Unsaturated soil shear strength is an important parameter in soil erosion and management. Measurement of unsaturated shear strength at field scale is difficult, time-consuming, and very costly. This study was conducted to investigate the relationship between unsaturated shear strength parameters and soil properties, and to predict the unsaturated shear strength parameters (effective cohesion, c', angle of effective internal friction, φ' and angle of internal friction related to matric suction, φb) using multiple-linear regression (MLR). Direct shear tests were performed at combinations of three normal stresses of 25, 50 and 100 kPa, and four matric suctions of 0, 10, 30 and 50 kPa (i.e., 12 tests per each soil) to determine the shear strength parameters in 14 soils. Soil properties including particle size distribution (sand, silt, and clay percentages or geometric mean diameter, dg and geometric standard deviation, σg), organic matter content (OM), calcium carbonate content (CaCO3), compactness indices (bulk density, ρb and, relative bulk density, ρb-rel), and mean weight diameter of aggregates (MWDdry, MWDwet), structural stability indices (aggregate stability, AS, stability index, SI and index of crusting, Ic) were determined and used as predictors in MLR models. Strong negative correlations were found between c' and φ'. The c' positively correlated with clay content. Significant negative correlation was observed between c' and sand fractions and dg. Significant positive correlation was obtained between φ' and fractions of sand and dg. The φ' negatively correlated with clay, fine silt content (FSi), MWDdry, and AS. The φb had no significant correlation with soil properties, indicating that φb is independent of soil properties and basically is affected by matric suction. Clay, coarse sand (CS) and very fine sand (VFS) were applied in the model for predicting c'. Clay had a positive and, CS and VFS had negative effects on c'. Pedotransfer functions (PTFs) using the fine sand (FS) and VFS as predictor could estimate the φ' accurately, so that they entered to PTFs with positive signs. In addition, the φb was predicted by parameter Ic only, so that it had negative effect on φb. Overall, better prediction models were developed for φ' than for c' and φb.

Keywords