In their Science article, Alan Robock of Rutgers and colleagues underscore that point. The group has used a NASA computer model to simulate the spraying of a sulfur dose roughly one-seventh as large as Pinatubo, released annually over the Arctic for two decades. (The comparison is messy, but deploying the Pinatubo Option to compensate for expected global warming would use roughly half as much sulfur as was released by the volcano in 1991.) Given the extreme variability in Earth's climate system, it might take an experiment of that size or bigger to get useful data on major side effects, Robock says, and the impacts of such a test would be profound. In the simulation, global rain and snowfall was reduced and the summer monsoon over Africa and Asia was weakened. Some models say that plants, including crops, might thrive with less heat; Robock fears testing the Pinatubo Option could affect "the food and water supplies of 2 billion people."
So one group of scientists argues that by gradually increasing the size of our experiments, we can get as much data as possible with minimal risk. Another says that only a dangerous, full-scale deployment can shed light on the crucial issue of how effective a particular dose will be. The point isn't who's right—it's that countries won't be willing to run these potentially harmful tests if scientists can't agree on how useful they'll be.
Indeed, those disputes might continue even after a given set of tests was completed. In the case of another potential form of geoengineering—adding trace amounts of iron to the ocean to grow algae to suck carbon out of the sky—we've already seen an example of how this plays out. A dozen small-scale experiments to grow algae blooms have been conducted around the world's oceans since 1993, and there's still no consensus among oceanographers as to what the results suggest. Some scientists say a few of the experiments have worked to permanently sequester some carbon in the deep ocean; others say that carbon was recycled back to the surface.
The authors of the Nature and Science papers hope that international agreements or treaties could be put in place ahead of time, to avert ecological harm or political strife. The surest way to minimize risks would be Keith's approach: Conduct internationally-coordinated, modest tests over a decade or more, teasing out the cooling signal over time. But given the inevitable political and scientific conflicts that would arise, such a long series of tests will just give countries more chances to bail on the project. That's why we may not see any medium-scale, informative tests until we're unlucky enough to face with looming, catastrophic changes in the climate. At that point, we'll have no choice but to go all out with a full deployment, with little more than computer-based risk estimates to guide us. From one dangerous global experiment—our current carbon binge—to another.