1. The geographical range sizes of individual species vary considerably in extent, although the factors underlying this variation remain poorly understood, and could include a number of ecological and evolutionary processes. A favoured explanation for range size variation is that this result from differences in fundamental niche breadths, suggesting a key role for physiology in determining range size, although to date empirical tests of these ideas remain limited.
2. Here we explore relationships between thermal physiology and biogeography, whilst controlling for possible differences in dispersal ability and phylogenetic relatedness, across 14 ecologically similar congeners which differ in geographical range extent; European diving beetles of the genus Deronectes Sharp (Coleoptera, Dytiscidae). Absolute upper and lower temperature tolerance and acclimatory abilities are determined for populations of each species, following acclimation in the laboratory.
3. Absolute thermal tolerance range is the best predictor of both species’ latitudinal range extent and position, differences in dispersal ability (based on wing size) apparently being less important in this group. In addition, species’ northern and southern range limits are related to their tolerance of low and high temperatures respectively. In all cases, absolute temperature tolerances, rather than acclimatory abilities are the best predictors of range parameters, whilst the use of independent contrasts suggested that species’ thermal acclimation abilities may also relate to biogeography, although increased acclimatory ability does not appear to be associated with increased range size.
4. Our study is the first to provide empirical support for a relationship between thermal physiology and range size variation in widespread and restricted species, conducted using the same experimental design, within a phylogenetically and ecologically controlled framework.
It has long been recognized that the geographical ranges of individual species vary enormously in extent, most taxa being endemic to relatively small areas, whilst comparatively few are widespread (Darwin 1859; Darlington 1957; MacArthur 1972; Gaston 1996, 2003). Within clades, species-range size distributions tend to be unimodal, with a strong right or positive skew, a pattern that appears to be almost universal, across a wide range of taxa and habitats (Gaston & Blackburn 2000; Gaston 2003). Despite our increasing knowledge of macroecological patterns, the factors determining the relative geographical range sizes of organisms are still poorly understood, with range size variation potentially resulting from a number of ecological and evolutionary processes (see Gaston 2003). These include interspecific differences in evolutionary age (e.g. Taylor & Gotelli 1994), population genetic structure and levels of gene flow (e.g. Kirkpatrick & Barton 1997; Gaston 2003), dispersal ability (DA) (e.g. Juliano 1983; Böhning-Gaese et al. 2006; Rundle et al. 2007a) and fundamental niche breadth (Brown 1984; Gaston & Spicer 2001). The last class of hypotheses proposes that species with broader fundamental niches will tend to achieve greater local densities, survive in more places, and so occupy wider geographical areas than narrow-niched relatives (Gaston & Spicer 2001) – the geographical range of a species being considered as a spatial reflection of its ecological niche (Lomolino, Riddle & Brown 2006).
The ecological impacts of recent climatic changes are becoming increasing well documented, and include apparent range expansions and shifts, genetic changes in natural populations and increased population extinction rates (e.g. Parmesan et al. 1999; Hill et al. 2002; Walther et al. 2002; Balanya et al. 2006). Despite this, there are few demonstrated examples where a species’ apparent response to climatic changes has been linked to a specific physiological or evolutionary mechanism (e.g. Crozier & Dwyer 2006; Pörtner & Knust 2007; Kearney et al. 2009). Attempts to elucidate the mechanisms behind range size variation need to consider both current and historical biogeography; restricted and widespread species are not evenly distributed across the globe, and species’ ranges are themselves dynamic. In the western Palaearctic, narrow-range endemic species are concentrated in lands around the Mediterranean, which are also widely recognized as having functioned as refugia for many temperate taxa during Quaternary glacial cycles (e.g. Taberlet et al. 1998; Hewitt 2004; Willis & Niklas 2004). The widespread species of present-day Europe have become widespread as a result of range expansion in the Holocene, and often belong to clades dominated by restricted southern endemics (e.g. Thompson 2005). In this context, why have some species been able to expand widely outside their Pleistocene refugia, whilst others remain largely restricted to these areas? Such questions apply not only to Europe, but to any areas where the extant ranges of organisms have been shaped by Quaternary climate shifts (Nilsson 1983), and may shed important light on the likely responses of widespread and restricted taxa to current climatic changes.
Here, we explore the relationships between thermal physiology and biogeography within a well-defined taxonomic assemblage of ecologically similar species; European diving beetles of the genus Deronectes Sharp (Coleoptera, Dytiscidae). Species of Deronectes occur in fast-flowing streams at intermediate elevations across the Palaearctic, with greatest diversity in the Mediterranean region (Franciscolo 1979; Millán & Ribera 2001; Ribera & Vogler 2004). The phylogenetic inter-relationships of European species are well documented based on mitochondrial cytochrome oxidase and 16s ribosomal DNA sequences (Ribera, Barraclough & Vogler 2001; Ribera & Vogler 2004). Most European species belong to a single clade, with the exception of the Deronectes latus group, which is more closely related to eastern Mediterranean taxa (Fery & Hosseinie 1998; Ribera et al. 2001). Deronectes species are generalist predators, feeding on a range of small aquatic invertebrates, and having similar general ecologies across their geographical ranges, although differing markedly in latitudinal range size. Whilst inhabiting a relatively thermally buffered aquatic medium (see Giller & Malmqvist 1998), in-stream temperatures experienced by these beetles vary both seasonally and latitudinally. In sites occupied by Deronectes in Andalucia, Spain, for example, water temperatures can vary between 6 and 30 °C annually (A. Millán, P. Abellán & D. Sánchez-Fernández, personal communication), whilst in southern England, temperatures in sites suitable for D. latus Stephens can fluctuate between 2 and 25 °C (D. T. Bilton, personal observation; Hildrew & Edington 1979). Widespread pan-European species of Deronectes experience greater macroclimatic variability than restricted southern endemics, and it is likely that widespread species also live under a greater range of microclimates, as species inhabit the same suite of microhabitats across their ranges (Sutton 1953).
We investigate the relationship between thermal physiology and latitudinal range extent and position amongst 14 species of Deronectes, whilst controlling for possible differences in dispersal ability, and the effects of phylogenetic relationship (Harvey & Pagel 1991; Garland, Bennett & Rezende 2005). In addition, we explore the relationships between species’ thermal biology and southern and northern range limits. Our study represents the most extensive empirical exploration to date of the relationship between thermal physiology and range size variation in widespread and restricted congeners, conducted using the same experimental design, examining relatively large numbers of individuals of both rare and common species, within a phylogenetically and ecologically controlled framework.
Materials and methods
Specimen collection, maintenance and preparation
Adult Deronectes were collected during spring and summer 2006 (see Appendix S1, Supporting Information) standardizing as much as possible for season of collection, working a D-Framed pond net (1 mm mesh, dimensions 20 × 25 cm) along stream banks. We studied adults as these are the life-history stage of longest duration (≥1 year), readily collected from the field, whilst larvae are short lived (c. 1–2 months), rarely collected due to their interstitial habits, and morphologically intractable. In addition, it is the adult stage which overwinters, and survives periodic droughts (Nilsson & Holmen 1995), making adult thermal tolerance most relevant here. All individuals collected were early post-teneral adults, minimizing any possible confounding effects due to age variation (Bowler & Terblanche 2008). All species were collected as close as possible to the central point of their latitudinal ranges, to avoid the possible confounding effects of local adaptation in range edge populations, and to ensure data were comparable for each species (Thompson et al. 1999). Given the largely allopatric occurrence of many species, and differences in the latitudinal position of their ranges, it is impossible to sample all taxa from the same latitude. Data on species’ geographical distributions were taken from Fery & Brancucci (1997) and Fery & Hosseinie (1998). Latitudinal range extents were calculated as the difference (in degrees latitude) between northern and southern distributional limits (Gaston 1994), and latitudinal range central points determined as the mid-point of each species’ latitudinal range extent (see Appendix S1, Supporting Information).
Thermal limits and acclimatory ability of Deronectes species
To define species’ thermal biology we employed upper and lower lethal thermal limits [defined as upper thermal limit (UTL) and lower thermal limit (LTL) subsequently], as these proved the most reliable, repeatable measure of thermal tolerance in diving beetles. Lethal limits were favoured amongst the various end-points which could be identified in thermal tolerance experiments, as they showed the lowest variance (see Lutterschmidt & Hutchison 1997a,b; Calosi et al. 2008a); however, the use of sublethal end-points (e.g. paralysis) did not change results.
Experiments commenced at the temperature to which individuals of a given subgroup had been acclimated. Thermal tolerance tests relied on a dynamic method, and were carried out in air in generic, 24-well (diameter = 12 mm, depth = 18 mm) plastic culture plates (Corning Ltd, Sunderland, UK), placed in a computer-controlled water bath (Grant LTC 6–30), heated and cooled, via a ramping program (±1 °C min−1) using the Grant Coolwise Software [Grant Instruments (Cambridge) Ltd, Herts, UK]. Experimental ramping rate and equilibration temperature can influence the outcome of thermal tolerance tests (Terblanche et al. 2007; Chown et al. 2009), and selecting an ecologically relevant ramping rate is difficult when comprehensive environmental data are lacking. Consequently, we employed an identical ramping rate, to allow comparisons amongst taxa, and with previous studies (Lutterschmidt & Hutchison 1997a,b).
Individuals were introduced, one per well, to a maximum of 12 individuals at any one time, with two investigators working together, for accurate determination of thermal tolerance limits. The actual temperature within each well was measured directly using a calibrated digital thermometer (Omega® HH11; Omega Engineering Inc., Stamford, CT, USA) equipped with an Omega®‘precision fine wire thermocouple’ (type K – dia./ga. 0·010 Teflon). Individuals were removed from their acclimation aquaria, quickly but carefully dried on absorbent paper and placed into a clean, dry, well. To avoid escape, well plates were covered with their plastic lids between addition of individuals. Once the experiment started, the lid was removed to allow full aeration and avoid the build-up of water vapour.
Thermal range (TR) was calculated as the difference between mean UTL at 20·5 °C and LTL at 14·5 °C, as these are likely to represent the most ecologically realistic measures of a species’ tolerance of high and low temperatures (as they follow high and low acclimation temperatures, respectively – see Calosi et al. 2008b– and, overall, species showed higher tolerance to heat and cold when acclimated at these temperatures). Upper and lower thermal tolerance acclimatory abilities (ΔUTL and ΔLTL) were estimated as the absolute difference between the thermal limits (for high or low temperatures respectively) measured at the two acclimation temperatures (Stillman 2003). A positive value for either ΔUTL or ΔLTL indicates a positive ability of a species to increase its mean UTL or mean LTL, following acclimation at a higher or lower temperature. After the experiments individuals were weighed (to ±0·001 g) using a Sartorius 1204 MP2 balance (Sartorius Ltd, Epsom, UK).
Obtaining accurate estimates of species’ relative dispersal ability (DA) is difficult (see Bilton, Freeland & Okamura 2001), as dispersal itself is an emergent trait, influenced by numerous morphological, physiological and ecological factors (Rundle, Bilton & Foggo 2007b). For aerial dispersers such as Deronectes species, however, wing size is an obvious feature which has been suggested to correlate with relative DA (e.g. Rundle et al. 2007a). This is particularly likely to hold in comparisons of closely related, otherwise similar, taxa, and the use of wing size as a surrogate of relative DA was adopted here. We specifically use wing length/body length ratio, as this is likely to provide a good comparative measure of the relative DA of diving beetle species (see Rundle et al. 2007a): other possible surrogates of DA (wing length, wing area/body mass ratio) were also explored, and gave the same results as those presented here. Individuals whose wings were to be examined were first photographed intact under a Leica MZ8 stereomicroscope (×50 mag) using a Nikon Coolpix 4500. The right wing was removed from 10 individuals of each species, digested in 10% potassium hydroxide for 15 min to increase flexibility, before being teased open and mounted in lactic acid solution (DL-lactic acid 85% w/w syrup – Sigma Chemical Co., St Louis, MO, USA) on a microscope slide. Disarticulated wings were examined and photographed as described above, and wing length estimated using UTHSCA Image Tool Version 3.0. Body length of each individual was measured from the front of the pronotum to the tip of the elytra (to avoid measurement error due to contraction of head into pronotum) using the same photomicroscopic approach.
Species’ body mass differed significantly amongst Deronectes (F13, 712 = 90·282; n = 726; P <0·0001), and was therefore considered as a covariate in subsequent analyses. The number of individuals studied ranged from 26 in Deronectes angusi Fery and Brancucci to 92 in D. hispanicus Rosenhauer. No significant correlation was found between the number of individuals of each species examined and any physiological, ecological or biogeographical trait (Pearson correlation minimum Z12 = 1·148; P =0·251), indicating that interspecific differences in sample size did not influence results. Mean body masses (±SE) and the number of individuals tested for each species are given in Appendix S2, Supporting Information.
Differences in mean UTL and mean LTL among species were first analysed separately using an ancova, including body mass as a covariate, whilst differences in mean DA among species were analysed using an ANOVA. We investigated factors influencing variation in latitudinal range extent and position, and both northern and southern range limits using a series of multiple regression models. Akaike’s Information Criteria (AIC) was used to select the best supported models, an approach which reduces problems associated with multiple testing and co-linearity of explanatory variables (Burnham & Anderson 2002). In each analysis, models were constructed using all combinations of experimental variables, and the five best models presented in the results. The single best supported model for each analysis was selected on the basis of the AIC weights, calculated to evaluate the relative likelihood of a model, given the data and the fitted model, scaled to one (Burnham & Anderson 2002). Model selection was performed using both raw data and independent contrasts (Felsenstein 1985) derived from mtDNA based phylogenies (Ribera et al. 2001; Ribera & Vogler 2004). Contrasts were produced using the CRUNCH algorithm of the CAIC software package (Purvis & Rambaut 1995), and regressions of contrast scores forced through the origin (Garland, Harvey & Ives 1992).
Species’ southern range limits were normally distributed, whilst latitudinal range extent and central point were normalized following log10 transformation. In the case of northern range limits, data were normalized following double log10 transformation. Normality in all cases was assessed via Shapiro–Wilks test; P >0·05. All statistical analyses were conducted using JMP IN® version 5.1, except for multiple regression models, which were run in R v.2.5.1 (R Development Core Team, 2007) and SPSS v.15.0.
Thermal limits and acclimatory abilities of Deronectes species
Upper and lower thermal limits differed significantly between species of Deronectes at all acclimations tested (ANCOVA minimum F13,173 = 8·797; P <0·0001 – see Appendix S2, Supporting Information for species’ values). Mean body masses ranged between 5·37 mg in D. platynotus platynotus Germar and 10·54 mg in D. opatrinus Germar (see Appendix S2, Supporting Information), but did not affect species’ ability to tolerate either high or low temperatures (ANCOVA maximum F13,160 = 0·623; P =0·431). Given that body mass did not influence thermal tolerance, we explored differences between species’ performances (excluding body mass from the analysis) via one-way ANOVAs, which showed that species differ significantly in both UTL and LTL (minimum F13,173 = 13·125; P <0·0001). ANOVAs were employed in conjunction with Tukey–Kramer tests. These analyses were carried out separately on UTL at 20·5 °C and LTL at 14·5 °C (see Fig. 1), focusing on these acclimations as these values are used in further analyses (see Methods).
Mean UTL ranged from 42·6 °C in D. semirufus Germar to 46·9 °C in D. latus, both following 20·5 °C acclimation (Appendix S2, Supporting Information; Fig. 1a), and mean LTL ranged from −3·4 °C in D. algibensis Fery and Fresneda, to −10·0 °C in D. latus, both following acclimation at 14·5 °C (Appendix S2, Supporting Information; Fig. 1b). Furthermore, phylogenetically independent contrasts reveal that the ability to tolerate heat and cold are significantly negatively correlated across the genus (Pearson correlation Z11 = 3·779; P =0·0004). Mean ΔUTL ranged from −1·13 °C in D. wewalkai Fery and Fresneda to 2·02 °C D. latus, whilst mean ΔLTL ranged from −1·99 °C in D. opatrinus to 1·42 °C in D. platynotus mazzoldi Fery and Brancucci (see Appendix S2, Supporting Information). Phylogenetically independent contrasts reveal that the ability to acclimate to heat and cold are significantly positively correlated across the genus (Pearson correlation Z11 = 7·679; P <0·000001). Wing length/body length ratio differed significantly amongst taxa (ANOVA F13,140 = 74·527; P <0·0001), ranging from 1·00 in D. angusi to 1·30 in D. fairmairei Leprieur (see Appendix S2, Supporting Information).
Thermal biology and range size and position
The best supported models examining range size and position always contained TR, which was the only parameter to be significant in all cases (Table 1). Model evidence suggested that latitudinal range extent was strongly positively related to a species’ TR, and this was the only parameter included in the best supported model (TR slope = 0·178, SE = 0·061; Fig. 2a). After controlling for phylogeny, the best predictor for latitudinal range extent was still TR; additionally in some models range size was found to be significantly negatively related to ΔLTL and/or positively to DA. In one case, a significant positive relationship between latitudinal range extent and ΔUTL was suggested, although this model had relatively low support (see Table 1). The best selected model included all parameters considered, however only TR (slope = 0·281, SE = 0·012) and ΔLTL (slope = −0·276, SE = 0·104), had significant slopes.
Table 1. Model selection to estimate factors influencing latitudinal range extent and latitudinal range central point in Deronectes species
In each case thermal range (TR), acclimatory ability of upper and lower thermal tolerances (ΔUTL & ΔLTL respectively), body mass (BM) and dispersal ability (DA) (measured as mean wing length/body length ratio) were included as independent variables. np = number of parameters; AIC = Akaike’s Information Criteria. AIC weight represents the likelihood of the model given the data. Parameters with a significant slope are highlighted in bold.
Latitudinal range extent
TR + BM
TR + ΔLTL
TR + DA
TR + ΔUTL
Latitudinal range extent (independent contrasts)
TR + ΔLTL + ΔUTL + BM + DA
TR + ΔLTL + BM + DA
TR + ΔLTL + DA
TR + ΔLTL + ΔUTL + DA
TR + ΔLTL + ΔUTL
Latitudinal range central point
TR + DA
TR + ΔLTL + ΔUTL + BM
TR + ΔUTL + DA
TR + ΔLTL + ΔUTL + BM + DA
Latitudinal range central point (independent contrasts)
TR + ΔLTL + ΔUTL + BM + DA
TR + ΔLTL + ΔUTL + BM
TR + ΔLTL + ΔUTL
TR + ΔLTL + BM
TR + ΔLTL + ΔUTL+DA
Latitudinal range central point was always significantly positively related to TR, species with more northerly ranges possessing broader TRs, whilst in one model latitudinal range central point was also negatively related to ΔLTL (Table 1). The best supported model included TR and DA, but only TR had a significant slope (TR slope = 0·016, SE = 0·005; Fig. 2b). After controlling for phylogeny, TR again emerged as the strongest predictor, being positively related to range position in all the best selected models (Table 1). In some models range position also appeared to be significantly negatively related to ΔLTL, and in three cases to ΔUTL (positively) and in one case to body mass (negatively), although most of these models had relatively weak support. The best supported model contained all parameters considered, but only TR (slope = 0·022, SE = 0·001), ΔLTL (slope = −0·019, SE = 0·007) and body mass (slope = −0·005, SE = 0·002) had a significant slope.
Thermal biology and range limits
All best supported models for southern range limit contained UTL, which was the only significant parameter in all cases (Table 2). In general, species of Deronectes with greater tolerance to high temperatures showed lower southern range limits. The best supported model explaining variation in southern range limit included UTL, ΔUTL and DA (Table 2), but only UTL showed a significant slope (UTL slope = −2·645, SE = 0·947; Fig. 3a). Following phylogenetic transformation both UTL and ΔUTL appeared as strong predictors of southern range limit. Again, species with higher UTL showed lower southern limits, whilst species with higher ΔUTL had higher southern limits. The best selected model included all parameters, but only UTL and ΔUTL had significant slopes (Table 2; UTL slope = −0·956, SE = 0·261; ΔUTL slope = 0·503, SE = 0·174).
Table 2. Model selection to estimate factors influencing southern and northern range boundaries in Deronectes species
In each case both absolute tolerance and acclimatory ability were included as independent variables (UTL and ΔUTL or LTL and ΔLTL respectively), together with body mass (BM) and dispersal ability (DA) (measured as mean wing length/body length ratio). np = number of parameters; AIC = Akaike’s Information Criteria. AIC weight represents the likelihood of the model given the data. Parameters with a significant slope are highlighted in bold.
UTL + ΔUTL + DA
UTL + ΔUTL + BM + DA
UTL + DA
UTL + ΔUTL
Southern limit (independent contrasts)
UTL + ΔUTL + BM + DA
UTL + ΔUTL + BM
UTL + ΔUTL + DA
UTL + ΔUTL
UTL + DA
LTL + BM
LTL + ΔLTL
LTL + DA
Northern limit (independent contrasts)
LTL + ΔLTL + BM
LTL + ΔLTL + BM + DA
LTL + ΔLTL + DA
LTL + ΔLTL
Finally, species with the greatest tolerance to low temperatures had the highest northern range limits. LTL was always significant if present (Table 2), and was the only parameter included in the best selected model, being negatively related to northern range limit (LTL slope = −6·661, SE = 1·593; Fig. 3b). Although this may appear counter intuitive, LTL values are negative, indicating that those species which are most tolerant to cold possess the highest northern limits. Following phylogenetic correction, LTL again emerged as the strongest predictor of northern range limits. Additionally, in most models, ΔLTL appeared to be significantly negatively related to northern range limits, as was body mass in two cases. Indeed the best selected model included LTL (slope = −9·582, SE = 0·776), ΔLTL (slope = −5·473, SE = 0·659) and body mass (slope = −7·753, SE = 1·235).
Thermal tolerance range represents an integrative estimate of absolute thermal niche breadth in Deronectes, and appears to be the best predictor of both their latitudinal range extent and position. In Deronectes, widespread, more northerly distributed species possessed broader thermal tolerance windows than their restricted, more southern relatives. Although broader thermal tolerance windows may have shaped range size evolution in these taxa, or vice versa, we favour the former explanation here, particularly as our data are based on studies of single, central populations. Local adaptation may lead to greater inter-population variation in thermal physiology in a species with a large geographical range, and as a consequence, some inter-populational differences in thermal physiology may be expected, particularly for the most widespread species of Deronectes, as has been reported in some other insect species (e.g. Garland & Adolph 1991; Huey, Partridge & Fowler 1991; Hoffmann & Watson 1993; Klok & Chown 2003; Terblanche et al. 2006). Whilst gene flow could result in broader thermal tolerance windows within central populations of such species, through the admixture of locally adapted genotypes, we feel this is unlikely to be the case here. Species of Deronectes are patchily distributed in suitable stream systems throughout their ranges, and, as with most lotic Coleoptera, show relatively strong spatial genetic structure (e.g. Ribera et al. 2001; Ribera, Bilton & Vogler 2003), suggesting gene flow levels are relatively low in these taxa. Within the genus, more northerly distributed species are also those which have the largest geographical ranges; latitudinal range extent and latitudinal range central point being positively related in Deronectes (Pearson correlation analyses: untransformed data Z12 = 2·275; P <0·0000001; phylogenetically independent contrasts Z11 = 59·182; P =0·023). These findings lend some support to Stevens (1989) ideas regarding the factors underlying the relationship between latitudinal range extent and position.
In conventional multiple regression models, species’ acclimatory abilities were not significant predictors of latitudinal range extent, and therefore results do not support Gaston & Spicer’s (2001) prediction that widespread species should have higher acclimation abilities than their restricted relatives. In many insects, acclimatory responses across temperature gradients are known to be nonlinear (e.g. Terblanche et al. 2006), and as a consequence results such as ours, which examine acclimation to the same two temperatures across a range of taxa, should be interpreted with some caution, as individual species may differ in the thermal windows over which they can effectively acclimate. It also should be acknowledged that the length of the experimental exposure window can influence absolute estimates of critical thermal limits (Terblanche et al. 2007; Chown et al. 2009). Assessments of acclimatory ability which rely on different experimental starting temperatures, as employed here, may be influenced by these effects, although such issues are unlikely to alter conclusions made from broad scale interspecific comparisons (Terblanche et al. 2007). In general, Deronectes species appeared to have limited acclimatory ability of either UTL or LTL, at least over the range of temperatures employed in our study and using the current experimental protocol (Calosi et al. 2008b). Analyses of independent contrasts data revealed an apparent negative relationship between ΔLTL and both latitudinal range extent and latitudinal range central point in Deronectes, suggesting that species with large, northerly ranges have poor acclimatory abilities at low temperatures. Whilst such a result may point to an evolutionary trade-off between thermal limits and their acclimation abilities (Janzen 1967; Stillman 2003; Somero 2005; Tewksbury, Huey & Deutsch 2008 and references therein), no evidence of this has been found with LTLs in Deronectes (see Chown & Terblanche 2007; Calosi et al. 2008b).
On the other hand, the fact that the ability to tolerate heat and cold are significantly negatively correlated across the genus may suggest a trade-off in the evolution of high and low temperature tolerance. Intraspecific differences in upper and lower thermal limits may not be uncommon in insects, as the physiological mechanisms underpinning upper and lower thermal tolerances are decoupled (e.g. Hoffmann et al. 2002; Chown & Nicolson 2004; Klok, Sinclair & Chown 2004; Terblanche et al. 2005), and apparent evolutionary trade-offs in thermal adaptation have been reported elsewhere (e.g. Bennett & Lenski 1993; Stillman 2003; Dixon et al. 2009; but see also Huey & Kingsolver 1993), and may be important in setting the limits of species’ distributions (see Pörtner et al. 2006; Chown & Terblanche 2007 and references therein). In our study, the most widespread species, D. latus, possesses the greatest tolerance of both high and low temperatures, contrary to the trend observed above. It seems likely that the wide thermal window of this species has been important in shaping its distribution, especially when one compares the performance of latus with its very close relative D. angusi (see Ribera & Vogler 2004), which is restricted to northern Iberia, and shows more limited tolerance of both high and low thermal extremes.
Possible differences in relative DA do not appear to be strongly related to biogeographical parameters in our analyses. Whilst DA appears in a number of our best supported models (see Tables 1 and 2), it only has a significant slope in relation to latitudinal range extent, and then with a relatively high standard error (see Results). Such findings contrast with a number of studies which have suggested a primary role for dispersal in shaping the range sizes of species within clades, across a number of taxa (Juliano 1983; Böhning-Gaese et al. 2006; Rundle et al. 2007a). Whilst we have no information on actual dispersal distances in these beetles, available data point to a limited role of interspecific differences in DA in shaping biogeography, particularly given the strong relationship with thermal biology demonstrated here. Species of Deronectes may simply be too similar in their relative dispersal abilities for such differences to be significant in shaping their potential ranges.
Mean UTL of Deronectes species ranged between 42·6 and 46·9 °C, a total of 4·3 °C, whilst mean LTL ranged over 6·6 °C, from −3·4 to −10·0 °C. In both cases, lethal limits fall well within the temperature interval reported for other temperate insects living at similar latitudes (Addo-Bediako et al. 2000; Chown & Nicolson 2004). However, compared to many temperate insects (see Addo-Bediako et al. 2000; Chown & Nicolson 2004) it would appear that Deronectes are not particularly cold hardy, something also observed in Agabus species (see Calosi et al. 2008b), which possess similar thermal limits, despite being up to eight times heavier. In both conventional and independent contrast regression models, absolute heat and cold tolerance in these insects appear to be related to the position of their southern and northern range limits respectively. Such findings are in general agreement with data from a range of taxa (e.g. Addo-Bediako et al. 2000; Ghalambor et al. 2006) but contrast with Huey et al. (2009), who found that low latitude forest lizards had the lowest CTMax values. Also in Deronectes, the use of independent contrasts suggests additional relationships between acclimatory abilities and body mass and range boundaries. Low-latitude southern range boundaries are apparently associated with a weaker ability to acclimate to high temperatures, whilst higher latitude northern boundaries are associated with poorer acclimation to cold, and smaller body size. As with range size and position, whilst such findings may point to evolutionary trade-offs between thermal limits and their acclimation abilities, this is apparently not the case in Deronectes (see Calosi et al. 2008b), making interpretation of these results difficult in the absence of further data on the mechanisms underlying thermal tolerance and acclimation in these beetles. The relationships between range boundaries and temperature tolerances occur despite the fact that in the Mediterranean region species may be expected to have ‘hard’ southern range boundaries, defined by barriers to dispersal for temperate freshwater taxa, such as the northern shore of the Mediterranean sea itself, or for trans-Mediterranean species, the Sahara (see Calosi et al. 2008a). Both summer air maxima and winter minima can exceed the lethal limits recorded here for Deronectes. For example, maximum recorded summer temperatures (from aerial weather stations) along the latitudinal gradient occupied by the species studied range from 34·5 °C (Jokkmokk, Sweden – 66°N; 257 m) to 48·1 °C (Marrakech, Morocco – 31°30′ N; 470 m) (Meteorological Office 1972, 1983). Comparably during winter, air temperatures in areas occupied by these species regularly drop well below their lethal limits, minimum recorded temperatures ranging from −10·8 °C (Bouârfa, Morocco – 32°N; 1310 m) to −43·4 °C (Stensele, Sweden – 65°N; 330 m) (Meteorological Office 1972, 1983). The direct relevance of extreme air temperatures in the ecology of these diving beetles may be limited, however, as these animals spend almost all of their life histories in a thermally buffered aquatic medium, only pupation occurring exclusively in air, and this taking place in bankside refugia, which are themselves thermally buffered (P. Abellán, A. Millán and D. Sánchez-Fernández personal communication). Additionally here, whilst our data broadly support Addo-Bediako et al. (2000) on the extent of variability of UTL and LTL with latitude, unusually for temperate insects (see Kingsolver & Huey 1998; Addo-Bediako et al. 2000; Goto, Kitamura & Kimura 2000; Klok & Chown 2003; Terblanche et al. 2005) and other ectotherms (Stillman 2002; Ghalambor et al. 2006; Deutsch et al. 2008), Deronectes species show only slightly higher interspecific variability in LTL compared to that observed for UTL. Again this may result from the fact that these insects live in a relatively thermally buffered environment, overwintering in water, where in-stream temperatures fluctuate 4–5 fold less than those observed in surrounding air, both daily and seasonally [southern Spain (A. Millán, P. Abellán and D. Sánchez-Fernández, personal communication), North-West Italy (S. Fenoglio and T. Bo, personal communication) and Australia (B.J. Robson, personal communication)]. In summer, the occurrence of occasional droughts, and the consequent lack of in-stream thermal protection, may have selected for heat tolerance, and it is interesting to note that species which live in the most drought-prone regions are amongst the most tolerant to high temperatures, including D. fairmairei, D. algibensis, D. opatrinus, D. hispanicus and D. moestus Fairmaire.
Given that widespread species of Deronectes are those that have been most successful in expanding their ranges outside Pleistocene refugial areas, our results suggest that thermal tolerance has influenced post-glacial colonization success. The genus has its centre of diversity in the Mediterranean region (Fery & Brancucci 1997; Fery & Hosseinie 1998), where many of the extant species seem to have evolved during the Pleistocene (Ribera et al. 2001; Ribera & Vogler 2004). Good post-glacial colonists seems to have been able to retain southern European populations whilst they expanded north: the most successful species of all in this regard being D. latus which has retained populations in the southern Balkans, at the same time as expanding as far as northern Scandinavia in the Holocene. Other species may be genuinely restricted by thermal physiology. Deronectes fairmairei, D. hispanicus and D. opatrinus, for example, all have relatively high mean UTLs, and are widespread in the western Mediterranean, but have poor cold tolerance, and reach their northern range limits at relatively low latitudes. Deronectes algibensis, the most narrowly endemic species of the genus in Europe, and also the one with the most southerly centered range has one of the highest mean UTLs in the genus, but has the least developed tolerance of low temperatures. Some other endemics may actually be trapped in southern refugia due to poor heat tolerance. Deronectes angusi and D. wewalkai, for example, are both relatively tolerant of low temperatures, but may have remained restricted to their southern European mountain ranges by a physiological inability to tolerate heat, and thus cross intervening lowland areas such as the Ebro valley, which may form effective barriers to their dispersal (see Ribera 2003). Such species may be particularly at risk in the face of ongoing and future climate changes, our results suggesting that restricted endemics tend to be more vulnerable in this regard than their widespread relatives. Given the added effect of climatic warming on habitat availability in southern mountains (Lovejoy & Hannah 2005), these taxa may be doubly at risk in the future. In contrast, more widespread species, which are relatively tolerant of high temperatures (such as D. latus, D. fairmairei, D. moestus, D. opatrinus and D. hispanicus) may have a better ability to retain their current geographical ranges in the face of ongoing global climate change.
We thank Pedro Abellán, Mark Briffa, Tiziano Bo, Stefano Fenoglio, Hans Fery, Andy Foggo, Garth Foster, Andres Millán, Ignacio Ribera, Brenda Robson, Simon Rundle, David Sánchez-Fernández, Jaroslav Š?astný, Richard Ticehurst, Ann Torr and Jon Webb for their assistance, help and support. We thank John S. Terblanche, Ray B. Huey, and two other anonymous reviewers, for their useful comments and constructive criticisms. In addition, we are grateful to Ted Garland for advice on phylogenetic analysis of comparative data. Finally, we thank Hans Fery, Shidi Hosseinie and Michael Brancucci for their outstanding systematic work on Deronectes– without which this study would have been impossible. This investigation was supported by an award from the Leverhulme Trust to DTB and JIS.