Functional evaluation of methods for predicting the soil water characteristic

Functional sensitivity analysis was used for 66 different soil horizons to evaluate methods of predicting the soil water characteristic (θ(ψ)). Error in predicted hydraulic properties was identified through examination of outputs from the SWIM soil water simulation model. The variability in these outputs arose solely from error in prediction of the soil hydraulic properties. The simulation scenario was the determination of drainage and evapotranspiration over 5-years from a perennial pasture in southern Victoria, Australia. The functional evaluation considered hydraulic conductivity as a function of water content (K(θ)) being independent of error in θ(ψ) prediction, as well as K(θ) being estimated from predicted θ(ψ) data. Inaccurate hydraulic property prediction caused error in drainage calculation approximately equal and opposite to that for evapotranspiration whilst profile water storage was almost unaffected. Water balance error resulting from the two-point method of θ(ψ) prediction was small with simulated drainage less than 4 mm/year different, on average, from that generated using measured θ(ψ) data. The two-point method will be sufficiently accurate for many simulation applications and the one-point method is a useful alternative. The multiple regression equations, which do not use measured hydraulic property inputs, were the least accurate of the three methods of θ(ψ) prediction and should be applied with caution. When K(θ) was estimated using θ(ψ) data generated with the multiple regression method the hydraulic property specification was sufficiently imprecise to cause large additional error in drainage prediction. Functional sensitivity analysis is useful for assessing pedotransfer functions and for determining the accuracy and precision required in hydraulic characterisation surveys.