Comparison of daily percentiles of streamflow and rainfall to investigate stream–aquifer connectivity

A frequency analysis approach was used to investigate the hydraulic connectivity between streams and aquifers, by comparing daily percentiles of streamflow and rainfall. Three Australian streams were examined – a dominantly gaining stream (Wilsons River, NSW), a dominantly gaining stream modified by significant water extraction (Ovens River, Victoria) and a dominantly losing stream (Mooki River, NSW). For the gaining stream examples, a lag is observed between the seasonal peak in the low-flow percentile curves and the seasonal peak in the daily rainfall percentile curve. Cross-correlation was used to calculate the time-shift that provides the best fit between the streamflow and rainfall percentile curves. There is a good correlation (r2 > 0.8) between the reference rainfall percentile curve and the shifted streamflow percentile curves for gaining streams. The lags evident between the rainfall and streamflow percentile curves represent the processes of first replenishing catchment storages (such as soil moisture and groundwater) and subsequent release to the stream. This is largely a function of catchment hydrogeology as well as climate, notably the magnitude and regularity of rainfall events. Catchment size is not a controlling factor. Analysis of these lags provides insights into the dynamics of groundwater recharge, storage and release. Changes in the lag times over the flow percentiles can reflect changes in the dominant catchment storage contributing to streamflow. For the Wilsons River, the contribution from a groundwater system with longer flow paths increases at lower flow percentiles. This can be critical when protecting minimum streamflows, as near-stream groundwater flow may not be the only determining factor. The impact of water extraction can be recognised in this analysis. For the Ovens River, streamflow deficits relative to the rainfall percentile curve correspond to the summer period of high irrigation demand. Such a deficit was also observed for the Wilsons River when only post-development monitoring was used in the analysis. For the losing stream example of the Mooki River, the relationship between the streamflow and rainfall percentile curves was poor due to the significant number of no-flow days and much greater variability in the gauging record. This is because the recharge component of rainfall is not returned as streamflow over the year and a portion of streamflow can become aquifer leakage. Hence, the calculated lags derived from this analysis have no hydrological significance for losing streams.