Up until now, we have discussed SST prediction in terms of extrapolating past estimates of either a trend or a current level into the future, and in terms of estimating future changes using the IPCC-based model ensemble. However, like hurricane numbers, SSTs may also experience multidecadal shifts, or oscillations. Note that these types of shifts can not be accounted for in the multi-model ensemble because the natural oscillations in these models are independent from the prescribed forcing and will average out in the ensemble mean. With the SST Climate Shift model we attempt to account for the possibility of a future shift in SSTs. There are five steps to making the SST Climate-Shift predictions (this is our most complex model). The first step is to isolate the multidecadal oscillation in the SST time-series. This is not an easy task as there is a trend in the time-series which must first be removed. It is obvious that removing a linear trend is not the correct thing to do because we expect that at least part of the trend in SSTs is due to anthropogenic forcing which we know has had a non-linear effect on the past record. To deal with this we use the IPCC ensemble mean as an estimate of the historical trend. Although not a perfect estimate, this is certainly better than removing a linear trend and it captures at least some of the non-linearities in the anthropogenic trend. Once this non-linear trend has been removed from the time-series, the second step is to estimate the current state and the low state as we did in section 5.1. This is done using the estimated warm and cold AMO periods from Trenberth and Shea (2006). The third step is to ask the experts to estimate the probability of being in the cold state next year and the probability of an AMO shift in the next five years. Given these values we estimate the probability of a shift for each of the following years and we calculate the average level of this oscillation for the next 5 years. The fourth step is to then add back the non-linear trend that we originally removed. This then gives us a prediction for the SSTs over the next five years. These SST predictions will vary depending on the probabilities that are given by the experts.
Was this article helpful?