Page 435

results of greenhouse warming itself, and which might result from attempts to mitigate it, include a shift to a glacial state and major shifts in ocean currents.

Our current models and understanding of geophysical systems do not allow us to predict such effects. Our understanding and modeling have so far not even permitted us to make a map of the possible states of the system. We might require a different modeling approach even to be able to do so.

It can be argued that, in the face of such uncertainty, we should not consider "tinkering" with the only earth we have. However, we are not entirely without understanding of this matter. The principal characteristic of chaos instability, for example, is that the behavior of states with only slightly different initial conditions may be totally different. This is frequently expressed by the statement that "the alighting of a butterfly may change the future of the earth." However, in the sense that we know something of the effects of various kinds of events on parts of the geophysical system, we do know a good deal about this.

For example, we know something of the effect of the dust and aerosols resulting from volcanic eruptions on the climate system and on atmospheric chemistry, and we know something of the effect of industrial sulfur emissions on the climate system. It seems reasonable to assume that mitigation systems that put dust or aerosols into the atmosphere at altitudes and in quantities that are within the bounds of the natural experiments or of previous experiments would not produce instabilities or effects that had not been produced before. This expectation could provide one criterion for use of a geoengineering option: the activity must be within the natural variability of the geophysical system. We could use natural variability, or what are effectively previous experiments, as tests of the stability of the geophysical system and as opportunities to search for possible side effects. However, we must also consider that the chemistry of the atmosphere is changing, particularly from the injection of chlorofluorocarbons (CFCs) and from the increased injection of other greenhouse gases, so past chemistry will be an incomplete guide to the future. We can use the past and our understanding of the nature of the physics and chemistry to guide us in looking for new effects as natural events occur: the next significant volcanic eruption, for example, can be used as an opportunity to extend our understanding of the effects of dust, sulfuric acid aerosol, and chemicals produced by volcanic eruptions on stratospheric chemistry and the climate system.

The possibility would have to be taken into account that a natural event occurring during a mitigation activity could push the system beyond its normal bounds. For example, a large volcanic eruption occurring while artificial volcanic dust was in place might result in a dust loading beyond that previously experienced. Given some knowledge of the statistics and occurrence of eruptions (but noting their current unpredictability in detail)