1 Progress in Oceanography 47 (2000) Have there been recent changes in climate? Ask the fish Gordon A. McFarlane *, Jacquelynne R. King, Richard J. Beamish Department of Fisheries and Oceans, Pacific Biological Station, 3190 Hammond Bay Road, Nanaimo, BC, Canada V9R 5K6 Abstract It is generally accepted that a climate shift occurred about 1977 that affected the dynamics of North Pacific marine ecosystems. Agreement on the possibility of further climate shifts in 1989 and the late 1990s is yet to be achieved. However, there have been changes in the dynamics of key commercial fishes that indicate changes in their environment occurred in the early 1990s, and possibly around One method of measuring climate change is to observe the dynamics of species that could be affected. Several studies have described decadal-scale changes in North Pacific climate ocean conditions. Generally, these studies focus on a single index. Using principal components analysis, we use a composite index based on three aspects of climate ocean conditions: the Aleutian Low Pressure Index, the Pacific Atmospheric Circulation Index and the Pacific Interdecadal Oscillation Index. We link this composite index (Atmospheric Forcing Index) to decadal-scale changes in British Columbia salmon and other fish populations. Around 1989 there was a change from intense Aleutian Lows (above average south-westerly and westerly circulation patterns and warming of coastal sea surface temperatures) to average Aleutian Lows (less frequent south-westerly and westerly circulation and slightly cooler coastal sea surface temperatures in winter). These climate ocean changes were associated with changes in the abundance and ocean survival of salmon (Oncorhynchus spp.), distribution and spawning behaviour of hake (Merluccius productus) and sardines (Sardinops sagax) and in recruitment patterns of several groundfish species Elsevier Science Ltd. All rights reserved. * Corresponding author. Tel.: ; fax: address: (G.A. McFarlane) /00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S (00)
2 148 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Contents 1. Introduction Key species Composite climate index Discussion References Introduction It was only a few years ago that the environment in which fish live began to be considered when assessing the impacts of fishing in fisheries management science. In general, it was believed that the effects of fishing outweighed the variability caused by the environment. If carrying capacity was considered, it was believed to be large relative to existing fish biomass. While fishing effects are undoubtedly important, in this paper we consider non-fishing effects on fish abundance and distribution. Most fisheries scientists now accept that the fishes environment can vary and that persistent patterns in fish abundance can be linked to patterns in climate-ocean conditions. Population dynamics and environmental conditions can be relatively stable on decadal scales, during periods called regimes, but they can abruptly shift from one state to another during a regime shift (Isaacs, 1975; Beamish, Neville & Cass, 1997a). Regime shifts have been identified as having occurred in 1925, 1947, 1977 (Minobe, 1997; Francis & Hare, 1994; Beamish, Noakes, McFarlane, Klyashtorin, Ivonov & Kurashov, 1999a) and 1989 (Beamish et al., 1999a). Preliminary reports indicated a further regime shift may have occurred recently (Beamish, King, Noakes, McFarlane & Sweeting, 1998; Ingraham, Ebbesmeyer & Hinrichsen, 1998). The regime concept forces scientists to examine the natural processes that regulate fish abundance, particulary those processes linked to climate ocean conditions. Matching productivity trends with climate ocean trends becomes a practical science when (and if) it can be used to forecast. From a fisheries management perspective, the issues are more related to these forecasts than debates about the existence of regimes. We need to be able to provide information on projected stock trends in a timely manner. To date, forecasting has not been particularly successful in fisheries management. As a consequence, large fluctuations have occurred without warning in catches in many of the major world fisheries, for example; salmon, walleye pollock (Theragra chalcogramma), sardines and Atlantic cod (Gadus morhua). These unexpected changes have had obvious economic consequences, as well as professional consequences. One view is that an understanding of the mechanisms is unnecessary, and these fluctuations in abundance can be managed simply by adjusting harvest rates (Walters & Parma, 1996). Even if this theory is correct, the
3 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) practical problem of reaching a consensus among managers and fishermen and restricting fisheries in time to adjust to the impacts of shifting regimes may be insurmountable. In addition, the ability to forecast trends in fish abundances would allow industry to make long-term economic and logistic plans. It is also clear that single-species management has outlived its usefulness. The public, in particular, are becoming increasingly alarmed by the threats of overfishing to ecosystems. Moving from single-species management to ecosystem management and improving forecasting techniques requires that factors regulating the abundance of fish populations naturally be much better understood. Recent studies have shown that large-scale climate changes can be associated with fluctuations in fish abundance that can be both large-scale and regional (Kawasaki & Omori, 1988; Beamish & Bouillon, 1993; Polovina, 1996; Mantua, Hare, Zhang, Wallace & Francis, 1997; Clark, Hare, Parma, Sullivan & Trumble, 1999). A number of indices of climate trends have been proposed for the North Pacific (Mantua et al., 1997; Minobe, 1997; King, Ivanov, Kurashov, Beamish & McFarlane, 1998; Beamish et al., 1999a). Indices of the Aleutian Low pressure system have been introduced (e.g. Beamish & Bouillon, 1993; Trenberth & Hurrell, 1995). The Pacific Interdecadal Index, is a widely used index that is strongly influenced by sea surface temperatures over a large area of the central North Pacific (Mantua et al., 1997). A new index (King et al., 1998), incorporates the atmospheric circulation patterns for the North Pacific ocean and characterises the flow of the westerly winds which determines both the directions of storm tracks and oceanic circulation in the North Pacific. We report trends in the productivities of a number of commercially important fish species, and changes in distribution and spawning behaviour of Pacific hake and sardines. We compare the combined responses in these fishes productivity with an index combining three climate indices, which we call the Atmospheric Forcing Index (AFI). We propose that this composite index AFI may provide a practical method of identifying when changes in the productivity of north Pacific fishes are to be expected. 2. Key species There has been a general overall decline in British Columbia fisheries which has been reflected both in the total catches and, perhaps more dramatically, in the value of the fish landed (Fig. 1). The value of landings is an indication of the change that has occurred in our expectations since the late 1980s. This indicator reflects not only changes in fish abundances, but also the social and economic implications for the fisheries. In the late 1980s the expectation was that the value of the fisheries would continue to grow or at least be sustained. However, a decline that began in the early 1990s has continued. For example, salmon catches have declined by approximately 60% since the late 1980s and early 1990, and at the same time those traditional groundfish (excluding hake) have declined by 30% and herring by 25%. Only catches of shellfish have been maintained, in part because of the development of new fisheries.
4 150 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 1. The landed value of British Columbia fisheries, 1988 and Note: 1998 data are estimated. Traditionally, such declines in fisheries would have been blamed on overfishing. However, the examination of the dynamics of the different stocks reveals they have responded synchronously so it seems probable that they are all subject to a single overriding factor in the ocean environment that has affected their productivity. Beamish and Bouillon (1993) reported that total Canadian salmon production, as indicated by catch, declined to low levels from the mid-1950s to the mid-1970s. Following the 1977 regime shift, salmon catches increased to historic high levels by the late 1980s. However, by the mid-1990s catches had declined to historically low levels, in part because of introduction of fishing restrictions in response to low abundances. It was not until 1994 that the declining trend became clear (Fig. 2), although the brood year strength for fish caught in 1994 would have been determined by conditions, depending on the species. Therefore, the decline in salmon production is both consistent with the regime change in 1989 and unprecedented in the available historical catch records. It was possible to examine productivity changes for sockeye salmon (O. nerka) from the Fraser River in relation to harvest rates (Beamish et al., 1997a). Despite an almost constant harvest rate (74.1%), there was an increase in the stock specific productivity after the 1977 regime shift indicating that the marine survival had improved. Probably the most dramatic shift in abundance has been for southern coho (O. kisutch; Fig. 3). Beamish, Noakes, McFarlane, Pinnix, Sweeting and King (1999c) showed a synchronous decline in the marine survival of stocks of coho entering the Strait of Georgia, Puget Sound and off the coast of Washington and Oregon beginning in the early 1990s. The abundances of the returning fish were so low that severe fishing restrictions were implemented in virtually all of these areas. In addition to these declines in marine survival, the behaviour of coho from the Strait of Georgia changed in association with a change in the timing of the spring freshet
5 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 2. The total Canadian catch of Pacific salmon (pink, sockeye, chum, coho and chinook), in the Fraser River, increased sea level heights, and a change in the direction of winter winds (Beamish, McFarlane & Thomson, 1999b). In a recent study, Bradford (1999) examined 45 coho populations from Alaska to Oregon to determine if these large-scale climate impacts that occurred in the ocean were also evident in fresh water. He found no evidence for a large-scale and synchronous fresh water impact on coho production, implying the large-scale climate impacts on the population dynamics of coho were primarily the result of changes in the ocean. Additional evidence for large-scale climate impacts occurring in the ocean is seen in the decline in hatchery fish returns, even though the numbers of hatchery fish entering the ocean have either been increased or have remained constant (Beamish, Mahnken & Neville, 1997b). The species presented in Table 1, represent over 90% of the current landings of groundfish off Canada s west coast. For the period the percentage of average or above average year classes ranged between 50 and 78% (with one exception Pacific hake from the Strait of Georgia). After 1988 (1989 to present), only 25 35% of the year classes were average or above average. Clearly the synchronous response in year class success across this diverse group of species is a reflection of large-scale climate ocean processes. For example, sablefish (Anoplopoma fimbria),
6 152 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 3. The standardised anomalies for the marine survival of coho salmon released into the (a) Strait of Georgia, (b) Puget Sound and (c) Oregon.
7 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Table 1 Numbers of strong, average and below average year classes for selected groundfish species during and 1989 to present Species to present Below Strong Average Strong Average average Below average Sablefish Pacific hake (offshore) Pacific hake (Strait of Georgia) Pacific cod Rock sole English sole Dover sole a Yellowtail rockfish b Pacific Ocean perch b a Based on relative year class abundance of US stocks (J. Fargo, Fisheries and Oceans Canada, Pacific Biological Station, Nanaimo, British Columbia, V9R 5K6, pers. comm.). b Estimates since 1989 are not available as year classes have not yet recruited to the fishery. which is currently the most valuable of groundfish species, shows clear decadalscale trends in year class strength from 1960 to 1998 (King, McFarlane & Beamish, 2000; Fig. 4). From , year classes were average or below average without any above average strength year classes. The 1977 year class was exceptionally Fig. 4. The index of relative year class success of sablefish
8 154 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) strong. From , there were 5 above average year classes and 3 year classes each of average and below average success. The 1989 and 1990 year classes were relatively very strong though not as exceptional as the 1977 year class. Since 1990, year class success has been below average except in The fishery for flatfishes in British Columbia is mainly a trawl fishery; more than 80% of the flatfish landings are either Dover sole (Microstomus pacificus), or English sole (Parophrys vetulus) or rock sole (Lepidopsetta bilineata). From 1970 to the mid-1980s, landings fluctuated between 3,000 and 7,000 tonnes, but increased dramatically in the late-1980s to the early-1990s. These are relatively long-lived species that have large fluctuations in their year class strengths (Fargo, 1998). Year class strength (estimated as the number of year 4 recruits) was generally above average in the 1980s but has been low since 1991 (Fargo, 1998). These species (Table 1) generally showed above average productivity from 1977 to 1989 but below average productivity in the 1990s (e.g. English sole, Fig. 5). Pacific cod (Gadus macrocephalus) is a short-lived, fast-growing species that matures at age 3 (Westrheim, 1996). In British Columbia, it is at the southern limit of its distribution. Spawning success of Pacific cod is sensitive to temperature (Alderdice & Forrester, 1971). Since species at the extremes of their distributional range tend to be more sensitive to the effects of climate, Pacific cod is likely to be affected by any warming of the ocean habitat in British Columbia. Its spawning stock biomass and recruitment has been estimated for Hecate Strait using a stock reconstruction method based on ages estimated from lengths (Haist & Fournier, 1998). Prior to 1993, there were no quotas for the fishery, although there were Fig. 5. Year class index (anomalies from long term average ) for English sole
9 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) restrictions relating to fishing times and areas. In the 1990s, recruitment was consistently poor (Table 1, Fig. 6) and in 1996, the fishery was closed, and only small catch was permitted in The pattern of recruitment (age 2+) in Hecate Strait from 1960 to 1988 was variable but since 1989 recruitment has been consistently poor (Fig. 6). In addition to the trends in catches of most species, there have been changes in their distributions that have significantly altered the relative species composition off the west coast of Canada. In 1992, Pacific sardines were once again captured off British Columbia, after a complete absence of 45 years (McFarlane & Beamish, 1999; Fig. 7a). Their abundances continued to increase and in 1995 an experimental fishery was initiated. The earlier disappearance of sardines from the waters of British Columbia waters corresponded to the 1947 regime shift. In 1997 the stock size off the west coast of Vancouver Island was estimated to be 60,000 tonnes (McFarlane & Beamish, 1999), exceeding that of Pacific herring. By the mid-1990s, sardines were found in the surface waters along the entire west coast. Additionally, in 1997 and 1998 sardines not only remained off Canada s west coast year round but also successfully spawned in Canadian waters for the first time ever recorded (McFarlane & Beamish, 1999; Fig. 7b). Pacific hake is a large migratory species, which during the 1960s, 1970s and 1980s spawned off Baja, California during the winter and then migrated north to summer feeding grounds (Francis, 1983). There are no data prior to this time documenting Fig. 6. Year class index (anomalies from long term average ) for Pacific cod
10 156 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 7. Changes in distribution and spawning area of Pacific Sardine during the 1990s illustrating (a) increased northward movement in Canadian waters since reappearance in 1992 and (b) a change in spawning areas during 1990s.
11 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) its distribution patterns. Approximately 25 30% of its stock of mature fish migrate into Canadian waters. Since the early 1990s the percentage of the stock migrating into the Canadian zone increased reaching approximately 40%; 400,000 tonnes. These fish have continued to move farther north (Fig. 8a) and now remain off the west coast of Canada year round. Since 1994, the hake have been spawning off the west coast of Vancouver Island (Saunders & McFarlane, 1996; Fig. 8b). The fishery for Pacific herring (Clupea harengus) is one of the most important off the west coast. It is managed as five distinct fisheries or stocks, and both the abundance trends and recruitment patterns differ between each stock. Synchronously there were strong year classes identified in all five stocks in 1977, 1985 and The stock in the Strait of Georgia is the most important, and this stock increased dramatically throughout the 1990s as a result of a number of strong year classes (Schweigert, Fort & Tanasichuk, 1998). However, the stock off the west coast of Vancouver Island has declined during this same time period (Schweigert et al., 1998). Northern stocks show the same differing dynamics (Schweigert et al., 1998). We used seven measures of year class recruitment for the various pelagic and groundfish species off the west coast of Canada. For salmon, pink (O. gorbuscha), sockeye and chum (O. keta) provide the majority of the Canadian catch, and the exploitation rates of all three species are high (e.g. 80%, Beamish & Bouillon, 1993). Catches can therefore be used as a surrogate for abundances. We time-lagged catches of pink, sockeye and chum (updated from Beamish et al., 1997b), by 1, 2, and 3 years respectively to correspond to their predominant age of ocean entry. The total lagged catch was used as an estimate of salmon production. For groundfish, we used recruitment estimates, based on year classes, using the available length-based or age-based models for Pacific cod, rock sole (Lepidopsetta bilineata), English sole and sablefish (Saunders & McFarlane, 1995; Fargo, 1998; Haist & Fournier, 1998; King et al., 2000). Recruitment time-series for two pelagic species, hake and herring were also available, from length- or age-based models (Dorn et al., 1999; Schweigert et al., 1998). All seven time-series for the fishes overlapped from and were assessed using a principal components analysis based on their correlation matrix. The time series for Pacific cod and hake had to be normalised by log transformation. The first component explained 34.89% of the variation, which, according to the broken-stick model (Frontier, 1976), is not considered to be significant (P 0.05). The component scores (Fig. 9) are influenced most by rock sole and English sole (l=0.56 and 0.55 respectively), followed by salmon (l=0.39) and Pacific cod (l=0.32). The scores are least influenced by sablefish, herring and hake (l=0.23, 0.20 and 0.03 respectively). The scores illustrate a below average recruitment from , above average from , and below average for (Fig. 9). The scores do not reflect the regime change in 1977 because of the high influence of rock and English sole, whose recruitment did not respond to the shift until In order to examine changes in recruitment between the regimes with the influence from each of the seven species equalised, we have used the approach of Ebbesmeyer, Cayan, Mclain, Nichols, Peterson and Redmond (1991) to identify the 1977 regime shift, which has subsequently been extended by Hare and Mantua (this volume). This
12 158 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 8. Change in distribution and spawning area of Pacific hake during the 1990s illustrating (a) increased northward movement in Canadian waters since 1990 and (b) a change in spawning area during 1990s.
13 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 9. The composite index of fish production represented as the standardised scores for the first component from a principal components analysis based on 7 fish time series. approach examines each regime shift separately, so we have been able to incorporate recruitment for halibut (Hippoglossus stenolepis) for which data are only available up to the 1990 year class (Clark et al., 1999). Each series was detrended by subtracting the series mean. To assess the recruitment before and after the regime shifts of 1977 and 1989, equal available years before and after each shift were selected (i.e. 9 and 4 years respectively). Each shift year was excluded, and each annual value was divided by the standard deviation of either all values before or all values after each shift. All the normalized series were then averaged into an aggregated time series (Fig. 10). This aggregated series shows that changes occurred in the mean standard deviations around 1977 (from 0.45 to 0.37, Fig. 10a) and around 1989 (from 0.67 to 0.55, Fig. 10b). 3. Composite climate index The development of the Aleutian Low pressure system makes the winter time in the North Pacific an energetic period. We have therefore chosen to focus on winter months (December through March) and have selected three indices that characterize the dominant features of atmospheric and oceanic processes. The Aleutian Low is the dominant atmospheric feature in winter. We obtained the Aleutian Low Pressure Index (ALPI, Fig. 11a) from the Pacific Biological Station website (http://www.pac.dfo-mpo.gc.ca/sci/sa-mfpd/english/clmfindx1.htm) which has been updated from Beamish et al. (1997).
14 160 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 10. The mean standard deviates of (a) the eight fish time series and (b) the seven fish (halibut removed) time series illustrating the overall increase in fish production around 1977 and overall decrease in fish production in the early 1990s.
15 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) Fig. 11. Indices of climate change: (a) Aleutian Low Pressure Index, (b) Pacific Circulation Index and (c) Pacific Interdecadal Index.
16 162 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) The intensity and position of the Aleutian Low determines the relative waviness of the westerlies. A relatively weak and northward positioned Aleutian Low results in a direct westerly flow in the atmosphere over the North Pacific (i.e. westerly winds). A relatively strong and southward positioned Aleutian Low would result in wavy westerlies that are from the southwest off the coast of North America (i.e. southwesterly winds). The Pacific Circulation Index (PCI, King et al., 1998) describes the atmospheric circulation patterns of the westerlies. The PCI (Fig. 11b) is a new atmospheric circulation index similar to the Atmospheric Circulation Index produced for the Atlantic (Girs, 1971) and is based on a simplified interpretation of daily atmospheric circulation patterns over the whole North Pacific. Daily atmospheric circulation patterns were determined using a manual synoptic classification scheme (Girs, 1971) based on sea level pressure data (see Yarnal, 1993, for a review of synoptic climatology techniques). The changes in both the daily frequency of westerly and southwesterly winds are reflected by changes in daily frequency of the negative anomalies of northwesterly winds (King et al., 1998). The negative northwesterly anomalies were also obtained from the Pacific Biological Station website based on original data supplied by the Arctic and Antarctic Research Institute (St Petersburg, Russia). The intensity of the Aleutian Low is associated with relative wind intensity and ocean circulation in surface waters in addition to the atmospheric circulation direction. When the Aleutian Low is intense the winds are strong, and there is enhancement of the upwelling of nutrient-rich cool water to the surface of the central subarctic Pacific. As a surrogate for mid-ocean upwelling, we have used the Pacific Interdecadal Oscillation index (PDO, Mantua et al., 1997). The PDO (Fig. 11c) was obtained from S. Hare (personal communication, International Pacific Halibut Commission, Seattle, Washington) via the website (http://www.iphc.washington.edu/staff/hare/html/decadal/post1977). The PDO is the first component of a principal components analysis on gridded sea surface temperatures of the North Pacific. Although it explains only 21% of the variance in sea surface temperatures (Overland et al., 2000), it typifies the variability in sea surface temperatures associated with mid-ocean upwelling. A negative phase (e.g ) reflects warmer sea surface temperatures in the central north Pacific and cooler temperatures along the coast of North America. The conditions are inverted during a positive phase (e.g ). From 1989 to 1997, there have been four years when PDO values have been close to zero, four years with negative PDO values and only one year with a positive PDO value. Similar shifts in the spatial patterns of sea surface temperature have been observed by Namias (1972) and Deser, Alexander and Timlin (1996). Hollowed and Wooster (1992) reported shifts in winter sea surface temperatures for the coastal and offshore areas off the coast of British Columbia. However, Freeland, Denman, Wong, Whitney and Jackques (1997) suggested an overall warming trend in sea surface temperatures for the British Columbia coastal waters. Average monthly sea surface temperatures are available for the coastal areas of British Columbia in 2 by 2 grids from the Comprehensive Ocean Atmosphere Data Set (COADS) and the Integrated Global Ocean Services System (IGOSS). If the sea
17 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) surface temperatures for the area bounded by 141 W, 123 W, 35 N, and 60 N are considered according to the winter and annual averages during the three regimes (pre-1977, and post-1989), a difference in warming is detectable (Fig. 12a,b,c). There was an overall warming trend in winter, exhibited by the warmer temperatures for the whole area in compared to (Fig. 12c), but the increase was greater during (Fig. 12a) than in (Fig. 12b). These differences in warming between regimes are not evident in data when averaged over the entire year (Fig. 12d,e,f). On an annual basis, there is a continuous increase in temperatures across regimes. The three climate ocean indices (ALPI, PCI, PDO), which overlapped from Fig. 12. The difference in average winter (December March) sea surface temperatures ( C) between (a) the period and the period, (b) the period and the period, and (c) the and the period. The difference in average annual temperatures were also calculated for similar comparisons: (d) the period vs. the period, (e) the period vs. the period, and (f) the vs. the periods. Differences in temperatures are reported for the centre of the 2 2 grid. Shaded grids represent areas in which the average temperature increased from the one period to the other.
18 164 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) to 1998, were combined using principal component analysis based on their correlation matrix. The first principal component (Fig. 13) scores positive for an intense Aleutian Low, an above frequency of south-westerly and westerly atmospheric circulation, and a general cooling in the central north Pacific but warming in the coastal areas. The composite index scores negative for a weak Aleutian Low, a decrease in south-westerly and westerly circulation, and a warming trend in the central north Pacific and cooling along the coast. The composite index, as the first principal component of the combined three indices expresses 70% of the total variation which, according to the broken-stick model (Frontier, 1976), is considered significant (P 0.05). Therefore, it can be used as a common measure of atmospheric forcing, and we refer to it as the Atmospheric Forcing Index (AFI). The AFI was about equally influenced by each of the 3 indices (ALPI l=0.81, PCI l=0.87, PDO l=0.81) indicating that each index may be an expression of an underlying (and as yet undetermined) global mechanism. The composite index changed pattern in the early 1920s, late 1940s, the mid- 1970s and the late 1980s. The times of these changes were consistent with regime shift years of 1925, 1947, 1977 proposed by Minobe (1997), Mantua et al. (1997) and Beamish et al. (1999a). Beamish et al. (1999a) proposed that a further shift occurred in 1989, but this was not recognised by Mantua et al. (1997). However, the 1989 regime shift was followed by a series of years with El Niño like conditions (Trenberth & Hoar, 1996). The extreme positive AFI value for 1998 (Fig. 13) may indicate the onset of a new regime. While this new regime will not necessarily reflect Fig. 13. A composite index (atmospheric forcing index, AFI) developed from the 3 indices in Fig. 11, as represented by standardised scores along the first principal component (PC1).
19 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) a complete reversal to pre-1989 conditions, it might result in a yet unobserved shift in the system s structure. It is important to note that the frequency of the changes in the composite index are different from the oscillations associated with El Niño Southern Oscillation events. It is the frequency associated with this atmospheric forcing index that we believe most influences the processes that regulate fluctuations in the abundances of key commercial species off Canada s west coast (Beamish, McFarlane & King, 2000). 4. Discussion The 1989 regime shift has been proposed by a number of researchers, but others remain unconvinced. Data aggregated for the eight species or species groups that contribute 90% of the commercial fish landings and 80% to the value of the fisheries off the west coast of Canada, show a response to the 1977 and 1989 regime shifts. If we ask the fish, then based on their productivity responses in the 1990s, there is little doubt that the changes observed in the climate-ocean system in 1989 represent a shift to another system state. We have used an aggregate time series of fish productivity in two different analytical approaches. Variability in this aggregated fish time series is a function of the duration of the time series, the accuracy of the measurements (or models) used to estimate productivity, and the biology or life histories of the species. For example, herring, which is a short-lived pelagic species, is unlikely to be influenced by the same processes as sablefish or halibut, which are long-lived demersal species. Given these differences, it is remarkable that all the species show similar patterns of variability. The differences also suggest that some statistical approaches may not be appropriate or should be used with caution, for example, the principal components analysis of the aggregate fish productivity series, which indicated a change in 1984 rather than in As mentioned above, the scores were most heavily influenced by the recruitment data for rock sole and English sole, whose year class response did not exhibit a change until after The principal component analysis allows the variation of each fish species to determine the amount of influence its series had on the resulting components scores. These artificial (though statistically sound) results are analogous to suggesting that one or two species represent, or drive, the response of the ecosystem. The second statistical approach used here (i.e. calculating average standard deviates), might be the more appropriate approach since it ensures that each fish time series has equal weight. Principal components analysis appears to be more appropriate for evaluating climate ocean or environment variables in which one or two series might best reflect system dynamics, especially when the important dynamics are known. In the climate-ocean principal component analysis used here, we selected the season and aspects of atmospheric and oceanic systems that we felt were most important in describing the North Pacific ecosystem dynamics. The results suggested that all three variables (ALPI, PCI, and PDO) reflect the variability in the North Pacific system states. This is not surprising given that all three reflect changes
20 166 G.A. McFarlane et al. / Progress in Oceanography 47 (2000) in the intensity of the Aleutian Low, so perhaps ALPI is, by itself, an appropriate measure of the North Pacific dynamics. We suggest that researchers must be aware of which statistical approach is most appropriate for the analysis of the variables they select for analysis. Attention should also be given to the number of variables selected relative to the number of observations available, especially when the number of variables begin to match or outweigh the number of observations (King & Jackson, 1999). The need to select appropriate analytical techniques will not override the necessity of understanding how fish regulate their abundance and behaviour naturally. The aggregated response of the fishes analysed in this study showed a response to the regime shifts that occurred in 1977 and 1989, but not all species responded to both shifts. For example, sardines only responded to the 1989 shift. Sardines had disappeared from Canadian waters following the 1947 shift, and only returned after the 1989 shift. This indicates that the linkage with climate is a specific response to a specific ecosystem change, which is coded in the responses of the biology and dynamics of individual species. Managing the impacts of climate change on commercial fisheries will require an understanding of these species specific mechanisms, to ensure species are protected or to avoid scientifically embarrassing mistakes such as listing sardines off of British Columbia as a species at risk (COSEWIC, 1998). The productivity and landings of British Columbia fisheries have undergone synchronised changes since These changes are consistent with a change in the composite index of atmospheric forcing. The mechanisms involved remain to be determined and probably differ between species. For example, for salmon the poor returns in the fisheries may reflect increased predation of the early juvenile stage in some of the species. However a large-scale response, as observed for coho, would suggest that it is unlikely that a suite of predators were synchronously able to feed more heavily on the coho. Beamish and Mahnken (1998) proposed that growth-based mortality that occurred in the fall and winter was related to growth during the summer months, which in turn may be affected by the availability of prey. According to their hypothesis, climate could be the common factor that would synchronously affect the carrying capacity of the ecosystem. An important consideration when attempting to explain the causal mechanisms for the fluctuations in British Columbia salmon and other species is the difference climate patterns that prevail to the north and south of Queen Charlotte Sound. Ware and McFarlane (1989), recognising the oceanographic responses, identified the ecosystems north and south of this boundary as two different domains. Moore and McKendry (1996) also identified responses to variations in precipitation. They noted that prior to 1977 there was a tendency for precipitation to be lighter in the north, and heavier in the south, but post 1977 the pattern reversed. This difference in north/south response is also detectable in the decadal-scale trends of the flow patterns of the large rivers (Mantua et al., 1997; Beamish et al., 2000). The change in AFI in the late 1970s also influenced salmon and groundfish productivity (Beamish & Bouillon 1993, 1995). Previous to this, any changes are difficult to identify because many of the fisheries had not developed and generally data are scarce. Another change may have occurred around But even if this change