# Synthetic WS4 Fixture This how-to generates and checks a deterministic synthetic landscape for the Nemora → WS4 build/simulate seam. It is a software integration fixture, not a calibrated forest and not a yield estimate. ## What it produces The fixture combines the existing synthesis modules: - `nemora.synthesis.tessellation` for a seeded Voronoi landscape; - `nemora.synthesis.stands` for fragment attributes; and - `nemora.synthesis.stems` for a placeholder tree list and structure summary. The checked-in fixture lives in `examples/ws4_synthetic/`: ```text examples/ws4_synthetic/ ├── README.md ├── recipe.xml ├── fragments.csv ├── schedule.csv ├── truth_table.csv ├── fixture.json └── landscape.geojson ``` Known properties: - seed `20261026`; - 12 fragments, 36.0 ha total; - ages 40, 60, and 80 (four fragments each); - natural volume curve `(0, 0)`, `(40, 120)`, `(80, 220)`; - expected first-year volume `6623.03115 m3`; - 200 trees per hectare target, placeholder fir structure. `truth_table.csv` is the published per-fragment truth: area, age, volume per hectare, expected volume, tree count, density, mean DBH, mean height, and basal area. `fixture.json` records the seed, generator parameters, totals, and limits. ## Generate The seed is required so the fixture cannot silently become a different landscape: ```bash python scripts/generate_ws4_fixture.py \ --output examples/ws4_synthetic \ --seed 20261026 \ --force ``` The generator also accepts `--count`, `--aspect-ratio`, `--area-ha`, and `--trees-per-ha`. Changing any of these is a different fixture; update the truth table and the docs together. ## Build and simulate With WS4 installed: ```bash ws4 build \ --xml examples/ws4_synthetic/recipe.xml \ --fragments examples/ws4_synthetic/fragments.csv \ --out tmp/ws4_synthetic/compiled ws4 simulate \ --dataset tmp/ws4_synthetic/compiled \ --schedule examples/ws4_synthetic/schedule.csv \ --years 2 \ --out tmp/ws4_synthetic/sim ``` The baseline schedule is empty. At offset `0`, `account_values.csv` should contain: ```text 0,feature.Volume,6623.031150000001 ``` The expected value is the sum of `area_ha * volume_per_ha` in `truth_table.csv`. ## Test ```bash pytest tests/test_synthesis_ws4_fixture.py -q ``` The tests check: - the generator is deterministic from the seed; - the checked-in fixture matches a fresh generation byte for byte; - the truth table reconciles with `fragments.csv` and the curve; - removing the seed or truth table fails validation; and - WS4 build/simulate reproduces the total volume within `1e-6 m3` when WS4 is installed. The WS4 test uses `pytest.importorskip("ws4")`, so a Nemora-only CI run skips the external build step rather than failing for a missing optional tool. ## Realism limits - The landscape is one species, three ages, and no mortality or treatment schedule. - The tree allometry is a placeholder from `nemora.synthesis.stems`; it is not calibrated for taper, biomass, or local yield. - The volume curve is the synthetic WS4 example curve, not a local yield estimate. - The fixture is not a real landscape and is not evidence about a real forest. Use it to catch broken build wiring, seed drift, truth-table drift, and volume accounting regressions. Use a real inventory and a validated yield model for a management decision.