USING THIS SCRIPT

Requires R software.
Re-running optimizations requires a Linux operating system.

Basic usage

  1. Start R.
  2. Set working directory to the folder where you unpacked all the files: menu File | Change dir... or type setwd('full-path-to-unpacked-files')?
  3. Load "run-optimizations.r": menu File | Source R code... or type source('run-optimizations.r')

To produce the plots presented in the article (and others not presented)

all plots are saved as TIFF files under the directory set above
  1. Choose option 1: "Explore results from paper".
  2. Choose one of the runs to plot. The main run that was thoroughly analysed in the paper is optim_020_100_100_1_400g.rdata. The other runs explore the influence on a priori assumptions on results (see below).
  3. Choose the restriction scenario
  4. Choose the type of plot

To examine the influence of a priori assumptions on solutions

  1. Choose option 4: "Influence of a priori assumptions"
  2. Choose option 4: "Compare solutions across different optimizations (signatures and ordination)"
  3. Choose one of the runs:

To run other optimizations

this requires a Linux operating system

All the code to conduct simulation-optimization is written within the original mco package, in a modified version. You must compile the modified version and install it, in order to run optimizations.

Compiling the C code

  1. Open a terminal
  2. Type the commands:

    3. NOTE: you must ensure you have write permissions in the R library directory. If any of the above fails, precede the commands with sudo, to run as root.

Making random base landscapes

you must create random landscapes (only once) prior to run any optimization
  1. Run the steps in Basic usage above
  2. Choose option 2: "Make random landscapes"
  3. Enter number of landscapes to produce
  4. Enter mean property area (main optimization needs a value of 20 here)
  5. Enter variance of property area (main optimization needs a value of 100 here)

    6. An excerpt of each landscape is printed on the screen, for progress and debugging purposes. Landscapes are saved in a .bin file (which may be quite huge).

Running optimizations

any of these will take days to complete
  1. Run the steps in Basic usage above
  2. Choose option 3 (main optimization) or 4 (influence of a priori assumptions).
  3. If choosing 4, then choose an option from 1 to 3. Note that you have to produce the base landscapes needed to run any of these; you'll be warned of missing files.

    4. Optimization results are saved in an .rdata file. You must copy it to Results folder to explore the results.

Examining solutions in detail

this consumes huge amounts of memory
  1. Run the steps in Basic usage above
  2. Choose option 5: "Simulate solutions along time"
  3. Choose one or more solutions to simulate, or type a solution manually. Do not run for more than 3 solutions simultaneously, due to high memory consumption.

Customization

This script, used as provided, is expected to reproduce the results in the article apart from minor variations due the stochastic nature. However, several options can be changed rather easily by editing "run-optimizations.r".

Optimization objectives

6 objectives were used in the article. The first 4 are fixed, the remaining 2 (biodiversity objectives, i.e. minimum area of given age classes) can be changed or even increased. For that, edit the following lines:

maxInterval=100
...
otherpars=list(
   pop=800L
   ,gener=as.integer(c(10, 20, 50, 100, 150, 200, 250, 300, 350, 400))
   ,intervals=as.integer(c(30,60,9999))
   ,minareaTSD=c(0.10,0.10)
   ,minInterval=3
   ,maxInterval=as.integer(maxInterval)
   ,n.years=100L
   ...
)

Where intervals are the limits of age classes (30-60 and >60 years in this case) and minareaTSD the minimum area proportion (relative to the whole landscape) that must be guaranteed during the whole simulation period (10% of each class in this case). You can add more classes and corresponding proportions, each will correspond to an objective. Note that the length of minareaTSD is equal to the length of intervals minus 1. Example:

,intervals=as.integer(c(10,20,30,50))
,minareaTSD=c(0.20,0.10,0.15)

Simulation and optimization options

Changing how many years the landscape simulator runs:

,n.years=100L

NOTE: you must append L to the number to specify that it is an integer

Changing the population and number of generations of the optimization algorithm:

,pop=800L
,gener=as.integer(c(10, 20, 50, 100, 150, 200, 250, 300, 350, 400))

Changing the minimum and maximum values that the prescribed management interval can take:

maxInterval=100
,minInterval=3

(note that the management interval is just the mean of a gamma random number generation)

Files contained

Main directory

/Results directory

/mcomp directory

Many files of the original mco package, plus new source files for the landscape generator/simulator/analyzer