Ordination analysis in sedimentology, geochemistry and palaeoenvironment—Background, current trends and recommendations

Abstract

Ordination is the name given to a group of methods used to analyse multiple variables without preceding hypotheses. Over the last few decades, the use of these methods in Earth science in general, and notably in analyses of sedimentary sources, has dramatically increased. However, with limited resources oriented towards Earth scientists on the topic, the application of ordination analysis is at times suboptimal and misuse by authors can occur. This text was written for researchers with little to no experience with ordination with the aim of exposing them to the utility and the pitfalls of this branch of exploratory statistics. To do so, a detailed review of three ordination methods is offered: principal component analysis, non‐metric multidimensional scaling and detrended correspondence analysis. A survey of 163 publications in Earth science is presented, in which these ordination methods were used together with a summary of how, why and on what type of data ordination was used. With common mistakes outlined and misuses in those publications identified. Notably, issues were found with reproducibility, documentation, data set dimensions and transformations. Based on this survey, a recommended workflow is offered for Earth scientists who wish to apply ordination. Additionally, this article is accompanied by highly annotated R scripts for novice users to use these methods.

Expanded North Pacific Subtropical Gyre and Heterodyne Expression During the Mid‐Pleistocene

“…To further display correlations between elements, we generated a correlation matrix following the ordination analysis workflow recommendations from Bialik et al. (2021) on all 11 elements (Figure 7; Taylor, 2021). Correlation matrix results show strong anticorrelation between Ca and the siliciclastic sourced elements (i.e., Al, Si, K, Ti, Fe, Rb, and Zr).…”

“…Additionally, Sr count data have statistically significant positive relationship with Ca throughout the record and previous works have demonstrated it as a carbonate-associated element (Bahr et al, 2014). To further display correlations between elements, we generated a correlation matrix following the ordination analysis workflow recommendations from Bialik et al (2021) on all 11 elements (Figure 7; Taylor, 2021). Correlation matrix results show strong anticorrelation between Ca and the siliciclastic sourced elements (i.e., Al, Si, K, Ti, Fe, Rb, and Zr).…”

Biotic Response of Plankton Communities to Middle to Late Miocene Monsoon Wind and Nutrient Flux Changes in the Oman Margin Upwelling Zone

“…However, a more conservative approach has recently been put forward, recommending that nMDS outputs exhibiting stress above 0.1 should be treated with extreme caution (Bialik et al, 2021). We, therefore, note the overall high compositional similarity of clusters, particularly sub-clusters, which is likely the cause of the high stress in the nMDS.…”

“…Percentage data were then arcsine-transformed before cluster analyses and ordination methods. The arcsine transformation was applied to generate a statistically viable dataset suitable for the applied clustering and ordination methods (Sokal and Rohlf, 1995;Hammer and Harper, 2006;Auer et al, 2014Auer et al, , 2019Bialik et al, 2021) and applies the universal paired group method with arithmetic mean (UPGMA) with Bray-Curtis distance. Cluster stability was further evaluated by using UPGMA clustering with Euclidian distance and Ward's method.…”

Palaeocene to Miocene southern Tethyan carbonate factories: A meta‐analysis of the successions of South‐western and Western Central Asia

“…no difference between Palaeocene and Miocene distribution of carbonate factories). The analysis was carried out following the recommendations in Bialik et al (2021) and normal distribution of the variable was tested using Shapiro–Wilk, Anderson–Darlin, Lilliefors and Jarque–Bera tests. PERMANOVA analysis was then carried out to test the dissimilarity between the data of each epoch.…”