GIS-based characterization of fault zones in South-Korea using information on seismicity, in-situ stress and slip tendency &ndash; Evaluation of respect distances for nuclear waste disposal site screening

Bredemeyer, Stefan; Yoon, Jeoung Seok; Xie, Linmao; Lee, Jeong-Hwan

doi:https://doi.org/10.5194/egusphere-2023-2674

Preprints

https://doi.org/10.5194/egusphere-2023-2674

Preprints

23 Feb 2024

| 23 Feb 2024

GIS-based characterization of fault zones in South-Korea using information on seismicity, in-situ stress and slip tendency – Evaluation of respect distances for nuclear waste disposal site screening

Stefan Bredemeyer, Jeoung Seok Yoon, Linmao Xie, and Jeong-Hwan Lee

Abstract. Identification of seismically active fault zones and the definition of sufficiently large respect distances from these faults which enable avoiding the damaged rock zone surrounding the ruptured ground commonly are amongst the first steps to take in the geoscientific evaluation of sites suitable for nuclear waste disposal. In this work we present a GIS-based approach, using the earthquake-epicentre locations from the instrumental earthquake record of South-Korea to identify potentially active fault zones in the country, and compare different strategies for fault zone buffer creation as originally developed for site search in the high seismicity country Japan, and the low-to-moderate seismicity countries Germany and Sweden. In order to characterize the hazard potential of the Korean fault zones, we moreover conducted slip tendency analysis, here for the first time covering the fault zones of the entire Korean Peninsula. For our analyses we used the geo-spatial information from a new version of the Geological map of South-Korea, containing the outlines of 11 rock units, which we simplified to distinguish between 4 different rock types (granites, metamorphic rocks, sedimentary rocks and igneous rocks) and the surface traces of 1,528 fault zones and 6,654 lineaments identified through years of field work and data processing, a rich geo-dataset which we will publish along with this manuscript. Our approach for identification of active fault zones was developed without prior knowledge of already known seismically active fault zones, and as a proof of concept the results later were compared to a map containing already identified active fault zones. The comparison revealed that our approach identified 16 of the 21 known seismically active faults and added 472 previously unknown potentially active faults. The 5 seismically active fault zones which were not identified by our approach are located in the NE- and SW-sectors of the Korean Peninsula, which haven’t seen much recent seismic activity, and thus are not sufficiently well covered by the seismic record. The strike directions of fault zones identified as active are in good agreement with the orientation of the current stress field of the peninsula and slip tendency analysis provided first insights into subsurface geometry such as the dip angles of both active and inactive fault zones. The results of our work are of major importance for the early-stage seismic hazard assessment that has to be conducted in support of the nuclear waste disposal siting in South-Korea. Moreover, the GIS-based methods for identification of active fault zones and buffering of respect areas around fault zone traces presented here, are applicable also elsewhere.

Received: 10 Nov 2023 – Discussion started: 23 Feb 2024

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Stefan Bredemeyer, Jeoung Seok Yoon, Linmao Xie, and Jeong-Hwan Lee

Status: final response (author comments only)

CC1:
'Comment on egusphere-2023-2674', Giacomo Medici, 22 Mar 2024

General comments
Good geophysical research on the management of faulted nuclear waste repositories. The manuscript can be improved following the specific comments below.

Specific comments
Lines 13-15. The first sentence of your abstract is too long. Please, split it in two parts.
Lines 99-140. Provide more detail on the fault zones. Either extensional or strike-slip faults? Which is the dominant type?
Lines 99-140. Can you provide more detail on the geometry of the major faults such as thickness of the damage zone? The reader can make a correlation between thickness of the damage zone and length (see Figure A1).
Lines 108-110. There are also well known examples of repositories in other countries with either crystalline or sedimentary rocks. See also my comments below.
Lines 390-440. If the methodology is divided in sub-paragraphs, the same should be for the data. There is no match between the Paragraphs 3 and 4 on this point.
Lines 524-527. Take into account references on migration of contaminants where highly permeable faults are mapped in the area of a proposed repository for nuclear waste. Sellafield (West Cumbria, NW England) with crystalline and sedimentary rocks, see references below:
- Gutmanis, J. C., Lanyon, G. W., Wynn, T. J., & Watson, C. R. (1998). Fluid flow in faults: a study of fault hydrogeology in Triassic sandstone and Ordovician volcaniclastic rocks at Sellafield, north-west England. Proceedings of the Yorkshire Geological Society, 52(2), 159-175.
- Medici, G., West, L. J. (2023). Reply to discussion on ‘Review of groundwater flow and contaminant transport modelling approaches for the Sherwood Sandstone aquifer, UK; insights from analogous successions worldwide’ by Medici and West (QJEGH, 55, qjegh2021-176). Quarterly Journal of Engineering Geology and Hydrogeology. 56 (1), qjegh2022–097. doi:10.1144/ qjegh2022-097.
Lines 760-938. Please, insert the relevant references suggested above on migration of contaminants in an area of a proposed nuclear waste repository cut by normal faults.

Figures and tables
Figure 1. Specify the number of faults in the rose diagrams that should be 457+163+128 to match figure 8.
Figure A1. There is a correlation between length of a fault and its maximum capacity in terms of magnitude of earthquakes. This point has not been fully discussed in your manuscript. Please, consider my point due to the fact that you have produced a very important figure.
Table 1a. The content under “Map Colour” in unclear. Please revise it or apply changes to the caption.

Citation: https://doi.org/10.5194/egusphere-2023-2674-CC1
- AC1: 'Reply on CC1', Stefan Bredemeyer, 03 May 2024
  
  Reply to CC1: 'Comment on egusphere-2023-2674' by Giacomo Medici, 22 Mar 2024, Citation: https://doi.org/10.5194/egusphere-2023-2674-CC1
  Dear Giacomo Medici,
  thank you for the appreciation of our work and for your helpful suggestions for the improvement of the manuscript. Sorry for letting you wait! We applied all changes as you suggested in the specific comments section of your comment. Applied changes are as follows:
  
  Specific comments
  Lines 13-15. The first sentence of your abstract is too long. Please, split it in two parts.
  Reply: You are right, we split the sentence into two.
  Changes in Line 13:
  Identification of seismically active fault zones and the definition of sufficiently large respect distances from these faults commonly are amongst the first steps to take in the geoscientific evaluation of sites suitable for nuclear waste disposal. These steps are of great importance as they enable to avoid construction of a repository in the damaged rock zones surrounding existing faults, particularly if these exhibit long-lived activity and thus are prone to rupture further.
  Similarly, we split the long sentence starting in Line 228:
  However, only 38,217 (i.e., 95 %) of the 40,134 fault zone vertices in the dataset were considered as a consequence of that 5 % of the fault vertices are located outside the boundaries of the here considered geological map. That is due to the fact that a few faults and lineaments are situated offshore, and some faults are located in a buffer zone along the northern border of the country which is not covered by information on lithology.
  
  Lines 99-140. Provide more detail on the fault zones. Either extensional or strike-slip faults? Which is the dominant type?
  Reply: The dominant type is strike-slip because of the transpressional setting in and around the Korean Peninsula. We added three sentences with corresponding information in Line 111 and one sentence in Line 158.
  Insertion in Line 111:
  The neotectonic stress regime in and around the Korean Peninsula was initiated in early Pliocene (i.e. about 5-3.5 Mio years ago) and is dominated by transpressional stresses imposed by the mostly NW-directed subduction of the Pacific and Philippine plates beneath the Eurasian plate and the far-field stresses imposed by the collision of the northward moving Indian plate with the Eurasian plate (e.g. Chough et al., 2000; Chough et al., 2010; Cheon et al., 2023). As a result, the crust of the Korean Peninsula is affected by both E-W oriented compression and N-S oriented extension, as e.g. was shown through measurements of dilatational stress rates obtained from 4 years of continuous GPS recordings acquired between 2000 and 2004 (Jin & Park, 2006). Focal mechanisms of recent earthquakes moreover indicate that the predominant faulting type is strike-slip with minor thrust components and an ENE-WSW oriented maximum horizontal stress (Park et al., 2007; Jun & Jeon, 2010; Soh et al., 2017).
  Insertion in Line 158:
  Strain energy mapped through continuous GPS measurements furthermore indicates that the areas of highest strain and hence highest earthquake potential are located in the central and northern parts and along the eastern shoreline of the Korean Peninsula (Jin & Park 2007). For this reason, many of the fault zones along the southern margin of the Precambrian Gyeonggi Massif, in the central parts of the metasedimentary/metavolcanic Okcheon Fold Belt and the Precambrian Yeongnam Massif, and all across the Cretaceous sedimentary Gyeongsan Basin either may be active or are at high risk to become reactivated (Figure 1).
  Corresponding references:
  Cheon, Youngbeom, Chang‑Min Kim, Jin‑Hyuck Choi, Sangmin Ha, and Seongjun Lee, Taehyung Kim, Hee‑Cheol Kang and Moon Son (2023). Near‑surface termination of upward‑propagating strike‑slip ruptures on the Yangsan Fault, Korea. Nature Scientific Reports | (2023) 13:9869 | https://doi.org/10.1038/s41598-023-37055-7
  Chough, S.K., S.-T. Kwon, J.-H. Ree, D.K. Choi (2000). Tectonic and sedimentary evolution of the Korean peninsula: a review and new view. Earth-Science Reviews 52 2000. 175–235. doi: 10.1016/S0012-8252(00)00029-5
  Chough, S.K. and Y.K. Sohn (2010). Tectonic and sedimentary evolution of a Cretaceous continental arc–backarc system in the Korean peninsula: New view. Earth-Science Reviews 101 (2010) 225–249. doi:10.1016/j.earscirev.2010.05.004
  Jin, Shuanggen and Pil-Ho Park (2006). Crustal stress and strain energy density rates in South Korea deduced from GPS observation. TAO, Vol. 17, No. 1, 169-178.
  Park, Jong-Chan, Woohan Kim, Tae Woong Chung, Chang-Eob Baag and Jin-Han Ree (2007). Focal mechanisms of recent earthquakes in the Southern Korean 2006 – Peninsula. Geophys. J. Int. (2007) 169, 1103–1114. doi: 10.1111/j.1365-246X.2007.03321.x
  Jun, Myung-Soon and Jeong Soo Jeon (2010). Focal Mechanism in and around the Korean Peninsula (Korean with English abstract). 지구물리와 물리탐사 Jigu-Mulli-wa-Mulli-Tamsa. Vol. 13, No. 3, 2010, p. 198~202
  Soh, Inho, Chandong Chang, Junhyung Lee, Tae-Kyung Hong and Eui-Seob Park (2017). Tectonic stress orientations and magnitudes, and friction of faults, deduced from earthquake focal mechanism inversions over the Korean Peninsula. Geophys. J. Int. (2018) 213, 1360–1373. doi: 10.1093/gji/ggy061
  
  Lines 99-140. Can you provide more detail on the geometry of the major faults such as thickness of the damage zone? The reader can make a correlation between thickness of the damage zone and length (see Figure A1).
  Reply: Information in this respect is merely available for major faults, which have been active recently, but we can provide information on the Yangsan fault which represents the longest surface trace in our dataset as reference in Line 144.
  Insertion in Line 144:
  Detailed information on damage zone widths and fault core widths usually is rarely available because of the tremendous efforts and costs of obtaining such information. Also in Korea, such information primarily exists for the major fault systems which recently have been active and thus have been subject to systematic palaeo-seismic studies and field surveys. The long-lived Yangsan fault in SE-Korea e.g. is one of these major structures, which recently went into the focus of such research efforts, as a consequence of the M 5.5 Geyongju earthquake in 2016 (e.g. Cheon et al., 2023). The surface trace of the ENE-SWS trending Yangsan fault extends about 200 km across the Gyeongsang basin (inset map in Figure 1) and above-mentioned studies have shown that its fault core is up to 100 m wide, while the damage zone extends up to several kilometres (Cheon et al., 2023). Having such numbers at hand as reference is useful for the determination of sufficiently large respect distances for both, active and inactive fault systems.
  Following the rationale of your request, we further added some information on the limitations of the dataset and future research requirements at the end of the discussion:
  5.4 Limitations of the available data and requirements for future research
  The fault zone buffering strategy selected here aims at far-field enveloping of unsuited areas based on available data. Currently available data however does yet allow to consider 1) subsurface geometry of fault planes at sufficient detail, 2) across-fault asymmetry and along-strike heterogeneities of corresponding damage zones, 3) differences in rock mechanical properties of different rock types affecting the width of damage zones, and 4) local anisotropies and heterogeneities of rock masses affecting their deformation and breaking behaviour. Furthermore, as mentioned in the data description above, the majority of minor faults in the length range <2 km seems to be unknown and thus hasn’t been mapped yet. Such information still requires being acquired and compiled, since this length range may comprise many of the splay faults defining the damage zones around larger fault strands. The buffer zones resulting from our analysis thus are intended to be refined by more detailed field investigations during a later stage of the site search, particularly where the fault zones cross areas deemed to be suited for repository construction.
  Corresponding references:
  Cheon, Youngbeom, Chang‑Min Kim, Jin‑Hyuck Choi, Sangmin Ha, and Seongjun Lee, Taehyung Kim, Hee‑Cheol Kang and Moon Son (2023). Near‑surface termination of upward‑propagating strike‑slip ruptures on the Yangsan Fault, Korea. Nature Scientific Reports | (2023) 13:9869 | https://doi.org/10.1038/s41598-023-37055-7
  Kim, T., Choi, J.H., Cheon, Y., Lee, T.H., Kim, N., Lee, H., Kim, C.M., Choi, Y., Bae, H., Kim, Y.S., Ryoo, C.R., and Klinger, Y. (2023). Correlation of paleoearthquake records at multiple sites along the southern Yangsan Fault, Korea: Insights into rupture scenarios of intraplate strike-slip earthquakes. Tectonophysics, 854, p.229817. https://doi.org/10.1016/j.tecto.2023.229817
  
  Lines 108-110. There are also well-known examples of repositories in other countries with either crystalline or sedimentary rocks. See also my comments below.
  Reply: That is a good point. We added some more countries as examples to show that consideration of crystalline and sedimentary rocks as potential host rock actually is (and should be) a common approach where these are available and referred to literature reviewing the geoscientific approaches of different countries for further reading.
  Insertion in Line 121:
  South-Korea plans to start their site search with a blank white map, and choses to investigate crystalline and sedimentary rock types as host rock for the deep geological repository (e.g. Kim et al., 2000), as is done in many other countries, e.g. in Finland, France, Belgium, Germany, Switzerland, UK, and USA (Faybishenko et al., 2016; Hartley et al., 2017).
  Corresponding references:
  Faybishenko, Boris, Jens Birkholzer, David Sassani, Peter Swift (2016). International Approaches for Deep Geological Disposal of Nuclear Waste: Geological Challenges in Radioactive Waste Isolation - Fifth Worldwide Review. Prepared for: U.S. Department of Energy, by Lawrence Berkeley National Laboratory and Sandia National Laboratories. DOI: 10.2172/1353043
  Hartley, Lee, David Roberts, Fiona Hunter, and Barbara Lambers (2017). Evidence - The early stages of implementing geological disposal: regulatory use of geoscientific information. Project Number: SC100011, Environment Agency - March 2017. ISBN: 978-1-84911-390-8
  Lines 390-440. If the methodology is divided in sub-paragraphs, the same should be for the data. There is no match between the Paragraphs 3 and 4 on this point.
  Reply: You are right, the description of the results of the earthquake cluster analysis rather is content for the results section and hence we moved it down to the results section 4.1 Earthquake clusters. The results of the fault zone classification are now in section 4.2 Potentially active and inactive fault zones.
  Deletion of the paragraph starting in Line 385, because of redundancy w.r.t. contents in section 4.2 of the results section:
  Comparing the extracted features from rules 1 and 2, there is a large intersection between the two groups, where faults fulfilled both rules 1 AND 2. 128 of the faults found by rule 1, additionally are within 1 km range of an earthquake epicentre and touch or intersect an earthquake cluster. Only 35 of faults picked by rule 2 do not touch or intersect pre-determined earthquake-epicentre clusters at the same time.
  Insertion of method description in Line 385 as replacement for the above deleted passage:
  The results of the feature extractions done by application of rules 1 and 2 were then compared by means of intersecting the two datasets in order to determine the overlap between the two groups of extracted fault segments, in order to find the fault zone traces which satisfied both rules 1 and 2, and those merely satisfying rule 2.
  Insertion in section 4. Results, Line 389:
  For this purpose, we first generated heatmaps of the earthquake epicenter locations in the catalogue which in turn were used to outline 1) earthquake clusters and 2) 2x2 km search kernels around each of the earthquake epicenters. The results of the earthquake cluster analysis are described in section 4.1. Both the outlines of the earthquake clusters and the search kernels were then used for extraction of potentially active fault zones. The results of the classification into potentially active fault zones and likely inactive fault zones are described in section 4.2.
  Insertion in Line 394:
  4.1 Earthquake clusters
  Insertion and adjustments of the text passage on earthquake cluster analysis results in Line 395:
  The outlining of earthquake clusters described in section 3.3 yielded outlines of 169 earthquake-clusters, 78 of which are located on South-Korean mainland (Figure 7). They cover (…)
  Insertion and adjustments in Line 395:
  4.2 Potentially active and inactive fault zones
  The feature extraction described in section 3.4 yielded 2 largely overlapping sets of in total 488 fault surface traces, which in the following are to be classified as “potentially active in recent times”, and a remainder of 1,040 fault zones which, at least according to the earthquake catalogue, are not located in immediate proximity to regions of recent earthquake activity (Figure 7).
  457 of the 1,528 known fault zones (…)
  
  Lines 524-527. Take into account references on migration of contaminants where highly permeable faults are mapped in the area of a proposed repository for nuclear waste. Sellafield (West Cumbria, NW England) with crystalline and sedimentary rocks, see references below:
  Gutmanis, J. C., Lanyon, G. W., Wynn, T. J., & Watson, C. R. (1998). Fluid flow in faults: a study of fault hydrogeology in Triassic sandstone and Ordovician volcaniclastic rocks at Sellafield, north-west England. Proceedings of the Yorkshire Geological Society, 52(2), 159-175.
  Medici, G., West, L. J. (2023). Reply to discussion on ‘Review of groundwater flow and contaminant transport modelling approaches for the Sherwood Sandstone aquifer, UK; insights from analogous successions worldwide’ by Medici and West (QJEGH, 55, qjegh2021-176). Quarterly Journal of Engineering Geology and Hydrogeology. 56 (1), qjegh2022–097. doi:10.1144/ qjegh2022-097.
  Reply: Thank you for hinting at these works. We read them with great interest and inserted them as reference at suggested instance in Line 527 and furthermore added a reference on fault damage zones. The sentence now reads as follows:
  Apart of determining whether fault zones are active or not, safety distances have to be assigned around fault zones (e.g. He et al., 2022) and for this purpose the areas of corresponding damage zones (e.g. Kim et al., 2004) have to be estimated in order to ascertain that the confinement functions of a repository are not lost due to nearby fault activity or by perturbations such as increased water, gas and contaminant permeability due to fault displacement (e.g. Gutmanis et al 1998; Medici & West 2022, 2023).
  Corresponding references additional to those proposed above:
  Kim, Young-Seog, David C.P. Peacock, David J. Sanderson 2004. Fault damage zones. Journal of Structural Geology 26 (2004) 503–517. doi: 10.1016/j.jsg.2003.08.002
  Medici, G and West, LJ (2022) Review of groundwater flow and contaminant transport modelling approaches for the Sherwood Sandstone aquifer, UK; insights from analogous successions worldwide. Quarterly Journal of Engineering Geology and Hydrogeology, 55 (4). qjegh2021. ISSN 1470-9236 https://doi.org/10.1144/qjegh2021-176
  
  Lines 760-938. Please, insert the relevant references suggested above on migration of contaminants in an area of a proposed nuclear waste repository cut by normal faults.
  Reply: Done as described in the reply to the previous comment.
  
  Figures and tables
  Figure 1. Specify the number of faults in the rose diagrams that should be 457+163+128 to match figure 8.
  Reply: The rose diagrams in Figure 1 show the fault orientations of all 1,528 fault surface traces available for analysis (where 488 are active and 1,040 are currently inactive according to our results). We now indicate that number on the rose diagrams in Figure 1 as we did in Figure 8 for the subsets of potentially active faults.
  
  Figure A1. There is a correlation between length of a fault and its maximum capacity in terms of magnitude of earthquakes. This point has not been fully discussed in your manuscript. Please, consider my point due to the fact that you have produced a very important figure.
  Reply: You are right, this correlation previously was addressed in the data description, but we now re-located it to the discussion. More specifically, we now hint at the correlation in the data description, and moved magnitude and maximum displacement estimates for Korean fault zones from the data description to a new subsection (section 5.2) of the discussion. Furthermore, we deleted the average earthquake magnitude estimates which previously were indicated in Tables 2 and A1 for the different rock types, since it doesn’t add too much value to the manuscript to know that each rock type hosts fault zones which at average may produce magnitude 6 earthquakes. Also since the magnitude estimates are now discussed later in the manuscript, the Figure A2 showing the estimates has been moved down and now is Figure A6 (the numbers of Figures A3-A6 have been changed accordingly).
  Insertion in Line 232:
  Considering that fault length distributions typically follow a power law having most fractures in the short length range it furthermore attracts the attention that the length range <2 km only comprises 238 fault segments in comparison to the 2 – 4 km range described above which comprises 485 fault segments (Appendix Figure A1). This discrepancy hints at that our dataset probably is largely incomplete in the length range <2 km, and that we presumably lack several hundred fault surface traces in this length range which haven’t been noticed and mapped yet for the determination of damage zone extents. Moreover, note that some of the major and minor local fault segments may be part of larger fault strands, which must be taken into account when applying fault zone scaling laws to determine e.g. the magnitudes of earthquakes, the maximum displacement, and the damage zone widths these fault zones can produce, if a future earthquake event ruptures the entire fault strand (see sections 5.2 to 5.4). The main fault strand of the Yangsan fault zone in our dataset e.g. is represented by a 126 km continuous segment line and some shorter segments to the NE and SW, rather than extending along the previously mentioned 200 km. Estimates obtained from application of scaling laws to this data thus inevitably will be underestimates.
  Text passage moved from the data description to section 5.2 starting in line 589:
  5.2 Capability of faults to rupture a canister for high-level radioactive waste disposal
  Earthquake magnitude commonly correlates with the geometric features of the rupture it produces, such as its rupture length, the average and maximum displacements along the fault plane and the slip rate (e.g. Wells and Coppersmith, 1994). Knowing the rupture length of a particular fault hence amongst others enables estimation of the magnitude it can produce. According to Wells and Coppersmiths (1994) empirically determined relationship between Moment Magnitude M and surface rupture length SRL of a fault (M = 5.08 + 1.16*log (SRL [km])), the fault zones investigated in this work, would be capable of producing earthquakes in the magnitude range 4.2-7.5 (Appendix Figure A6). The largest earthquake magnitude of M 7.5 was obtained for the 126 km long surface rupture trace of the Yangsan fault. Despite being an underestimate, since we only consider a fraction of the actually 200 km long fault strand, this estimate is quite a bit larger than the maximum magnitude of 6.8 estimated from fault displacement by Kyung et al. (2010) for the Yangsan fault zone. Modern canisters for disposal of spent nuclear fuel as e.g. used in Finland and Sweden, and as planned to be used also in Korea (Choi et al., 2013) consist of a copper-shell and a massive cast iron inset, which are designed to withstand an arbitrary shear of 0.05 m (SKB, 2010). Thus, the maximum displacement on a fault cutting through the disposal site shall not exceed 0.1 m according to the results of the safety assessments performed by SKB (SKB, 2010). Relating maximum displacement Dmax to the surface rupture lengths SRL of the faults in our dataset utilising the regression line (log(Dmax [m]) = -1.38 + 1.02*log(SRL [km])) of Wells and Coppersmiths (1994), all of the faults considered here would be capable to produce a maximum displacement which exceeds 0.1 m.
  
  Table 1a. The content under “Map Colour” in unclear. Please revise it or apply changes to the caption.
  Reply: We revised the caption and table heading to show more clearly that we provide the hexadecimal colour codes and the RGB colour codes of the colours used for rock type distinction.
  
  Additional to the requested changes we made minor corrections.
  Line 216: correction/harmonization of the spelling of the name of the Okcheon fold belt, previously “misspelled” in this instance as Ogchon belt
  Line 325: The number of the fault zone segments picked by the circular 1-km search kernel is 142 and not 128. We corrected that number accordingly and the sentence now reads: “This slight enlargement of the search kernel area in comparison to a more circular area if using a smaller pixel size of 0.1 km, resulted in picking 163 instead of only 142 earthquake events in the vicinity of known faults in South-Korea.” This change does not cause any conflict with any other contents or statements in this work.
  Line 340: We inserted “and objects” into “A contour map was then created from this heatmap raster image data (Figure 6), in order to enable working with the heatmap information as a tool for selecting defined areas and objects in other datasets such as our fault zone traces located within earthquake clusters”
  Line 495: We moved the contents of the brackets one word further to the back to place it behind the word it refers to (yellow and red lines in Figure 7).
  Line 497: We inserted “usually” into “…, which is consistent with the fact that fractures optimally oriented for slip usually are highly dipped for the strike-slip stress regime (Xie & Min, 2016).”
  Line 640: We inserted another “potentially active” into the fault length+activity classification which we proposed at the end of the discussion of our manuscript and simplified the length description for the category of minor inactive faults from “(10m-1km)” to (<1 km). The sentence now reads as follows: “The authors of this work thus propose using 3-km-buffers applied both to potentially active supra-regional and regional fault zones (>10 km length), 1-km-buffers applied to potentially active major and minor fault zones (<10 km length) and further to apply 100-m-buffer zones to inactive regional faults (> 10 km length), 50-m-buffers to inactive major faults (1-10 km length), and 5-m-buffers to inactive minor faults (<1 km length).”
  
  Citation: https://doi.org/10.5194/egusphere-2023-2674-AC1
CC2:
'Comment on egusphere-2023-2674', Ki-Bok Min, 30 Apr 2024

This is an interesting and useful work for the siting of geological repository for nuclear waste. Slip tendency analysis has not been actively used in the community of nuclear waste repository community and this work can contribute to paving the way for more wider application.
The merit of this work is that it can cover entire country by utilizing GIS based approach and this work can be a decent starting point for initializing site investigation.
One need to bear in mind that there are uncertainties associated with slip tendency as the input parameter such as friction coefficient and in situ boundary stresses often have uncertainties.
This work has never been used in Korea, and, overall, this is a nice contribution.

Citation: https://doi.org/10.5194/egusphere-2023-2674-CC2
- AC2: 'Reply on CC2', Stefan Bredemeyer, 02 Aug 2024
  
  Dear Ki-Bok Min,
  thank you for your positive feedback to our work!
  Slip tendency analysis to our knowledge until now rather was conducted at a later stage of the site searches, as e.g. was done for the Yucca Mountains in the United States of America by Ferril et al. (1996) and for the HLW repository site in China (Qin et al., 2024).
  However, as we were able to demonstrate in this work and as is shown in the work of Röckel et al., 2022, who presented a comprehensive slip tendency analysis for the fault zones of Germany, it may be of high value to use this type of fault zone characterization already during earlier stages of the site selection procedure in order to avoid late surprises regarding the suitability of an area.
  Furthermore, it is advisable to repeat the entire analysis along with the acquisition of new information obtained through field studies and to update analysis results as soon as new data is available. Also, at the current stage of the site search in South-Korea there still exists a large literature body with details on fault geometry which has not been included in the fault zone dataset provided here.
  Thank you also, for pointing out the uncertainties associated to the input parameters of the slip tendency analysis. The uncertainties of the stress-field data used here, are tackled in detail by Soh et al. (2018). They may be reduced by including more than the 152 well constrained focal mechanisms Soh et al. (2018) used for the inversion.
  
  Best regards,
  Stefan Bredemeyer
  
  Citation: https://doi.org/10.5194/egusphere-2023-2674-AC2
RC1:
'Comment on egusphere-2023-2674', Tiegan Hobbs, 14 May 2024
I wish to thank the authors and editor for the chance to review this very engaging paper, however, I find that I must recommend that this paper not be published in its current format. The manuscript uses a 32-year seismic catalogue of under 900 earthquakes with average magnitude 2.6 and a national database of bedrock faults to try to identify active faults in South Korea, for application to selection of nuclear waste disposal sites. The method used to process the seismic data for comparison to fault locations is not aligned with industry best practice for smoothing seismicity catalogues for hazard assessment, and gives no consideration to 3D structure of these faults at depth. Additionally, and most critically, this catalogue is inherently far too short to draw any reasonable conclusion about the kind of long-term hazard (thousands of years) that must be understood in order to site facilities for nuclear waste storage at shallow depth in the earth’s crust. Finally, the discussion of buffer widths is interesting, but does not present enough information to helpfully allow a reader to understand what buffers may be most appropriate in South Korea. The paper concludes with what I would describe as a fairly subjective recommendation about a set of active faults and buffer widths to be used for nuclear storage site screening in North Korea, without sufficient analysis or discussion of hazard from unmapped faults.

I worry that policy makers or stakeholders may misinterpret this paper as an authoritative seismological screening of safe sites for nuclear waste storage, which would require at least a probabilistic assessment of seismic hazard over very long return intervals, further paleoseismic investigation of a region surrounding any site of interest, and 3D geological models. However, I applaud the intention of wanting to find a technique for creating a national inventory of active faults for public safety using widely available data.

For publication in future, I would want to see the analysis redone with all seismic data (going back hundreds of years), using a well-established seismicity smoothing method (or implementing a national seismic hazard model), and potentially including the impact of, say, age of deposition/emplacement/latest deformation for geological units and perhaps geodetic strain accumulation in determining which faults are likely to rupture in the current tectonic regime. I believe conclusions should be toned down in either case, presented alongside a full discussion of known unknowns when it comes to active fault inventories, and described only as a possible approach for very high-level screening with large uncertainties that should be dealt with at each site of interest.

I regret that this review is quite negative, but feel that nuclear storage represents a very high stakes application that warrants a high degree of scrutiny. If you have any questions, please don’t hesitate to reach out.

Kind regards,
Tiegan Hobbs

Comments, in addition to in-text revisions:
MAJOR: The definition of ‘active’ fault already exists: a fault with evidence of earthquake rupture in the Holocene. There should be a fuller discussion of how and why the criteria for ‘activity’ were chosen for this paper, and I think it would be appropriate to use a different word than ‘active’ for the faults meeting this criteria. This is in particular because the criteria used in this paper are very relaxed and somewhat imprecise (e.g. an earthquake within 1km of a fault makes that fault active, without regard for the 3D structure of that fault or depth of the earthquake.

MINOR: Please describe what ‘elaborate’ field work is, as this term is used repeatedly.

MAJOR: How does the seismic hazard model of South Korea compare to the regions identified as potentially suitable for nuclear waste disposal? This analysis does not consider regions which may not have current/modern seismicity but which are capable of rupturing in the future, which would be very important for nuclear waste storage. For example: blind active faults (therefore not mapped), faults with very long recurrence which have minimal/no ongoing seismicity until the next major event.

MINOR: Fix Figure 2 to make it “flow”. See suggestions in text.

MAJOR: Relocation procedure and final spatial uncertainty of the KMA catalogue need to be stated explicitly, as it is different than the resolution of the data. Same for the older datasets that were excluded. You should also indicate the magnitude of completeness for each catalogue and show the seismic stations used in the relocation. If there are fewer stations in the north mainland it could explain the apparent paucity of events. I also think it would be critical to include all older data, in order to have as much information as possible about the long-term seismicity in South Korea, particularly for a nuclear waste application. There could have been a swarm of seismicity in the north that just hasn’t been active in the very short 32 year period of consideration (many fault lines have recurrence intervals of hundreds to thousands of years).

MAJOR: The use of Kernel Density Estimation to create a seismicity model is out of line with current best practice from the Seismic Hazard Modelling community. They commonly use a smoothed seismicity approach (e.g. https://doi.org/10.1785/BSSA0860051372, https://doi.org/10.1785/0220180040, https://doi.org/10.1016/j.jafrearsci.2020.103894, https://doi.org/10.1177/87552930231215428). If the authors wish to break from the best practice, I think they should explain the physical reason for the change in approach. As it stands, the authors are using a very small/short-term seismic catalogue, with very few details given about its magnitude of completeness, relocation, declustering, etc., and applying a KDE method that ends up representing seismicity over a broad area in a misleadingly small pixel. These small pixels, representing large regions, are then used to identify which faults are potentially active with implications for nuclear waste storage. As the authors identify 457 fault segments based on this approach, with roughly 900 earthquakes of mostly small magnitude, they are therefore identifying faults with an average of 2 events per fault. This type of analysis should be handled by observational earthquake seismologists with experience in seismic hazard modelling, and preferably should be treated by consideration of a fully probabilistic seismic hazard model designed for assessment of long recurrence hazard sources. This will mean including seismicity from a much longer interval and likely consideration of seismic source zones as well as regions which may be capable of seismicity even if they have not had a modern earthquake recorded instrumentally.

MAJOR: Figure 6 colormap is wrong, according to caption, and also shows light yellow areas where there is less than 1 but more than 0 earthquakes. This is impossible. See note in text.

MINOR: Figure A3 needs to be part of the main text.

MAJOR: The 1km analysis seems both (1) needlessly complicated and (2) insufficient to actually help identify causative faults of these earthquakes. If you want to find faults within 1km of an earthquake, you just need to draw a 1km buffer around each epicenter and intersect it with fault lines, not obscure it with a heatmap and then intersect that polygon. That analysis is different than finding faults within 1km of an earthquake. Secondly, this analysis doesn’t take into account three dimensional structure of the faults at depth – it may be that they actually are much more than 1km apart. What you are really doing here is seeing if the surface trace of a bedrock fault is within 1 km of the surface projection of earthquakes. This is not a meaningful assessment of whether a fault is seismically active. Finally, your average earthquake size is magnitude 2.6, which is very small. As a seismologist, I wouldn’t necessarily consider that every small event needs to occur on a major (capable of M5+) fault. Therefore, labelling a fault as active based, potentially, on a single small earthquake is not valid. I must therefore say that I don’t feel this analysis has sufficient scientific merit to be presented in a published paper as is.

MAJOR: Line 403 states that faults which did not meet the criteria used herein (proximity to clusters of seismicity or a single event from a 32-year catalogue of less than 900 events) are “classified as not currently active”. This is potentially dangerously misleading, as the analysis of the potential seismic hazard from these faults has not been appropriately nuanced/sufficient to make this claim. But stakeholders such as nuclear energy operators, dam operators, etc. may cite this study to justify not completing further investigations. It may also lead stakeholders or the public to believe that only known faults are capable of hosting earthquakes in the future, which we know to be untrue. Therefore, I must object in very strong terms to this analysis and phrasing being published as is.

MAJOR: The slip tendency analysis does not use any known dip angles, but instead calculates the slip tendency values at representative dips. The authors then conclude that because their faults show high slip tendency at high dip angle that their faults are indeed active, and that since they are active they are probably high angle (circular logic). I don’t find this particularly convincing without knowing the true dip angle of these faults. They then investigate the slip tendency value of the “not active” faults as well. It seems that this analysis just shows that almost any fault with high dip angle will have high slip tendency, rather than proving something about this particular subset of “active” faults.

MAJOR: This work talks about “behavioural segments” for buffer width determination, but the current fault dataset may not be segmented in the same way. Usually behavioural segments are based on a detailed knowledge of the behaviour of the fault (e.g. creeping rate, seismicity rate, characteristic rupture extent, multi-event rupture segmentation, location of asperities), whereas in my experience, bedrock fault databases are segmented somewhat arbitrarily. Therefore, the authors should prove that this dataset has sufficient behavioural segmentation information for this method to be applied. They should also address the issue of faults which are capable of rupturing together, such as is seen in California (UCERF model), Alaska (Denali-Tintina Fault), etc.

MINOR: Ensure all maps have lat/lon markings at the periphery.

MAJOR: When establishing buffer zones, the authors have not discussed horizontal uncertainty in the location of mapped faults. Depending on when they were mapped (were handheld GPS units used to denote high-precision locations or were geologists triangulating off of nearby landmarks to get an approximate location?), what criteria was used to determine their presence (was a fault zone observed or was a change in lithology assumed to be due to a fault which was then placed somewhere between the last location of lithology A and first location of lithology B), and the attention paid by cartographers digitizing paper maps (how were overlapping lithology boundaries harmonized? What behaviour was assumed under water bodies that were not mapped?), there can be significant uncertainties in a faults mapped location — certainly larger than 100m. The suggested guidelines are therefore unrealistically precise given our knowledge of faults.
Citation: https://doi.org/10.5194/egusphere-2023-2674-RC1
- AC4:
  'Reply on RC1', Stefan Bredemeyer, 12 Sep 2024
  Dear Tiegan Hobbs,
  
  Thank you for the appreciation of our work and the helpful suggestions for further improvement of the analysis and manuscript. We agree that the earthquake catalogue used here should be extended towards the beginning of instrumental seismic recordings that is ~1905 to repeat the analysis presented here. For this purpose, we now started compiling respective catalogue. Using the M>2 had the purpose to reduce the catalogue to those events that eventually may produce a rupture extending to earth’s surface, since we made a comparison the epicentre locations to the surface fault zone traces. 3D structures are not known yet for most of the major fault zones apart of some spatially confined information obtained during the “Active Fault Zone studies” conducted by the Korean Ministries for identification of Quaternary active faults.
  Consideration of the historical earthquake record ranging back to 2 A.D. will require a different approach.
  We agree that the discussion of the different buffer zones would benefit from further explanation which buffers are suited for South-Korea and better justification of the buffer zone choice we made in the end of the manuscript.
  Since this is not a probabilistic work, and the identification of fault zone traces we conducted here is supposed to support further mapping studies onsite which includes identification of further fault zone traces, the work presented in this manuscript should not at all be understood as a complete seismic hazard analysis from which we directly can derive the repository site. We agree that should be stated more clearly at the end of the manuscript.
  Doing a full probabilistic analysis of the earthquake hazard for future scenarios is out of scope of this work. Similarly, the large literature body existing on both active and inactive fault zones still requires being compiled for in depth studies of the features fault zone and earthquake features you mention.
  
  Comments, in addition to in-text revisions:
  MAJOR: The definition of ‘active’ fault already exists: a fault with evidence of earthquake rupture in the Holocene. There should be a fuller discussion of how and why the criteria for ‘activity’ were chosen for this paper, and I think it would be appropriate to use a different word than ‘active’ for the faults meeting this criteria. This is in particular because the criteria used in this paper are very relaxed and somewhat imprecise (e.g. an earthquake within 1km of a fault makes that fault active, without regard for the 3D structure of that fault or depth of the earthquake.
  
  A: Yes, you are right, we add more detail with respect to activity definition and the time scale relevant for repository safety.
  Changes in Lines 49-53: For this reason, active fault zones, i.e. fault zones which might experience slip within the time period of concern for the safety of the repository have to be distinguished from inactive or extinct faults in mechanically more stable crustal blocks (e.g. Rybicki et al., 1985; Muir Wood & Mallard 1992). It has to be noted though, that faults which appear to have been inactive for a long time do not necessarily stay inactive, since intra-continental earthquakes typically occur through reactivation of previously inactive faults, which particularly applies to faults in collision belts such as found on the Korean Peninsula (e.g. Kwon et al., 2009; de Jong et al., 2015). Such inactive faults readily can be reactivated by stress redistribution due to change in direction of the crustal movement, the injection of fluids, or by triggering through distant earthquakes or movements along connected fault systems (e.g. Sibson, 1985; Cappa & Rutqvist, 2011). Well-sealed and healed inactive fractures in contrast may develop a strength that even exceeds that of the surrounding rocks (Townend and Zoback, 2000; Børlykke 2001).
  Moreover, the time period of concern to assess fault activity, the capability of a fault and the safety of a repository should be considered depending on the level of tectonic activity in a region (IAEA, 2010). More specifically, while short time intervals of 100k years reaching back to the Upper Pleistocene may be appropriate for the assessment of fault activity in high seismicity regions with short earthquake recurrence intervals such as Japan (Kim & Chang, 2020), much longer time intervals in the range of Mio of years have to be considered in tectonically less active regions covering periods spanning from the Quaternary to the present with possible extension to the Pliocene in stable cratonic regions (IAEA, 2010). The major fault zones of Korea are characterized by slow slip and very long recurrence intervals of large magnitude earthquakes (Choi et al., 2015) and the current stress regime, i.e. the neo-tectonic setting in and around the peninsula was initiated in the early Pliocene. Active fault zone studies conducted for the repository site search in South-Korea thus must consider the period ranging from Quaternary to present (NEMA, 2012; KIGAM, 2021).
  Bjørlykke, Knut (2001). How faulting keeps the crust strong: Comment and Reply. Geology, February 2001, 29 (2): 189–190. DOI: 10.1130/0091-7613(2001)029<0189:HFKTCS>2.0.CO;2
  Choi, Jin-Hyuck, Young-Seog Kim, Sung-Ja Choi (2015). Identification of a suspected Quaternary fault in eastern Korea: Proposal for a paleoseismic research procedure for the mapping of active faults in Korea. Journal of Asian Earth Sciences 113 (2015) 897–908. Doi: 10.1016/j.jseaes.2015.09.014
  IAEA (International Atomic Energy Agency), 2010. Seismic Hazards in Site Evaluation for Nuclear Installations. Chapter 8: Potential for Fault Displacement at the Site. : Specific Safety Guide No. SSG-9, ISBN . ISBN: 978-92-0-102910-2. STI/PUB/1448.
  Kee, W.-S., Kihm, Y.H., Lee, H., Cho, D.L., Kim, B.C., Song, K.-Y., Koh, H.J., Lee, S.R., Yeon, Y.-K., Hwang, S., Park, K.G., Seong, N.-H. (2009). Evaluation and database construction of quaternary faults in SE Korea. Korea Institute of Geoscience and Mineral Resources, IP2006-047-2009(1) (in Korean).
  KIGAM 2021. KIGAM Annual Report. Korea Institute of Geosciences and Mineral Resources, November 11, 2021. https://www.kigam.re.kr/
  Kim, Jiyeon and Chandong Chang (2020). Evolution of slip susceptibility of Quaternary faults in southeastern Korea from Quaternary to present day. Geosciences Journal Doi: 10.1007/s12303-020-0053-4
  Townend, John, Mark D. Zoback (2000). How faulting keeps the crust strong. Geology; May 2000; v. 28; no. 5; p. 399–402. DOI: 10.1130/0091-7613(2000)28<399:HFKTCS>2.0.CO;2
  MINOR: Please describe what ‘elaborate’ field work is, as this term is used repeatedly.
  
  A: We refer to structure geological and palaeoseismic studies. We will add that information.
  MAJOR: How does the seismic hazard model of South Korea compare to the regions identified as potentially suitable for nuclear waste disposal? This analysis does not consider regions which may not have current/modern seismicity but which are capable of rupturing in the future, which would be very important for nuclear waste storage. For example: blind active faults (therefore not mapped), faults with very long recurrence which have minimal/no ongoing seismicity until the next major event.
  
  A: This work focusses on identification of potentially active surface traces, i.e. the seismic hazard model does not exist yet. And we advise not to understand our results as recommendation for a potentially suited site, since there are many other factors which additionally have to be considered including other natural and human related hazards and planning related factors.
  MINOR: Fix Figure 2 to make it “flow”. See suggestions in text.
  
  A: The intention was to show the results in the order they were derived from previous analysis results. But we can apply changes as you suggest.
  MAJOR: Relocation procedure and final spatial uncertainty of the KMA catalogue need to be stated explicitly, as it is different than the resolution of the data. Same for the older datasets that were excluded. You should also indicate the magnitude of completeness for each catalogue and show the seismic stations used in the relocation. If there are fewer stations in the north mainland it could explain the apparent paucity of events. I also think it would be critical to include all older data, in order to have as much information as possible about the long-term seismicity in South Korea, particularly for a nuclear waste application. There could have been a swarm of seismicity in the north that just hasn’t been active in the very short 32 year period of consideration (many fault lines have recurrence intervals of hundreds to thousands of years).
  
  A: We agree that the manuscript would benefit from adding more detail such as the magnitude of completeness and spatial uncertainties. The earthquake catalogue for the entire instrumental period reaching back to ~1900 still needs to be compiled, but we started doing this already. The short period we previously considered here clearly shows that most of the Quaternary active fault zones likely have been active in these 32 years, i.e. it could be that recurrence intervals of earthquakes in South-Korean fault zones actually are shorter than previously thought.
  MAJOR: The use of Kernel Density Estimation to create a seismicity model is out of line with current best practice from the Seismic Hazard Modelling community. They commonly use a smoothed seismicity approach (e.g. https://doi.org/10.1785/BSSA0860051372, https://doi.org/10.1785/0220180040, https://doi.org/10.1016/j.jafrearsci.2020.103894, https://doi.org/10.1177/87552930231215428). If the authors wish to break from the best practice, I think they should explain the physical reason for the change in approach. As it stands, the authors are using a very small/short-term seismic catalogue, with very few details given about its magnitude of completeness, relocation, declustering, etc., and applying a KDE method that ends up representing seismicity over a broad area in a misleadingly small pixel. These small pixels, representing large regions, are then used to identify which faults are potentially active with implications for nuclear waste storage. As the authors identify 457 fault segments based on this approach, with roughly 900 earthquakes of mostly small magnitude, they are therefore identifying faults with an average of 2 events per fault. This type of analysis should be handled by observational earthquake seismologists with experience in seismic hazard modelling, and preferably should be treated by consideration of a fully probabilistic seismic hazard model designed for assessment of long recurrence hazard sources. This will mean including seismicity from a much longer interval and likely consideration of seismic source zones as well as regions which may be capable of seismicity even if they have not had a modern earthquake recorded instrumentally.
  
  A: We did not aim to produce a seismic model, nor this is a probabilistic approach assessing future earthquake probabilities in fault zones. We thus add more information on the physical reason for doing this spatial comparison of earthquake epicentres and surface traces of fault zones.
  MAJOR: Figure 6 colormap is wrong, according to caption, and also shows light yellow areas where there is less than 1 but more than 0 earthquakes. This is impossible. See note in text.
  
  A: We will check whether the colorscale is correctly displayed and crop any unnecessary color-information from the map to avoid such confusion.
  MINOR: Figure A3 needs to be part of the main text.
  
  A: Ok, we add that figure and corresponding information to the main text.
  MAJOR: The 1km analysis seems both (1) needlessly complicated and (2) insufficient to actually help identify causative faults of these earthquakes. If you want to find faults within 1km of an earthquake, you just need to draw a 1km buffer around each epicenter and intersect it with fault lines, not obscure it with a heatmap and then intersect that polygon. That analysis is different than finding faults within 1km of an earthquake. Secondly, this analysis doesn’t take into account three dimensional structure of the faults at depth – it may be that they actually are much more than 1km apart. What you are really doing here is seeing if the surface trace of a bedrock fault is within 1 km of the surface projection of earthquakes. This is not a meaningful assessment of whether a fault is seismically active. Finally, your average earthquake size is magnitude 2.6, which is very small. As a seismologist, I wouldn’t necessarily consider that every small event needs to occur on a major (capable of M5+) fault. Therefore, labelling a fault as active based, potentially, on a single small earthquake is not valid. I must therefore say that I don’t feel this analysis has sufficient scientific merit to be presented in a published paper as is.
  
  A: The 1km analysis was done to capture all seismic events which occurred near a fault zone, in order to add those to the fault zones intersecting or touching earthquake clusters. Since our analysis focussed on M>2 earthquakes which according to Zielke et al., (2001) typically produce ruptures up to earth’s surface, they very likely can be associated to a nearby fault zone trace. The hypocentre locations of major earthquakes at greater depth which may occur further away from the fault trace at the surface are covered by the earthquake cluster analysis. Our analysis intends to cover all surface traces in range of such clustered seismic activity, in order to provide a basis for further field investigations that aim at amongst other the identification of Quaternary active fault zones.
  MAJOR: Line 403 states that faults which did not meet the criteria used herein (proximity to clusters of seismicity or a single event from a 32-year catalogue of less than 900 events) are “classified as not currently active”. This is potentially dangerously misleading, as the analysis of the potential seismic hazard from these faults has not been appropriately nuanced/sufficient to make this claim. But stakeholders such as nuclear energy operators, dam operators, etc. may cite this study to justify not completing further investigations. It may also lead stakeholders or the public to believe that only known faults are capable of hosting earthquakes in the future, which we know to be untrue. Therefore, I must object in very strong terms to this analysis and phrasing being published as is.
  
  A: You are right we tone down our classification of faults as currently not active, because they are simply not covered by the seismic catalogue we used for production of our results.
  MAJOR: The slip tendency analysis does not use any known dip angles, but instead calculates the slip tendency values at representative dips. The authors then conclude that because their faults show high slip tendency at high dip angle that their faults are indeed active, and that since they are active they are probably high angle (circular logic). I don’t find this particularly convincing without knowing the true dip angle of these faults. They then investigate the slip tendency value of the “not active” faults as well. It seems that this analysis just shows that almost any fault with high dip angle will have high slip tendency, rather than proving something about this particular subset of “active” faults.
  
  A: The dip angles of major fault zones in fact are currently not known in sufficient detail for all fault zones in South-Korea. Thus, the slip tendency analysis presented here aims at showing how slip tendency of fault zones varies at different dip angles. Furthermore, we can show there is a number of inactive fault zones with high slip potential, which did not show seismic activity in the past three decades.
  In order to avoid that our statement regarding the dip angle in a strike-slip stress regime is misunderstood as circular logic the sentence requires being rephrased.
  The slip tendency generally increases with the dip angle for the current study, which is consistent with the fact that fractures optimally oriented for slip typically are highly dipped for the strike-slip stress regime (Xie & Min, 2016).
  MAJOR: This work talks about “behavioural segments” for buffer width determination, but the current fault dataset may not be segmented in the same way. Usually behavioural segments are based on a detailed knowledge of the behaviour of the fault (e.g. creeping rate, seismicity rate, characteristic rupture extent, multi-event rupture segmentation, location of asperities), whereas in my experience, bedrock fault databases are segmented somewhat arbitrarily. Therefore, the authors should prove that this dataset has sufficient behavioural segmentation information for this method to be applied. They should also address the issue of faults which are capable of rupturing together, such as is seen in California (UCERF model), Alaska (Denali-Tintina Fault), etc.
  
  A: You are right, the behavioural segmentation of the dataset still requires being adjusted, which however will require thorough compilation of information on individual fault zones, as previously mentioned. Similarly, the buffers in general have to be adjusted throughout the process of adding detail to the datasets.
  MINOR: Ensure all maps have lat/lon markings at the periphery.
  
  A: Ok, we add lat-lon markings in all maps, where missing.
  MAJOR: When establishing buffer zones, the authors have not discussed horizontal uncertainty in the location of mapped faults. Depending on when they were mapped (were handheld GPS units used to denote high-precision locations or were geologists triangulating off of nearby landmarks to get an approximate location?), what criteria was used to determine their presence (was a fault zone observed or was a change in lithology assumed to be due to a fault which was then placed somewhere between the last location of lithology A and first location of lithology B), and the attention paid by cartographers digitizing paper maps (how were overlapping lithology boundaries harmonized? What behaviour was assumed under water bodies that were not mapped?), there can be significant uncertainties in a faults mapped location — certainly larger than 100m. The suggested guidelines are therefore unrealistically precise given our knowledge of faults.
  
  A: We will add some more information on the horizontal uncertainty in the location of mapped faults and provide a link for downloading the metadata of the dataset including such descriptions. Most water-bodies in South-Korea are rivers which typically are filled with sediments.
  
  Last but not least we aim at conduction of the minor improvements and addition of information which you suggested in the appended manuscript. We are sorry for the delayed reply, but the main author experienced major computer problems throughout the writing of this reply.
  
  Kind regards,
  Stefan Bredemeyer
  
  Citation: https://doi.org/10.5194/egusphere-2023-2674-AC4

RC2: 'Comment on egusphere-2023-2674', Ramon Arrowsmith, 07 Jun 2024

Review of

GIS-based characterization of fault zones in South-Korea using

information on seismicity, in-situ stress and slip tendency - Evaluation

of respect distances for nuclear waste disposal site screening

by

Bredemeyer et al

Reviewer

Ramon Arrowsmith

Summary

Interesting paper presenting a methodological exploration largely geospatially-based to identify potentially active faults.

I appreciate the detailed work and the clear evidence of the basis on rich datasets. The readability can be significantly improved (and thus the paper made nore effective) if the writing were streamlined by trimming substantial extraneous detail from the main text and placing it accessibly in the appendix. This reads in places like a report to an agency and in that sense is thorough. However, many readers will struggle to find what are the important contributions if they are diluted.

I am concerned about the definition of recently active faults as presented in the paper. What we learn is that there are faults, variably oriented relative to the current stress state, which are associated with instrumental seismicity. It is logical that they may have been active in the past, and will be again in the future, but just because a fault not producing seismicity over that last few decades, does not mean it is inactive. There are many cases of seismically quiet faults (locked) which have produced great earthquakes. This is particularly a concern in my view for the intraplate low deformation rate setting of the Korean peninsula. So, the semantics of active faults need to be addressed.

General comments

1) While I appreciate data rich papers--and this is one--there is too much detail in places making it hard to read and diluting the impact. For example there are places I have tried to highlight with unnecessary detail in the text. I have noted for example, that Tables 1a and 1b are not apparently needed for the main text. I think the text or caption should better make the case for the inclusion of such detail in the main text. Otherwise, put it in the appendix and refer to it. In other cases, we have long and dense enumeration of data that can be stated more simply for the reader. In another example, Why is knowing that the interquartile range of the fault surface trace segmentation important? How does it aid in understanding the processes and the hazard? If you cannot clearly demonstrate the relevance, then it should be cut. THis comment also refers to at least some of the seemingly trivial logical expressions used in QGIS.

2) I am not comfortable ascribing even a generic probability increase of rupture, given the significant emphasis on disciplined probabilistic analysis in geological repository safety assessment. See for example, "Updated Implementation Guidelines for SSHAC Hazard Studies" NUREG-2213. Where in that structured approach, might this analysis fit, and how indeed does it indicate a probability increase for rupture?

3) The discussion of the comparison of the buffering is interesting. However, why is it a problem to exclude some percentage of the territory? That is the point of the exercise. And, there is still at least 70% remaining. I think this point needs to be more carefully stated.

4) It is not clear what the slip tendency adds. I think it is a nice addition, but it needs to be integrated with the screening and the buffer analysis somehow, otherwise it appears like an unjustified addition to the paper.

Specific comments tied to the manuscript by element

Title--The title is ok but I don't like acronyms in titles and would think of restructuring it. Perhaps:

Characterization of fault zones in South-Korea using information on seismicity, in-situ stress and slip tendency - Evaluation of respect distances for nuclear waste disposal site screening

Abstract--Generally adequate abstract. However at lines 20-25, I would like to see a sentence or two more on the method details and how the results were integrated, and then compact or shorten other aspects of the abstract that are more general.

Line 40--Perhaps this is a bit pendantic, but GIS is really just a tool to do geospatial analysis. I think a bit more general tone about what it is about the study of the structures that is needed and their integration, which then requires the GIS toolkit is a bitter logic flow; rather than making it sound like the GIS is a hammer looking for a nail to pound.

Line 45--"which will be" what does that mean and when? Does it mean when the paper is published or some other time?

Lines 49-53--This is awkwardly stated. The active faults are those which might experience slip within the time period of concern for the repository. There activity in the (late) Quaternary is what we would like to use to anticipate their (relative) activity.

Lines 54-56--Important indeed to recognize the potential for reactivation of structures for the reasons noted; can also cite Townsend and Zoback; weak faults in strong crust?

Line 60--Could mention that the reason to do the field work and have the lidar is in part to characterise the landscape and surficial geology which may indicate a relationship between the faults and those features.

Lines 65-67--I get the idea of the association between the seismicity and well oriented faults, but I am not comfortable ascribing even a generic probability increase of rupture, given the significant emphasis on disciplined probabilistic analysis in geological repository safety assessment.

Line 70--I think it is important to acknowledge that the instrumental seismicity reflects a very short portion of the likely earthquake recurrence rate of an intraplate setting like Korea. There is a lot of literature on this point. Yes, you can use it because it is available and of high quality, but we know that it only gives an incomplete in time record of activity.

Lines 73-97--Good summary of the analyses to be done.

Line 99--Geology no capitalization

Line 105--data are plural

Line 135--Need to start a new paragraph here "A first concept..."

Tables 1a and 1b are not needed in the main text. The detail distracts from a straightforward summary of the work and the reader can get a sense of what the point is by review of figure 1.

Lines 165-175--This might be better as a small table.

Lines 191-195--I think this is too much detail. Why is knowing that the interquartile range of the fault surface trace segmentation important? How does it aid in understanding the processes and the hazard? If you cannot clearly demonstrate the relevance, then it should be cut.

Paragraph at line 205--again, why is this detail on lineament lengths important?

Line 233--amongst this detail, we find what is interesting that indeed the tectonic domains become evident.

Table 2--Ok it is useful to see the geometrically defined possible magnitude of about M6.

Table 3--I don't see the importance of this table for the main text. What is the point of it? It shows the thorough analysis but...

Table 4--ok but there is extraneous information. Why is the perimeter important? THe percentages are useful, but not to two significant figures.

Paragraph at line 290--Heatmaps are a standard product so some of the basic description of them could be trimmed. It is really just a visualizatoin of the analysis which is the spatial kernal density.

Paragraph at line 390--Here we get to the results. I would be very careful with the language here. We have a horizontal spatial association between fault traces and instrumental seismicity. Indeed that is a reasonable assumption that at least some of the earthquakes occurred along those faults. We don't know how the fault systems vary with depth and we know that the earthquakes are occurring in the upper crust so this 2D association is a first step. "Potentially active in recent times" is ok but recent here has different meanings. Based on the analysis, it is strictly within the time since the earthquakes occurred locally (or catalog age overall). It does not mean recent geologically unless you want explicitly state the assumption that activity in the instrumental period is consistent with longer time scale activity in the past (and the future).

Table 6--I am not convinced that this is very helpful for the main text. It is useful I suppose to have a sense of which rocks have more and less potentially active faults, but that information by itself, without other context could actually hinder an assessment of optimal siting regions if it is applied blindly.

Line 455--good result that they compare well but it is not so evident in figure 10.

Section 5.2--Slip tendency analysis is a useful complement to the other elements of the research presented here.

Line 483--Stress state reference ok but how about a quick reference to other data sets including the world stress map?

Line 490--I am concerned about the lack of knowledge on dip of the faults as analyzed here, given its importance in the slip tendency calculations.

Lines 497 and else where--"highly dipped" is not a standard nomenclature in my view; I would expect to read "steeply dipping"

Lines 515 and vicinity--This paragraph needs to be rewritten a bit more simply. There is some repetition and the long sentences make it hard to read. I would also write "low dip" and I think the deviatoric stress being a quantity should be referred to as "greater" or "smaller" or "lower"

Line 520--I don't think you should proceed here without summarizing briefly what we have learned with the slip tendency? It looks like maybe the steeper faults within the seismicity zones show a greater overall slip tendency. Can this be quantified? maybe a histogram of slip tendency to complement at least some of the panels in Figures 11 and 12?

Tables 7 and 8 and 12 and 13--The main result for the text narrative is that the polygon coverage in total is about 2.65% or 5.26% or 9.29% or 20.45% of the onshore portion of South Korea. I don't read a justification for the rest of the detail in this figure to be included in the main text.

Table 10--Significant figures are too many. We don't need to know to the 1000th of a percent how much the rock types are covered. There is various error in the computations that probably exceeds some of the sig figs.

Line 595 area--THe QGIS logical expressions are trivial and the reader would normally expect that this was how it was done. i would put a summary of these expressions in the appendix

Line 620 and area--Here we get something interesting appearing: the discussion of the comparison of the buffering. However, why is it a problem to exclude some percentage of the territory? That is the point of the exercise. And, there is still at least 70% remaining. I think this point needs ot be more carefully stated.

Appendix--This has useful information. More from the main text could be placed here to make the main text more readable and effective. I would like to see the appendix begin with a short overview as to what is contained within it.

Line 760--Strange formatting errors. Did we also lose the beginning of the references?

THe references would be much easier to read if they had a hanging indent.

Figures

Figure 1--This is not a geologic map but rather a lithologic map with overlain faults. It seems to be ok for the purposes of this study.

Line 115--I don't think it is necessary to provide the detail on the number of polygons, fault traces, and lineaments. It is distracting.

Line 121--Parenthetic statement about the data sources is appreciated, but is too much detail for this caption. I would have a brief reference to the appendices here and then in the appendix, provide this kind of information for the rare reader who might want to recreate the figure.

Figure 2--I like the flow chart. I think some of the directionality at the lower right is not clear. I would think that the final output would be the discrimination of potentially active faults form all of the faults, followed by buffers around them. Thus, it should be more clear that the slip tendency is part of the screening and that the buffers are the final outputs.

Figure 3--nice looking figure.

Caption ending with detail on datasets may not be needed if that information can be presented in the appendix.

Figure 4--ok; useful to see the seismicity along with the fault network.

FIgure 5--ok; but the color bar does not need to be nearly as large as the map. However, given that FIgure 6 shows similar, and is used in subsequent analyses, I would cut figure 5.

FIgure 6--ok; the color bar is way too large. I would cut all contours except the ones that are used for the intersection and then the reader can see the color map underneath and that will be sufficient. Delete figure 5

Figure 7--This is an important figure. Mask the offshore given that it is not part of the analysis.

Figure 8--lots of detail; could simplify for main text to highlight the main point?

Figure 9--Seems like a useful figure

FIgure 10--Good but hard to get the comparison between the datasets

Figure 11--THis is an interesting figure. Please only use one color bar scale (given that the range is the same for all) and make it about half the current size that will clean up the figure a bit.

Figure 12--THis is an interesting figure. Please only use one color bar scale (given that the range is the same for all) and make it about half the current size that will clean up the figure a bit.

Figures 13 and 14 and 16--This is ok but not that helpful. The map scale is such that the lineweight seems to obscure the buffer polygons and so it does not actually add anything. We get slightly more out of the /50 case. Can this be combined with a reference to earlier fault maps and maybe then here provide some zooms to interesting regions? If you must keep these figures, you could probably make them two or three panels of a single figure given the large amount of repeated information?

Figure 15--This is a helpful zoom. But, it needs some lat longs and indication box in a prior figure so we know where it is.

Figure 16--At least here the buffers are large enough that they can be seen at this scale. But again, maybe combine this somehow into a multipanel figure with Figures 13, 14, and 16? Provide similar zooms as FIgure 15?

Review questions:

Does the paper address relevant scientific and/or technical questions within the scope of NHESS? YES

Does the paper present new data and/or novel concepts, ideas, tools, methods or results? Marginally. I think that the GIS analysis of the rich datasets is interesting and correct but I don't see significant novelty.

Are these up to international standards? YES

Are the scientific methods and assumptions valid and outlined clearly? Mostly. See my concerns about definition of active faults.

Are the results sufficient to support the interpretations and the conclusions? YES

Does the author reach substantial conclusions? YES

Is the description of the data used, the methods used, the experiments and calculations made, and the results obtained sufficiently complete and accurate to allow their reproduction by fellow scientists (traceability of results)? YES there are but are perhaps documented with obfuscating detail.

Does the title clearly and unambiguously reflect the contents of the paper? YES

Does the abstract provide a concise, complete and unambiguous summary of the work done and the results obtained? NO; see comments above

Are the title and the abstract pertinent, and easy to understand to a wide and diversified audience? YES

Are mathematical formulae, symbols, abbreviations and units correctly defined and used? If the formulae, symbols or abbreviations are numerous, are there tables or appendixes listing them? YES

Is the size, quality and readability of each figure adequate to the type and quantity of data presented? NO

Does the author give proper credit to previous and/or related work, and does he/she indicate clearly his/her own contribution? YES

Are the number and quality of the references appropriate? YES

Are the references accessible by fellow scientists? YES

Is the overall presentation well structured, clear and easy to understand by a wide and general audience? NO

Is the length of the paper adequate, too long or too short? Far too long for the result.

Is there any part of the paper (title, abstract, main text, formulae, symbols, figures and their captions, tables, list of references, appendixes) that needs to be clarified, reduced, added, combined, or eliminated? YES; see my comments above

Is the technical language precise and understandable by fellow scientists? YES

Is the English language of good quality, fluent, simple and easy to read and understand by a wide and diversified audience? YES except for a few places I have highlighted.

Is the amount and quality of supplementary material (if any) appropriate? YES, but some may in the end not be totally relevant. The authors should justify to the reader why the data are included, either in the main text or the appendix. THe reader gets the sense that the data and the analysis of high quality but there is a lot of repetition and detail that may be extraneous to the results.

Citation: https://doi.org/10.5194/egusphere-2023-2674-RC2

AC3: 'Reply on RC2', Stefan Bredemeyer, 11 Sep 2024

Dear Ramon Arrowsmith,

Thank you for the appreciation of our work and your valuable suggestions regarding further improvement of the manuscript. We agree that the readability of the manuscript can significantly be improved by removing the extraneous details which you mention from the main text and by moving some of the statistics provided in the main text to the appendix. Also, we agree that the result which we present does not necessarily cover all the currently active fault zones in South-Korea and that the semantics of recently active fault zones as defined here, therefore should be explained in more detail.

We address requested improvements as follows:

Furthermore, we removed the extraneous enumeration from the presentation of our fault length statistics previously given in brackets, i.e.:

(at average 7.81 km, standard deviation is 10.30 km, median is 4.42 km) (Coefficient of Variation: 1.3189, First quartile: 2.75 km, Third quartile: 8.85 km, Interquartile Range (IQR): 6.099 km)

We provided insight into the seemingly trivial logic expressions used in Q-GIS in order to demonstrate the simplicity of the buffering approaches utilized by most countries during the first stages of excluding unsuited areas, also to the non-expert stakeholders interested in the application of the exclusion criteria.

2) I am not comfortable ascribing even a generic probability increase of rupture, given the significant emphasis on disciplined probabilistic analysis in geological repository safety assessment. See for example, "Updated Implementation Guidelines for SSHAC Hazard Studies" NUREG-2213. Where in that structured approach, might this analysis fit, and how indeed does it indicate a probability increase for rupture?

A2): Using the term probability in this regard is misleading because we here do not refer to the probability of a future earthquake occurrence, but rather to the generally increased likelihood that a fault zone is active, if it overlies a seismogenic zone and if it has an orientation that corresponds to the current stress field of the region.

Within the framework defined by the Senior Seismic Hazard Analysis Committee (SSHAC), our work primarily contributes to the compilation of information relevant for site search and the localization of seismically active areas and active fault zones unsuited for construction of a repository. Our analysis results highlight those fault zones which have seen enhanced seismic activity in the past two decades. That is, we retrospectively shed light on the seismic hazards of the past two decades, and how these cluster in time and space, enabling to extend the findings of palaeoseismological studies which previously identified fault zones as active in the Quaternary, to the present. Such results, additional to those obtained from palaeoseismological studies provide the basis for the safety assessments that aim at determination of hazard probability in future scenarios. We intentionally did not refer to a probability of future earthquake activity in the fault zones our approach identified as recently active, because such probability assessment and the analysis of earthquake catalogue completeness was not part of the work. However, we may tentatively assign the observation of repeated locally confined earthquake activity to an enhanced earthquake probability in these locations.

3) The discussion of the comparison of the buffering is interesting. However, why is it a problem to exclude some percentage of the territory? That is the point of the exercise. And, there is still at least 70% remaining. I think this point needs to be more carefully stated.

A3): Thank you for your interest in our discussion regarding the different fault zone buffering strategies. We totally agree with you, that excluding some 10 % of the country’s area from the site screening process does not hurt at all. On the contrary, we believe that the removal of any unsuited areas should be done in a generous way regarding the width of the exclusion zones and as early as possible in the exclusion process.

4) It is not clear what the slip tendency adds. I think it is a nice addition, but it needs to be integrated with the screening and the buffer analysis somehow, otherwise it appears like an unjustified addition to the paper.

A4): By including the slip tendency analysis results we aimed at adding physical evidence for the activity of the fault zones we classified as active and the potential of future reactivation of currently inactive fault zones.

Abstract--Generally adequate abstract. However at lines 20-25, I would like to see a sentence or two more on the method details and how the results were integrated, and then compact or shorten other aspects of the abstract that are more general.

A: Ok, we added one sentence providing more detail on the slip tendency analysis.

In order to characterize the hazard potential of the Korean fault zones, we moreover conducted slip tendency analysis, here for the first time covering the fault zones of the entire Korean Peninsula. The analysis was performed for three different scenarios of fault dip (15°, 45° and 75°) using the approach presented by Röckel et al. (2022).

For our analyses we used the geo-spatial information from the current version of the 1:250k Geological map of South-Korea, containing the outlines of 11 rock units, which we simplified to distinguish between 4 different rock types (granites, metamorphic rocks, sedimentary rocks and igneous rocks) and the surface traces of 1,528 fault zones and 6,654 lineaments identified through years of field work and data processing, a rich geo-dataset which we will publish along with this manuscript. Our approach for identification of active fault zones was developed without prior knowledge of already known seismically active fault zones, and as a proof of concept the results later were compared to a map containing already identified active fault zones. The comparison revealed that our approach identified 16 of the 21 known seismically active faults and added 472 previously unknown potentially active faults. The 5 seismically active fault zones which were not identified by our approach are located in the NE- and SW-sectors of the Korean Peninsula, which haven’t seen much recent seismic activity, and thus are not sufficiently well covered by the seismic record. The strike directions of fault zones identified as active are in good agreement with the orientation of the current stress field of the peninsula and slip tendency analysis provided first insights into subsurface geometry such as the dip angles of both active and inactive fault zones. The results of our work are of major importance for the early-stage seismic hazard assessment that has to be conducted in support of the nuclear waste disposal siting in South-Korea. Moreover, the GIS-based methods for identification of active fault zones and buffering of respect areas around fault zone traces presented here, are applicable also elsewhere.

Line 40--Perhaps this is a bit pendantic, but GIS is really just a tool to do geospatial analysis. I think a bit more general tone about what it is about the study of the structures that is needed and their integration, which then requires the GIS toolkit is a bitter logic flow; rather than making it sound like the GIS is a hammer looking for a nail to pound.

A: You are right, the beginning of the introduction likely is a bit too prominent location for this section about the usage of GIS and thus we move it further back to Line 63 in order to introduce and justify the objectives of our work.

The reason for putting it there originally was to shortly introduce the status of site search in Korea and to clarify that the development of a GIS usually is part of the process. We are aware that the usage of this tool including the processes of database compilation, conduction of geospatial analyses, documentation the progress of the site search, and provision of the information to the stakeholders in principle is of little scientific interest.

Furthermore, we applied changes in the section as is described in the response to your following comment.

Line 45--"which will be" what does that mean and when? Does it mean when the paper is published or some other time?

A: The GIS database of the web-GIS is still under construction. In order to clarify that we add a “currently” to the sentence. Current versions of the datasets required for site search (KIGAM, 2019) are accessible at http://data.kigam.re.kr/gives/ . We specify that accordingly in respective sentence as follows.

Changes in Line 45: The planning of site investigations for nuclear waste disposal and the spatial assessment of the onsite geological conditions is commonly done by means of Geographic Information Systems, commonly abbreviated as GIS (e.g. Mays et al, 2012; Silva et al., 2015; Bilgilioglu 2022). Such systems enable compiling extensively large databases containing all spatial information required to assess the suitability of a site, such as e.g. performing seismic hazard analysis (Lee & Oh, 2022; Sun & Kim, 2017), and at the same time provide a means to document the progress of the site search and to communicate the results to the public (e.g. Cheon et al., 2022). For this purpose, the Korea Institute of Geoscience and Mineral resources KIGAM currently develops a web-GIS called ‘Geo-environmental Information Verification System’ (GIVES). The database of the GIVES which contains the datasets of the thematic maps required for site search is accessible at http://data.kigam.re.kr/gives/ (KIGAM, 2019).

Lines 49-53--This is awkwardly stated. The active faults are those which might experience slip within the time period of concern for the repository. There activity in the (late) Quaternary is what we would like to use to anticipate their (relative) activity.

A: Yes, you are right, we reformulate respective sentences and add more detail with respect to the time scale relevant for repository safety.

Changes in Lines 49-53: For this reason, active fault zones, i.e. fault zones which might experience slip within the time period of concern for the safety of the repository have to be distinguished from inactive or extinct faults in mechanically more stable crustal blocks (e.g. Rybicki et al., 1985; Muir Wood & Mallard 1992). It has to be noted though, that faults which appear to have been inactive for a long time do not necessarily stay inactive, since intra-continental earthquakes typically occur through reactivation of previously inactive faults, which particularly applies to faults in collision belts such as found on the Korean Peninsula (e.g. Kwon et al., 2009; de Jong et al., 2015). Such inactive faults readily can be reactivated by stress redistribution due to change in direction of the crustal movement, the injection of fluids, or by triggering through distant earthquakes or movements along connected fault systems (e.g. Sibson, 1985; Cappa & Rutqvist, 2011). Well-sealed and healed inactive fractures in contrast may develop a strength that even exceeds that of the surrounding rocks (Townend and Zoback, 2000; Børlykke 2001).

Moreover, the time period of concern to assess fault activity, the capability of a fault and the safety of a repository should be considered depending on the level of tectonic activity in a region (IAEA, 2010). More specifically, while short time intervals of 100k years reaching back to the Upper Pleistocene may be appropriate for the assessment of fault activity in high seismicity regions with short earthquake recurrence intervals such as Japan (Kim & Chang, 2020), much longer time intervals in the range of Mio of years have to be considered in tectonically less active regions covering periods spanning from the Quaternary to the present with possible extension to the Pliocene in stable cratonic regions (IAEA, 2010). The major fault zones of Korea are characterized by slow slip and very long recurrence intervals of large magnitude earthquakes (Choi et al., 2015) and the current stress regime, i.e. the neo-tectonic setting in and around the peninsula was initiated in the early Pliocene. Active fault zone studies conducted for the repository site search in South-Korea thus must consider the period ranging from Quaternary to present (NEMA, 2012; KIGAM, 2021).

Bjørlykke, Knut (2001). How faulting keeps the crust strong: Comment and Reply. Geology, February 2001, 29 (2): 189–190. DOI: 10.1130/0091-7613(2001)029<0189:HFKTCS>2.0.CO;2

Choi, Jin-Hyuck, Young-Seog Kim, Sung-Ja Choi (2015). Identification of a suspected Quaternary fault in eastern Korea: Proposal for a paleoseismic research procedure for the mapping of active faults in Korea. Journal of Asian Earth Sciences 113 (2015) 897–908. Doi: 10.1016/j.jseaes.2015.09.014

IAEA (International Atomic Energy Agency), 2010. Seismic Hazards in Site Evaluation for Nuclear Installations. Chapter 8: Potential for Fault Displacement at the Site. : Specific Safety Guide No. SSG-9, ISBN . ISBN: 978-92-0-102910-2. STI/PUB/1448.

Kee, W.-S., Kihm, Y.H., Lee, H., Cho, D.L., Kim, B.C., Song, K.-Y., Koh, H.J., Lee, S.R., Yeon, Y.-K., Hwang, S., Park, K.G., Seong, N.-H. (2009). Evaluation and database construction of quaternary faults in SE Korea. Korea Institute of Geoscience and Mineral Resources, IP2006-047-2009(1) (in Korean).

KIGAM 2021. KIGAM Annual Report. Korea Institute of Geosciences and Mineral Resources, November 11, 2021. https://www.kigam.re.kr/

Kim, Jiyeon and Chandong Chang (2020). Evolution of slip susceptibility of Quaternary faults in southeastern Korea from Quaternary to present day. Geosciences Journal Doi: 10.1007/s12303-020-0053-4

Townend, John, Mark D. Zoback (2000). How faulting keeps the crust strong. Geology; May 2000; v. 28; no. 5; p. 399–402. DOI: 10.1130/0091-7613(2000)28<399:HFKTCS>2.0.CO;2

Lines 54-56--Important indeed to recognize the potential for reactivation of structures for the reasons noted; can also cite Townsend and Zoback; weak faults in strong crust?

A: Yes, we insert the reference hinting at the increased strength a fully sealed and healed fracture may have in comparison to the surrounding rocks. See reply to comment above.

Line 60--Could mention that the reason to do the field work and have the lidar is in part to characterise the landscape and surficial geology which may indicate a relationship between the faults and those features.

A: You are right, we added that information.

Insertion in Line 60: The reason to do the field work and have the LiDAR is in part to characterize the landscape and surficial geology which may indicate a relationship between the faults and those features, while the seismological studies allow for spatially confined insights into geometry and slip mechanisms of subsurface structures.

Lines 65-67--I get the idea of the association between the seismicity and well oriented faults, but I am not comfortable ascribing even a generic probability increase of rupture, given the significant emphasis on disciplined probabilistic analysis in geological repository safety assessment.

A: You are right, using the term probability in this regard is misleading , since our approach is not a probabilistic approach and because we here do not refer to the probability of a future earthquake occurrence, but rather to the generally increased likelihood that a fault zone is active, if it overlies a seismogenic zone and at the same time has an orientation that corresponds to the current stress field of the region. We weaken the statement found in SGD (2020) accordingly.

Changes in Lines 65-67: The likelihood that a fault zone is active is quite large, if 1) the fault zone overlies a currently seismically active zone and if 2) the fault orientation corresponds to the orientation of the recent regional stress field (e.g. SGD 2020).

Line 70--I think it is important to acknowledge that the instrumental seismicity reflects a very short portion of the likely earthquake recurrence rate of an intraplate setting like Korea. There is a lot of literature on this point. Yes, you can use it because it is available and of high quality, but we know that it only gives an incomplete in time record of activity.

A: We integrated that information and adapted the rest of the paragraph accordingly.

Following this rationale, and since the historical and modern instrumental earthquake records in many regions of the world is the only available information on whether a region is seismically active or not, our approach was to use the well-constrained earthquake-epicentre locations of Korea’s most recent instrumental earthquake record to identify potentially active fault zones in South-Korea, according to their location within areas showing high earthquake-density and to validate the results by means of slip tendency analysis. Here we used 32 years of the instrumental record provided by the Korea Meteorological Administration (KMA) covering the M > 2 earthquakes which occurred during the period 1991-01-01 to 2023-05-10. We focus on M > 2 earthquakes, since such larger magnitude earthquakes typically occur at larger depths than the planned repository levels of several 100 m to 1 km (e.g. Chapman & Hooper, 2012) and have the potential to rupture a fault up to the ground surface (e.g. Figure 1 in Zielke et al., 2015 and Choi et al., 2017), i.e. to rupture through the repository. A more complete earthquake catalogue reaching back to the start of instrumental recordings still needs to be compiled for the site search in Korea and the KMA catalogue containing more refined earthquake epicentre locations currently is under construction (pers. comm. Youngchai Kim KIGAM). In this respect, it is important to acknowledge that the instrumentally recorded seismicity of any region in the world at most may cover the past 100 years, and thus generally merely reflects a very short portion of the commonly much longer earthquake recurrence intervals observed in a region. The latter is particularly true for intraplate settings as found e.g. in South Korea.

In order to find the seismically active regions in Korea, we generated heatmaps (...)

Chapman, Neil and Alan Hooper (2012). The disposal of radioactive wastes underground. Proceedings of the Geologists’ Association 123 (2012) 46–63. doi:10.1016/j.pgeola.2011.10.001

Zielke, O., Klinger, Y. and Arrowsmith, J.R. (2015). Fault slip and earthquake recurrence along strike-slip faults - Contributions of high-resolution geomorphic data. Tectonophysics, 638, 43-62. doi: 10.1016/j.tecto.2014.11.004

Choi, Jin-Hyuck, Young-Seog Kim, Yann Klinger (2017). Recent progress in studies on the characteristics of surface rupture associated with large earthquakes. Journal of the Geological Society of Korea, Geological Society of Korea, 2017, 53 (1), pp.129-157. Doi: ff10.14770/jgsk.2017.53.1.129

Lines 73-97--Good summary of the analyses to be done.

A: Thank you for acknowledging that.

Line 99--Geology no capitalization

A: Done.

Line 105--data are plural

A: Done.

Line 135--Need to start a new paragraph here "A first concept..."

A: Done.

Tables 1a and 1b are not needed in the main text. The detail distracts from a straightforward summary of the work and the reader can get a sense of what the point is by review of figure 1.

A: Good point, we moved the Tables 1a and 1b to the appendix.

Lines 165-175--This might be better as a small table.

A: Good, we present that list as a table.

Table 1. Datasets used for identification of active fault zones.

No.	Dataset	Description
1.	Fault zone surface traces (.shp)	most recent line segment data (KIGAM, 2019)
2.	Earthquake catalogue (.csv)	of instrumental earthquake-record covering M > 2 earthquakes of the period between 1991-01-01 and 2023-05-10 in and around South-Korea, obtained from the National Earthquake Comprehensive Information System (NECIS) operated by the Korea Meteorological Administration (KMA), amongst others containing information on: a. Time of occurrence b. epicentre Latitudes, Longitudes (and depths, where available), c. earthquake-magnitude scale
3.	Country outline for masking purposes (.shp)	unprojected national (level-0) vector data in decimal degrees provided by GADM (Global Administrative Areas) https://gadm.org/download_country.html (gadm40_KOR_0.shp, gadm40_PRK_0.shp)
4.	Rock unit polygons (.shp)	most recent polygons (KIGAM, 2019)

Lines 191-195--I think this is too much detail. Why is knowing that the interquartile range of the fault surface trace segmentation important? How does it aid in understanding the processes and the hazard? If you cannot clearly demonstrate the relevance, then it should be cut.

A: We removed the statistic values previously given in brackets, i.e.:

(at average 7.81 km, standard deviation is 10.30 km, median is 4.42 km) (Coefficient of Variation: 1.3189, First quartile: 2.75 km, Third quartile: 8.85 km, Interquartile Range (IQR): 6.099 km)

Paragraph at line 205--again, why is this detail on lineament lengths important?

A: The lengths of the lineaments have relevance in those cases were these correspond to fault zones that have not been identified as fault zones, yet. We notify that accordingly.

Insertion following line 205: The lengths of the lineaments have relevance in those cases were these correspond to fault zones that have not been identified as fault zones, yet.

Line 233--amongst this detail, we find what is interesting that indeed the tectonic domains become evident.

A: Ok, we reformulate the sentence accordingly.

Table 2--Ok it is useful to see the geometrically defined possible magnitude of about M6.

A: Ok, then we insert these values again.

Table 3--I don't see the importance of this table for the main text. What is the point of it? It shows the thorough analysis but...

A: You are right. We remove it from the manuscript.

Table 4--ok but there is extraneous information. Why is the perimeter important? THe percentages are useful, but not to two significant figures.

A: Ok, we remove the perimeter and we round the percentage values.

Paragraph at line 290--Heatmaps are a standard product so some of the basic description of them could be trimmed. It is really just a visualizatoin of the analysis which is the spatial kernal density.

A: Yes, we can shorten that a bit.

Paragraph at line 390--Here we get to the results. I would be very careful with the language here. We have a horizontal spatial association between fault traces and instrumental seismicity. Indeed that is a reasonable assumption that at least some of the earthquakes occurred along those faults. We don't know how the fault systems vary with depth and we know that the earthquakes are occurring in the upper crust so this 2D association is a first step. "Potentially active in recent times" is ok but recent here has different meanings. Based on the analysis, it is strictly within the time since the earthquakes occurred locally (or catalog age overall). It does not mean recent geologically unless you want explicitly state the assumption that activity in the instrumental period is consistent with longer time scale activity in the past (and the future).

A: Good then we add more detail with respect to the spatial correlation in 2D, which due to the low spatial resolution of both the earthquake epicentre locations and the fault zone traces allows for a good first order matching of the two datasets.

Table 6--I am not convinced that this is very helpful for the main text. It is useful I suppose to have a sense of which rocks have more and less potentially active faults, but that information by itself, without other context could actually hinder an assessment of optimal siting regions if it is applied blindly.

A: You are right we add more detail in this respect. The purpose of having the table was indeed to get an idea which rock types have the most potentially active fault zones and to distinguish which of the fault zones are confined to one rock type and which of the fault zones cross through different rock types. The numbers will look different, if applied higher resolved lithological maps. The intention was not to use this information as basis for decision on which rock type is suited best for a repository.

Line 455--good result that they compare well but it is not so evident in figure 10.

A: Ok then we add another line object beneath the active fault zones which shows the active fault zones our analysis identified, or compare the two maps side by side.

Section 5.2--Slip tendency analysis is a useful complement to the other elements of the research presented here.

A: Thank you for notifying that. The slip tendency analysis amongst others had the purpose to identify all those fault zones which have high slip tendency despite not having shown any activity in the time span the earthquake catalogue covers.

Line 483--Stress state reference ok but how about a quick reference to other data sets including the world stress map?

A: Yes, we can include the references on other available datasets such as the world stress map.

Line 490--I am concerned about the lack of knowledge on dip of the faults as analyzed here, given its importance in the slip tendency calculations.

A: The knowledge of the dip of the fault zones and geometries at depth is indeed still lacking for many fault zones apart for locally confined knowledge on fault dip for those fault zones which already have been identified to have been active in Quaternary times. Since this type of information has not been integrated yet into the fault zone dataset, we compared the results for the 3 scenarios of fault dip in this work.

Lines 497 and else where--"highly dipped" is not a standard nomenclature in my view; I would expect to read "steeply dipping"

A: We will apply according changes and make sure to use the standard nomenclature, i.e. amongst others replace the term “highly dipped" by "steeply dipping" as suggested.

Lines 515 and vicinity--This paragraph needs to be rewritten a bit more simply. There is some repetition and the long sentences make it hard to read. I would also write "low dip" and I think the deviatoric stress being a quantity should be referred to as "greater" or "smaller" or "lower"

A: Ok, we simplify the paragraph and check the nomenclature, also elsewhere.

Line 520--I don't think you should proceed here without summarizing briefly what we have learned with the slip tendency? It looks like maybe the steeper faults within the seismicity zones show a greater overall slip tendency. Can this be quantified? maybe a histogram of slip tendency to complement at least some of the panels in Figures 11 and 12?

A: You are right. The slip tendency generally is greater for the steeper faults.

Tables 7 and 8 and 12 and 13--The main result for the text narrative is that the polygon coverage in total is about 2.65% or 5.26% or 9.29% or 20.45% of the onshore portion of South Korea. I don't read a justification for the rest of the detail in this figure to be included in the main text.

A: Good we move the tables to the appendix then.

Table 10--Significant figures are too many. We don't need to know to the 1000th of a percent how much the rock types are covered. There is various error in the computations that probably exceeds some of the sig figs.

A: We round them to next figure.

Line 595 area--THe QGIS logical expressions are trivial and the reader would normally expect that this was how it was done. i would put a summary of these expressions in the appendix

A: We provided insight into the seemingly trivial logic expressions used in Q-GIS in order to demonstrate the simplicity of the buffering approaches utilized by most countries during the first stage of excluding unsuited areas, also to the non-expert stakeholders interested in the application of the exclusion criteria.

Line 620 and area--Here we get something interesting appearing: the discussion of the comparison of the buffering. However, why is it a problem to exclude some percentage of the territory? That is the point of the exercise. And, there is still at least 70% remaining. I think this point needs ot be more carefully stated.

A: We will change the statement that it does not sound like we claim it would hurt to exclude some 10% of the investigated area.

Appendix--This has useful information. More from the main text could be placed here to make the main text more readable and effective. I would like to see the appendix begin with a short overview as to what is contained within it.

A: We include more of the information from the main text to the appendix as detailed in previous comments and write a short overview summarizing the appendix contents.

Line 760--Strange formatting errors. Did we also lose the beginning of the references?

A: Thank you for hinting at these text fragments. We removed them. None of the references were lost due to the not intended insertion of the fragments.

The references would be much easier to read if they had a hanging indent.

A: A hanging indent does not comply with the format given in the template, but we agree with you that it would improve readability.

Figures

Figure 1--This is not a geologic map but rather a lithologic map with overlain faults. It seems to be ok for the purposes of this study.

A: We apply changes to the caption accordingly.

Line 115--I don't think it is necessary to provide the detail on the number of polygons, fault traces, and lineaments. It is distracting.

A: You are right that information is already given in the main text and we remove the information accordingly.

Line 121--Parenthetic statement about the data sources is appreciated, but is too much detail for this caption. I would have a brief reference to the appendices here and then in the appendix, provide this kind of information for the rare reader who might want to recreate the figure.

A: You are right, we remove the dataset names from the caption and merely refer to the Appendix Table A11.

A: The buffers in principle are independent from the slip tendency analysis outputs, but

Figure 3--nice looking figure.

Caption ending with detail on datasets may not be needed if that information can be presented in the appendix.

A: We remove the details on datasets from the caption and refer to Table A11 in the Appendix, were all datasets are listed.

Figure 4--ok; useful to see the seismicity along with the fault network.

A: Thank you for notifying that.

FIgure 5--ok; but the color bar does not need to be nearly as large as the map. However, given that FIgure 6 shows similar, and is used in subsequent analyses, I would cut figure 5.

A: Good we remove Figure 5 then.

A: We will apply changes to the figure as requested.

Figure 7--This is an important figure. Mask the offshore given that it is not part of the analysis.

A: Ok, we mask or tone down the offshore earthquake clusters.

Figure 8--lots of detail; could simplify for main text to highlight the main point?

A: Ok, we reduce the number of diagrams in the figure.

Figure 9--Seems like a useful figure

A: Thank you for appreciating that!

FIgure 10--Good but hard to get the comparison between the datasets

A: Ok, we insert a figure showing identified fault zones aside of that figure.

Figure 11--THis is an interesting figure. Please only use one color bar scale (given that the range is the same for all) and make it about half the current size that will clean up the figure a bit.

A: Ok.

Figure 12--THis is an interesting figure. Please only use one color bar scale (given that the range is the same for all) and make it about half the current size that will clean up the figure a bit.

A: Ok.

A: Yes the line weight obscures the buffer widths of the short fault zones.

Figure 15--This is a helpful zoom. But, it needs some lat longs and indication box in a prior figure so we know where it is.

A: Ok, we insert the lat longs and the indication box.

A: Combining into a multi-panel figure would obscure the detail again.

Best regards,

Stefan Bredemeyer

Citation: https://doi.org/10.5194/egusphere-2023-2674-AC3

Stefan Bredemeyer, Jeoung Seok Yoon, Linmao Xie, and Jeong-Hwan Lee

Viewed

Total article views: 714 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
473	182	59	714	23	22

HTML: 473
PDF: 182
XML: 59
Total: 714
BibTeX: 23
EndNote: 22

Views and downloads (calculated since 23 Feb 2024)

Month	HTML	PDF	XML	Total
Feb 2024	62	14	3	79
Mar 2024	51	21	3	75
Apr 2024	61	26	6	93
May 2024	83	33	9	125
Jun 2024	68	28	8	104
Jul 2024	21	21	4	46
Aug 2024	44	12	3	59
Sep 2024	45	10	22	77
Oct 2024	33	13	1	47
Nov 2024	5	4	0	9

Cumulative views and downloads (calculated since 23 Feb 2024)

Month	HTML	PDF	XML	Total
Feb 2024	62	14	3	79
Mar 2024	51	21	3	75
Apr 2024	61	26	6	93
May 2024	83	33	9	125
Jun 2024	68	28	8	104
Jul 2024	21	21	4	46
Aug 2024	44	12	3	59
Sep 2024	45	10	22	77
Oct 2024	33	13	1	47
Nov 2024	5	4	0	9

Viewed (geographical distribution)

Total article views: 713 (including HTML, PDF, and XML) Thereof 713 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 17 Nov 2024

Short summary

Siting of deep geological repositories for spent nuclear fuel requires large volumes of intact rock, which must not be damaged due to seismic activity in order to ensure safe containment of the toxic waste. Here we present an approach to identify large-scale fractures in Earth's surface, which may rupture in the event of future earthquakes and defined safety zones around the fractures in which a deep geological repository for storage of spent nuclear fuel should not be built.


Total:	0
HTML:	0
PDF:	0
XML:	0