A comparison of univariate and multivariate methods for analyzing clinal variation in an invasive species

The evolution of clinal variation has become a topic widely studied for invasive species. Most studies of this kind have found significant correlations between latitude and various plant traits, usually using univariate analytic methods. However, plants are composed of multiple, interacting traits, and it is this correlation among traits that can affect how quickly or even whether the populations of invasive plants adapt to their local climatic conditions. We used data from a common garden experiment to determine the possible formation of latitudinal clines in invasive North American populations of Lythrum salicaria L. (purple loosestrife) from the central portion of its invasive range. Analyses were conducted using the more common univariate approach (nested and oneway ANOVAs; linear regression) on individual plant traits (e.g., time to flowering, plant height, various mass measures, and growth rate) and then a multivariate approach (principle components analysis followed by redundancy analysis). Significant among-population differences (P < 0.01) were noted when using both the nested and oneway ANOVAs, and multivariate techniques. However, there were no significant relationships between individual plants traits to latitude when using linear regressions, most likely as a result of the small number of populations used in the study (n = 4). On the contrary, the multivariate analyses showed a significant effect of latitude (P < 0.001) on the invasive populations, but this explained only 4% of the variance; latitude explained 8% of the variance when both invasive and native populations were analyzed. Because of the integrated nature of plant phenotypes, a multivariate approach should provide a clearer and deeper understanding of population responses to changing conditions than univariate techniques.