Spatial Evolutionary and Ecological Vicariance Analysis
The Spatial Evolutionary and Ecological Vicariance Analysis (SEEVA) methodology was first developed by Lena Struwe, Richard Lathrop, Scott Haag, and Peter Smouse at Rutgers University, USA, and Einar Heiberg at Lund University, Sweden.
Further methodological improvements have been developed by Lena Struwe, Peter Smouse, Marcelo Reginato (The New York Botanical Garden), and Donald Walker (Tennessee Tech University).
This website is under revision... please be
patient with us while we update it with our improved new
SEEVA software for Matlab by
Einar Heiberg, is available here: http://seeva.heiberg.se/
SEEVA software for R by
Marcelo Reginato, is available on this website were you are now:
References and sources:
Struwe, L., P. E. Smouse, E. Heiberg,
S. Haag, & R. G. Lathrop. 2011. Spatial evolutionary and
ecological vicariance analysis (SEEVA), a novel
approach to biogeography and speciation research, with an example
Brazilian Gentianaceae. Journal of Biogeography 38: 1841-1854.
Struwe, L., S. Haag, E. Heiberg, &
J. R. Grant. 2009. Andean speciation and
vicariance in neotropical Macrocarpaea (Gentianaceae-Helieae).
Annals of Missouri Botanical Garden 96: 450-469.
Walker, D. M., L. A. Castlebury, A. Y.
Rossman, & L. Struwe. 2013. Host
conservatism or host specialization? Patterns of fungal
influenced by host specificity in Ophiognomonia
Diaporthales). Biological Journal of Linnean Society. 111: 1-16.
Walker, D. M, P. E. Smouse, M. Reginato, L. A. Castlebury, & L. Struwe. Does phyletic radiation in fungal Ophiognomonia (Gnomoniaceae, Diaporthales) show evidence of climatic niche vicariance? (submitted)
SHORT INTRODUCTION TO
Fig. 1. Showing the complex interaction between geography (distribution. space), phylogeny (evolution, ancestry), and ecology (environment, climate) through time for a lineage.
In contrast to vicariance biogeography, which assumes geographic separation of populations, the Spatial Evolutionary and Ecological Vicariance Analysis approach allows researchers also to look at ecological vicariance (differences) of sympatric and allopatric species and clades. This method can utilize GIS-derived dataset of collection-associated ecological and environmental data in combination with phylogenetic data to investigate trends in speciation using statistical methods with spatial interpretations. The method can also be used for other kinds of comparisons between groups and clades, in areas such as coevolution, diseases, morphological evolution, and niche comparisons.
Generally, SEEVA works by using measurements gathered from individuals of species or populations, and these measurements are then analyzed statistically for differences between groups (species) and/or clades. Two statistical test are being employed, the Divergence Index (to measure differences between groups or clades) and Fisher’s Exact test (the latter to provide a p-value for tests with small sample sizes).
Environmental variables (data columns) are divided into categories either as non-ordered, qualitative categories (e.g., soil types) or ordered, quantitative sections representing subsets of the total variation in the variable (e.g., precipitation amounts). A statistical test is performed to investigate if the distribution of species or monophyletic clades in different categories of environmental variables shows a random or non-random pattern, with taxonomic group vs. counts for collections for the categories for one variable in an X x Y multi-way table.
Example of X x Y multiway table showing skewed character state distributions for two different groups using 4 categories (states) for one variable. The numbers inside the table are number of observations, i.e., collections.
A non-random (skewed) pattern is a stronger association for an environmental character state(s) with a specific group/clade, both historically (evolutionary) and presently. The classification of the environmental variables is rather coarse, but these tests provide a way of looking for broad patterns.
The Divergence Index (D) is a measure from 0 to 1 on how skewed a distribution is between two groups or clades, and D is independent of sample-size and can therefore be compared between groups/clades as well as variables. D is also independent from p-values, which indicates if the data distribution at a node is statistically significantly different between the two groups/clades, or not. We suggest that measures are taken to adjust the significance value of the p-value (usually 0.05) to adjust for multiple sampling from the same dataset, using for example the Bonferroni correction.
We have previously used another index of skewness, the Impact Index (I), but the Index of Divergence (D) is superior to I and replaces this. However, the Impact Index is still listed in the result outputs from the SEEVA software (as are chi-square statistics).
The SEEVA method can be enhanced when combined with biogeographic and phylogeographic spatial analysis, ancestral area analysis, dating methods, geographic mapping of populations, endangered species analysis, and ecological niche analysis. The method will work on any kind of data including absence/presence of diseases, morphological or phytochemical measurements, pollinator type, color morphs – as long as individuals are measured.
The species can be grouped in two ways for the SEEVA analysis: 1) species-by-species (Manual Analysis); and 2) by sister clades (Tree-based). The latter approach includes phylogenetic information, since species data from two (or more, if a polytomy) clades will be compared and analyzed, and environmental trends and reactions over the time of the evolution of a group can be assessed by comparing impact values between nodes.
Caveats and notes: The method works with a minimum of one record for each species, however, results based on such a small sample should be evaluated with caution. In general, correlations between divergence and environmental variables can be inferred as trends and tendencies within phylogenetic lineages, and not as the definite cause for the divergence until further research. It should not be assumed that environmental variables are independent, in fact, many of them are not, but an assumption of independency is not necessary for this analysis.
Macro written by Scott Haag for ArcGIS to export environmental data into existing Excel spreadsheets with collection data: link (text file)