modleR: a modular workflow to perform ecological niche modeling in R

Ecological niche models (ENM) use the environmental variables associated with the currently known distribution of a species to model its ecological niche and project it into the geographic space. Widely used and misused, ENM has become a common tool for ecologists and decision-makers. Many ENM platforms have been developed over the years, first as standalone programs, later as packages within script-based programming languages and environments. The democratization of these programming tools and the advent of Open Science brought a growing concern regarding the reproducibility, transparency, robustness, portability, and interoperability in ENM workflows.ENM workflows have some core components that are replicated between projects. However, they have a large internal variation due to the variety of research questions and applications. Any ecological niche modeling platform should take into account this trade-off between stability and reproducibility on one hand, and flexibility and decision-making on the other. Here, we present modleR, a four-step workflow that wraps some of the common phases executed during an ecological niche model procedure. We have divided the process into (1) data setup, (2) model fitting and projection, (3) partition joining and(4) ensemble modeling (consensus between algorithms). modleR is highly adaptable and replicable depending on the user9s needs and is open to deeper internal parametrization. It can be used as a testing platform due to its consistent folder structure and its capacity to control some sources of variation while changing others. It can be run in interactive local sessions and in high-performance or high-throughput computational (HPC/HTC) platforms and parallelized by species or algorithms. It can also communicate with other tools in the field, allowing the user to enter and exit the workflow at any phase, and execute complementary routines outside the package. Finally, it records metadata and session information at each step, ensuring reproducibility beyond the use of script-based applications.