The effect of temperature on swimming performance and oxygen consumption in adult sockeye (Oncorhynchus nerka) and coho (O. kisutch)salmon stocks

Temperature has been coined the `ecological master factor' for fish(Brett, 1971), and important physiological functions such as growth, swimming performance and active metabolic rate can have species-specific temperature optima that are near a species-preferred or acclimated temperature(Fry, 1947;Brett, 1971;Dickson and Kramer, 1971;Beamish, 1978;Houston, 1982;Bernatchez and Dodson, 1985;Johnston and Temple, 2002). Thus, when fish are exposed to temperature changes, they can obtain optimal performance by altering either their behaviour (preference/avoidance) or their physiology (adaptation and acclimation), when the temperature change is sufficiently long. Certain short-term variations in temperature may be unavoidable, however, and this is particularly the case for adult migratory salmon that are returning to their natal streams to spawn. For example, water temperatures in one of the world's greatest salmon-bearing rivers, the Fraser River, BC, Canada, may vary annually on a given date by as much as 6°C. Furthermore, the river temperatures encountered by the Early Stuart stock of Fraser River sockeye salmon Oncorhynchus nerka during its 25-day migration can vary by as much as 10.5°C(Idler and Clemens, 1959) and reach up to 22°C (Rand and Hinch,1998).

Given the adult salmon's short migration window and its exposure to a wide variation in temperature, it is possible that acclimation mechanisms that would normally compensate for temperature change may be incomplete. Conversely, Guderley and Blier(1988) suggest that swimming performance and most of its components demonstrate thermal compensation on an evolutionary time scale (i.e. adaptation) such that optimal performance and lowest thermal sensitivity are typically within the temperature range most frequently encountered by the organism. In the case of adult salmon stock, the prediction is that they would retain sufficient physiological flexibility to accommodate the range of temperatures most frequently encountered during their river migration; otherwise intolerance of non-optimal temperatures in reaching spawning grounds (Macdonald et al.,2000) could hamper spawning success.

While considerable information on the temperature effects on swimming (e.g. critical swimming speed, U_crit) and oxygen consumption(Ṁ_O2) exists for juvenile Pacific salmon (Oncorhynchus spp.) (e.g.Brett et al., 1958;Brett, 1971;Griffiths and Alderdice, 1972;Beamish, 1978), only four studies have measured Ṁ_O2 in adult,wild Pacific salmon (Brett and Glass,1973; Jain et al.,1998; Farrell et al.,1998,2003). One of these studies(Brett and Glass, 1973)established a temperature optimum of 15°C for both U_crit and maximum Ṁ_O2(Ṁ_O2max). All the same, important intraspecific (between stocks) as well as interspecific differences in swimming energetics with respect to temperature are anticipated. Different salmon stocks migrate to different spawning streams in the Fraser River watershed, resulting in dissimilar up-river migration costs due to different water temperatures, coupled to variation in migration timing as well as unequal migration distances in the presence of differing hydraulic impediments. Indeed, juvenile salmonids reared or held under laboratory conditions can show intraspecific differences among populations and strains(Tsuyuki and Williscroft,1977; Thomas and Donahoo,1977; Taylor and McPhail,1985). Our focus was on whether performance differences exist among adult, wild salmon stocks.

Berst and Simon (1981)suggested that field-based rather than laboratory-based studies are more likely to reveal any differences among species or stocks, because animal transportation is minimized and natal river water can be used. While U_crit has been previously measured in adult salmonids under field conditions (e.g. Jones et al.,1974; Brett, 1982;Williams et al., 1986;Farrell et al., 2003), only two field studies have previously reported active Ṁ_O2 for adult salmon (Farrell et al., 2003;C. G. Lee, A. P. Farrell and R. H. Devlin, manuscript submitted for publication). Because no field study has comprehensively examined the effects of temperature on swimming energetics in adult salmon, the present study considered: (1) how the temperature affects swimming energetics of adult salmon from the Fraser River watershed, and (2) whether intraspecific differences in swimming energetics exist with respect to temperature. Assuming that natural selection acts strongly on the physiology associated with up-river migration, we predicted that swimming ability (as measured by U_crit and Ṁ_O2max) should increase with migration distance and difficulty among stocks of sockeye salmon that were studied, one of which was a coastal (short-distance migrating) stock while the other stocks (long-distance migrating) were from the interior of the province of British Columbia (BC).

Experiments were conducted in 2000 and 2001 on sockeye salmon Oncorhynchus nerka Walbaum and coho salmon O. kisutchWalbaum intercepted during their up-river migration in the Fraser River watershed (Fig. 1), which provides considerable variability in the timing, temperature, distance and difficulty of migration among salmon stocks and species as a study system because of its large size, numerous tributaries and hydrological challenges(Gilhousen, 1990; G. T. Crossin, S. G. Hinch, A. P. Farrell, D. A. Higgs, A. G. Lotto, J. D. Oakes,and M. C. Healey, unpublished observations). Experiments were performed at both field (i.e. near the sites of fish capture and using natal stream water)and laboratory (i.e. captured fish returned to Cultus Lake laboratory,Fisheries and Oceans Canada, Chilliwack, BC, Canada, or the home laboratory at Simon Fraser University) locations using mobile Brett-type respirometer swim tunnels (Farrell et al.,2003). Comparisons were made among three sockeye salmon stocks,hereafter identified by name and relative river migration distance from the mouth of the Fraser River to the spawning stream [Early Stuart (ES), 1100 km;Gates Creek (GC), 400 km; Weaver Creek (WVR), 100 km], and one coho salmon(O. kisutch) stock [Chehalis River (CHE), 110 km] (seeTable 1). In addition, the ambient water temperatures encountered by these stocks in their natal rivers showed some degree of overlap (Table 1). The sockeye stocks that migrate to the interior of the province (GC and ES) also negotiate particularly demanding hydrological challenges (e.g. Hell's Gate and Saddle Rock) during their upstream migration(Fig. 1)(Hinch and Rand, 1998;Hinch and Bratty 2000). Compared with most sockeye salmon stocks, which migrate during the warmer summer and fall months, Fraser River coho salmon encounter cooler temperatures during their fall and winter migration(Groot and Margolis, 1991). Therefore, while coastal stocks of CHE coho salmon and WVR sockeye salmon experience a comparable migration distance and difficulty, they encounter different river temperatures because of differences in run timing.

Field tests were always performed at ambient natal river water temperature and fish were transferred directly to the swim tunnel from the nearby stream or creek. Laboratory tests at both Simon Fraser University (with dechlorinated municipal water) and the Cultus Lake laboratory (with lake water) were performed either at the ambient temperature of the natal stream or at an adjusted temperature (Table 1). Laboratory-tested fish were transported in a 330 litre insulated tank containing oxygenated water, dilute Marinil anaesthetic (0.02 mg l^–1 metomidate hydrochloride, Syndel International Inc.,Vancouver, BC, Canada) to calm the fish, and block ice to chill the water. Fish were held in 1000 litre aquaria, one of which was maintained at the ambient temperature of the fish's natal stream. The water temperature was adjusted (via mixing warm surface and cold deep Cultus Lake water) in the other aquaria at a rate of approximately 1°C per day over 5 days to extend the temperature range slightly beyond the ambient temperatures, after which two fish were tested daily in one of two swim tunnels receiving water at the fish's adjusted temperature. A longer temperature acclimation period was not used because some of these mature fish were within weeks of spawning. For all tests, fish were given a practice swim after 1 h and then allowed to recover in the swim tunnel overnight. Water delivery was at a rate of 30 l min^–1via a submersible sump pump to ensure that dissolved oxygen was normally >90% of air saturation.

Experiments were conducted on ten male and ten female adult GC sockeye salmon at the BC Hydro Seton Dam near Lillooet, BC(Fig. 1) in mid-August 2000. Fish were dip-netted on-site at the top of a fish ladder and immediately placed into a swim-tunnel. Fish, which were ripe and within 1–2 weeks of spawning, were not killed after the experiments to comply with the sampling permit and so gonad mass was not measured. In early August 2001, six male and eleven female GC sockeye salmon were again collected at the Seton Dam site,but were transported (1.5 h) to the Cultus Lake laboratory. Three fish were tested at the ambient temperature of the Seton River at that time(15.0±1.0°C), while six fish were tested at a colder and eight fish at a warmer temperature.

Experiments were performed in October 2000 on six male and six female WVR sockeye salmon at the Chehalis River Fish Hatchery, which is situated <5 km from Weaver Creek, their natal stream. Fish were dip-netted at the spawning creek, transported to the hatchery and immediately placed in the swim tunnel for overnight recovery. Experiments were also conducted at Simon Fraser University (SFU) on five male and seven female WVR sockeye salmon, captured by beach seine from the Harrison River, BC, Canada(Fig. 1) in September 2000. Transportation to SFU took 1 h, where fish were held at 13.0±0.2°C for a minimum of 3 days before testing commenced. An additional five male and three female WVR sockeye salmon were collected from Weaver Creek viadip-net and transported (0.5 h) to the Cultus Lake laboratory in October 2001. Two fish were tested at the ambient water temperature at Weaver Creek(12°C), while five fish were tested at a warmer temperature.

A small number of ES sockeye salmon were dip-netted from the Fraser River near Yale, BC, Canada (Fig. 1)in early July 2000, and transported (1.5 h) to SFU, where they were held for a minimum of 3 days before testing commenced. These fish were 4–5 weeks from spawning and were beginning to exhibit secondary sexual characteristics.

Experiments were conducted in November 2000 on seven male and six female CHE coho salmon that were captured with a knotless cotton dip-net from the stream entering at the Chehalis River Fish Hatchery. Fish were immediately placed into a swim tunnel for overnight recovery. Additional experiments were conducted at the Chehalis River Fish Hatchery on six male and six female CHE coho salmon in January 2001. Experiments were also conducted in November 2001 on four male and three female CHE coho salmon after transportation (0.5 h) to Cultus Lake laboratory. Two fish were at the ambient temperature of the Chehalis River (9°C), while five fish were tested at a warmer temperature.

The 272 litre and 471 litre swim tunnels (afterGehrke et al., 1990),described in Farrell et al.(2003; www.sfu.ca/biology/faculty/farrell/swimtunnel/swimtunnel.html)were mounted on trailers to facilitate transportation to the field locations. The 124.3 cm long transparent swim chamber had an internal diameter of 20.3 cm for the small tunnel and 25.4 cm for the large swim tunnel. A `shocking' grid(2–10 V; 0.4–2.0 W), made of graphite rods and mounted at the rear of the swim chamber, was utilized briefly at higher water velocities to promote swimming in some fish. Water flow in the swim tunnels was driven by a 29 cmdiameter fiberglass centrifugal impellor pump and a 7.5 hp three-phase motor, controlled by a Siemens Midimaster Vector frequency drive (PLAD,Coquitlam, BC, Canada). Water velocity was calibrated against the motor frequency (Farrell et al.,2003). Throughout the course of an experiment, water temperature in the swim tunnel did not fluctuate by more than 0.5°C.

The practice swim involved water velocity increments of 0.15 body lengths(BL) s^-1 every 2 min until failure and was used to familiarize naïve fish to the swim tunnel and also provide an estimate of the U_crit (Jain et al., 1997). The following day, each salmon was tested with a ramp-U_crit protocol(Jain et al., 1997), in which the water velocity was ramped up in 5 min increments of 0.15 BLs^-1 up to approximately 50% of the fish's maximum speed attained in the practice swim. Water velocity increments of 0.15 BLs^-1 then followed every 20 min until the fish ceased swimming. Testing was terminated when the fish failed to move off the rear grid for 20 s. Water velocity was then reduced to 0.30-0.45 BL s^-1 for a 45 min recovery period, after which the a second ramp-U_crit protocol was performed. Approximately half the fish swam intermittently as they recovered, while the remainder rested on the bottom of the swim chamber for the entire recovery period. U_crit values were calculated as in Brett(1965): U_crit=U_f+(t_f/t_iU_i),where U_f is the water velocity of the last fully completed increment; t_f is the time spent on the last water velocity increment; t_i is the time period for each completed water velocity increment (20 min); and U_i is the water velocity increment (0.15 BL s^-1). U_crit was corrected for the solid blocking effect as outlined by Bell and Terhune(1970). A streamline shape factor was used in the correction equation U_F=U_T(1+ϵ_s), where U_F is the corrected flow speed, U_T is the speed in the tunnel without a fish in the swim chamber andϵ _s is the fractional error due to solid blocking.ϵ _s is defined for each fish byϵ _s=τλ(A_o/A_T)^1.5,where τ is a dimensionless factor depending on swim chamber cross section(equivalent to 0.8 in this study), λ is the shape factor for the fish(λ=0.5 body length/body thickness), A_o is the cross sectional area of the fish, and A_T is the cross sectional area for the swimming chamber. The U_crit correction averaged 16.2±0.1%. The second U_crit test examined the ability of fish to recover and re-perform. A recovery ratio (RR) expressed the ratio of the two swimming performance tests:RR=U_crit2/U_crit1. Thus, when RR=1, the U_crit performance was identical for both swim tests.

The oxygen consumption measured immediately prior to the initial U_crit swim test was assigned as routine Ṁ_O2(Ṁ_O2routine). We did not attempt to estimate standard metabolic rate either by eliminating data for fish that were active, as others have done (seeBrett and Groves, 1979), or by extrapolating to zero velocity, because of concerns regarding this method of extrapolation (see Thorarensen et al.,1993; Farrell et al.,2003). Oxygen consumption rates during swimming were measured for every other water velocity increment during both swim tests. The Ṁ_O2 measured at U_crit was designated maximum Ṁ_O2(Ṁ_O2max). We distinguish Ṁ_O2max from active metabolic rate, which is defined as the Ṁ_O2 during maximum sustained activity (i.e. steady state swimming for >200 min;Brett and Groves, 1979). The designation of Ṁ_O2max during swimming at U_crit in salmonids was rationalized because both cardiac output and venous oxygen partial pressure can plateau before U_crit is reached, and arterial oxygen partial pressure can decrease (Thorarensen et al.,1993; Gallaugher et al.,1995; Farrell and Clutterham,2003). In addition, some fish can show a plateau in Ṁ_O2 measurements before U_crit is reached (see data for GC sockeye). We did not calculate metabolic scope (defined as active metabolic rate - standard metabolic rate; Brett and Groves,1979). Instead, we calculated scope for activity (from Ṁ_O2max-Ṁ_O2routine). Cost of transport, COT, was calculated from Ṁ_O2/Ufor each swimming speed, U, and net cost of transport,COT_net, was calculated from(Ṁ_O2-Ṁ_O2routine)/U. The minimum costs of transport were interpolated from the curves fitted to these data. Ṁ_O2measured immediately prior to the second U_crit test was termed Ṁ_O2recovery, and was compared with Ṁ_O2routine to determine the degree of recovery from the first swim test.

Values are means ± s.e.m. and P<0.05 was used as the level of statistical significance. Intraspecific statistical comparisons between the first and second swim trial and between the laboratory-based and field-based measurements were performed with paired and unpaired students t-tests, respectively. Statistical comparisons among all fish stocks were accomplished using a parametric analysis of variance (ANOVA). In cases where the ANOVA reported significant differences, a pairwise post-hoc Tukey test was used to determine specifically which groups were different. For the relationship between swimming speed Uand Ṁ_O2,regression analysis used exponential equations, based on previous findings(e.g. Webb, 1971), although preliminary analysis indicated that power functions (e.g. a 4-parameter Lorentzian regression) also produced similar r² values (to within 10%). For the relationships with water temperature, exponential regressions were used for Ṁ_O2routine and bell-shaped regression for Ṁ_O2max and U_crit, based on previous findings(Brett and Groves, 1979).

U_crit values for adult salmon tested at ambient water temperature are presented in Table 2. Overall, fish swam just as well on the second test because no significant differences were observed between U_crit1 and U_crit2 (first and second swim tests, respectively) for any of the salmon stocks. In fact, the first and second U_critvalues rarely differed by more than 5% and, as a result, the recovery ratios(RR) for all groups of salmon were not significantly different from unity(Table 3). Similarly, the Ṁ_O2max and scope for activity for the two swims did not differ(Table 2). Interestingly,swimming performance was repeatable without Ṁ_O2 being restored to within 5% of Ṁ_O2routine in 77 out of the 107 tests (35 out of 37 tests for GC sockeye salmon, 6 out of 6 tests for ES sockeye salmon, 21 out of 32 tests for WVR sockeye salmon, and 14 out of 32 tests for CHE coho salmon). Because fish performed equally well on their first and second swim tests, averaged U_crit and Ṁ_O2max values for the two tests were used to analyze the effects of temperature.

ES sockeye salmon had a significantly higher U_crit(P<0.05) than either GC or WVR sockeye salmon stocks but at a comparable ambient water temperature (Table 3). Conversely, CHE coho salmon swam as well as WVR sockeye(P>0.05), but at a lower ambient temperature(Table 3). Therefore,stock-specific differences existed independent of temperature differences.

The swimming performance of some stocks did not vary between field and laboratory tests (Tables 1,2,3). In addition, preliminary laboratory experiments conducted with CHE coho salmon (N=4) showed that routine Ṁ_O2, U_crit and Ṁ_O2max(2.59±0.14 mg O₂ kg^-1 min^-1;1.72±0.12 BL s^-1; 9.19±0.61 mg O₂kg^-1 min^-1, respectively) were not statistically different compared with field tests (Tables1,2,3). Similarly, preliminary laboratory experiments with GC sockeye salmon (N=4) showed that U_crit and Ṁ_O2max(2.15±0.11 BLs^-1; 14.71±0.69 mg O₂ kg^-1 min^-1) were not statistically different compared with field tests (Tables2,3), although Ṁ_O2routine(3.31±0.43 mg O₂ kg^-1 min^-1) was significantly (P<0.05) lower than field tests(Table 2).

The two sets of field measurements for CHE coho salmon were pooled for subsequent analyses because there were no significant differences (Tables1,2,3). In contrast, WVR sockeye salmon tested at the SFU laboratory had a significantly higher U_crit (23%), Ṁ_O2max (19%) and scope for activity (25%) compared with the same stock tested in the field when the fish were in a slightly more mature state and also at a temperature 4°C colder (Tables 1,2,3). Consequently, the two data sets for WVR sockeye salmon were treated separately for subsequent analyses.

Ṁ_O2routinemeasured at ambient temperature could vary significantly, but not always among stocks, between years and between species(Table 2). To investigate the influence of ambient water temperature on Ṁ_O2routine, all stocks were pooled and a statistically significant (P<0.05)exponential relationship existed between Ṁ_O2routine and ambient water temperature that accounted for 65% of the variation in the individual data (Fig. 2). Addition of temperature-adjusted fish to this pooled data set slightly weakened the relationship (r²=0.52; P<0.05)(Fig. 2).

Regression analysis was performed for three salmon stocks (GC, WVR and CHE), revealing significant (P<0.05) bell-shaped relationships between Ṁ_O2maxand temperature (Fig. 3A) when data from temperature-adjusted fish were included. Temperature optima for Ṁ_O2max were interpolated from the regression equations (GC=17.5°C; WVR=15.0°C;CHE=8.5°C) and were found to correspond closely to the ambient water temperature for each stock (Table 1; Fig. 3A). Furthermore, when individual Ṁ_O2max values were compared among sockeye stocks and at common ambient temperatures, there were clear differences between stocks (Fig. 3A). These results suggest that important stock-specific differences existed for Ṁ_O2max and its thermal sensitivity. Similarly, significant (P<0.05) bell-shaped regressions were found between scope for activity and temperature for each salmon stock (Fig. 3B). The temperature optimum for scope for activity was either similar to that for Ṁ_O2max (CHE stock), or 1°C lower (WVR and GC stocks)(Fig. 3B), reflecting the important contribution of temperature on Ṁ_O2routine.

For GC and WVR sockeye salmon stocks, there were significant(P<0.05) bell-shaped regressions between U_critand temperature, with U_crit falling off at temperatures>19°C and >16°C, respectively(Fig. 4). For CHE coho salmon,the regression between U_crit and temperature was not significant (P=0.71) (Fig. 4).

The increase in Ṁ_O2 with swimming speed is illustrated for three salmon stocks from the various test locations(Fig. 5A). As expected, Ṁ_O2 varied exponentially (r²=0.99) with swimming speed for WVR sockeye salmon and CHE coho salmon. However, for GC sockeye, the data were not satisfactorily fitted by an exponential relationship and a sigmoidal regression (r²=0.99) was required to account for the plateau in Ṁ_O2prior to U_crit. Only Ṁ_O2routine and Ṁ_O2max were measured for ES sockeye salmon and these values are included inFig. 5A-C.

Ṁ_O2 differed significantly (P<0.05) among the salmon stocks at intermediate swimming velocities, with GC sockeye salmon having the highest Ṁ_O2 values and CHE coho salmon the lowest values for a given swimming velocity. The cost of transport showed typical U-shaped curves with the exception of GC sockeye salmon (Fig. 5B), because GC sockeye salmon were tested at the highest temperature and Ṁ_O2routineincreased exponentially with temperature. GC sockeye salmon were the least economical swimmers. The difference among stocks could simply reflect a higher Ṁ_O2routine. However, this was found not to be the case because the net cost of transport was also elevated for GC sockeye salmon(Fig. 5C). In contrast, because WVR sockeye salmon and CHE coho salmon had a similar net cost of transport,the small differences in the cost of transport between these two stocks were likely a result of temperature effects on Ṁ_O2routine. The minimum cost of transport occurred at around 1 BL s^-1 for all three salmon stocks (Fig. 5B).

This study is the first to extensively examine the role of temperature on swimming energetics within and among different stocks of adult Pacific salmon under field and laboratory settings. With a total of 107 adult salmon tested,it is also the most comprehensive study of adult salmon swimming performance to date.

Although the blocking effect for a few of the fish was high, the U_crit and Ṁ_O2max data obtained here for sockeye salmon are entirely consistent with earlier laboratory and field studies involving adult Pacific salmon (e.g.Brett and Glass, 1973;Jones et al., 1974;Williams et al., 1986). For example, U_crit (2.41 BL s^-1, N=8) reported for smaller (1.65±0.07 kg) adult sockeye salmon(Brett and Glass, 1973) lies in the upper end of our U_crit range, while our Ṁ_O2max data tend to be higher than theirs at corresponding temperatures(Fig. 3C). Ṁ_O2max (13.83 mg O₂ kg^-1 min^-1) and U_crit(2.33 BL s^-1) for pink salmon(Williams et al., 1986) are comparable to the present study. The exponential relationships between Ṁ_O2max and temperature reported earlier for adult sockeye salmon either lie below(Davis, 1966) or above(Brett and Glass, 1973) a significant exponential relationship (r²=0.63;Fig. 3C) that could be fitted to our Ṁ_O2maxdata. (Note: this exponential regression tended to over-represent WVR sockeye salmon and under-represent both ES and GC sockeye salmon, i.e. there was a poor fit for any of the individual salmon stocks.) The high quality of the present data was also illustrated by the repeatability of the swim tests,because RR decreases significantly (Jain et al., 1998; Tierney,2000) when rainbow trout Oncorhynchus mykiss and sockeye salmon are either sick or have been challenged by toxicants.

The final temperature preferendum paradigm, proposed by Fry(1947), embodied three principal inferences: a species-specificity to the final temperature preferendum; a relationship between the final temperature preferendum and field distribution; and a relationship between final temperature preferendum and the temperature at which centrally important processes take place at maximum efficiency. Indeed, the concept of temperature optima for physiological processes related to swimming is well documented (see reviews byBeamish, 1978;Houston, 1982;Guderley and Blier, 1988;Hammer, 1995;Kelsch, 1996;Johnston and Ball, 1997;Kieffer, 2000). Furthermore,there is emerging evidence that maximum cardiac performance and the oxygen supplying the cardiac tissue may be `centrally important processes' (Farrell,1997,2002), though other processes are likely to be important (Pörtner,2002). The present results therefore extend the idea of temperature optima to include the possibility of stock-specific temperature optima, in addition to confirming an important temperature effect on the physiological processes that determine Ṁ_O2max, scope for activity and U_crit. Three stocks of adult salmon demonstrated distinct temperature optima for Ṁ_O2max and scope for activity, while GC and WVR sockeye salmon also exhibited temperature optima for U_crit. In contrast, U_critfor CHE coho salmon displayed low temperature sensitivity. Temperature optima around 15°C have been reported previously for Ṁ_O2, metabolic scope and sustained cruising speed with juvenile and adult sockeye salmon(Brett and Glass, 1973). While this temperature is very close to the temperature optima reported here for GC sockeye salmon, there were clear differences in the temperature optima for WVR sockeye salmon (Fig. 3). Furthermore, the temperature optima adult CHE coho salmon are considerably lower than that reported earlier for juvenile coho salmon (approx. 20°C;Brett et al., 1958), a difference that could reflect either a stock-specific effect or developmental effect.

Although temperature optima were clearly established for the salmon stocks,a measure of temperature insensitivity for peak swimming capability is likely to be critical for these salmon stocks because they routinely face varying water temperatures. To gauge temperature insensitivity we used the regression equations to arbitrarily estimate the temperature range over which a salmon stock could reach at least 90% of its peak Ṁ_O2max. These temperature ranges were: 14.7-20.3°C for GC sockeye salmon,12.7-17.3°C for WVR sockeye salmon and 5.0-11.4°C for CHE coho salmon. Using a similar analysis for scope for activity, the temperature ranges for the three stocks were similar to those for Ṁ_O2max, but marginally cooler and/or narrower (13.9-19.3°C for GC sockeye salmon,12.8-16.2°C for WVR sockeye salmon and 6.6-8.9°C for CHE coho salmon). This analysis clearly shows that these three salmon stocks can approach their respective peak aerobic activity over a temperature range spanning as much as 5°C. Such physiological flexibility may be adaptive (see Guderley and Blier, 1998) because the water temperature in the Fraser River may vary annually on a given date by as much as 6°C, perhaps even providing sufficient flexibility to handle all but the most extreme temperature conditions encountered in the Fraser River during a particular migration window. This does not mean that extreme temperature and/or hydrological conditions (known to occur in certain years) would not impose difficulties for migration (e.g. ES sockeye have faced water temperatures reaching 22°C and flows of 9000 m³ s^–1;Macdonald et al., 2000). But it does mean that physiological information provided here could be useful in predicting which river conditions are more likely to impair passage and reduce spawning success.

An equally important discovery was that these temperature optima correlated very closely with the ambient water temperature of the natal river for individual salmon stocks. This meant that the lowest thermal sensitivity of peak aerobic performance occurred around the ambient temperature of the natal stream. While environmental variability can differentially influence the ability of organisms to survive and reproduce, tailoring populations to their respective environmental niches (Cooke et al., 2001), and the `stock concept' suggests adaptation to local conditions (Berst and Simon,1981), migratory salmon spend most of their life away from the natal streams. Whether the present correlation between temperature optima and natal stream temperature is a reflection of adult salmon being pre-adapted to water temperatures likely to be encountered during river migrations, or is coincidental with the temperature preferendum of the species (e.g. for sockeye salmon 14.5°C, Brett, 1952;10.6-12.8°C, Horak and Tanner,1964), will require further study. Further work will also need to tackle the possibility of rather rapid thermal compensation during the actual in-river migration. For some sockeye salmon stocks, the timing of migration seems to have been far too restrictive for thermal compensatory processes to take full effect. For example, ES sockeye move from seawater in the Georgia Strait, where they encounter temperatures likely to be no warmer than 13°C, into river water as high as 18°C, and then within 4 days face one of their most difficult in-river swimming challenges, Hell's Gate.

Although the majority of tests were performed at ambient water temperature,small temperature adjustments were used to extend the ambient temperature range. Acclimations to these temperatures were necessarily short (5 days)because fully ripe salmon have compromised swimming ability(Williams et al., 1986). While the short acclimation period is a concern, the extent of the temperature change (<6°C) was not unusual compared with changes naturally encountered, because ES and Chilko stocks of adult sockeye salmon routinely face temperature changes of 1°C daily and as much as 7°C over 1 day during their migrations (Idler and Clemens, 1959). In addition, individual fish tested at either their ambient temperature or an adjusted temperature showed a reasonable overlap of Ṁ_O2maxvalues (Fig. 3A). A second concern is that we did not consider sex differences in swimming energetics. This concern is offset by the fact that we used equal numbers of male and female fish in many test groups. Furthermore, physiological telemetry studies of migrating ES sockeye salmon and Seton River pink salmon have revealed little difference between sexes in terms of the overall cost of transport to the spawning site, although males were less efficient at migrating through hydraulic obstacles (Hinch and Rand,1998; Standen et al., 2002). The present study could be used as a framework for future studies of sexual dichotomy in swimming capabilities.

The finding that Ṁ_O2routineincreased exponentially with temperature is consistent with previous studies showing exponential relationships for both Ṁ_O2routine and standard metabolic rate (rainbow trout;Dickson and Kramer, 1971),brown trout Salmo trutta; Butler et al., 1992), sockeye salmon(Brett and Glass, 1973),tilapia Sarotherodon mossambicus;Caulton, 1978) and largemouth bass Micropterus salmoides; Cooke et al., 2001). As expected, our Ṁ_O2routine values for adult sockeye salmon were higher than the standard metabolic rate previously reported (Brett and Glass,1973). Some of this difference could be attributed to the on-going gonadal development in the mature fish used in the present study. It is also possible that the overnight recovery is insufficient (seeFarrell et al., 2003) and adult salmon are more restless than less mature fish. Williams et al.(1986) noted that adult pink salmon were more restless than sockeye salmon in swim tunnels.

The intraspecific differences in migration capacity were sometimes correlated with in-river migration distance and difficulty. For example, ES sockeye salmon, the furthest migrating stock of any of the Fraser River salmon, attained a significantly higher U_crit at a 5°C cooler temperature and were more efficient swimmers at U_crit because of a lower Ṁ_O2max compared with GC sockeye salmon. These attributes of ES sockeye salmon may be advantageous because they migrate almost three times the distance up the Fraser River compared with GC sockeye salmon(Table 1). Similarly, both ES and GC sockeye salmon migrate much longer distances and negotiate more severe hydraulic challenges compared with the coastal WVR sockeye salmon and correspondingly had a larger scope for activity at comparable water temperatures. Moreover, almost all of the GC and ES fish repeated their swimming performance without recovering Ṁ_O2 to within 5%of Ṁ_O2routine. The differences in swimming energetics found between CHE coho salmon and WVR sockeye also probably reflect species-specific adaptations. Yet, because these two salmon stocks face almost identical in-river migration distances and conditions, other factors must be involved in these adaptations. Thus, the suggestion that distance and/or difficulty of migration are powerful selective factors acting on salmonids (Bernatchez and Dodson, 1985; G. T. Crossin, S. G. Hinch, A. P. Farrell, D. A. Higgs, A. G. Lotto, J. D. Oakes and M. C. Healey, unpublished observations) is supported by the present study.

Intraspecific adaptation of maximum swimming ability has been previously established for juvenile salmonids either held or reared in a laboratory, but to our knowledge not for adult, wild salmon. For example, juvenile Pacific coho salmon from an interior river had an inheritable trait that resulted in a lower initial acceleration for a fast start, but a longer time-to-fatigue at a constant swimming speed compared with coho salmon from a coastal river(Taylor and McPhail, 1985). Similarly, Pacific steelhead trout O. mykiss from an interior river also had a greater time-to-fatigue for an incremental swimming speed test and allelic differences in the lactate dehydrogenases compared with a coastal stock (Tsuyuki and Williscroft,1977). Nevertheless, because Bams(1967) found that rearing conditions could alter swimming performance of sockeye salmon fry, phenotypic rather than genotypic expression could have contributed to the differences we observed.

Paradoxically, GC sockeye salmon were less efficient swimmers than either WVR sockeye salmon or CHE coho salmon, and why this is so is unclear. GC sockeye salmon were unusual in another regard, the plateau in Ṁ_O2 as the fish neared U_crit. In fish, Ṁ_O2 typically increases exponentially with swimming speed (seeBeamish, 1978) to overcome drag, which is exponentially related to water velocity(Webb, 1975). Rarely is a plateau observed in Ṁ_O2 even though fish progressively increase the anaerobic contribution to swimming at 75%U_crit (Brett and Groves, 1979; Burgetz et al.,1998). Consequently, the plateaus for both Ṁ_O2 and the cost of transport curve near U_crit for GC sockeye salmon point to an unusually high contribution of anaerobically fueled locomotion. This possibility is further explored in the accompanying paper(Lee et al., 2003b), in which excess post-exercise oxygen consumption is examined as a measure of the anaerobic swimming activity.

In summary, we conclude that variation in Ṁ_O2routine among adult salmon stocks was primarily due to differences in water temperature. In contrast, distinct temperature optima for Ṁ_O2max were evident among salmon stocks, which when combined with differences in scope for activity and U_crit, suggest stock-specific as well as species-specific differences in the temperature sensitivity of the physiological mechanisms that underpin oxygen delivery during swimming in adult Pacific salmon.

The work was supported by research grants from the Forestry Renewal BC program (to A.P.F.) and a strategic grant from the Natural Sciences and Engineering Research Council of Canada (to M.H., S.H. and A.P.F.). Fisheries and Oceans Canada, BC Hydro and First Nations (Cayoose, Chehalis and Yale Bands) contributed significantly by permitting use of their facilities,providing fish, or providing support in kind. We are indebted to the Science Workshop and Electronics Workshop at Simon Fraser University who built the entire apparatus, assisted in redesigning the swim tunnels and trailers, and developed the software for data acquisition.