Classification of Rainfall Regimes over North-Western Australia in Observations and CMIP5 Models

Since 1950, annual precipitation has been increasing in north-western Australia. Recent research has been investigating factors that have contributed to the trend, which have focused mainly on changes in circulation over longer time scales, or changes in the length of the wet season. The work presented here firstly investigates the causes of rainfall by identifying rainfall regimes from daily rainfall patterns. Initially, tropical cyclone days are removed and assigned to one of two regimes based on their location. K-means clustering is used to classify the remaining daily rainfall observations into regimes, before demonstrating their link to known synoptic structures in northern Australia and attributing the trend in rainfall to changes in the frequency and intensity of these structures. Secondly, the ability of CMIP5 models to replicate these regimes is investigated by highlighting differences in the frequency, intensity and location of maxima of rainfall regimes in the models. This is investigated by 3 methods using daily rainfall from the historical runs: projecting model rainfall onto observed clusters to highlight differences in frequency and intensity, clustering model rainfall directly to determine differences in position of the rainfall regimes, and clustering a combination of model and observed days to further highlight differences between the location and frequency of model and observed days. In the observations, six rainfall regimes were identified in addition to the two tropical cyclone regimes. The total of eight regimes fall into four groups: two associated with localised thunderstorms, two associated with tropical cyclones, two associated with monsoon lows and two involving interactions with the mid-latitudes. The main driver of the rainfall trend over the region is found to be an increase in the frequency of tropical cyclones over the Pilbara region of north-western Australia, with an increase in the frequency of monsoon lows and mid-latitude interactions also occurring. The trend is therefore attributed to an increase in the number of organised synoptic structures as opposed to an increase in rainfall intensity associated with these structures. In CMIP5 models, it is found that some synoptic structures are often, but not always, modelled to be in the correct location. However, the models generally produce rain from organised synoptic structures too rarely, although with a higher intensity than is observed. Changes in the regimes in future scenarios will also be examined and discussed to determine whether the north-western Australian rainfall trend will continue.