The entire future of human space exploration may rest on a patch of lunar ice. For the past two years NASA has focused on designing a new crew vehicle and launch system that could return astronauts to the moon by 2018. The agency's ultimate goal is to establish a permanent lunar base and use the program's technology to prepare a human mission to Mars. But the grand plan hinges on a risky prediction: that NASA will find water ice in a permanently shadowed crater basin at one of the moon's poles.

Plentiful ice deposits would be a boon for lunar colonists, who could use the water for life support or convert it to hydrogen and oxygen rocket fuel. And two orbiters sent to the moon in the 1990s, Clementine and Lunar Prospector, found evidence of ice in perpetually shadowed polar areas where consistently frigid temperatures would preserve the water carried to the moon by comet and meteorite impacts. But some scientists have disputed Clementine's radar data, and the anomalous neutron emissions observed by Lunar Prospector could have been caused by atomic hydrogen in the lunar soil instead of ice.

In an attempt to settle the question, NASA plans to launch the Lunar Reconnaissance Orbiter (LRO) in 2008. Traveling in a polar orbit only 50 kilometers above the moon's surface, the one-ton, $400-million probe will train a high-resolution neutron sensor on the suspected ice deposits to determine their locations more precisely. The LRO will also carry a radiometer to measure surface temperatures, an ultraviolet detector to peer into the shadowed crater basins, and a laser altimeter and camera to map the polar regions and to scout possible landing sites.

But because the ice is probably buried and mixed with the lunar dirt, NASA will need to land a probe that could dig up and analyze soil samples. This mission, scheduled for 2011, is a challenging one given that instruments operating in shadowed areas cannot use solar power. The craft could land at a sunlit site and send a battery-powered rover into a dark crater, but the batteries would quickly die. A radioisotope thermal generator could provide electricity using heat from plutonium decay, but NASA is leaning against this option because it is expensive and controversial.

Another idea under consideration is sending a probe that could hop from place to place on the lunar surface by restarting its landing rockets, which could lift the craft up to 100 meters above its original landing site and move it to another spot in the crater basin to hunt for ice. Investigating more than one site is crucial because the ice may be unevenly distributed. Yet another alternative would be to fire ground-penetrating instruments at several places in the shadowed basin, either from a lander at the crater's rim or from an orbiting craft.

The major pitfall of NASA's strategy is the possibility that its probes will find no ice or discover that the ice is too sparse to be a useful resource. "I'm a little worried that they're counting too much on finding water in a usable form," says Wesley T. Huntress, Jr., a former NASA science chief who now leads the Carnegie Institution's Geophysical Laboratory. If extracting ice from the moon proves infeasible, the space agency may have to choose new landing sites and exploration goals for the human missions. But NASA believes the robotic craft will be worthwhile even if they do not find ice. Says Mark Borkowski, head of the Robotic Lunar Exploration Program: "We can't help but get science from these measurements."