**4. Results**

#### *4.1. Drainage and Irrigation Recharge Using Partial Calibration*

An initial calibration of the AWB-TF models was undertaken using drainage data. It was found that the *Kv* of the clay layer was well constrained by the drainage targets and only minor adjustments were required from the default value of 0.03 cm/day (109.5 mm/year) (Table 2). Non-negligible drainage rates would occur in areas where drainage infrastructure is absent when the *Kv* was decreased substantially from the default value. Any substantial increases to the *Kv* were constrained by the observed drainage data at Loxton.


**Table 2.** Calibrated hydraulic conductivity ( *Kv*) of Blanchetown Clay from use of drainage data.

Figure 5 shows a comparison of the recharge and drainage rates derived from the existing methods and datasets to those derived by the new AWB-TF models developed. The AWB-TF models show recharge and drainage outputs that increased substantially in the late 1950s, reflecting the main period of irrigation development and accounting for a short time lag (<5 year), a period of stabilisation between 1960 and 2000, and a decline post-2000 as irrigation e fficiency measures were introduced. The calibration focused on the period of late 1990s onwards, as this was where there was more confidence in the available datasets.

**Figure 5.** Comparison of recharge and drainage rates derived from existing methods and datasets (plotted in orange) to those derived the new agronomic water balance (AWB)-transfer functions (TF) models (plotted in blue) for the Loxton irrigation area, showing (**a**) techarge rates, (**b**) drainage rates, and (**c**) total drainage and recharge rates.

When recharge and drainage were combined, the outputs of the AWB-TF models compared reasonably to the combined LB2011 recharge and observed drainage rates (Figure 5c). However, there was divergence in the 1960–2000 period when recharge and drainage were considered separately. The LB2011 model had recharge declining from 1980 onwards at a period when drainage rates were increasing. This is counter-intuitive because if drainage rates were increasing then it would be expected that recharge was also increasing or was at least stable. It may be the result of the regional groundwater mound being intercepted by the drainage infrastructure and the LB2011 model needing to reduce recharge to account for this. Although this is possible, the LB2011 model does not model these processes explicitly, nor does its documentation describe these circumstances as the justification for reducing recharge over a period when drainage was increasing. In contrast, the AWB-TF model estimates gross recharge, and if the interception of the regional groundwater mound is occurring then it can be modelled explicitly by the groundwater model. As shown in this example, the new approach provides a much clearer representation of the irrigation actions affecting groundwater recharge, ultimately creating transparency in the estimation of salinity impacts to the river.

#### *4.2. Comparison of Integrated Models*

The integrated models obtained comparable calibration statistics to the existing method. The scaled residual mass statistic (SRMS) was 1.6% for the LB2011 model, 1.8% for TF-A/B, and 1.7% for TF-C. The calibrated horizontal hydraulic conductivities, *Kh*, of the regional water table aquifer are shown in Figure 6. The spatial distribution of *Kh* under TF-B was broadly similar to LB2011. Given that the conductivity field of TF-B was derived independently of LB2011, this provides some confidence in the new method.

**Figure 6.** Comparison of calibration for horizontal hydraulic conductivity (m/d) of the Loxton Sands (regional water table aquifer) for the three models evaluated: LB2011/TF-A model, TF-B model, TF-C model.

In TF-C, a zone of high *Kh* was derived in the south of the model domain, which was not replicated by the other models. Closer inspection revealed a complication in this area of the model, with simultaneous calibration of drain conductance and efficiencies in the AWB, whereby greater recharge could be produced by the transfer function only to be offset by increases to drainage conductance in the MODFLOW model. This was a product of the automated calibration procedure, and further work is required to mitigate this effect, but this complication is not reflective of the integrated modelling approach being tested.

There were some key differences in how irrigation recharge was distributed in space and time. The inclusion of the AWB-TF components to the modelling tended to minimize the spatial variability in recharge rates, which could be considerable in the existing method (Figure 7). The recharge rates of the integrated models were also more stable temporally (see Figure 5a).

**Figure 7.** Distribution of irrigation recharge in 1985 within the existing LB2011 model, the TF-A/B model, and TF-C model.

Figure 8 compares the irrigation efficiency factor (*E*ff ) used in the integrated models. In the TF-A/B model, *E*ff was taken from the default values of [26] and was not varied during the calibration, whereas in TF-C it was varied as part of the calibration. The recharge rates and estimated groundwater returns can be quite sensitive to *<sup>E</sup>*ff, and the large increase in the early 2000s was coincidental with the pressurisation of the irrigation supply system at Loxton.

**Figure 8.** Irrigation efficiency factor used in the TF-A/B and TF-C models.

Figure 9 shows the projected salt flux to the river for the different models used. It is highest under the existing method.

**Figure 9.** Estimated salt flux associated with groundwater returns to the river for the different models used. The modelled salt flux was determined using the calibrated models up until 2011. For LB2011, future projections were determined by assigning a constant irrigation recharge of 100 mm/year. For the other models, future projections were determined in accordance with the AWB and calibrated irrigation efficiency for 2011 and running this for 100 years through the TF models to generate a recharge rate into the future.

The integrated approach offers a transparent method of determining an appropriate recharge rate for estimating future salinity impacts. The basis of the method is to set the irrigation accession rate in accordance with the AWB and calibrated irrigation efficiency for 2011, and to run this into the future. By comparison, the existing methods subjectively select 100 mm/year as the recharge rate into the future on the basis of expert opinion rather than local datasets.

For TF-A/B, the continued *E*ff was 85% for both districts. For TF-C, the *E*ff was only adjusted slightly (87% for Loxton and 83% for Bookpurnong) but its sensitivity was such that small changes can result in large changes to the *RZD*. For instance, at Loxton, an adjustment in *E*ff from 85% to 83% resulted in a 13% decrease in the *RZD* (from 129 to 112 mm/year). This influenced the projected groundwater returns and shows the importance of having an objective procedure to calibrate *<sup>E</sup>*ff.
