Global patterns of thermal tolerances and vulnerability of endotherms to climate change remain robust irrespective of varying data suitability criteria

In their comment to our paper on the global variation in thermal tolerances and the vulnerability of endotherms to climate change [1], Wolf et al. [2] have expressed concerns that errors in data compilation undermine the credibility of our analyses and conclusions. In a nearly identical comment, the same authors [3] have also criticized our paper on the phylogenetic and environmental determinants of the variation in the thermal traits of endotherms [4]. The main issue raised is potentially inadequate measures of species' upper critical temperatures (UCT) and thus of their thermo-neutral zone (TNZ) due to small numbers of individuals in the experiments or to an experimental range of ambient temperatures that Wolf et al. [2] deem insufficient to determine an UCT value. Having responded in detail to these criticisms elsewhere [5], we will here only briefly repeat our main arguments. Specifically, these refer to (i) our own assessment of data quality and suitability and (ii) the robustness of our results and conclusions to varying the data included in the analyses.

While we admit that our paper and especially the data compilation and documentation may have profited from more transparency and, for a small proportion of the data points, also from stricter inclusion criteria, we disagree in large parts with Wolf et al.'s assessment of data quality. In particular, we disagree with their conclusion that approximately 71% of the data used are not suitable for our analyses. In general, whether data are suitable for a specific study depends on its scope and aim. For instance, coarse-resolution data may be useful for large-scale assessments of potential impacts from climate change threats [6,7], but are not suitable for predicting species-specific extinction risk in a certain region. Our paper [1] is a global-scale, multi-species analysis on macrophysiological patterns of thermal tolerances. Given this scope and scale, we invested much effort to compile a dataset that represents the global diversity of birds and mammals, thereby taking a rather inclusive approach [8] and thus prioritizing generality over detail. Accepting such a prioritization renders most species categorized as ‘insufficient data’ by Wolf et al. as generally reasonable to use for an analyses as ours, and the proportion of potentially problematic data decreases to the species classified as ‘No UCT’ by Wolf et al. [2] (58% of birds and 43.4% of mammals; see electronic supplementary material, tables S1 and S2 for details).

After a point-by-point re-check of all data classified as ‘No UCT’ by Wolf et al. [2], this classification is questionable. In most of the problematic cases (birds: 72% and mammals: 78% of species within Wolf et al.'s ‘No UCT’ category; our categories 3A–C; see electronic supplementary material, tables S1 and S2), TNZ limits are explicitly reported in the text of the respective original studies or the data have been extracted from unpublished theses. If quantitative data values are reported in the text of a peer-reviewed paper even though the figures depicting the individual experimental results do not support these values, it remains a challenge to evaluate the validity of the data points. While we are willing to admit that our data documentation may have profited from more such details on the data sources, Wolf et al.'s critique may also call for a reflection on the review and publication standards and practices in their own field (see also the debate in [8,9]). However, as long as these issues are still debated, we think that peer-reviewed information on physiological traits can and should be used with care as long as the data are adequate for the aim and scale of the respective study (e.g. [10–12]).

Based on these arguments, we do in fact think that most of the data (our data categories 1, 2 and 3A–C; see electronic supplementary material, figure S1 and table S1 for details on categories 3B and C) are in fact suitable for the analyses in our paper. However, we also admit that we may have been too generous in several cases (category 3D; 26 bird species = 16.1% of the original dataset; 24 mammal species = 8.1%) for which our UCT data represent values of the highest ambient temperatures used in the respective experiment or for which we were not able to retrieve the original publication source (category 3E, 4 mammal species = 1.4% of the original dataset). However, most category 3D data, which may serve as minimum estimates of UCT, might also actually be potentially useful for our analyses.

Nevertheless, irrespective of which data may be judged as most suitable, we repeated the statistical analyses of our paper with datasets of different categories. For several species, we found minor mistakes in the geographical coordinates of the sites where the animals for the physiological experiments were captured. These are now corrected in the updated data file (electronic supplementary material, table S2). Regarding the test of the climate variability hypothesis, we found that despite some variation in the strength of the effect, the overall result of a stronger relationship between climatic variability and thermal tolerance breadth in birds than in mammals remains largely robust (table 1; with one exception: in one of the mammal datasets, we found a significant relationship between precipitation seasonality and TNZ breadth). The patterns of potential vulnerability to climate change based on the mismatch between species' UCT and current or future temperature conditions locally (electronic supplementary material, figure S1) or throughout distributions (electronic supplementary material, figure S2) are also robust irrespective of the dataset used.

Table 1.

Association of thermo-neutral zone (TNZ) breadth of birds and mammals with climatic variability, for different categories of data quality. TNZ breadth was modelled using pgls (phylogenetic generalized least squares), as a function of body mass and acclimatization, while we estimate Pagel's λ and set it to its maximum-likelihood value (ML λ). After controlling for body mass, acclimatization and phylogeny, we individually added temperature seasonality, precipitation seasonality or radiation seasonality to the model. Italic values indicate associations where estimated parameters (B) are significantly different from 0. Seasonality variables are calculated as the coefficient of variation of monthly climate variables. See electronic supplementary material, figure S1 and table S1 for the definition of the different data categories.

	birds				mammals
	B	λ	R2	p-values	B	λ	R2	p-values
category 1
mass	0.14 ± 0.05	0.00	0.54	0.011	0.07 ± 0.03	0.62	0.04	0.02
acclimatization	0.26 ± 0.17	0.00	0.54	0.14	0.07 ± 0.13	0.62	0.04	0.58
absolute latitude	0.02 ± 0.006	0.00	0.54	0.002	−0.005 ± 0.004	0.62	0.04	0.33
temperature seasonality	0.15 ± 0.08	0.00	0.44	0.07	−0.04 ± 0.06	061	0.03	0.47
precipitation seasonality	−0.02 ± 0.12	0.22	0.38	0.86	−0.004 ± 0.10	0.62	0.02	0.96
radiation seasonality	0.41 ± 0.13	0.00	0.52	0.005	−0.12 ± 0.10	0.63	0.03	0.23
annual temperature range	0.38 ± 0.19	0.00	0.44	0.05	−0.09 ± 0.14	0.60	0.03	0.49
categories 1 and 2
mass	0.24 ± 0.03	0.25	0.43	<0.001	0.10 ± 0.02	0.51	0.06	<0.001
acclimatization	0.28 ± 0.12	0.25	0.43	0.027	−0.03 ± 0.09	0.51	0.06	0.73
absolute latitude	0.009 ± 0.004	0.25	0.43	0.04	0.0004 ± 0.003	0.51	0.06	0.90
temperature seasonality	0.04 ± 0.05	0.26	0.39	0.36	0.01 ± 0.04	0.52	0.06	0.70
precipitation seasonality	−0.06 ± 0.08	0.34	0.39	0.47	−0.06 ± 0.06	0.48	0.07	0.34
radiation seasonality	0.13 ± 0.09	0.30	0.40	0.14	0.01 ± 0.06	0.52	0.06	0.81
annual temperature range	0.14 ± 0.13	0.26	0.40	0.28	0.04 ± 0.11	0.53	0.06	0.69
categories 1, 2 and 3A–C
mass	0.25 ± 0.03	0.30	0.39	<0.001	0.12 ± 0.01	0.24	0.22	<0.001
acclimatization	0.13 ± 0.09	0.30	0.39	0.13	−0.05 ± 0.07	0.24	0.22	0.47
absolute latitude	0.01 ± 0.002	0.30	0.39	<0.001	0.003 ± 0.002	0.22	0.22	0.26
temperature seasonality	0.07 ± 0.03	0.27	0.33	0.03	0.01 ± 0.03	0.20	0.23	0.55
precipitation seasonality	−0.08 ± 0.06	0.46	0.30	0.21	−0.16 ± 0.05	0.00	0.20	0.001
radiation seasonality	0.16 ± 0.06	0.24	0.34	0.009	0.010 ± 0.05	0.10	0.23	0.052
annual temperature range	0.11 ± 0.10	0.33	0.31	0.25	−0.069 ± 0.07	0.00	0.16	0.34

For these reasons, the conclusions of our paper [1] remain unchanged: (i) the climatic variability hypothesis receives support from our analyses for birds, but not for mammals; (ii) future temperature increases may be within the thermal tolerance limits for many of the investigated species; and (iii) from a temperature perspective alone, tropical species are more vulnerable from global warming than species in temperate areas.

All of our arguments and re-analyses render the doubts by Wolf et al. [2] about the credibility of our analysis ill-founded. Their suggestions to extend or refine analyses such as ours with additional information, e.g. on functional traits, are certainly valuable, but depending on the scope of a certain study as well as journal space restrictions the possibilities to consider additional details are, of course, limited. However, we hope that Wolf et al.'s suggestions encourage further studies extending and refining our analyses—irrespective of whether they support or question the broad conclusions of our paper.

To conclude, we agree that future macrophysiological analyses will benefit from more rigorous and transparent criteria for data selection. However, we also emphasize that prior to final statements about the credibility of any scientific study, it is important to analytically re-evaluate the results and to interpret them in the light of the scope, aim and scale of the respective analyses. Moreover, perspectives from different disciplines or approaches may lead to different decisions about whether certain data are suitable for a specific study. However, as long as debates on these perspectives are constructive, we hope they will stimulate exchange of data and approaches among different disciplines. We think that more of such efforts for integration are urgently needed to successfully address the scientific challenges in global change ecology and beyond.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material, or were part of the electronic supplementary material of [1].

Authors' contributions

C.H. and I.K. coordinated the study, I.K. and C.H. ran analyses. C.H. wrote the paper, with the help if I.K. All authors commented on the manuscript and gave their final approval for publication.

Competing interests

We have no competing interests.

Funding

No funding has been received for this article.