 10l).
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig10_HTML.gif
Fig. 10

Chlorophyll concentration (mg m−3) distribution in the Northwest Atlantic and the Labrador Sea for February–July (af); shading added to enhance value contrast. Correlation between North America one sea-winter whole weight at West Greenland and chlorophyll concentration for February–July (gl). Shading and P value details as for Fig. 8

Bloom indices reflecting changes in bloom timing, duration, and extent were highly patterned in the Northwest Atlantic, but timing appears to be of no consequence to salmon growth, and bloom dimension indices reflect the patterns already detectable in chlorophyll concentration. The spring bloom in the Labrador Sea generally began between day of the year 100–120, with the most prominent delay associated with the central portion of the Labrador Sea (Fig. 11a). The correlation field is unstructured and marked by a single negative correlation (Fig. 11e). Regions of the Labrador Sea with the most delayed bloom start dates were associated with the longest bloom durations, with blooms lasting on the order of 60 days (Fig. 11b); the correlation field was characterized by low, un-patterned correlation (Fig. 11f). The bloom intensity and magnitude indices (Fig. 11c, d) both show high values associated with the area off West Greenland; there was high bloom intensity associated with the Scotia Shelf and high magnitude associated with the central Labrador Sea. The only significant growth correlation was associated with bloom magnitude off West Greenland. This was negative and indicated that higher salmon growth was associated with lower intensity, and less protracted blooms (Fig. 11g, h).
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig11_HTML.gif
Fig. 11

Spring phytoplankton bloom start day (a), duration of the spring bloom in days (b), intensity of chlorophyll concentrations in (mg m−3) during the spring bloom (c), and magnitude of the spring bloom (mg m−3) day (d) distributions. Correlation between North America one sea-winter whole weight at West Greenland and bloom dynamics parameters start day, during, intensity and magnitude (eh, respectively). Shading and P value details as for Fig. 8

Net primary productivity and salmon growth

The highest levels of NPP were associated with the highest concentrations of chlorophyll during the months of maximum solar radiation. NPP was lowest and spatially unstructured during February and March (Fig. 12a, b), but centers of high NPP associated with the West Greenland and Scotian Shelf spring bloom areas developed subsequently (Fig. 12c). The correlation fields yielded no significant pattern for February through April (Fig. 12g–i). May NPP was highest in the area associated with West Greenland (Fig. 12d) and correlations in this region were negative, with two separate locations at a significant level (Fig. 12j). By June, NPP at West Greenland had declined and higher productivities were observable south of Greenland (Fig. 12k); as for the other plankton productivity data, significant correlations were negative in sign (Fig. 12k). NPP was lower in July and uncorrelated with salmon growth (Fig. 12f, l, respectively). As for chlorophyll concentrations, those between salmon growth and NPP concentration were negative and relatively weak with significant effects only at a limited number of locations.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig12_HTML.gif
Fig. 12

Net primary production (mg C m−2 day−1) distribution in the Northwest Atlantic and the Labrador Sea for February–July (af); shading added to enhance value contrast. Correlation between North America one sea-winter whole weight at West Greenland and net primary production for February–July (gl). Shading and P value details as for Fig. 8

Zooplankton and salmon growth

The seasonal zooplankton data provided contrasting associations with the weight of salmon at Greenland. With the exception of copepod abundances in SAHFOS area E9, summer abundance indices were higher than winter indices (Table 2). Small copepods ranged from 39 to 755 m−3, with winter abundances higher in the southwesterly areas and a reverse summer pattern of higher abundances in northeasterly areas. Large copepods ranged from 6 to 154 m−3, with little contrast in winter distribution but a distinct abundance peak in area D7 during summer. Amphipods and euphausiids both ranged from minima of <1 to maxima just >10 m−3; summer amphipods peaked in area D7, whereas euphausiids were more abundant in northern areas. The winter CPR zooplankton data likely overlap the distribution of salmon very well, but the overlap erodes with the summer data, though not to any marked extent. By following the February–July progression of SST data (Fig. 8a–f), the overlap between the SAHFOS areas and the range of 4–8°C water remains significant in May, but by July the relevant SST ranges—and likely the fish—are further north and west in the Labrador Sea. It is undeniable that the CPR data provide incomplete coverage of salmon habitat in the Northwest Atlantic. But that the CPR data are available for at least their southern range is both valuable and informative, especially for the winter months. There were no significant correlations between winter zooplankton abundances and salmon growth (Fig. 13). However, the summer zooplankton means were significantly correlated in multiple SAHFOS areas, most notably the significant negative correlations between euphausiid abundance and salmon weight in areas C7, D7, and E8. Additionally, large copepods and amphipods were negatively correlated with salmon weight in area E8.
Table 2

Mean monthly winter and summer concentrations of zooplankton (number m−3) by SAHFOS area

SAHFOS

Small copepods

Large copepods

Amphipods

Euphausiids

Area

Winter

Summer

Winter

Summer

Winter

Summer

Winter

Summer

B7

61.99

755.26

27.48

84.54

0.05

2.05

1.61

14.24

C7

38.91

573.17

12.92

99.24

0.12

4.06

6.36

11.94

D7

84.27

719.28

25.64

154.04

0.81

10.14

3.81

13.79

D8

68.76

536.90

23.75

91.13

1.54

8.21

3.23

14.32

E8

158.95

338.40

45.46

57.97

2.29

4.26

5.00

9.79

E9

517.00

160.99

18.68

5.64

1.45

2.32

0.67

1.22

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-011-1105-z/MediaObjects/300_2011_1105_Fig13_HTML.gif
Fig. 13

Correlation between North America one sea-winter whole weight at West Greenland and the abundance of four function taxonomic groupings of zooplankton for February–April (a) and May–July (b) for six SAHFOS standard areas in the Northwest Atlantic (areas B7, C7, D7, D8, E8, and E9). Asterisks mark taxa where autocorrelation corrected correlations are significant at *P = 0.05

Discussion

There has been an excursional change in weight at age of Atlantic salmon in the Arctic foraging areas in the Labrador Sea. This transition involves periods of declining and increasing weight at age, thus increasing our confidence in associative relationships spanning the excursion and lowering our confidence in associations testing only a monotonic portion of the excursion. Metrics related to the thermal environment span the full excursion of the data and suggest that weight at age is related (directly or indirectly) to Arctic temperature and ice conditions. Warmer ocean climate conditions are associated with greater growth of salmon, and the indicators of thermal conditions suggest multiple and potentially interacting mechanisms by which these conditions affect growth. Shorter time series related to production at the base of the food web overlap only a period of increasing weight-at-age; the indication from these tests of association suggests that primary production and other metrics related to plankton biomass and bloom timing are of no consequence to the observed change in size. Perhaps the key (unquantified) component here is the importance of the availability of small forage fish (e.g. barracudina) to salmon feeding.

The identification of transitional states of population response in salmon and the more general ecosystem observations are concordant with the emerging picture of regime shift in the Arctic and its manifest effects both on the physical environment and ecosystem responses (Greene et al. 2008). The 1990s regime shift in the Northwest Atlantic has been associated with a transitional state of the North Atlantic Oscillation and a change in atmospheric forcing resulting in an eastward shift of the low-pressure system over the Northwest Atlantic and a change in the meridional component of the winds (Drinkwater 2004). The regime shift period of the early 1990s also is associated with an expansion of the cold intermediate (depth) layer, which provides an environmental index of thermal and production conditions of the eastern Canadian shelves associated with the Labrador Sea (Colbourne 2004). None the less, an intractable analytical problem remains because larger salmon typically are piscivorous as well as preying opportunistically on larger zooplankton (Haugland et al. 2006), and no data are available on likely Northwest Atlantic forage fish species such as barracudina (Jensen and Lear 1980).

The change in winter thermal habitat associated with weight-at-age suggests a spatial dimension to the temperature effect on Atlantic salmon, but this possible impact may be the least plausible of competing hypotheses. The concept of aquatic thermal habitat has been applied variously to a range of problems as a means of characterizing both physical constraints of the environment and associated niche space. These impingements are more obvious and acute in freshwater aquatic systems so, as expected, the effects of habitat constraints on distribution caused by shifting temperature conditions often are associated with the creation and re-distribution of non-contiguous habitats (Eaton and Scheller 1996). Furthermore, shifts in the quantity of thermal habitat—as associated with variation in water column stratification—have been linked with growth and reproduction of freshwater fishes (King et al. 1999). Thermal habitat is finding application in marine environments as it relates to the definition of niche space. For example, Pacific salmonids display age-related thermal habitat use (Morita et al. 2010) and the vertical distribution of striped bass (Morone saxatilis) appears limited by coastal thermal habitat (Nelson et al. 2010). To view thermal habitat as a constraint on salmon growth, the habitat would have to be limiting in a compensatory fashion. But this argument fails to explain the growth responses of Atlantic salmon, because over the excursion of weight-at-age, there has been a multi-decadal decline in salmon abundance (Friedland et al. 2009a, b); thus, large extents of thermal habitat area, associated with higher growth, have been alternately associated both with high and low salmon abundance. Density dependence has been demonstrated for high abundance salmonids (Wertheimer et al. 2004; Martinson et al. 2008), but Atlantic salmon are orders of magnitude less abundant than Pacific species and represent only a minor component of the food web thus making it difficult to support density dependence in the same context.

The distribution of ocean thermal conditions may be affecting salmon directly by eliciting physiological and behavioral compensations that in turn affect growth. For temperature to produce a direct effect, salmon would have to navigate to a specific habitat during the overwintering and summer growth seasons. Salmon may well select habitats during this period on the basis of the distribution of prey resources, which themselves would have to show distributions unaltered by thermal regime, thus causing salmon to migrate into varying thermal conditions. Because adult salmon feed opportunistically on pelagic prey, such static and deterministic habitat associations seem unlikely (Jacobsen and Hansen 2001; Dempson et al. 2010). Alternatively, if salmon, and their epipelagic prey, seek (or are confined to) a preferred thermal regime, then the direct effect of temperature is relegated to the behavioral expense of modifying migration to locate the desired thermal conditions. In the case of Atlantic salmon in the Northwest Atlantic, warm conditions would be associated with a longer, faster migration because desired thermal conditions would be distributed further into the Labrador Sea; however, warmer conditions are associated with greater growth, so the data do not support this hypothesis either. The greater likelihood is that SST conditions are indicative of factors impacting properties of the pelagic ecosystem that alter feeding opportunities for salmon.

The possibility of recent changes in the trophic position within the food web occupied by Atlantic salmon in the North Atlantic has been investigated recently by Sinnatamby et al. (2009). Their analyses of stable isotope variation for multiple North American stocks, and a single European stock, indicate little evidence of significant shifts in trophic ecology; this might lead to the conjecture that the large-scale time-series shifts in growth response of North American fish reported here are more likely attributable to marked changes in the relative abundance and availability of preferred prey species rather than to systemic alterations in the trophic levels at which salmon are operative. Notwithstanding the lack of data for potential forage fish species, the nature of such a change in feeding opportunities for Atlantic salmon is not obvious from our analysis of zooplankton dynamics and primary productivity; yet, there is comparative precedent in the region and associated areas to presume there have been changes higher up the food web. The Labrador Sea has some dominant elements in the patterning of the spring bloom and primary production. The ice retreat influences the commencement of the spring bloom throughout most of the region (Wu et al. 2007) and the dominant production area is the West Greenland shelf, where primary productivity is greatly influenced by freshwater-mediated stratification (Frajka-Williams and Rhines 2010). The central portion of the Labrador Sea is more classically controlled by the time course of irradiation in the water column and tends to be less productive. The absence of positive associations between the size and phenology (Koeller et al. 2009) of phytoplankton blooms in the region suggest these factors are not influencing salmon growth; neither does the rate of primary production during the summer season appear to be important.

However, it would be imprudent to dismiss variation in food web productivity as a factor controlling the growth of salmon. The present time-series tests of association between growth and plankton production are restricted in the number of years and do not cover the full excursion of the growth data. Furthermore, the Arctic ecosystem has changed in response to thermal variation, and associated melting and salinity effects, within a time frame and including regime break points that closely resemble the pivot points in the salmon growth time series (Greene et al. 2008). The associated effects usually involve changes in salinity and result in alterations of water column structure, but these effects have not been uniform in water bodies influenced by the Arctic (Drinkwater et al. 2009). Nor has the effect of Arctic ice melt been simple to quantify because, in some cases, this actuates complex trophic cascades in food webs (Shackell et al. 2010) and may include as yet unconsidered aspects of ecosystem dynamics, such as the location and stability of ocean frontal features (Belkin et al. 2009). It would be prudent to assume that transitions in Arctic thermal conditions have driven changes to the ecosystem used by Atlantic salmon and have thus modified their feeding opportunities. Such changes may well embrace multiple components of the food web, including the replacement and re-distribution of zooplankton and nekton species and at multiple trophic levels, and thereby critically affecting this generalist, opportunistic predator.

Many fisheries for Atlantic salmon utilized size-selective gear and thus may have influenced phenotypic size at age through selection. Selection that acts on the faster-growing segment of a population will have an effect on the frequency of genes related to growth (Conover and Munch 2002; Reznick and Ghalambor 2005). Thus, the decline in size-at-age seen in the early segment of the time series may be related to selective forces of the combined exploitation of North American stocks in Greenland and home waters (Jensen 1990). Additionally, the actual exploitation regime changed in the early 1990s and at around the same time of the reversal from a trend of decreasing to increasing size-at-age (ICES 2010). Conover et al. (2009) assert that following the removal of size-selective pressures, growth characteristics can be restored; but the effect is not immediate and requires the accumulated contribution of upwards of twelve generations. Salmon size-at-age increased markedly in the early 1990s, well before any reasonable expectation that a genetic effect could be applicable.

Incisive and informative analyses of correlations between physical environmental variations and putative biological responses of populations and ecosystems to those changes are fraught with difficulties, but the goal of a mechanistic understanding of these undoubtedly complex processes is not necessarily unattainable (Wells et al. 2008). The relationship between the observed growth changes for Atlantic salmon and Arctic climatic fluctuations are consistent both with a direct physiological effect on growth, as well as indirect effects through changes in ecosystem dynamics including prey availability. Our conclusion accords with Michaud et al. (2010) for the related Arctic charr (Salvelinus alpinus) in Labrador, in that this coherence in response does not necessarily allow for definitive separation of direct physiological effects versus indirect food web effects. It does, however, increase our confidence that the growth response is biological instead of fisheries related because the fishery effects would not be equivalent between these two species. Other species in the region experienced a decline in growth during the time period entering the 1990s, but not all populations experienced as marked a rebound in growth as salmon. Cod (Gadus morhua) stocks on the Newfoundland and Scotian shelves experienced a growth minimum centered on 1992; however, a failure to return to robust stock and individual condition levels has been attributed to multiple causes (Sherwood et al. 2007; Hutchings and Rangeley 2011). Likewise, capelin (Mallotus villosus) showed growth minima variously centered on 1992 or 1995 and have only partially improved in size-at-age, which might be related to a change in their distribution and habitat (Carscadden 2005). Changes in marine fish production can, in part, be attributed to population changes in top consumers such as birds (Gaston et al. 2009) and, to a lesser extent, marine mammals because these populations have increased markedly over the past half century (Sjare and Stenson 2010).

The improved growth of salmon at West Greenland would be expected to increase, or at least maintain, the reproductive contribution of spawners returning to home waters; however, the continued decline in recruitment of adults over the past decade has shown that any increase in spawning that could be attributed to this growth effect has been overshadowed by mortality in other segments of the life history. The ocean climate signal affecting salmon growth during the first sea-winter is not in synchrony with that affecting post-smolt survival, and this would appear to be of little consequence because the persistence of stocks may be determined by the trajectory of climate in another region of the ocean during another segment of the year.

Acknowledgments

We thank colleagues contributing to the North Atlantic Salmon Working Group in their continuing effort to understand the life history of salmon at sea. We also thank David Johns of the Sir Alister Hardy Foundation for Ocean Science (SAHFOS) for the provision of zooplankton data.

Copyright information

© US Government 2011