Predicting three-region unsaturated hydraulic conductivity from three soil-water retention points

Reliable prediction models for unsaturated hydraulic conductivity (K) based on a minimum of easily measurable input parameters are valuable for assessing water and chemical transport in the vadose zone. Recently, the Three Region Campbell (TRC) model for predicting K in undisturbed soil in three separate pore size regions (corresponding to 0 to -10, -10 to -350, and -350 to -15,000 cm H2O of soil-water matric potential) was suggested and calibrated, but not tested. In this study, the TRC model is tested against independent data for 36 undisturbed soils from the Netherlands, and the number of necessary input parameters is reduced<br/>Reliable prediction models for unsaturated hydraulic conductivity (K) based on a minimum of easily measurable input parameters are valuable for assessing water and chemical transport in the vadose zone. Recently, the Three Region Campbell (TRC) model for predicting K in undisturbed soil in three separate pore size regions (corresponding to 0 to -10, -10 to -350, and -350 to -15,000 cm H2O of soil-water matric potential) was suggested and calibrated, but not tested. In this study, the TRC model is tested against independent data for 36 undisturbed soils from the Netherlands, and the number of necessary input parameters is reduced. The TRC model yielded accurate K predictions for the 36 soils in all three pore-size regions. Two different methods for using the TRC model with reduced sets of input parameters were proposed. Method I requires two measurements of hydraulic conductivity (at saturation and at -10 cm H2O) and three measurements of soil-water retention (swr) (at -10, -100 and about -250 cm H2O). Method II requires only a measurement of soil total porosity and the three swr measurements, and it is based on estimating saturated and near-saturated hydraulic conductivity from the air-filled porosity at -100 cm H2O (representing pores larger than 30 mum equivalent pore diameter). Root mean square error (RMSE) of prediction of Log(K) for the 36 soils increased by 18% (Method I) and 78% (Method II) compared with using the nonreduced parameter input in the TRC model. However, the prediction accuracy (typically around one order of magnitude) is still sufficient for many modeling purposes and, in comparison, the RMSE (Method II) was still four times lower than the original Campbell single-region K model. Data from six additional soils were used to evaluate the possibility of adding a fourth region (film flow, &lt;- 15,000 cm H2O) in the TRC model. The results imply that the TRC model concept, with its advantage, could be expanded to a four-region model when additional data for K in very dry soil become available