Journal of Fish Biology (2015) 86, 1139–1152 doi:10.1111/jfb.12634, available online at wileyonlinelibrary.com
Thermal stress in Arctic charr Salvelinus alpinus broodstock: a 28 year case study H. Jeuthe*, E. Brännäs and J Nilsson Wildlife, Fish, and Environmental Studies, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Sweden (Received 30 June 2014, Accepted 20 December 2014) Temperature and egg viability data from an Arctic charr Salvelinus alpinus hatchery covering a period of 28 years were analysed. During the study period, there was a significant increase in the mean water temperature in May, July, August and September of c. 2∘ C. Independent of year, the egg viability showed a negative correlation with the mean monthly temperatures in July, August and September as well as with the temperature difference between October and November. The negative effect of high summer temperatures was further supported by a comparison of egg viability from replicate broodstock reared at two sites differing mainly in summer water temperature. The eggs from the colder site were, on average, significantly larger (4⋅4 mm compared with 4⋅0 mm) and had higher hatching rates (57% compared with 37%). These results suggest that unfavourable temperature conditions during the summer and autumn can explain much of the excessive egg mortality experienced at the main facility used for the Swedish S. alpinus breeding programme. The main effect was supra-optimal temperatures during the period July to September, but there also appears to have been an effect from the temperature regime before and during spawning (October to November) that was unrelated to the summer temperatures. These findings emphasize the importance of site selection and sustainable management of aquaculture hatcheries in the light of the ongoing climate change. © 2015 The Fisheries Society of the British Isles
Key words: aquaculture; climate change; egg size; egg viability; hatching rate; reproductive success.
INTRODUCTION Seasonal changes in temperature and day length are two of the most important abiotic factors influencing the timing and success of reproduction in fishes inhabiting temperate regions (Pankhurst & King, 2010). In addition, these factors become increasingly important with higher altitude and for more pronounced cold-water species (Pankhurst & Porter, 2003). In general, for autumn-spawning fishes, the shortening of the day is the abiotic factor that controls the production and maturation of eggs and sperm (Bromage et al., 2001) while a decrease in temperature triggers the onset of ovulation and spermiation (Gillet, 1991). There is an increasing focus and concern regarding the effects of climate change on fish populations (Elliott & Elliott, 2010; Pankhurst & Munday, 2011; Jeppesen et al., 2012). Because fishes are ectothermic and often have narrow ranges for temperature tolerance, their viability, especially as eggs and juveniles, and physiological performance are very sensitive to temperature changes in their *Author to whom correspondence should be addressed. Tel.: +46 907868571; email: [email protected]
1139 © 2015 The Fisheries Society of the British Isles
1140
H . J E U T H E E T A L.
environment (Pankhurst & King, 2010). In wild populations, fishes can, at least to some extent, avoid suboptimal environmental conditions through behavioural changes, e.g. moving to deeper and colder water. For farmed fishes, behavioural responses are very limited. In sufficiently deep-net pens, there may be a vertical temperature gradient, but in land-based aquaculture, the fishes are totally susceptible to the conditions determined by the facility and its operators. The effects of elevated temperature can be positive for on-growing and include higher growth and, hence, shorter production time from stocking-sized juveniles to slaughter-sized fishes (King et al., 2007). The temperature tolerance for sexual development and reproduction, however, is much lower than for growth and survival (Elliott & Elliott, 2010; Pankhurst & King, 2010). Therefore, the temperature requirements for an on-growing facility and a hatchery may be very different. In autumn-spawning salmonids, the major stages of gamete development, e.g. vitellogenesis, occur during the summer, making this period a potential bottleneck for reproductive success in the farming of cold-adapted species (Pankhurst & King, 2010). The Atlantic salmon Salmo salar L. 1758 is a well-studied example. The broodstock of farmed S. salar in Tasmania may be exposed to summer temperatures as high as 22∘ C, which results in reduced embryo survival compared with broodstock reared at lower temperatures, i.e. 14 and 18∘ C (King et al., 2003; Anderson et al., 2012). Elevated temperatures disturb gonadal steroidogenesis connected to the development of germ cells (Pankhurst & King, 2010). King et al. (2007) exposed female S. salar in Tasmania to elevated temperatures during different periods of vitellogenesis. They identified the summer month, 3 months prior to spawning, as the most sensitive period for elevated water temperature. The Arctic charr Salvelinus alpinus (L. 1758) is the northernmost distributed of all freshwater fishes and has the lowest warm-water tolerance among salmonid species (Elliott & Elliott, 2010). Studies on the effects of high temperature on S. alpinus broodstock and their reproductive success and egg quality parallel studies of S. salar in Tasmania. High summer temperatures affect embryonic survival by disturbing oocyte development and the timing of spawning and even result in the inhibition of ovulation (Gillet, 1991; Gillet & Breton, 1992; Jobling et al., 1995; Gillet et al., 2011). Salvelinus alpinus have been farmed in Sweden since the early 1900s. For many years, the main purpose was restocking into natural habitats but farming is now of increasing commercial interest (Eriksson et al., 2010). One of the major bottlenecks that have been identified for the future expansion of the industry is the reliable supply of eggs and juveniles (Jobling et al., 1998). In addition, the location of farms, hatcheries in particular, is restricted by access to sufficiently cold water because of the species’ cold-stenothermy. With ongoing climate change, access to cold water is expected to become even more limited with time. The effects of temperature on egg viability have mainly been studied using experimental designs where broodfish are held at different, and most often constant, temperature conditions, and differences in hormonal status, egg quality and embryonic survival are documented (Pankhurst & King, 2010). The relevance of results from laboratory experiments on suboptimal rearing conditions, when they are extrapolated to the farming industry, however, is sometimes questionable; the actual effects when applied to real-world situations are poorly documented. In this study, the effects of thermal stress on S. alpinus reproduction were assessed under routine conditions at a commercial hatchery. Egg survival and daily temperature data from a period of 28 years (1986–2013) were used to quantify the long-term temperature change in incoming
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1141
water and to evaluate the effect of annual temperature variation on egg viability. In addition, a comparative study of egg viability was performed using duplicate broodstock of equal family distribution, reared at different farming sites. MATERIALS AND METHODS H AT C H E RY Aquaculture Centre North (ACN) is situated in Kälarne, central Sweden (62∘ 59′ N; 16∘ 06′ E), and is a combined research station and commercial hatchery. It is a land-based, flow-through facility that takes water from a nearby lake. Until 2010, the water inlet was situated near the surface of the lake, which resulted in high water temperatures during warm summers. In April 2011, a deeper inlet was installed to provide cooler water during the summer. The temperature of the incoming water was measured daily throughout the study period. The S. alpinus broodstock was reared indoors in concrete tanks with depths of 1–3 m and volumes of 7⋅5–14 m2 . Rearing follows standard farming routines with automatic feeders, artificial lighting at natural day length (62∘ N) and a maximum rearing density of 25 kg m−3 . The hatchery is subjected to strict disease control routines, and there were no disease outbreaks during the study period. B RO O D S TO C K The ACN is the main facility for the Swedish S. alpinus breeding programme and the Arctic superior strain, which is selectively bred for the aquaculture industry. In addition to the selectively bred strain, S. alpinus from three natural populations in Lakes Ottsjön, Rensjön and Hornavan have been bred for restocking in the wild. The Arctic superior consists of descendants of the Lake Hornavan population that have been selectively bred over seven generations for growth, flesh pigmentation and age at sexual maturity (Nilsson et al., 2010). During the early parts of this study period, the broodstock consisted of fish from the natural (unselected) populations but were gradually replaced by the Arctic superior. The egg viability data were collected over a long time (28 years) and from an environment that was not manipulated to suit any research purpose. Conditions that could not be controlled, such as feed formulation, rearing techniques and stripping and fertilization procedures, may have changed during this time and affected the reproductive success of the broodstock. The age distribution of the broodstock has varied over the years. Often, several age classes were used in parallel and the same individuals were reused for up to five consecutive spawning seasons. In addition, age was not determined for first-generation individuals born in the wild. Therefore, age could not be included as a factor in the analyses covering the entire study period. A separate analysis, however, was performed for the parts of the dataset that included age information. It is known that female age has a significant positive effect on egg viability in S. alpinus (Jeuthe et al., 2013), and the uneven and sometimes unknown age distribution, together with the other factors mentioned previously, may have had confounding effects on the results. R E P RO D U C T I O N The broodstock was sorted into males and females in early September, and the sexes were then reared separately to avoid aggression and spontaneous spawning. The fish were examined for ovulation and spermiation once or twice per week. Stripping and fertilization were usually concentrated in late October and early November but could extend from the last week in September to the first week in December. The stripping procedure was generally the same throughout the study period and followed standard methods for dry fertilization (Shepherd & Bromage, 1988). The number of parents contributing to each fertilization batch, however, has varied. Most fertilizations were achieved using mixed gametes from, typically, three to four dams and one to three sires. These pooled egg batches were then incubated in separate hatching trays. When new generations were created for the selective breeding programme every 4–5 years, one male was used to fertilize eggs from two to three females separately, forming half-sibling families that
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1142
H . J E U T H E E T A L.
were incubated in separate hatching trays. Analyses of the individually fertilized egg batches from the breeding programme, including female age and egg size, have been published separately (Jeuthe et al., 2013). After fertilization and swelling, mean egg size and total volume in each hatching tray were estimated and used to calculate the number of eggs according to the method by Brofeldt (1935), and dead (white) eggs were removed and counted. Early in the incubation stage, the eggs were treated with malachite green (3–5 mg l−1 ) or, after the year 2000, with formaldehyde (1 ml l−1 25% formaldehyde for 30 min) twice per week to prevent fungal colonization. When the eggs reached the eyed stage, they were mechanically shocked to induce coagulation in the dead and unfertilized eggs, which were then removed and counted. Egg viability was defined, in this study, as the proportion of eggs that successfully reached the eyed stage. Eggs were sometimes sold to other hatcheries at this stage, which limited the survival data to hatching. Additionally, most mortality occurred before the eyed stage. Egg viability data were collected annually from 1986 to 2013 and are highly variable with respect to the genetic variability and number of fertilizations. There are no data from 1998 and 2002 as most of the broodstock died due to high summer temperatures. Percentages were calculated using the total number of eggs stripped and the number of eggs that successfully reached the eyed stage.
S I T E C O M PA R I S O N In the autumn of 2004, 46 families were created to form a new generation for the breeding programme. One year later, when the fish had reached a mean mass of 40 g, they were freeze-branded for family identification, and the families were equally distributed into two groups. One group remained at ACN in Kälarne, and the other group was transported to a net-pen farm in Slussfors in northern Sweden (Fig. 1). The same type of feed was used at both sites (Skretting, 3–6 mm in size; www.skretting.com). In mid-September of 2008, the fish were sexually mature, and the broodstock was transported from Slussfors to a hatchery in Vilhelmina, located 100 km to the south (Fig. 1), where they were sorted by sex and kept in circular tanks (3 m in depth and 5 m in diameter). Stripping and fertilization were performed on two occasions in Vilhelmina, 9 and 15 October, at water temperatures of 8⋅2 and 5⋅1∘ C, respectively, using protocols and personnel from ACN. The broodstock housed at ACN was stripped on 11 occasions from 17 October to 11 November, at temperatures ranging from 7⋅2 to 3⋅7∘ C. After the fertilization process, which included size and count estimates, was completed in Vilhelmina, the eggs were transported to ACN on the same day and incubated alongside the eggs fertilized there. All of the offspring used in the site comparison study were kept at ACN post-hatching, which enabled the use of survival to hatching as the output variable rather than the eyed stage variable used in the long-term study. Between the two rearing sites, the two main differences in environmental factors affecting reproduction are temperature and light conditions. Summer water temperatures were generally lower at the Slussfors site, which is located in a river regulated by hydroelectric power stations. At ACN, the broodstock was kept indoors under artificial lighting while the Slussfors broodstock was outside in natural light (65∘ N). The ACN facility is supported by water from a lake and the Vilhelmina facility is supported with water from a small river, which is why the autumn temperature decreases faster in Vilhelmina (Fig. 2). In addition, both the broodstock and eggs of the Slussfors and Vilhelmina group were transported while the ACN group was stationary in Kälarne. There is a chance that stress due to transportation had a negative effect on egg viability. As it is hypothesized that the egg viability is higher at the colder Slussfors and Vilhelmina site, the potential effect of transport induced stress can only increase the risk of a type II error and is therefore only a problem if no difference between the sites is found. D ATA A N A LY S I S Temperature trends All statistical analyses were performed using Minitab statistical software 16 (www.minitab. com). Linear regression was used to determine if the monthly temperatures of the incoming water at ACN changed from 1986 to 2010 and correlations between monthly temperatures
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1143
Polar circle
Slusstors Vilhelmina
Kälarne
Second farm site and hatchery
Main hatchery
60° N N
17° E
Fig. 1. Location of farming sites involved in the study: Aquaculture Centre North (ACN) in Kälarne in central Sweden is the main facility for the Swedish Salvelinus alpinus breeding programme. In the site comparison study, sibling replicates of the selected S. alpinus strain were reared in Slussfors (net pens) and Vilhelmina (hatchery).
were estimated using Pearson’s correlation. The P-values attained were corrected for the false discovery rate (FDR) resulting from multiple comparisons using the method by Benjamini & Hochberg (1995) and are presented as q-values. Temperature data from 2011 to 2013 were not included in this analysis because of the new inlet installed in 2011. The data from 2011 to 2013 were only used to analyse the relationship between mean monthly temperature and egg viability.
Egg viability The relationship between monthly mean temperatures and egg survival was analysed using Pearson’s correlation; P-values were then corrected for FDR. The percentage data showed normal distribution (probability plot, P > 0⋅05). Only the egg viability data from the Arctic superior strain from 1994 to 2013 (excluding 1997, 1998 and 2002) contained reliable information on brood fish age. Using this sub-sample, a multiple linear regression model
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1144
H . J E U T H E E T A L.
18 16 Mean temperature (° C)
14 12 10 8 6 4 2
Ja
nu Fe ary br ua ry M ar ch A pr il M ay Ju ne Ju l A y ug Se us pt em t b O er ct o b N ov er em D be ec r em be r
0
Month Fig. 2. Water temperatures (based on monthly means) at the farming sites in Kälarne ( ) and Slussfors and Vilhelmina ( ) during 2008. The Slussfors broodstock was moved to the hatchery (Vilhelmina) in September ( ), temperatures presented before this point in time are from Slussfors and later temperatures are from Vilhelmina.
was constructed to evaluate the importance of the different temperature variables combined with female age in predicting egg viability. The 7 and 8 year-old fish were excluded from the analysis because previous research suggests that there is no improvement in egg viability past the age of c. 6 years (Jeuthe et al., 2013). The model was constructed by first including all mean monthly temperatures, broodstock generation and age as independent variables and total annual age-specific egg survival to the eyed stage (%) as the output variable. The egg survival data were normally distributed (probability plot, P > 0⋅05). Non-contributing variables (P > 0⋅05) were then removed one by one until the model contained only statistically significant variables. Egg viability was positively correlated with October temperature (P < 0⋅05) and showed a tendency of negative association with November temperature (P = 0⋅05). This was interpreted as a positive correlation with the distinct autumn cooling, and the two variables were replaced with the difference in mean temperature between October and November. A similar model was then constructed with data from all years, with pooled results for the different populations and without taking age into consideration. The populations were tested for differences in egg survival, but none were found [analysis of variance (ANOVA), P > 0⋅05], and therefore the data were pooled in the model. Linear regression was used to test whether there was any change in annual egg viability over the study period. Models were constructed with both age as a variable (1994–2013) and without age (1986–2013).
Site comparison The eggs from the two replicate broodstocks were tested for differences in viability and size using two sample t-tests. The analyses were limited to broodstock families that were successfully stripped at both localities, which gave equal family distribution in both groups. Viability was recorded as the overall survival rate of individual egg batches; zero-success fertilizations were excluded. The analysis was performed using arc-sine square-root transformed proportion data to meet the normal distribution requirement (P > 0⋅05). The group means presented, however, were calculated using untransformed data.
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1145
Table I. Mean monthly temperature (maximum and minimum of daily temperatures) between 1986 and 2010 at the main hatchery for the Salvelinus alpinus breeding programme in Sweden (Kälarne, see Fig. 1). Linear regression was used to estimate t-, P- and R2 -values for the change in mean monthly temperatures over time. Significant P-values were corrected for multiple comparison (q) using the method described by Benjamini & Hochberg (1995). The significant relationships are plotted in Fig. 2 Month January February March April May June July August September October November December
Temperature (∘ C) mean (minimum–maximum)
t-value
P-value (q)
R2
2⋅0 (0⋅7–3⋅2) 2⋅2 (0⋅2–3⋅2) 2⋅2 (0⋅8–3⋅8) 2⋅7 (1⋅2–4⋅7) 5⋅5 (1⋅9–11⋅8) 11⋅5 (4⋅7–17⋅8) 15⋅5 (10⋅4–20⋅5) 16⋅0 (11⋅5–20⋅8) 12⋅4 (6⋅8–18⋅4) 7⋅4 (2⋅4–12⋅9) 3⋅0 (0⋅8–6⋅8) 2⋅0 (0⋅5–3⋅4)
−0⋅589 −0⋅597 −0⋅695 0⋅737 2⋅657 1⋅924 4⋅085 3⋅192 3⋅074 1⋅609 0⋅975 −0⋅046
>0⋅05 >0⋅05 >0⋅05 >0⋅05 0⋅05 0⋅05
0⋅015 0⋅015 0⋅021 0⋅023 0⋅235 0⋅139 0⋅420 0⋅307 0⋅291 0⋅101 0⋅04 0⋅002
RESULTS From 1986 to 2010, the mean monthly temperature of the incoming water at ACN increased significantly in May, July, August and September (linear regression, q = 0⋅042, 0⋅0012, 0⋅02 and 0⋅02, respectively; Table I). There was no significant change in temperature for the other months during the same time period. The temperature increase during the 25 year period was nearly 2∘ C in May, July, August and September [Fig. 3(a)]. The highest temperatures were noted in 2003 with daily peaks in July and August reaching as high as 20⋅1 and 20⋅2∘ C, respectively, as well as 20⋅8∘ C in August 1997 and 20⋅5∘ C in July 1999. The number of days with temperatures exceeding 15∘ C showed an increasing trend during the study period [Fig. 3(b)]. Because the new water inlet was installed in 2011, the mean monthly temperatures have been below 14∘ C and the maximum temperature below 15∘ C (2011–2013, Fig. 3). High water temperatures in July were followed by high temperatures in August (P < 0⋅001, r = 0⋅740) and September (P < 0⋅01, r = 0⋅563) but not in October (P > 0⋅05) or November (P > 0⋅05). No temporal trend in egg viability was evident over the study period, either with or without age as a factor in the model (linear regression, t = 1⋅21, P > 0⋅05 and t = −1⋅49, P > 0⋅05, respectively, Fig. 4). Significant negative correlations were found between egg survival and the temperature in July, August and September (Table II), and the strongest correlation was found for September (Fig. 5). A multiple-regression model based on the parts of the dataset that included age information, i.e. the Arctic superior data from 1994 to 2013, showed that 77% of the interannual variation in egg viability could be explained by female age, mean temperature in August and September and the temperature difference between October and November (Table III, model I). The equivalent model for the entire dataset, with no regard to
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1146
H . J E U T H E E T A L.
(a) 20 Mean temperature (° C)
18 16 14 12 10 8 6 4 2
(b) 90 80 70
Days
60 50 40 30 20 10 0 1985
1990
1995
2000
2005
2010
2015
Year Fig. 3. The temperature conditions in a Salvelinus alpinus hatchery located in central Sweden (see Fig. 1) from 1986 to 2013. , the installation of a new deeper water inlet providing colder water to the facility during summer. (a) The mean monthly temperatures ( , , , ) from 1986 to 2013 and linear regressions ( ) showing the general temperature change from 1986 to 2010 for the 4 months with significant increase: May ( ), September ( ), July ( ) and August ( ). The curves were fitted by: May y = 0⋅07212x − 138⋅6, July y = 0⋅1008x − 185⋅9, August y = 0⋅1038x − 191⋅3 and September y = 0⋅07943x − 146⋅3. (b) The number of days with water temperatures exceeding 15∘ C from 1986 to 2013.
age, resulted in the loss of significance for the August temperature (Table III, model II). Egg viability in different age groups of the Arctic superior strain can be seen in Fig. 6. The rearing of the sibling broodstock at different sites from age 0+ years had a pronounced effect on egg viability and egg size. During the summer before the fish were stripped (at age 3+ years), the water temperature never exceeded 15∘ C at the Slussfors site but reached 19⋅1∘ C in late July at the Kälarne hatchery site (Fig. 2). The mean of 57% of the viable eggs from the broodstock kept at the Slussfors site was significantly higher compared with the mean of 37% at the Kälarne hatchery site (t-test, t = 4⋅94, d.f. = 122, P < 0⋅001). The eggs from the broodstock in Slussfors were also significantly larger with a mean diameter of 4⋅4 mm compared with 4⋅0 mm from the broodstock kept at Kälarne (t-test, t = −11⋅46, d.f. = 122, P 0⋅05 >0⋅05 >0⋅05 >0⋅05 >0⋅05 >0⋅05 0⋅05
0⋅290 0⋅283 0⋅265 0⋅221 0⋅148 0⋅164 0⋅486 0⋅453 0⋅522 0⋅000 0⋅158 0⋅170
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1148
H . J E U T H E E T A L.
Frequency of eyed eggs (%)
70 60 50 40 30 20 10 0 9
10
11
12
13
14
15
Mean temperature in September (° C) Fig. 5. Annual egg survival (% eyed eggs) in relation to mean temperature in September of the same year. Significant correlations were also found for May, July and August (Table II).
late April, which results in only small temperature variations during this time of the year. The lack of a clear temperature increase in June could be explained by annual variations in the timing and extent of the spring mixing of the lake, which determines when a stable warm surface layer can be established. With the surface ice melting in April, a warmer surface layer most likely starts to form in May but is lost again in June due to mixing with the cold, deep water. This would explain why the warming trend is evident in May but not in June. Table III. Multiple-regression models predicting annual egg viability of temperature-stressed Salvelinus alpinus broodstock. Model I is based on the selected strain, Arctic superior, from 1994 to 2013 and includes information on broodstock age, R2 = 0⋅77. Model II is based on the entire dataset from 1986 to 2013 with pooled populations and no regard for age, R2 = 0⋅40 Model
Predictor
I
Constant Age August temperature September temperature Temperature difference from October to November Constant August temperature September temperature Temperature difference from October to November
II
Coefficient
s.e. of coefficient
t-Value
P-value
187⋅5 6⋅29 −4⋅86 −10⋅88 7⋅06
30⋅2 1⋅87 1⋅46 2⋅86 2⋅79
6⋅21 3⋅37 −3⋅32 −3⋅80 2⋅53
Thermal stress in Arctic charr Salvelinus alpinus broodstock: a 28 year case study H. Jeuthe*, E. Brännäs and J Nilsson Wildlife, Fish, and Environmental Studies, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Sweden (Received 30 June 2014, Accepted 20 December 2014) Temperature and egg viability data from an Arctic charr Salvelinus alpinus hatchery covering a period of 28 years were analysed. During the study period, there was a significant increase in the mean water temperature in May, July, August and September of c. 2∘ C. Independent of year, the egg viability showed a negative correlation with the mean monthly temperatures in July, August and September as well as with the temperature difference between October and November. The negative effect of high summer temperatures was further supported by a comparison of egg viability from replicate broodstock reared at two sites differing mainly in summer water temperature. The eggs from the colder site were, on average, significantly larger (4⋅4 mm compared with 4⋅0 mm) and had higher hatching rates (57% compared with 37%). These results suggest that unfavourable temperature conditions during the summer and autumn can explain much of the excessive egg mortality experienced at the main facility used for the Swedish S. alpinus breeding programme. The main effect was supra-optimal temperatures during the period July to September, but there also appears to have been an effect from the temperature regime before and during spawning (October to November) that was unrelated to the summer temperatures. These findings emphasize the importance of site selection and sustainable management of aquaculture hatcheries in the light of the ongoing climate change. © 2015 The Fisheries Society of the British Isles
Key words: aquaculture; climate change; egg size; egg viability; hatching rate; reproductive success.
INTRODUCTION Seasonal changes in temperature and day length are two of the most important abiotic factors influencing the timing and success of reproduction in fishes inhabiting temperate regions (Pankhurst & King, 2010). In addition, these factors become increasingly important with higher altitude and for more pronounced cold-water species (Pankhurst & Porter, 2003). In general, for autumn-spawning fishes, the shortening of the day is the abiotic factor that controls the production and maturation of eggs and sperm (Bromage et al., 2001) while a decrease in temperature triggers the onset of ovulation and spermiation (Gillet, 1991). There is an increasing focus and concern regarding the effects of climate change on fish populations (Elliott & Elliott, 2010; Pankhurst & Munday, 2011; Jeppesen et al., 2012). Because fishes are ectothermic and often have narrow ranges for temperature tolerance, their viability, especially as eggs and juveniles, and physiological performance are very sensitive to temperature changes in their *Author to whom correspondence should be addressed. Tel.: +46 907868571; email: [email protected]
1139 © 2015 The Fisheries Society of the British Isles
1140
H . J E U T H E E T A L.
environment (Pankhurst & King, 2010). In wild populations, fishes can, at least to some extent, avoid suboptimal environmental conditions through behavioural changes, e.g. moving to deeper and colder water. For farmed fishes, behavioural responses are very limited. In sufficiently deep-net pens, there may be a vertical temperature gradient, but in land-based aquaculture, the fishes are totally susceptible to the conditions determined by the facility and its operators. The effects of elevated temperature can be positive for on-growing and include higher growth and, hence, shorter production time from stocking-sized juveniles to slaughter-sized fishes (King et al., 2007). The temperature tolerance for sexual development and reproduction, however, is much lower than for growth and survival (Elliott & Elliott, 2010; Pankhurst & King, 2010). Therefore, the temperature requirements for an on-growing facility and a hatchery may be very different. In autumn-spawning salmonids, the major stages of gamete development, e.g. vitellogenesis, occur during the summer, making this period a potential bottleneck for reproductive success in the farming of cold-adapted species (Pankhurst & King, 2010). The Atlantic salmon Salmo salar L. 1758 is a well-studied example. The broodstock of farmed S. salar in Tasmania may be exposed to summer temperatures as high as 22∘ C, which results in reduced embryo survival compared with broodstock reared at lower temperatures, i.e. 14 and 18∘ C (King et al., 2003; Anderson et al., 2012). Elevated temperatures disturb gonadal steroidogenesis connected to the development of germ cells (Pankhurst & King, 2010). King et al. (2007) exposed female S. salar in Tasmania to elevated temperatures during different periods of vitellogenesis. They identified the summer month, 3 months prior to spawning, as the most sensitive period for elevated water temperature. The Arctic charr Salvelinus alpinus (L. 1758) is the northernmost distributed of all freshwater fishes and has the lowest warm-water tolerance among salmonid species (Elliott & Elliott, 2010). Studies on the effects of high temperature on S. alpinus broodstock and their reproductive success and egg quality parallel studies of S. salar in Tasmania. High summer temperatures affect embryonic survival by disturbing oocyte development and the timing of spawning and even result in the inhibition of ovulation (Gillet, 1991; Gillet & Breton, 1992; Jobling et al., 1995; Gillet et al., 2011). Salvelinus alpinus have been farmed in Sweden since the early 1900s. For many years, the main purpose was restocking into natural habitats but farming is now of increasing commercial interest (Eriksson et al., 2010). One of the major bottlenecks that have been identified for the future expansion of the industry is the reliable supply of eggs and juveniles (Jobling et al., 1998). In addition, the location of farms, hatcheries in particular, is restricted by access to sufficiently cold water because of the species’ cold-stenothermy. With ongoing climate change, access to cold water is expected to become even more limited with time. The effects of temperature on egg viability have mainly been studied using experimental designs where broodfish are held at different, and most often constant, temperature conditions, and differences in hormonal status, egg quality and embryonic survival are documented (Pankhurst & King, 2010). The relevance of results from laboratory experiments on suboptimal rearing conditions, when they are extrapolated to the farming industry, however, is sometimes questionable; the actual effects when applied to real-world situations are poorly documented. In this study, the effects of thermal stress on S. alpinus reproduction were assessed under routine conditions at a commercial hatchery. Egg survival and daily temperature data from a period of 28 years (1986–2013) were used to quantify the long-term temperature change in incoming
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1141
water and to evaluate the effect of annual temperature variation on egg viability. In addition, a comparative study of egg viability was performed using duplicate broodstock of equal family distribution, reared at different farming sites. MATERIALS AND METHODS H AT C H E RY Aquaculture Centre North (ACN) is situated in Kälarne, central Sweden (62∘ 59′ N; 16∘ 06′ E), and is a combined research station and commercial hatchery. It is a land-based, flow-through facility that takes water from a nearby lake. Until 2010, the water inlet was situated near the surface of the lake, which resulted in high water temperatures during warm summers. In April 2011, a deeper inlet was installed to provide cooler water during the summer. The temperature of the incoming water was measured daily throughout the study period. The S. alpinus broodstock was reared indoors in concrete tanks with depths of 1–3 m and volumes of 7⋅5–14 m2 . Rearing follows standard farming routines with automatic feeders, artificial lighting at natural day length (62∘ N) and a maximum rearing density of 25 kg m−3 . The hatchery is subjected to strict disease control routines, and there were no disease outbreaks during the study period. B RO O D S TO C K The ACN is the main facility for the Swedish S. alpinus breeding programme and the Arctic superior strain, which is selectively bred for the aquaculture industry. In addition to the selectively bred strain, S. alpinus from three natural populations in Lakes Ottsjön, Rensjön and Hornavan have been bred for restocking in the wild. The Arctic superior consists of descendants of the Lake Hornavan population that have been selectively bred over seven generations for growth, flesh pigmentation and age at sexual maturity (Nilsson et al., 2010). During the early parts of this study period, the broodstock consisted of fish from the natural (unselected) populations but were gradually replaced by the Arctic superior. The egg viability data were collected over a long time (28 years) and from an environment that was not manipulated to suit any research purpose. Conditions that could not be controlled, such as feed formulation, rearing techniques and stripping and fertilization procedures, may have changed during this time and affected the reproductive success of the broodstock. The age distribution of the broodstock has varied over the years. Often, several age classes were used in parallel and the same individuals were reused for up to five consecutive spawning seasons. In addition, age was not determined for first-generation individuals born in the wild. Therefore, age could not be included as a factor in the analyses covering the entire study period. A separate analysis, however, was performed for the parts of the dataset that included age information. It is known that female age has a significant positive effect on egg viability in S. alpinus (Jeuthe et al., 2013), and the uneven and sometimes unknown age distribution, together with the other factors mentioned previously, may have had confounding effects on the results. R E P RO D U C T I O N The broodstock was sorted into males and females in early September, and the sexes were then reared separately to avoid aggression and spontaneous spawning. The fish were examined for ovulation and spermiation once or twice per week. Stripping and fertilization were usually concentrated in late October and early November but could extend from the last week in September to the first week in December. The stripping procedure was generally the same throughout the study period and followed standard methods for dry fertilization (Shepherd & Bromage, 1988). The number of parents contributing to each fertilization batch, however, has varied. Most fertilizations were achieved using mixed gametes from, typically, three to four dams and one to three sires. These pooled egg batches were then incubated in separate hatching trays. When new generations were created for the selective breeding programme every 4–5 years, one male was used to fertilize eggs from two to three females separately, forming half-sibling families that
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1142
H . J E U T H E E T A L.
were incubated in separate hatching trays. Analyses of the individually fertilized egg batches from the breeding programme, including female age and egg size, have been published separately (Jeuthe et al., 2013). After fertilization and swelling, mean egg size and total volume in each hatching tray were estimated and used to calculate the number of eggs according to the method by Brofeldt (1935), and dead (white) eggs were removed and counted. Early in the incubation stage, the eggs were treated with malachite green (3–5 mg l−1 ) or, after the year 2000, with formaldehyde (1 ml l−1 25% formaldehyde for 30 min) twice per week to prevent fungal colonization. When the eggs reached the eyed stage, they were mechanically shocked to induce coagulation in the dead and unfertilized eggs, which were then removed and counted. Egg viability was defined, in this study, as the proportion of eggs that successfully reached the eyed stage. Eggs were sometimes sold to other hatcheries at this stage, which limited the survival data to hatching. Additionally, most mortality occurred before the eyed stage. Egg viability data were collected annually from 1986 to 2013 and are highly variable with respect to the genetic variability and number of fertilizations. There are no data from 1998 and 2002 as most of the broodstock died due to high summer temperatures. Percentages were calculated using the total number of eggs stripped and the number of eggs that successfully reached the eyed stage.
S I T E C O M PA R I S O N In the autumn of 2004, 46 families were created to form a new generation for the breeding programme. One year later, when the fish had reached a mean mass of 40 g, they were freeze-branded for family identification, and the families were equally distributed into two groups. One group remained at ACN in Kälarne, and the other group was transported to a net-pen farm in Slussfors in northern Sweden (Fig. 1). The same type of feed was used at both sites (Skretting, 3–6 mm in size; www.skretting.com). In mid-September of 2008, the fish were sexually mature, and the broodstock was transported from Slussfors to a hatchery in Vilhelmina, located 100 km to the south (Fig. 1), where they were sorted by sex and kept in circular tanks (3 m in depth and 5 m in diameter). Stripping and fertilization were performed on two occasions in Vilhelmina, 9 and 15 October, at water temperatures of 8⋅2 and 5⋅1∘ C, respectively, using protocols and personnel from ACN. The broodstock housed at ACN was stripped on 11 occasions from 17 October to 11 November, at temperatures ranging from 7⋅2 to 3⋅7∘ C. After the fertilization process, which included size and count estimates, was completed in Vilhelmina, the eggs were transported to ACN on the same day and incubated alongside the eggs fertilized there. All of the offspring used in the site comparison study were kept at ACN post-hatching, which enabled the use of survival to hatching as the output variable rather than the eyed stage variable used in the long-term study. Between the two rearing sites, the two main differences in environmental factors affecting reproduction are temperature and light conditions. Summer water temperatures were generally lower at the Slussfors site, which is located in a river regulated by hydroelectric power stations. At ACN, the broodstock was kept indoors under artificial lighting while the Slussfors broodstock was outside in natural light (65∘ N). The ACN facility is supported by water from a lake and the Vilhelmina facility is supported with water from a small river, which is why the autumn temperature decreases faster in Vilhelmina (Fig. 2). In addition, both the broodstock and eggs of the Slussfors and Vilhelmina group were transported while the ACN group was stationary in Kälarne. There is a chance that stress due to transportation had a negative effect on egg viability. As it is hypothesized that the egg viability is higher at the colder Slussfors and Vilhelmina site, the potential effect of transport induced stress can only increase the risk of a type II error and is therefore only a problem if no difference between the sites is found. D ATA A N A LY S I S Temperature trends All statistical analyses were performed using Minitab statistical software 16 (www.minitab. com). Linear regression was used to determine if the monthly temperatures of the incoming water at ACN changed from 1986 to 2010 and correlations between monthly temperatures
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1143
Polar circle
Slusstors Vilhelmina
Kälarne
Second farm site and hatchery
Main hatchery
60° N N
17° E
Fig. 1. Location of farming sites involved in the study: Aquaculture Centre North (ACN) in Kälarne in central Sweden is the main facility for the Swedish Salvelinus alpinus breeding programme. In the site comparison study, sibling replicates of the selected S. alpinus strain were reared in Slussfors (net pens) and Vilhelmina (hatchery).
were estimated using Pearson’s correlation. The P-values attained were corrected for the false discovery rate (FDR) resulting from multiple comparisons using the method by Benjamini & Hochberg (1995) and are presented as q-values. Temperature data from 2011 to 2013 were not included in this analysis because of the new inlet installed in 2011. The data from 2011 to 2013 were only used to analyse the relationship between mean monthly temperature and egg viability.
Egg viability The relationship between monthly mean temperatures and egg survival was analysed using Pearson’s correlation; P-values were then corrected for FDR. The percentage data showed normal distribution (probability plot, P > 0⋅05). Only the egg viability data from the Arctic superior strain from 1994 to 2013 (excluding 1997, 1998 and 2002) contained reliable information on brood fish age. Using this sub-sample, a multiple linear regression model
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1144
H . J E U T H E E T A L.
18 16 Mean temperature (° C)
14 12 10 8 6 4 2
Ja
nu Fe ary br ua ry M ar ch A pr il M ay Ju ne Ju l A y ug Se us pt em t b O er ct o b N ov er em D be ec r em be r
0
Month Fig. 2. Water temperatures (based on monthly means) at the farming sites in Kälarne ( ) and Slussfors and Vilhelmina ( ) during 2008. The Slussfors broodstock was moved to the hatchery (Vilhelmina) in September ( ), temperatures presented before this point in time are from Slussfors and later temperatures are from Vilhelmina.
was constructed to evaluate the importance of the different temperature variables combined with female age in predicting egg viability. The 7 and 8 year-old fish were excluded from the analysis because previous research suggests that there is no improvement in egg viability past the age of c. 6 years (Jeuthe et al., 2013). The model was constructed by first including all mean monthly temperatures, broodstock generation and age as independent variables and total annual age-specific egg survival to the eyed stage (%) as the output variable. The egg survival data were normally distributed (probability plot, P > 0⋅05). Non-contributing variables (P > 0⋅05) were then removed one by one until the model contained only statistically significant variables. Egg viability was positively correlated with October temperature (P < 0⋅05) and showed a tendency of negative association with November temperature (P = 0⋅05). This was interpreted as a positive correlation with the distinct autumn cooling, and the two variables were replaced with the difference in mean temperature between October and November. A similar model was then constructed with data from all years, with pooled results for the different populations and without taking age into consideration. The populations were tested for differences in egg survival, but none were found [analysis of variance (ANOVA), P > 0⋅05], and therefore the data were pooled in the model. Linear regression was used to test whether there was any change in annual egg viability over the study period. Models were constructed with both age as a variable (1994–2013) and without age (1986–2013).
Site comparison The eggs from the two replicate broodstocks were tested for differences in viability and size using two sample t-tests. The analyses were limited to broodstock families that were successfully stripped at both localities, which gave equal family distribution in both groups. Viability was recorded as the overall survival rate of individual egg batches; zero-success fertilizations were excluded. The analysis was performed using arc-sine square-root transformed proportion data to meet the normal distribution requirement (P > 0⋅05). The group means presented, however, were calculated using untransformed data.
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
T H E R M A L S T R E S S I N S A LV E L I N U S A L P I N U S B R O O D S T O C K
1145
Table I. Mean monthly temperature (maximum and minimum of daily temperatures) between 1986 and 2010 at the main hatchery for the Salvelinus alpinus breeding programme in Sweden (Kälarne, see Fig. 1). Linear regression was used to estimate t-, P- and R2 -values for the change in mean monthly temperatures over time. Significant P-values were corrected for multiple comparison (q) using the method described by Benjamini & Hochberg (1995). The significant relationships are plotted in Fig. 2 Month January February March April May June July August September October November December
Temperature (∘ C) mean (minimum–maximum)
t-value
P-value (q)
R2
2⋅0 (0⋅7–3⋅2) 2⋅2 (0⋅2–3⋅2) 2⋅2 (0⋅8–3⋅8) 2⋅7 (1⋅2–4⋅7) 5⋅5 (1⋅9–11⋅8) 11⋅5 (4⋅7–17⋅8) 15⋅5 (10⋅4–20⋅5) 16⋅0 (11⋅5–20⋅8) 12⋅4 (6⋅8–18⋅4) 7⋅4 (2⋅4–12⋅9) 3⋅0 (0⋅8–6⋅8) 2⋅0 (0⋅5–3⋅4)
−0⋅589 −0⋅597 −0⋅695 0⋅737 2⋅657 1⋅924 4⋅085 3⋅192 3⋅074 1⋅609 0⋅975 −0⋅046
>0⋅05 >0⋅05 >0⋅05 >0⋅05 0⋅05 0⋅05
0⋅015 0⋅015 0⋅021 0⋅023 0⋅235 0⋅139 0⋅420 0⋅307 0⋅291 0⋅101 0⋅04 0⋅002
RESULTS From 1986 to 2010, the mean monthly temperature of the incoming water at ACN increased significantly in May, July, August and September (linear regression, q = 0⋅042, 0⋅0012, 0⋅02 and 0⋅02, respectively; Table I). There was no significant change in temperature for the other months during the same time period. The temperature increase during the 25 year period was nearly 2∘ C in May, July, August and September [Fig. 3(a)]. The highest temperatures were noted in 2003 with daily peaks in July and August reaching as high as 20⋅1 and 20⋅2∘ C, respectively, as well as 20⋅8∘ C in August 1997 and 20⋅5∘ C in July 1999. The number of days with temperatures exceeding 15∘ C showed an increasing trend during the study period [Fig. 3(b)]. Because the new water inlet was installed in 2011, the mean monthly temperatures have been below 14∘ C and the maximum temperature below 15∘ C (2011–2013, Fig. 3). High water temperatures in July were followed by high temperatures in August (P < 0⋅001, r = 0⋅740) and September (P < 0⋅01, r = 0⋅563) but not in October (P > 0⋅05) or November (P > 0⋅05). No temporal trend in egg viability was evident over the study period, either with or without age as a factor in the model (linear regression, t = 1⋅21, P > 0⋅05 and t = −1⋅49, P > 0⋅05, respectively, Fig. 4). Significant negative correlations were found between egg survival and the temperature in July, August and September (Table II), and the strongest correlation was found for September (Fig. 5). A multiple-regression model based on the parts of the dataset that included age information, i.e. the Arctic superior data from 1994 to 2013, showed that 77% of the interannual variation in egg viability could be explained by female age, mean temperature in August and September and the temperature difference between October and November (Table III, model I). The equivalent model for the entire dataset, with no regard to
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1146
H . J E U T H E E T A L.
(a) 20 Mean temperature (° C)
18 16 14 12 10 8 6 4 2
(b) 90 80 70
Days
60 50 40 30 20 10 0 1985
1990
1995
2000
2005
2010
2015
Year Fig. 3. The temperature conditions in a Salvelinus alpinus hatchery located in central Sweden (see Fig. 1) from 1986 to 2013. , the installation of a new deeper water inlet providing colder water to the facility during summer. (a) The mean monthly temperatures ( , , , ) from 1986 to 2013 and linear regressions ( ) showing the general temperature change from 1986 to 2010 for the 4 months with significant increase: May ( ), September ( ), July ( ) and August ( ). The curves were fitted by: May y = 0⋅07212x − 138⋅6, July y = 0⋅1008x − 185⋅9, August y = 0⋅1038x − 191⋅3 and September y = 0⋅07943x − 146⋅3. (b) The number of days with water temperatures exceeding 15∘ C from 1986 to 2013.
age, resulted in the loss of significance for the August temperature (Table III, model II). Egg viability in different age groups of the Arctic superior strain can be seen in Fig. 6. The rearing of the sibling broodstock at different sites from age 0+ years had a pronounced effect on egg viability and egg size. During the summer before the fish were stripped (at age 3+ years), the water temperature never exceeded 15∘ C at the Slussfors site but reached 19⋅1∘ C in late July at the Kälarne hatchery site (Fig. 2). The mean of 57% of the viable eggs from the broodstock kept at the Slussfors site was significantly higher compared with the mean of 37% at the Kälarne hatchery site (t-test, t = 4⋅94, d.f. = 122, P < 0⋅001). The eggs from the broodstock in Slussfors were also significantly larger with a mean diameter of 4⋅4 mm compared with 4⋅0 mm from the broodstock kept at Kälarne (t-test, t = −11⋅46, d.f. = 122, P 0⋅05 >0⋅05 >0⋅05 >0⋅05 >0⋅05 >0⋅05 0⋅05
0⋅290 0⋅283 0⋅265 0⋅221 0⋅148 0⋅164 0⋅486 0⋅453 0⋅522 0⋅000 0⋅158 0⋅170
© 2015 The Fisheries Society of the British Isles, Journal of Fish Biology 2015, 86, 1139–1152
1148
H . J E U T H E E T A L.
Frequency of eyed eggs (%)
70 60 50 40 30 20 10 0 9
10
11
12
13
14
15
Mean temperature in September (° C) Fig. 5. Annual egg survival (% eyed eggs) in relation to mean temperature in September of the same year. Significant correlations were also found for May, July and August (Table II).
late April, which results in only small temperature variations during this time of the year. The lack of a clear temperature increase in June could be explained by annual variations in the timing and extent of the spring mixing of the lake, which determines when a stable warm surface layer can be established. With the surface ice melting in April, a warmer surface layer most likely starts to form in May but is lost again in June due to mixing with the cold, deep water. This would explain why the warming trend is evident in May but not in June. Table III. Multiple-regression models predicting annual egg viability of temperature-stressed Salvelinus alpinus broodstock. Model I is based on the selected strain, Arctic superior, from 1994 to 2013 and includes information on broodstock age, R2 = 0⋅77. Model II is based on the entire dataset from 1986 to 2013 with pooled populations and no regard for age, R2 = 0⋅40 Model
Predictor
I
Constant Age August temperature September temperature Temperature difference from October to November Constant August temperature September temperature Temperature difference from October to November
II
Coefficient
s.e. of coefficient
t-Value
P-value
187⋅5 6⋅29 −4⋅86 −10⋅88 7⋅06
30⋅2 1⋅87 1⋅46 2⋅86 2⋅79
6⋅21 3⋅37 −3⋅32 −3⋅80 2⋅53
