Writing in the New York Times (September 19, 2008) Thomas Homer-Dixon and David Keith argue for the importance of settling some of the uncertainty about geoengineering climate solutions by experimentation. They write: “Of course, flooding the atmosphere with man-made particles poses real risks. So to reduce the uncertainty surrounding geo-engineering, research should include real-world tests of various technologies that poke the climate system just a little. At first, tests might use existing research aircraft like NASA’s ER-2, a heavy version of the U-2, to release small payloads of particles and then measure the effects on solar radiation and the ozone layer. If these early tests showed the risks were low, enough material could then be released to have a detectable climate impact, while still keeping the amount substantially less than that needed to offset all human-driven global warming. For the second stage of tests, we might use high-altitude aircraft to deliver a larger quantity of particles at about 65,000 feet in the tropics, which would then be carried much higher and toward the poles by the natural overturning circulation in the stratosphere. The reduction in climate risk from even a small-scale sun-shading scheme could easily be larger than the increase in risk from the scheme’s possible side effects. And in any case the effort would cost only a tiny fraction of the expense of meaningful efforts to reduce man’s carbon emissions.”
So far so good. But before we start down this track it is worth asking just what information we might expect to get to settle the question of whether or not to intervene at a level of intensity and scale that could actually make a difference. The problem is two-fold. In the natural progression of scientific inquiry, mathematical modeling might be expected to be followed by limited experimentation which (if the results merit it) can then be implemented at scale. In the case of aerosols, ‘limited experimentation’ comes in two forms:
1. Geographically limited interventions.
2. Low concentration insertions.
But in both cases, we ought to ask if they are feasible and if so, what sort of information they might be capable of yielding. The first raises issues of safety - could such experimental interventions be contained? The second raises issues of scale – could the results of low concentration interventions be reliably projected for higher level concentrations? These are not killer objections, but they merit attention – if only because they may matter less for some kinds of interventions as compared to others.