Four seasons, 1951–2040
Every season is warmer. Spring has changed fastest.
Each disc is a tree section. One ring is one year, from 1951 next to the KG Labs emblem at the centre to 2040 at the edge; the outer band after the gap is the climate models' projection. Each slice is one river basin, read clockwise from the top: Talas, Chu, Issyk-Kul, Tarim · Sary-Jaz, Naryn, Kara-Darya, South Fergana, Alai.
A century of rivers, 1910–2021
Two kinds of river answer heat in opposite ways
Rivers fed mostly by snow peak between April and early summer and run thinner in late summer after a hot season. Rivers with more glacier ice in their basin peak in July and August and hold up in hot summers, because the heat melts ice they will not get back.
Drag from the coolest summers to the hottest: two rivers 50 km apart answer in opposite ways
Each solid line is the average monthly flow of eight summers, ranked by the basin's June–August temperature (ERA5-Land); the dashed line is the usual year. Turquoise: more water than usual; brown: less. Gauge years 1970–2017. Kara-Balta is compared within the periods before and after its 1987 gauge change.
Brown: less water than usual. Turquoise: more. Gauges sorted by when their record begins; the share of the basin under glacier ice is shown after each name. Where a gauge was changed (a break detected in its record), it is compared with its own usual flow before and after the change. Records published so far end in 2017–2021.
The river's year: share of annual flow in each month
Twelve gauges in Chuy, Jalal-Abad and Naryn. Ochre: basins under 2 % glacier; grey: 2–5 %; turquoise: 5–11 %. The snow-fed Jalal-Abad rivers peak in April–June, Kara-Balta in June–July, and the glacier-fed Chuy rivers in July–August. Hover or tap a line to name the river.
No river-flow projection to 2040 is shown. Our glacier model responds to warm summers about twice as strongly as the measured glaciers do, so it has not passed its check.
Rain and snow
Precipitation is where the evidence is weakest, and we show that
The climate grid follows wet and dry years closely, but its long-run level drifts wetter than the rain gauges. On the map, each year's precipitation is therefore compared with its own surrounding 30 years, not with 1951–1980. For the future, the climate models disagree on the direction of change in most seasons.
Precipitation vs 1951–1980, by decade: rain gauges and the grid at the same sites
Projected precipitation change around 2040 (vs 1995–2014)
How this page is built
Five layers under every pixel
The map is the surface of a longer chain. Each value has passed through measurement, a climate grid, climate models, checks against independent data and expert reading before it is drawn.
KG Labs research
AI and climate research for Kyrgyzstan
KG Labs Public Foundation builds the research and computing layer for understanding water, climate and natural hazards in Kyrgyzstan: national datasets, reproducible pipelines, tests of AI and weather models, and briefs for the people who need them.
Figures from KG Labs records, 7–13 October 2026. Image-tile and GPU-training figures are from the September 2026 national run.
Collaborate
Bring water and climate intelligence to your decisions
We work with government agencies, cities, universities, water and energy operators, donors and international programmes on monitoring, data sharing and research.