Update ENM Script

For testing on the cluster today use sdm1.R provided by your instructor. Save this to your R repository’s ‘R’ folder.

You will later be updating this script to work with higher resolution climate data and the future climate data for climate change projections.

HPC Setup

To run our distribution modeling code on the cluster we first need to interact with the cluster a bit to make sure that everything is set up and working that will be needed for the analysis. This will include making sure the correct R packages are installed, copying scripts from the git repository, and setting up a script to submit the R jobs to the queue.

Log in

Use ‘ssh’ to log in using the terminal of your choice.

ssh yourusername@scc1.bu.edu

Set up R packages

Since the cluster is used by many users in different fields the administrators maintain software modules that are called on the fly by users. This way modules for several versions of a program can be maintained without conflict across the system.

Load R using the command ‘module load’

module load R #call the system module for the current R software

# Run an R environment
R

The cluster R environment may not have everything needed for these analyses. Test the packages that are needed for the modeling script.

In the R prompt run:

library(raster)
library(ggplot2)
library(rasterExtras)
library(RSpatial)
library(spocc)
library(dplyr)
library(ENMeval)

Any of those that fail need to be resolved. ‘raster’, ‘ggplot2’, ‘spocc’, ‘dplyr’, and ‘ENMeval’ can be installed with install.packages().

‘rasterExtras’ and ‘RSpatial’ need to be installed as:

devtools::install_github('rsh249/rasterExtras')
devtools::install_github('oshea-patrick/RSpatial', force=TRUE)

Clone git repo

To copy your up-to-date git repository to the cluster use:

git clone 'https://url.to.github.com/username/repository'

AFTER the initial copy, all other updates can be retreived with:

cd /path/to/repository/on/cluster

git pull

Submitting a Job

Do NOT attempt to run the R script in an interactive R session when logged in to the cluster. To work well within the cluster system we have to submit the R script as a job to the queue.

To do this we need to write a qsub submission script. For our analysis this should look like: link to instructor copy

#!/bin/bash -l
#
#Number of cores requested
#$ -pe omp 12

#Give the name to the job
#$ -N SDM_ENMeval

#Send an email when the job is finished (or aborted)
#$ -m ae

#Join the error and output file
#$ -j y


# Set the runtime limit (default 12 hours):
#$ -l h_rt=12:00:00

# Specify your project
#$ -P ct-shbioinf

module load R/3.6.0

cd /project/ct-shbioinf/rharbert/bio331-geospatial

Rscript R/sdm1.R

Retrieving figure files:

You can use more git commands to retrieve the figure files that were generated. Run the following:

git add figures
git commit -m 'collect figures from HPC'
git push

When prompted, enter your GitHub username and password.

Then in RStudio make sure to ‘pull’ the new version of your repository and check your local ‘figures’ folder.

Homework

Modify the R script provided for HPC modeling to do the future projection assigned for homework last time. Test on the BU cluster using a qsub submission script. Report errors and successes to Slack #general