STREM

Supplementary code for the Jousimo et al. paper "A spatio-temporally explicit random encounter model for large-scale population surveys".

Requirements

To speed up the computations, the STREM framework is designed to support HPC clusters. However, the cluster requirements are recommendations and the code can be run in non-distributed systems as well.

• Linux (or possibly other UNIX alike) system HPC cluster
• 32 GB of main memory in each HPC cluster node
• 16 CPUs in each HPC cluster
• 500 GB of shared disk space in the HPC cluster
• 50 GB of local disk space in each HPC cluster node
• R
• Several R packages
• Python
• C/C++ compiler
• GDAL
• GEOS
• GRASS

Installation

Run R and install the devtools package with install.packages("devtools") and the STREM R package with install_github("statguy/STREM"), FMI API package with install_github("statguy/STREM") and follow the installation instructions for the testing version R-INLA here. For the rest of the dependencies, consult the installation instructions of the respective libraries.

Setup

Create configuration file to your git-directory ~/git/STREM/setup/WTC-boot.R and add the following lines:

wd <- "~/STREM" # Replace with data directory
wd.figures <- file.path(wd, "figures")
wd.data <- file.path(wd, "data")
wd.data.raw <- file.path(wd.data, "raw")
wd.data.processed <- file.path(wd.data, "processed")
wd.data.results <- file.path(wd.data, "results")
wd.scratch <- file.path("/tmp/scratch") # Replace with scratch directory for large files
grassLocalTempDir <- "/tmp/STREM" # Temporary directory for GRASS
fmiApiKey <- "XYZ" # Replace with your FMI API key

where wd is the base directory for input and output data and wd.scratch is a scratch directory for storing large files. grassLocalTempDir is a temporary directory for processing rasters with GRASS in a HPC cluster. Replace fmiApiKey with your API key to the FMI open data (not required for simulated data).

STREM uses HPC (=high-performance computing) cluster for parallel computations. There are a bunch of scripts that should be installed from Github to use HPC. The scripts are configured for the author's HPC cluster and a custom module for your HPC cluster should be provided.

External data installation

• Survey routes: copy file from STREM/data/inst/Intersections-simulation-Finland.RData to the directory pointed by the variable wd.data.processed.
• CORINE land cover: Download the raster from http://wwwd3.ymparisto.fi/d3/Static_rs/spesific/clc2006_fi25m.zip, unzip the file and copy the TIFF file to the directory pointed by the variable wd.scratch as HabitatRaster-Finland-cropped.tif.
• Study area boundary polygon: Downloaded automatically from http://gadm.org.

Usage

Scripts to generate simulation data, estimate models and verify estimations are found in the directory STREM/inst/simulations.

Author

Jussi Jousimo, jvj@iki.fi