Development of terrain classifications as a proxy to estimate VS30

While geomorphological maps have generally been used to delineate the boundaries of homogeneous landforms, they have also been used as the basis to estimate the susceptibility of terrains to natural hazards, and in particular, on terrains where urban infrastructures are situated. As these terrain maps are predominantly based on the properties of surface materials, terrain types can be useful indicators for rock erosion resistance and other ground conditions located within the same climate/geological provinces. VS30, the time-averaged shear-wave velocity of the upper 30-m from the surface, is an important engineering index for estimating ground conditions that affect seismic amplification and has often been predicted using terrain-based models. To further the use of these proxy-based approaches, Iwahashi et al. (2018) recently developed a polygon-based model of global terrain classification using a combination of slope gradient, surface texture, and local convexity derived from a 280-m digital elevation model (DEM) interpolated from the Multi-Error-Removed Improved-Terrain (MERIT) DEM (Yamazaki et al., 2017). Based on visual comparisons, the polygon model can generally be considered suitable for distinguishing bedrock mountains, hills, large highland slopes, intermediate landforms (plateaus, terraces, large lowland slopes), and plains. However, the model cannot detect the presence of narrow valley bottom plains or slight rises in gentle plains where many metropolitan areas are situated. We find artifacts, relating to unevenness in the DEMs, adversely affect the accurate classification of natural landforms, which is a major drawback of the 280-m DEM. We then proceed to analyze higher resolution DEMs and include the Height Above Nearest Drainage (HAND) parameter. Based on visual comparisons of our results to manual geomorphological and geologic maps of the Kanto Plain, Los Angeles basin, and San Francisco Bay area, we conclude that the Light-Detection-And-Ranging-scale DEMs and the addition of the HAND metric substantially aide in the detection of the narrow and gentle plains. Finally, we calibrate the new terrain model against nearly 4,000 measured VS30 values for the Kanto Plain, San Francisco Bay area, and Los Angeles basin and obtain the representative VS30 estimates for each terrain category.