A Statistics-Based Method for Estimating the Soil Water Retention Curve and Unsaturated Shear Strength in Engineering Practice

Unsaturated soil mechanics is rarely applied by geotechnical engineers working within the construction industry. This could be due to a poor understanding of the subject area, a lack of suitable unsaturated testing data, or a lack of suitable procedures and tools required to apply the theory in practice. The aim of this research is to show how the soil water retention curve (SWRC) and unsaturated shear strength of a soil can be estimated using standard site investigation data and then applied to geotechnical engineering problems in practice. This includes the development of a SWRC prediction procedure using 102 soil datasets from the UNSODA database. Statistical analysis is undertaken to compare the prediction of the SWRC using the Arya and Paris (1981) model (AP), Modified Kovacs Model (Aubertin et al., 2003) (MK) and the Perera et al. (2005) model (PM) with the measured drying SWRC from the database. The 5 and 95 percentiles of the error between the predicted and measured suction (suction error) are calculated to assess the performance of each method for different soil types and later used as confidence limits for soils not included in the dataset. Analysis shows that all three SWRC predictive methods can reasonably predict the SWRC of sands, but due to a lack of plasticity data in the database, only the Arya and Paris (1981) model can be used to estimate the SWRC of cohesive soils. The SWRC estimation procedure is validated using two soil samples from the literature, a sandy clay soil and a sand soil. A method to estimate the increase in shear strength due to soil suction is presented using each predicted SWRC, along with the the upper and lower confidence limits of the SWRC, for a typical geotechnical engineering slope stability problem. The use of this research is demonstrated via a two-dimensional PLAXIS finite element model showing how the factor of safety (FoS) of the slope increases as a result of using the SWRC to estimate changes in shear strength using the Fredlund et al. (1996) and Vanapalli et al. (1996) equations. By taking soil suction into account, the FoS of the slope can be significantly increased, with an improvement of 0.24 over the simulation that ignored suction when using the SWRC estimated using the AP model. By using the predicted SWRC upper and lower confidence limits, it is shown that the estimated increase in shear strength is not highly sensitive to the choice of values of soil suction.