Source code and materials accompanying the paper "Planning cost-effective operational forest inventories" (2024)

This repository contains all source code and files used to draw the conclusions of the paper "Planning cost-effective operational forest inventories" (2024) by Santeri Karppinen, Lovisa Engberg-Sundström, Liviu Ene and Juha Karvanen.

Quick start

1. Clone this repository:

git clone https://github.com/skarppinen/cost-eff-forest-inv.git

2. Install the R programming language (code written using version 4.3.2)

• Install required R packages by executing Rscript install-r-packages.jl in the root folder.

3. Install the Julia programming language (code written using version 1.8.5)

• Install required Julia packages by executing julia install-julia-packages.jl in the root folder.

Description of files

• /data/test-data-4.jld2: the dataset in the paper containing the 100 Swedish tracts.

• /data/planning-problem-data.jld2: the contents of test-data-4.jld2 + the list of inventory decisions whose PoV value needs to be computed to solve the planning problem as discussed in the paper.

• /no-vc/output contains the numerical results of all experiments in the paper as raw (JLD2, Julia's data format) datafiles.

• /src/c/random-sweep contains the implementation of the random sweep algorithm in C. The experiments (see below) call this implementation from Julia. See separate compilation instructions below.

• /src/bash a folder containing SLURM (job scheduler) files (in ZIP files) for running the experiments of the paper (on a computing cluster). These document the arguments with which the experiments of the experiments folder were run in order to produce the results of the paper. Note that all results are available in no-vc/output.

• /src/julia/experiments: Folder containing the source code for launching the experiments of the paper. Each "experiment script" begins with "run-" and has a separate help menu (execute julia run-*experiment* --help on the command line) which lists the available arguments.

• /src/julia/lib: Julia code for calling the C code, reading data, handling jobs, etc. A limited implementation of the random sweep algorithm can be found here as well. Note that the Julia code calling C requires that the associated C code has been compiled to a shared library that Julia can find. (see separate instructions for compiling below)

• /src/julia/scripts/gen-planning-problem-data.jl generates /data/planning-problem-data.jld2 from /data/test-data-4.jld2.

• /src/julia/scripts/gen-task-lists.jl creates the contents of the /src/bash/ folder.

• /src/julia/scripts/figs contains source code to draw the figures of the paper. Each script producing a figure begins with "plot-" and produces a figure to a newly created folder /plots. The remaining files are configuration for drawing the figures.

• /figs contains an additional figure with the tract locations shown on the map, together with the "optimal" decision obtained with the normal model.

Compiling the C code

Running the Makefile /src/c/random-sweep/Makefile with make compiles a shared library libpov.so that contains the implementation of the random sweep algorithm from the paper and a function to evaluate the posterior value of inventory decision (see pov.c and /src/julia/lib/pov.jl for example of calling from Julia). By default, the GNU C compiler gcc is used for compilation. After compiling the library, move the shared library to a folder that the operating system searches when resolving shared libraries (or modify system settings such that it is found).

The C code has two dependencies:

1. The GNU Scientific library (for random number generation)
2. OpenBLAS (or some other BLAS implementation) (for linear algebra)

These dependencies have to be installed and discoverable by the operating system for compilation to succeed. The Makefile uses the flags -lgsl and -lopenblas to link to these libraries.

Contact

Santeri Karppinen (skarppinen 'at' iki 'dot' fi)