Rate-of-change analysis in palaeoecology revisited: a new approach

Abstract

Dynamics in the rate of compositional change beyond the time of human observation are uniquely preserved in palaeoecological sequences from peat or lake sediments. Changes in sedimentation rates and sampling strategies result in an uneven distribution of time intervals within stratigraphical data, which makes assessing rates of compositional change and the detection of periods with a high rate-of-change (RoC) or ‘peak-points’ challenging. Despite these known issues and their importance, and the frequent use of RoC in palaeoecology, there has been relatively little exploration of differing approaches to quantifying RoC.Here, we introduce R-Ratepol (an easy to use R package) that provides a robust numerical technique for detecting and summarising RoC patterns in complex multivariate time-ordered stratigraphical sequences. We compare the performance of common methods of estimating RoC using simulated pollen-stratigraphical data with known patterns of compositional change and temporal resolution. In addition, we apply our new methodology to four representative European pollen sequences.Simulated data show large differences in the successful detection of known patterns in RoC peak-point detection depending on the smoothing methods and dissimilarity coefficients used, and the level density and their taxonomic richness. Building on these results, we propose a new method of binning with a moving window in combination with a generalised additive model for peak-point detection. The method shows a 22% increase in the correct detection of peak-points and 4% lower occurrence of false positives compared to the more traditional way of peak selection by individual levels, as well as achieving a reasonable compromise between type I and type II errors. The four representative pollen sequences from Europe show that our methodological combination also performs well in detecting periods of significant compositional change including the onset of human activity, early land-use transformation, and changes in fire frequency.Expanding the approach using R-Ratepol to the increasingly available stratigraphical data on pollen, chironomids, or diatoms will allow future palaeoecological and macroecological studies to quantify, and then attribute, major changes in biotic composition across broad spatial areas through time.