In Neves-P/DAISIErobustness: Test the Robustness of DAISIE to Geodynamics and Traits

Introduction

What is DAISIErobustness?

DAISIErobustness is an R package for testing of the robustness of the island biogeography model "DAISIE" (Dynamical Assembly of Islands by Speciation, Immigration and Extinction) to more complex and potentially more realistic evolutionary models. Different measures of error of number of species, endemics, non-endemics and evolutionary trajectories are used to determine whether the alternative models can influence the inference capabilities of the current DAISIE model.

DAISIErobustness pipeline

DAISIErobustness consists of a pipeline designed to measure the error one creates when extending the standard DAISIE model with new features. Examples of such new additions include the modelling of island ontogeny, as per the General Dynamic Model¹, sea level changes², and non-oceanic scenarios. The error measure is obtained by simulating and comparing DAISIE data using simulation code that builds upon the existing DAISIE simulations by including geodynamic processes.

Usage

Running the pipeline

The full pipeline can be called by the external function run_robustness(). If the save_output argument is set to TRUE, a folder named results will be created in the current working directory, into which a folder named after the selected parameter space with an .RData file will be saved. This .RData file will contain the results for all the replicates of a given parameter set. If save_output is set to FALSE, the results object will be returned and not saved to a file.

library(DAISIErobustness)
run_robustness(
  param_space_name = "continental",
  param_set = 1,
  replicates = 2,
  distance_method = "abs",
  save_output = FALSE
)

RNG concencerns when batching jobs

While this package can run locally without issues, it was designed primarily to run in batch jobs on HPCCs, in particular the University of Groningen Peregrine HPCC. For this purpose, and assuming a user has access to the Peregrine Cluster, there are pre-prepared submission bash scripts in the bash/ folder. However, care should be taken when running this package (through the run_robustness() function) in a distributed computing setting. The RNG algorithm used (Mersenne-Twister) is seeded within each run by a combination of the system time and the index of the current parameter set. For this reason, if jobs for multiple parameter spaces (which will always share parameter set indices) are started simultaneously, said jobs WILL be running with the same seed! To prevent this, make sure all jobs from a given parameter space have started (and are not on queue) before submitting the next parameter space.

References

¹Whittaker, Robert J., Kostas A. Triantis, and Richard J. Ladle. "A general dynamic theory of oceanic island biogeography." Journal of Biogeography 35.6 (2008): 977-994.

²Fernández‐Palacios, José María, et al. "Towards a glacial‐sensitive model of island biogeography." Global Ecology and Biogeography 25.7 (2016): 817-830.

Neves-P/DAISIErobustness documentation built on May 22, 2024, 4:26 p.m.