Is there potential to inform time varying conceptual rainfall runoff model parameters using groundwater data?

Abstract

Declining groundwater levels in easy-to-access shallow aquifers have been observed in many locations. Trends in these types of aquifers are likely to affect the streamflow response, particularly in catchments where baseflow is a significant proportion of the total streamflow. Conceptual rainfall runoff models used to model streamflow typically fail to represent this non-stationarity, which may be caused by long periods of below average rainfall, land use change or over abstraction from the aquifer for consumptive use. To improve the ability of conceptual models to represent these changes, an investigation has been carried out into whether groundwater data can be used to inform time varying parameter values for conceptual rainfall runoff models. Since the late 1990s, a significant change in the rainfall-runoff relationship has been observed in Mosquito Creek, flowing from Victoria into the South East of South Australia. The timing of this change corresponds to a decline in groundwater levels at a number of wells in the catchment, particularly those in the riparian zone, defined as within 1 km of the creeks. The GR4J model was selected to investigate the ability of conceptual models to represent this change in the catchment response, as it is a parsimonious model that explicitly accounts for 'leaky' catchments, and has demonstrated good performance for Australian conditions. Calibration to the period where groundwater records are available indicated that the model under predicted the relatively high runoff period prior to 1996 and over predicted the observed flows during the more recent period. Calibration to a rolling seven year period indicated that capacity of the production store was largely stationary over time. Both the parameters involved in the catchment water exchange were found to change over the period to represent the change in the hydrologic response in the catchment over time. The correlation between these two parameters (x2 and x3) was found to be high (R=0.96), as was the correlation with the catchment averaged groundwater level based on wells in the riparian zone. This relationship suggests that there is potential to relate one or both of these parameters directly to groundwater data, and, as such, represent non-stationary groundwater processes in catchments where baseflow is a significant component of the total streamflow.