Using indices of atmospheric circulation to refine southern Australian winter rainfall climate projections.

Persistent shifts in mean rainfall have wide-ranging impacts to hydrology and water availability, and a reliable set of climate projections of change to mean rainfall is a useful tool for future planning. Most climate models project a decrease in winter rainfall in southern Australia, however there is a wide model range and there is not yet a robust assessment of underlying physical processes that can inform and constrain projections. Here, a multiple linear regression model between indices of atmospheric circulation and gridded rainfall in observations and in CMIP5 climate models is developed for July, representing the peak of winter. The regression is used as an evaluation tool for models and a basis to select models. Spatial distributions of the coefficients from the regression illustrate the relative important of different circulation features for rainfall across the region, and illustrate where climate models have deficiencies. As an additional check of projections, historical years that are an analogue for the projected future mean state of the atmospheric circulation are identified and the rainfall anomaly during those years is examined. Both approaches broadly agree and support previous work in suggesting a constraint on rainfall change to a decrease only. The regression analysis also suggests that the median projection for southwest Western Australia should be revised lower than the median of all climate models. The results demonstrate value in applying statistical techniques to understand relationships of rainfall to circulation and to refine confidence in regional climate projections. [ABSTRACT FROM AUTHOR]