*2.4. Hydrogeology*

The regional water table occurs in Loxton Sands unit, which is in direct hydraulic connection with the alluvial floodplain aquifers that interact with the river. Groundwater in the Loxton Sands is highly saline (7000–50,000 mg/L). The Murray River is generally gaining along this reach, with groundwater (and salt) entering the river and floodplain by lateral flow from the Loxton Sands and some upwards leakage from deeper units in the Murray Basin [21].

In the areas of irrigation, the vadose zone is characterised by three layers in order of depth: loamy sands of the Woorinen Formation, the Blanchetown Clay (which may or may not be present), and the unsaturated upper sections of the Loxton Sands.

#### *2.5. Irrigation and Drainage*

The region is well suited to irrigated horticulture, having abundant sunshine, suitable soils, and access to surface water. Irrigation along the Murray River in South Australia commenced in the 1890s, with more intense development after the Second World War [25]. Most of the development in Loxton occurred in 1948 with the establishment of the Loxton Irrigation Trust, a government-sponsored initiative [26]. The Bookpurnong area was established more recently, from the 1980s onwards, as a private initiative.

The e fficiency of irrigation has improved over time [25]. Prior to the 1970s, irrigation practices were either by flood or furrow irrigation with water supplied on a set roster. This has shifted progressively to sprinklers and drippers, and the pressurisation of o ff-take systems from the river. This has resulted in lower rootzone drainage rates relative to the application rates. The author of [26] estimated irrigation efficiency factors at decadal intervals since irrigation development. The estimates were based on a combination of anecdotal evidence from early time periods, as well as some selected irrigator surveys and benchmarking studies for particular crop types and application methods from later time periods [26]. The estimates are more accurate since 2000 but are uncertain for all time periods.

Rootzone waterlogging became evident soon after the development of irrigation, due the formation of perched water tables on the Blanchetown Clay. Sub-surface drainage systems were installed to manage the issue.

At Loxton, a comprehensive drainage scheme (CDS) was installed. This features a series of gravity flow pipelines that receive water from tile drainage networks on each property and delivers this water to a series of open bottom caissons, where water collects and is pumped to the floodplain for disposal. Data have been collected on the volumes of water intercepted. Drainage volumes increased until the late 1990s and have steadily declined since.

In Bookpurnong, drainage issues do not occur as commonly as at Loxton. About 30% of the irrigated area requires drainage and this is confined mostly to the southwest of the district. The drainage of these properties has been managed privately, in which a tile drainage network transfers water to sumps that are pumped from for disposal. There are no data available on the cumulative drainage volumes.

#### *2.6. Groundwater Returns and Salinisation*

Irrigation has caused recharge rates to increase substantially from background levels. Prior to European settlement, the region was covered by native Mallee vegetation that allowed very little water past its rootzone [27]. The clearing of native vegetation for dryland agriculture resulted in recharge rates increasing from 0.1 to 10 mm/year, and irrigation development has resulted in recharge rates exceeding 100 mm/year [28].

The higher recharge has led to the development of a groundwater mound within the Loxton Sands aquifer, resulting in increased salt loads to the river and the floodplain (see Figure 2). The lateral flow component from the Loxton Sands is thought to be the main mechanism for salt entering the floodplain and the river, and a network of dedicated pumping wells (a salt interception scheme, SIS) was installed in 2006 to actively reduce the groundwater returns to the river.

**Figure 2.** Conceptual model of groundwater returns to the Murray River from the Loxton and Bookpurnong irrigation districts, sourced with permission from [29]. Note that the Agronomic Water Balance used in this paper does not include irrigation drainage, which is addressed as an output of the transfer function.

As illustrated in Figure 2, the quantum of irrigation recharge is the key driving force for groundwater returns, and there are numerous processes that can influence it. These can be separated into three domains: (1) an agronomic water balance (AWB) that occurs at the surface/rootzone and defines the rootzone drainage rate; (2) the deeper vadose zone where variably saturated conditions, including perching, occur; and (3) the saturated zone.

#### *2.7. Existing Methods to Model Groundwater Returns*

The salinity managemen<sup>t</sup> framework requires that groundwater returns (and their salt loads) to the river be quantified over history and into the future on the basis of past and present actions that include irrigation development, the implementation of efficiency measures, and SIS pumping. A 3D MODFLOW groundwater model [30] has been developed for this purpose and is updated over time to provide more refined salt load estimates and to account for any changes in managemen<sup>t</sup> or observed data [21,31]. The domain of the groundwater model is primarily that of the saturated zone shown in Figure 2.

Estimates of rootzone drainage are obtained from an AWB undertaken on a district-wide scale at annual intervals [26]. Inputs to the water balance are based on measured rainfall and irrigation diversions. Outputs are based largely on an assumed irrigation efficiency factor, which represents the percentage to total water applied that is consumed by crop evapotranspiration. The residual of the water balance is the rootzone drainage rate. It is a comparatively small term relative to the other components (which have large error bands) and is therefore highly uncertain, particularly in early time periods for which there are limited data. It is also heavily dependent on the irrigation efficiency factor used in the calculation.

Although the estimates of rootzone drainage are available, they have not been used directly by the groundwater model. This is due largely to the lack of a simple, ye<sup>t</sup> appropriate, vadose zone modelling tool at the time of the groundwater model development that could simulate the influence of perching of the transfer of rootzone drainage to recharge.

The recharge rates have been derived using the inverse method in which observed groundwater level fluctuations and aquifer pumping test data are used to constrain the hydrogeological parameters so that recharge may be derived inversely during model calibration, which was undertaken manually. The spatial zonation of recharge is based initially on the timing of irrigation development. The zonation evolves during the inversion process and is not highly constrained. The recharge rates over the whole irrigation areas are compared to the root zone drainage rates post-calibration as a check that the rates derived are comparable, but there is significant spatial variability between recharge zones. Other datasets are used, wherever possible, to constrain the model, such as under-river geophysics and river salinity data.

The calibrated groundwater model is then used to predict groundwater returns and salt loads into the future for a range of managemen<sup>t</sup> scenarios. Under these scenarios, a subjective recharge rate of 100 mm/year has been selected to simulate ongoing irrigation.

The existing method is logical but has some significant limitations related to (1) the non-uniqueness of the recharge–aquifer parameter relationship and its associated uncertainty, and (2) the indirect link between on-ground actions and the derived recharge rates. These limitations result in a model that does not use all of the datasets available to constrain it, and one that relies on a subjective approach for the recharge rates used in predicting of groundwater returns (and salt loads). As such, the method is susceptible to bias.
