Lack of coherence in the warming responses of marine crustaceans

Authors


Summary

  1. Understanding the extent to which organisms are affected by climate change and are capable of adapting to warming is essential for managing biodiversity. Recent macrophysiological analyses suggest that range-related responses to warming may be more coherent (less variable) and predictable in marine than in terrestrial systems.

  2. To examine this generalization, we investigate basal upper thermal tolerances (measured as CTmax), the extent of their phenotypic plasticity and the impacts of different rates of temperature change on these tolerances, in five species of intertidal crustaceans from three distinct thermal regimes, incorporating South African (RSA) shores and sub-Antarctic Marion Island (MI).

  3. For all species, lower rates of change resulted in lower CTmax, while acclimation resulted in varied responses depending on the rate of temperature change. At fast rates of temperature change, higher temperature acclimation resulted in elevated CTmax, while at slow rates of change, acclimation had no effect or resulted in a decline in CTmax.

  4. Maximum habitat temperatures recorded at the organisms' microsites were lower than the CTmax for the MI populations but were above CTmax at slow rates of change for RSA populations. Thus, populations from more equatorward locations have a lower tolerance of extremes than those from cooler regions. In addition to reduced warming tolerance, RSA populations had a lower acclimation capacity than their sub-Antarctic counterparts.

  5. We find substantial differences in long-term responses among groups in different areas as a consequence of spatial variation in the interactions among basal tolerance, phenotypic plasticity and thermal environments. These outcomes emphasize the significance of examining forecasts using a range of data and approaches so that their certainty can be established to inform key policy decisions in a spatially appropriate context.

Introduction

Humans are affecting biodiversity in every area of the planet. Several of these impacts can be alleviated or are potentially reversible over short time-scales (Halpern & Warner 2002; Gaston et al. 2008; Simberloff et al. 2013). By contrast, mitigating the effects of climate change appears much less tractable given current CO2 levels, the growing pace of carbon emissions, and forecasts for increasing energy demand (Rockström et al. 2009; IEA 2012; Peters et al. 2012). As a result, understanding of the extent to which organisms will be affected by and are capable of adapting to such change is essential for conservation (Chevin, Lande & Mace 2010; Dawson et al. 2011; Hoffmann & Sgrò 2011; Bellard et al. 2012).

For terrestrial systems, comprehension of responses to warming is emerging rapidly. From physiological and evolutionary perspectives, it is evident that limited scope exists for alteration of high temperature tolerance (Addo-Bediako, Chown & Gaston 2000; Hoffmann, Chown & Clusella-Trullas 2013; Araújo et al. 2013). Species in subtropical to tropical regions have the lowest tolerance of warming (Deutsch et al. 2008; Huey et al. 2009; Duarte et al. 2012) and are most at risk of extinction (Sinervo et al. 2010). Moreover, the extent to which other factors, such as rates of warming, and other stressors, such as habitat alteration, might interact with thermal tolerance and its plasticity to exacerbate or ameliorate extinction risk is being increasingly well explored (Chown et al. 2009, 2010; Kearney, Shine & Porter 2009; Terblanche et al. 2011; Overgaard, Kristensen & Sørensen 2012). This work also suggests that links between physiological tolerances and changes in species range boundaries may be complicated by various factors such as tolerance to desiccation and spatial variation in climates and the strength of biotic interactions (Bonebrake & Mastrandrea 2010; Clusella-Trullas, Blackburn & Chown 2011; Bonebrake & Deutsch 2012; Sunday, Bates & Dulvy 2012). Hence, forecasts of range shifts, important for conservation, are unlikely to be straightforward (Araújo et al. 2013).

Recent macrophysiological analyses suggest that range-related responses to warming may be more coherent (i.e. less variable) and more predictable in marine than in terrestrial systems because of stronger relationships between thermal tolerances and range boundaries in marine than in terrestrial species (Sunday, Bates & Dulvy 2012). In terrestrial taxa, responses may be more variable (less coherent) because of the important roles of factors other than thermal limits in setting range boundaries. While a useful first approximation, data on which this idea has been built, and a range of other studies, suggest that it requires further exploration.

First, strong relationships between upper thermal tolerance limits and equatorward range boundaries are apparent for tropical marine species. However, these relationships show much greater variation extratropically (Sunday, Bates & Dulvy 2012). Thus, substantial differences might exist between tropical and non-tropical organisms, as is found in terrestrial systems (Deutsch et al. 2008; Kellermann et al. 2012). Such latitudinal variation in aspects of thermal tolerance (notably its range and extent of acclimation) has been widely discussed for terrestrial and marine species and is encapsulated in the macrophysiological literature as Janzen and Vernberg's rules, respectively (Gaston et al. 2009). Janzen's rule stipulates that thermal tolerance range and latitude should be positively related, while Vernberg's rule posits a positive relationship between the extent of acclimation and latitude (Gaston et al. 2009).

Secondly, marine species with the greatest thermal tolerance capacity appear to have the least phenotypic plasticity or ability to adjust upper thermal tolerance over the short term, usually reversibly (Stillman 2003). The exceptions are Antarctic stenothermic species which typically have limited tolerance and plasticity (Peck, Morley & Clark 2010). Although exploration of this finding (Gause's hypothesis, or a ‘negative relationship between acclimation ability and extent of initial tolerance’, Gaston et al. 2009) is growing, its generality for marine organisms has a narrow empirical basis (Somero 2011). Nonetheless, plasticity might affect substantially the ways in which environmental change is translated to range shifts (Chevin, Lande & Mace 2010; Chown et al. 2010; Sunday, Bates & Dulvy 2012).

Thirdly, experimental rates of temperature change have a profound effect on estimates of upper thermal tolerance limits in the species examined to date (the rate hypothesis – see Terblanche et al. 2011 for an overview). In several groups of marine organisms, slower rates of change lead to substantially lower limits than do faster rates of change (Peck et al. 2009; Nguyen et al. 2012; Richard et al. 2012). If general, then estimates of range filling, based on typically fast-rate experiments (Sunday, Bates & Dulvy 2011, 2012), may be confounded, unpredictably if the magnitude and direction of the effects are species specific, as is the case for terrestrial ectotherms (Allen, Clusella-Trullas & Chown 2012). Moreover, interactions between rates of change and estimates of plasticity may further complicate predictions (Chown et al. 2009).

In consequence, despite significant advances in understanding the responses of marine organisms to warming (Helmuth, Kingsolver & Carrington 2005; Pörtner & Farrell 2008; Somero 2012; Sunday, Bates & Dulvy 2012), current generalities are at best provisional. Here, to develop them further, we investigate basal upper thermal tolerances (measured as critical thermal maximum), the extent of their phenotypic plasticity (i.e. how the across-genotypes phenotypic mean of a trait changes with the environment, Pigliucci 2005) and the impacts of different rates of temperature change on these tolerances, in six populations of five species of intertidal crustaceans from three distinct thermal regimes. We focus especially on tolerance of ambient thermal extremes because of their growing significance in a climate change context (see Helmuth, Kingsolver & Carrington 2005; Clusella-Trullas, Blackburn & Chown 2011; Hansen, Sato & Ruedy 2012). In doing so, we also test explicitly Janzen, Vernberg and Gause's macrophysiological ‘rules’ related to thermal tolerance and its plasticity (Gaston et al. 2009). We use a range of experimental temperature change rates, from those typically used in past assessments and incorporated into macrophysiological studies (e.g. Sunday, Bates & Dulvy 2011, 2012), to much slower rates of change (0·17 °C per day) more likely to be experienced within generations of long-lived species.

Materials and methods

Study Species and Sites

Two Hyale amphipod (Hyalidae) and three Exosphaeroma isopod (Sphaeromatidae) species from the intertidal zone were selected (Table 1). The assumption has been made that coherent responses to change are as likely in these groups as for others (though prediction of habitat temperature may be more complicated, Sunday, Bates & Dulvy 2012). In addition, the intertidal zone is a marine environment with considerable climate change vulnerability (Somero 2012). Restricting the investigations to these groups also narrows the likely influence of widely differing life histories associated with marine higher taxa (Marshall et al. 2012).

Table 1. Locations and habitat temperature parameters for Hyale and Exosphaeroma spp. from Marion Island (MI) and South Africa (RSA). Latitude and longitude in decimal degrees are given in parentheses. Parameters denote: mean habitat temperature ± standard deviation (SD), Min: absolute minimum temperature, Max: absolute maximum temperature, Mean RoI: mean of daily maximum rates of temperature increase, Max RoI: absolute maximum rate of temperature increase. Rates of temperature increase were calculated as the difference in temperature between successive increasing records and divided by the period between records. For H. grandicornis, habitat temperatures were taken from Hangklip given the similarity and proximity of the two locations
SpeciesLocationCollection dateShore positionMean ± SD (°C)Min (°C)Max (°C)Mean RoI (°C min−1)Max RoI (°C min−1)
Hyale hirtipalma Trypot, MI (46·9S; 37·9E)April–May 2010Semi-exposed5·4 ± 2·2−1·414·20·0080·05
Exosphaeroma gigas Trypot MIApril–May 2010Submerged5·7 ± 0·8−0·99·10·0020·04
H. grandicornis Muizenberg, RSA (34·1S; 18·5E)February–June 2011Semi-exposed17·5 ± 2·79·733·10·0450·2
Exosphaeroma laeviusculum Hangklip, RSA (34·4S; 18·8E)January–June 2011Semi-exposed17·5 ± 2·79·733·10·0450·2
E. antikraussi Hangklip, RSAJanuary–June 2011Semi-exposed17·5 ± 2·79·733·10·0450·2
E. laeviusculum Lamberts Bay, RSA (32·2S; 18·3E)February–July 2011Semi-exposed15·9 ± 3·07·631·70·0270·1

Two environments with temperate and variable conditions (South Africa – hereafter RSA) and one more stable environment (sub-Antarctic Marion Island – hereafter MI), were selected to examine variation across different non-tropical sites (Table 1). To assess site-related microclimate variation, I-button data loggers enclosed in highly conductive, protective capsules [DS1922L (accuracy ± 0·5 °C) & DS1907, Dallas Semiconductor Maxim, USA; silicone capsule SL-ACC06, Signatrol, Tewkesbury, UK] were deployed at exposed, semi-exposed and submerged positions of the shore at the three sites: Trypot Beach (MI), Hangklip (RSA) and Lamberts Bay (RSA). Microsite temperature data (hereafter habitat temperatures) were collected at 1 h 30 intervals from May 2009 to September 2010 and December 2010 to August 2011 for the MI and RSA sites, respectively. For each species, microsite data were analysed from the shore position where most individuals were found (Table 1) and which corresponds to the species’ preferred habitat reported in the literature (Day 1969; de Villiers 1976; Kensley 1978; Branch et al. 1991). All species were typically found among algal fronds and boulders where semi-exposed loggers were deployed. Exosphaeroma gigas on MI favoured the substrate under boulders, and therefore, data from submerged loggers were used. Time series for submerged sites were screened for outliers indicating emersion from the water (n = 11 of 2359 data points for Hangklip and n = 29 of 2176 data points for Lamberts Bay). The mean, the absolute maximum and minimum, the range and the mean and maximum rate of increase were calculated for each temperature time series (Table 1).

Individuals were collected by hand from the sites (Table 1). For the MI amphipod population, the critical thermal maximum (CTmax) experiments at fast rates were conducted at the research station laboratory <5 km from the collection site. For the remainder of the experiments on MI populations, individuals were transported to South Africa by ship, maintained in temperature-controlled plastic tanks containing aerated sea water (mean temperature during the 9 day voyage ± SD = 5·9 ± 1·6 °C; I-button DS1922L and silicone capsule SL-ACC06) and provided with refuges (algae and rocks) and food (Durvillaea antarctica kelp from MI) ad libitum. No mortalities were recorded during the voyage. Individuals were then transferred for recovery and acclimation treatments to the laboratories at the Seapoint Research Aquarium (Cape Town) where the majority of work was undertaken. Individuals from the RSA populations were transported to these facilities within 2 days of collection using plastic containers with aerated sea water and kept at the temperatures of site of collection. During recovery, individuals were maintained in filtered (8–12 and 0·02 μm, MI and RSA, respectively), aerated sea water at temperatures resembling the locality of origin: 6·5 ± 0·4 and 15·1 ± 0·3 °C for MI and RSA, respectively (measured using silicone waterproofed I-buttons). Individuals were kept with a photoperiod of 12 L/12 D, and provided once with a standard amount of food (seaweed and kelp from their habitats) during recovery and then acclimation, and fed every 10 days during slow-rate trials. During the full period that individuals spent in the laboratory, fresh, filtered (0·02 μm) sea water at the required temperature was provided every 3 days, and sea water quality (temperature, salinity and dissolved oxygen) was monitored and maintained (see Appendix S1, Supporting Information for details).

Experimental Design and Analysis

Following field collection, individuals were given a recovery period of 1–3 days (fast rates) or 2–11 days (slow rates). During recovery, mortality was noted and organism identity reconfirmed (Appendix S1). Individuals were then randomly divided into three acclimation treatments, simulating a low, average and high sea water temperature at the locality of origin. These temperatures were 3·2 ± 0·5, 6·6 ± 0·3 and 11·6 ± 0·9 °C for MI populations, and 11·2 ± 0·2, 15·0 ± 0·3 and 19·2 ± 0·4 °C for RSA populations (measured using waterproofed I-buttons). Acclimation periods (following evidence that acclimation proceeds rapidly in most ectotherms, Claussen 1980; Weldon, Terblanche & Chown 2011; and also Gaston & Spicer 1998) were 5–12 days for fast-rate experiments and 13–15 days for slow-rate experiments.

After acclimation, CTmax was determined at the five rates of temperature change. Experiments were categorized into those with slow warming rates (1 °C per day, 1 °C per 3 days and 1 °C per 6 days corresponding to 0·0001, 0·0002 and 0·0007 °C min−1) and those with rapid warming rates (0·1 and 0·5 °C min−1) (Fig. S1). For the fast-rate trials, five individuals were placed individually in mesh vials housed in a plastic container (900 mL) containing filtered, aerated sea water immersed in a water bath (Grant Instruments GP 200-R4, Cambridge, UK). For all rate trials, experimental start temperatures were maintained for 30 min at 7 and 15 °C for MI and RSA populations, respectively. Thereafter, temperature ramping, controlled by the water bath, commenced. During experiments, individual responses to mechanical stimuli were monitored continuously and daily for fast- and slow-rate experiments, respectively. The temperature at which no response occurred was deemed CTmax. CTmax is typically taken as the temperature at which an organism no longer shows a coordinated locomotory response (Lutterschmidt & Hutchison 1997). In crustaceans, it usually approaches lethal temperatures (Somero 2012). Organisms that had reached CTmax were removed from the experiment and placed at their original start temperatures. Potential recovery was determined after 10 and 60 min. Fast-rate trials were also undertaken for individuals returned to the laboratory without acclimation, but with 1–3 days recovery period following collection (hereafter ‘field fresh individuals', Fig. S1). Each slow-rate trial was undertaken in a jacketed, Perspex tank containing filtered, aerated sea water, the temperature of which was controlled by a water bath (Fig. S2 in Supporting information). Owing to the low number of individuals in the field, Hyale grandicornis acclimated at 19 °C was not tested at 0·0001 and 0·0002 °C min−1 and slow-rate trials for Exosphaeroma antikraussi were not completed.

A total of c. 9–36 individuals were examined at each rate by acclimation treatment for each population (total of 1576 individuals for all populations). Wet mass (AE163 Mettler, Sartorius Analytic balance, Göttingen, Germany; ±0·0001 g) after blotting excess water with a paper towel and sex (using a Stemi 2000-C dissecting microscope; Zeiss, Barrington, NJ, USA) of all individuals were determined.

During acclimation and slow-rate trials, control individuals were maintained at a constant temperature of either 7 °C (MI populations), 11 °C (H. grandicornis) or 15 °C (Exosphaeroma laeviusculum). Controls were treated in the same way as experimental individuals and their response to mechanical stimuli was checked daily. To test if captivity period confounded results, the CTmax of 15 control individuals was determined at 0·5 °C min−1 within 2–3 days of the end of each slow-rate experiment. Control CTmax was compared to the CTmax of field fresh individuals at 0·5 °C min−1. Due to logistic constraints, these experiments were only performed on South African species. To establish if nutrition status confounded results, the lipid content of control and field fresh individuals was also determined. Dried individuals were dismembered and rinsed in three changes (24 h each) of a 1 : 2 chloroform–methanol solution (Bligh & Dyer 1959). Lipid content was estimated by determining dry mass before and after lipid extraction, except for E. gigas owing to scarcity in the field.

Prior to analysis, to adjust for mortality of weak or old individuals in slow-rate experiments, data >1 SD from the mean were removed from all rate of temperature change × acclimation combinations (representing a maximum of 8–15% of the data across populations). Plotting techniques (Faraway 2005; Crawley 2007), implemented in r (version 2.10.1; R Development Core Team 2009), were used to assess assumptions of normality and homogeneity of variances and revealed that only in a few instances data followed a uniform (short-tailed) distribution, for which the consequences of non-normality in general linear models (GLM) are minor (Faraway 2005). In a few cases, mass did not follow a normal distribution and the data were log10-transformed prior to analysis. To determine whether the period of captivity influenced the CTmax of slow-rate experimental individuals, GLMs were used to examine the effect of treatment (controls and field fresh groups), mass, sex and their interactions on CTmax. Similarly, GLMs were used to assess the effect of treatment (controls and field fresh groups), dry mass and treatment × dry mass interaction on lipid content. Typically, significant effects either had a small size in the case of CTmax (<1 °C) or laboratory individuals had somewhat higher lipid contents than those in the field (see Appendix S2, Tables S1 and S2). In consequence, experimental data were not further adjusted in any way for time in the laboratory.

To explore the rate hypothesis and its interaction with acclimation, the effects of rate of temperature change, acclimation, mass, sex and their interactions were assessed for each population using GLMs. Sex and mass were included in the models because both have been shown to have effects on tolerance in other species (Sprague 1963; Gaston & Spicer 1998). The E. laeviusculum populations were treated separately because of significantly different upper thermal tolerances among them (see Identification section in Appendix S1). Backward stepwise model simplification was used, and the minimum adequate models are presented (Crawley 2007).

To explore Janzen's rule (or at least the upper thermal limits component thereof), relationships between tolerance of thermal extremes (calculated as CTmax minus maximum habitat temperature, hereafter ‘tolerance of extremes’) and maximum habitat temperature (rather than latitude) were explored, for two rates of temperature change closest to those experienced by populations in the field (0·0007 and 0·1 °C min−1), using Pearson's product moment correlation implemented in r (Hmisc library). Thermal extremes were chosen here because they are becoming more frequent (Hansen, Sato & Ruedy 2012) and may be critically important for intertidal populations (Helmuth, Kingsolver & Carrington 2005). This approach was repeated using the more commonly adopted ‘warming tolerance’ (Deutsch et al. 2008), defined as the difference between CTmax and mean habitat temperature. Vernberg's rule assumes a positive relationship between latitude and extent of acclimation ability (with the exception of polar stenothermic marine organisms, Gaston et al. 2009), which we tested as a negative relationship between extent of acclimation (calculated as CTmax at the highest acclimation temperature minus CTmax at the lowest acclimation temperature) and mean habitat temperature, using the 0·1 °C min−1 rate of change data (as it lies between the maximum rates of change found for MI and RSA) and Pearson's product moment correlation. The Gause's rule presumes a negative relationship between basal tolerance and extent of plasticity, which was examined using the correlation between extent of acclimation and basal tolerance (assessed as the CTmax of individuals acclimated at mean habitat temperature, i.e. the intermediate acclimation treatment), using the 0·1 °C min−1 data.

Results

Rate of temperature change had a significant, positive effect on CTmax for all of the populations investigated (Table 2; Fig. 1) – the faster the rate of change, the higher the CTmax, with the difference among the fastest and slowest rates being c. 10 and 20 °C, in the MI and RSA populations, respectively (see Table S3 for population means). Acclimation had significant (except in H. grandicornis) effects that varied based on rate of temperature change (Table 2; Fig. 1). Typically, at fast rates of temperature change (0·1 and 0·5 °C min−1), acclimation to higher temperatures was related to an increase in CTmax, but these differences were not always significant (Fig. 1). At slow rates of temperature change (0·0001, 0·0002 and 0·0007 °C min−1), acclimation to higher temperatures either had no effect or resulted in a decline in CTmax (Fig. 1). Mass and sex generally had no significant effects on CTmax (Table 2). In the Hyale species, however, larger individuals had lower CTmax than smaller individuals (Table 2), whereas sex effects varied based on rate of temperature change (Hyale hirtipalma and E. laeviusculum from Lamberts Bay) or acclimation (E. laeviusculum populations) (Table 2). However, effect sizes were typically small, especially compared with the effects of rates of temperature change.

Table 2. Results of general linear models testing the effects of rate of temperature change, acclimation temperature, sex, mass and interactions on CTmax. Minimum adequate models are presented for each species
SpeciesCollection sited.f. F P
Exosphaeroma gigas MI
 Rate4283·52<0·001
 Acclimation22·750·066
 Rate × Acclimation913·75<0·001
Hyale hirtipalma MI
 Rate4207·69<0·001
 Acclimation21·500·22
 Sex10·110·74
 Mass115·94<0·001
 Rate × Acclimation88·58<0·001
 Rate × Sex42·670·032
H. grandicornis Muizenberg
 Rate4746·60<0·001
 Mass119·63<0·01
Exosphaeroma laeviusculum Hangklip
 Rate4780·71<0·001
 Acclimation22·290·104
 Sex13·850·051
 Rate × Acclimation94·513<0·001
 Acclimation × Sex32·9630·033
Lamberts Bay
 Rate4101·66<0·001
 Acclimation21·7910·169
 Sex122·83<0·001
 Rate × Acclimation99·235<0·001
 Rate × Sex45·226<0·001
Figure 1.

Effect of rate of temperature change and acclimation on the CTmax of Marion Island (MI) and South African species. (a) Exosphaeroma gigas (MI); (b) Hyale hirtipalma (MI); (c) H. grandicornis (South Africa); (d) Exosphaeroma laeviusculum (Hangklip, South Africa); and (e) E. laeviusculum (Lamberts Bay, South Africa). Box plots represent median ± 25th and 75th percentiles. FF denotes field fresh individuals. Numbers indicate sample sizes within each treatment. Differing upper and lower case letters indicate significant differences among rates of temperature change, and among acclimations within each rate of temperature change, respectively. The dashed line represents the maximum habitat temperature, and the arrow the maximum rate of temperature increase in an organism's habitat.

Microclimate data for the shore position at which the species were most typically found (Table 1) indicated that the maximum ambient rates of temperature increase lay between the slow and fast experimental rates of change for the MI populations, but within the fast experimental rates of change for the RSA populations. Similarly, maximum habitat temperatures lay well below all CTmax values for the MI populations, but lay between the CTmax at slow and fast rate values for the RSA populations (Fig. 1). Thus, at a rate of change of 0·1 °C min−1, populations from MI have a tolerance of thermal extremes (calculated as CTmax minus maximum habitat temperature) of c. 22·0–25·5 °C, while those from South Africa have a lower tolerance of extremes of c. 17·8–22·9 °C. At the slower rate of 0·0007 °C min−1, the difference between the temperate and sub-Antarctic populations is more significant in terms of a threshold effect. That is, the MI populations remain with substantial tolerance of extremes (c. 17·1–21·6 °C), whereas those from South Africa either have CTmax values approaching maximum habitat temperature (c. 2·5 °C), or lying well below it (by 14·5–16·2 °C; Fig. 1). These trends are consistent even if organisms use behaviour to buffer thermal extremes and move to submerged sites (Table S4).

At both rates of temperature change, negative relationships were found between tolerance of thermal extremes and maximum habitat temperature (Pearson: 0·1 °C min−1 r = −0·95, P = 0·0034; 0·0007 °C min−1 r = −0·89, P = 0·044), following Janzen's rule. For warming tolerance, the relationships were negative but not significant (Pearson: 0·1 °C min−1 r = −0·70, P = 0·12; 0·0007 °C min−1 r = −0·67, P = 0·21), although at the slowest rate of temperature change (0·0001 °C min−1), the relationship was significantly negative (Pearson: 0·0001 °C min−1 r = −0·95, P = 0·012). In addition, at the two slowest rates of change, the MI populations retain a CTmax greater than maximum habitat temperature, whereas none of the RSA populations do (Fig. 1). The relationship between phenotypic plasticity of CTmax (measured at 0·1 °C min−1) and mean habitat temperature was negative (Pearson: 0·1 °C min−1 r = −0·96, P = 0·002, Fig. 2), in keeping with Vernberg's rule. The relationships between basal tolerance and extent of acclimation were also negative and significant at 0·1 °C min−1 (Pearson: 0·1 °C min−1 r = −0·83, P = 0·043, Fig. 3) in accordance with Gause's rule.

Figure 2.

The extent of acclimation as the change in CTmax over an 8 °C acclimation range (ΔCTmax) against habitat temperature. ΔCTmax is the difference between mean CTmax when acclimated to low [3 and 11 °C for Marion Island (MI) and South African species, respectively] and high (11 and 19 °C for MI and South African species, respectively) temperatures. Data at the rate of temperature change of 0·1 °C min−1 were used. The equation for the fitted line is: y = −0·18x + 3·21; R2 = 0·92; = 0·002. ○ Exosphaeroma gigas (MI); □ Hyale hirtipalma (MI); ■ H. grandicornis (Muizenberg, South Africa); ▲ Exosphaeroma laeviusculum (Hangklip, South Africa); ♦ E. laeviusculum (Lamberts Bay, South Africa); and ● E. antikraussi (Hangklip, South Africa).

Figure 3.

The relationship between the extent of acclimation and basal thermal tolerance, measured at 0·1 °C min−1. Basal thermal tolerance corresponds to the tolerance of individuals acclimated to average habitat temperature. The equation for the fitted line is: y = −0·20x + 7·83; R2 = 0·68; = 0·043. ○ Exosphaeroma gigas (Marion Island); □ Hyale hirtipalma (Marion Island); ■ H. grandicornis (Muizenberg, South Africa); ▲ Exosphaeroma laeviusculum (Hangklip, South Africa); ♦ E. laeviusculum (Lamberts Bay, South Africa); and ● E. antikraussi (Hangklip, South Africa).

Discussion

A growing body of work, on terrestrial and marine species, is showing that rates of temperature change have pronounced impacts on thermal tolerance and estimates of their heritability (reviewed in Terblanche et al. 2011). For marine species, across a wide range of latitudes, slower rates of change result in lower physiological limits to high temperature (Peck et al. 2009; Nguyen et al. 2012; Richard et al. 2012). These differences can be profound, illustrated by the 10–20 °C differences in CTmax found among rate treatments in the populations investigated here. Clearly, when interpreting the results of temperature tolerance studies, the rate of warming needs to be taken into account. However, these data do not suggest that all previous estimates of upper thermal limits may be in error or that assessments of the consequences of ocean warming have necessarily been underestimated. Rather, they emphasize the importance of understanding organismal microclimates and the significance of behavioural regulation. That is, what rates of change are experienced (Helmuth, Kingsolver & Carrington 2005), and the extent to which sublethal or lethal conditions can be avoided (Huey 1991). Both are significant in the context of understanding the extent to which a given population might be capable of tolerating warming.

In the populations examined here, at realistic warming rates, that is, at the maximum or average rates currently experienced, the sub-Antarctic populations have considerable tolerance of thermal extremes, in the range of 8–20 °C, whereas those in the more equatorward, temperate South African locations, have very limited tolerance of extremes. Indeed, at the slowest rates of warming, their tolerance of habitat extremes seems to be greatly exceeded. In other words, if extreme temperatures are realized relatively slowly, as might be the case in buffered habitats, but have a long duration (so meaning high temperatures are encountered), tolerance may be substantially reduced. Such lack of tolerance of extremes is important given that extreme temperature conditions are increasing in frequency and duration (Meehl & Tebaldi 2004; Hansen, Sato & Ruedy 2012) and constitute significant sources of mortality for marine intertidal (Helmuth, Kingsolver & Carrington 2005), subtidal (Richard et al. 2012) and terrestrial species (see overview in Clusella-Trullas, Blackburn & Chown 2011).

The small difference between the CTmax and maximum habitat temperature in the RSA populations could potentially mean that population extirpation, as has been found for terrestrial lizards (Sinervo et al. 2010), has taken place. However, this need not necessarily have occurred. First, all of the experiments undertaken here remain short term, by comparison with realized rates of ocean warming (Domingues et al. 2008), and organisms were provided with no opportunity for behavioural regulation during trials. In consequence, the slower rates of change trials were both measuring the accumulation of thermal stress as a time-by-temperature interaction (see discussion in Hochachka & Somero 2002) and neglecting the potential for evolutionary change that might enable organisms to overcome thermal stress over the long term (Somero 2012). Secondly, in the field, individuals will also have the opportunity to move away from thermally stressful situations. However, this may expose them to poorer feeding situations or higher predation risk (Angilletta 2009), so lowering the likelihood of long-term population persistence. Finally, if mean habitat (microsite) temperatures are considered (as is typically done for warming tolerance – see Deutsch et al. 2008), then only at the slowest rates of warming are differences between the various sites sufficient to result in a significant negative relationship between warming tolerance and mean habitat temperature. By contrast, tolerance of thermal extremes was negatively related to maximum habitat temperature at all rates of warming. In other words, impacts might only be realized during extreme events. Clearly, understanding the relative contributions of extreme events relative to persistent sublethal stress is a significant area for future investigation.

Nonetheless, the key finding here is a substantial difference in the tolerance of extremes and warming tolerance, in thermal regimes appropriate to the organisms, between those populations from more equatorward locations compared with those from cooler regions, reflected by the negative relationship between maximum habitat temperature and warming tolerance. In other words, temperate populations (i.e. closer to the equator) appear to be more at risk from increasing temperature, because of lower warming tolerance of high temperatures, than are populations from cooler regions such as the sub-Antarctic.

Populations form the mainstay of biodiversity and are the focal level of evolutionary responses (Ricklefs 2008; Reed, Schindler & Waples 2011). Therefore, how these differences in tolerances and plasticity among regions will affect responses to climate change will depend on connectivity among populations of the species we examined and the overall variation in physiological limits found across the populations (see e.g. Chevin, Lande & Mace 2010; Bellard et al. 2012; Hoffmann, Chown & Clusella-Trullas 2013). Much of the information required to understand these processes is absent for these species. However, their biogeography and emerging work on related taxa provides a concerning perspective. The sub-Antarctic species, which show most warming tolerance, are not only widely distributed (Kenny & Haysom 1962; Brandt & Wagele 1989), but related species show substantial biogeographic connectivity owing to ongoing dispersal via drifting kelp (Nikula et al. 2010; Fraser, Nikula & Waters 2011). By contrast, although the warming intolerant South African taxa also show fairly wide distributions (Griffiths 1973; Kensley 1978), related species show very limited gene flow and substantial population structuring (Teske et al. 2006, 2007). Thus, both evolutionary rescue and recolonization are likely to be limited. In consequence, local extirpation risks may be compounded in something of a positive feedback loop (see Marshall et al. 2010).

In addition to reduced tolerance in the lower latitudes, South African populations also have a lower acclimation capability than do those at sub-Antarctic MI. Such a negative relationship between basal tolerance and phenotypic plasticity has also been recorded in Petrolisthes crabs along the west coast of North America (Stillman 2003). In consequence, it appears that this relationship, first formulated by Gause (see discussion in Gaston et al. 2009), may be a common feature of intertidal species.

In terms of the hypotheses or ‘rules’ we set out to examine, it is clear that the Gause, Janzen and Vernberg's ideas could not be rejected. That is, negative relationships were found between basal tolerance and plasticity, and between plasticity and habitat temperature. Thermal tolerance also shows substantial latitudinal variation, in keeping with a variation of Janzen's original idea (see Gaston et al. 2009). Thus, the responses of marine organisms to ocean warming may be somewhat less coherent and predictable than initial analyses (Sunday, Bates & Dulvy 2012) have suggested. In other words, substantial differences in long-term responses may be found among groups in different areas as a consequence of spatial variation in organismal traits and thermal environments, acclimation ability (or phenotypic plasticity) and evolutionary potential. Such an outcome does not imply that initial generalizations based on macrophysiological investigations should be dismissed. Rather, it emphasizes the significance of examining forecasts using a range of data and approaches so that their certainty can be established to inform key policy decisions in a spatially appropriate context.

Acknowledgements

We thank the Seapoint Research Aquarium, Cape Town, and Erika Nortje, Charlene Janion, Richard Viljoen and Kriek Bekker for logistic support. Neil Bruce and Charles Griffiths provided help with species identifications. This work was supported by the British Antarctic Survey Grant 41326, by National Research Foundation Grant SNA2007042400003 and by the South African National Antarctic Programme. The research was undertaken under permit RES2010/35&60 and RES2011/22, Department of Agriculture, Forestry and Fisheries, Republic of South Africa.

Ancillary