Anomalous Ocean Conditions May Explain the Recent Extreme Variability in Fraser River Sockeye Salmon Production

Abstract Record low returns of sockeye salmon Oncorhynchus nerka to the Fraser River in 2009 were followed by record high returns to the river in 2010, providing an unprecedented opportunity to examine links between oceanic factors and the survival of Pacific salmon stocks. The low returns in 2009 indicated poor early marine survival of juvenile sockeye salmon in 2007. The poor survival was likely due to low food levels arising from unfavorable wind and runoff conditions in the Strait of Georgia and the Queen Charlotte Sound-Hecate Strait region in the spring of 2007. Conversely, the high returns in 2010 were associated with a large smolt output from the Fraser River and good early marine survival in 2008. This enhanced survival was likely associated with adequate food levels arising from favorable oceanic conditions in the Strait of Georgia and the Queen Charlotte Sound-Hecate Strait region in the spring of 2008. We speculate that ocean factors during the subsequent marine years also affected brood year strength. Specifically, the back-to-back La Nina winters of 2007–2008 and 2008–2009 would have negatively influenced the survivability of the 2007 entry stocks, while the El Niño winter of 2009–2010 would have positively affected the survivability of the 2008 entry stocks. We conclude that poor early marine survival leads to low production. However, if large numbers of healthy fish survive the early marine entry, and if conditions during at least one of the two ocean winters in the Gulf of Alaska are favorable to stock survivability, then returns to the river can be high.

salmon spawning populations have been identified in the Fraser River drainage system. Juveniles rear in about 24 nursery lakes, although most (about 90%) of the production is centered on fewer than 10 nursery lakes. Individual Fraser River sockeye salmon populations have characteristic timings when adults return to spawn and are broadly classified into four groups or "runs" for management purposes: early Stuart, early Summer, summer, and late (Gable and Cox-Rogers 1993).
Most juvenile sockeye salmon from the Fraser River leave their nursery lakes 1 year after they emerge from the gravel redd. They then enter the Strait of Georgia ( Figure 1) over a 2-month period from late March until late May (Preikshot et al. 2012, this volume) and subsequently move northward through the Strait of Georgia over 5-6 weeks from roughly mid-May to mid-July (Beamish et al. 2012, this volume;Preikshot et al. 2012). Juveniles leave the Strait of Georgia to the north through the "Inside Passage" that takes them through Johnstone Strait, Queen Charlotte Strait, Queen Charlotte Sound, and Hecate Strait (Groot and Cooke 1987;Tucker et al. 2009;Welch et al. 2009Welch et al. , 2011. In the summer and fall, the juveniles move offshore and eventually into the Gulf of Alaska (Hartt and Dell 1986;Tucker et al. 2009), where they remain until they begin their return migration to the Fraser River in the spring and summer of their second year in the ocean. As a consequence of this life history, adult FIGURE 2. Total returns of sockeye salmon to the Fraser River. Dark bars indicate years when late-run Adams River stocks made a major contribution to the total salmon return. Return years 2007Return years , 2008Return years , and 2009Return years (brood years 2003Return years , 2004Return years , and 2005 were the lowest on record. The highest recorded return was in 2010. sockeye salmon have a dominant age of four years and stocks undergo persistent cycles of abundance within this 4-year pattern ( Figure 2). The present-day migration northward out of the Strait of Georgia may represent a change in direction from the westward migration through Juan de Fuca Strait postulated earlier by Healey (1980). This change would have occurred sometime in the late 1970s and could be related to the 1977 oceanic "regime shift" in the North Pacific (Preikshot et al. 2012).
As with other adult sockeye salmon, total returns of Fraser River stocks are highly variable throughout their distribution due, in part, to different production cycles and levels of marine survival. Ricker (1987) proposed that total "big year" returns in the early 1900s may have been around 100 million fish; other years presumably had substantially fewer returning fish. In the past 60 years, total annual returns have averaged approximately 7.8 million fish, with those in the big years (or dominant years, as they are now called) ranging from about 3.4 million to 22 million fish. This marked variability in adult returns is likely related to fishing and natural cycles (there is no consensus on the reasons for these cycles) and to the effects of climate. The most extreme variability in adult returns occurred between years 2009 and 2010, corresponding to the 2005 and 2006 brood years and the 2007 and 2008 ocean entry years, respectively. For 2009 and 2010, the expected returns were in the range of 10-11 million fish, whereas the actual returns were about 1.4 million and 29.6 million fish, respectively. The poor returns in the fall of 2009 led to a Canadian federal commission of inquiry (www.cohencommission.ca) while the good returns in the fall of 2010 (which included dominant-year late-run Adams River fish) established a record high for returning Fraser River stocks.
In this study, we show that the extreme, and unexpected, difference between the 2009 and 2010 sockeye salmon returns to the Fraser River was likely a consequence of several major factors, including environmental conditions in coastal British Columbia and the Gulf of Alaska, the freshwater production of smolts, and a life history strategy that requires that juveniles grow rapidly when they first enter the ocean. Our work extends the "critical size-critical period" hypothesis of Beamish and Mahnken (2001), which proposes that Pacific salmon brood year strength is determined in two stages during the first year in the ocean. According to this hypothesis, there is a large early-marine mortality that occurs shortly after juvenile salmon enter the ocean. This is followed by a physiologically based mortality during the first ocean winter that affects individuals that did not grow to a critical size and were unable to accumulate the necessary energy reserves during the previous summer and fall. The results presented here suggest that conditions in the second marine winter may also contribute to salmon survival in the Gulf of Alaska.

METHODS
Estimation of sockeye salmon smolts produced in the Fraser River drainage.-As noted in the introduction, ocean entry years 2007 and 2008-the years for which we wish to estimate the number of smolts produced in the Fraser River drainage system-correspond to spawning years 2005 and 2006 and to adult return years 2009 and 2010, respectively. The numbers of returning adults can be used to estimate the survival of the smolts to the adult stage.
Sockeye salmon smolts that enter the Strait of Georgia originate from a number of populations throughout the Fraser River drainage system. One of the major populations is from the Chilko Lake drainage (see Figure 1a for the location of the lake). Production from this system averaged 18.8% of the total returns to the Fraser River over the past 50 years (Grant et al. 2010). The number of sockeye salmon smolts leaving the drainage is estimated each year, and random samples are measured for length throughout the smolt migration out of the lake (Grant et al. 2010;T. Cone, Department of Fisheries and Oceans, personal communication). We used these estimates, and the relative proportion of the spawning population that produced the smolts to the total escapement in the Fraser River, to estimate the total number of sockeye smolts entering the Strait of Georgia from the Fraser River in 2007, 2008, and 2009. We acknowledge that determination of the number of smolts entering the strait is approximate and does not take into account freshwater mortality.
The estimated total marine survival for Chilko Lake smolts for return years 2009 and 2010 was 0.3% and 5.5%, respectively (Pacific Salmon Commission; http://www.psc.org). Assuming that these values were representative of all populations in the Fraser River drainage enables us to calculate the total sockeye salmon smolt production for the drainage region for the 2 years. A second method uses a proportional estimate of production. For example, if the sockeye salmon originating from Chilko Lake represent 10% of all adults returning to the Fraser River, then the number of smolts leaving the lake 2 years earlier would be 10% of all smolts produced. The method assumes that freshwater rearing conditions are similar 418 THOMSON ET AL. for all populations and that marine survival is similar for all populations. Although there are obvious problems with both methods, they are intended only to provide an approximate estimate of the number of smolts that entered the ocean.
Trawl surveys for juvenile Fraser River sockeye salmon.-Trawl surveys were conducted in June and July 2007, 2008 in the Strait of Georgia, Queen Charlotte Sound, and Hecate Strait. All surveys followed standard track lines, with net design and survey methodology reported in Beamish et al. (2000), Sweeting et al. (2003), andTucker et al. (2009). The modified midwater trawl net had an approximate 30-m-wide, 15-m-deep opening and was fished for 30 min at an average speed of 2.6 m/s (5 knots) with headrope depths ranging from the surface to 30 m. Most sets were in the top 15 m of the water column. Catches for sets shorter than 30 min were standardized to a 30-min catch. Sets in the Strait of Georgia fished the top 30 m, whereas those in Queen Charlotte Sound, Queen Charlotte Strait, and Hecate Strait fished only the top 15 m. As a consequence, we compare catches for the surface 15 m only. For the 3 years being considered, the only survey in Queen Charlotte Strait was in 2009. The catches and the lengths of juvenile sockeye salmon for the Queen Charlotte Strait sample in 2009 were combined with those for Queen Charlotte Sound. Samples for DNA analysis are reported separately. Average catch is defined as the sum of the catch for each set divided by the number of sets. Fork lengths were measured from a subset of the catch. The stock composition of sockeye salmon in the trawl catches was determined with microsatellite DNA using the methods of Beacham et al. (2005).
Environmental data.-The physical environmental data examined in this study include daily mean wind velocity from meteorological buoys; daily salinity and sea surface temperature (SST) from British Columbia lighthouse stations; monthly mean discharge for the Fraser River and other rivers bordering the Strait of Georgia and southern Queen Charlotte Sound (Whitfield and Spence 2011); gridded (1.9 • latitude × 1.9 • longitude) monthly mean National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR) Reanalysis-1 atmospheric sea level pressure and wind stress (Kistler et al. 2001); and gridded (1 • latitude × 1 • longitude) monthly mean Optimum Interpolation SST Version-2 data (Reynolds et al. 2002). We also examined monthly mean satellite-derived chlorophyll (ocean color) time series for British Columbia coastal waters derived from the Seaviewing Wide Field-of-View Sensor and Moderate Resolution Imaging Spectroradiometer (G. Borstad and L. Brown, ASL Environmental Sciences, Inc., 2011, personal communication). However, given the numerous gaps in the satellite data and the unknown relationship between surface chlorophyll concentration and depthintegrated production (total plankton biomass), we have left detailed investigation of these data to future study.
The meteorological time series serve as proxy variables for oceanic factors such as surface currents that could be affecting fish production (see Thomson and Hourston 2011). We use wind stress, τ = ρ air C D |U | U , rather than wind velocity (U) to characterize the effects of surface currents because it is wind stress that appears as a driving force in the Navier-Stokes equations that govern physical oceanographic dynamics (here, ρ air is the air density and C D is the drag coefficient). Wind stress is available as part of the Reanalysis data; for meteorological buoys, we have estimated wind stress from wind velocities assuming neutral stability conditions (Smith 1988). We have also determined a pseudo-mixed layer depth derived from conductivity-temperature-depth (CTD) profiles collected by the Canadian Navy at Nanoose Bay in the west-central Strait of Georgia (http://www.pac.dfo-mpo.gc.ca/science/oceans/datadonnees/search-recherche/profiles-eng.asp). We use the prefix "pseudo" here because the minimum mixed layer depth we can derive from the CTD temperature, salinity, and density profiles is 2.5 m. This is the minimum depth that can be resolved by our mixing layer detection algorithm, which uses a split-and-merge curve fitting method to detect inflection points in water property profiles (Thomson and Fine 2003). Profiles which show no surface mixed layer were specified to have a mixed layer depth of 0 m. Close examination of the profile data for the strait shows that the absence of a surface mixed layer was typically due to a high surface density gradient, coinciding with a near-zero mixed layer depth. We note that each individual profile used in the derived monthly averages was carefully examined and any erroneous data (such as those with inverted surface density gradients) were rejected.

Fraser River Sockeye Salmon Smolts Entering the Strait of Georgia in 2007 and 2008
Smolts entering the Strait of Georgia in 2007 were produced from a total spawning escapement of 3,307,950 adult fish in 2005, of which Chilko Lake sockeye salmon accounted for 15.7%. In 2007, the estimated number of sockeye smolts leaving Chilko Lake was 77,128,000, about twice the previously recorded high for this lake. When we combine the 15.7% Chilko Lake fish for the total 2005 escapement return with the observed number of smolts leaving the lake in 2007 (and assume smolt-to-adult uniformity among stocks), we obtain an estimate of 491,563,000 sockeye salmon smolts produced in the entire watershed. Our second estimate of smolt production uses the estimation of 0.3% for the total marine survival of Chilko Lake smolts that went to sea in 2007. If we assume that this survival was representative of the survival of all sockeye salmon smolts and use the total observed return of 1,443,000 adult fish in 2009, we find that approximately 416,729,000 smolts entered the ocean in 2007. The average of the two estimates, roughly 454,146,000 smolts, represents the possible number of sockeye salmon smolts produced in the Fraser River drainage in 2007.
A total of 73,000,000 sockeye salmon smolts was counted leaving Chilko Lake in 2008. As with 2007, this was roughly twice the previously recorded high. The percentage of Chilko Lake sockeye salmon in the 2006 escapement was 15.2%. Using this percentage to determine the total smolt production, we calculate that 480,565,000 sockeye salmon smolts were produced in the Fraser River drainage in 2008. Our second estimate was obtained from an estimated marine survival of 5.4% for the Chilko Lake sockeye salmon smolts that went to sea in 2008. If we assume that this marine survival is representative of all stocks, approximately 556,944,000 smolts would have needed to enter the ocean in 2008 to produce the total observed return of 29,600,000 adults in 2010. The average of the two estimates yields roughly 518,754,000 sockeye salmon smolts produced in the Fraser River drainage in 2008, or 14.2% more smolts than in 2007.
In 2009, a total of 27,515,000 smolts was estimated leaving Chilko Lake. The adult return to Chilko Lake in 2007 was 28.9% of all returns to the Fraser River. Using these values, and applying the same methodology that we used for the 2007 and 2008 entry years, we estimated that 95,009,000 smolts were produced in the Fraser River drainage in 2009. This amounts to only 21% of the number of smolts produced in 2007. (The numbers of smolts produced in 2009 could not be estimated using the second method, as estimates of the marine survival for 2011 were not available at the time of this study.)

Lengths of Juvenile Sockeye Salmon Leaving Chilko Lake
In all 3 years, the samples measured for length were collected throughout the migration out of Chilko Lake, which started in  was 81.9 mm (SD = 4.7 mm) and in 2009, 82.9 mm (SD = 4.8 mm).

Average Catches of Juveniles and Stock Composition in the Trawl Surveys (2007-2009)
In 2007, 74 trawl sets were conducted in the Strait of Georgia from July 8 to July 15. Twenty-one of these sets captured juvenile sockeye salmon and 53 had zero catch. All 65 fish were caught in the 0-15-m stratum. No individuals sampled were analyzed for stock composition. The average catch in the surface 15 m was 1.9 juvenile sockeye salmon per 30-min set ( Figure 4a).
In 2008, 90 trawl sets were conducted in the Strait of Georgia from June 27 to July 6. Forty-four of these sets captured juvenile sockeye salmon and 46 sets had zero catch. A total of 1,597 sockeye salmon (96%) were captured in the 0-15-m depth range and 65 fish (4%) in the 15-30-m depth range. The average catch in the surface layer was 34.0 sockeye salmon per 30-min set ( Figure 4a). A total of 167 fish was analyzed for stock composition, and all (100%) were from the Fraser River. The stock composition according to run timing was as follows: late, 73%; early summer, 20%; and summer, 7% ( Figure 5a). Similar compositions were obtained for the Gulf Island region ( Figure 5b). These percentages were generally similar to the expected stock percentages for the adult return for this brood year ( Figure 5e); specifically, 82% late, 9% early summer, 8% summer, and 1% early Stuart.
In 2009, 83 trawls sets were undertaken in the Strait of Georgia from June 26 to July 7. Of these, 53 sets captured juvenile sockeye salmon and 30 sets had zero catch. There were 1,354 fish (85%) captured in the surface layer (0-15 m), 218 (14%) in the 15-30-m depth range, and 16 fish (1%) in water deeper than 30 m. The average catch in the surface layer was 33.9 fish per 30-min set ( Figure 4a). A total of 141 fish was analyzed for stock composition for the Strait of Georgia and all were from the Fraser River ( Figure 6a). The stock composition according to run timing was as follows: summer, 51%; late, 30%; early summer, 18%; and early Stuart, 1% ( Figure 6a). The stock composition for the Gulf Islands ( Figure 6b) differed slightly from that for the Strait of Georgia. However, for both the open strait and the Gulf Islands, the stock composition was similar to the expected percentages for adults from this brood year, specifically, summer, 47%; late, 37%; early summer, 15%; and early Stuart, 1% ( Figure 6f).
There were 56 trawl sets in Queen Charlotte Strait, Queen Charlotte Sound, and Hecate Strait from 2007 to 2009, and all sets were conducted at the end of June or early July (Table 1)

Size of Juvenile Sockeye Salmon Entering the Strait of Georgia and in the Trawl Catches
In all years, the juvenile sockeye salmon in Hecate Strait were significantly larger than those in all other areas (Table 2; analysis of variance [ANOVA], P < 0.01). Juveniles were the smallest in Hecate Strait in 2007 (Table 2; P < 0.01) and in Queen Charlotte Sound in 2007 (Table 2; P < 0.05). The average size of juvenile sockeye salmon in Queen Charlotte Sound was larger than that in the Strait of Georgia in all years, but only significantly larger in 2008 and 2009 (Table 2; P < 0.05). The average length of juvenile sockeye salmon in the Strait of Georgia was the smallest in 2008, but not significantly smaller (Table 2; P > 0.05).

Timing of the Spring Transition
Although details vary from region to region, meteorological time series for the West Coast of British Columbia (Figure 8) revealed that return years 2007, 2008, and 2009 were associated with anomalously late reversals in the alongshore winds over southern coastal British Columbia in spring 2 years earlier, i.e., at the times of marine entry in years 2005, 2006, and 2007, respectively. (The mean ± SE for the wind reversals for this region is April 16 ± 5 d for the period for which there are meteorological buoy observations [1980 to 2011]. There is no detectable long-term trend.) In each of these 3 years, the "spring transition" from predominantly northward directed (i.e., "southerly") winds in winter to predominantly southward directed (i.e., "northerly") winds in summer (see Strub et al. 1987;Thomson and Ware 1996) was delayed by about 1 month relative to the mid-April long-term mean. Moreover, the prevailing northerly winds that followed the transition in 2007 were generally weaker than those of other years. Conversely, the high returns in 2010 were associated with an anomalously early spring transition during the 2008 marine entry year. The transition occurred over several weeks beginning around mid-January and was characterized by anomalously weak southerly winds punctuated by periods of moderately strong northerly winds. Although this variability makes it difficult to pinpoint the exact time of the 2008 spring transition, it is clear from Figure 8 that the reversal to northerly winds in 2008 occurred up to several months earlier than normal ( Table 3). The prevailing northerly wind that followed the   Figure 1). Inverted triangles mark the approximate times of the "spring transition," when the prevailing winds switch from southeasterly (to the northwest) to northwesterly (to the southeast). The mean date is April 16 ± 5 d (see text). Similar wind timing conditions occurred within the Strait of Georgia and central coastal regions of Hecate Strait and Queen Charlotte Sound. The spring transition is generally followed by a reduction in rainfall and cloud cover as well as accelerated snowmelt. Prior to the transition, the prevailing southeasterly winds in the Strait of Georgia are conducive to the northwestward transport (and hence retention) of freshwater runoff entering the strait from the Fraser River and from the numerous smaller rivers located along the East Coast of Vancouver Island and the mainland side of the basin (Figure 9). In winter, the volume of freshwater entering the Strait of Georgia from pluvial-fed rivers can be equivalent to that entering the strait from higher elevations in the Fraser River drainage system. In the spring and summer months following the spring transition, the total discharge into the strait (Figure 9a) becomes dominated by snowmelt-derived discharge from the Fraser River. Similar wind and runoff conditions apply to Queen Charlotte Sound and Hecate Strait (Figure 9b). A principal difference with the Strait of Georgia is that most of the discharge on the central British Columbia coast originates from the mainland side. As a consequence, the southerly winds that prevail prior to the transition will tend to drive the brackish surface waters entering from the mainland inlets northward in Queen Charlotte Sound (Hannah et al. 1991). Conversely, the northerly winds that follow the transition will help drive the outflow from the coastal inlets into the sound (Borstad et al. 2011).

Effect of the Spring Transition Timing on Juvenile Salmon Survival
To determine the possible contribution of early marine conditions to the extreme difference in fish production observed  (d) show the corresponding monthly mean anomalies. Anomalies were obtained by subtracting the seasonal cycle (mean monthly values) from each data series. The total discharge into the Strait of Georgia consists of the Fraser River discharge recorded at the eastern end of the Fraser River valley at Hope, that from the central and lower portions of the river downstream of Hope and its tributaries, and that from freshwater inputs from Central Vancouver Island and the British Columbia mainland. The total discharge into the Queen Charlotte Sound-Strait region is comprised of freshwater inputs from the Wannock River (mainland British Columbia) to the east of buoy 46204 in Figure 1a The anomalous wind and runoff conditions that prevailed along the British Columbia coast in early 2007 (prior to the time that juvenile Fraser River sockeye salmon were resident in the different regions) likely account for the low surface salinities observed throughout much of the region that spring and summer ( Figure 11). Freshwater retention in the Strait of Georgia also accounts for the magnitudes of the observed surface salinity anomalies (see appendix) and for the anomalously shallow pseudo-mixed layer observed at Nanoose Bay in the central sector of the strait throughout the spring and summer of 2007 ( Figure 12). The shallow mixed layer was unprecedented in its timing and duration for the period 1980 to 2010, for which we have nearly continuous data. Moreover, satellite-derived chlorophyll data for the Strait of Georgia for the period 1997 to 2010 ( Figure 13 Figure 1). Positive alongshore directions are shown in brackets in degrees true compass bearing. The cross-shore components of the wind stress are much weaker and therefore not presented. weaker entrainment of nutrient-rich water from the underlying marine layer, limiting the duration and vertical extent of upper layer primary productivity. These factors, in conjunction with possibly reduced light penetration due to higher fluvial-derived particle concentrations, may have confined plankton growth to a shallow (and therefore relatively low, depth-integrated biomass) surface layer around the time that the smolts were entering the strait. The cloudy and overcast "winter" skies that typically precede the spring transition also may have helped confine plankton productivity to a relatively shallow surface layer in the strait.  Figure 1). Panel (a) shows the mean anomalies, panel (b) the corresponding standard deviations. The anomalies were obtained by subtracting the mean monthly values for the full observation period (1969-2010) from the monthly mean values. The term "pseudo" is used to describe the mixed layer depth because estimates may include the use of CTD observations which show no indication of a true mixed layer near the surface, either because the water is too stratified or the winds too weak for a true mixed layer to form (see text for further explanation). Similar mixed layer depth values were obtained using temperature and salinity profile data separately. sustained plankton production in the spring and summer. We expect that food levels for the migrating juvenile sockeye salmon stocks would have been considerably higher than those for the 2007 juveniles. The physical factors can account for the relatively high surface salinities observed throughout the strait in the early spring of 2008 (Figure 11a), for the deep to average pseudo-wind-induced mixed layer depths observed at Nanoose Bay in the central strait for that time period (Figure 12), and the delayed spring bloom in the Strait of Georgia that year (see Collins et al. 2009; Figure 13). The flushing mechanism would have been most effective during neap tides when turbulent tidal mixing within passages through the Gulf and San Juan islands exerts minimal "hydraulic control" on water exchange between the Strait of Georgia and Juan de Fuca Strait (Griffin and LeBlond, 1990;Masson and Cummins 2000;Thomson et al. 2007).

Effect of Gulf of Alaska Conditions on Salmon Survival
Plankton productivity in the Gulf of Alaska is maintained by the vertical flux of nutrients into the upper mixed layer in winter and early spring (Brodeur and Ware 1992;Ware and McFarlane, 1989;Thomson and Fine 2009). Westward propagating mesoscale eddies originating off northwestern British Columbia and southeastern Alaska (Tabata 1982;Thomson and Gower 1998) are also capable of delivering nutrients to the region (Ladd et al. 2009). Nutrient levels diminish through the summer and are renewed in winter through wind-driven upwelling of deep (>100-m), nutrient-rich waters below the permanent pycnocline (∼100-200 m depth). Averaged over time and space, this upward winter flux of nutrients is dependent on the location and intensity of the Aleutian Low pressure system and the maximum depth of the surface mixed layer in late winter. Here, it is the curl of the wind stress associated with the cyclonic Aleutian Low and not the wind stress directly which determines the degree of upwelling (Cummins and Lagerloef 2002). In particular, the positive curl in the wind stress associated with the Aleutian Low leads to divergence of the surface wind-driven Ekman layer and an uplift of the subsurface isopycnal surfaces; a negative curl in the wind stress (such as that associated with the North Pacific High pressure system to the south) leads to convergence of the wind-generated surface Ekman layer and a depression of isopycnal surfaces.
Based on the monthly mean property distributions presented in Figures 14 and 15, Fraser River sockeye salmon from the 2007 entry year would have spent their first marine winter in the Gulf of Alaska during a period of relatively low sea surface temperatures (SSTs) and moderately unfavorable winds associated with the 2007-2008 La Niña. These factors are assumed to have had a slightly negative effect on stock survival. While SST anomalies in the gulf were again negative in the second marine winter, wind conditions in the region were highly anomalous due to the extreme La Niña winter of 2008-2009, the second most intense La Niña recorded in the Pacific Ocean. The Aleutian Low shifted to the Bering Sea and anomalously weak cyclonic winds prevailed over the Gulf of Alaska from December 2008 through April 2009 (Figure 15b). This would have led to greatly  (Figures 14c, 15c) would have supported intensified wind-driven upwelling in the Gulf of Alaska. This would have helped lift nutrients (and, possibly, overwintering plankton) higher into the water column so that they were more readily entrained in the surface mixed layer. We assume that this enhanced plankton productivity in the ensuing spring and had a positive effect on the survival of the 2008 and 2009 marine entry stocks. 430 THOMSON ET AL.

DISCUSSION
We have estimated that about 14% more sockeye salmon smolts entered the Strait of Georgia from the Fraser River in 2008 than in 2007. However, trawl survey catches of juvenile sockeye salmon were approximately 400% larger in Queen Charlotte Sound and 1,800% larger in the Strait of Georgia in 2008 than in 2007. These comparisons between years and areas are problematic because of small sample sizes in Queen Charlotte Sound and sampling dates in the Strait of Georgia. Nonetheless, it does appear that the early marine survival in the Strait of Georgia was substantially greater in 2008 than in 2007, resulting in larger catches in the Strait of Georgia and in more fish surviving to travel into Queen Charlotte Sound in 2008. The results of DNA analysis and the trawl catches in 2007, 2008, and 2009 show that juvenile sockeye salmon from the Fraser River are distributed from the Gulf Islands in the south to Hecate Strait in the north by late June and early July. This indicates that the migration out of the Strait of Georgia involves both dispersion throughout the Strait of Georgia and movement out of the strait. This pattern of migration also confirms that the patterns of juvenile sockeye salmon migration are complex and that all areas should be sampled repeatedly during the dispersal and migration phases. Until such studies are carried out, we are left to extrapolate from the existing data.
The catches of juvenile sockeye salmon in the trawls were much larger in 2008 than in 2007. Considering the relatively small (roughly 14%) difference in the smolts produced between the 2 years, we interpret the larger catches as representing much larger early marine survival. Moreover, ocean entry years 2007 and 2008 produced adult returns that were dramatically different despite this apparent similarity in total smolt production. The total return of 1,443,000 adult fish from ocean entry year 2007 (return year 2009) was the lowest in history, while the total return of 29,600,000 adult fish from ocean entry year 2008 (return year 2010) was the highest in nearly 100 years. Beamish et al. (2012) find that all juvenile Pacific salmon Oncorhynchus spp. that entered the Strait of Georgia in 2007 and survived to the trawl survey in mid-July showed signs of encountering a major shortage of prey. Coho salmon O. kisutch and Chinook salmon O. tshawytscha were the smallest in the 11 years of research surveys, resulting in the lowest condition since the trawl surveys started in 1998. These fish also had the highest percentage of empty stomachs in the 11 survey years, indicating that they were probably having difficulty finding prey. The early marine survival for coho salmon from ocean entry until mid-September in 2007 was also very poor, which most likely explains the poor return of adults in 2008   (Cleary et al. 2009;Beamish et al. 2012).
Juvenile Pacific salmon and herring represent 98% of all the fish caught in the surface 30 m of the Strait of Georgia in the daytime trawls since 1998 ). Thus, the generally poor growth and/or poor survival of all juvenile Pacific salmon and herring in 2007 is convincing evidence of a major change in the Strait of Georgia ecosystem early that year. We propose that the ocean conditions described in this report resulted in a shift to exceptionally low production of the prey normally consumed by these fish during the critical early marine period for Pacific salmon. This is consistent with the generally accepted notion that it is during this early marine period that most of the marine mortality occurs for Pacific salmon and many other species of fish (Hjort 1914;Houde 1987;Pearcy 1992;Beamish and Mahnken 2001;Quinn et al. 2005;MacFarlane 2010), except that the conditions causing the mortalities in 2007 in the Strait of Georgia were extreme and possibly unique.
The juvenile sockeye salmon that were in the Strait of Georgia in late June and early July 2008 were smaller than those in 2007.
The small size appears to be a consequence of the small ocean entry size compared with 2007. The small size of these fish could also result from an increased density of other competing species, as ocean entry year 2008 was also a year when there were large numbers of juvenile pink salmon O. gorbuscha entering the Strait of Georgia. If adjustments were made to standardize the lengths of trawl-caught fish with the ocean entry size of Chilko Lake smolts, the juvenile sockeye salmon in the Strait of Georgia in 2008 would be larger than the fish in 2007. We do not know whether using the lengths of Chilko Lake smolts in 2008 to standardize the average length of all smolts in the trawl catch would be representative of the population, but it is an indication that all juvenile sockeye salmon in 2008 grew more rapidly than the juveniles in 2007. If this standardization is representative of the population, then the juvenile Fraser River sockeye salmon that were distributed from the Strait of Georgia to Hecate Strait towards the end of June 2007 had the smallest size among the years 2007, 2008, and 2009. The exceptionally large adult return of pink salmon in 2009 of this brood year is evidence that conditions for marine survival for juvenile Pacific salmon were much improved in the spring of 2008. We note that the survival of juvenile Pacific herring was also much larger than in 2007 .
The large numbers of smolts that were produced in freshwater in 2007 and 2008 would experience some mortality during their migration down the Fraser River, but there would still be hundreds of millions that entered the Strait of Georgia. It is to be expected that there will a large initial mortality in the Strait of Georgia. However, it is during the residence time in the Strait of Georgia of approximately 43-54 d (Preikshot et al. 2012) that we propose that many of the sockeye salmon that went to sea in 2007 died or became increasingly susceptible to sources of mortality because of their poor condition. Brood year strength is likely determined early in the marine period (Beamish and Mahnken 2001), even if the actual mortality occurs later, indicating that not all of the mortality would have to occur in the Strait of Georgia. Many of the juvenile sockeye salmon that survive their residence period in the Strait of Georgia presumably do not survive through the rest of the year and following winter because of their poor condition. It is not known how long the juvenile sockeye salmon are resident in Queen Charlotte Sound and Hecate Strait, but it is known that the juveniles migrate into the Gulf of Alaska, where they spend the first ocean winter. As indicated by our analyses, the 2007 entry year fish would also have experienced poor feeding conditions in Queen Charlotte Sound, Hecate Strait, and the Gulf of Alaska during their first and second winters, thereby exacerbating the poor survival conditions experienced in the Strait of Georgia. This interpretation is supported by studies showing that reduced early marine growth is also associated with increased overwinter mortality for coho salmon and pink salmon (Beamish et al. 2004;Moss et al. 2005). Farley et al. (2011) further show that the first marine winter also may be a critical period for juvenile sockeye salmon survival in the Bering Sea; years when the juveniles had low energy density were years when the brood year had low marine survival. Rensel et al. (2010) suggest that mortalities of juvenile sockeye salmon in the Strait of Georgia in 2007 resulted from an accumulation of toxic algae. Toxic algae may have contributed to mortality in some areas of the Strait of Georgia; however, the nearly complete collapse of the 2007 Pacific herring yearclass indicates that the cause of the generally poor survival of all juvenile fish in the surface 30 m must have been prevalent throughout the Strait of Georgia. Thus, the high mortality was probably due to a major failure of the production of prey required by juvenile Pacific salmon when they begin feeding in early spring in the Strait of Georgia. Rensel et al. (2010) show that from 1994 to 2007 there was a strong correlation between the total marine survival of Chilko Lake sockeye salmon and the early marine survival of age-0 Pacific herring in the Strait of Georgia. We suggest that this is additional evidence that conditions within the Strait of Georgia have a major influence on the brood year strength of Fraser River sockeye salmon. Furthermore, the relationship would indicate that the influence occurs from about mid-May until early July, when the Chilko Lake fish are in the Strait. It is likely during this time that most of the juvenile Pacific herring died.
The almost constant average catch of juvenile Fraser River sockeye salmon in 2007, 2008, and 2009 in Hecate Strait (Figure 4b) is difficult to explain, considering the variation in average catches in the other areas. On the other hand, the average catch was small relative to that in Queen Charlotte Sound, indicating that the fish that migrated into Hecate Strait were not a major percentage of the surviving population. The fish were also consistently larger than those in Queen Charlotte Sound and the Strait of Georgia. The large size, as well as the stock composition of the fish in Hecate Strait, may be a consequence of faster growth, older ages, or both. For example, approximately 5% of Chilko Lake sockeye salmon become smolts at 2 years of age and not at 1 year like most salmon (Henderson and Cass 1991). Chilko Lake sockeye salmon were estimated using DNA analysis to comprise 22% of the Fraser River sockeye salmon caught in Hecate Strait in 2007 and 29% of the Fraser River sockeye salmon caught in 2008, higher than the average 15-16% of Fraser River drainage escapement accounted for by the Chilko Lake population. Similarly, the Cultus Lake and Birkenhead River populations display higher proportions of 2-year-old fish than is typically observed in other populations. These two populations accounted for 11% and 4% of the Fraser River sockeye salmon sampled in Hecate Strait in 2007 and 2008, respectively, higher than would be expected based upon their relative contributions to drainage escapement. It thus seems likely that the Fraser River sockeye salmon sampled in Hecate Strait in June were derived disproportionately from pop-ulations where individuals display older ages and larger sizes at smoltification. Unfortunately, as ages were not determined for any of the individuals sampled, it was not possible to confirm that the individuals sampled in Hecate Strait in late June were typically older than those sampled in the Strait of Georgia in July. We suspect that the larger fish migrated out of the Strait of Georgia faster because of their size in freshwater, but this interpretation remains speculative until the age composition of the fish in Hecate Strait is determined. It is relevant that, in 2007, the percentage of sockeye salmon from the Fraser River in Hecate Strait was the smallest in the 3 years (Table 1), possibly indicating that there were fewer juveniles entering Hecate Strait from the Strait of Georgia in 2007 than in 2008 and 2009, which is consistent with the interpretation that there was a larger early marine mortality in the Strait of Georgia in 2007.
In 2009, the estimated production of sockeye smolts in the Fraser River drainage was about one-fifth that for 2007 and 2008. Juvenile sockeye salmon in the 2009 trawl survey displayed the largest body size of the 3 years in all areas. Unlike in 2007 and 2008, average catches in the Strait of Georgia were larger than those in Queen Charlotte Sound. Because of the limited number of sets in Queen Charlotte Sound, it is difficult to interpret the relatively large catch in the Strait of Georgia, other than to suggest that there had been a higher early marine survival or that the migration out of the Strait of Georgia in 2009 was later than in 2008. The average catch in the Strait of Georgia in 2009 was 58% larger than in Queen Charlotte Sound, perhaps indicating that the migration out of the strait was slower and that residence times vary.

Effects of Ocean Conditions on Salmon Survival
The effects of oceanic conditions on the survival of Fraser River sockeye salmon are expected to differ among the different coastal and offshore regions. These regions are examined in order of importance to marine survival.
Strait of Georgia.-Based on the winds recorded at Halibut Bank in the central Strait of Georgia (not shown but similar to the events shown in Figure 8 for the outer coast), winds capable of flushing brackish water from the Strait in the spring of 2007 occurred only during short-term events in late March and mid-May 2007 (data from mid-May to mid-June are missing). The first prolonged event did not occur until the first half of July, after most of the fish had left the strait. The wind events in the spring of 2007 occurred during periods of waning spring tides, so it is possible that the flushing events, which would help reduce surface stratification in the strait, were partially blocked by strong mixing in the passes (Griffin and LeBlond 1990). Thus, it appears that winds in early 2007 were unable to break down the intensified surface stratification, resulting in shallowly confined plankton growth at the time the juvenile sockeye salmon were entering the strait that year. The shallow mixing layer and thinly distributed plankton biomass in the spring of 2007 could have led to low sockeye prey concentrations during the 2007 entry year and ultimately contributed to the historically low returns in 2009. Although highly speculative, it is also feasible that the low surface salinity in 2007 served as a stressor that triggered an enhanced autoimmune response to possible diseases carried by the smolts as they migrated downstream from the spawning grounds to the marine environment (Miller et al. 2011). Response to disease would have further compromised the health of the out-migrating stocks. While disease may have been a contributor to the mortality of sockeye salmon in 2007, it is important to note that there were also exceptionally large mortalities of juvenile Pacific herring, which hatched throughout the Strait of Georgia that year.
In the spring of 2008 there were numerous northwesterly wind events, so that the freshwater was flushed out on a more regular basis, diminishing the surface stratification (and possibly the light-absorbing surface fluvial silt concentration), thereby enabling plankton growth over a greater depth range. The moderate northwest winds, combined with low runoff, would have given rise to a relatively deep mixing layer and, presumably, relatively high food production in the late spring and summer of 2008. Unlike in 2007, the smolts entering the strait in 2008 were less likely to have been compromised by a lack of prey prior to their northward migration through the strait.
Queen Charlotte Sound-Hecate Strait.-The ocean conditions that negatively impacted the juvenile Fraser River sockeye salmon within the Strait of Georgia in the spring of 2007 also negatively impacted the surviving fish that entered the central coastal waters. The persistent southeasterly winds that prevailed within southeastern Queen Charlotte Sound at that time ( Figure  10b) would have driven the surface waters northward into Hecate Strait (Hannah et al. 1991), generating low salinities ( Figure 11b) and presumably strongly stratified surface waters along the eastern boundary of the sound. As a consequence, freshwater entering the sound from the mainland rivers would not have contributed significantly to the salinity structure and establishment of an optimally deep mixed layer depth (see Gargett 1997) in the central and seaward portions of the sound where the juvenile fish were migrating. Thus, in contrast to the situation in the Strait of Georgia, southerly winds over the central coast led to weak vertical stratification and a less than optimally deep surface mixed layer in Queen Charlotte Sound. This argument is consistent with the low surface chlorophyll concentrations reported throughout southern Queen Charlotte Sound in April and May 2007 (Borstad et al. 2011). The reverse conditions would have occurred during the northerly to light southerly winds and relatively low runoff (Figure 9c, d) that prevailed in the late spring of 2008. These winds would have made it easier for brackish surface water emanating from the mainland inlets to move seaward toward the entrance to Queen Charlotte Sound (Figure 1), leading to the vertical entrainment of nutrients and aiding in the establishment of a more optimally favorable mixed layer depth. Formation of a moderately deep surface mixed layer in Queen Charlotte Sound would be conducive to high plankton productivity. The seaward flushing of high chlorophyll concentration surface waters from the inlets is likely to have contributed to high productivity in the region (G. Borstad, ASL Environmental Sciences, Inc., personal communication).
Gulf of Alaska.-The juvenile Fraser River sockeye salmon that survived the 2007 marine entry period likely had low lipid reserves due to the adverse conditions they had encountered in British Columbia coastal waters. These fish  Research, Boulder, Colorado, and Trenberth and Hurrell (1994); they can be obtained from: http://www.cgd.ucar.edu/cas/ jhurrell/indices.data.html.