Porting notes

Source model

• Primary source: legacy/CoCoNet V3_rubble.nlogo (code section before the first @#$#@#$#@ delimiter).
• Backup or optimized NetLogo variants were reviewed to confirm that logic differences are mostly profiling or refactoring.

Porting approach

• Preserve the NetLogo annual execution flow (setup, go, nested ensemble / year loops).
• Preserve key NetLogo behaviours:
  • reseeding random-seed ((ensemble + 1) * year) in procedures,
  • ordered reef iteration (creation order),
  • one-of, random, random-float semantics,
  • cone / radius dispersal geometry for recruitment,
  • intervention ordering and thresholds.
• Keep output schema aligned with legacy output files.

Configuration in Python

The Python model supports:

1. Legacy parameter CSVs (same style as NetLogo parameter files / output metadata block),
2. YAML config files for modern workflows,
3. Environment variable overrides with the COCONET_ prefix.

You can drive these from the CLI (files and flags) or from the library (load_coconet_config, CoconetConfig.from_file); see Getting started.

Known implementation choices

• Internal state is held in NumPy arrays with explicit reef / site indexing.
• A synthetic NetLogo-style who numbering scheme is reconstructed so site-based remainder who 11 rubble behaviour is retained.
• predate_mort_* variables are carried through output as in the legacy model (site values remain zero unless explicitly changed by logic).

Validation strategy

Recommended parity checks:

1. Run legacy NetLogo and Python with the same parameter CSV.
2. Compare output distributions and time series for:
   • C_*, C_reef, R_reef,
   • S_*, S_manta,
   • E_*, G_*, catches,
   • bleaching / cyclone mortality summaries.
3. Validate intervention scenarios independently (control, seeding, shading, pH, fishing changes).

See also

• Architecture for RNG and state layout.
• Repository README for parameter semantics tied to legacy labels.