Conclusions

This work highlights that trends in streamflow can be difficult to detect, even using a range of methods. The analysis is sensitive to the aggregation scale, the noise in the data and the length of the data series. In this work, only very small downward trends in the streamflow for 13 hydrological reference stations in Australia were observed between 1970 – 2010, and only four stations indicated consistent statistically significant trends across the different methods applied. This is despite evidence of increasing trends in temperature in all catchments, and declining rainfall in the southern part of the continent. While the analyses identified trends in the 1970 – 2010 data, the Mann Kendall analysis under scaling hypothesis did not always identify significant trends, meaning that we cannot yet draw a definite conclusion on the long term decline in streamflow, but can be definite about trends in the 41 years under investigation.
In the catchments where a statistical downward trend was detected, only a small part of the trend could be directly explained by changes in rainfall and temperature. As a result, extrapolation of models calibrated on current datasets might bias future predictions, specifically if the assumption is that the runoff response to rainfall remains stationary.
Currently there is no clear biophysical explanation for the remaining negative trends after accounting for rainfall and temperature changes (the observed amplification). Given that the catchments were chosen based on homogeneous landcover, this suggests that possible vegetation effects impact the observed trends. An alternative explanation is that timing and distribution of rainfall affects the rainfall runoff response. Both these areas will require further research.