“Aral System” is a public service for monitoring the Aral Sea’s decline and for early warning about dust storms across the whole Aral region. It covers 10 settlements in the three coastal countries and the exposed seabed: Uzbekistan, Kazakhstan, Turkmenistan. Two rivers with different fates feed the sea: the Amu Darya feeds the Greater (Southern) Aral, which keeps drying up, while the Syr Darya feeds the Lesser (Northern) Aral, restored by the Kokaral Dam. The rivers themselves rise in the upper basin, in the mountains of Tajikistan and Kyrgyzstan. The system combines open satellite and model data, ML forecasts (PM10, mirror area, anomalies) and an analytical LLM summary with a deterministic fallback.
- Open data
- In-house ML models
- AI interpretation
Disclaimer. This is a demonstration project, not an official warning system. Data comes from open model sources and may differ from ground measurements. For safety decisions, rely on your official hydrometeorology and civil-protection services.
| Data | Source | Note |
|---|---|---|
| PM10 / PM2.5 / dust / AOD | Open-Meteo Air Quality API (CAMS) | 10 / 31 days |
| Wind, temperature, humidity, precipitation, soil moisture | Open-Meteo Forecast API | — |
| Amu Darya & Syr Darya discharge (m³/s) | Open-Meteo Flood API (GloFAS) | Амударья · у Нукуса · Сырдарья · у Казалинска |
| Climate projection to 2050 | Open-Meteo Climate API (CMIP6) | MRI_AGCM3_2_S |
| Sea mirror area 2005–2018 | Landsat · open-data-kazakhstan | Open dataset |
| Coastlines and contours | Natural Earth | Vector contours |
- PM10 statuses. Thresholds: watch from 55, warning from 155, emergency from 355 µg/m³. A feed event is created automatically on ingest — from the “warning” status.
- ML PM10 forecast. A short-term model on hourly air-quality history with a confidence band; anomalies — outlier detection (Isolation Forest) on the same history.
- Sea area. History from Landsat, projection to 2030 — a regression with an uncertainty band; the map contour is an illustrative interpolation between reference geometries.
- Dust risk. Topsoil moisture of the exposed seabed below 0.08 m³/m³ counts as a “dry seabed” — in strong wind the risk of salt-dust uplift is elevated.
- Discharge → area. The link between river inflow and sea area is shown as a risk factor (narrative), not a hydrological calculation: Amu Darya → Greater Aral (decline), Syr Darya → Lesser Aral (recovery via the Kokaral Dam).