**5. Discussion**

#### *5.1. Accuracy of Modelling*

Modelling will necessarily involve simplifications of reality. For example, (1) all surfaces are assumed to be topographically flat; and (2) layers are considered to be physically homogeneous with known properties. The values of hydraulic properties need to be considered to be conceptual in nature, rather than being true values. The study here has tried to incorporate the key processes. The comparison between models tests mathematical errors.

The comparison of model outputs show that the outputs were consistent in most aspects for the 1D modelling. Two exceptions were (1) the transfer function for Experiment 4; the situation involving interception of the ponded head with the upper boundary condition; and (2) the drainage volumes for Experiment 4. These discrepancies were related in that the PerTy3 output showed that the perched head reached the upper boundary condition about five years later than the FEFLOW output. This has led to drainage volumes occurring five years later, although volumes were similar. It also meant that recharge has equilibrated at about the same time as this interception for PerTy3; while recharge has equilibrated some years after the perched head has intercepted the upper surface, causing a distortion of the transfer function. Despite these discrepancies, the consistency of the model shows that the representation of the processes are accurate.

The comparison of model outputs for 2D modelling does not show the same level of consistency. In particular, PerTy3, compared to FEFLOW, overpredicts (1) the delay in recharge occurring external to the irrigated field; and (2) the exponential decay in transfer function. Both may be due to ignoring the additional ponded head on recharge external to the irrigated field. The mound extends beyond the irrigated field [24] and this reduces delays for recharge to occur, causing recharge to increase continuously rather than be delayed. The mounding also changes the shape of the moisture storage in the unsaturated zone, which will a ffect the exponential decay [19].

The outputs for FEFLOW and PerTy3 2D models are consistent in the partitioning of recharge between under the irrigated field and external to the irrigated field; the reduction of the perched head with *B,* and the exponential decay of perched head and wetted width. Thus, while some 2-dimensional aspects are not adequately modelled, most other aspects appear to be adequate for the objectives of the transfer function.

The comparison of model outputs with field observations will be discussed in [20].
