Statistical characterisation and attribution of recent rainfall and runoff in the Murray-Darling Basin

It is widely recognised that the Murray-Darling Basin (MDB) in south-eastern Australia has been in the grip of a serious drought for many years. Over the period 1997−2006, mean annual rainfall has been approximately 16 percent lower than the long-term (1895−2006) average across the MDB. Runoff has been 39 percent lower over the same period. The continuing drought in the MDB has brought the need for a thorough understanding of how the current drought compares to the longer-term variability of rainfall and runoff in the MDB. In this paper, we: (1) examine the historical variability of annual rainfall and runoff in the MDB using a statistical “changedetection” methodology; (2) quantify, in a probabilistic sense, the severity of the recent dry sequence using average recurrence intervals (ARIs); and (3) quantify the hydroclimatic causes of low runoff in the southern MDB, using the Campaspe basin as a representative example, with the lumped conceptual daily rainfallrunoff model SIMHYD and four scenarios of rainfall and potential evapotranspiration. In terms of annual rainfall and runoff in the MDB, the largest signal is an increase in both rainfall and runoff after the mid-1940s. However, a closer analysis using a multiple change-point test identifies several other dry sequences, notably at the start of the twentieth century (the “Federation drought”), a long period of low rainfall around 1936−1945, and the ten years of 1997−2006. Most of the MDB demonstrates a downward trend in rainfall since 1950, with practically no upward trends in rainfall detectable after the 1970s. Over the period 1997−2006, rainfall and runoff are substantially lower over much of the MDB. This lower rainfall and runoff is predominant in the north-east part of the MDB and the southernmost parts. In the southern MDB, the average recurrence intervals of the reductions in rainfall are mostly 20−50 years, with small areas having ARIs more than 100 years. In the southern MDB, the average recurrence intervals of the runoff reductions over 1997−2006 are much larger than the corresponding average recurrence intervals for rainfall reductions, with large areas of the southern parts of the MDB having ARIs greater than 300 years, which indicates that the recent runoff reduction is larger than expected based on the reduction in annual rainfall alone. Several reasons have been suggested for this including: larger proportional decreases in autumn and winter rainfall; higher temperatures; and less interannual variability of rainfall compared to previous dry sequences. We attribute the lower runoff to these hydroclimatic features of the recent dry sequence, and find that the changed seasonality of rainfall has the largest effect on runoff, followed by historical variability (which includes interannual variability) of rainfall, then higher potential evapotranspiration. Using this attribution methodology we explain 71% of the observed reduction in runoff over 1997−2006. The residual is most likely attributable to interactions between different effects.