Hydraulic conductivity prediction for sandy soils

The potential for petroleum-contaminated soils to impact ground water is evaluated using the soil leachability model in South Carolina and Georgia. In this model, the Green and Ampt (1911) equation is used to estimate unsaturated flow with saturated hydraulic conductivity (K(S)) values obtained using the Rawls and Brakensiek (1989) equation from inputs of percent sand- and clay-sized particles. However, many soils have > 70% sand-sized particles, which is the maximum amount for which the Rawls and Brakensiek equation is valid. Therefore, 70 sets of K(S) and particle-size data from the literature for southeastern United States sandy soils were analyzed to develop a new equation for estimating K(S). A multiple linear regression model with an adjusted R2 = 0.65 (p < 0.0001) was developed from percent clay- and sand-sized particle data. Eight additional sets of data were used to validate the model. The root mean square deviation and maximum squared difference, (yi - yi(2)), values for the new model are smaller than those obtained using the Rawls and Brakensiek equation, or Rosetta, a neural network-based model (Schaap 1999). The new model provides estimates that are within an order of magnitude for all but one of the test data sets. Rosetta predicts K(S) within one order of magnitude of measured values for six test data, and predicts K(S) within two orders of magnitude for the other two. The Rawls and Brakensiek predictions are within one order of magnitude for four test data, but are two or more orders of magnitude too high for the remaining four points. The new equation is recommended for estimating K(S) for sandy southeastern United States soils for which inputs are limited to percent sand and clay.