Monday, April 04, 2011, 12:00PM - 1:00PM
Meaningful projections of sea level rise – forecasts of sufficient accuracy that engineers and planners can design adaptations within acceptable cost limits – are ideally delivered as probability distribution functions, with the probability of occurrence of any given total sea level rise expressed as a distribution in which the most likely sea level rise (or most likely range of sea level rise) is identified as well as information on outliers or ‘worst-case’ scenarios. Sea level rise projections will thus ideally be delivered not as simple time series but as sequences of probability distribution functions defined for certain times in the future.
Observations of the components of present-day sea level rise are incomplete, and even today we cannot say exactly how rapidly sea level is rising or what the sources of sea level rise are. Our knowledge of the basic physics governing critical sea level rise processes, most notably the potential dynamic behavior of land-based ice, is also incomplete, and numerical models of glacier and ice sheet response to climate forcing have not progressed to the point where robust simulations of land ice contributions to sea level will be operational in the near future. One option for short-term projection is extrapolation using present-day loss rates and rates of change, with the principal emphasis on evaluating the growth of uncertainty during the extrapolation, and on generating distributions of uncertainty. I review the use of this technique in publications over the past 6 years, its potential as a primary tool for sea level forecasts in the near-term future, and the primary weaknesses of this approach.