In this experiment we tested seven hatchery-sourced populations of juvenile Chinook salmon from the states of Washington, Oregon and California in the United States of America (Figure 1). Of the seven populations, five exhibit the Fall-run migratory phenotype [Priest Rapids (WA), Trask (OR), Elk River (OR), Trinity (CA), Feather Fall-Run (CA)]. Adult Fall-run Chinook salmon return to freshwater during the Fall months and spawn shortly after arrival to their spawning grounds (43). Embryos and alevin incubate through the winter months and the young predominantly outmigrate during their first spring. Additionally, we tested two populations of early migrating Chinook salmon from California, the threatened Central Valley Spring-run from the Feather River hatchery and the critically endangered Sacramento River Winter-run from the Livingston Stone hatchery. Both populations exhibit early migration with adults entering freshwater during either the spring or winter months, respectively (43, 44). Both populations historically migrated to high elevation headwaters. Spring-run accessed snow-fed rivers in the Sierra Nevada and Southern Cascades while the Winter-run, endemic to northern California, migrated to cold spring-fed rivers in the upper Sacramento River watershed. Components of this research as well as greater detail on these studied populations has been published in (17, 66).

Chinook salmon smolts (N = 470, mass 22.7 ± 4.4 g; fork length 12.4 ± 0.8 cm; μ ± S.D., n = 46–61 per population) from seven hatchery populations (Figure 1) were acclimated to one of three temperatures (11, 16, or 20°C), for a total of 21 treatment groups (n = 8–22 per treatment group) following methods in Zillig et al. (17). Temperatures were chosen to be ecologically relevant to the rearing and outmigration conditions that a naturally-spawned juvenile Chinook salmon may encounter (37). Experiments were conducted between 2017 and 2019 with two populations assessed in 2017 and 2018 and three in 2019. All experiments were conducted in accordance with the Institutional Animal Care Committee of UC Davis Protocol # 19928.

The studied populations (Figure 1) came from seven hatcheries in six defined Chinook salmon evolutionary significant units [ESU, (45), Trask and Elk River populations share an ESU]; each ESU is a conservation unit managed as a distinct taxonomic unit (46). Hatcheries strip spawned returning adults and incubated the fertilized embryos in systems using natural surface water from each populations resident watershed. We received progeny as either eyed embryos or young fry. Eyed eggs from the Priest Rapids population were received via overnight mail, surface sterilized with iodophor upon arrival, and incubated at 9°C until they hatched and began to feed. Fish from all other populations were acquired from their respective hatcheries when of transportable size (~1–2 g) and trucked to the Center for Aquatic Biology and Aquaculture (CABA) at the University of California, Davis campus in a 765 L tank. Dissolved oxygen was maintained through controlled bubbling of compressed air and water temperatures (7–11°C) were reduced with bagged ice as needed during transit. At CABA, rearing and treatment tanks were supplied with temperature-controlled, flow-through fresh water from a dedicated well and continuously aerated with air stones. Collected and hatched fry from all populations were reared at 11–13°C until distributed into their acclimation treatment tanks (2–89 days depending on population and size at acquisition). Each treatment group (population × acclimation temperature) was housed in two, replicate, 400 L, with 55–105 fish per tank. Fish were exposed to natural photoperiods (38°55′N) and automated belt feeders delivered an excess ration (4% of body mass per day) of 2–4 mm Sinking Salmon Feed (Skretting, USA) during daylight hours. Rations were updated biweekly to account for fish growth and tank density. Acclimation temperatures were achieved by adjusting tank temperature by ~1.5°C per day. Once tanks achieved their specific acclimation temperature, fish were acclimated for at least 60 days prior to any CT_MAX trials. Mean tank temperatures (± SEM) were 11.1 ± 0.02°C (n = 14), 16.0 ± 0.03°C (n = 14) and 19.8 ± 0.03°C (n = 14). Tank temperatures were maintained for the duration of the experiments (4–9 months) and fish were tested once they reached the target size (~20 g). As fish growth is temperature dependent [see (17) for growth rates], fish acclimated to 20 and 16°C were acclimated to fewer days (average acclimation: 118 and 119 days, respectively) than fish acclimated to 11°C (150 days) prior to their exposure.

Critical thermal maxima (CT_MAX) trials were conducted according to established methods (47) to determine a fish's maximum acute thermal tolerance. Fish were tested individually, each in their own beaker. Six or seven 4L Pyrex beakers were contained within a fiberglass bath (1 × 2 × 0.2 m) and individually aerated with an air stone to ensure normoxic conditions throughout the trial. The volume of water in each beaker (~2.5 L) was calibrated to ensure even heating across all CT_MAX beakers (0.33°C min⁻¹). Two pumps (PM700, Danner, USA) were used to circulate water: one pump moved water across three heaters (Process Technology S4229/P11), and the other recirculated heated water through the CT_MAX bath via a submerged manifold. Bath temperature for each trial began at the acclimation temperatures (11, 16, or 20°C) of the tested fish. Prior to CT_MAX trials, fish were transferred from treatment tanks to separate tanks for fasting. Fish reared at 20 and 16°C were fasted for 24 h and 11°C fish were fasted for 48 h to account for their slower metabolic rate and ensure they were in a fasted state. After fasting, individual fish (n = 8 to 22 per treatment, 22.72 ± 4.40 g mean ± S.D.) were gently netted and placed into individual chambers within the CT_MAX bath. After a 30-min acclimation period (48), the CT_MAX trial began.

The test chamber temperature was recorded every 5 min using a thermocouple (Omega HH81A) routinely calibrated to a standardized mercury thermometer. Fish were continually observed for signs of distress and loss of equilibrium (LOE). When a fish lost equilibrium the CT_MAX trial was concluded and the temperature of the CT_MAX beaker was recorded (15, 49). The fish were then immediately transferred to a recovery bath at their acclimation temperature where they were individually housed without food. The following day (μ ± S.D.: 24 h 12 m ± 1 h 1 m, range: 22 h 2 m−26 h 55 m) fish were retested using the same protocol and placed within the same test chamber to measure their heat-hardened CT_MAX. Once completed, fish were again returned to their recovery chambers for ~24 h. After the final 24-h recovery period, fish were humanely euthanized in a buffered solution of MS-222 (0.5 g L⁻¹ buffered with 0.42 g L⁻¹ of NaHCO3 and 6.0 g L⁻¹ of NaCl) and then weighed (wet mass ± 0.01 g) and measured (fork length ± 0.1 cm).

A total of 470 fish were tested for heat hardening, but fish whose trials were impacted by experimenter error (n = 11), exhibited unusual behavior (n = 8, discussed further in results) or did not recover from either the first (n = 33) or second (n = 33) CT_MAX trial were not included in the analysis, resulting in 385 fish that recovered within 24 h from both the first and second CT_MAX trial. We quantified the heat hardening response (ΔCTM) as the difference between an individual fish's second CT_MAX trial and its first, with positive ΔCTM values indicating an increase in CT_MAX from the first trial to the second. Gunderson (30) highlights the possibility of spurious thermal trade-off results (30) due to regression to the mean (RTM). If the correlation of two datasets is not perfect, then there will be RTM. In the case of repeated CT_MAX trials RTM would be expected to yield results where fish with higher initial CT_MAX will exhibit a relatively worse performance on their 2nd CT_MAX trial then fish with initially poor CT_MAX values. Taken collectively, this result would mimic a thermal trade-off, where the plastic response (i.e., ΔCTM) of fish or populations with high CT_MAX values is less than that of fish or populations with low CT_MAX values. We adjusted a fish's second CT_MAX value using the RTM correction to adjust CT_MAX data. Put briefly, this statistical transformation approach assesses the change in variance in CT_MAX values between the first and second trials to calculate the strength of RTM effect [please refer to Kelly and Price (78) for the mathematical theory and Gunderson (30) for its application to CT_MAX data from multiple ectothermic species]. We investigated thermal trade-offs at two scales. The first was at the organismal scale where each fish's ΔCTM value was compared against its first to test for a tradeoff among treatment groups. We then investigated whether populations with a higher initial CT_MAX exhibited a reduced ΔCTM by using the model estimated mean values for each treatment (population × acclimation temperature).

We tested for relationships between physiological performance and seven environmental parameters including population latitude and six stream temperature metrics (Table 1). Population latitude, a common metric for assessing biogeographic patterns in physiology (12, 15), was defined by the location of each population's hatchery using Google Earth. Stream temperature metrics were estimated from the predictions of a spatial stream network (SSN) model (37, 50). The SSN model is a tailored regression model that incorporates the complex spatial autocorrelation of streams due to the direction of flow and nested connectivity. The SSN model accurately predicted stream temperatures across the western U.S. throughout the year (R2 = 0.928 for the out-of-sample testing dataset), and mean monthly stream temperatures were predicted for 465,775 river km in the western U.S. (37). We extracted mean monthly stream temperature predictions from the SSN (average of, 2002–2011) for each river kilometer within a population's current natural (i.e., not the hatchery) and historical natal distribution (Figure 2). Juvenile natal rearing habitat was defined for each population from observed spatial distributions of natural spawning (37). To estimate each population's historical pre-dam thermal regime, we defined potential rearing habitat upstream of dams until habitats were bounded by river slope, flow, natural barriers, or intermittency (37, 50–52). From this dataset we calculated the annual maximum mean monthly temperature (A_MAX), the maximum mean monthly temperature during the rearing period (R_MAX), and the thermal range between the annual maximum and minimum mean monthly temperatures (A_RANGE). These were quantified from current (-Cur) and historical (-His) ranges (e.g., Historical annual maximum mean monthly temperature: A_MAX-His). For greater detail see description in (17).

CT_MAX and heat hardening response were modeled with linear mixed effect models implemented using the Bayesian statistical package “brms” (53, 54) in R (version 4.2.2). We used the packages “ggplot2” (55) to visualize model output and “emmeans” (56) to calculate marginal means of the models. Visual inspection of Q-Q plots indicated several outliers which could not be explained as erroneous data. We therefore modeled the observed data using a Student's t-distribution as opposed to a Gaussian distribution as the former produces more robust estimates in the presence of outliers. Priors were uniformed and centered on zero. Models were built using a forward selection of potential predictor variables guided by the biology of the fish and experimental design. We used leave-one-out cross-validation (LOO) to determine the best-fitting models, selecting the model with the lowest LOO score. We assembled three suites of models: “Treatment models,” “Trade-off models,” and “Environmental models,” discussed in detail below.

During CT_MAX experiments, acute thermal tolerances ranged from 20.8 to 31.0°C (μ = 28.6 ± 1.2°C; Table 2). We found that the CT_MAX values were sensitive to the population of origin, the acclimation temperature, and whether the fish had undergone a prior CT_MAX experiment. CT_MAX estimates generally increased with acclimation temperature, with fish acclimated to 20°C exhibiting the highest thermal tolerances (μ = 29.5 ± 1.0°C), particularly during their second CT_MAX challenge (1st CT_MAX: 29.3 ± 0.9 and 2nd CT_MAX: 29.6 ± 1.0°C). The overall heat hardening response (ΔCTM) was nearly zero (−0.07 ± 1.23°C, μ ± S.D.), however this response was also found to vary by population and acclimation temperature, with fish acclimated to cold temperatures (11°C) exhibiting primarily negative ΔCTM values (−0.8 ± 1.4°C) and those acclimated to warmer temperatures (16 or 20°C) exhibiting primarily positive values (16°C: 0.2 ± 0.9°C, 20°C: 0.6 ± 0.8°C, μ ± S.D.). More detail on the observed data can be found in Table 2.

Bayesian models were used to determine the mean acute thermal tolerances for each treatment group, and we used LOO to determine the best model (Supplementary Table 1). There were two models of fish CT_MAX for which LOO values were not significantly different. The difference between the predicted acute thermal tolerances of these two models was 0.001 ± 0.029°C. Therefore, we conducted our analysis with the lowest LOO model which included a three-way interaction of categorical predictors for acclimation temperature, hatchery, and whether it was a fish's first or second trial. There was an additional fixed effect for fish mass and a random effect for fish ID. The discarded model was identical with an additional random effect for CT_MAX chamber. The highest model-estimated CT_MAX values were exhibited by the Trask River population reared at 20°C during the second CT_MAX trial {30.7°C [89% credible interval (C.I.) 30.5–30.9°C]}. The lowest values were exhibited by the Priest Rapids and Winter-run populations acclimated to 11°C during their second CT_MAX trial [27.1°C (89% C.I. 26.8–27.4°C) and 27.3°C (89% C.I. 27.0–27.6°C), respectively]. Acclimation duration was not a predictor in the lowest-scoring model, nor did it yield any significant associations with CT_MAX in models that contained it.

The capacity for individual juvenile Chinook salmon to physiologically improve their CT_MAX performance upon their second trial varied among populations and acclimation conditions with capacity increasing up to 3.2°C (Feather Spring-run individual acclimated to 20°C) and other fish losing capacity by up to −7.9°C (Winter-run individual acclimated to 11°C). There were two models of individual ΔCTM for which LOO values were not significantly different. Comparing the predicted values between these models revealed a mean difference of 0.016 ± 0.12°C. Therefore, we conducted our analysis with the lowest LOO-scoring model which included a three-way interaction of a fish's initial CT_MAX, acclimation temperature, and hatchery, an additional fixed effect for fish mass, and a random effect for CT_MAX chamber. Acclimation duration was not a predictor in the lowest-scoring ΔCTM model, nor did it yield any significant associations with ΔCTM in models that contained it.

Across all populations, when fish were acclimated to 11°C their 2nd CT_MAX was lower than their first (ΔCTM = −1.1° to −0.2°C). In contrast, the heat hardening response was positive with higher 2nd CT_MAX measurements among 12 of 14 treatment groups that were acclimated to 16 or 20° (Figure 3A) with the Trinity population being the only exception. Populations also differed in the amount of change in ΔCTM associated with changes in acclimation. For instance, the Priest Rapids hatchery exhibited a 2.5°C change in the heat hardening response between fish acclimated to 11 vs. 20°C. Meanwhile, the Trinity population exhibited a small but positive (0.5°C) change in ΔCTM from fish acclimated to 11–20°C (Figure 3B).

Populations differed in the amount of acclimation capacity they possessed (defined as the difference between treatment mean model-estimated CT_MAX of 11°C fish on their first thermal trial and the treatment mean model-estimated CT_MAX of 20°C on their second thermal trial). The Priest Rapids and Trask hatchery populations exhibited the greatest amount of acclimation capacity [2.6°C (89% C.I. 2.3–2.9°C) and 2.4°C (89% C.I. 2.1–2.7°C), respectively], while the Feather Fall-run and Trinity populations exhibited the least [1.2 (89% C.I. 0.9–1.5°C) and 0.9°C (89% C.I. 0.5–1.3°C), respectively].

We found evidence supporting the TOH between CT_MAX and ΔCTM among individual fish (Figure 4A). Only the Trinity hatchery, across all acclimation temperatures, exhibited non-significant associations (where the 89% credible interval overlapped 0) between a fish's initial CT_MAX and its ΔCTM, and therefore, no evidence of a tradeoff. For all other populations we found that a fish's initial CT_MAX value was significantly negatively associated with ΔCTM. We did not, however, find evidence of the TOH between modeled estimated CT_MAX and ΔCTM among treatment groups (i.e., population × acclimation temperature), indicating that populations exhibiting high thermal tolerance did not exhibit a reduced heat hardening response (Figure 4B).

In addition to our trade-off models, we investigated whether ΔCTM and acclimation capacity were associated with environmental characteristics and latitude of each population. The best fitting (lowest LOO-score) environmental model of ΔCTM included an interaction between the historical maximum mean monthly temperature experienced during the rearing period (R_MAX-His) and acclimation temperature, and a fish's condition factor as well as random effects of population and CT_MAX beaker (see Supplementary Tables 4, 5 for all environmental predictor models). Latitude as a predictor was only found in low LOO-scoring models alongside R_MAX-His, in these models, latitude alone did not yield a significant association with ΔCTM. The relationship between R_MAX-His and ΔCTM varied by acclimation temperature (Figure 5). Among fish acclimated to 11°C there was a significant negative association between R_MAX-His and ΔCTM (Figure 5). The relationship among fish acclimated to 16°C was slightly positive but was not significantly different from 0. When fish were acclimated to 20°C, the relationship was positive and significant. However, the 16 and 20°C acclimation groups were not significantly different from each other (Figure 5).

The lowest LOO-scoring model for acclimation capacity and environmental traits included population latitude and the R_RANGE-His. There was a significant positive association between acclimation capacity and the R_MAX-His, with populations from historically warmer habitats exhibiting a larger acclimation capacity (Figure 6A). Separately, latitude was also a significant predictor of acclimation capacity, with populations from higher latitudes exhibiting greater acclimation capacity (Figure 6B; r = 0.53). All other associations between acclimation capacity and current or historical environmental predictors were not significantly distinguishable from zero (Figure 6C).

Experimental mortality was slightly less during the first CT_MAX trial (n = 33 of 461, 7.2%) than during the second trial (n = 33 of 418, 7.9%). Of the 33 fish that did not recover from their second CT_MAX trial, 27 had initial CT_MAX values that were outside of the 89% credible interval of their respective treatment's modeled mean performance (13 below, 14 above). All of the eight fish that were discarded for unusual behavior involved fish losing equilibrium during the acclimation period prior to their second CT_MAX trial or early in the trial when water temperatures were within 3°C of their initial temperature.

In October 2018, one tank of 20°C acclimated Winter-run Chinook salmon experienced an outbreak of columnaris and subsequent mortality (n = 7). The mortality in this tank is theorized to be a result of temperature stress after being housed at 20°C for 202 days. Collection of CT_MAX data preceded disease onset by 41 days and data are presumed to be unaffected. However, the loss of animals limited our ability to continue to gather CT_MAX data and so reduced the treatment's sample size (Winter-run acclimated at 20°C, n = 8).

The vast majority of previous research on heat hardening and our estimates of heat hardening among Chinook salmon acclimated to 16 or 20°C have demonstrated an improvement in acute thermal tolerance (Table 3). We found that while Chinook salmon populations differed in their ΔCTM performance, there was a consistent effect of acclimation temperatures across populations, with fish acclimated to 16 or 20°C exhibiting a positive ΔCTM. However, our discovery of significant, “heat weakening” (i.e., negative ΔCTM) exhibited by all populations when acclimated to 11°C is unique. To our knowledge, the only other examples of “heat weakening” occurred when Red Shiner (Notropis lutrensis) and Fathead Minnow (Pimephales promelas) were acclimated to temperatures ~17°C below their initial CT_MAX values, but this “heat weakening” was not significant (20). Our 11°C acclimated Chinook salmon were also acclimated to ~17°C below their eventual CT_MAX (average 28.0°C), but “heat weakening” in our study was significant. It may be that exposing fish to near-lethal thermal extremes when their thermal history is so ill-matched incurs additional physiological damage leading to the “heat weakening” effect we encountered.

The Trade-off Hypothesis (TOH) (25, 26) posits that organisms with a high heat tolerance will exhibit a reduced plasticity, arising from the idea that organisms exhibiting a higher CT_MAX are likely closer to a hypothetical physiological maximum and can therefore exhibit a reduced plastic capacity (e.g., ΔCTM or acclimation capacity). We used repeated CT_MAX trials to assess the TOH among juvenile Chinook salmon both within and between populations. Among individual fish, we found a consistent tradeoff wherein a fish with a higher initial CT_MAX exhibited a lower ΔCTM. These results may indicate the presence of a “concrete ceiling” of acute thermal performance within a population (64), with individuals losing plasticity as their tolerance approaches a fixed limit. A fixed thermal ceiling may indicate a limited adaptive capacity within a population to evolve greater thermal tolerance.

Interestingly, we did not find evidence for the TOH when populations are compared against one another. After controlling for acclimation temperature, there was no evidence for a trade-off between a population's CT_MAX and its average ΔCTM. One interpretation of this mixed support for TOH between individuals and populations is that while populations may be approaching their respective thermal ceilings, the species as a whole retains adaptive capacity, potentially enabling Chinook salmon to adapt their acute thermal tolerance during future colonization and radiation.

Salmonid physiology has been linked to population-specific environmental characteristics (17, 38, 65). In this experiment, we tested seven populations of juvenile Chinook salmon that differed in their migratory phenology (five fall-run populations, one spring-run population and one winter-run population) and thermal physiology (17, 66). In the case of acute thermal maxima, past research by Zillig et al. (17) on a subset of these populations indicated that modeled historical measures of stream temperatures were better predictors of population performance than current estimates. Likewise, work by Eliason et al. (38) found that the metabolic performance of a population of adult Sockeye salmon (O. nerka) were better suited to historical river temperatures rather than current ones. We expanded upon these results, documenting how the heat hardening response and acclimation capacity in CT_MAX exhibited by different populations was associated with several environmental characteristics, particularly the historical maximum monthly mean temperature experienced during the rearing period (R_MAX-His). We found that when fish were acclimated to 11°C, populations from habitats with cooler R_MAX-His performed better than those populations from warmer habitats. Among fish acclimated to 20°C, the opposite pattern was observed, where fish from cold habitats elicited worse ΔCTM than fish from historically warmer habitats. These results are consistent with local adaptation of ΔCTM, wherein fish that may experience warmer habitats exhibit an improved capacity to withstand those conditions. Furthermore, we find a positive association between R_MAX-His and acclimation capacity indicating that fish which may have historically encountered higher temperatures are more capable of acclimating to those regimes. Taken together, these results point toward populations from warmer habitats exhibiting greater capacity to both acclimate and withstand thermal extremes, traits that may become increasingly important in a warming world, whereas populations from unusually cold habitats may be expected to exhibit reduced thermal tolerance and acclimation capacity.

Counterintuitively, we also found a significant, positive association between population latitude and acclimation capacity, with populations from higher latitudes exhibiting greater acclimation capacity than those from lower latitudes. This may be explained by the hydrologic idiosyncrasies of our selected populations which include several Northern Californian populations resident to snowmelt-fed rivers and downstream of large, deep storage reservoirs, rain-fed coastal populations (Trask and Elk River) and the Priest Rapids population downstream of run-of-the-river hydropower dams (67). This result highlights the value of local population-specific knowledge and the risk of relying solely upon broad-scale geographical patterns (e.g., latitude) when predicting population response (37).

For instance, the Priest Rapids population was identified as being susceptible to future warming conditions due to its relatively reduced growth when acclimated to 20°C and long migration route (17). In the present study, the Priest Rapids population exhibits both the greatest acclimation capacity (2.1°C) and also a significant loss in ΔCTM among fish acclimated to cold temperatures (11°C). This duality highlights the importance of assessing multiple thermal performance metrics, and additionally, the risks in predicting species performance without a thorough understanding of thermal physiology. Therefore, it is important to note that the Priest Rapids population come from the warmest habitat per the SSN model (R_MAX-His: 20.4 ± 0.5°C) and when acclimated to 20°C elicits an above average CT_MAX and the 2nd highest ΔCTM. Therefore, the species apparent plasticity when acclimated to warm temperatures may be more relevant to their performance in the wild than their poor performance when acclimated to cold temperatures. Furthermore, these results support ideas of local adaptation in regards to acute thermal tolerance where populations perform optimally under conditions similar to their local environment.

On the other hand, the Trinity hatchery population exhibited both the smallest response to acclimation temperature, the smallest acclimation capacity, and no significant heat hardening effects at any acclimation temperature. The lack of a heat hardening response as well as the overall limited response with acclimation temperature support the idea that the Trinity population possesses very little overall acclimation capacity, a trait that makes it vulnerable to future environmental warming. The modeled environmental traits of the Trinity population identify it as one of the coldest-rearing populations in our study, and therefore this population may lack the acclimatory response to warm temperatures exhibited by other populations. Still, a larger sample size may elucidate individuals with positive heat hardening responses, and further research linking thermal physiology to population genetics would provide insight into these individual and population differences.

Other work on Chinook salmon (New Zealand population) demonstrated that fish acclimated to 13°C were able to improve their CT_MAX by 0.6°C after exposure to a salinity stressor or air exposure relative to an unstressed control group (68). This improvement in CT_MAX is greater than the differences observed in our study, particularly among fish acclimated to 11°C which elicited a mean ΔCTM of −0.6 ± 1.4°C. It is interesting that the non-thermal stressors (a salinity challenge and air exposure) tested in Rodgers and Gomez Isaza (68), led to greater acute thermal tolerance than the thermal trials conducted presently. Variation in the size of CT_MAX increase among stressors [air, salinity or acute thermal exposure (present study)] may be due to the activation of a general protective stress response (69, 70) which may be somewhat mitigated in the present study due to physiological damage associated with a prior near-lethal acute thermal stress which may not be triggered by air exposure or salinity challenges. Given the effect of acclimation temperature as well as interpopulation variation, further study of differences in multi-stressor responses among salmonid populations would be informative for understanding population-specific responses to environmental change.

An important caveat to this work is the timing of CT_MAX trials. We exposed fish to a second CT_MAX ca. 24 h after their first. While this time frame has some ecological relevance (thermal extremes are typically consistent in their timing day to day), it may be that the heat-hardening response of juvenile Chinook salmon occurs more quickly. In other teleosts, the heat hardening effect was found to peak a few hours after the initial exposure (20, 71). Presumably, acclimation temperature may also impact the timeframe of a heat hardening response as fish acclimated to warm temperatures may be capable of physiological adjustment quicker than those acclimated to cold temperatures. Testing the speed at which a fish mounts a heat hardening response was beyond the scope of this study but would be an interesting avenue for future investigation. Additionally, fish used in this study had no prior experience with fluctuating temperatures whereas wild fish may be able to respond to transient periods of thermal stress and physiologically prepare for future incidents (69, 72). At the time of testing, the fish used in this study were of the smolt size class. Smolting in salmonids is a time of rapid physiological change as they prepare for migration to the ocean. In other salmonids, CT_MAX declined gradually as the fish aged (73), however the impact of smoltification on their heat hardening response is unstudied.

Finally, the fish used in this experiment were all of hatchery origin with free-swimming parents who may also have been of hatchery origin. Hatchery embryos are incubated in constant, cold temperatures with high oxygen saturation levels to maximize survival, and juveniles are reared in warmer temperatures with high food intake to maximize growth. These artificial, unnatural conditions can result in juveniles entering the river system with lower fitness, and the effect of hatchery domestication on salmonids has been demonstrated to modify populations and lower fitness over just a handful of generations (32, 74, 77). Hatchery selection may have altered thermal physiology as well, as supported by a study on zebra-fish (Danio rerio), where captive, laboratory-reared fish had reduced thermal plasticity when compared to wild rearing counterparts (27). Indeed, a parent's thermal history can significantly affect the physiological responses of their offspring (34, 35). However, the influence of hatchery domestication on thermal physiology of Pacific salmon is poorly understood and requires comparisons of natural- and hatchery-reared fish, which may be infeasible for populations that are primarily hatchery-origin. Regardless, while hatcheries may have unresolved phenotypic impacts on the thermal physiology of young salmon, our study finds that ancestral thermal history (i.e., historical, natural rearing temperatures) alongside current thermal exposure (i.e., acclimation temperatures) were associated with heat hardening responses and acclimation capacity, indicating that both ancestral and current thermal exposures shape thermal physiology.