Accounting for skill in trend, variability, and autocorrelation facilitates better multi-model projections: application to the AMOC and temperature time series.

We present a novel quasi-Bayesian method to weight multiple dynamical models by their skill at capturing both potentially non-linear trends and first-order autocorrelated variability of the underlying process, and to make weighted probabilistic projections. We validate the method using a suite of one-at-a-time cross-validation experiments involving Atlantic meridi- onal overturning circulation (AMOC), its temperature-based index, as well as Korean sum- mer mean maximum temperature. In these experiments the method tends to exhibit superior skill over a trend-only Bayesian model averaging weighting method in terms of weight assignment and probabilistic forecasts. Specifically, mean credible interval width, and mean absolute error of the projections tend to improve. We apply the method to a prob- lem of projecting summer mean maximum temperature change over Korea by the end of the 21st century using a multi-model ensemble. Compared to the trend-only method, the new method appreciably sharpens the probability distribution function (pdf) and increases future most likely, median, and mean warming in Korea. The method is flexible, with a potential to improve forecasts in geosciences and other fields.

Figure 8. Probabilistic projections of summer mean maximum temperature change 1973–2005 to 2081–2100 over Korea under the RCP8.5 emissions scenario using “trend” and “trend+var” methods. Vertical lines are the means and the 90% posterior credible intervals.