# Independent aquifer evidence: availability, qualification and next concept

**The strongest supported concept is a connected, heterogeneous alluvial aquifer with low-permeability interbeds that become more important away from coarse fan deposits. It should cross the political border. A precise replacement contact map is not yet qualified.** Natural emergence should follow groundwater pressure and local outlets, not define the material zones itself.

This is an E0 read-only evidence task. No model parameter, geometry or owner decision was changed. `evidence_inventory.csv` identifies each source, geographic coverage, useful properties, limitations and intended use. PDF pages below are one-based file pages, not printed page numbers.

## What is actually available

Measured by `.venv/bin/python docs/records/seepage-storage-aquifer-review-2026-09-28/aquifer_evidence/probe_evidence.py`:

- BGS Appendix3, PDF p62: **25 table rows, 24 unique wells**, across four cross-section groups. **18 rows** report upper-aquifer base depth, spanning **45–170 m**. The table was parsed afresh and ground elevation minus depth checked against reported upper-base elevation. Duplicate well 2538 is not an additional observation.
- BGS PDF p63–66 contains lithological well sections with screen drawings. The table and A–B section were visually inspected. Logs are therefore available; it would be wrong to claim none exist. However, BGS PDF p29 explains that some coordinate pairs are orthogonal projections onto a section. PDF p31 describes a local project grid. A registered basemap alone would not identify every actual well position. The prior `physical_basis/bgs_historical_well_table_probe.json` records the same limitation; this task independently reproduces its counts.
- Tazhiyev 2025 Table2, PDF p5–6 supplies **11 intake coordinates** with geological aquifer classes. All intersect active grid cells; **3** fall within the existing Georgiev reconnaissance screen: **Kordai, Sarybulak, Kainar**. These are recorded source locations, not new observations. The CSV and GeoJSON preserve the table source and classification. They do not imply verified command-area membership, surveyed screens, hydraulic conductivity or pumping-test parameters.
- OIEau lithological classes intersect **6,469 of 6,470 KGZ cell centres**, but only **30 of 10,219 KAZ centres**. In the existing Georgiev screen they cover **122 of 960 centres**. Small cross-border overlaps do not establish KAZ coverage. Attributes supply broad material/age classes, not hydraulic properties or buried-contact elevations.

## What the primary sources support

Read BGS PDF p19–20 and Fig3.3: northward, low-permeability horizons become more numerous/thicker; southward they are sparse or discontinuous and the system approaches a coarse unconfined aquifer. That supports variable connectivity, not one impermeable confining sheet or one fixed layer interface throughout Chu.

Read BGS PDF p25: a bulk transmissivity estimate of 4,000–6,000 m²/day and an assumed vertical/horizontal permeability ratio underpin an illustrative leakage calculation. The ratio and resulting vertical permeability are an estimate, not a basin-wide measured Kv. Read Appendix2 PDF p52: the questionnaire gives south/north transmissivity estimates and ticks storativity >0.1, but the source-detail fields are blank. This is weak context for unconfined drainage storage, not a measured confined-storage coefficient. Do not equate it automatically with numerical-layer specific storage.

Read Morris primary PDF p1 and p3–5: tracer and screen evidence supports induced infiltration and hydraulic connection through the Bishkek system. Its urban/periurban observations cannot locate a KAZ aquitard contact.

Read Tazhiyev PDF p13–14: the Kazakhstan continuation contains alluvial/proluvial deposits with alternating water-bearing and poorly permeable strata; younger riverbed alluvium and deeper Neogene material are distinguished. This independently supports a layered regional system on the KAZ side. Its survey coordinates are useful retrieval anchors, but the inspected intake table reports water demand/reserves and geology, not K/T/storage tests. Administrative reserve figures must not become model conductivities or abstraction observations.

CAIAG monitoring context is available, but screened intervals and consistent head datum remain unresolved. The retained regional synthesis supplies additional literature hypotheses; its detailed numerical priors are not upgraded here to independently qualified spatial evidence.

## Feasibility before geometry

| Possible input | What can be built now | What cannot yet be justified |
|---|---|---|
| Existing native grid and aquifer arrays | Country-independent diagnostic end members; parameter identities and footprints are already measured | Calling a uniform array an observed aquifer |
| BGS local logs/sections | Local stratigraphic table and relative section interpretation | Model-cell contact elevations before grid, section-position and vertical-datum qualification |
| KAZ intake coordinates | Source-backed point layer and targeted requests for original well logs/tests | A continuous confined/unconfined boundary interpolated from aquifer names |
| OIEau map | KGZ lithological context with explicit missing coverage | A full-domain geological partition or copying KGZ contacts into unassessed KAZ |
| DEM and mapped rivers | Terrain/channel-distance covariates and natural-outlet screening, as already used in the seepage investigation | Buried aquitard contacts, saturated thickness or K inferred from surface wetness alone |

## Proposed alternatives for review

**Preferred physical concept: connected alluvial facies with variable vertical resistance.** Keep the same interpretation everywhere: shallow water-table deposits above deeper water-bearing material; coarse fan settings can be strongly connected, while fine-grained plain settings contain stronger interbed resistance. River alluvium may form local high-conductivity corridors where independently supported. Use a numerical interface only as a computational device until logs locate physical units. The material transition is uncertain; no administrative or hand-drawn flood polygon selects it.

For a first defensible spatial candidate, qualify the BGS sections in a local Bishkek pilot and retrieve matched KAZ logs using the coordinate-bearing intakes, including the Georgiev leads. The comparison should bracket plausible interbed connection and thickness, with every observation retaining its well identity, depth/screen, date and datum. A whole-domain interface is premature until those spatial gaps are addressed.

**Immediately buildable diagnostic alternative: border-independent homogeneous end members.** Apply shared bulk-property sets to the same native domain to test how much of the remaining seam depends on inherited regionalization. This is a deliberately simplified null comparison, not the preferred geology or a production recommendation. It can discriminate inherited property jumps but cannot validate a facies map. Use regional literature as broad ranges only; do not calibrate these arrays to the crude flooding polygons. Storage changes should remain separate from geometry changes so their effects can be distinguished.

**Not supported:** nearest-zone extrapolation into KAZ, smoothing existing Kh/Kv solely to erase the visible seam, extending the rough upwelling polygons, or drawing an aquitard contact from irrigation extent. Each would replace a known weak boundary with an unverified one.

## Review recommendation

Review the storage experiment separately. For aquifer structure, accept the common connected/facies-variable concept as the next hypothesis, while keeping spatial qualification open. The most useful next evidence is georeferenced actual borehole locations, screens and lithological intervals on paired fan-to-plain transects crossing the border; local aquifer-test records; and dated shallow/deep heads. Existing literature supports this direction, but does not yet supply a ready-to-adopt whole-domain geometry.

Outputs: `evidence_inventory.csv`; parsed `bgs_appendix3_wells.csv`; `tazhiyev_intake_coordinates.csv/.geojson`; `oieau_cell_coverage.csv`; `receipt.json`; `probe_evidence.py`; inspected PDF-page PNGs. Full text extractions are search aids, not newly qualified source maps.