10:45 AM - 11:00 AM
[HDS11-06] Prediction accuracy and error factors of active fault maps at rupture location: verification using surface ruptures associated with the 2016 Kumamoto earthquake
Keywords:Fault displacement hazard, Active fault mapping, 2016 Kumamoto earthquake, Surface rupture, Fault uncertainty zone
At inland earthquakes, buildings can be damaged by fault displacements as well as ground shaking (e.g., Zhou et al., 2010). Thus, as measures against fault displacement hazards, several regions have adopted regulation zones around active faults. For example, in California, United State, a 150–200 m-wide regulation zone and a 15 m setback are defined on either side of major fault traces after 1973 (Bryant and Hart, 2007). The data on mapping errors of active fault maps (Petersen et al., 2011; Scott et al., 2023) are crucial to establish these regulation zones. Activity and mapping criteria of active faults and preservation state of tectonic landforms vary by region. Thus, the existing data on mapping errors are not always applicable to all regions. Studies using surface rupture data with high accuracy are needed especially in Japan where erosion and deposition rate are fast and active faults are clustered.
Mapping errors for the 2016 Kumamoto earthquake that generated the longest surface ruptures in Japan were calculated and evaluated in this study. The mapping errors have been already calculated by Konno and Toda (2017). In this study, other two maps and factors related to mapping errors were newly verified.
Pre-earthquake active fault maps verified were Watanabe et al. (1979) and Watanabe (1984), Ikeda et al. (2001), and Nakata and Imaizumi (2002). Surface ruptures at the 2016 earthquake were from Kumahara et al. (2022). The method of Scott et al. (2023) was followed to assess mapping errors. Point data was created for every 1 m on the surface ruptures and pre-earthquake active fault traces in QGIS. The horizontal distances between the ith rupture point and all fault trace points were calculated, and the minimum value among them was taken as the mapping errors for the ith rupture point in MATLAB. Surface ruptures were grouped into “principal” and “distributed” and “predicted” and “unpredicted”. Surface ruptures were also classified by the rough age of the surface (Hoshizumi et al., 2004, 2014; AIST, 2023) on which they appeared. The mapping area of Ikeda et al. (2001) did not cover parts of the 2016 surface ruptures, and thus these ruptures were excluded in the analysis for Ikeda et al. (2001).
The medium and maximum values of mapping errors for all ruptures are ~72–280 m and ~4000–9000 m, respectively. The 150 m regulation zone width in California covers ~40–65% of the 2016 ruptures. The medium and maximum values of mapping errors for predicted principal ruptures are ~23–48 m and ~227–272 m, respectively. The 150 m width covers ~88–96% of the predicted principal ruptures.
Scott et al. (2024) showed that the mapping errors for ruptures on the Holocene surface are larger than those on the older units. In this study, the maximum mapping errors for the 2016 surface ruptures on the Late Pleistocene surface are also larger (Fig. 1). Unpredicted faults at the 2016 earthquake (e.g., Goto et al., 2017) on the Late Pleistocene units might be indistinct in landforms even for the time scale after the Late Pleistocene due to the low activity. Focusing on predicted principal ruptures, mapping errors for those on the Late Pleistocene are smallest. This quantitatively supports that the Late Pleistocene surface is a critical marker of rupture locations.
The dextral strike-slip at the 2016 earthquake generated left-stepping echelon alignments at various scales. Several to several hundred-meter scale en echelon arrays were distinguished from the 2016 surface ruptures as one of the factors related to mapping errors. The main faults were estimated by the linear approximation of the all or central rupture points on each en echelon rupture. The distance between the main faults and the edge rupture points on each en echelon rupture was calculated as the half width of the en echelon patterns. The half widths were mostly under 50 m. For regulation zones around strike-slip faults, this type of width should be considered.
Mapping errors for the 2016 Kumamoto earthquake that generated the longest surface ruptures in Japan were calculated and evaluated in this study. The mapping errors have been already calculated by Konno and Toda (2017). In this study, other two maps and factors related to mapping errors were newly verified.
Pre-earthquake active fault maps verified were Watanabe et al. (1979) and Watanabe (1984), Ikeda et al. (2001), and Nakata and Imaizumi (2002). Surface ruptures at the 2016 earthquake were from Kumahara et al. (2022). The method of Scott et al. (2023) was followed to assess mapping errors. Point data was created for every 1 m on the surface ruptures and pre-earthquake active fault traces in QGIS. The horizontal distances between the ith rupture point and all fault trace points were calculated, and the minimum value among them was taken as the mapping errors for the ith rupture point in MATLAB. Surface ruptures were grouped into “principal” and “distributed” and “predicted” and “unpredicted”. Surface ruptures were also classified by the rough age of the surface (Hoshizumi et al., 2004, 2014; AIST, 2023) on which they appeared. The mapping area of Ikeda et al. (2001) did not cover parts of the 2016 surface ruptures, and thus these ruptures were excluded in the analysis for Ikeda et al. (2001).
The medium and maximum values of mapping errors for all ruptures are ~72–280 m and ~4000–9000 m, respectively. The 150 m regulation zone width in California covers ~40–65% of the 2016 ruptures. The medium and maximum values of mapping errors for predicted principal ruptures are ~23–48 m and ~227–272 m, respectively. The 150 m width covers ~88–96% of the predicted principal ruptures.
Scott et al. (2024) showed that the mapping errors for ruptures on the Holocene surface are larger than those on the older units. In this study, the maximum mapping errors for the 2016 surface ruptures on the Late Pleistocene surface are also larger (Fig. 1). Unpredicted faults at the 2016 earthquake (e.g., Goto et al., 2017) on the Late Pleistocene units might be indistinct in landforms even for the time scale after the Late Pleistocene due to the low activity. Focusing on predicted principal ruptures, mapping errors for those on the Late Pleistocene are smallest. This quantitatively supports that the Late Pleistocene surface is a critical marker of rupture locations.
The dextral strike-slip at the 2016 earthquake generated left-stepping echelon alignments at various scales. Several to several hundred-meter scale en echelon arrays were distinguished from the 2016 surface ruptures as one of the factors related to mapping errors. The main faults were estimated by the linear approximation of the all or central rupture points on each en echelon rupture. The distance between the main faults and the edge rupture points on each en echelon rupture was calculated as the half width of the en echelon patterns. The half widths were mostly under 50 m. For regulation zones around strike-slip faults, this type of width should be considered.