Easy, fast and reproducible Stochastic Cellular Automata with 'chouca'

Abstract

Stochastic cellular automata (SCA) are models that describe spatial dynamics using a grid of cells that switch between discrete states over time. They are widely used to understand how small-scale processes scale up to affect ecological dynamics at larger spatial scales, and have been applied to a wide diversity of theoretical and applied problems in all systems, such as arid ecosystems, coral reefs, forests, bacteria, or urban growth. Despite their wide applications, SCA implementations are often ad-hoc, lacking performance, guarantees of correctness and poorly reproducible. De novo implementation of SCA for each specific system and application also represents a major barrier for many practitioners. To provide a unifying, well-tested technical basis to this class of models and facilitate their implementation, we built chouca, an R package that translates definitions of SCA models into compiled code, and runs simulations in an efficient way. chouca supports SCA based on rectangular grids where transition probabilities are defined for each cell, with performance typically two to three orders of magnitude above typical implementations in interpreted languages (e.g. R, Python), all while maintaining an intuitive interface in the R environment. Exact and mean-field simulations can be run, and both numerical and graphical results can be easily exported. Besides providing better reproducibility and accessibility, a fast engine for SCA unlocks novel, computationally intensive statistical approaches, such as simulation-based inference of ecological interactions from field data, which represents by itself an important avenue for research. By providing an easy and efficient entry point to SCAs, chouca lowers the bar to the use of this class of models for ecologists, managers and general practitioners, providing a leveled-off reproducible platform while opening novel methodological approaches.