Modelling impacts of regulation on flows to the Lowbidgee floodplain of the Murrumbidgee River, Australia

Abstract

Summary Major wetland systems have been significantly affected by alteration of flows by dams and subsequent abstraction upstream around the world. Estimating the level of this impact is particularly difficult where there is high flow variability, such as dryland rivers in Australia. Such information remains critical for assessing ecological impacts to ecosystems (e.g. long-lived flood-dependent trees). To determine effects of flow reduction to a large floodplain wetland, we built statistical flow models, integrated flow and flood modelling (IFFM), for the extensive and ecologically important Lowbidgee wetland, supplied by the Murrumbidgee River, based on local annual rainfall and upstream flow data (1880–2010). Large volumes of water are diverted upstream primarily for irrigation and to Australia’s capital city, Canberra, achieved with 26 large dams: Burrinjuck Dam, Snowy Mountains Scheme dams and other upper catchment dams. We identified two periods using structural change analysis, low (before 1957) and high (after 1958) regulated periods; the year marked significant alteration in monthly flow at Redbank gauge, within the Lowbidgee wetland, after which most major dams were built. To determine differences in flow between these periods, we developed two models for three flow gauges on the lower Murrumbidgee River: Hay, Maude and Redbank. The latter two were within the floodplain wetland. The models were based on annual rainfall from stations in the upper catchment and flow, using LOESS and leave-one-out samples without overfitting. This flexible local polynomial regression method was a useful approach to modelling complex processes without theoretical models. The proposed low and high regulated models performed well using the goodness of fit (Nash–Sutcliffe coefficient of efficiency ⩾91.5%) and Kolmogorov–Smirnov test ( p > 0.9999), i.e., there was no significant difference between the observed and fitted distributions of flow data at the Hay, Maude and Redbank flow gauges for both before 1957 and after 1958 periods. And these models were used to extend and predict annual low and high regulated flows from 1880 to 2010, where the Kolmogorov–Smirnov test ( p