*5.4. Data Requirements*

Superficially, the model appears to require many parameters and hence should be considered as complex. For example, the model, in principle, requires several soil parameters for each layer. However, many of these parameters are based on texture and can be approximated based on lithology information and data [23]. The saturated vertical hydraulic conductivity for layer 2 and horizontal conductivity for layer 1 appear to be critical parameters and are reflected in the dimensionless variables *A* and *B*, which portray the range of behavior expected in the field [19]. A critical input to the model is the irrigation accession, *IA*. This also is incorporated in the dimensionless variable *A*. Where district-scale flux data is used e.g., surface water diversions, length of channels, this can only be converted to *IA* through a good knowledge of the irrigation area and how it has changed over the history of the irrigation district. In some cases, agronomic experiments have been conducted to estimate the field water balance and hence water-use e fficiency factors. Information of this type may be able to highlight how irrigation accessions have changed over time in response to changed technology, rainfall and availability of surface water. While PerTy3 has not represented this component of the water balance, its accuracy is dependent on good information on surface water balance. The need for drainage is a sign that there is rejected recharge and is useful for the calibration of soil hydraulic parameters [19].

## *5.5. Further Work*

The work above has shown that the processes are well understood; that the 1D modelling of these processes appears to be adequate, but the 2D modelling is only adequate for some aspects. In particular, the assumption that the ponded head is e ffectively zero external to the irrigated field is clearly inadequate. Some improvement to the model are, therefore, required, perhaps by considering the quasi-steady state analytical models [26], in which a non-zero perched head is incorporated. The 2D transfer function does not vary significantly from the 1D modelling, suggesting that the one transfer function may be adequate for a wide range of situations, and that an exponential model appears to be an adequate approximation. This suggests that such approximants may be useful for estimating recharge, but this requires more work. Brownfield irrigation developments appear to only di ffer from greenfield developments for large B parameters and for separations of less than 250 m. This suggests that treating developments as independent may be a reasonable approximation of a range of situations. However, more work would be required for the other situations to find the best estimation method.
