By the early 1990s, at the onset of GLOBEC, researchers studying the responses of marine ecology to physical environmental forcing began to consider the large-scale climate variability (Forchhammer and Post 2004), typically expressed in terms of changes in climate indices related to atmospheric pressure patterns, such as the ENSO, NAO, or the North Pacific Index (NPI). The NPI is the area-weighted sea-level pressure over the region 30°N–65°N, 160°E–140°W. Other climate indices, such as the Pacific Decadal Oscillation (PDO) or the (p.13) AMO, were developed based on the intensity of sea temperature patterns. Further descriptions of several of these indices are provided in Box 2.1. While the observed climate-induced changes in marine ecosystems are indeed responses to local conditions, large-scale indices often account for significant portions of this local or regional variability. This is because the local climate changes are often a response to the large-scale processes, and also because large-scale indices are related to several physical elements (e.g. air and ocean temperatures, sea ice, winds, etc.) and as such they can be more representative of the climate forcing than any single local variable (Stenseth et al. 2003). Also, local indices can vary significantly while large-scale indices are smoother, with less noise. Thus, in some cases large-scale indices can account for as much, or at times more, of the variance of ecosystem elements than local indices (e.g. Drinkwater et al. 2003). Using large-scale climate indices potentially allows linking climate-induced ecological dynamics over a range of trophic levels, species, and geographic locations. Some examples of the ecological impacts of the variability in these indices are provided in Section 2.7.

Linkages between climate indices, weather events, or climate patterns at far distant locations are called teleconnections. Such teleconnections can be within ocean basins or between oceans. For example, Müller et al. (2008) found that at decadal to 20-year time scales, the NAO in the North Atlantic tends to be out of phase with ENSO in the equatorial Pacific and with the PDO in the North Pacific. Better known, through ENSO forcing, are the associations between heavy rains along the west coast of South America and droughts in Indonesia (McPhaden et al. 2006), or, in the Atlantic Ocean, between years of reduced ice in the Barents Sea and heavy ice off Labrador due to NAO variability (Deser et al. 2000). Because the ecological literature often puts emphasis on climate and ecosystem (p.14) (p.15) (p.16) teleconnections, we will discuss this in greater detail with a focus on northern Hemisphere atmospheric teleconnections.

Box 2.1 Climate indices

Several indices of large-scale atmospheric pressure patterns have been developed and used to compare with the variability of marine ecosystems. The most common have been the El Niño-Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), Atlantic Multidecadal Oscillation (AMO), and the Pacific Decadal Oscillation (PDO).

ENSO

El Niño events are characterized by significant warming of equatorial surface waters over an area extending from the International Date Line to the west coast of South America. The atmospheric phenomenon related to El Niño is termed the Southern Oscillation (SO), which is a global-scale east-west see-saw in atmospheric mass involving exchanges of air between eastern and western hemispheres centred in tropical and subtropical latitudes. During an El Niño event, the sea level atmospheric pressure tends to be higher than usual over the western tropical Pacific and lower than usual over the eastern Pacific, and warmer-than-average SSTs cover the near equatorial Pacific. Warm ENSO events are those in which both a negative SO extreme and an El Niño occur together. The interval between ENSO events is typically 2 to 7 years. During the warm phase of ENSO, the increased upper layer sea temperatures in the central and eastern tropical Pacific shifts the location of the heaviest tropical rainfall eastward from its usual position centred over Indonesia and the far western Pacific. This shift in rainfall and the associated latent heat release alters the heating patterns that force large-scale waves in the atmosphere, producing an amplification of the Pacific-North America (PNA) pattern (Hoerling et al. 1997). Since the PNA pattern and PDO are linked, this has led to the PDO being described as either a long-lived El Niño-like pattern of Indo-Pacific climate variability or a low-frequency residual of ENSO variability on multidecadal time scales (Newman et al. 2003).

The NAO

The long-term monthly mean surface atmospheric pressure pattern over the North Atlantic is dominated by a subpolar (Icelandic) Low and a subtropic (Azores) High (Barnston and Livezey 1987), which tend to intensify or weaken at the same time. This variability is known as the North Atlantic Oscillation or NAO (Hurrell 1995). The most commonly used NAO index is the average of the December to March surface atmospheric pressure difference between Iceland and Lisbon in Portugal (Hurrell 1995). In the positive NAO phase, the Icelandic Low is lower than normal, the Azores High is higher than normal and the westerly winds across the Atlantic intensify with more storms of higher intensity (Box 2.1, Fig. 1). Stronger southerly flows produce warmer than normal air temperatures and higher precipitation over the eastern United States and northern Europe, whereas the northerly winds between eastern Canada and Greenland and in the vicinity of the Mediterranean cause cooler than normal air temperatures and drier conditions in these locations. The opposite pattern occurs during the negative phase of the NAO (Box 2.1, Fig. 1).

Box 2.1, Figure 1 The atmospheric conditions during the positive (left) and negative (right) phases of the NAO. (Taken from http://www.ldeo.columbia.edu/res/pi/NAO/). (See Plate 4).

Box 2.1, Figure 2 The time series of the annual (dashed line) and 5-year running mean (solid line) of the NAO Index. (Based on Hurrell 1995).

Box 2.1, Figure 3 The time series of the AMO (temperature anomalies in °C). The index is the deviation of the annual mean SSTs between 0–60°N and 75–7.5°W from its long-term mean, and low past filtered with a 37-point Hendersen filter and detrended. (Taken from Sutton and Hodson 2005 reprinted with permission from AAAS.)

There are periods when the NAO exhibits persistence from one winter to the next. For example, during the 1960s there was a long period of intense and principally negative NAO values followed by a period through to the mid-1990s of increasing values on top of strong decadal variability (Box 2.1, Fig. 2). The NAO has a strong influence on the atmosphere and hence the physical oceanography of the North Atlantic. Comprehensive reviews of ecological responses in the North Atlantic to NAO are provided by Drinkwater (2003) and Stenseth et al. (2004).

The AMO

The AMO (Kerr 2000) is defined by the annual mean SST averaged over the North Atlantic after the linear trend has been removed (Delworth and Mann 2000; Sutton and Hodson 2005). A positive AMO index is associated with higher-than-normal SSTs throughout most of the North Atlantic and a negative index with colder temperatures. The AMO index has varied between warm and cool phases with a period of about 60–80 years since the late 1880s (Box 2.1, Fig. 3). The temperature changes associated with the AMO can alternately obscure and exaggerate the increase in temperature due to anthropogenic-induced warming (Latif et al. 2004). The AMO index is correlated to air temperatures and rainfall over much of the northern Hemisphere, in particular, North America and Europe (Sutton and Hodson 2005). For example, the two most severe droughts of the twentieth century in the United States occurred during the positive AMO between the 1920s and the 1960s, that is the dust bowl of the 1930s and the 1950s drought. In contrast, extreme rainfall fell during this time in Florida and the Pacific Northwest. In the marine environment there are strong ecological signals associated with the AMO, examples of which are described by Sundby and Nakken (2008) as well as by Brander (2003a) and Drinkwater (2006) and references therein.

The PDO

The PDO index is based on the observed monthly SST anomalies in the North Pacific north of 20°N

Box 2.1, Figure 4 The annual (dashed line ) and 11-year running mean (solid line) of the PDO index. (from http://www.beringclimate.noaa.gov/data/)

after removal of the long-term warming trend. The monthly PDO indices are often averaged to create seasonal or annual indices (Box 2.1, Fig. 4). During the positive phase of the PDO, the Aleutian low pressure system deepens and shifts southward, winter storminess in the North Pacific intensifies, SSTs are anomalously warm along the coast of North America and cool in the central and western North Pacific, the subpolar gyre intensifies, the mixed-layer depth shoals within the Alaska gyre enhancing upwelling, the Oyashio pushes farther southward and there is an intensification of the Kuroshio Current (Box 2.1, Fig. 5). During the negative phase of the PDO, conditions are generally the opposite of those in the positive mode. Changes in the sign of the PDO have been associated with regime shifts (Moloney et al., Chapter 7, this volume) in the North Pacific (Mantua et al. 1997).

Box 2.1, Figure 5 The atmospheric and oceanic conditions as well as the phytoplankton response during the positive phase of the PDO. (Modified from Chavez et al. 2003 with perission from AAAS). (See Plate 5).

James E. Overland

James E. Overland

2.3.1 Atmospheric teleconnections

The concept of atmospheric teleconnections and their influences on marine ecosystems has been explored by a number of authors inside the GLOBEC programme (e.g. Bakun 1996; Alheit and Bakun 2010; Overland et al. 2010; Schwing et al. 2010. Overland et al. 2010 noted that stationary atmospheric waves are formed in winter in the northern Hemisphere due to continent-ocean heating contrasts and the influences of large mountain ranges such as the Rockies and Himalayas. These waves generate regions of atmospheric low and high pressure which are linked, often with opposite conditions, over large spatial scales. The most prominent (p.17) teleconnections over the northern Hemisphere are the NAO and the Pacific-North America (PNA) pattern (Barnston and Livezey 1987; also see Box 2.1). Analyses of northern Hemisphere pressure patterns indicate that the NAO, PNA, and similar large-scale climate patterns represent about one-half of the climate variability signal in the northern Hemisphere, the remainder being ‘climate noise’ (Overland et al. 2010). Although atmospheric teleconnections between ocean basins have been investigated (e.g. Honda et al. 2005), no consistent covariability between ocean basins is apparent when the full records for the twentieth century are considered (Overland et al. 2010). Evidence for atmospheric teleconnections within ocean basins is more convincing, in particular for marine populations across the North and South Pacific Ocean (e.g. Alheit and Bakun 2010; Schwing et al. 2010).

Drinkwater et al. (2010a) provide examples of ecosystem impacts in response to climate variability and the possible mechanisms by which the response is generated. It must be remembered, however, that each of these individual processes usually act in concert with several others to produce the observed response on marine ecosystems.

2.6.1 Sea temperature

(p.21) 2.6.1.5 Distribution

The most convincing evidence of the effects of climate change on marine ecosystems comes from distributional shifts of marine organisms. They generally are most evident near the northern or southern boundaries of the species' geographic range with warming usually causing poleward movement and cooling an equatorward movement. Examples abound of northward distributional shifts in response to increasing temperatures, for example, C. finmarchicus and C. helgolandicus in the north-east Atlantic (Beaugrand et al. 2002c; Box 2.2); several zooplankton species in the north-east Pacific (Mackas et al. 2007); Atlantic cod off West Greenland, (p.22) Iceland, and in the Barents Sea (Drinkwater 2006); capelin off the east coast of Canada (Frank et al. 1996), Iceland (Sæmundsson 1934), and the Barents Sea (Vilhjálmsson 1997b); and numerous fish species during recent warming in the North Sea (Perry et al. 2005), along the continental shelf from the Iberian Peninsula to west of Scotland (Brander et al. 2003a) and in the Bering Sea (Mueter and Litzow 2008). The spawning location of Atlantic cod off Norway has been found to vary in concert with the long-term temperature changes such that the cod tend to favour more northern spawning during warm conditions and more southern spawning under cold conditions (Sundby and Nakken 2008). The spawning areas of squid (Todarodes pacificus) off Japan expand in the Sea of Japan and East China Sea during warm periods and retract to the East China Sea during cold years (Sakurai et al. 2000; Box 2.3). Temperature changes can have a strong, especially negative, effect on species whose distribution is tied to specific geographic features (Pershing et al. 2004) such as fish species that spawn on a few fixed submarine banks (Mann and Lazier 1996), seals whose breeding grounds are geographically limited, and the nesting grounds of marine birds (Graybill and Hodder 1985).

Box 2.2 North-east Atlantic zooplankton

Continuous Plankton Recorder (CPR) surveys have collected plankton data since the 1930s (Reid et al. 2003a). The CPR is a towed body that funnels water past a moving filter of silk. Phytoplankton and zooplankton that are retained by the silk are later counted and identified to species. Species abundances can then be determined as a function of location knowing the ship's track and its speed. Through a concerted effort to maintain systematic sampling in space and time in over 70 years, CPR data have proven increasingly useful for describing seasonal changes, interannual changes, long-term trends, and species' distributional changes (Brander et al. 2003b). The longest and most complete spatial coverage of CPR data is in the north-east Atlantic.

The CPR data from the North Sea revealed significant changes in the plankton in response to climate variability in the late 1980s (Box 2.2, Fig. 1). A large influx of warm saline water entered the northern North Sea due to an increase in the northward flowing shelf edge current to the west of the British Isles (Reid et al. 2001a,b). In response, the North Sea phytoplankton biomass doubled (Reid et al. 2001a), the plankton community structure changed, in particular the calanoid-copepod diversity (Beaugrand and Ibanez 2002), and the phytoplankton blooms of several of the dominant species peaked 1 to 3 months earlier than usual (Edwards et al. 2002). Seabird abundances also increased and large numbers of horse mackerel (Trachurus trachurus) entered the northern North Sea subsequently resulting in high

Box 2.2, Figure 1 The mean number of southern shelf edge, pseudo-oceanic temperate, cold temperature, and subarctic species in the north-east Atlantic for different time periods from the 1960s to the late 1990s. (From Beaugrand et al. 2002c.) (See Plate 6).

catches by the fishing fleet (Reid et al. 2001a,b). These ecosystem changes led Reid et al. (2001a) to label this a regime shift.

The CPR data have also revealed longer-term trends in the plankton community linked to multidecadal climate change, in particular the AMO (Box 2.1). A decline in the zooplankton population of cold-water species such as C. finmarchicus and a rise in the neritic (coastal) species from the late 1950s to the end of the century were observed in the North Sea (Beare et al. 2002). West of the British Isles, the number of warm-temperate and temperate pseudo-oceanic species assemblages increased northwards by over 10° of latitude while the number of cold-mixed water and subarctic species assemblages decreased (Box 2.2, Fig. 1; Beaugrand et al. 2002c). All assemblages showed consistent long-term changes that appear to reflect the response of the marine ecosystems towards a warmer dynamic regime. Subsequent studies revealed that the biodiversity of calanoid copepods responded quickly to the SST rise by moving geographically northward at a rapid rate of up to about 23 km/year (Beaugrand et al. 2009). Increases in the occurrence of subtropical fish paralleled these changes and some studies reported latitudinal movements of fish species distribution that have been mainly related to SST changes (Quéro et al. 1998; Brander et al. 2003a; Perry et al. 2005).

Consistent with this shift, along the continental shelf from the English Channel to Scotland and in large parts of the North Sea, C. finmarchicus were replaced by the smaller warm-water species C. helgolandicus (Beaugrand et al. 2002c). Helaouët and Beaugrand (2007) showed that among the various physical parameters they tested, the distribution of these two copepod species was primarily linked to temperature, although part of the temperature change was likely due to redistribution of water masses through changes in circulation. Beaugrand et al. (2003a) concluded that the shift from the spring spawning C. finmarchicus to the autumn spawning C. helgolandicus led to a mismatch of the spring-spawned Atlantic cod (Gadus morhua) larvae with their food. This mismatch, and an observed reduction in prey biomass and mean size of prey, are believed to have resulted in reduced survival of the young cod and contributed to the observed decrease in the population of cod in the North Sea.

Gregory Beaugrand

The reasons fish change their distribution is often unclear. Adult fish may actively move to seek their preferred thermal range, to follow their prey or to avoid their predators. Larval survival may improve in more northern areas with atmospherically induced temperature increases, but temperature (p.23) (p.24) changes may also be associated with modification in circulation patterns, which in turn may carry larvae into areas that they previously had not occupied.

Box 2.3 Japanese common squid

The Japanese common squid, Todarodes pacificus (Cephalopoda: Ommastrephidae), is found in the Sea of Japan, the East China Sea, and in the Oyashio and Kuroshio current systems off eastern Japan. There are three subpopulations with different peak spawning seasons; summer, autumn and winter, with the later two the largest and most important (Murata 1989). Autumn spawning occurs mainly in the southern Sea of Japan, including in the Tsushima Strait between Japan and the Republic of Korea, while winter spawning occurs in the East China Sea off Kyushu Island in southern Japan. This species has a 1-year life cycle with the autumn-spawned squid migrating to the northern Sea of Japan and then back again to spawn before dying. Some of the winter-spawned squid migrate to feed in the northern Sea of Japan while others migrate in the waters off eastern Japan to the Oyashio region where they feed during the summer and autumn. These squid feed on large zooplankton and small fish, and can impact the recruitment of walleye pollack by feeding on their juveniles (Sakurai 2007). After hatching, the squid paralarvae ascend to the surface layer above the continental shelf and slope and are advected into convergent frontal zones (Yamamoto et al. 2002, 2007). Laboratory studies reveal that hatchlings (<1 mm mantle length) ascend to the surface successfully at temperatures between 18 and 24°C, and especially between 19.5 and 23°C (Sakurai et al. 1996; Sakurai and Miyanaga, unpublished data).

Low-pass filtered SST data off south-western Japan show alternate warm and cold periods at decadal scales that correspond with the variability of the PDO. For example, there was a sudden decrease in SST in the late 1970s (shift to positive PDO) and an increase in SST (negative PDO) during the mid to late 1980s (Senjyu and Watanabe 2004). Note that the positive phase of the PDO corresponds to cool conditions on the western side of the Pacific and warm on the eastern side and vice versa in its negative phase. During the cool period around Japan in the late 1970s and early- to mid-1980s, the annual catch of Japanese common squid decreased and only began to

Box 2.3, Figure 1 Annual fluctuations in the catch by Korea and Japan of the Japanese common squid and periods of cold and warm regimes. (Adapted from Sakurai et al. Changes in inferred spawning areas of Todarodes pacificus (cephalopoda: ommastrephidae) due to changing enivronmental conditions. ICES Journal of Marine Science, 57, 24–30.)

Box 2.3, Figure 2 The potential distribution of spawning squid during 2 cold years (1984 and 1988) and 2 warm years (1989 and 1994) based on temperature preference. The spawning locations are denoted by the grey areas. (Taken from Sakurai 2007.)

increase after the shift to the warm period (Box 2.3, Fig. 1; Sakurai et al. 2000, 2002, 2003). The catch is considered a good proxy for abundance and hence reproductive success (Choi et al. 2008). One reason for the increased larval survival during warm periods might be related to an increase in the size of the spawning areas. In the warm periods, the winter spawning area extends northward into the Tsushima Strait and farther eastward to the continental slope (Box 2.3, Fig. 2). For the latter, the inner flow of the Kuroshio helps transport the hatchlings into the Oyashio region to feed. Thus, during warm periods more squid may be transported to the feeding grounds via this circulation. During cool periods there is less spawning along the continental slope and the squid are deeper due to an increased mixed-layer depth caused by stronger winds. This may result in less squid being transported into the Oyashio region. Fluctuations in the strength of the northward flowing Tsushima Warm Current influences the area of the warm water region in the Sea of Japan and also matches the variability in squid catches from this region (Choi et al. 2008). The current's effect on squid is believed to be through its prey, as the changes of annual zooplankton biomass are also closely connected to the fluctuations in the Tsushima Warm Current. This too may contribute to the increased survival of the Japanese common squid in the Sea of Japan during warm years.

Yasunori Sakurai

One of the consequences of distributional shifts of individual species is the possibility of changes in the composition of ecosystem assemblages, as observed in the fish community around Britain (Attrill and Power 2002; Genner et al. 2004) and in the intertidal regions off California (Helmuth et al. 2006). These changes can occur due to differential rates of movement of various species in response to ocean climate conditions (Mueter and Litzow 2008), through invasion of new species (Stachowicz et al. 2002), or when some species disappear either due to the new temperatures exceeding the species thermal tolerance, a reduction of their prey, or an increase in their predators or competitors. These changes in community structure can also result in changes in ecosystem function, that is, who is feeding on whom.

2.6.2 Vertical stratification and mixed-layer depth

The intensity of the vertical density stratification and the depth of the mixed layer have important implications for the marine ecosystem, especially primary production. Primary production depends on the amount of solar radiation and the availability of nutrients, both of which can vary with the intensity of upper layer stratification and the mixed-layer depth. Strong stratification tends to favour a pelagic-dominated system, where energy is recycled within the upper layers, while weaker stratification favours (p.26) a more demersal or benthic-dominated system, especially on the continental shelves (Frank et al. 1990). Arguably, the most dramatic ecosystem effects of changes in stratification and mixed-layer depth occurs during El Niño events. These events result in a deepening of the thermocline and an intensification of the stratification between the surface and subsurface layers, as well as rapid warming of the waters off western South America. As a consequence of the increase in stratification, there is much less upwelling of nutrients. The directly driven biological responses include a decrease in the local primary production, collapses of a variety of small planktonic herbivore and low-trophic-level carnivore populations, and dramatic declines of the normally dominant Peruvian anchoveta population (Overland et al. 2010). Once an El Niño event has subsided, conditions begin to return to normal, although sequences of transient responses of varying duration can carry effects forward into ensuing years (Overland et al. 2010). Such a process of transient responses to frequent El Niño events may represent a continuous ‘resetting’ of the Peru system by recurring El Niño episodes before internal non-linear feedback responses can arise, possibly switching the system into a different state (Overland et al. 2010). On the opposite side of the tropical Pacific during an El Niño, the thermocline shallows resulting in a vertical extension of the mid-depth temperature habitats of yellowfin and bigeye tuna, changes that are favourable for the tuna fisheries (Lehodey 2004; see Box 2.4). During La Niña (opposite phase to an El Niño), the reverse patterns prevail on both sides of the Pacific. These ENSO patterns also seriously impact the ecosystem in the northern portion of the North Pacific subtropical gyre. There, a La Niña tends to result in an increase in vertical stratification during winter leading to a drop in nutrients and subsequently plankton productivity (Polovina et al. 2008; Ottersen et al. 2010).

(p.27)

Box 2.4 Tuna in the equatorial Pacific

Fisheries for tunas and tuna-like species have existed for centuries and now extend throughout the tropical and temperate regions of the oceans and seas using multiple types of fishing gear. The tropical species of skipjack (Katsuwonus pelamis) and yellowfin (Thunnus albacares) tuna provide the bulk of the world tuna catch and are exploited by surface fishing gear. Both species are characterized by rapid growth, early age-at-first maturity (~9–15 months), and a short to medium lifespan (~4–6 and ~6–8 years for skipjack and yellowfin, respectively). Bigeye tuna (Thunnus obesus) have a wider habitat distribution (from tropical to subtropical and temperate latitudes), a longer lifespan (>10 years), and later age at maturity (2–3 years). Adult yellowfin and bigeye are exploited by subsurface longline fisheries for fresh fish and sashimi markets.

Climate variability affects the migration and abundance of tropical tuna as well as their catchability. For example, the spatial distribution of purse seine catches of tuna in the equatorial Pacific clearly shifts during different phases of the ENSO (Box 2.4, Fig. 1; Lehodey 2000; Lehodey et al. 1997). Catches of skipjack tuna are highest within the western equatorial Pacific warm pool (surface temperatures > 29°C). During the La Niña phase of the ENSO, the easterly trade winds over the Pacific confine the warm pool to the western Pacific. At such times the purse seine fleet is mainly distributed west of 165°E where the catch per unit effort (CPUE) of skipjack and juvenile yellowfin tuna is high. With the development of an El Niño, the trade winds relax, and the warm pool spreads eastward towards the central Pacific (McPhaden and Picaut 1990). Tagging data show that the tuna follow the warm water eastward (Lehodey et al. 1997). The fishing fleet in turn follows the tuna and, although usually found west of the Date Line (180° meridian), significant numbers of tuna were caught east of the Date Line during the most intense El Niño (1997–8).

El Niño events have a strong effect on the vertical structure of the equatorial Pacific Ocean,

Box 2.4, Figure 1 ENSO and movements of skipjack tuna in the Pacific Ocean. Left: Observed movements of tagged skipjack during El Niño and La Niña phases. (Redrawn from Lehodey et al. 1997.) Tagging data were compiled from records of a large-scale tagging programme carried out by the Secretariat of the Pacific Community. Right: Predicted distribution of biomass of skipjack (age cohort 9 months) in November 1997 (El Niño phase) and November 1998 (La Niña phase) in the Pacific Ocean. Circles and arrows represent random (diffusion) and directed (advection) movements of population density correspondingly and averaged by 10 degree squares. (Redrawn from Lehodey et al. 2008 with kind permission from Elsevier). (See Plate 7).

in particular a deepening of the thermocline in the eastern Pacific during El Niño events and a rising in the west. As skipjack and yellowfin tuna tend to spread vertically over the entire upper mixed-layer depth, a deepening of the mixed layer reduces the concentrations of fish. This reduction is considered to be one of the factors responsible for the drop in catch rate of yellowfin by the eastern Pacific fleets during the 1982–3 El Niño and that led the US fleet to move into the western central Pacific, where the thermocline tends to rise during El Niño events and the mixed layer is shallow. Analyses of the CPUE time series in this region (Lehodey 2000) showed higher catch rates of yellowfin by surface gear (pole and line and purse seine) during the El Niño phases, which was attributed to increased concentrations of fish through the contraction of the vertical habitat of this species. The displacement of the United States purse seine fleet triggered by the powerful 1982/3 El Niño coincided with the development of the industrial purse seine fishery in the western central Pacific that is now the most productive tuna fishing ground of the World Ocean.

Thus, ENSO influences tuna species and their fisheries in the tropical Pacific but the impacts show different effects depending on the biology (e.g. lifespan) and ecology (e.g. habitat temperature) of the species, as well as the type of fishing gear used. A direct positive effect due to the vertical change in the thermal structure during El Niño decreases yellowfin catch rates of purse seiners in the eastern Pacific. For skipjack, large fluctuations in the catch and catch rates are driven by environmentally-induced changes in stock size with a short delay, and by horizontal spatial extension (El Niño) or contraction (La Niña) of the habitat.

Patrick Lehodey

(p.28) 2.6.3 Sea ice

In polar regions, changes in air and ocean temperatures produce variability in the seasonal abundance of sea ice, which in turn impacts their ecosystems. The timing of the ice retreat and its associated melt water affects the timing intensity, speciation, and fate of the spring bloom (Hunt et al. 2002; also see detailed discussion in Section 2.7.1). For example, sea-ice algae can produce intense blooms before zooplankton have developed, resulting in much of the production sinking out of the pelagic zone onto the ocean floor for use by the benthos (Carroll and Carroll 2003; Grebmeier et al. 2004). Along the Antarctic Peninsula, cold years tend to result in a krill-dominated ecosystem, benefiting from overwintering under the abundant sea ice. In contrast, warm years with limited sea ice result in fewer krill and a salp-dominated ecosystem (Atkinson et al. 2004). Sea ice also provides necessary habitat for many species of marine mammals, including several species of seals for breeding and protection of their young, walrus for resting areas, and polar bears for hunting and resting. Significant reductions in sea ice-coverage can pose hardships on these animals (Loeng et al. 2005).

2.6.4 Turbulence

Turbulence increases the contact rate between plankton predators and prey (Rothschild and Osborn 1988), which in turn can increase their feeding rates (MacKenzie and Leggett 1991; Sundby et al. 1994; Saiz and Kiørboe 1995). At the same time, small-scale turbulence can change the behaviour of copepods between ‘slow’ and ‘fast’ swimming speeds (Costello et al. 1990; Margalef 1997). A parabolic relationship between feeding efficiency and turbulence exists with the shape and threshold limits dependent upon the species, their stage, and also the region (Cury and Roy 1989). Higher turbulence levels can also cause an increase in total metabolism (Alcaraz et al. 1994) and wind-induced turbulence and high temperatures during winter can reinforce the energetic imbalance of copepods and potentially decrease fecundity.

2.6.5 Advection

Dispersion of fish eggs and larvae from their spawning ground is considered a key aspect of recruitment success in several fish stocks as currents may carry them into or away from favourable nursery areas. Numerical models are well suited for the study of transport of fish larvae, for example, for gadoids on Georges Bank (Werner et al. 1993), in the North Sea (Gallego et al. 1999), in the Baltic (Voss et al. 1999; Hinrichsen et al. 2003), and in the Barents Sea (Vikebø et al. 2005) and for flatfish in the Bering Sea (Wilderbuer et al. 2002). These studies indicate that recruitment success has a (p.29) strong dependency on wind-dependent drift. Advection through entrainment by Gulf Stream rings also has been shown statistically to lead to reduced recruitment levels of several groundfish stocks off the eastern United States and Canada (Myers and Drinkwater 1989). This is believed to be due to increased mortality of the larvae in the offshore waters through insufficient food quantity or quality or increased predation pressures (Drinkwater et al. 2000).

Zooplankton are also advected, leading to important ecological consequences. On the Faroe Plateau the transport of adequate numbers of C. finmarchicus onto the Faroese Shelf regions is required to ensure high recruitment of Atlantic cod (Hansen et al. 1994; Steingrund and Gaard 2005). The zooplankton overwinter in the deep waters of the Norwegian Sea Basin, rise in the spring and are then carried by the flows through the Faroes-Shetland Channel and the Faroe Bank Channel before reaching the vicinity of the Faroe Plateau (Gaard 1996). Similarly, advection of C. finmarchicus from the Norwegian Sea into the Barents Sea by the Atlantic water inflow impacts ecological production of phytoplankton and fish in that region (Skjoldal et al. 1987; Sakshaug 1997; Sundby 2000; Ottersen and Stenseth 2001). The strength of the Atlantic inflow is related to both local and the large-scale wind patterns.

Modification in circulation patterns can produce significant shifts in water mass distributions, which in turn can influence ecosystem dynamics. For example, the abundance of C. finmarchicus off northern Iceland has been linked to interchanges in the distribution of polar, Arctic, and Atlantic water masses. Higher C. finmarchicus abundance is generally associated with increased presence of higher nutrient Atlantic waters. The connection may be either through increased levels of associated primary production brought about from the increase in nutrient concentrations, direct transport of C. finmarchicus with the Atlantic water mass, or a combination of these (Ástthórsson and Gislason 1995).

Benthos also has been observed to undergo distributional changes in response to water mass shifts. Increased Atlantic water flow during the warm period of the 1920s to the 1960s resulted in the northward extension of its associated benthic fauna and flora by 500 km along western Spitzbergen (Blacker 1957) and eastward into the Barents Sea (Nesis 1960; Cushing 1982).

Previous sections have demonstrated the diverse and multiple impacts of climate on marine ecosystem processes. In this section we provide further examples, with an emphasis on insights gained using the comparative approach applied by GLOBEC.

2.7.1 Subarctic ecosystems

The subarctic ecosystems of the Bering (Alaska) and Barents (Norway) Seas are strongly impacted by climate, but there are significant differences in how they respond to climate forcing (Mueter et al. 2009), including different responses to sea-ice variability and possibly to differences in food web structures (e.g. Ciannelli et al. 2004).

Sea ice is a crucial aspect of the physical environment of the continental shelves in both the Bering and Barents Seas. The Bering Sea is typically ice-free from June through to the autumn when cold arctic winds cool the water and ice begins to form, first in the north and western shelves and later in the east (Pavlov and Pavlov 1996). Throughout winter, prevailing winds advect the ice southward into warmer water where it melts, cooling and freshening the seawater (Pease 1980; Niebauer et al. 1999). The date of the retreat of the sea ice, as well as the timing of the last major winter storms, determine the timing of spring primary production (Sambroto et al. 1986; Stabeno et al. 1998, 2001; Eslinger and Iverson 2001) and the ambient water temperatures in which grazers of the bloom must forage. In a typical light-ice year, the ice retreat occurs prior to mid-March, at which time there is insufficient light within the water column to support net primary production (Fig. 2.1; Hunt et al. 2002). The spring bloom is delayed until May or June, after winter winds have ceased and thermal stratification stabilizes the water column (Stabeno et al. 1998, 2001; Eslinger and Iverson 2001). In a heavy-ice year, melting is usually delayed until April or May and the ice-associated bloom occurs immediately after ice melt and hence earlier than in light-ice years. Although wind (p.30)

Figure 2.1 The relationship between the timing of the ice retreat and the spring bloom. When the ice retreat is early, the bloom occurs later, the water is warm, a large copepod biomass develops, and a pelagic food web is favoured (top panel). When the ice retreat is late, the bloom occurs earlier, the water is cold, much of the production goes into a benthic food web and the copepod biomass is small. (Modified from Hunt et al., 2002 with kind permission from Elsevier; F. Mueter, University of Alaska, Fairbanks, personal communication). (See Plate 3).

The late bloom under warm conditions leads to increased abundance of small neritic zooplankton (copepod) species while the early ice-associated bloom favours the larger shelf copepod species, Calanus marshallae (Baier and Napp 2003). The timing and ice-association of the bloom also influences the fate of carbon between the pelagic and benthic components (Walsh and McRoy 1986; Alexander et al. 1996; Mueter et al. 2007). Under an early bloom, there is less zooplankton to crop the phytoplankton production and thus a significant portion of this production falls to the bottom of the ocean resulting in more food for the benthos (Fig. 2.1). On the other hand, if the bloom is late the primary production is mainly consumed by the zooplankton community and less makes it to the bottom. This hypothesis is supported by the inverse relationship between survival anomalies of walleye pollock (Theragra chalcogramma) whose young feed on zooplankton, and of yellowfin sole (Limanda aspera) that feed on the benthos (Mueter et al. 2007). In light-ice years with late blooms, more production remains in the upper layers, the populations of small copepod species increase and early stages of pollock thrive, while in heavy-ice years with earlier blooms, more of the production makes its way to the benthos and there are more yellowfin sole. However, in the very warm years with little ice, the large copepod, C. marshallae, (p.31) has reduced recruitment, as is apparently also true for the shelf euphausiid, Thysanoessa raschii (Coyle et al. 2008; Hunt et al. 2008; Moss et al. 2009). Under these circumstances, in summer middle shelf populations of pollock of all ages lack zooplankton prey, and cannibalism of smaller pollock is increased. Thus an initial bottom-up limitation results in a subsequent top-down limitation of pollock recruitment. Survival of age-0 pollock is further limited in warm years because they invest more energy in growth than in storage, and therefore enter winter with very low energy supplies, leading to overwinter mortality (Moss et al. 2009).

In the Barents Sea, which lies approximately 15° of latitude farther north than the Bering Sea, approximately 40% is ice covered on an annual basis but with extensive seasonal variability (Loeng 1979; Vinje and Kvambekk 1991). Minimum ice coverage occurs in August/September but by the end of October new ice begins to form, eventually leading to peak coverage in March or April, although in some years not until early June. Approximately 60% of the sea is covered at the time of maximum extent, mainly in the northern and eastern regions. The interannual variability of ice coverage in the Barents Sea generally reflects local temperature conditions. In contrast to the Bering Sea, phytoplankton blooms in the northern Barents Sea tend to coincide with the disappearance of the ice, resulting in a northward progressing spring bloom, typically during June and July (Rey and Loeng 1985; Wassmann et al. 1999). The differences between ecosystems will likely depend on the relative timings of the increase in solar radiation, decline in wind strength (i.e. the number and intensity of the storms), and melting of sea ice. However, similar to the Bering Sea in cold years, a significant portion of the phytoplankton bloom in the seasonally ice-covered Arctic waters settles to the seabed (Sakshaug and Skjodal 1989; Sakshaug 1997). During extremely cold winters in the Barents Sea there is often a more southern distribution of ice than usual. If it extends far enough south to reach the warm Atlantic waters, it will melt and the resultant stratification will tend to initiate an earlier-than-usual phytoplankton bloom in these areas (Rey et al. 1987). As in the Bering, early blooms usually result in lower secondary production.

Another difference between the two seas is that in eastern shelf of the Bering Sea, most zooplankton are recruited from local populations, whereas in the Barents Sea advection plays an important role in the abundance of zooplankton. In the northern Barents Sea, the dominant copepods are Arctic species, which are advected in Arctic water masses, whereas in the southern Barents Sea, the advection of zooplankton, mainly C. finmarchicus, in Atlantic water plays an important role in their productivity, and ultimately the abundance of higher trophic levels (Skjoldal et al. 1992; Sundby 2000; Wassmann et al. 2006).

The Barents Sea ecosystem also differs from that of the eastern Bering Sea in that the upper trophic levels in the former appear to be strongly influenced by downward cascading trophic effects, that is, as defined by alternating patterns of abundance across more than one trophic level. Herring and cod prey upon capelin (Hjermann et al. 2004), and capelin heavily impact their zooplankton prey (Hassel et al. 1991). Since the mid-1960s, there have been alternations between periods when herring abundance was high and capelin abundance was low and vice versa. When capelin stocks were low, large zooplankton such as krill and amphipods increased to up to 10 times their former abundance owing to a release of predation pressure (Skjoldal and Rey 1989; Dalpadado and Skjoldal 1996; Mueter et al. 2009).

2.7.2 Upwelling regions

Small pelagic fishes (sardine and anchovy) in many of the large-scale upwelling regions of the world are one of the most remarkable examples of marine populations showing low frequency environmentally-driven multidecadal biomass variations. Kawasaki (1983) first noticed that catch time series of the world's largest populations of sardine, Sardinops sagax, from the Pacific (north-east, south-east, and north-west) Ocean showed large synchronous fluctuations. The Scientific Committee on Oceanic Research (SCOR) Working Group (WG) 98 investigating ‘Worldwide Large-Scale Fluctuations of Sardine and Anchovy Populations’ adopted the term ‘Regime Problem’ to describe an alternative to the hypotheses that the variability in fish stocks was (p.32) caused either by the fisheries or by recruitment-induced variability. WG98 documented that some small pelagic fish populations grow for a period of 20 to 30 years and then collapse for similar periods until practically disappearing. Further, low sardine abundance regimes have often been marked by dramatic increases in anchovy populations (Fig. 2.2), and similar but opposite phase fluctuations were apparent in the Benguela system in the Atlantic Ocean (high anchovy abundance during sardine periods in the Pacific, and vice versa), suggesting a worldwide, rather than a basin-scale phenomena (Lluch-Belda et al. 1989, 1992).

A permanent concern in fisheries science is the use of catch as a stand-in for biomass, and no catch-independent biomass reconstructions exist for all upwelling systems over the same periods. Some independent evidence suggests that large multidecadal fluctuations, and the alternation between

Figure 2.2 Historical catch series of sardine and anchovies from the main worldwide fishing systems and regime indicator series (RIS), a proxy of the synchronic variability.

The underlying mechanisms of these alternating large-scale fluctuations remain a mystery, but several hypotheses have been proposed including ecological mechanisms, direct physical forcing, and recently some integrative frameworks. Ecological mechanisms include both top-down (predatory outbreaks) and bottom-up (primary productivity) trophic controls, sudden reductions in predation pressure resulting from environmental shifts that allow rapid increases of small pelagics (environmental ‘loopholes’; Bakun and Broad 2003), competition between species (Matsuda et al. 1992), and the ‘school-mix feedback’ hypothesis to explain alternation between species and cycling though the schooling behaviour (Bakun 2001; see Moloney et al., Chapter 7, this volume). While some of these mechanisms might be operating, they are extremely hard to test, and accounting for longer than interannual variability and synchrony between regions is problematic.

The first physical variable to be related to the observed fluctuations in small pelagics was temperature (Lluch-Belda et al. 1989), and detailed thermal habitat analyses were carried out for some of the systems and periods (e.g. Lluch-Belda et al. (p.33) 1991; Hardman-Mountford et al. 2003). Takasuka et al. (2007) proposed that the dominant mechanism of fish regime shifts may be through direct temperature impacts on survival during early life stages. They documented that in the western North Pacific the temperature optimum range for sardine is colder and wider than that of anchovy, the opposite to what happens in the eastern North Pacific (Lluch-Belda et al. 1991).

Habitat expansion and contraction have been related to abundance fluctuations. For many years it was thought that when the Japanese sardine population increase, they expand offshore, whereas when they decrease, they are confined to a more coastal distribution (Kawasaki 1983; Watanabe et al. 1996). In contrast, in California and the other eastern boundary systems, sardine population growth coincides with latitudinal expansion poleward, and declines with area contraction equatorwards (Lluch-Belda et al. 1992). This notion was re-examined by Logerwell and Smith (2001) who postulated two principal spawning habitats in the California Current: inshore and offshore, the former resulting in continuing moderate sized cohorts, and the latter associated with oceanic eddies and fronts, allowing sporadic occurrence of large size cohorts. Recently, Barange et al. (2009) demonstrated that anchovy and sardine in the Benguela, California, Humbodt, and Kuroshio regions have a positive relationship between stock abundance and distributional area, but suggested that habitat availability may not be a prerequisite for sardine growth in some areas, while anchovy may require habitat to become available for its population to grow.

Ocean dynamics and large-scale circulation patterns have also been suggested as influencing the sardine populations. Lluch-Cota et al. (1997) proposed a hypothetical mechanism where shifts in large-scale ocean circulation and global average temperature conditions would drive low-frequency small pelagic regimes, with strong ENSO events as the possible trigger to shift from one regime to another. Support for this idea was provided by Parrish et al. (2000) who found relationships between the mid-latitude wind stresses and the major climate variations in the North Pacific and to the decadal variations in several fisheries, including sardines and anchovies.

Optimistically, we should see significant progress in the next few years in understanding the low-frequency fluctuations in sardine and anchovy populations and their synchronous behaviour based on available databases, modelling capabilities, and several interdisciplinary analysis and integration efforts presently underway.

2.7.3 Atlantic cod

The above examples have compared ecosystem responses to climate variability between similar types of ecosystems. However, important insights have also been gained within GLOBEC by comparing the response of a particular species within different oceanographic conditions. One of the primary species that has been studied is Atlantic cod (Gadus morhua), which inhabits the continental shelves of the North Atlantic, including those off the north-eastern United States, Canada, West Greenland, Iceland, in the Barents Sea and south to the North Sea, the Irish Sea, and the Celtic Sea. One of the best studied of all of the world's fish species, evidence abounds that climate strongly influences cod abundance.

Fish recruitment is the number of young that survive through to the fishery and is typically measured for cod at age 1 to 3. The recruitment response as a function of temperature shows that some cod stocks exhibit increasing recruitment while others show decreasing recruitment or no relationship. By comparing the different cod stocks it was found that the temperature-recruitment relationship depended upon the mean bottom temperature the cod occupied. Increasing recruitment with increasing temperatures occurred for cold water stocks (mean annual bottom temperatures of approximately 0–5°C), decreasing recruitment for warm-water stocks (>8°C), and no relationship at intermediate temperatures (Ottersen 1996; Planque and Fredou 1999; Drinkwater 2005).

At multidecadal time scales, as expressed by the Atlantic Meridional Circulation (AMO; see Box 2.1), temperature has clear influences on the distribution, abundance and growth of Atlantic cod. For example, as temperatures warmed during the 1920s and 1930s, cod along West Greenland extended their distribution approximately 1,200 km northwards and (p.34)

Figure 2.3 The relative size of an average 4-year-old Atlantic cod for different stocks as a function of their mean bottom temperature. (Drinkwater 2000 from data by Brander 1994.)

During the warm phase of the AMO (1930s–1960s) recruitment in the northern range of the cod's distribution (Barents Sea, Icelandic waters, and off West Greenland) was the highest on record (Drinkwater 2006). Also, since the growth of cod is temperature-dependent (Fig. 2.3), growth rates of individual cod increased throughout these northern regions during the warm period. For example, the weight of spawning cod off Norway increased, on average, by over 50% between the early 1900s and the 1950s (Drinkwater 2006). The successful recruitment and improved growth were linked not only to the warm temperatures but also to accompanying increases in phytoplankton and zooplankton production (Drinkwater 2006).

At decadal time scales, several cod stocks respond to the variability in the NAO. Ottersen and Stenseth (2001) demonstrated a positive association between the NAO and Barents Sea cod recruitment that accounted for 53% of the recruitment variability. The mechanistic link was thought to be through effects on regional sea temperatures and food availability. With a higher NAO index there is increased Atlantic inflow, which transports warmer waters and more C. finmarchicus into the Barents Sea and hence more food for cod. On the other hand, the year-class strength of cod in the much warmer North and Irish Seas is negatively related to the NAO index, which is believed to result from a limitation in energy resources necessary to achieve higher metabolic rates during NAO-induced warm years (Planque and Fox 1998). Brander and Mohn (2004) found that the NAO has a significant positive effect on cod recruitment in the North, Baltic, and Irish Seas, and a negative effect at Iceland. Later, Brander (2005) showed that the effects on these four stocks plus the Baltic and west of Scotland stocks were significant only when the stock biomass is low. Growth rates also vary with the NAO. For example, the variability in the growth rate of northern cod off Labrador and Newfoundland is negatively associated with the NAO (Drinkwater 2002). The NAO accounted for 66% of the variance in the weight gain between ages 3 to 5 of the northern cod over the years 1980 to 1995 and is believed to be related through NAO influences on temperature and food, similar to the case in the Barents Sea.

2.7.4 Pacific salmon

Pacific salmon have also received attention from GLOBEC, with comparisons made between the dynamics of different salmon species and their response to climate forcing. Salmon spawn in freshwater, where they stay for up to 3 years, before moving into the marine environment. They remain in the ocean for 1 to 5 years depending on the species, during which time they mature sexually. They then return to their natal river to spawn and all Pacific species die immediately after spawning. Catches of salmon species in Alaskan waters (principally pink, Onchorhynchus gorbuscha, and sockeye, O. nerka) vary synchronously but inversely with stocks along the west coast of the United States, most notably chinook, O. tshawytscha, and coho, O. kisutch (Mantua et al. 1997). Salmon production is linked to variability in the PDO, with higher production in Alaska during the positive phase of the PDO (Mantua et al. 1997), when SSTs along the periphery of the north-east Pacific are relatively warm. Greater salmon production along the west coast of the United States coincides with the (p.35)

Figure 2.4 North Pacific salmon catch time series. Black (grey) bars denote catches greater (less) than the long-term medium for Alaska stocks. The shading convention is reversed for Washington-Oregon stocks (lower two panels). Vertical dotted lines denote regime shifts in the Pacific Decadal Oscillation (PDO; top panel) in 1925, 1947, and 1976. Alaska (Washington, Oregon) catches are higher during the positive (negative) phase of the PDO. (Figure from Mantua et al. 1997 © American Meteorological Society. Reprinted with permission.)

Gargett (1997) suggested the fishery responses in the late 1970s climate regime shift were due to bottom-up forcing by changes in nutrient availability that led to increased biological productivity at multiple trophic levels. Adjustments between the atmosphere and upper ocean due to heat exchange and wind-driven ocean currents contributed to the anomalously warm SSTs extending from the Bering Sea and Gulf of Alaska along the North American coast (Parrish et al. 2000). While basin-scale climate (p.36) variations set the conditions for the decadal regime shifts, the ecosystem responded to local and regional perturbations in the environment. Adjustments in ocean mixing and basin circulation resulted in a 20–30% shallower mixed layer in the subarctic North Pacific after the 1970s, contributing to greater water column stability. This enhanced primary production and doubled the zooplankton biomass in the Gulf of Alaska (Brodeur and Ware 1992). The coincident deepening and strengthening of the thermocline and nutricline in the California Current (Palacios et al. 2004), reduced the availability of nutrients to the euphotic zone and led to a sharp decline in local zooplankton biomass (Roemmich and McGowan 1995). This contributed to the decline in salmon along the west coast of the United States. Peterson and Schwing (2003) suggested that this west coast salmon production shifted back to higher levels after 1997, coincident with a return to the negative PDO phase and higher ecosystem productivity in the California Current. This lasted up until 2002, after which conditions in the California Current have undergone year-to-year variability with no consistent periods of good or bad years for west coast salmon.