Higher temperatures lower rates of physiological and niche evolution

Abstract

Understanding rates and patterns of change in physiological and climatic-niche variables is of urgent importance as many species are increasingly threatened by rising global temperatures. Here, we broadly test several fundamental hypotheses about physiological and niche evolution for the first time (with appropriate phylogenetic methods), using published data from 2059 vertebrate species. Our main results show that: (i) physiological tolerances to heat evolve more slowly than those to cold, (ii) the hottest climatic-niche temperatures change more slowly than the coldest climatic-niche temperatures, and (iii) physiological tolerances to heat and cold evolve more slowly than the corresponding climatic-niche variables. Physiological tolerances are significantly and positively related to the corresponding climatic-niche variables, but species often occur in climates outside the range of these tolerances. However, mismatches between climate and physiology do not necessarily mean that the climatic-niche data are misleading. Instead, some standard physiological variables used in vertebrates (i.e. critical thermal maxima and minima) may reflect when species are active (daily, seasonally) and their local-scale microhabitats (sun versus shade), rather than their large-scale climatic distributions.

1. Introduction

As global temperatures rise, the survival of many species may hinge on their ability to rapidly shift their realized climatic niches to accommodate these new climatic conditions [1–5]. These niche shifts may occur through plastic changes (e.g. behavioural avoidance of high temperatures) or evolutionary change [6–8]. A crucial trait that may need to evolve rapidly to allow species to persist in warmer climates is their physiological tolerance to high temperatures [5]. Therefore, it is urgently important to understand evolutionary rates in physiological tolerances, rates of climatic-niche change, and the interrelationships between climatic niches and physiology.

Some papers have now addressed this intersection of physiological and climatic-niche evolution. However, most have not done so using appropriate phylogenetic methods. For example, in an influential and well-cited paper, Araújo et al. [9] suggested that physiological tolerances to heat evolve more slowly than tolerances to cold (e.g. due to different constraints on physiological processes at very high and low temperatures). They reviewed physiological data from ectotherms for critical thermal minima (CT_min) and maxima (CT_max), and similar variables for endotherms (upper and lower thermal neutral zones; UT_max and UT_min). They concluded that ‘heat freezes niche evolution’ (i.e. lowers the evolutionary rate), based on less variability among species in upper thermal tolerances (CT_max/UT_max) than in lower thermal tolerances (CT_min/UT_min). However, they did not infer rates of physiological or climatic-niche evolution. They also suggested that data on species' realized climatic niches may give misleading inferences about topics related to climatic niches (e.g. responses to climate change, niche evolution, invasive species), based on differences between climatic and physiological data. Grigg & Buckley [10] found greater phylogenetic signal in CT_max than CT_min in lizards. However, signal is not an estimate of rates for continuous variables [11]. Similarly, Hoffmann et al. [12] inferred greater constraints on upper than lower thermal limits in insects and squamate reptiles, again based on phylogenetic signal. Muñoz et al. [13] used phylogenetic methods [14] to directly compare evolutionary rates in upper and lower physiological tolerances among seven Anolis lizard species. They found lower rates for CT_max than CT_min.

We suggest that several important questions about the evolution of physiological tolerances and climatic niches should be broadly addressed using appropriate phylogenetic approaches. (1) Do physiological tolerances to heat (CT_max) evolve more slowly than physiological tolerances to cold (CT_min), as suggested in these previous studies? (2) Does heat freeze niche evolution? For example, do the hottest temperatures in species’ realized climatic niches change more slowly than the coldest temperatures, possibly because of these asymmetric rates in physiological tolerances? (3) Does physiological evolution constrain climatic-niche evolution? Three subquestions may help resolve this. (a) Do physiological tolerances to extreme temperatures (CT_max, CT_min) evolve more slowly than the corresponding climatic-niche temperatures (hottest annual, coldest annual)? We expect lower rates in these physiological variables if their slow rates limit (constrain) rates of climatic-niche evolution. (b) Second, are there strong relationships among species between physiological tolerances to extreme heat and cold and corresponding niche temperatures? This was addressed by Araújo et al. [9], but without incorporating phylogeny. They concluded that physiological tolerances to cold were correlated with minimum niche temperatures, but tolerances to heat tended to be uncorrelated with maximum temperatures. (c) Do these physiological variables seem to limit where species can occur climatically? Specifically, do species occur only where maximum temperatures are less than their physiological tolerances to heat, and where minimum temperatures exceed their physiological tolerances to cold? (4) How are body temperatures related to physiological tolerances and climatic temperatures? For example, does variation in body temperatures across environments lower evolutionary rates in physiological tolerances by buffering individuals from extreme temperatures (e.g. [13,15])? Based on this idea, we might expect strong relationships between body temperatures and climatic variables and lower evolutionary rates in physiological tolerances than body temperatures. Alternatively, different species might maintain similar body temperatures across diverse environments, yielding weak relationships and similarly low rates in both body temperatures and physiological tolerances. To our knowledge, most of these questions have not been directly addressed before, or only at limited taxonomic scales. However, a recent study [16] found faster rates of niche evolution for coldest temperatures than hottest temperatures (in plants and vertebrates) but did not examine physiological variables.

Here, we address these questions using phylogenetic methods and physiological and climatic data from tetrapods. First, we compile physiological and climatic data from the literature for hundreds of amphibians, squamates, birds and mammals. We assemble time-calibrated phylogenies for each group (amphibians [17]; squamates [18]; birds [19]; mammals [20]). We then test our hypotheses using phylogeny-based methods designed to test for significant differences in evolutionary rates between variables [14] and infer relationships between variables [21]. Our results show lower evolutionary rates for physiological tolerances and climatic-niche variables associated with maximum temperatures, that climatic variables change more quickly than physiological variables, and climatic variables and physiological tolerances are significantly related.

2. Material and methods

(a). Physiological data

Physiological data for each species and corresponding references are given in electronic supplementary material, table S1. Body temperatures (T_b) reflect behavioural thermoregulation and available environmental temperatures (review in [22]). Data on T_b were for active animals in the field. Thermal tolerance limits in ectothermic vertebrates are typically estimated based on body temperatures associated with the loss of righting responses at upper and lower thermal limits [23,24]. These are referred to as critical thermal maxima (CT_max) and minima (CT_min). Endotherms can maintain a high and relatively constant active body temperature (T_a) through metabolic heat production (e.g. [25]). Birds and mammals with a broad thermoneutral zone can maintain their basal metabolism over a wide range of temperatures. Metabolic rates rise when environmental temperatures are above (UT_max) or below (UT_min) the thermoneutral zone [25]. Data were obtained on T_b (amphibians, squamates), T_a (birds, mammals), CT_min and CT_max (amphibians, squamates), and UT_min and UT_max (birds, mammals). We recognize that the physiological variables in ectotherms and endotherms are not strictly equivalent, and we do not compare them across groups. However, we treat them as conceptually similar (following [9]).

For amphibians, we obtained physiological data from literature sources [26–28]. We checked species names [29] and found 30 synonymous species. We deleted species with the same data under different names, and averaged values across conspecific taxa. We obtained data from 222 species, but removed 40 species absent from the tree, leaving physiological data from 182 species for one or more variables.

For squamates, we assembled published physiological data [3,10,27,28,30]. In one study [3], most T_b were from multiple populations, and we used these values when possible. We averaged T_b values across populations to obtain a single value for each species. Again, we checked species names (using [31]), found 80 synonyms, and renamed taxa and combined data from synonymous species. CT_min values from Brattstrom [30] were unusually low for 16 species. Specifically, values were less than 0°C (including species known to be active at high temperatures), whereas CT_min values from confamilial species were much higher. We therefore deleted CT_min data for these 16 species, as have other authors [27]. We retained a very low CT_min value (−3.7°C) from Sceloporus jarrovii, a species active in winter at high elevations [32]. We also removed 90 species absent from the tree, leaving 620 species. We also included the rhynchocephalian Sphenodon punctatus, but use ‘squamates’ for convenience.

We assembled bird data from various sources [25,33,34]. We preferentially used T_a data from Clarke & Rothery [34], who averaged values between sexes. We removed potentially erroneous UT_max values from 102 species [35]. We checked all species names [36] and deleted or combined 42 synonyms. We removed 42 species absent from the tree, leaving data for 618.

We summarized mammal data from various sources [25,28,34,37]. We renamed or combined 96 synonymous species [38], and removed 4 species absent from the tree. We removed UT_max data from 189 species (following [35]), leaving data for 628 species.

(b). Climatic data

Climatic data for each species are given in electronic supplementary material, table S2. We used three climatic variables: annual mean temperature (Bio1), maximum annual temperature (Bio5) and minimum annual temperature (Bio6). Most climatic data were from past studies of squamates [39], birds [40] and mammals [41]. These studies used the WorldClim dataset [42], using mean values across grid cells from range maps for each species and variable. For amphibians, we obtained range maps from IUCN [43] then used QGIS [44] to extract mean values from grid cells (from [42]) at 30″ spatial resolution.

There were also 313 squamate species, 130 birds and 36 mammals that lacked data in these datasets. We obtained distribution maps for birds [36], and obtained climatic data as described above (except for two, deleted Antarctic species, Aptenodytes forsteri and A. patagonicus). We obtained georeferenced locality data for squamate and mammal species from GBIF.org. We confirmed that sampled localities spanned each species' range and were not outside it (using [31,43]). We calculated mean climatic values across localities using QGIS and the WorldClim dataset. However, seven squamates had less than five georeferenced localities and were deleted.

(c). Phylogenies

The tree with maximum taxon sampling for each group is given in electronic supplementary material, datafiles S1–S4. We used Mesquite 3.04 [45] to prune trees to include only species with relevant data for each analysis. For birds, we used the consensus tree (containing all species) from the posterior distribution of 10 000 trees (from [19]; backbone tree from [46]).

We chose phylogenies that were time-calibrated (necessary for estimating rates) and had extensive species sampling. For some groups, phylogenies with less taxon sampling but better gene sampling suggest some alternative higher-level relationships (e.g. [47,48]). However, rate estimates should depend more on relationships and divergence times among closely related species, not higher-level relationships.

We also performed analyses using an alternative phylogeny for each group. These alternative trees (electronic supplementary material, Datafiles S5–S8) and results are described in electronic supplementary material, appendix S1. We also provide further justification there for the choice of trees used in the main analyses.

(d). Data analyses

For each group, we separately compared evolutionary rates in physiological tolerances to heat (CT_max/UT_max) and cold (CT_min/UT_min), body temperatures (T_b/T_a), and climatic-niche variables associated with extreme heat (Bio5) and cold (Bio6). We compared maximum-likelihood estimates of evolutionary rates (σ²) from the Brownian motion model [14]. For each pair of variables for each group, we performed a likelihood-ratio test (LRT) comparing the likelihood for the different, observed rates for each variable and a model with constrained rates (equal between variables). Tests were performed using R code [14] in R 3.3.0 [49]. Rate comparisons were only made when comparing identical sets of species (i.e. with data for both variables).

We used phylogenetic generalized least-squares regression (PGLS; [21]) to test relationships between physiological and climatic variables. PGLS was performed in R using caper v.1.0.1 [50]. We used phylogenetic paired t-tests to evaluate whether climatic-niche variables (Bio5/Bio6) were significantly outside the range of physiological variables (CT_max/CT_min, UT_max/UT_min), using the R package phytools v.0.6–60 [51]. We applied a sequential Bonferroni correction [52,53] to tables of results with p-values.

Following most authors, we use ‘niche evolution’ to describe rate estimates for climatic-niche variables. However, we recognize that changes in the realized climatic niche among species can also reflect non-evolutionary processes, such as plasticity, extrinsic biotic factors and non-climatic abiotic factors.

(e). Methodological caveats

Here, we address several potential concerns about the data and methods used in our study. In general, there are many sources of error, but we were still able to find significant results in most groups. Furthermore, many limitations here are shared with earlier studies.

First, our taxon sampling is limited within each group, since relatively few species have physiological data. Nevertheless, our sampling is generally improved over previous studies. The least data were available for comparing CT_min/UT_min to CT_max/UT_max. Araújo et al. [9] had data for 26 amphibians, whereas we have 30. We also have more squamates (135/147) and birds (70/75). We have fewer mammals (133/321), but many mammalian data in previous studies [9] were problematic [35]. Our data for each group contains a mixture of closely related and distantly related species.

Second, incomplete sampling might bias rate estimates. However, our rate comparisons are only between sets of species with identical taxon sampling, and so this cannot bias these comparisons. More broadly, incomplete sampling does not appear to bias rate estimates [16,54]. Furthermore, limited sampling did not generally prevent us from finding significant patterns.

Third, incomplete sampling might influence PGLS analyses. However, simulations show that limited sampling (6% of species) does not bias phylogeny-based correlations [55] and has no impact on type-1 error rates (false positives). Limited sampling does reduce power [55], but we often obtained significant PGLS results.

Fourth, we only included climatic data for species with physiological data. Nevertheless, we included more than 600 species for these comparisons for most groups (but 182 for amphibians), and obtained some significant results for all groups. Our climatic datasets were broadly similar to our physiological datasets.

Fifth, there are potential geographical mismatches between physiological and climatic data. Physiological data were generally from few localities per species, whereas climatic data were from across the species's range. Furthermore, even for a single species, data from different physiological variables were sometimes from different localities. Thus, some mismatch between variables might be caused by mismatched variation across localities. This should be a source of random error, rather than bias. Nevertheless, we performed a set of analyses using only climatic data from populations from which physiological data were obtained (electronic supplementary material, table S5; details in electronic supplementary material, appendix S2). The results were similar to the main analyses, and did not show dramatically stronger physiology–climate relationships.

Sixth, we used a single model and rate estimate for each variable in each tree. Different models and rates might apply to different parts of each tree within each group (e.g. [39]). However, the approach we used [14] should be the most practical for comparing overall rates across large-scale trees for each group.

Seventh, our physiological data came from many sources. There may be some differences in how physiological variables were measured among studies. Previous authors combined data from the same sources used here (e.g. [9,27]). However, we removed problematic UT_max data for many birds and mammals used in earlier studies (following [35]). We also removed some problematic squamate data (see above). We still have concerns about the CT_max/CT_min data for amphibians, in terms of quality and taxon sampling. The amphibian results should, therefore, be taken with some caution.

Finally, we did not correct for potentially suboptimal acclimation temperatures on CT_max and CT_min measurements. However, analyses that did found very limited impact on CT_max and no impact on overall results for both variables [27].

3. Results

Physiological and climatic data are summarized in electronic supplementary material, table S3. The analyses (table 1) show that physiological tolerances to high temperatures evolve more slowly than tolerances to low temperatures. Rates (σ²) for CT_min/UT_min are faster than for CT_max/UT_max in all groups (figure 1a). However, the difference in amphibians was not significant, and the sample size was limited (n = 30).

Table 1.

Comparison of evolutionary rates among physiological and climatic-niche variables. Physiological variables include body temperatures (T_b/T_a for ectotherms/endotherms), lower thermal tolerances (CT_min/UT_min) and upper thermal tolerances (CT_max/UT_max). Climatic variables include maximum annual temperatures (Bio5) and minimum annual temperatures (Bio6). Sample size (n) is the number of species included, and varies because only species with data for both variables were compared. Rates are not identical for the same variable in the same group when different numbers of species are included. Significant p-values (p < 0.05) are boldfaced. A sequential Bonferroni correction does not impact which results are considered significant in this table.

group	comparison	n	σ2	LRTd.f. = 1	p
amphibians	CTmin versus CTmax	30	0.38 versus 0.29	0.58	0.4476
	Bio5 versus Bio6	182	1.69 versus 1.51	0.61	0.4354
	CTmin versus Bio6	43	0.32 versus 1.94	27.37	<0.0001
	CTmax versus Bio5	96	0.24 versus 2.90	122.97	<0.0001
	Tb versus CTmax	10	0.61 versus 0.07	9.69	0.0019
squamates	CTmin versus CTmax	147	0.29 versus 0.15	16.42	<0.0001
	Bio5 versus Bio6	613	0.66 versus 1.26	62.72	<0.0001
	CTmin versus Bio6	177	0.40 versus 1.71	85.43	<0.0001
	CTmax versus Bio5	258	0.17 versus 0.72	123.39	<0.0001
	Tb versus CTmin	124	0.29 versus 0.40	2.92	0.0874
	Tb versus CTmax	173	0.28 versus 0.16	13.23	0.0003
birds	UTmin versus UTmax	75	2.65 versus 0.74	28.80	<0.0001
	Bio5 versus Bio6	636	3.96 versus 25.09	479.01	<0.0001
	UTmin versus Bio6	207	5.10 versus 8.48	13.27	0.0003
	UTmax versus Bio5	78	0.72 versus 3.91	50.04	<0.0001
	Ta versus UTmin	47	0.05 versus 2.61	120.64	<0.0001
	Ta versus UTmax	27	0.02 versus 0.25	35.88	<0.0001
mammals	UTmin versus UTmax	133	0.92 versus 0.49	12.89	0.0003
	Bio5 versus Bio6	628	3.82 versus 26.42	513.07	<0.0001
	UTmin versus Bio6	364	2.20 versus 6.18	93.13	<0.0001
	UTmax versus Bio5	134	0.49 versus 0.75	6.11	0.0135
	Ta versus UTmin	307	0.08 versus 2.18	620.38	<0.0001
	Ta versus UTmax	114	0.05 versus 0.43	106.69	<0.0001

Figure 1.

Comparisons of estimated evolutionary rates (σ²) among physiological and climatic variables. (a) upper thermal tolerances (CT_max/UT_max) versus lower thermal tolerances (CT_min/UT_min), (b) maximum temperatures (Bio5) versus minimum temperatures (Bio6), (c) upper thermal tolerances versus maximum temperatures and (d) lower thermal tolerances versus minimum temperatures. Triple asterisks indicate significantly different rates (table 1). Rates can differ when different numbers of species are included. (Online version in colour.)

Our results (figure 1b; table 1) also show that the hottest temperatures in species’ realized climatic niches change significantly more slowly than the coldest temperatures, except in amphibians. In amphibians, hottest temperatures change faster, but the difference is not significant.

Our results show mixed support for the idea that physiological evolution constrains climatic-niche evolution. Physiological tolerances to extreme temperatures do evolve more slowly than extreme climatic-niche temperatures (table 1): CT_max/UT_max evolve significantly more slowly than maximum temperatures in all groups (figure 1c), and CT_min/UT_min evolve significantly more slowly than minimum temperatures (figure 1d).

Furthermore, there are significant, positive relationships between physiological tolerances and extreme niche temperatures (table 2; figure 2). This includes both high temperatures (CT_max/UT_max versus maximum temperature: amphibians: r² = 0.09; squamates: r² = 0.02; birds: r² = 0.21; mammals: r² = 0.11) and low temperatures (CT_min/UT_min versus minimum temperature: amphibians: r² = 0.53; squamates: r² = 0.09; birds: r² = 0.12; mammals: r² = 0.07). We did not find a strong negative relationship between CT_max and maximum temperatures in squamates (as in [9]). Nevertheless, despite being positive and significant, many relationships had low r². Intriguingly, there are strong relationships between UT_max and UT_min in both birds and mammals, but not between CT_max and CT_min in squamates or amphibians (table 2).

Table 2.

Relationships between pairs of climatic and physiological variables, based on phylogenetic regression (PGLS). Variables as in table 1. We did not include a relationship between T_b and CT_min for amphibians because only two species had data for both variables. Significant associations (p < 0.05) are boldfaced, but p-values not significant after sequential Bonferroni correction are asterisked.

models	n	slope	r2	p
amphibians
CTmax ∼ Bio5	96	0.25	0.09	0.0025
CTmin ∼ Bio6	43	0.25	0.53	<0.0001
CTmax ∼ CTmin	30	0.04	<0.01	0.7143
Tb ∼ CTmax	10	−0.03	<0.01	0.9679
Tb ∼ Bio1	83	0.54	0.25	<0.0001
Tb ∼ Bio5	83	0.43	0.16	0.0002
Tb ∼ Bio6	83	0.30	0.14	0.0005
squamates
CTmax ∼ Bio5	258	0.07	0.02	0.0181*
CTmin ∼ Bio6	177	0.15	0.09	<0.0001
CTmax ∼ CTmin	147	0.04	<0.01	0.4126
Tb ∼ CTmax	173	0.52	0.15	<0.0001
Tb ∼ CTmin	124	0.11	0.02	0.1641
Tb ∼ Bio1	518	0.12	0.03	0.0001
Tb ∼ Bio5	518	0.23	0.09	<0.0001
Tb ∼ Bio6	518	0.02	<0.01	0.3028
birds
UTmax ∼ Bio5	78	0.33	0.21	<0.0001
UTmin ∼ Bio6	207	0.23	0.12	<0.0001
UTmax ∼ UTmin	75	0.39	0.43	<0.0001
Ta ∼ Bio1	474	<0.01	<0.01	0.4781
Ta ∼ Bio5	474	0.01	<0.01	0.3111
Ta ∼ Bio6	474	<0.01	<0.01	0.7508
Ta ∼ UTmax	27	0.02	0.01	0.6808
Ta ∼ UTmin	47	<0.01	<0.01	0.9043
mammals
UTmax ∼ Bio5	134	0.27	0.11	<0.0001
UTmin ∼ Bio6	364	0.15	0.07	<0.0001
UTmax ∼ UTmin	133	0.44	0.52	<0.0001
Ta ∼ Bio1	571	−0.03	0.02	0.0004
Ta ∼ Bio5	571	−0.03	<0.01	0.0244*
Ta ∼ Bio6	571	−0.02	0.03	0.0001
Ta ∼ UTmax	114	0.04	0.01	0.1985
Ta ∼ UTmin	307	<0.01	<0.01	0.8851

Figure 2.

Relationships between physiological and climatic variables, including upper thermal tolerances (CT_max/UT_max) versus maximum annual temperatures (Bio5), and lower thermal tolerances (CT_min/UT_min) versus minimum annual temperatures (Bio6). For clarity, results are from the least-squares regression for raw data (PGLS results in table 2).

On the other hand, species climatic distributions were not necessarily constrained by these physiological variables. In all four groups, species sometimes have higher values of maximum annual temperatures than CT_max/UT_max (amphibians = 6/96 species; squamates = 2/258; birds = 7/78; mammals = 35/134). Nevertheless, species have (overall) significantly higher values of CT_max/UT_max than maximum temperatures in all groups but amphibians, based on phylogenetic t-tests (electronic supplementary material, table S4). Species often have lower values for minimum temperatures than their CT_min/UT_min (amphibians = 16/43; squamates = 127/177; birds = 192/207; mammals = 354/364). Minimum temperatures were significantly lower than CT_min/UT_min in birds, but not in amphibians, squamates or mammals (electronic supplementary material, table S4).

Body temperatures (T_b/T_a) evolved more slowly than CT_min/UT_min in all groups (except in amphibians, with too few species: table 1). By contrast, T_b evolved significantly more quickly than CT_max in amphibians and squamates, whereas T_a evolved significantly more slowly than UT_max in birds and mammals. We found (table 2) that T_b was not significantly related to CT_max or CT_min, or with only weak relationships (r² < 0.05), except for T_b and CT_max in squamates (r² = 0.15; see also [10]). In amphibians, T_b was positively related to climatic variables (table 2). But for other groups, relationships were also generally weak and/or non-significant between T_b/T_a and climatic variables (table 2). In summary, these patterns support the idea that body temperatures change slowly (slower than physiological tolerances or intermediate between them), and are generally only weakly related to climate (except in amphibians). For brevity, we do not address these patterns further in the Discussion.

Overall, similar results were obtained using alternative trees (electronic supplementary material, appendix S1). These analyses confirmed that physiological tolerances to high temperatures evolve significantly more slowly than those to low temperatures in all four groups (even in amphibians, unlike the main analyses), and the hottest climatic-niche temperatures change more slowly than coldest temperatures. Physiological tolerances to extreme temperatures evolve more slowly than extreme climatic-niche temperatures (but not significantly for maximum temperatures in mammals). There were significant, positive relationships between physiological tolerances and extreme niche temperatures in all four groups. Other results were also broadly similar.

We also tested relationships between extreme physiological and climatic variables when data were obtained from the same localities, focusing on ectotherms (electronic supplementary material, appendix S2). These analyses confirmed that (i) extreme climatic-niche variables evolve faster than their corresponding physiological tolerances, (ii) CT_min and coldest temperatures are significantly related, (iii) species rarely occur where hottest temperatures are hotter than their CT_max, but often occur where coldest temperatures are lower than their CT_min. A notable difference is that CT_max and maximum temperatures were not significantly related in this dataset, in squamates or amphibians. Thus, weak relationships between physiological and climatic variables in the main analyses are not necessarily caused by a mismatch between localities for climatic and physiological data, since relationships can be weak when these data are matched.

4. Discussion

The relationships between physiological tolerances, climatic niches and evolution are of urgent importance as global warming potentially threatens much of global biodiversity. Here, we broadly test several hypotheses about rates of physiological and climatic-niche evolution. We find that rates of physiological evolution for upper thermal tolerances are slower than for lower thermal tolerances (figure 1a). We show that rates of change in maximum temperatures in the realized climatic niche are significantly slower than rates for minimum temperatures, and that rates of physiological evolution are slower than rates of climatic-niche evolution for these variables (figure 1b–d). We also show that physiological tolerances to high and low temperatures are significantly, positively related to both maximum and minimum temperatures (respectively) in the realized climatic niche (figure 2). By contrast, Araújo et al. [9] stated that tolerances to higher temperatures tend not to covary with maximum temperatures. However, we found that these relationships are often weak, and climatic-niche values are often outside the bounds of these physiological variables. Below, we address the potential causes of these patterns. We also address the idea that mismatches between physiological and climatic-niche values mean that climatic-niche data are misleading.

(a). Why are rates lower for higher temperatures?

Our results show that physiological tolerances to higher temperatures change more slowly than tolerances to lower temperatures, and that rates of niche change are slower for hottest temperatures than coldest temperatures. What might explain these patterns?

One potential explanation is that there are asymmetric constraints on rates of physiological evolution, and that physiological tolerances then limit the corresponding rates of climatic-niche evolution. Thus, there may be less change in maximum niche temperatures among species because the limited evolution of physiological tolerances to high temperatures constrains such shifts, whereas rapid evolution of physiological tolerances to cold allows species to adapt to environments with very different minimum temperatures. This is a particularly important hypothesis, because it implies that the limited rate of evolution of physiological tolerances to high temperatures will make species especially vulnerable to anthropogenic warming [9]. However, this hypothesis also raises the question of what explains the asymmetry in rates of physiological evolution. Araújo et al. [9] suggested that shared tolerances to high temperatures reflect negative impacts of high temperatures on membranes and proteins [56], whereas variation in lower thermal tolerances reflects different thermodynamic effects on reaction rates, especially those associated with ion homeostasis. They suggested that the former impacts may be more conserved.

Surprisingly, the hypothesis that these physiological variables explain variation in climatic niches has only mixed support in our results. We found lower evolutionary rates in physiological variables than in their corresponding climatic variables. We also found that relationships between climatic and physiological variables were positive and significant, but often weak (most r² < 0.20; table 2). More importantly, species often occur where minimum temperatures are lower than their lower physiological tolerances (CT_min/UT_min). This pattern implies that this physiological variable does not determine where species occur. By contrast, species often occur where maximum annual temperatures are below their upper physiological tolerances (CT_max/UT_max).

Why are these physiological variables uncoupled from these estimates of large-scale climate? For CT_min, many species of ectotherms may simply be inactive during cold weather. Thus, if the coldest temperatures are too low for locomotion, it may have few consequences. Similarly, species can occur in hotter climates than suggested by their CT_max by being active at cooler temperatures (e.g. at night or during rainy weather in amphibians). More generally, behavioural thermoregulation (e.g. occurring in cooler microclimates during hot weather) may allow species to occur in climatic conditions that differ from the range of values suggested by these physiological variables (e.g. [7,57]). Seasonal migration might also help explain these patterns in birds, especially for UT_min in birds (e.g. bird species that winter in the tropics can avoid cold winter temperatures).

Our results also imply an alternative (but non-exclusive) hypothesis to explain the low rates of physiological and niche evolution for maximum temperatures. Based on electronic supplementary material, table S3, the range of values among species for minimum niche temperatures is almost twice as large as for maximum temperatures, for all four groups (ranges for maximum temperatures = 25–32; ranges for minimum temperatures = 46–61). Thus, minimum temperatures may decrease strongly from the equator to the poles, but maximum temperatures can overlap broadly between temperate and tropical regions (e.g. [58,59]). Therefore, an alternative explanation for slower rates of physiological and niche evolution for high temperatures is that maximum temperatures are less variable across the globe than minimum temperatures. Indeed, previous studies have shown how (at least in lizards) CT_min changes strongly across latitudes whereas CT_max does not [57,60].

We acknowledge the potential ‘chicken and egg’ problem in discriminating between these two hypotheses, however. In other words, is there limited change in maximum temperatures among species because of limited evolution in physiological tolerances (constraining where a species can occur), or is there limited change in physiological tolerances to high temperatures because of limited variation in maximum temperatures across the globe? It is possible that both factors contribute to the lower rates for upper temperatures for the physiological and climatic variables, and that the importance of each factor varies among species. Intriguingly, estimates of σ² in squamates and mammals [61] suggest that species' maximum latitudes (i.e. in cooler climates) change more rapidly among species than their minimum latitudes (i.e. in warmer climates), showing how rates of change in distributions parallel the relevant rates in physiological variables.

A third, non-exclusive hypothesis is that many ectotherms use behavioural thermoregulation to avoid high temperatures, which helps constrain the evolution of CT_max across different environments (e.g. [13,15]). Thus, species may behaviourally avoid the highest temperatures rather than adapting to them physiologically. Conversely, behavioural thermoregulation may be less likely at lower temperatures, when individuals are inactive and unable to move (e.g. sleeping or seasonally inactive). Although this hypothesis is appealing, one might also argue that species in different environments could (in theory) avoid physiological evolution for different lower thermal tolerances by only being active at similar temperatures, the same as for upper temperatures. Overall, our results show a strong difference between rates of evolution for upper and lower temperature variables for both physiological tolerances and climatic niches, but the causes of these differences will require further study to elucidate.

(b). Do mismatches between climatic niches and physiological tolerances mean that climatic-niche data are ‘misleading’?

Our results show relatively weak relationships between physiological variables and maximum and minimum climatic-niche temperatures (table 2; figure 2). Araújo et al. [9] suggested that conservatism in physiological tolerances to high temperatures would lead to mismatches between realized and fundamental niches. They also suggested that in studies of niche evolution ‘if the currency of interest is rates of change in physiological aspects of the niche’ then mismatches between physiology and climate would lead to ‘erroneous conclusions' about niche conservatism and evolution based on climatic-niche data, including overestimating rates of niche evolution.

We think that many researchers analyse climatic-niche data not to estimate rates of physiological evolution, but rather to understand the causes and consequences of species distributions over space and time. Thus, researchers analyse climatic-niche data to address topics such as species responses to climate change (e.g. [5,62,63]), speciation (e.g. [64,65]), large-scale diversification patterns among clades (e.g. [40,41]), biogeography and species richness (e.g. [66,67]) and the distributions of introduced species (e.g. [68,69]). Physiological data might also be very relevant to explaining these distribution patterns, but understanding physiological evolution is not the primary goal.

Rather than necessarily indicating that climatic-niche data are wrong, mismatches between physiology and climatic distributions might instead indicate that some physiological variables do not necessarily determine species' climatic and geographical distributions. For example, our results show that many species occur where winters are colder than their CT_min, but it would be nonsensical to suggest that these climatic data were, therefore, misleading because of this mismatch with the physiological data.

We follow the idea here that the realized climatic niche is the set of climatic conditions where a species occurs (potentially determined by physiological tolerances, species interactions and other factors), whereas the fundamental climatic niche is the set of conditions where the species can occur given constraints of physiology alone [70]. Araújo et al. [9] implicitly treated species values of CT_max and CT_min as the species' fundamental climatic niche. However, CT_max and CT_min simply reflect the temperatures at which species can be active (or active without metabolic cost, for UT_max/UT_min). Importantly, these physiological variables may sometimes be more relevant to when species are active (daily and seasonally) and exactly where (e.g. microhabitat, sun versus shade [57,71]) rather than to large-scale climatic distributions and climate change.

In summary, mismatches between physiological and climatic-niche data do not necessarily mean that the physiological data are correct and climatic-niche data are misleading. Our results imply that these physiological variables may not reflect the large-scale climatic conditions where species can occur. This seems problematic for studies of species responses to climate change. Indeed, some studies suggest that recent climate-related species declines and extinctions may often be associated with changes in species interactions, not physiological tolerances (e.g. [72,73]). Thus, predicting species responses to climate change based on physiology alone might be problematic. Furthermore, many analyses of climate change that use physiological data incorporate additional factors, such as microclimate, behavioural thermoregulation, impacts of temperatures on activity schedules and acclimation (e.g. [3,27,60,74]). There are also alternative physiological measurements that may be even more relevant to species activity patterns and climatic distributions than CT_max/UT_max and CT_min/UT_min, such as voluntary thermal maxima (e.g. [75]). In theory, lethal thermal limits might predict species climatic distributions better than CT_max/UT_max and CT_min/UT_min, but this seems unlikely for upper limits (since most ectotherms occur at temperatures below CT_max, based on our results).

(c). Implications of evolutionary rates for climate change

Our results show that physiological tolerances to heat evolve relatively slowly, whereas maximum niche temperatures change more quickly. How do these rates compare to rates of climate change? Our rate estimates here do not have a straightforward interpretation in terms of units. However, analyses using alternative methods (with rates in °C per year) show that past rates of realized-niche change are much slower than rates of projected climate change, often by approximately 100 000-fold or more (tetrapods [76]; plants and animals [16,54]). Our results here show that physiological tolerances evolve more slowly than these realized climatic-niche variables. Thus, both physiological and niche rates seem too slow to keep pace with rapid, projected climate change. However, in contrast to these results, analyses of introduced vertebrate populations show that very rapid climatic-niche shifts are possible, much faster than projected rates of climate change [77]. These include shifts into regions with much hotter maximum temperatures than the native distributions. It seems unlikely that these maximum-temperature shifts generally occurred through rapid physiological evolution, since most occurred as species were introduced (‘instantaneous’ niche shifts). Further, analyses of recent responses to climate change suggest that rapid niche shifts in maximum annual temperatures may be crucial to the survival of plant and animal populations and species [5], but the mechanisms underlying these shifts remain unclear. Overall, we caution that neither realized climatic-niche data nor physiological data alone may fully predict how species respond to climate change. Slow rates of evolution for physiological tolerances to heat and maximum climatic temperatures may both be important, but are only two among many relevant factors.

(d). Summary

Here, we show that physiological tolerances to heat evolve more slowly than those to cold, that hottest temperatures in the climatic-niche evolve more slowly than coldest temperatures, and that physiological tolerances evolve more slowly than their corresponding climatic-niche variables. We also show significant positive relationships between physiological and climatic-niche variables, but these relationships are generally weak, and species can occur in climates outside those implied by these physiological variables. Therefore, we caution against equating species-level data for these physiological variables with the fundamental climatic niche, and using mismatches between climatic and physiological data to argue that climatic-niche data are misleading.

Data accessibility

All data are available as electronic supplementary material and from the Dryad Digital Repository: https://dx.doi.org/10.5061/dryad.m63xsj3zv [78].

Authors' contributions

Y.-F.Q. and J.J.W. designed the study and wrote the paper. Y.-F.Q. performed the analyses.

Competing interests

We have no competing interests to declare.

Funding

Y.-F.Q. was funded by the National Natural Science Foundation of China (grant nos 31670422 and 31770443), the Natural Science Foundation of Jiangsu Province (grant no. BK20161556) and the Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions. J.J.W. was supported by the U.S. National Science Foundation grant no. DEB 1655690.