Uncertainty Projected Mapping: A Bayesian tool for generating site response maps with statistically significant resolutions   

Abstract

Site response maps are mostly proxy-based. The map resolutions are driven by the resolutions of the digital elevation model. Although high-resolution maps are seemingly more enriched with local information, these details are not always supported with in-situ data. The high-resolution maps are reliable only when the in-situ data supports it. Without in-situ data available, a low-resolution map might be more reliable. Depending on the availability of in-situ data, a site response map with spatially varying map resolutions would better represent the actual ground conditions. In this study, we introduce uncertainty projected mapping (UPM) to generate statistically significant map resolutions. UPM is Bayesian-based and considers the statistical significance of differences in neighborhood values in determining the posterior site response. The study area is in Osaka, Japan, where dense borehole data from the Kansai Geo-informatics Network is available. In the Bayesian framework of UPM, the site responses estimated using 1D seismic ground response analysis at this borehole network constitute the likelihood. In the first case study, a non-informative prior (uniform) is employed to generate the posterior UPM site response map. The UPM map shows the presence of statistically significant map resolutions, which in-situ data can explain. In the second case study, the available proxy-based site responses are employed as an informative prior to generate the posterior UPM map. The results show that proxy-based site responses have been updated only at meshes with in-situ data. However, these updates also show statistically significant resolutions explainable by the in-situ data. The results of both case studies show that the statistically significant map resolutions of the UPM site response map better represent the in-situ data.