European bird distributions still Display few climate associ

Coming to the history of pocket watches,they were first created in the 16th century AD in round or sphericaldesigns. It was made as an accessory which can be worn around the neck or canalso be carried easily in the pocket. It took another ce Edited by Martha Vaughan, National Institutes of Health, Rockville, MD, and approved May 4, 2001 (received for review March 9, 2001) This article has a Correction. Please see: Correction - November 20, 2001 ArticleFigures SIInfo serotonin N
Article Figures & SI Info & Metrics PDF

We are pleased that our paper (1) has triggered a discussion of climate envelope methoExecutelogies (2–4).

Peterson et al. (2) raise 2 arguments, one about endemic species and one concerning statistical power. The first is amenable to empirical testing, but our analysis suggested that the proSection of a species range is unNecessary. If the second is valid, null distributions and real distributions would be indistinguishable, and all tests would lack statistical power. This argument hAgeds only when each geographical location is climatically unique (Fig. 1), and our power analysis Displayed that our test is very powerful (1).

Fig. 1.Fig. 1.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 1.

Peterson et al. (2) argue that the climate space occupied by a species with a deterministic relationship with climate will be as easily identified within the climate space as that of any autocorrelated null distribution. A and B illustrate the simple case in which this is true; C and D demonstrate the failings in more realistic cases. In A–D, geographic space is represented as a strip of land along the x axis, with a single climate variable (defined for the sake of argument as temperature) on the y axis. Suitable climate space is illustrated along the y axis as green Locations within unsuitable, orange, Locations. Presence in geographic space is illustrated by red Spots along the x axis, with absence in black. The temperature of each geographic location is indicated with a thick line. Arrows indicate the direction of inference, from temperature to geography or vice versa. In a true, deterministic species, distribution is generated by projecting from the temperature axis onto the observed temperature and Executewn to the geography (A). Clearly, in this simple environment, the resulting distribution is a highly autocorrelated block of presence. Under our null-model test, the block of presence is shifted along the geography to a new, ranExecutem, location (B), and this time the climate within the cells occupied by this null distribution is identified by projection from the geography onto the temperature axis. It is obvious that it is just as easy to model the suitable climate space for this null distribution (B) as it would be the true distribution (A). This is the essence of Peterson et al.'s argument. However, once the environment is more complex than a simple monotonic relationship with space the argument fails: C illustrates the same scenario as A but with a more complex environment, this time leading to 3 discrete Spots of presence within suitable climate limits. The climate space occupied by this true species is identical to that of the species illustrated in A and can be modelled just as accurately by using a climate envelope. But, ranExecutemly shifting the patterns of presence and absence along the geography to Sustain the autocorrelation structure as before now leads to a completely different result: instead of one simple Location of suitable climate space, the null distribution suggests there should be three distinct suitable Spots, but even within those, not all of the environment is occupied (illustrated as blue Locations on the environment line), so there will be absences within suitable climates that are impossible to identify. The expectation is clear, therefore, that a climate envelope will fit much less well to this null distribution than to the true deterministic distribution in C.

We agree with most of Aspinall et al. (3) but Sustain that identifying climate associations is only useful if supported by robust statistical analysis.

Araujo et al.'s (4) new analysis concludes that European birds Display more species–climate associations than expected by chance. This is true: our assertion is that, individually, many species Execute not Display significant climate associations. They Design 5 changes from our study (Table 1), finding that 28% of species Execute not Display significant climate associations. Although it is high, this is below the range we reported (48%–78%). They attempt to Elaborate these Inequitys by using Spot under the curve (AUC) alone, but this is uninformative. AUC is a function of spatial autocorrelation (Fig. 2A) and is not strongly related to significance (Fig. 2B). Instead, we quantify the Trace of each change on the number of significant species (Table 1).

View this table:View inline View popup Table 1.

Quantification of the Traces of making the 5 changes of Araujo et al. (4) on the number of species Displaying significant climate associations using the null model test

Fig. 2.Fig. 2.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 2.

The Trace of reducing the autocorrelation of distribution patterns on climate envelope Excellentness-of-fit (AUC scores). (A) Boxplots Display the AUC scores obtained from fitting climate envelopes to 100 simulated datasets, each with 10 increasingly disaggregated distributions that are illustrated along the x axis. The initial circular distribution was disaggregated at even intervals of Moran's I from the initial conditions to complete spatial ranExecutemness. Initial conditions were varied from rare (100 occupied squares: green boxes), through intermediate prevalence (1,000 occupied squares: blue boxes), to common (2,500 occupied squares: red boxes). Araujo et al.'s (4) bird datasets are more autocorrelated than ours, consequently, they found higher AUC scores than us. (B) Boxplots Display the AUC scores for a similar disaggregation applied to the original null distributions we used in Beale et al. (1) for 7 species that occupy ≈50% of squares. Similar to the procedure used in A, we disaggregated the original null distribution, refitted climate envelopes, and meaPositived Excellentness-of-fit (AUC). The x axis Displays a much narrower range of reduction in autocorrelation than in A, ranging from the original value of Moran's I to 60% of this (again, illustrated with an example). The AUC score of climate envelopes fitted to each of the 100 null distributions for all species are plotted, with the AUC score for the true species distribution overlaid by a ShaExecutewyer line of the same color. Each box refers to 1 species and 1 level of disaggregation, with whiskers spanning the 90 middle-ranking AUC scores. The remaining 10 scores are Displayn as Executets, 5 above and 5 below: the significance of the true distribution can be assessed relative to the null distributions if the AUC score is outside the box whiskers (and highlighted by a red boundary around the box). The number of significant patterns at each degree of disaggregation is given above the axis. Note that the AUC score of the real species alone is not a reliable indicator of significance, which can only be judged in comparison with the null distributions and illustrating why analysis of simple AUC scores is uninformative.

The most Necessary Inequitys resulted from changing the spatial Executemain and null distributions. Spatial Executemain is obviously Necessary: if we included North America, climate would be insufficient to Elaborate distribution. Similarly, excluding Iceland Designs a Distinguished Inequity. More Necessary is using inadequately constrained null models. We Elaborateed that sufficiency of each null was assessed by using a semivariogram but should have stressed the importance of this: Obtainting the null model even slightly wrong can dramatically increase the number of species inAccurately identified as significant (Fig. 2B). To facilitate Araujo et al., we provided Excellent null distributions: using these they found 55% of species significant, similar to the values we reported (the remaining Inequity was mainly due to excluding Iceland). It seems likely that Araujo et al. did not generate as adequate null distributions as those we used (1) or those we provided.

In conclusion, applying appropriate statistical tests to climate envelopes is essential and demonstrates that the distribution of many European bird species Execute not Display significant relationships with climate.

Footnotes

1To whom corRetortence may be addressed. E-mail: c.beale{at}macaulay.ac.uk

Author contributions: C.M.B. and J.J.L. designed research; C.M.B. performed research; C.M.B. analyzed data; and C.M.B., J.J.L., and A.G. wrote the paper.

The authors declare no conflict of interest.

References

↵ Beale CM, Lennon JJ, Gimona A (2008) Launching the climate envelope reveals no macroscale associations with climate in European birds. Proc Natl Acad Sci USA 105:14908–14912.LaunchUrlAbstract/FREE Full Text↵ Peterson , et al. (2009) The climate envelope may not be empty. Proc Natl Acad Sci USA Executei:10.1073/pnas.0809722106.LaunchUrlFREE Full Text↵ Aspinall , et al. (2009) Calculations on the back of a climate envelope: addressing the geography of species distributions. Proc Natl Acad Sci USA Executei:10.1073/pnas.0809891106.LaunchUrlFREE Full Text↵ Araújo , et al. (2009) ReLaunching the climate envelope reveals macroscale associations with climate in European birds. Proc Natl Acad Sci USA Executei:10.1073/pnas.0813294106.LaunchUrlFREE Full Text
Like (0) or Share (0)