Applicability of Richardsʹ equation models to predict deep percolation under surface irrigation

Deep percolation under irrigation is often indirectly estimated from Richardsʹ equation models. A Richardsʹ equation model requires soil hydraulic and crop specific parameters that may be estimated from soil moisture data at different soil depths. This paper examined whether parameters of a Richardsʹ equation model estimated from soil moisture data can be used to adequately predict deep percolation under surface irrigation. For this purpose, we used data from a lysimeter experiment on a loam soil under surface irrigation. Monte Carlo simulations of a 1-D Richardsʹ equation model were generated using a set of model parameters sampled from a feasible range of parameter values. For each Monte Carlo simulation (a set of model parameters), adequacy of model predictions was evaluated against lysimeter data on soil moisture and independently measured deep percolation. Adequacy was assessed by fitting a linear regression of observed data on simulated values and testing for the null hypotheses of unit slope and zero intercept. Using this approach, we did not succeed in predicting simultaneously soil moisture and deep percolation. This lack of simultaneous fit likely indicates an inadequate model structure as data uncertainty was of secondary importance in our lysimeter study. For surface irrigation, preferential infiltration can probably not be avoided even when soil cracking is not apparent. Under such conditions, the use of a simple Richardsʹ equation model appears inadequate. A conceptual model that was published earlier is considered to be more appropriate as it adequately predicted deep percolation, effectively accounts for preferential infiltration during surface irrigation and required less data than the Richardsʹ equation model.