# Substriva synthetic sample dataset v1

**SYNTHETIC DEMONSTRATION DATA.** The place, sources, custodians and records are all fictional. The geometry is placed in open water in Bass Strait so it can't be read as a real reserve. Nothing here is a council, agency or real-site record, and no participation or endorsement is implied. Context, not clearance. UNKNOWN remains UNKNOWN.

## Files
- `substriva_synthetic_sample_v1.csv`: one row per record. Line 1 is a `#` notice, so skip it when loading (e.g. `pandas.read_csv(..., comment="#")`).
- `substriva_synthetic_sample_v1.geojson`: the same records as points, plus the fictional place polygon (WGS84, EPSG:4326).

## What the 8 records demonstrate
| Record | Demonstrates |
|---|---|
| 0001 | A published, dated, sourced record with its limits stated |
| 0002 | Date precision kept as "month", not invented to a day |
| 0003 | Reviewed but unpublished; date and material source stay UNKNOWN |
| 0004 | Awaiting second-person review, so not public |
| 0005 | A recorded nil search: "not located" is not evidence that nothing occurred |
| 0006 | The publisher withholds an authoritative record (a valid outcome) |
| 0007 | A correction creates a new version; the original stays in history |
| 0008 | Evidence from a different custodian sits alongside the record and isn't merged into it |

## Schema
| Field | Type | Meaning |
|---|---|---|
| `synthetic` | bool | Always `true` in this dataset |
| `place_id` | string | Managed-place identifier (`SYN-PLACE-001`) |
| `record_id` | string | Stable record identifier |
| `work_type` | enum | `boundary_reference`, `vegetation_works`, `treatment_record`, `mulch_or_material_placement`, `survey_observation`, `correction` |
| `event_date` | ISO date or `UNKNOWN` | When the work or event happened, to the stated precision |
| `date_precision` | enum | `day`, `month`, `year`, `unknown` |
| `source_type` | enum | `document`, `field_observation`, `third_party_dataset`, `none_located` |
| `source_title` | string | Human-readable source (fictional) |
| `source_reference` | string | Source identifier, or `NONE_LOCATED` |
| `retrieved_at` | ISO datetime (UTC) | When the source was retrieved or checked |
| `custodian` | string | Organisation responsible for the source (fictional) |
| `review_state` | enum | `awaiting_review`, `reviewed` |
| `publication_state` | enum | `unpublished`, `published`, `withheld_by_publisher` |
| `uncertainty` | string | What the source does and does not establish |
| `unknown_fields` | list (`;`-separated in CSV) | Fields the evidence does not establish. These stay UNKNOWN |
| `lon`, `lat` | number | Fictional point location (CSV only; GeoJSON uses geometry) |
| `note` | string | What the row is there to teach |

## Rules this data encodes
- UNKNOWN is a value, not a gap to be filled.
- Review is not publication.
- A nil search result is recorded, but it never becomes a claim of absence.
- Corrections are new versions.
- Nothing here establishes safety, treatment status, clearance or causation.

## Licence
Suggested licence: CC BY 4.0, attributed to Substriva (synthetic demonstration data). Confirm before publishing.