stmove

The goal of stmove is to make more accessible and transparent standard spatio-temporal approaches to interpreting movement data before getting into more challenging aspects of deconstructing movement trajectories. Analyses in this package expect “clean, regular, movement data time series” which consists of a sequence of points (x, y, time) where time = 0, 1, 2, 3,..., T and all missing points have been interpolated and filled in.

For a detailed review of stmove’s motivation and functionality, please see our preprint available on Biorxiv.

Installation

stmove is in active development and not yet available on CRAN. To download the current version of the package use the following code:

If you encounter bugs or have features you would like to see incorporated in future version of stmove, please open an issue.

Usage

The primary function of this package is build_report which, given a clean regularized trajectory, will deliver a .Rmd and .pdf report including the results of the following computations.

Basic path distributions:

  1. Generate step size time series \(S={(t,s(t)) | over [0,T]}\) and plot step-size histogram
  2. Generate turning angle time series \(A={(t,a(t)) | over [0,T]}\) and plot turning angle distribution

Basic path statistics:

  1. If data is sub hourly, plot running means, standard deviations, and auto correlations for s(t) and a(t), plus cross-correlation s(t)-a(t) (basic stats collection: BSC) using a 3 hour (or 6 hour if data is only hourly) “sliding time window” (STW)
  2. Generate and plot 12-hourly BSC using a 12 hour “jumping time window” (JTW)
  3. Generate and plot 14/15 day BSC following a new-full-new moon sequence using a JTW
  4. Generate and plot seasonal BSC for how ever many seasons are available using a JTW

Basic path visualizations.

  1. Plot the trajectory over space
  2. Generate a wavelet plot that allows us to visualize possible periodic components in s(t) and a(t), auto and cross correlation coefficients.

Basic space constructions:

  1. Construct 25%, 50% and 95% home range isopleths (optionally by season) using two methods:
    1. k-LoCoH hull sets
    2. autocorrelated utilization distribution analysis implemented with ctmm::akde

Once these are done, one can then pursue various kinds of analysis that address questions of interest (e.g., GLM models of location and landscape, HMM modeling, step section analysis), but this package strives to set a standard for what is minimally needed before embarking on such analyses.