Skip to content

About

Moving beyond species-based conservation planning by simulating the effect of diverse local conservation measures on the global preservation of biotic interaction networks.

Topics

Resources

Stars

0 stars

Watchers

1 watching

Forks

Latest commit

 

History

45 Commits

Folders and files

Repository files navigation

Ecological networks for conservation

A simulation framework (Scala 3 / sbt, analysis in Python) asking whether conservation plans that prioritise biotic interactions protect ecological networks better than plans that prioritise species richness.

Each run builds a virtual food web (niche model, 300 species), places populations in a continuous landscape, keeps an interaction only where predator and prey home ranges overlap, and lets energy flow from a fixed basal biomass decide which populations are viable. The landscape is then split into square management tiles; a fraction of them is protected by one of two greedy strategies, primary extinctions hit unprotected populations harder than protected ones, and an energetic cascade follows. The outputs are the percentage of species and of species-level interactions lost.

  • SR — protect the tiles with the highest species richness.
  • IR — protect the tiles with the highest egocentric interaction richness (interactions with at least one partner in the tile).

Supported by the Beca Leonardo de Investigación Científica y Creación Cultural 2024, Fundación BBVA.

Where to start

If you want to… Read
understand the project, its results, weak points and a proposed article narrative docs/report/project_assessment.pdf
know every parameter and the paper configuration docs/MODEL_PARAMETERS.md (top table and §B′)
see why the model is designed as it is, with the evidence docs/MODEL_DESIGN_AND_RESULTS.md
reproduce the sweeps and figures analysis/README.md
read the hand-over notes written before the clean-up docs/ASSESSMENT_GUIDE.md

Layout

src/main/scala/econetcons/   model, runners and checks (current model only)
analysis/paper_sweep/        Python scripts for the paper sweep
analysis/figures/            figures produced by those scripts
data/paper_sweep/            paper sweep outputs (54,000 runs), population-level export, trophic-class sweep
data/emergent_design/        small diagnostic runs behind the parameter choices
docs/                        model documentation, hand-over guide, assessment report (LaTeX + PDF)
legacy/                      earlier model versions: runners, workflows, scripts, notes (see legacy/README.md)

Code map (src/main/scala/econetcons/):

Files Role
World, WorldParameters, Species, Population, PopulationWebGenerator, Utils, SquareGrid metaweb, placement, spatial realisation of links, extinction phases
EnergyFlow emergent energetics (assembly and cascades)
ManagementLandscape, ManagementArea, ConservationStrategy, ConservationParameters, TileSelection, ProtectionStatus tiles, tile scores, protection plans
Observables paired per-strategy outcomes on a shared world
PaperSweep, ProtectedSetExport, TrophicLossSweep runners that produced data/paper_sweep/
AssemblyDiagnostic, WebStructureSweep (only metrics is current) community diagnostics
ModelChecks consistency checks of the model

World.scala and the strategy code still contain the abundance models, cascade rules and strategies of earlier versions behind parameter switches. The paper uses abundanceModel = "emergent" and the two strategies above (PaperSweep.worldParameters, Observables.paperStrategies).

Quick start

Requirements: JDK 17+, sbt 1.10, Python 3 with the packages in analysis/requirements.txt.

# consistency checks (about 3 minutes)
sbt -J-Xmx40g "runMain econetcons.ModelChecks"

# reproduce the summary tables from the stored sweep
python3 analysis/paper_sweep/summary.py

# a small sweep (32 runs, under a minute); its rows are identical to those in data/paper_sweep/paper_sweep.csv
sbt -J-Xmx40g "runMain econetcons.PaperSweep --out /tmp/mini.csv --seeds 2 --cMeta 0.05,0.2 \
     --tileSize 1,4 --fracSteps 2 --pExt 0.3,0.8 --threads 8"

The full paper sweep takes about 4.5 h on 20 threads; commands for every step are in analysis/README.md. Run everything from the repository root.

Conventions that are easy to get wrong

  • Losses are percentages of the community after stabilisation; contrasts are IR − SR in percentage points, so negative means IR loses less.
  • "IR" in current data and figures is the egocentric variant. In legacy/ files IR_* is an older, discarded variant and EIR_* is today's IR.
  • Tile size is given in median home-range diameters.
  • There are no unit tests; ModelChecks and the --verify modes of the exporters are the safety net.

About

Moving beyond species-based conservation planning by simulating the effect of diverse local conservation measures on the global preservation of biotic interaction networks.

Topics

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages