Data from: How density dependence, genetic erosion, and the extinction vortex impact evolutionary rescue
Data files
Nov 01, 2023 version files 103.87 MB
-
alldata_n100_a000_hivar.csv
-
alldata_n100_a000_lowvar.csv
-
alldata_n100_a035_hivar.csv
-
alldata_n100_a035_lowvar.csv
-
alldata_n20_a000_hivar.csv
-
alldata_n20_a000_lowvar.csv
-
alldata_n20_a035_hivar.csv
-
alldata_n20_a035_lowvar.csv
-
bholt_robust_check.csv
-
longsims_n100_a000_hivar.csv
-
longsims_n100_a000_lowvar.csv
-
longsims_n100_a035_hivar.csv
-
longsims_n100_a035_lowvar.csv
-
longsims_n20_a000_hivar.csv
-
longsims_n20_a000_lowvar.csv
-
longsims_n20_a035_hivar.csv
-
longsims_n20_a035_lowvar.csv
-
nloci_robust_check.csv
-
README.md
-
sig2e_robust_check.csv
-
theta_robust_check.csv
-
w_max_robust_check.csv
-
wfitn_robust_check.csv
Abstract
Following severe environmental change that reduces mean population fitness below replacement, populations must adapt to avoid eventual extinction, a process called evolutionary rescue. Models of evolutionary rescue demonstrate that initial size, genetic variation, and degree of maladaptation influence population fates. However, many models feature populations that grow without negative density dependence or with constant genetic diversity despite precipitous population decline, assumptions likely to be violated in conservation settings. We examined the simultaneous influences of density-dependent growth and erosion of genetic diversity on populations adapting to novel environmental change using stochastic, individual-based simulations. Density dependence decreased the probability of rescue and increased the probability of extinction, especially in large and initially well-adapted populations that previously have been predicted to be at low risk. Increased extinction occurred shortly following environmental change, as populations under density dependence experienced more rapid decline and reached smaller sizes. Populations that experienced evolutionary rescue lost genetic diversity through drift and adaptation, particularly under density dependence. Populations that declined to extinction entered an extinction vortex, where small size increased drift, loss of genetic diversity, and the fixation of maladaptive alleles, hindered adaptation, and kept populations at small densities where they were vulnerable to extinction via demographic stochasticity.
README: How density dependence, genetic erosion, and the extinction vortex impact evolutionary rescue
https://doi.org/10.5061/dryad.zgmsbccjj
Simulation code and data from the simulated experiment in the manuscript "How density dependence, genetic erosion, and the extinction vortex impact evolutionary rescue." Data is in .csv
format and code is in .R
format.
Description of the data and file structure
There are two main types of .csv
s with results for main text and a third type .csv
for the robustness analysis. These files are described below:
alldata_n{20/100}_a{000/035}_{hi/low}var.csv
These files are the "main" batch of simulations with all relevant variables recorded. There are 4000 per parameter combination. All trials were run for up to 15 timesteps. The name implies the treatment: initial size (n
) 20 or 100, alpha (a
) 0 or 0.0035, and genetic variation (var
) high or low.
Columns:
trial
trial number (replicate) 1 through 4000 within that treatmentgen
generation (time step); here presented as 1-16 because the initial generation is recorded as1
; our analysis makes the processing step of subtracting 1 from thegen
column such that populations are initialized at step zero and simulations run until at most time step 15n
population size at that time stepgbar
mean genotype in the populationzbar
mean phenotype in the populationwbar
mean intrinsic fitness within the populationpfem
proportion of the population that is femalep.fix.pos
proportion of loci that are at fixation for the positive allelep.fix.neg
proportion of loci that are at fixation for the negative allelev
additive genetic variance within the populationn.pop0
initial size treatment (20/100)alpha
density dependence parameter (0/0.0035)low.var
boolean for if the population is in the low genetic variance treatmentext.gen
the generation at which the population went extinct; populations that did not go extinct have value15
hereextinct
boolean for if the population went extinct during the simulation or not
Note that in this data file, data for a trial is truncated at extinction. That is, the time step in which they went extinct (went to size zero) is not included, and neither are time steps after that. For analysis that treated extinct populations as size zero, we added this data to the data frame.
longsim_n{20/100}_a{000/035}_{hi/low}var.csv
The "long" simulations lasting up to 50 generations. There are 1000 simulated trials per treatment here. Naming conventions are the same here as for the main simulation output described above.
The fields are the same as above, but without the columns ext.gen
and extinct
.
{variable}_robust_check.csv
Output data for the robustness checks. One file for each of the six variables tested (sig2e
, theta
, w_max
, wfitn
, nloci
, and bholt
- the final of these, bholt
is the Beverton-Holt density dependence). Each file contains the eight-treatment simulated experiment, run at several levels of the specified variables.
Fields:
gen
generation/time step (as specified above)n
population sizealpha
density dependence parameter (0/0.0035)n.pop0
initial size treatment (20/100)low.var
boolean for if the population is in the low genetic variance treatmenttrial
trial number - note that unlike in thealldata
scripts, thetrial
is not repeated across treatments heresig.e
(filesig2e_robust_check.csv
only) standard deviation of environmental contribution to phenotypic variance in the trialtheta
(filestheta_robust_check.csv
andw_max_robust_check.csv
only) value of environmental change used in the trialw.max
(filew_max_robust_check.csv
only) value of w_max (W_{max} in the manuscript) (maximum intrinsic fitness) used in the trialwfitn
(filewfitn_robust_check.csv
only) value ofwfitn
(w in manuscript), the variable quantifying the width of the selection surface used in the trialn.loci
(filenloci_robust_check.csv
only) value of number of loci (m in manuscript) used in the trialdelta
(filebholt_robust_check.csv
only) value of delta, the degree of overcompensation used in the trialbeta
(filebholt_robust_check.csv
only) value of beta, measuring strength of density dependence in Beverton-Holt model (see Supporting Information) used in the trial
Code/Software
All simulations were run in R
version 4.1.2. Simulations and analysis use the packages dplyr
(v. 1.0.7), tidyr
(1.1.3), and analysis additonally uses rstanarm
(v. 2.21.2) for the Bayesian modeling. Plots are generated with ggplot2
(v. 3.3.4) and cowplot
(v. 1.1.1) . The robustness analysis was run with the package parallel
(v. 4.1.2) but can be run in serial with modification to the code (changing mclapply
to lapply
).
Each of the .csv
files above is generated by a single R script. Each script begins by sourcing a file (sim_functions.R
) with wrapper functions for implementing the simulations. The scripts then generate a mesh of parameter combinations for the simulations, runs a large batch of simulations, and then aggregates the output into a summary data frame for exporting.
The files take the following forms:
- The main batch of simulations (starting with
alldata_
) are run by the filessim_alldata_n{20/100}_a{000/035}_{hi/low}var.R
- The batch of longer (up to 50 generation) simulations are run by the files
longsim_n{20/100}_a{000/035}_{hi/low}var.R
- The robustness checks are run by the files
robust_check_{varname}.R
Sharing/Access information
All associated data and code can also be found on GitHub at the following address: https://github.com/melbourne-lab/evo_rescue_ndd_erosion. The files for analyzing this data are also stored here.
Contact information
Scott Nordstrom (scottwatsonnordstrom@gmail.com)
Methods
Data enclosed is generated from individual-based stochastic simulation; code to generate and analyze this data is included with the repository.