Low-flow watch Beta
Each gauge is compared with its own history for the same time of year (2007–2020). Low = below the 10th percentile, Watch = below the 25th. The dry-season outlook gives the chance of being below the 10th percentile in 30 days. It is shown only where it beat both simple baselines (climatology and "same as today") on held-out years 2021 onwards.
| Gauge | Status | Percentile | 10th / 25th pct today | 30-day outlook | Outlook shown | Test Brier: model vs climatology / persistence |
|---|---|---|---|---|---|---|
| ChiangSaen | Watch | 11 | 2.41 / 3.12 m | 22% | Public | 0.154 vs 0.1871 / 0.1915 (n=1212) |
| Vientiane | Low | 9 | 2.95 / 3.31 m | 38% | Public | 0.0105 vs 0.0151 / 0.0239 (n=1212) |
| KhongChiam | Low | 7 | 3.99 / 4.55 m | – | Admin-only | 0.044 vs 0.038 / 0.0621 (n=1212) |
| Pakse | Low | 7 | 2.89 / 3.39 m | – | Admin-only | 0.0348 vs 0.0301 / 0.0464 (n=1210) |
| StungTreng | Low | 9 | 3.82 / 4.31 m | 43% | Public | 0.0014 vs 0.0079 / 0.01 (n=1212) |
| Kratie | Low | 7 | 10.19 / 11.34 m | 48% | Public | 0.0027 vs 0.0089 / 0.01 (n=1212) |
| TanChau | Low | 10 | 1.93 / 2.23 m | 18% | Public | 0.1321 vs 0.2032 / 0.1478 (n=1202) |
| ChauDoc | Low | 10 | 1.83 / 2.05 m | 57% | Public | 0.1232 vs 0.1717 / 0.1371 (n=1202) |
Caveats
- Test days are consecutive and strongly correlated, so the effective sample is far smaller than n.
- Dam operations can change dry-season flows quickly; the model knows nothing about planned releases.
- Not an official drought warning. Data: MRC PMFM public dashboard (daily series not redistributed).
Protocol docs/v85_lowflow_protocol.md, SHA-256 fa7cad6d1ff4ecb9…, locked 2026-11-01T04:14:30+00:00 before fitting.