The Influence of Late Quaternary Climate-Change Velocity on Species Endemism

D. Currie presents three main criticisms of our recent Report (1). First, he says that we focus most of our attention on the role of climate-change velocity, when in fact the strongest predictor of endemism is the extent of analogous contemporary climate. Here, our philosophical viewpoints seem fundamentally different. Currie refers to velocity and extent as competing explanations, whereas we see them as complementary, based on our expectation that patterns of endemism in amphibians, mammals, and birds would not be entirely attributable to any one explanatory variable (2, 3). We were thus not interested in eliminating all but one explanation, but rather in assessing the support for several possible explanations and the interactions among them. If models must compete against one another, we suggest comparing the model sets with and without each candidate predictor variable, as we did when calculating the summed Akaike weights for each variable (2). This gives a measure of the indispensability of each predictor variable, and shows that multiple factors, both historical and contemporary, are needed to adequately describe endemism patterns.

Second, Currie notes that, to the extent that historical and contemporary climate variables are collinear, it is difficult to assess their relative importance; this is a general problem in comparative studies. However, there was not a strong multicollinearity between velocity and our other predictor variables (tolerance values for velocity: amphibians = 0.51, mammals = 0.49, birds = 0.52, values differ due to the use of different predictor variable subsets). As always, it is impossible to ever be certain that any particular predictor is causally connected with a response; there could always be an unmeasured collinear explanation. Thus, the best practice is to consider and test such possible alternative explanations, as we did with topographic heterogeneity and temperature anomaly. This challenge is by no means limited to our study, nor to historical explanations in particular. We also note that the within-region r values reported in our Report and quoted by Currie are global means. We argue that the local importance of velocity should vary strongly, as indeed it does [Fig. 3A (1)]. Under conditions where velocity was predicted to be most important, the local relationships do become much stronger [Fig. 3, B–D (1)].

Finally, Currie suggests that our explanation demands numerous species extinctions since the Last Glacial Maximum (LGM). There is little evidence for massive numbers of global species extinctions during this time period, except for megafaunal extinctions, whose causes are only controversially linked to climate change (4). However, as stated in our Report (1), the spatial pattern of warming from the LGM to the current period is likely to resemble warming patterns following previous glacial periods (5, 6). The climate-change velocity used in our Report thus likely reflects a much longer period than 21,000 years, probably the whole Late Quaternary, and perhaps even beyond. Additionally, extinctions below the species level are documented in this time period (e.g., 7) and, as argued in our Report, may contribute to the observed velocity-endemism relationship.

We do not claim that a historical explanation for endemism patterns should supplant a contemporary one. Rather, we show that adding a mechanistically based historical predictor greatly improves our ability to explain modern endemism patterns.

B. Sandel, J.-C. Svenning

Ecoinformatics and Biodiversity Group, Department of Bioscience, Aarhus University, Aarhus 8000 C, Denmark.

L. Arge

Center for Massive Data Algorithmics (MADALGO), Department of Computer Science, Aarhus University, Aarhus 8000 C, Denmark.

B. Dalsgaard, W. J. Sutherland

Conservation Science Group, Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, UK.

R. G. Davies

School of Biological Sciences, University of East Anglia, Norwich NR4 7TJ, UK.

K. J. Gaston

Environment and Sustainability Institute, University of Exeter, Cornwall TR10 9EZ, UK.

References and Notes

1. B. Sandel et al., Science 334, 660 (2011).

2. K. P. Burnham, D. R. Anderson, Model Selection and Multi-Model Inference: A Practical Information-Theoretic Approach (Springer, New York, ed. 2, 2002).

3. J. B. Johnson, K. S. Omland, Trends Ecol. Evol. 19, 101 (2004).

4. A. D. Barnosky, P. L. Koch, R. S. Feranec, S. L. Wing, A. B. Shabel, Science 306, 70 (2004).

5. R. Jansson, Proc. R. Soc. London Ser. B 270, 1515 (2003).

6. Records of sea surface temperature from oceanic sediment cores, for example, show that the magnitude of warming following several previous glaciations are well-correlated (www.ncdc.noaa.gov/paleo/recons.html).

7. L. Dalén et al., Proc. Natl. Acad. Sci. U.S.A. 104, 6726 (2007).

Conflict of Interest:

None declared

B. Sandel et al. recently argued that areas of the Earth where rates of climate change since the Last Glacial Maximum (LGM) were largest have few endemic (small-ranged) species due to elevated extinction rates ("The influence of Late Quaternary climate-change velocity on species endemism," Reports, 4 November 2011, p. 660, published online 6 October 2011). The hypothesis is intriguing. However, Sandel et al. focus on results that support their hypothesis, rather than testing it against its most obvious competitor: that endemic species occur in places where contemporary patches of climatically appropriate habitat are small. In this light, Sandel et al. found that endemism is more strongly correlated with the area of analogous contemporary climate (r² = 0.415 - 0.538) than with climate change velocity (r² = 0.283 - 0.385) for all taxa they studied. In multiple regressions, effect sizes for contemporary patch size were nearly twice those for historical climate change. Sandel et al. report that "models incorporating velocity were always strongly preferred over equivalent models without velocity." However, the same is true for several contemporary habitat variables. Given that historical and contemporary variables are strongly collinear, there is no reason to attribute causation to any one of them. Analyses that were restricted to 10° x 10° regions showed that "high velocity areas within regions were associated with low endemism." In fact, within regions where collinearity between climate change and contemporary climate is likely weaker, the correlation between endemism and climate change nearly disappears: r² = 0.008 - 0.026.

Finally, if extinctions produced the gradients of endemism observed in this study, then very large numbers of species would have to have gone extinct since the LGM. Published literature shows dramatic shifts in species' ranges as glaciers retreated, but little evidence of mass extinctions during that period (1). If, however, extinctions were only local, then species distributions simply shifted in response to shifting climate. In sum, the study presents no evidence to favor a historical explanation of endemism over a contemporary one.

David Currie

Biology Department, University of Ottawa, Ottawa, ON K1N 6N5, Canada.

Reference

1. J. W. Williams, B. N. Shuman, T. Webb, P. J. Bartlein, P. L. Leduc, Ecol. Monogr. 74, 309 (2004).

Conflict of Interest:

None declared