[mmyrte]

Stacked on #49, which provides the exported single-period allocators and per-run trans_pot_t that the vignette uses. Once #49 merges, retarget this PR to develop.

Removal

Nothing read intrv_meta_t or intrv_masks_t, and intervention machinery isn't meant to live in the package. This PR removes:

• R/intrv_meta_t.R, R/intrv_masks_t.R, their tests and .Rd pages;
• the evoland_db table bindings, the Collate entries and the NAMESPACE exports (as_intrv_*, create_intrv_meta_t).

It also rewords the remaining mentions: util_download, the create_lulc_meta_t / create_pred_meta_t docs (which said "intervention specifications"), the allocator docs, the intro vignette and the README.

Breaking: databases that already hold intrv_* tables keep them in the catalog, but the package no longer binds them.

Vignette: vignettes/policy-mechanisms.qmd

Shows how scenario-specific policy is expressed with the tables the simulation already reads. Each mechanism goes on a run of its own and is compared against a business-as-usual run. It's named after the policy represented rather than the model step, since potentials are adjusted before allocation:

Mechanism	Example in the vignette	Scenario-specific table
Settlement form	compact development (3× mean urban patch size)	alloc_params_t
Spatial steering	building zones (urban potentials ×0.1 outside), nature conservation (urban/arable potentials ×0 inside)	trans_pot_t, after prediction and before allocation
Exogenous change	a reservoir project flooding a strip	lulc_data_t, edited before committing

The vignette also explains:

• why the potential adjustments are multiplicative: they survive the rescaling to the demanded rates, and zero is an exact exclusion;
• that under a fixed demand they relocate change rather than add to it;
• that the whole trans_pot_t slice must be written back;
• the commit → upsert_new_neighbors() order.

A short note maps the terms to SSP-CH's "pre-allocation / allocation / post-allocation" interventions.

It reuses the synthetic landscape from stochastic-allocation-sensitivity.qmd (code folded) and renders with quarto render. On that render:

• the building-zone share of new urban land rises from 0.68 to 0.88;
• conversions inside the protected area fall from 41 to 0;
• compact settlement gives 33 new urban patches (mean 3.3 cells) instead of 82 (mean 1.3);
• the reservoir floods 88 cells.

Tests

tinytest::build_install_test(): all ok, 619 results.

🤖 Generated with Claude Code

https://claude.ai/code/session_019fog72qNWbuSjiGBWuP3np

---

Generated by Claude Code