Technical Comments

Comment on “Climatic Niche Shifts Are Rare Among Terrestrial Plant Invaders”

Science  12 Oct 2012:
Vol. 338, Issue 6104, pp. 193
DOI: 10.1126/science.1225980

Abstract

Petitpierre et al. (Reports, 16 March 2012, p. 1344) conclude that niche shifts are rare for terrestrial plant invaders and that this justifies the use of correlative modeling to project species geographic ranges for biological invasions and climate change. We draw attention to the limitations of their conceptual assumptions and the importance of niche shifts excluded from their analyses.

Ecological niche modeling for invasive species and species range shifts under climate change relies on the ability of the model to project robustly into novel climates (1). This, in turn, depends entirely on the extent to which a species’ range, or part of its range, is defined by abiotic factors that can be revealed by inferential modeling techniques, or by other limiting factors that may not be detectable a priori (2). Petitpierre et al. (3) contend that only expansion to novel climates that are available in the native range (i.e., the available niche space) should be considered unequivocally as niche shifts, assuming that expansion into nonanalog climates [i.e., those beyond the native realized niche (Fig. 1)] is somehow different because it may represent the filling of a preadapted niche. We disagree with this thesis and counter that exaptation [as used in (4)] constrained by biotic interactions or dispersal limitation could just as easily explain niche expansion into available climatic space in the native range as it could expansion into nonanalog climates. At the heart of our disagreement are two different aspects of a species’ Grinnellian realized niche: first, our ability to estimate it by different means, and second, its definition.

Fig. 1

Niche shifts for native (green circle) and alien (brown oval) realized niches of a hypothetical species as defined by two axes of a PCA of climatic variables. Relative occupancy of the realized niches is shown along each axis for the native (green shading) and alien (brown shading) realized niche. Full and 75th percentile lines (solid and dashed, respectively) are depicted for the entire range of conditions available in the native (green) and alien (brown) regions of study as defined for modeling purposes. Niche shift kernel analyses were conducted by Petitpierre et al. using the 75th percentile. Along the axes, “unfilling” (light blue), “stability” (yellow) and “expansion” (orange) regions as defined by Petitpierre et al. are shown alongside additional unfilling (blue) and expansion (red) in nonanalog climates that are not considered by their niche shift kernel technique. The star indicates niche shift to an unoccupied area within all possible combinations of the PCA values for the native realized niche. [Adapted from figure 10.2 in (13) with permission]

A recent review concluded that species’ niches were remarkably stable across a variety of scales (5) but did not go on to conclude that climatic niche expansion was similarly limited. Within its native range, a species is likely to be constrained by both biotic and abiotic factors, and possibly by dispersal limitations (6, 7). When introduced into a new region beyond its native range, it is sometimes the case that species appear to expand their realized climatic niche when assessed using modeling tools. Such expansion can arise when the new location has nonanalogous climates but where the values of the key climatic variables [raw or principal component analysis (PCA) transformed] still fall within the species’ physiological limits. For example, a species that evolved in a Mediterranean climate may thrive in a subtropical climate that was not within or accessible to its native range. Using correlative species distribution models relating species distribution records to Bioclim or PCA-transformed climate variables, any subtropical climates should be outside of the climate space used to train the model and may not be correctly identified as suitable by a model trained on the native range data alone. From a practical perspective, this example may represent a realized niche expansion, and this possibility could be tested using mechanistic models, including data based on the native and alien ranges. The interpretation of this niche expansion is a somewhat separate issue. It could arise because the species was exapted to the novel climatic conditions—its ability to survive the stressful Mediterranean climatic conditions conferring on it an ability to tolerate subtropical ecosystems—or it could depend on other factors, such as the release from natural enemies (8) or a favorable habitat disturbance pattern. Irrespective of the mechanism, the point is that these estimates of niche shift represent an expansion only in relation to our ability to define the shift based on niche modeling tools [including the niche shift kernel method (3, 9)].

In arguing that a shift to only those climates that are available in the native range can unequivocally define evidence of niche expansion when occupied in the alien range, Petitpierre et al. are ignoring the ordinal nature of the PCA-transformed climate space and the ecological law of tolerance (10). One cannot limit the term “niche expansion” to the PCA space (and thus climate space) shared between the native and alien ranges (orange zone, Fig. 1) without also considering the more extreme PCA space occupied by the species (red zone, Fig. 1) as also constituting an expanded realized niche. Furthermore, by explicitly constraining their analyses to the former highly conservative component of niche expansion in PCA space (Fig. 1), Petitpierre et al. actively exclude a critical component of niche shifts when modeling species for invasions and with climate change. When introduced into a new location or with climate change, a species’ niche can expand by both infilling into novel combinations of covariates within the range available to the species and by expansion into novel covariate values beyond those currently experienced (Fig. 1). We regard niche shifts relating to the former situation as more likely to be driven by geographic barriers and nonclimatic abiotic factors, whereas examples of the latter are more likely to be driven by changes in biotic interactions. Furthermore, elements of infilling (star, Fig. 1), as quantified by Petitpierre et al. in their figure 3, are arguably uninformative to understanding the potential niche of the species. Although we suggest that niche conservatism for invasive aliens will be the exception rather than the rule, without careful consideration of model performance in both analog and nonanalog (i.e., interpolation and extrapolation) climate space, it is impossible to assess the ecological plausibility of any model output.

It seems clear to us that the constrained niche shift kernel concept employed by Petitpierre et al. greatly diminished their ability to identify niche shifts for alien species (and native species) under climate change. The Petitpierre et al. study’s concepts of unfilling and expansion may be useful in some contexts, but their limited concept of expansion does not allow them to exhaustively explore whether species niches appear to shift or change in nonanalog environments. Addressing potential niche shifts by alien or native species requires robust models that permit exploration of nonanalog climates (11). We are similarly concerned, therefore, that the subsequent ecological niche models of Petitpierre et al. for alien species were specifically constrained to avoid any extrapolation beyond the training domain (i.e., available niche space in the native range). We argue that their conclusion that ecological niche models are suitable “for the prediction [sic] of both biological invasions and responses to climate change” does not hold, and we find no support for any change in the caution (12) to be extremely wary when using correlative ecological niche models to project potential distributions for alien species or with climate change.

References

View Abstract

Subjects

Navigate This Article