Meta-analysis of 301 slope failure calculations. I: database description

Since the early part of the twentieth century, two-dimensional limit equilibrium (2DLE) analysis has been the scientific community's primary means of slope stability calculation. However, it is well established that the input parameters to 2DLE, namely, soil strength and anisotropy, slope geometry, pore water pressures, failure surface geometry, applicable correction factors, and loading conditions are all inherently uncertain. Effective modeling must account for these uncertainties statistically. Unfortunately, most of the key statistical parameters, such as the safety factor statistical distribution and standard deviation (s_d), are unknown and must be estimated by the analyst. In response to this growing need for statistical information, a database was established from the literature of 157 different failed slopes and the corresponding published 301 safety factor (SF) calculations. The database, which covered more than five decades of slope stability research, also included a number of the slope stability factors, including analytical method used, stress approach (effective versus total), assumed slip surface geometry, slope type, applied correction factors, and soil Atterberg limits. A temporal analysis found no evidence that SF prediction or deviation had significantly changed. A log (base 10) normal distribution was found to adequately describe the SF data, with a (nontransformed) mean of 1.03 and a (transformed) s_d of 0.087, but the pronounced curvature of the residuals indicated significant, unresolved slope factors, further investigated in the companion paper.