The R package metaRange is a framework that allows you to build process based species distribution models that can include a (basically) arbitrary number of environmental factors, processes, species and species interactions. The common denominator for all models built with metaRange is that they are grid (i.e. raster) and population (i.e. not individual) based.
Install the stable version from CRAN with:
install.packages("metaRange")Or the development version from github with:
devtools::install_github("metaRange/metaRange")Ecologists and programmers that are interested in studying and predicting species distributions and interactions under the influence of a changing environment on an intermediate time scale (years to hundreds of years).
To give users with different experience levels an easy introduction to metaRange, it is accompanied by vignettes / articles that give in-depth explanations and examples on how to setup different types of simulations. That being said, a basic knowledge of general R programming and “classic” object-oriented programming (OOP) concepts as well as some familiarity with geographic (raster) data processing and species distribution models (SDM) might be beneficial. In case one is unfamiliar with any of these topics here are some reading recommendations:
| Topic | Recommended introduction |
|---|---|
| general R programming | Hadley Wickham’s “Advanced R” |
| OOP / R6 package | “Advanced R” Chapter “R6” |
| terra & geodata handling | rspatial.org |
| SDMs | Dormann et al. 2012 (Ref. 5) |
The package supports an unlimited number and type of environmental variables as for example climate data, land-use data, habitat suitability maps or any other kind of data that may influence the species in the simulation. Environment variables can be supplied as raster data (specifically as a SpatRasterDataset (SDS) created by the terra package) (Hijmans 2023). The layers in the SDS correspond to / represent the different time steps of the simulation which enables simulations with varying environmental conditions.
A simulation can contain multiple species, as well as interactions between them. Every simulated species (i.e. each species object) consists of a set of processes that describe its relationship with time, itself, the abiotic environment and other species and traits, which can be any type of data that can be accessed or changed by the processes.
Species have implied populations, each of which inhabit one grid cell of the landscape. On a computational level this means that the traits of a species will in most cases be stored in a matrix with the same size as the landscape it is simulated in, where each value in the matrix represents the trait value of a population.
The processes of all the species are executed in each time step, based on a user defined priority (i.e. the user can choose which process of which species should be executed at what time within a time step). This gives enormous flexibility in the type of simulations the user wants to perform. One can simulate and study single species range dynamics, species interactions, age-structured populations, invasion dynamics, climate and land-use change or a combination of any of those factors on a population / meta-population level.
To provide the user a quick start into building models with metaRange, it includes common ecological functions that can be used to model a variety of species such as kernel-based dispersal, a negative exponential function to calculate such a kernel, environmental suitability estimation based on cardinal values, reproduction models in form of an Ricker model and a Ricker model with Allee effects, as well as metabolic scaling of multiple parameter based on the body mass and temperature.
The performance critical parts of the package have been implemented in C++ with the help of the Rcpp (Eddelbuettel & Balamuta 2018) and RcppArmadillo (Eddelbuettel & Sanderson 2014) packages. The package can handle simulations of a large spatial (millions of grid cells) and temporal extent (> hundreds of years / time steps) on regular consumer grade hardware.
Thanks to mikefc / coolbutuseless for the tutorial on “Modifying R6 objects after creation” and to Tyler Morgan-Wall for the R package rayimage which gave a good example of a Rcpp implementation of an image convolution function.
Eddelbuettel D, Balamuta J (2018). “Extending R with C++: A Brief Introduction to Rcpp.” The American Statistician, 72(1), 28-36. doi:10.1080/00031305.2017.1375990
Eddelbuettel D, Sanderson C (2014). “RcppArmadillo: Accelerating R with high-performance C++ linear algebra.” Computational Statistics and Data Analysis, 71, 1054-1063. doi:10.1016/j.csda.2013.02.005
Chang W (2021). R6: Encapsulated Classes with Reference Semantics. R package version 2.5.1, https://CRAN.R-project.org/package=R6.
Hijmans R (2023). terra: Spatial Data Analysis. R package version 1.7-46, https://CRAN.R-project.org/package=terra.
Dormann, C.F., Schymanski, S.J., Cabral, J., Chuine, I., Graham, C., Hartig, F., Kearney, M., Morin, X., Römermann, C., Schröder, B. and Singer, A. (2012), Correlation and process in species distribution models: bridging a dichotomy. Journal of Biogeography, 39: 2119–2131. doi:10.1111/j.1365-2699.2011.02659.x