Tilapia Growth In Fluctuating Thermal Regimes Biology Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Fish species have evolved in environments where water temperature largely fluctuates between hours of the day, whereas the bulk of knowledge on their thermal biology originates from laboratory experiments at constant temperatures. Hence, it is uncertain whether fish get a benefit or incur a penalty when exposed to temperature fluctuations. Here, we measured the growth of Nile tilapia Oreochromis niloticus L. (20mg-20g) fed in slight excess during the hours of light, following their exposure to various thermal regimes fluctuating around the thermal optimum for growth (T°opt = 30°C), i.e. 27°L:33°N, 28.5°L:31.5°N, 30°L:30°N, 31.5°L:28.5°N and 33°L:27°N (two replications per treatment, six rearing weeks, growth controls at weekly intervals). For each thermal regime, a log-log model was constructed between the fish wet body mass (WM) and the specific growth rate (SGR, % WM day-1), i.e. log SGR = a + b log WM. The a and b coefficients of the five models were equated to the daily thermal amplitude (from -6°C to +6°C). For both parameters, the resulting models were third order polynomials with a parabolic shape. These models (R2>0.99) indicated that juvenile O. niloticus benefit from daily oscillations around T°opt until a pivot size of circa 160 mg, above which oscillating temperatures impact negatively on their growth. Slower growth under fluctuating temperatures is accompanied by greater size dispersal. Functional hypotheses on how temperature fluctuations affect fish growth are discussed, together with their implications in natural habitats and aquaculture systems with different contexts of food availability. Methodological aspects, especially the necessity of equating the pivot temperature of fluctuating regimes to T°opt are also debated.

Key-words fluctuating thermal regimes, size heterogeneity, optimum temperature, Oreochromis niloticus


Temperature is presumably the most directing environmental factor (sensu Fry 1971) in heterothermal animals as it intimately governs their physiology, growth, distribution area, habitat use and behaviours. Proportionally, aquatic animals can be more affected by changes in temperature than their terrestrial counterparts, because water has a greater specific heat than air. The majority of animal species have evolved under temperatures that fluctuate between years and seasons, but also during the daily cycle. The amplitudes of daily thermal fluctuations vary between latitudes, climates and seasons, but also between habitats, as they are inversely proportional to depth and water current. Similarly, the vast majority of aquaculture takes place in waters that are not thermoregulated and undergo more or less pronounced daily variations. Quite paradoxically, while thermal fluctuations prevail in the wild and in most culture systems, the bulk of knowledge on the thermal biology of aquatic heterotherms originates from laboratory studies under constant or almost constant temperatures. Without proper validation, it is uncertain whether results obtained at constant temperatures can be extrapolated to fluctuating environments (see Crawshaw, 1977).

Previous comparisons between the effects of stable and fluctuating thermal regimes on fish have produced contrasting conclusions. Some studies demonstrated that the exposure to fluctuating thermal regimes enhanced the fish tolerance to high temperatures (mosquitofish Gambusia affinis (Baird & Girard), Otto, 1974; sheepshead minnow Cyprinodon variegatus variegatus Lacépède, Bennett & Beittinger, 1997; cutthroat trout Oncorhynchus clarkii clarkii (Richardson), Johnstone & Rahel, 2003; spikedace Meda fulgida Girard, Carveth et al., 2007), whereas no difference was found in others (Currie, 1995). In several instances, daily fluctuations were found to boost growth (rainbow trout Oncorhynchus mykiss (Walbaum), Hokanson et al., 1977; striped bass Morone saxatilis (Walbaum), Cox & Coutant, 1981; brown trout Salmo trutta Linnaeus, Spigarelli et al., 1982; coho salmon Oncorhynchus kisutch (Walbaum) and goldfish Carassius auratus (Linnaeus), Konstantinov et al., 1990; spikedace, Carveth et al., 2007). In other studies, fluctuating thermal regimes were reported to produce slower growth or lower growth efficiencies than constant temperatures (Tahoe sucker Catostomus tahoensis Gill & Jordan, Vondracek et al., 1982; Arctic charr Salvelinus alpinus (Linnaeus), Lyytikäinen & Jobling, 1998, 1999; brown trout, Flodmark et al., 2004). Finally, in other species, no significant effect was documented (Lahontan cutthroat trout, Dickerson & Vinyard, 1999; Japanese medaka Oryzias latipes (Temminck & Schlegel), Dhillon & Fox, 2007; blue tilapia Oreochromis aureus, Steindachner Baras et al., 2000).

Certainly, part of this variation originated from genuine differences between species or populations. On the other hand, the protocols that were used for testing the effects of daily thermal fluctuations varied substantially. In some instances, fluctuating regimes were compared to constant temperatures corresponding to the upper and lower values of the fluctuation. In other instances, the constant temperature that served as control stood at mid of the fluctuating thermal range. Furthermore, the thermal regimes under study were rarely equated with the thermal optimum for growth (hereafter T°opt), which also corresponds to the final thermal preferendum, i.e. the temperature at which fish placed in a thermal gradient eventually congregate, if given time (Jobling, 1981). During the ontogeny, the ratio between gill surface and body volume varies allometrically, so T°opt is size-dependent in most fish species (synthesis in Jobling, 1994), and in general small fish prefer (McCauley & Huggins, 1979; Baras & Nindaba, 1999; Hernández et al., 2002) and perform better at warmer temperatures than larger conspecifics (Pedersen & Jobling, 1989; Imsland et al., 1996, 2006). Cichlids are no exception to this general rule (Oreochromis mossambicus Peters, Mironova, 1976; Nile tilapia O. niloticus (Linnaeus), Mélard, 1986; blue tilapia, Baras et al., 2002). The breadth of the growth-to-temperature response also varies during the ontogeny (e.g. stone loach Barbatula barbatula (Linnaeus), Elliott et al., 1996; blue tilapia, Baras et al., 2002). Hence, the effect of fluctuating temperatures on growth is also expected to vary during the ontogeny of fishes, but evidence is tremendously lacking.

This study aimed to test the effects of constant and oscillating thermal regimes on the survival, growth and size heterogeneity of the Nile tilapia Oreochromis niloticus, and whether these effects were size-dependent (size range of 20 mg to 20 g). In echo to the aforementioned paragraph, the pivot temperature for the oscillating regimes was set as close as possible to T°opt. By definition, T°opt is the temperature that produces the fastest growth, but also the value around which variations in temperature normally cause the smallest variations of fish growth, unless the fluctuation itself produces an intrinsic positive or detrimental effect.

Material and methods


The fish were Oreochromis niloticus Maryut, from a captive population, which originated from the Marine Centre of Tajoura (Lybia), and was transferred in 1999 to the Aquaculture Research Station of the Tunisian National Institute for Marine Sciences and Technologies at Bechima Gabès (Turki & Kraïem, 2002). In each experiment of this study, the fish were full siblings that were collected as hatchlings from the mouth of a mouthbrooding female. Water temperature in the broodfish tank, and rearing facilities prior to the experiments was 30±1°C, i.e. the very same average temperature as during the experimental periods.

Experimental design

The study aimed at testing the effects of controlled daily oscillations of water temperature on the survival, growth, and size heterogeneity of young Nile of different ages and sizes (from circa 20 mg to circa 20 g). Two daily thermal amplitudes (3 and 6°C) were evaluated and compared with a control, where temperature was maintained as constant as technically possible. In two fluctuating thermal regimes, temperature was warmer during the day than during the night, as it normally is in the wild. In addition to these "natural" regimes, we also evaluated two mirror treatments with the very same daily amplitudes (3 or 6°C), but where water temperature was warmer during the night than during the day (Figure 1). This design enabled testing whether the effect of daily thermal amplitude was dependent on the coupling or decoupling of light and temperature. It also permitted testing whether the variables under scrutiny were dependent on the temperature at the time of feeding, since fish were fed exclusively during the day in this study (see Rearing conditions). For all five thermal regimes, two replicate groups were used. In the rest of the text, each thermal treatment is designated as X°L:Y°N, where X and Y are the temperatures during the hours of light (L) and night (N), respectively.

The mean daily temperature in all groups was identical and set as 30°C, which corresponds to the thermal optimum for growth (T°opt) in young O. niloticus Maryut (Azaza et al., 2008). The maximum thermal amplitude that was evaluated here was 6°C (i.e. 27-33°C), a value that is frequently observed in natural environments, as well as in culture ponds in tropical regions. No amplitude greater than 6°C was evaluated, essentially because it was important that the maximum daily temperature did not exceed the temperature at which food intake is maximal (a few degrees above T°opt; Brett, 1979; Jobling, 1997), otherwise feeding declines steeply and affects growth directly, as well as indirectly, due to the degradation of water quality that ensues from the decay of uneaten food. The temperature at which food intake is maximal is currently unknown for the Maryut population, but in view of previous studies on the thermal biology of this population (Azaza et al., 2008), it is certainly not exceeded at 33°C for fish ranging from 20 mg to 20 g.

The periods of thermal transition were synchronised with the light cycle and were programmed to last for two hours each (06:00-08:00 h and 18:00-20:00 h). This schedule permitted maintaining water temperature as constant as possible during the 10-hour feeding period (08:00-18:00 h). At first sight, thermal transitions as steep as 3°C per hour might look exaggerated, but they do not exceed the rapidity of thermal transitions of natural nursery habitats on sunny days (e.g. Baras & Nindaba, 1999; Finlay et al., 2000).

Rearing conditions

This experimental design was implemented in two experiments (duration: 3 weeks each) with fish of identical origins but of different sizes and ages. The first experiment started at mid of the first week of exogenous feeding, after fish had fully exhausted their yolk (mean wet body mass [WM] ± SD = 22.3 ± 2.3 mg, mean skewness of -0.337, groups of 100 fish each). The second experiment started with fish about 100 times larger (mean WM ± SD = 2.27 ± 0.17 g, mean skewness of-0.046, groups of 30 fish each), which were about the same size as at the end of the first experiment. Altogether, these two experiments enabled testing the effects of fluctuating thermal regimes over three orders of magnitude (1000:1 ratio) for body mass, which represents more than 50% of the species size range when plotting body mass on a logarithmic scale (i.e. Nile tilapia rarely exceed 2 kg). In both experiments, fish were sedated with tricaine methanesulfonate (50 ppm), weighed individually to produce groups where mean size and size heterogeneity were as similar as possible. Thereafter, the groups of fish were randomly allocated to the different thermal regimes.

The five thermal regimes were evaluated in five indoor water recirculating systems in the rearing facilities of the Aquaculture Research Station of Bechima Gabès (Tunisia). Each system comprised two 85-L rearing aquaria (35 x 80 x 45 [h] cm), a 180-L reservoir tank for filtration and thermoregulation (with a 2-kW thermostatic immersion heater), and a 20-L biofilter with UV-sterilization lamp (Oase, model Filtoclear UWC 9/11W). The water flow in the aquaria was 1.5-2.0 L min-1, and 2.0-4.5 L min-1 (start to end of experiment), in the first and second experiments, respectively. Supplemental aeration was provided by individual air stones to maintain the oxygen level as close as possible to saturation. Water pH and alkalinity averaged 7.5 and 105 mg CaCO3 L-1, respectively. Day length was maintained at 12L:12D, with lights on from 08:00 to 20:00 h (light intensity of 800 Lx at the surface of the water, as measured with a Digital Instrument LX-101).

In the systems with a fluctuating thermal regime, the setting of the thermostat was changed at 06:00 and 18:00 h in order to attain the dedicated temperatures at 08:00 and 20:00 h, and thus to fit the thermal regime to the light cycle (Figure 1). The experiments were conducted during wintertime (November 2007 and January 2008), when the air temperature inside the rearing facility was cool enough to enable the return of the temperature to the daily minimum within 2 hours. However, every day, it was verified that the fluctuations followed the desired patterns, and, for example, small amounts of cool water was added to cool down the system within time on slightly warmer days.

Fish were fed formulated feed, which were distributed with automatic feeders from 08:00 to 18:00 h, during the period when lights were on and water temperature was constant. The restriction of the feeding period to the hours of light was based on the common sense hypothesis that O. niloticus is a diurnal species, that feeds essentially during the day (even though juveniles can equally benefit from night-time feeding; Baras et al., 1995). This restriction was further motivated by reducing as much as possible the risk of water degradation during the night if, for any reason, fish consumed less food than expected in a particular treatment. Food composition stood as 42% protein, 8% fat and 18.9 kJ.g-1 gross energy for the first experiment on small juveniles, and 36% protein, 6% fat and 17.7 kJ.g-1 gross energy for the second experiment on larger fish. These compositions were found to produce fast growth in O. niloticus Maryut (Azaza et al., 2005). The diameter of pressed pellets was less than 0.25 mm, and 1.0-1.5 mm for the first and second experiments, respectively. Food rations were calculated after Mélard (1986) in order that fish be fed in slight excess throughout. They were incremented every day, partly based from calculations, partly from the observation of the amount of the uneaten food at the end of each feeding period. Uneaten food was removed by siphoning during the hour following the end of food distribution, always with the objective of preventing any degradation of water quality that might have interfered with the genuine effect of daily thermal fluctuations.

Aquaria were searched for dead fish twice a day, before and after the period of food distribution. Water temperature in the reservoir tank of each recirculating system was recorded automatically every hour, and dissolved oxygen twice a day (08:00 and 18:00 h) with a digital thermo-oxymeter (WTW, MIQ/C184, accuracy of 0.1 °C and 0.1 mg O2.L-1). Total ammonium and nitrite concentrations were measured with a spectrophotometer on the days of fish measurement (D8, D15 and D22). Thereafter, water was added to compensate for evaporation during the past rearing week, so the values of water quality that were measured here are slightly pessimistic since they were measured on the day when the water volume in the recirculating system was lowest.

Each experiment lasted for three weeks, with weekly controls. On the morning of the control days, fish were captured with a dipnet, sedated with 50 ppm tricaine methanesulfonate, weighed (nearest 0.1 mg and nearest 0.01 g for the first and second experiment, respectively), and returned to their aquarium. The sole difference between the two experiments resided in that all survivors (n ≤ 30) were measured at each control in the second experiment, whereas in the first experiment, 30 fish were randomly sampled from each aquarium. Food distribution on the control day was suspended during the morning, and resumed in the early afternoon, at least 3 hours after the last fish were controlled, in order to minimise the effects of handling on food intake.

Data analysis

Survival rates were compared between groups with chi-square analysis. The effect of the thermal regime on growth was tested with one-way analyses of variance (ANOVA) and Scheffe post-hoc tests for comparisons of means. Kruskal-Wallis and Mann-Whitney U-tests were used wherever parametric analyses could not be applied (between-treatment comparisons of coefficient of variation of wet body mass, skewness coefficients, minimum and maximum body mass within each group of fish). For each rearing week, the average specific growth rate (SGR) was calculated as SGR (% WM d-1)= 100 (Ln WM2 - Ln WM1) (t2-t1)-1, where WM2 and WM1 are the mean wet body mass at times t2 and t1, respectively. Here, time was expressed by reference to the number of feeding days, thus 7 days for the first week of each experiment, and 6.5 days for the next two weeks of each experiment. For each thermal regime, a SGR-to-WM model was produced by simple linear regression analysis (after a logarithmic transformation of the two variables), using the data from both experiments (6 weeks and two groups per thermal regime). Null hypotheses were rejected at p<0.05.


In both experiments, the experimental settings were well respected. Fish size at the start of the experiments did not differ significantly between groups (mean WM of 22.1 to 22.4 mg, and 2.25 to 2.30 g), nor did the coefficient of variation of the distributions of wet body mass (CV WM of 10.0 to 11.3%, and 6.1 to 8.4%) and the skewness coefficients (from -0.60 to -0.06, and -0.38 to +0.41). Temperature never deviated from the experimental targets by more than 0.05°C on average, and by more than 0.3°C punctually, and thermal transitions always took place within 2.0 ±0.2 hours (Table I). Similarly, oxygen levels never dropped below 70% saturation, and the concentration of nitrogenous compounds remained very low in all treatments, during each week of the two experiments (Table I). In both experiments, survival on D22 was high and did not differ significantly between replicates or treatments (i.e. 92-95 %, and 90.0-93.3% in the first and second experiments; chi-square of 1.372 and 1.087, P=0.9980 and P=0.9992, respectively).

At the end of the first experiment (D22), fish size was proportional to the temperature during the hours of light, when food was distributed. However, between-treatment differences were tenuous (mean WM of 1637 to 1802 mg) and not significant, due to strong size heterogeneity within groups (mean CV WM of 31.1%; Fig. 2). Kruskal Wallis tests revealed no significant treatment effect on the CV WM (H=5.24) or skewness coefficients (H=1.29), and on the sizes of the smallest (WMmin; H=4.58) and largest individuals (WMmax; H=0.19). No single significant difference between replicates or treatments was observed for any of the aforementioned variables on D8 and D15 either. This was partly due to the fact that the CV WM soared in all groups during the first rearing week (on average from 10.6 to 32.3%), and remained high thereafter. Skewness also increased in all groups during the experiment, but essentially during the second and third rearing weeks (mean skewness of -0.35, -0.24, -0.03 and +0.55, on D1, D8, D15 and D22, respectively).

Size heterogeneity and skewness also increased rapidly during the second experiment, as CV WM passed from an average of 7.4 to 25.1%, and skewness from -0.03 to +0.40 during the first rearing week. At this moment (D8), there was no significant difference between groups as regards fish size (Figure 3) and size dispersal (CV WM [Kruskal Wallis test, H=0.27], skewness [H=1.64], Wmmin [H=1.19] and Wmmax [H=0.96]). By contrast, at the end of the second rearing week (D15), fish raised under strongly fluctuating temperatures (33°L:27°N and 27°L:33°N) were significantly smaller than those raised at constant or slightly oscillating temperatures. These differences amplified during the third rearing week, and on D22 fish raised under 27°L:33°N and 33°L:27°N averaged no more than 11.1 and 12.8 g, while the others averaged 15.4-17.7 g (Fig. 3). Between-treatment differences were also conspicuous for the smallest and largest individuals within each group, thereby indicating that all size classes were negatively affected by marked thermal oscillations (i.e. WMmin of 8.00-8.31 g versus 8.62-10.43 g; Kruskal Wallis test, H=6.87, p=0.0322; WMmax of 20.30-23.08 versus 23.08-24.70 g; Kruskal Wallis test, H=6.12, p=0.0468). The CV WM and skewness of size distributions on D22 were significantly higher when fish were raised under strongly fluctuating temperatures than under more stable thermal regimes (CV WM of 24.0-28.1% versus 20.5-23.6%; Kruskal Wallis test, H=6.55; p=0.0379; skewness of 0.55 to +2.25 versus -0.29 to +0.62; Kruskal Wallis test, H=7.28; p=0.0262).

SGR-to-WM relationships were calculated for all five thermal regimes, and compared in order to determine the advantage or penalty for a fish of a particular size to be exposed to moderate or strong daily thermal oscillations (Table II). The greater the daily thermal amplitude, the higher the intercept and the steeper the slope of the log-log relationship. The intercepts and slopes of these growth models were equated with the amplitude of the daily thermal oscillations, and for both parameters, consistent parabolic relationships were found (Figure 4). These models clearly indicate that small juveniles exposed to temperatures oscillating around T°opt get a growth advantage over those raised at constant T°opt. However, because of the greater steepness of the slopes under fluctuating temperatures, this advantage vanishes rapidly and above a pivot size of circa 160 mg, fish exposed to oscillating temperatures incur a growth penalty, the severity of which increases with fish size (and with the amplitude of the thermal oscillation).


Methodological aspects

In both experiments and for all thermal regimes, the presence of small amounts of wasted food in each tank at the end of all rearing days indicated that fish had indeed been fed in excess throughout. Despite fish were fed in excess, the concentrations of nitrogenous compounds remained low, and highly satisfactory by reference to the recommendations for rearing Nile tilapia (Ballarin & Hatton, 1979). Furthermore, the measurements of water quality were done on a weekly basis, prior to adding clean water, so they give a pessimistic view of the actual water quality during each rearing week. On all rearing days, the oxygen levels at the start of the feeding period were close to saturation, and they never dropped values that were found to affect food intake or food conversion in O. niloticus (40-50% saturation; C. Mélard, O. Plunus & E. Baras, unpublished data). The adequacy of the experimental conditions is further attested by the low fish mortality in all experimental groups (< 0.5% d-1), and by the fact that growth was fast, by reference to previous studies on the Maryut population (Azaza et al., 2008), but also by reference to other populations of O. niloticus raised in similar experimental conditions (e.g. Manzala population, Mélard, 1986). The mean diurnal or nocturnal temperatures never deviated from the target temperatures by more than 0.05 °C, and the slopes of thermal transitions did not depart from the targeted values by more than 0.03°C h-1. Altogether, these observations support the idea that the results documented in this study reflected the intrinsic effects of temperature oscillations on the growth of O. niloticus, rather than discrepancies between experimental tanks, or side effects resulting from treatment-dependent alterations of water quality. The similarity between the results observed under mirror regimes of identical daily thermal amplitudes suggests that the coupling between light and temperature has no major importance on the growth of O. niloticus. A similar conclusion was drawn in a study on O. aureus where mirror regimes fluctuating between 28 and 35°C were evaluated (Baras et al., 2000), thereby indicating that this feature is shared by several fish species, at least among the tilapiines, but this is not systematical in fish (see the contrasting responses of Theragra chalcogramma and Anoploma fimbria, Sogard & Olla, 1998).

Yet, two particular aspects of the experimental design might be pointed out as possible artefacts. At first, the fish used in this study had been raised at almost constant temperatures around T°opt prior to the experiments. It cannot be ruled out that the absence of acclimation to fluctuating temperatures might have been responsible for the slower growth under strong daily thermal oscillations. However, if the lack of acclimation had been the key to the differences observed in this study, then the differences would have been expected to soar during the first rearing week of the experiment, which was not the case in any of the two experiments of this study. It cannot be strictly ruled out that the lack of acclimation to fluctuating temperatures impacted on fish growth, but if it did, the effect was insignificant or extremely transient.

The second possible artefact refers to the fact that food was distributed exclusively during the hours of light during this study. The absence of night-time feeding was deliberate and aimed at preserving water quality, the degradation of which might have impacted on fish growth to a greater extent than the variable under study, especially in a context where fish were fed in slight excess. A direct consequence of this design was that fish from different treatments were fed at different temperatures (i.e. 27, 28.5, 30, 31.5 and 33°C). It is notorious that the growth of fish raised at constant temperatures is slower if the temperature departs from T°opt (e.g. Jobling, 1994). Hence, before debating the possible effects of temperature fluctuations on fish growth, it was necessary testing whether the growth rates observed at the five thermal regimes under study was similar to those that would have been expected under constant 27, 28.5, 30, 31.5 and 33°C. Expected growth were modelled from the study by Azaza et al. (2008) who raised O. niloticus Maryut under constant temperatures ranging from 22 to 34°C. Starting from the raw data of this study, it was possible producing a general model of growth (SGR, % d-1) against wet body mass (WM, g) and water temperature (T°, °C) for fish fed in slight excess, i.e. Log (1+SGR) = -18.65+20.17 Log T° - 3.16 (Log T°)3 - 0.15 (Log WM) (Log T°)2. In order to facilitate the comparisons between the study by Azaza et al. (2008) and this study, all data were expressed by reference to a growth at constant 30°C (i.e. the sole experimental conditions that were common to the two studies). Fish growth at constant 30°C in the study by Azaza et al. (2008) was slightly slower than here, so expected growth rates were adjusted to permit straightforward comparisons. The predictions of this model are shown in Figure 5, together with the values observed under temperatures fluctuating around 30°C in this study. In both experiments of this study, fish raised under strongly fluctuating thermal regimes attained sizes that departed substantially from those that would have been expected if their growth had been exclusively governed by the temperature during the period of food distribution.

This analysis further emphasizes that the growth of O. niloticus is less influenced by the temperature at the time of feeding than by the average daily temperature (which was identical in all treatments here) and by the amplitude of the daily thermal oscillation. It is probable that this tendency is not restricted to O. niloticus. The finding that the average daily temperature has a predominant effect on growth (notwithstanding the genuine effect of thermal fluctuations) might contribute to account for the contrasting conclusions of studies where thermal regimes oscillating between two extreme temperatures were compared to constant regimes at these two extreme temperatures (Figure 6). With such protocols, it can be predicted on a conceptual basis that the growth under oscillating temperatures would be intermediate if the two extreme temperatures were either below or above T°opt (Figure 6.C and 6.F). By contrast, the growth under the oscillating regime would be faster than under the two constant temperatures if these stood apart from T°opt (Figure 6.I ). If T°opt stood in between one constant temperature and the pivot temperature of the oscillating regime, then the growth at the oscillating regime would similar to the aforementioned constant temperature, and faster than at the other constant regime (Figure 6.L). This functional explanation does not overlook the possibility that fluctuating regimes have a genuine (positive or negative) impact on growth. However, it suggests that this effect might be masked or amplified depending on the thermal range under study and its position by reference to T°opt. Figure 6 provides, on a conceptual basis, illustrations of such possible confusions between negative and neutral effects (Figure 6.C versus 6.G), or between neutral and positive effects of daily oscillations (Figure 6.E versus 6.L). In some cases, positive and negative genuine effects of thermal oscillations might even produce growth performances that look most similar (Figure 6.E versus 6.P). A practical example of such possible confusion can be found in Baras et al. (2000), where juvenile O. aureus (12-1500 mg) raised under 35°L:28°N or 28°L:35°N grew at a similar rate as under constant 35°C. At first sight, this suggests a neutral effect of oscillating temperatures on the growth of this species. However, the mean T°opt over the 12-1500 mg size range in this species was later found to average circa 31.5°C (Baras et al., 2002), thus the value of the pivot temperature of the oscillating regimes in the study by Baras et al. (2000). If the daily thermal fluctuation had no impact on the growth of O. aureus, fish raised under thermal regimes oscillating around 31.5°C should have grown faster than at constant 35°C, whereas their growth was slightly slower. All in all, the growth penalty for living under a 7°C daily oscillation amounted to no less than 13.5% in this species, but this penalty passed unnoticed before the temperatures under study were equated to T°opt.

Until this study, the variable effects of fluctuating temperatures had been essentially attributed to the extent of the amplitude, i.e. with moderate amplitudes having a positive effect, whereas higher amplitudes had a negative effect on growth (Konstantinov et al., 1987, 1990; Meeuwig et al., 2004; Dong & Dong, 2006). The conceptual approach illustrated in Figure 6 as well as this example from the authors' data clearly indicate that a reliable statement on the positive, neutral or negative effect of thermal oscillations can hardly be derived from a straightforward comparison between growth rates under oscillating and constant extreme temperatures. Similarly, any protocol testing the effects of fluctuating thermal regimes of different amplitudes around a particular pivot temperature can produce contrasting responses depending on whether the pivot temperature is colder, warmer or close to T°opt, as already expounded by Jobling (1997). These arguments strengthen the protocol that was developed in this study, where the pivot temperature was selected as close as possible to the T°opt of the size range under study. They also emphasize the difficulty of drawing meaningful comparisons with other studies on fluctuating temperatures where growth at the constant pivot temperature was not evaluated or equated with T°opt.

Size-dependent effects of thermal oscillations

This study demonstrated that the effects of thermal oscillations around T°opt were size-dependent. The modelling approach (Figure 4) indicated that juvenile O. niloticus benefit from oscillating regimes until a pivot size of 160 mg, above which oscillating temperatures start impacting negatively on their growth. The absence of significant between-treatment differences during the first experiment of this study can be accounted for by the fact that the geometric mean size of fish during this experiment was close to the pivot size, whereas in the second experiment, when fish were above the pivot size, growth was significantly dependent on the amplitude of daily thermal fluctuations. The parabolic shape of the models between the daily amplitude of the thermal oscillation and the intercept or slope of the growth models is consistent with the well admitted parabolic nature of the relationships between temperature and growth in fish (Jobling, 1994).

As a matter of fact, no between-treatment difference was observed at the end of the first rearing week of the second experiment, despite fish were 10 times larger than the pivot size at the start of this experiment. It is suggested that the effect of fluctuating temperatures during this period was masked by the settlement of dominance hierarchies. It is frequent that dominance hierarchies settle or be reinstalled soon after fish have been rearranged into groups, especially when group size is low and fish are homogenous in size, as was the case in this study. The settlement of a dominance hierarchy is generally accompanied by heterogeneous food intake and growth, generally to the detriment of the smallest fish, thereby resulting in slower growth, greater size dispersal and increased skewness. All three characteristics were observed at the end of the first week of the second experiment. Thereafter, size heterogeneity remained almost stable and skewness increased slightly, and growth rebounded, as it is frequently the case after the hierarchy has settled. However, size heterogeneity and skewness were higher, and growth was slower among the groups that were raised under strongly fluctuating temperatures.

A functional explanation behind the greater size dispersal under fluctuating temperatures, lies in that fish exposed to fluctuating regimes spend a greater part of the daily cycle at temperatures that are further from T°opt. There is now a growing number of evidences that size dispersal is lower among fish raised at temperatures close to T°opt than at colder or warmer temperatures (for O. niloticus, see Azaza et al., 2008). The functional hypothesis behind this phenomenon invokes the possibility that T°opt varies between individuals, for example if the gill surface varies between fishes of identical sizes. If this hypothesis is valid, size heterogeneity is expected to increase if the rearing temperature departs from T°opt. By virtue of the parabolic shape of the relationship between growth and temperature, size heterogeneity is expected to rise faster if the rearing temperature is warmer than T°opt, than if it were cooler than T°opt (E. Baras & M. Daffé, personal communication). A similar explanation might apply for thermal regimes fluctuating around T°opt, since the departure of the rearing temperature from T°opt increases with the amplitude of the daily thermal fluctuation. Additionally, it can be put forward that not all fish of a particular size equally tolerate strong daily thermal fluctuations. Both hypotheses remain to be tested experimentally, but in both cases, a formal demonstration would require cloning protocols. In addition or alternatively to these physiological factors, it cannot be excluded either that fluctuating temperatures impacted on the behavioural register, and exacerbated dominance hierarchies through enhanced aggressiveness. Recent experiments on two tilapia species (O. niloticus and Sarotherodon melanotheron) indeed provided evidence that aggressiveness was higher, and could produce higher mortality when fish were raised at warm temperatures (E. Baras & M.S. Azaza, unpublished data).

The reasons for why the effect of thermal oscillations shifted from a benefit to a disadvantage during the ontogeny of O. niloticus, remain obscure as well. However, it should be reminded that the thermal range under evaluation in this study remained constant (27-33°C), whereas T°opt and the thermal comfort range probably shifted towards slightly cooler temperatures as fish grew bigger, by analogy to studies on other cichlids. For example, T°opt in Oreochromis aureus was found to decrease by circa 1°C for each 10-fold increase of the wet body mass (Baras et al., 2002). Hence, the warmest temperatures, which were presumably closer to the T°opt of small juveniles, became increasingly unsuitable, and potentially stressful to fish of increasing size. It cannot be excluded either that oscillations of identical amplitudes progressively became increasingly unsuitable to fish of increasing size.

Stress parameters were not measured in this study, so we cannot claim that fish exposed to greater daily thermal amplitudes were more stressed than others. However, it is worth remembering that abrupt thermal shocks induce stress in fish (Reaves et al., 1968; Bevelhimer & Bennett, 2000). Furthermore, a recent study on the sea cucumber Apostichopus japonicus (Echinodermata) provided evidence that individuals exposed to strongly fluctuating temperatures around T°opt (18°C) exhibited a lower hexokinase activity but a higher super oxide dismutase activity, which suggests an oxidative stress (Dong et al., 2008). Dong et al. (op. cit.) also reported a higher level of heat-shock proteins (HSP70) among the animals that were temporarily exposed to the warmest temperatures, thereby suggesting a greater level of protein damage that might have accounted for their slower growth. There is currently very little counterpart information on fish raised under fluctuating temperatures (salmonids: Thomas et al., 1986; Shrimpton et al., 2007), so it can only be speculated that these mechanisms might have intervened here.

Almost independently from the physiological mechanisms that govern the tolerance of fish of various sizes to water temperature, parallels can be drawn between this experimental study and field studies on the ontogenetic variations of habitat use by fish species. It is frequent that young, small fishes with low swimming capacities but high T°opt live in shallow habitats with slow water velocity, which undergo strong daily thermal fluctuations on sunny days. By contrast, larger fishes tend to occupy deeper or faster flowing habitats, which are thermally more stable. A series of factors, including predation pressure, ontogenetic variations of oxygen transport and thermal preferenda have been invoked to account for these size-dependent variations in habitat use and daily migrations (e.g. Baras & Nindaba, 1999). In view of the results of this study, which show that small fish tolerate thermal fluctuations better than larger / older conspecifics, it cannot be excluded either that the search for environments that are more thermally stable be an additional factor that fosters the shift towards deeper and faster flowing habitats, which provide such characteristics.

Synthesis, conclusions, perspectives

This study did not pioneer the field of fluctuating thermal regimes, but it provided substantial evidence, on a conceptual basis, that no sound conclusion can be drawn on whether constant or fluctuating temperatures impact on fish growth unless the temperatures under study have been equated with the thermal optimum for growth. It also highlighted that the life in thermally stable or fluctuating environments can have a cost or bring a benefit, depending on fish size and age, and that the growth benefit brought about by environments that offer a greater thermal stability might be among the keys that force fish of increasing size to deeper and faster flowing habitats. The models provided in this study probably echo to a number of fish species, at least as regards their general nature (i.e. young fish being more tolerant than large fish to marked daily thermal fluctuations). However, data on how fluctuating temperatures impact on fish of different sizes are still tremendously lacking for other fish species, from other taxa or latitudes.

As regards aquaculture, the information provided in this study indicates clearly that the growth observed under almost constant temperatures cannot be extrapolated to situations where fish would undergo marked daily fluctuations. It should also be reminded that fish the differences observed in this study referred to a context where fish were fed in slight excess. If the life under fluctuating temperatures is more energy-demanding than under more stable thermal regimes (see Lyytkaïnen & Jobling, 1998), the growth penalty might be more severe in actual rearing conditions, where the feeding level generally lies in between the optimum and the maximum food rations to avoid food wastage. Future studies should aim at investigating this issue in contexts of different food availability. In addition to their interest for aquaculture, these studies would also contribute to document the respective advantages or drawbacks of living in highly fluctuating thermal environments with contrasting trophic levels.


This study benefited from the bilateral collaboration between Tunisia (INSTM) and France (IRD), under the auspices of the French Ministry of Foreign Affairs. We want to thank Mrs S. Kalboussi and K. Elebdelli for technical assistance throughout the period of this study. Thanks are also given to M. Dominique Caseau for checking the English.