Identification of the optimal ground motion intensity measure and input parameters for assessing liquefaction-induced lateral spreading based on the generalized additive method

Abstract

Seismic intensity measures (IMs) play an important role in predicting liquefaction-induced lateral spreading. Many studies identified the optimal IMs for estimating lateral spreading without considering soil and topography parameters, and their findings are based on numerical simulation. These inadequate considerations may lead to impractical and incomplete results. In this paper, therefore, a large amount of real historical data containing multiple factors is collected, in which abnormal or inappropriate data are removed. Thirty-one IMs are calculated using the bidirectional ground motion records from historical earthquake records. Based on historical field data and the corresponding 31 IMs, the optimal IM is identified according to some criteria considering soil and topography parameters including correlation, efficiency, proficiency, and sufficiency analysis. The results show that the composite acceleration intensity I_a satisfies the four criteria. Additionally, the key input parameters (including I_a) affecting lateral spreading were analyzed using the generalized additive model (GAM). The average fine content and mean grain size are proposed to be removed in the construction of models, which could significantly reduce testing costs and be more conducive to engineering applications. The proposed GAM with four input parameters, I_a, T₁₅ (cumulative layers thickness with (N₁)₆₀ < 15), S (Slope), and W (free face radio), performs the best after comparing with other machine learning methods and existing empirical models. A flowchart of GAM usage was provided to the engineers.