Sensitivity of streamflow to rainfall and temperature in south-eastern Australia during the Millennium drought

Abstract

The Millennium drought, generally defined from the mid-to-late 1990s until late 2010, is arguably the most severe drought in the historical record in south-eastern Australia (SEA). Reductions in rainfall between the years of 1997-2008 in SEA are comparable in magnitude to the reductions seen during the World War II Drought, and the Federation Drought. However, particularly in the southern Murray- Darling Basin, reductions in runoff during the Millennium drought were comparatively greater than during these two other 20th Century droughts. Increases in maximum air temperatures (Tmax) have been suggested as a primary reason for the severity of this drought. Other causes have also been suggested including changes in rainfall characteristics (e.g. variability and seasonality), loss of groundwater connectivity, as well as changes in land-use practice. We examine the annual rainfall-temperature-streamflow relationships in 34 largely unregulated catchments in SEA. Using a regression approach, we test whether the annual rainfall-streamflow relationship and the annual rainfall-Tmax relationship were significantly different during the drought (1997-2008) compared to the pre- drought data. Approximately two-thirds of the catchments showed evidence of significantly lower streamflow for the same annual rainfall, whereas a little over half of the catchments had significantly increased residual Tmax (i.e. higher annually averaged daily maximum temperatures for the same annual rainfall). However, the spatial coherence between significant test results was low, and a chi-squared test suggests that the incidence of significantly lower streamflow and higher residual Tmax can be considered independent. In other words, the proportion of catchments with significantly less streamflow was not statistically higher in the group of catchments with increased residual temperatures. We develop regression-based estimates of rainfall and temperature elasticities of streamflow, or sensitivities of annual streamflow to annual rainfall and annual Tmax anomalies. Most of the rainfall elasticities of streamflow lie within the range of about 1.5 to 3.0, meaning that a 10% decrease in annual rainfall leads on average to a 15%-30% reduction in annual streamflow, which is consistent with other estimates in SEA. More arid catchments generally have both larger rainfall elasticities and larger uncertainties in these estimates, although all rainfall elasticities are statistically significant. In contrast, only 19 out of 34 catchments have statistically significant temperature elasticities of streamflow; less arid catchments are more likely to have significant temperature elasticities. The mean of the significant temperature elasticities of streamflow is −0.19°C −1 . However, the catchments with larger average annual Tmax anomalies during the drought tend to have smaller (absolute) values of temperature elasticity. In those catchments with average Tmax anomalies during the drought greater than 0.6oC, the mean of significant temperature elasticities of streamflow is only −0.12°C −1 . We obtain estimates of the effect on streamflow of rainfall and Tmax anomalies by multiplying the observed mean annual anomalies during the drought by the long-term rainfall and temperature elasticities of streamflow. There is some variability of these estimates between catchments, but when averaged over large regions, the proportion of the streamflow reduction explained by this method works out remarkably similar. Over all 34 catchments, the catchment area weighted average streamflow reduction is 46% of the long-term mean. Approximately two-thirds of this is explained by the observed reduction in annual rainfall during the drought, and 7% is explained by the increase in annually averaged maximum temperatures. When averaged over all catchments, the estimated effect of a 1oC increase in maximum temperatures is a 5% reduction in streamflow.