Using assemblage data in ecological indicators: A comparison and evaluation of commonly available statistical tools

Ecological Indicators
By:  and 

Links

Abstract

Ecological indicators are science-based tools used to assess how human activities have impacted environmental resources. For monitoring and environmental assessment, existing species assemblage data can be used to make these comparisons through time or across sites. An impediment to using assemblage data, however, is that these data are complex and need to be simplified in an ecologically meaningful way. Because multivariate statistics are mathematical relationships, statistical groupings may not make ecological sense and will not have utility as indicators. Our goal was to define a process to select defensible and ecologically interpretable statistical simplifications of assemblage data in which researchers and managers can have confidence. For this, we chose a suite of statistical methods, compared the groupings that resulted from these analyses, identified convergence among groupings, then we interpreted the groupings using species and ecological guilds. When we tested this approach using a statewide stream fish dataset, not all statistical methods worked equally well. For our dataset, logistic regression (Log), detrended correspondence analysis (DCA), cluster analysis (CL), and non-metric multidimensional scaling (NMDS) provided consistent, simplified output. Specifically, the Log, DCA, CL-1, and NMDS-1 groupings were ≥60% similar to each other, overlapped with the fluvial-specialist ecological guild, and contained a common subset of species. Groupings based on number of species (e.g., Log, DCA, CL and NMDS) outperformed groupings based on abundance [e.g., principal components analysis (PCA) and Poisson regression]. Although the specific methods that worked on our test dataset have generality, here we are advocating a process (e.g., identifying convergent groupings with redundant species composition that are ecologically interpretable) rather than the automatic use of any single statistical tool. We summarize this process in step-by-step guidance for the future use of these commonly available ecological and statistical methods in preparing assemblage data for use in ecological indicators.

Additional publication details

Publication type Article
Publication Subtype Journal Article
Title Using assemblage data in ecological indicators: A comparison and evaluation of commonly available statistical tools
Series title Ecological Indicators
DOI 10.1016/j.ecolind.2011.06.009
Volume 13
Issue 1
Year Published 2012
Language English
Publisher Elsevier
Contributing office(s) Coop Res Unit Leetown
Description 10 p.
First page 253
Last page 262
Google Analytic Metrics Metrics page
Additional metadata about this publication, not found in other parts of the page is in this table