A geo-statistical framework to reduce uncertainty in predictions of V_S30 and other geotechnical variables

Geo-statistical modelling challenges arise when regional models in data-scarce regions are required. This is relevant when evaluating and dealing with geotechnical uncertainty in earthquake engineering applications. From a geotechnical earthquake engineering perspective, VS30 (shear-wave velocity in the upper 30m of soil) predictions must cover regional scale study areas. A systematic lack of data can significantly limit obtaining a satisfactory dataset for accurate soil amplification definition. This paper implements a novel geo-statistical framework to create VS30 mapping. Such a framework employs an approach of Bayesian Kriging, implemented initially within the context of petroleum reservoir modelling and recently applied to the case of Kathmandu Valley. The approach uses primary and secondary data to apply either debiasing or declustering to deal with typical issues of data availability in data-scarce regions. The multidisciplinary use of this analysis method provides an assessment of uncertainty, and an informed quantification of geotechnical parameters where traditional statistical methods may not produce sufficiently acceptable results. In this paper, a new set of secondary data using the case study of Kathmandu Valley (Nepal) is employed to demonstrate the flexibility of the geo-statistical framework developed in a previous study by the same authors.