Integrating Fundamental Concepts of Ecology, Biogeography, and Sampling into Effective Ecological Niche Modeling and Species Distribution Modeling

Correlative techniques for estimating environmental requirements of species— variably termed ecological niche modeling or species distribution modeling—are becoming very popular tools for ecologists and biogeographers in understanding diverse aspects of biodiversity. These tools, however, are frequently applied in ways that do not fit well into knowledge frameworks in population ecology and biogeography, or into the realities of sampling biodiversity over real-world landscapes. We offer 10 ‘fixes’—adjustments to typical methodologies that will take into account population ecological and biogeographic frameworks to produce better models. Introduction The past 15 years have seen a massive increase in the popularity of techniques that link known occurrences of species with environmental variation across landscapes to estimate ecological niches and geographic distributions, generally termed ecological niche modeling or species distribution modeling (for further discussion regarding this latter term, see below). The literature taking advantage of this novel analytical functionality has increased massively (Figure 1), and two book-length syntheses have now appeared (Franklin 2010, Peterson et al. 2011). Two recent papers have seen massive citation in the field—Elith et al. (2006) has been cited 1050 times, and Phillips et al. (2006) has been cited 842 times (Web of Science, consulted 30 January 2012)—such massive attention in the literature indicates considerable popularity. More importantly, these coarse-resolution summaries of ecology and distribution have been incorporated into the basic ‘toolkit’ of the macroecologist and biogeographer, such that optimizing their use and implementation becomes critical. Many uses of niche modeling in the literature, however, have been rather inappropriate. That is, the computational tools that have been developed for niche modeling are easily used, and frequently have been used in ways that are not in good accord either with what is known of population biology of species, ideas from modern biogeography, or the realities of sampling biodiversity phenomena across real-world landscapes. These misuses, unfortunately, detract from the genuine potential utility of the tools, and cause mistrust and misunderstanding on the part of the broader biodiversity science community. A comparison with another field of biodiversity science is perhaps illustrative. Modern phylogenetics can arguably be stated to have begun with the publication of Willi Hennig’s framework for reconstruction of evolutionary history (Hennig 1950). This thinking framework preceded by 2-3 decades the appearance of computational tools for implementation and use of phylogenetic thinking in systematics—the first software packets for cladistic analyses did not appear until the 1980s (Wiley 1981), and were not broadly available for several years more (Felsenstein 1986, Berlocher and Swofford 1997). As such, the thinking framework for cladistics was available long before the technique was easy to implement. Niche modeling, however, has seen the opposite evolutionary trajectory: tools that, in effect, estimate niches have been around for decades (Nix 1986, Austin et al. 1990, Stockwell and Noble 1992), yet a conceptual framework for the technique has been much slower to appear (Soberón and Peterson 2005, Soberón 2007, Soberón and Nakamura 2009, Godsoe 2010, Soberón 2010, Peterson et al. 2011). We argue that this mismatch between practice and theory has handicapped the development of this emerging field, and has limited the inferences that have been possible. In this contribution, we outline 10 critical considerations that must be taken into account in development of ecological niche models as powerful tools in ecology and biogeography. In each case, the consideration is not widely appreciated in this field; some have appeared in the literature, and others will appear soon. The point, nonetheless, is that, for lack of a solid conceptual framework for the field, key conceptual-to-empirical links have failed, and the field has been handicapped as a consequence.