Hannah Holm of the Water Center at Colorado Mesa University argues that agricultural water conservation doesn’t really save water:

When water is diverted from a stream and put onto the land, part of that water is taken up by plants, part of it evaporates, and part of it makes its way back to the stream. With flood irrigation, a lot of the water diverted from a stream is simply used to push water to the end of the ditch, after which it makes its way back to the stream. Seepage will also eventually return to a stream, in some cases sustaining late season flows. Increasing efficiency through a sprinkler or drip system may require less diversion of water out of the stream to transport water to the plants, but the plants will consume just as much as before.

To actually “save water” that can then be available to other uses, you have to reduce the amount of water that’s actually consumed, either by plants or through evaporation. That means changing to a less thirsty crop, reducing your acreage, or giving your plants less water than they really want — which is likely to lead to lower crop yields. Apart from measures to reduce evaporation and weed growth, there’s not really any way to reduce actual water use and keep getting the same production as before.

Macarena Dagnino and Frank Ward down at New Mexico State University made a technical version of this argument last year (pdf) in the context of subsidies for drip irrigation:

[A]n unexpected result is that water conservation subsidies that promote conversion to drip irrigation can increase the demand for water depleted by crops. Our findings show that where water rights exist, water rights administrators will need to guard against increased depletion of the water source in the face of growing subsidies for drip irrigation.