Several of Asia's largest rivers begin in ice that is now losing every year.
The immediate effect on flow is the opposite of what most people expect.
Melting ice increases flow for several decades before the supply declines.
therefore pass through a period of apparent that encourages exactly the wrong investments.
Irrigation systems, and cropping patterns are sized for a flow that will not persist.
The timing of the flow matters more than the total volume for most users.
Glacier melt arrives late in the dry season, precisely when the has not yet come and demand is highest.
Losing that contribution shortens the growing season even if annual rainfall is unchanged.
Modelling this is unusually difficult because the ice is in a region with sparse measurement.
Satellite estimates of loss agree on the direction and differ on the rate by a factor that matters for planning.
Governments planning a fifty-year reservoir therefore choose a number from a range and rarely publish which one.
arrangements complicate every response.
Rivers that begin in one state and feed three others cannot be managed by any one of them.
Existing treaties allocate volumes rather than shares of a variable flow, which is the wrong instrument for a shrinking river.
Rewriting them requires trust between states that are usually negotiating about something else.