Every year at the annual Advanced Research Projects Agency for Energy Summit (which took place this week) you can count on there being at least one left-field idea that just might work—or might go horribly wrong.

Here’s my nominee for this year’s so-crazy-it-might-work award.  On the first day of the conference, Robert Conrado, a senior fellow at ARPA-E, took the stage to describe his idea for addressing a big problem with agriculture and biofuels: plants use a huge amount of water just to stay cool. Only 1 percent of water they take up goes to making the carbohydrates and other materials that make up the plant. The rest, Conrado says, is for temperature regulation. Worldwide, 90 percent of water use goes to watering plants, which of course limits food production. It also limits where biofuel crops can be grown.

The solution Conrado proposed has to do with the fact that plants can use only a narrow part of the solar spectrum to conduct photosynthesis. They can’t use infrared and ultraviolet light, which just go to heating the plant up, forcing it to suck up water to cool off.

Conrado put up a slide showing a peacock, a butterfly, and a strangely colored berry. These all grow microscopic structures that are tuned to interact with specific colors of light, producing brilliant coloration.

He asked, why not engineer plants to grow similar structures designed to reflect the light they can’t use? That would reduce the cooling needs of the plants. Reflecting infrared light would reduce water consumption by one-third and would make it possible to grow plants for biofuels in currently water-constrained areas. Some of the light would reflect back into space, he suggested, helping to offset global warming. (Much of the reflected light would be absorbed by water vapor in the atmosphere, however.)

What could go wrong?

Seriously—I’m curious about what people think of this idea. What should researchers watch out for if they try to do this?