Time Series for Decadal Means of Temperature and Precipitation Under Global Warming in the Context of Observed Trends

Probabilistic projections of change in temperature and precipitation over the globe for the decade 2090-2099 relative to a base climate from 1980-1999 have been derived from the CMIP3 simulations for the A1B scenario. By combining these with the best estimate for global warming in intervening years, time series representing the range of forced change for a location can be constructed. This range can be augmented to allow for natural variability of decadal means. The approach is applied to the cases of locations in both the US and southeastern Australia, where temperatures have risen and rainfall declined sharply in the past two decades. It is then extended to depict the full uncertainty in change relative to an unforced base climate assumed to apply up to 1900. By putting these series in the context of the full observational record, a further representation of the absolute values of the quantities is made. This provides a determination of the likely range of the unforced climate itself. A Bayesian theory is applied to the estimation of the local climate sensitivity for Victoria, and also the central Murray-Darling Basin. The theory indicates that the observed values are largely compatible with those from CMIP3, given their statistical uncertainty. Based on the time series, one can predict, for example, that the Victorian temperature for the coming decade has a probability of 0.5 of being even higher than the record value of the past decade. Rainfall is unlikely to be as low as it was then, but has a probability of 0.7 of being lower than the average for the rather moist 1961-1990 period.