A new application of principal response curves for summarizing abrupt and cyclic shifts of communities over space

Abstract

Abstract. There is a growing need to easily describe and synthesize the dynamics of ecosystems' components in space and time. Most multivariate analyses provide ordination diagrams or biplots that are too cluttered to allow simple reading and are unfamiliar to most users. To overcome such difficulties, a novel application of principal response curves (PRCs) is proposed. Principal response curves are traditionally used to assess treatment effects on community structure measured repeatedly over time. In this new application, the tested factor and the repeated-observation axis are replaced by time and space, respectively. The georeferencing of sampling sites permits to produce an easy-to-read map that summarizes both the temporal dynamics of the community and the contribution of each species to these dynamics at each sampling site. A 24-yr-long time series of scientific surveys monitoring 77 fish and cephalopod species in the Eastern English Channel is used to illustrate the novel application of the PRC method. This new application could prove a relevant tool for the ecosystem approach to human activities management by providing spatialized indicators of community changes, as spatial monitoring tools are increasingly recommended for measuring the effectiveness of management actions.

Effect of insect herbivory on plant community dynamics under contrasting water availability levels

“…Statistically, this method represents a special case of partial redundancy analysis (RDA), assessing the response of community composition to treatments as deviations from a control at each point in time. PRC method allows to better summarize and visualize the partial RDA results compared to traditional ordination biplots (Auber, Travers‐Trolet, Villanueva, & Ernande, ). PRC analysis plots the canonical coefficients of the treatment effects along the first canonical axis of the partial RDA against time, directly displaying the differences in species composition between the treatments and the control at each date.…”

Ecological and life history traits explain a climate-induced shift in a temperate marine fish community

“…A high (3 m) vertical opening bottom trawl (GOV) with a 10-mm-stretched-mesh-size 101 codend is used. The stratified sampling scheme manages to complete 90 to 120 hauls per year 102 depending on weather conditions, and we removed all sites that had not been visited for at least 103 three consecutive years (Auber et al 2017 When combining species' abundances and traits, ecologists generally have two choices: 125 calculate community-weighted mean (CWM) trait values or use the abundances of functional 126 groups. The advantage in using CWM trait values is that continuous data are not broken into 127 categories and thus no information is lost, however examination of the underlying changes in trait 128 values is not possible.…”

“…The PRC analysis was 214 performed using the function prc in the R package vegan. Significant changes in functional 215 structure between the two periods were then tested at each sampling site using Monte-Carlo 216 permutation tests designed to correct for the increase in the family-wise type 1 error rate due to 217 multiple comparisons across sampling sites (see Auber et al [2017] for full details and R code). 218 219…”

“…Auber et al 2017). Despite the utility of PRC for describing community shifts, this method 83 has been largely underused and has not yet been applied to ecological traits.…”

Measuring change in biological communities: multivariate analysis approaches for temporal datasets with low sample size

“…Furthermore, PRC allows for only one ‘treatment’ variable and each analysis focusses on the change in one sample, relative to control samples. However, Auber et al (2017) have recently generalised this method so that it can be used to analyse compositional change in multiple samples, but this application removes the ability to consider an environmental or treatment effect. Thus, a principal response curve analysis yields a line graph showing the relative change in taxonomic composition for each ‘treatment’ along one compositional gradient relative to a control ( Table 2 ).…”