# Outlet evidence and observation priorities

E0 reconnaissance; no production boundaries changed. All numerical statements below are measured by `.venv/bin/python scripts/audit_seepage_outlet_evidence.py` and reproduced in `receipt.json`. The audit reads the existing mapping snapshots; it is not a field survey or a new remote-sensing classification.

## Finding

The audited hydrography contains 157 drain-tagged and 502 ditch-tagged OSM ways intersecting the active model. Thus drainage locations are partly mapped, not completely unknown. The source carries line positions and descriptive tags, but does not carry surveyed invert elevations, hydraulic capacities, downstream water levels, or verified groundwater connections. No line has been promoted to a hydraulic model boundary.

| Cell-centre country | Active cells | Cells intersecting river/stream | Cells intersecting drain | Cells intersecting ditch |
|---|---:|---:|---:|---:|
| KAZ | 10219 | 707 | 26 | 28 |
| KGZ | 6470 | 2075 | 102 | 261 |

Counts describe mapping support, not drainage density or completeness. A way can cross several cells; river/stream, drain and ditch counts can overlap. Political country is used only for reporting; the classification rule is identical everywhere. A canal can supply or receive water, and a ditch can irrigate or drain: a tag alone cannot establish either function. No mapped feature means unknown, not absent.

## Georgiev

The union of the existing Georgiev screening polygon and cells whose centres are within 5 km of named Georgiev OSM lines includes 960 active cells. Of these, 14 intersect drain-tagged ways, 11 intersect ditch-tagged ways, and 150 intersect rivers/streams. This is an assumed reconnaissance screen, not a verified irrigation command or drained area. It must not become an automatic drainage mask.

## Terrain and winter-water leads

The inherited KAZ-only terrain inventory has 4827 arcs, of which 151 are marked for winter review. Every receiver-status value is `nearest_only_not_connected`. Proximity to a reach does not establish an outlet, direction, elevation or capacity. This reconnaissance was not conducted symmetrically in KGZ; lack of a KGZ flag is unassessed, not dry. The user-described crude, hand-drawn KGZ polygons are not used in the inventory or in the ranking.

## Reproducible field/review shortlist

`observation_priorities.csv` and its GIS GeoJSON contain 28 stratum-specific entries covering 23 distinct model cells. Repeated cells deliberately retain their different regional purposes. Coordinates are model-cell centres, not measured outlet or well locations. Targets are selected separately for mapped drains, natural channels and cells without mapped channel/drain/ditch, in each country, near the border and around Georgiev. Selection ranks historical model exceedance, with an assumed 5 km spacing within each group and at most two entries per group. Border proximity means distance to an opposite-country active cell centre; it is not a surveyed border distance. No observations have been invented.

At the shortlisted cells:

1. Check source imagery and field access, then locate actual emergence, collectors and their receiving channels. Do not survey the cell centroid merely because it is listed.
2. Measure shallow and deeper groundwater levels on the same dates to distinguish upward flow from locally supplied irrigation water. Repeat during spring recharge and late irrigation-season conditions.
3. Survey local low ground, channel/drain bottom, outlet and downstream water level in a common vertical datum; verify whether an apparent outlet is blocked or ends in a depression.
4. Measure spring/collector discharge and document canal operation and irrigation timing. Capacity and seasonal flow constrain the allowed sink; a visible line alone cannot do so.
5. Compare observed wetness against groundwater, irrigation and river conditions. Permanent free drainage is inappropriate where water must accumulate or is held back by downstream stage.

## Consequence for the candidate

Continue a common natural-seepage sensitivity formulation, explicitly labelled potential discharge. These data do not justify assigning 2.5 m drains to irrigated land or fitting seepage boundaries to the hand-drawn polygons. Use mapped outlets as reconnaissance leads and retain uncertainty in resistance and receiving conditions until hydraulic evidence qualifies them. Existing explicit model boundaries are not newly validated by their coincidence with a mapped line.

## Files

- `cell_outlet_evidence.csv`: complete active-cell inventory, original model quantities, mapping support, coordinates and regional screens.
- `mapped_line_inventory.csv`: intersecting source OSM ways with retained descriptive attributes.
- `mapped_line_cell_intersections.csv`: source identity, cell and clipped length; reproducible join evidence.
- `country_support.csv`: country summaries.
- `observation_priorities.csv` / `.geojson`: investigation targets, not observations.
- `receipt.json`: input identities, hashes, measured counts and interpretation.
- `verification.json`: executed integrity checks.