Joint species movement model (JSMM) R-code, instructions on how to use it and input data from the manuscript Otso Ovaskainen, Danielle Leal Ramos, Eleanor M. Slade, Thomas Merckx, Gleb Tikhonov, Juho Pennanen, Marco Aurélio Pizo, Milton Cezar Ribeiro, and Juan Manuel Morales. Joint species movement modeling: how do traits influence movements?
Abstract. Joint species distribution modelling has enabled researchers to move from species-level to community-level analyses, leading to statistically more efficient and ecologically more informative use of data. Here, we propose joint species movement modelling (JSMM) as an analogous approach that enables inferring both species- and community-level movement parameters from multi-species movement data. The species-level movement parameters are modelled as a function of species traits and phylogenetic relationships, allowing one to ask how species traits influence movements, and whether phylogenetically related species are similar in their movement behavior. We illustrate the modelling framework with two contrasting case studies: a stochastic redistribution model for direct observations of bird movements, and a spatially structured diffusion model for capture-recapture data on moth movements. In both cases, the JSMM identified several traits that explain differences in movement behavior among species, such as movement rate increasing with body size in both birds and moths. We show with simulations that the JSMM approach increases precision of species-specific parameter estimates by borrowing information from other species that are closely related or have similar traits. The JSMM framework is applicable for many kinds of data, and it facilitates a mechanistic understanding of the causes and consequences of inter-specific variation in movement behavior.
We provide here the code for the JSMM (JSMM.html), which consists mainly of one likelihood function and a MCMC sampling scheme with a Metropolis-Hastings step. We also provide a dataset on bird movement in a heterogeneous landscape in Brazil to demonstrate the model – the sequence of movement steps of 43 bird species, their traits and phylogenetic relationship, and the landscape covariates. The code estimates the movement parameters posteriors and their convergence, and produces a figure to show the influence of species traits on species-specific parameters.
The JSMM can be used for many kinds of movement models and movement data by adapting the likelihood function, which should use 3 inputs: the species movement data, the environmental covariates, and the parameters to be sampled. We illustrate it with a simple step-selection model to describe bird movement. As a contrasting example, we show how to fit the model to the data of moths movement, which is based on a spatial capture-recapture study.
The code for the birds case is in the JSMM.html file. Use the link to view the rendered version: http://htmlpreview.github.io/?https://github.com/LEEClab/JSMM/blob/master/JSMM.html. The code for the moths case is in the moths_case/JSMM_fitting_moth_data.html, and the rendered version is here: http://htmlpreview.github.io/?https://github.com/LEEClab/JSMM/blob/master/moths_case/JSMM_fitting_moth_data.html. The code to create the simulated datasets and to fit the model to it is in the "simulateddata_case" folder: https://github.com/LEEClab/JSMM/blob/master/simulateddata_case/simulatedDataset-generate_and_fit.R .