Numerically Predicting Seepage Gradient Forces and Erosion: Sensitivity to SoilHydraulic Properties

Keywords: Groundwater seepage Pedotransfer function ROSETTA Seepage erosion Uncertainty analysis s u m m a r y Research has suggested that streambank seepage can be an important mechanism of bank instability; however, limited information is available on the level of soil characterization necessary to accurately predict seepage gradient forces and erosion. The objective of this research was to quantify the expected range of predicted seepage gradients for various degrees of site characterization. Uncertainty analysis on seepage gradient predictions was performed relative to variability in soil hydraulic properties. A two-dimensional unsaturated/saturated groundwater flow model was used to simulate a homogeneous soil layer for sand and loamy sand soils packed at various bulk densities, q b. A pedotransfer function (ROSETTA), designed to estimate unsaturated hydraulic properties from surrogate soil data (i.e., texture and bulk density), was used to derive the saturated hydraulic conductivity, K s , and water retention parameters for various levels of site information (i.e., only textural class; percent sand, silt, and clay (%SSC); %SSC and q b ; and %SSC, q b , and K s). Statistical distributions were derived for each soil hydraulic parameter and Monte Carlo simulations were performed to generate distributions of maximum seepage gradient. The deviation in predicted seepage gradient was calculated using assumed baseline conditions. Ranges in predicted soil hydraulic parameters and maximum seepage gradients were considerably reduced when using %SSC as compared to soil texture. Therefore, at a minimum, soil samples should be taken for particle size analysis. For q b between 1450 and 1500 kg m À3 , soil hydraulic parameters could be derived using ROSETTA and inputting %SSC, with little additional benefit provided by measuring q b and/or K s. When the q b was less than 1450–1500 kg m À3 , inputting q b and/or K s consistently reduced the magnitude of deviations from the baseline and therefore should be measured from undisturbed soil samples. The opposite was observed for q b greater than 1450–1500 kg m À3 due to discrepancies between ROSETTA-derived and actual values of soil hydraulic parameters other than K s. Considerable deviations (i.e., around 20%) were observed in seepage gradients under this scenario. When ROSETTA-derived and actual values of soil hydraulic parameters more closely matched, inputting q b and/or K s benefitted seep-age gradient predictions as deviations in seepage gradients were less than 5% for the sand and loamy sand soils. Therefore, it is vital to …