The continuum of elegance: Insight and ecological judgement in the development of quantitative ecology

Background/Question/Methods

Robert McIntosh played a critical role in the initiation of quantitative ecology, both in creating new methods and establishing criteria for the evaluation of current and future methods. Arguably, few areas of ecology have seen more rapid expansion and growth than quantitative ecology, with developments both from within ecology and adopted from statistics and mathematics. In this paper I address: 1) What are the roots of current methods?  2) What primary paths have the trajectories of development followed?  3) What are the common elements among methods, and what are the distinguishing characteristics?  More importantly, 4) how do current methods meet McIntosh's criteria of "... validity, generality, and prediction ability with respect to ecological reality"?  As McIntosh noted, "... mathematical elegance per se is of little worth to the ecologist."  Following McIntosh's own interests the emphasis will be on ordination/classification of communities, species distribution models, and quantitative community assembly rules.

Results/Conclusions

In general quantitative ecology has increased exponentially in sophistication, fueled in part by increasing computational abilities. In ordination we have seen a progressive development from classical mathematical forms, acclaimed for their rigor, through somewhat ad hoc methods developed by ecologists to new methods better suited to the sometimes peculiar distributions of ecological data. Community classification has received a renewed emphasis from the desire for more effective approaches to conservation and planning. The classical hierarchical methods are showing their age, supplanted by model-driven or optimizing algorithms used in conjunction with mathematical assessments of species discriminatory power. Classification approaches are inherently combinatorial and the infusion of ideas from machine learning have also played a role. Species distribution and community assembly modeling have profited from the rapid development of regression-based approaches through generalized linear to hierarchical models with explicit consideration of the rampant overdispersion of most ecological data. Despite the considerable advancements, McIntosh's admonishment that "... the answer lies in the insight of the ecologist, not in the method. Ecological judgment, has not, as some have feared, become computerized ..." is still valid today. Methodological advancement has sometimes superseded ecological rationale, but we have in fact come a very long way.