Thomas et al. (2004) examined the connection between extinctions and climate change via the well-known power-law relationship that describes the number of species found in a given area, where current habitat areas of species are obtained from the results of ENMs based on "climate envelopes" that define the sizes and locations of these areas.  These results are then compared with the future sizes and locations of habitat areas that are predicted by ENMs under various global warming scenarios, from which a new species number can be calculated, which when compared with the original number provides an assessment of the number of extinctions expected to be caused by the specified future warming.

One of several complications associated with this approach is our lack of knowledge of the unique dispersal ability of each species.  Hence, two extreme dispersal scenarios are used to derive upper and lower bounds on extinction predictions: free dispersal, which assumes all species are free to migrate to the full extent of their future climatically-suitable habitats, and no dispersal, which assumes that species cannot move at all from their current locations.

Both scenarios incorporate only range-contracting species in the methodology of Thomas et al., as species with expanding ranges are discounted on the basis that they are not at risk of extinction due to climate change.  This approach, however, ignores species that are currently threatened with extinction by non-climatic factors, and which could therefore benefit from an expanded potential habitat and so escape extinction in the new CO2/climate regime.  For example, a CO2- or climate-driven range expansion would clearly help species that are threatened with extinction due to increasing habitat loss attributable to expanding urbanization and agricultural activities; while it may help other species that are threatened with extinction by habitat fragmentation to cross geographical barriers that were previously insurmountable obstacles to them.  Hence, by neglecting the many species that fall into these and other like categories, no decrease in extinctions is possible under Thomas et al.'s approach to the problem, even under the free dispersal scenario, with the result that a massive increase in extinctions is a foregone conclusion.

The no dispersal scenario also forces an unrealistic decrease in range with any climatic change that shifts habitat area without reducing it; while "overfitting" reduces ranges even more, producing systematic errors on the order of 10-20%, particularly with smaller data sets, deficiencies in data sampling and modelling methods, and the inclusion of irrelevant variables (Stockwell and Peterson 2002a, 2002b, 2003).  In the study of Bakkenes et al. (2002), for example, two independent climate variables adequately explain 93% of the variation in their dependent variable; while the use of more climate variables ends up incorporating more random variation than it does actual signal, leading to a contraction of the climate envelope and a systematic bias towards smaller predicted ranges.  It comes as no surprise, therefore, that in this study and that of Peterson et al. (2002) -- which comprise two of the six major studies on which the analysis of Thomas et al. is based -- the use of only two climate variables by the two studies yields extinction percentages of 7% and 9%, while the four additional studies upon which Thomas et al. rely (which use from 3 to 36 independent variables) yield extinction percentages ranging from 20% to 34%, consistent with what would be expected from errors associated with statistical overfitting.

Because ecological models are notoriously unreliable, the common sense response when extreme results such as those of Thomas et al. are encountered would be to attempt to verify some aspect of them with independent data.  However, their single attempt to do so with a real-world extinction supposedly caused by global warming (Pounds et al., 1999) has been satisfactorily explained by changes in local weather patterns due to upwind deforestation of adjacent lowlands (Lawton et al., 2001).  Hence, Thomas et al. have a dearth of pertinent hard data to support their contentions; and while the absence of evidence does not necessarily disprove a claim, the lack of any real extinction data to support the results of their analysis certainly suggests that the models they are using are not "tried and true."

In the real world, as readers of CO2 Science Magazine are aware, species survive in a fabric of positive and negative interactions, and changes are not uniformly deleterious.  Hence, a balanced approach should be taken in modelling the effects of climate change on biodiversity.  While Thomas et al. speculate that displaced species are hampered by habitat fragmentation, have difficulty persisting due to newly-encountered competition from other species, and are universally stressed by increases in atmospheric temperature and CO2 levels, there are many documented instances of species tolerating, or even benefiting from, some of these very factors (Idso et al., 2003).

In summary, Thomas et al. (2004) seek to create the impression of impending ecological disaster due to CO2-induced global warming, claiming their results justify mandating reductions of greenhouse gas emissions.  However, it is clear that their results are forced by the calculations, confounded with statistical bias, lack supporting real-world evidence, and are perforated with speculation.  It would thus appear that their doctrine of "massive extinction" is actually a case of "massive extinction bias."

References
Bakkenes, M., Alkemade, J.R.M., Ihle, F., Leemans, R. and Latour, J. B. 2002. Assessing effects of forecasted climate change on the diversity and distribution of European higher plants for 2050.  Global Change Biology 8: 390-407.

Idso, S.B., Idso, C.D. and Idso, K.E.  2003.  The Specter of Species Extinction: Will Global Warming Decimate Earth's Biosphere.  Center for the Study of Carbon Dioxide and Global Change, Tempe, AZ, USA.  www.co2science.org

Lawton, R.O., Nair, U.S., Pielke Sr., R.A. and Welch, R.M.  2001.  Climatic impact of tropical lowland deforestation on nearby montane cloud forests.  Science 294: 584-587.

Peterson, A.T., Ortega-Heuerta, M.A., Bartley, J., Sánchez-Cordero, V., Soberón, J., Buddemeier, R.H. and Stockwell D.R.B.  2002.  Future projections for Mexican faunas under global climate change scenarios.  Nature 416: 626-629.

Pounds, J. A., Fogden, M. L. P. and Campbell, J. H.  1999.  Biological response to climate change on a tropical mountain.  Nature 398: 611-615.

Stockwell D.R.B. and Peterson, A.T.  2002a.  Controlling bias during predictive modeling with museum data.  In: Predicting Species Occurrences: Issues of Scale and Accuracy (Scott, J.M., Heglund, P.J., Morrison, M., Raphael, M., Haufler, J. and Wall, B., Eds.).  Island Press, Covello, CA.

Stockwell, D.R.B. and Peterson, A.T.  2002b.  Effects of sample size on accuracy of species distribution models.  Ecological Modelling 148: 1-13.

Stockwell, D.R.B. and Peterson, A.T.  2003.  Comparison of resolution of methods for mapping biodiversity patterns from point-occurrence data.  Ecological Indicators 3: 213-221.

Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Barend, F., Erasmus, N., Ferreira de Siqueira, M., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L. and Williams, S.E.  2004.  Extinction risk from climate change.  Nature 427: 145-148.