CR 26:233-249 (2004)  -  doi:10.3354/cr026233

ABSTRACT: A statistical method, based on the idea of analogous synoptic situations, is used to study the observed rainfall trends in the SW of Western Australia. The method has been developed and optimised by relating patterns of atmospheric fields (predictors) to station records of rainfall (predictands), for both rain occurrences and daily rain amounts. When tested on observational data, the analogue method is able to reproduce the trend observed globally during the past 50 yr, showing the sensitivity of the statistical approach to changed climatic conditions. The choice of predictor has an important impact on the results: mean sea level pressure (MSLP) and vertically integrated precipitable water (PWTR) are both important in explaining the observed local rainfall trend. The model is then applied to 3 coupled models¹ simulations of the current climate. The improvement gained when using the statistical downscaling tool over direct model grid-average outputs is substantial. Most of the model biases are removed and realistic probability distribution functions at the station level are obtained. The method is then applied to altered climate conditions; the impact of model sensitivities on large-scale responses is quantified using the various coupled models. Future trends for rainfall, estimated by direct climate model outputs, are compared with downscaled projections. A general agreement is found; uncertainties are smaller amongst projections when the downscaling technique is used. All projections agree on a mean reduction of total rainfall in winter and spring. The possible impact of future drying trends induced by global warming is compounded by changes in mean and extreme rainfall, such as long wet spells and large yearly totals.

KEY WORDS: Statistical downscaling · Climate change · Rainfall · Australia