When an irrigator applies water to a field, some of it transpires through the plants, and some evaporates. The rest -- the return flow -- seeps across the floodplain or through the aquifer to recharge the river's loss.  One farmer's backwash becomes another's water source.

But from an upstream farmer's perspective, water that evaporates, percolates or runs back to the river is a loss -- one that modern technology is designed to curtail. In 1997, when the Farm Bill began providing financial incentives for efficiency upgrades, Tongue River farmers replaced flood-irrigation systems with center-pivot sprinklers, which promised higher yields and water savings. Sprinklers applied water evenly, and crops flourished. In theory, irrigators were conserving water: The plants absorbed more of what was applied, so they didn't need to take as much from the river.

In practice, it wasn't so simple. Flood a field with 100 units of water, the plant takes 40. Sprinkle it with 80, the plant takes 60. More water stays in the river initially, but more is also consumed in the field. And when less returns to the river, in the long run, the basin sees a loss.

Both Montana and Wyoming know how much water irrigators take, but not how much crops are consuming or what volume returns to the river. Agencies, in turn, regulate the quantity of water a user diverts from a source, not the amount consumed. "If water rights were based on water depleted rather than water applied, then there wouldn't be this problem," says Frank Ward, professor of water resource economics at New Mexico State University. "But depletion is really tricky to measure."

Sprinklers have grown popular in both Montana and Wyoming. But the federal subsidies that lubricated the switch were given with different conditions. In Montana, farmers must agree to return saved water to its source or take less from the river. In Wyoming, if farmers have water left, they can use it on a field they may have neglected during drier years, leaving less for downstream users.

"A general rule in water law is that you can't force someone to not improve efficiency," says Cosens. "If they do a better job on the acres they irrigate, you can't do anything about it." That's been true within states. But now the Supreme Court has stretched this rule across borders; if Wyoming farmers have a right to a certain amount of water, who is to tell them they can't consume it all?

Montana attorneys are at least relieved to have won on coalbed methane, which poses a greater water sharing problem than sprinklers. A year ago, near the Powder River headwaters in Wyoming, L.J. Turner's well went dry; the aquifer beneath his ranch had dropped five feet. Nearby, the drawdown was more than 600 feet. "In the summertime, you used to hear the frogs crying for rain." he says. "Now there isn't anything." There are more than a dozen coalbed methane wells on Turner's ranch. Down slope, a tawny haze hangs over Campbell County's gas rigs. Drilling began here in earnest in the mid-'90s, and in the last few years, companies have pumped between 20 and 30 billion gallons of groundwater annually.

If groundwater pumping reduces surface flows, the special master concluded, then those wells must be considered post-1950 uses under the Yellowstone River Compact. But in order to force Wyoming to curtail groundwater pumping for coalbed methane in times of scarcity, Montana will have to prove that Wyoming's deep aquifers do, in fact, feed its rivers. That won't be easy. Groundwater flow, guided by local geologic formations, is tricky to map. And water takes time to move underground, making it difficult to demonstrate cause and effect.

In the absence of data, the debate has pivoted more on faith than fact. "My sense is that coalbed methane aquifers are deeper aquifers. I do not believe that they're affecting surface water flows," says Harry LaBonde, Wyoming's deputy state engineer. His logic may stem from the fact that water law and hydrologic science have not advanced at an equal pace. Most Western states still appropriate surface and groundwater separately, under the assumption that aquifers are disconnected from streams -- an idea hydrologists dispute. But even Wyoming is starting to see flaws in this thinking. "As water tables drop more and more, that's when conflicts arise between users," says LaBonde. "And that's when we become very concerned."