Siluro-Devonian Communities

Lithofacies and temporal variation predict composition of Siluro-Devonian vertebrate, invertebrate, and plant communities. Includes multiple imputation of community data using chained equations.

lead: Corey Bradshaw
collaborators: Richard Cloutier, Clive Burrett, Mongkol Udchachon, Per Ahlberg, John Long

Abstract

While Siluro-Devonian vertebrate communities have been well-described, co-variation in invertebrate and plant communities from the same period has not been adequately assessed. Using methods from community ecology, we tested how the combined vertebrate, invertebrate, and plant communities from 105 fossiliferous sites around the world varied relative to lithofacies and time. We built logistic linear models to examine variation in the three main communities (vertebrates, invertebrates, plants). We found that the type of lithofacies (6 categories) explained the most variation (% deviance explained [%DE] = 6.4), followed by geological period (4 categories); however, community type had little explanatory power (%DE = 0.05). Variation among most lithofacies expalined variation in presence probability. There was little support for a period×lithofacies interaction. Lithofacies explained most variation among communities when divided by major palaeo-environmental category (marine, alluvial/deltaic, freshwater/estuarine). We then applied a stochastic variant of a permutation analysis of variance to the entire community (10 vertebrate, 20 invertebrate, 7 plant taxa) to examine the relative explanatory power of lithofacies and temporal variation on composition. There was an effect of the lithofacies (R²_perm = 14.0%; p_perm = 0.0002; driven mainly by limestone) and categorical period (R²_perm = 10.3%; p_perm = 0.0001; driven mainly by the Silurian, Middle Devonian, and Late Devonian) (but not their interaction: p_perm = 0.1652) on community composition. Treating time as a continuous variable (age) also demonstrated an effect on community composition (R²_perm = 14.7%; p_perm = 0.0004), with incidence tending to rise in most taxa from 435 to 360 million years ago. Model predictions can now be used to show spatial variation in the relative presence probability for any of the taxa we modelled provided detailed distribution maps of lithofacies are available.

Scripts

• silurodev.R: main code to generate results
• r.squared.R: functions to estimate goodness of fit for linear models
• newlmerAIC_tables3.r: functions to compare linear models
• palaeocoords.ipynb: Jupyter Notebook with Python code to project current-era lat/lon coordinates to palaeo-coordinates, with corresponding plate tectonics, for anytime in the last 1 billion years (modified code supplied by @amer7632; detailed instructions here)

Data

• sildevcomp4.txt: fossiliferous site database
• plate-model-repo: data for hindcasted tectonic reconstruction and generation of palaeo coordinates (maintain sub-folder hierarchy & unzip files in 'Topologies' sub-folder)
• sitecoords: .csv files with site coordinates — all: sitelocs.csv; and divided by major period: Silurian (SIL), Lochkovian (LOCH), Emsian (EMS), Eifelian (EIF), Givetian (GIV), Frasnian (FRAS), Famennian (FAM)

R libraries

performance, sjPlot, lme4, ggplot2, stringr, vegan, parallel, vcdExtra, plyr, dplyr, mice, VIM

GPlately

Install GPlately, a Python interface to accelerate spatio-temporal data analysis leveraging pyGPlates and PlateTectonicTools (references: Merdith et al. 2021 Earth-Science Reviews 214:103477; Mather et al. 2024 Geosciences Data Journal 11:3-10)

Python dependencies

pyGPlates, plate-model-manager, Shapely, NumPy, SciPy, Matplotlib, Cartopy, Pooch, GeoPandas, netCDF4