5B.2
Using the replicate Earth paradigm to interpret and reduce climate change uncertainty

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 25 January 2011: 1:30 PM
Using the replicate Earth paradigm to interpret and reduce climate change uncertainty
608 (Washington State Convention Center)
Craig H. Bishop, NRL, Monterey, CA; and G. Abramowitz

Climate is defined by the probability density function (pdf) giving the probability of observing ranges of values of temperature, wind, rain or even particular environmental phenomena such as tropical cyclones or floods conditioned on, say, greenhouse gas concentrations. If climate were not changing, it could be reasonably approximated by probability distributions of observations. However, in changing climate each observation belongs to a different climate so it's impossible to get an observational record long enough to define the climate pdf. Climate change hides the climate pdf. If the universe contained a large number of Earth replicates experiencing a set of evolving global climate forcing parameters similar to those experienced by our Earth then the climate pdf would be defined by the frequencies with which categories of weather (drought, floods, El Nino, tropical cyclones, etc) occurred on the replicate Earths. In the absence of Douglas Adam's mythical planet where Slartibartfast and colleagues built replicate Earths, replicate Earth's are unavailable. At hand are ensembles of climate forecasts and, in this talk, we discuss how they may be transformed into ensembles of forecasts that, under certain statistical measures, are indistinguishable from an ensemble of replicate Earths. To be specific, the transformation yields pseudo-replicate Earth ensembles whose mean is the minimum error variance estimate and whose time averaged variance equals the time averaged error variance of the mean. We apply this ensemble transformation strategy to the IPCC fourth assessment report models and demonstrate that it affords considerable performance improvements both globally and regionally in out-of-sample experiments.