Abstract
Recent research has shown that decadalโtoโmultidecadal (D2M) climate variability is associated with environmental changes that have important consequences for human activities, such as public health, water availability, frequency of hurricanes, and so forth. As scientists, how do we convert these relationships into decision support products useful to water managers, insurance actuaries, and others, whose principal interest lies in knowing when future climate regime shifts will likely occur that affect longโhorizon decisions? Unfortunately, numerical models are far from being able to make deterministic predictions for future D2M climate shifts. However, the recent development of paleoclimate reconstructions of the Atlantic Multidecadal Oscillation (AMO) (Gray et al., 2004) and Pacific Decadal Oscillation (PDO); (MacDonald and Case, 2005) give us a viable alternative: to estimate probability distribution functions from long climate index series that allow us to calculate the probability of future D2M regime shifts. In this paper, we show how probabilistic projections can be developed for a specific climate modeโthe AMO as represented by the Gray et al. (2004) treeโring reconstruction. The methods are robust and can be applied to any D2M climate mode for which a sufficiently long index series exists, as well as to the growing body of paleoโproxy reconstructions that have become available. The target index need not be a paleoโproxy calibrated against a climate index; it may profitably be calibrated against a specific resource of interest, such as stream flow or lake levels. Copyright ยฉ 2006 Royal Meteorological Society