ORIGINAL RESEARCH

published: 27 October 2020

doi: 10.3389/feart.2020.581605

Probability of Occurrence and

Displacement Regression of
Distributed Surface Rupturing for
Reverse Earthquakes
Fiia Nurminen 1*, Paolo Boncio 1,2, Francesco Visini 3, Bruno Pace 1,2, Alessandro Valentini 1,
Stéphane Baize 4 and Oona Scotti 4
1
DiSPUTer, University “G. d’Annunzio” of Chieti-Pescara, Chieti, Italy, 2 CRUST—Centro inteRUniversitario per l’analisi
SismoTettonica tridimensionale con applicazioni territoriali, Chieti, Italy, 3 INGV—Istituto Nazionale di Geofisica e Vulcanologia,
Sezione di Pisa, Pisa, Italy, 4 IRSN—Institut de Radioprotection et de Sûreté Nucléaire, Fontenay-aux-Roses, France

Edited by:
Chong Xu, Probabilistic fault displacement hazard analysis provides a systematic approach to
Ministry of Emergency Management,
China estimate the likelihood of occurrence and expected amount of surface displacement
Reviewed by: during an earthquake on-fault (principal fault rupturing) and off-fault (distributed rupturing).
Qi Yao, The methodology is based on four key parameters describing the probability of occurrence
China Earthquake Networks Center,
and the spatial distribution of the displacement both on and off-fault. In this work we
China
Zhikun Ren, concentrate on off-fault rupturing, and develop an original probability model for the
China Earthquake Administration, occurrence of distributed ruptures and for the expected displacement distribution
China
Gang Rao, based on the compilation and reappraisal of surface ruptures from 15 historical crustal
Zhejiang University, China earthquakes of reverse kinematics, with magnitudes ranging from Mw 4.9 to 7.9. We
*Correspondence: introduce a new ranking scheme to distinguish principal faults (rank 1) from simple
Fiia Nurminen
distributed ruptures (rank 2), primary distributed ruptures (rank 1.5), bending-moment
fi[email protected]
(rank 21) and flexural-slip (rank 22) and triggered faulting (rank 3). We then used the rank 2
Specialty section: distributed ruptures with distances from the principal fault ranging from 5 to 1,500 m. To
This article was submitted to minimize bias due to the incomplete nature of the database, we propose a “slicing”
Geohazards and Georisks,
a section of the journal approach as an alternative to the “gridding” approach. The parameters obtained from
Frontiers in Earth Science slicing are then is then combined with Monte Carlo simulations to model the dependence
Received: 09 July 2020 of the probability of occurrence and exceedance with the dimensions and position of the
Accepted: 17 September 2020
site of interest with respect to the principal fault, both along and across strike. We applied
Published: 27 October 2020
the probability model to a case-study in Finland to illustrate the applicability of the method
Citation:
Nurminen F, Boncio P, Visini F, Pace B, given the limited extend of the available dataset. We finally suggest that probabilistic fault
Valentini A, Baize S and Scotti O (2020) displacement hazard model will benefit by evaluating spatial distribution of distributed
Probability of Occurrence and
Displacement Regression of rupture in the light of spatial completeness of the input data, structural complexity and
Distributed Surface Rupturing for physics observables of the causative fault.
Reverse Earthquakes.
Front. Earth Sci. 8:581605. Keywords: probabilistic fault displacement hazard analysis, surface faulting, distributed rupturing, displacement
doi: 10.3389/feart.2020.581605 regression, reverse earthquakes

Frontiers in Earth Science | www.frontiersin.org 1 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

INTRODUCTION introducing the least interpretation possible. The events, their key
parameters used in this study, and the referenced documents used
Probabilistic fault displacement hazard analysis (PFDHA) is a are listed in Table 1.
method developed for characterizing the expected amount and The ruptures in the database were distinguished between
distribution of co-seismic fault displacement at the surface. In a principal fault rupture (PF) and DR following the definition of
PFDHA, the probability of exceeding a certain amount of Youngs et al. (2003) where the PF tend to have rather continuous
displacement at a site is represented as a function of the fault traces hosting the major displacements, whereas DR are
earthquake magnitude frequency distribution, the ensuing characterized often by shorter, discontinuous traces of smaller
probability of surface rupturing, and distribution of distances displacement in respect to the PF. However, in order to be able to
from the fault trace to the site. The key element in surface clearly identify and possibly exclude data affected by structural
rupturing analysis is the differentiation between the principal complexities in the derivation of PFDHA empirical regressions,
fault rupturing (PF), and distributed rupturing (DR). PF is the several sub-ranking categories for DR are also proposed beyond
fault plane along which the energy releases during the seismic the classical PF and DRs categories (Figure 1). Distinguishing
event; DR (also known as secondary rupturing) refers to all the between these is seen as important due to the different
other surface ruptures, often less continuous, occurring away from mechanisms behind them (Boncio et al., 2018). Consequently,
PF. The PFDHA methodology was developed for normal faulting for the generalized analysis, we are able to exclude the complex-
environments by the working group of Youngs et al. (2003) and type DR, that are likely related to the local structural setting,
developed further for strike-slip faults by Petersen et al. (2011). namely the occurrence of active folding related to the earthquake
For the case of reverse faults, Moss and Ross (2011) worked as illustrated in Figure 1. In the simplest case, rupturing along PF
specifically on PF data whereas Takao et al. (2013) worked on both (rank 1) provokes spontaneous and discontinuous rupturing
strike-slip and reverse faults but using data only from Japanese beyond this PF with a displacement significantly smaller which
earthquakes. In depth analysis of the DR data for reverse faulting is referred to as “simple DR” (rank 2). By this we mean the
earthquakes has not received much attention from the scientific rupturing occurred as a direct response to the surface rupturing
community. The present work is a first step toward filling this gap. along the PF in unpredictable locations. In favorable conditions,
In the following, this article describes the construction of the surface rupturing may take place also on pre-existing faults
database of surface ruptures for reverse faulting earthquakes, the connected to the PF at depth along so called “primary DR”
methodology used for obtaining the empirical parameters, and (rank 1.5), or along a disconnected, distant triggered fault (rank
the mathematical approach used to model the occurrence and 3). Primary DRs may actually host significant displacement, and
displacement distribution of DR based on the database. In the their surface traces are often relatively continuous. Triggered
final part, the new regression parameters are applied to model the surface faulting takes place typically along pre-existing fault at a
distributed surface rupturing hazard along the Suasselkä post- remarkable distance from the PF, but the surface fault is usually
glacial fault (PGF) situated in Lapland, Northern Finland. It is a rather discontinuous. Furthermore, reverse faulting may lead to
several tens of kilometres long reverse fault along which several bending-moment (rank 21) and flexural-slip (rank 22) DR as
geological and geophysical research projects are ongoing. This defined by Philip and Megharoui (1983) and Yeats (1986); both of
fault was chosen as an example case because the Suasselkä PGF these being highly influenced by the specific geological structures
may be considered still active today (Ojala et al., 2019a) and thus a favoring this type of faulting (especially active folding in relation
potential source of a surface rupturing earthquake. to earthquake faults). The database contents by earthquake and
fault ranking are listed in Supplementary Table S2.
The dataset is based on empirical field work, and the locations
CONSTRUCTING THE DATABASE AND of the earthquakes and their surface deformation zones vary from
RANKING THE OBSERVATIONS very urbanized to hardly reachable mountain areas. During field
survey after a surface-rupturing earthquake, geological
The work initiated by developing a database of surface faulting investigations mainly focus on tracing and measuring PFs.
data of well documented reverse faulting earthquakes. The Moreover, surface extension and spatial density of the DR are
displacement parameters and the spatial distribution of surface not always uniform along the ruptures of the corresponding PF
faulting were analyzed from georeferenced maps and published and there is a large variability among the 15 earthquakes. We can
datasets. Following the SURE database architecture (Baize et al., attribute the variability in the database to a) the spatial
2019), we built a database of surface rupturing displacement completeness of data available, or b) physical reasons. For
reported in publications for 15 historical medium-large reverse example, the spatial completeness can be affected by some
earthquakes of moment magnitude from Mw 4.9 to Mw 7.9, areas not being accessible for field surveying, or the surface
occurred between 1970 and 2019. The selected events were all extension of the PF was too large to map all the DRs in detail.
studied with an acceptable accuracy on site, and there are good Also, the length of the DRs is a subject of mapping accuracy and
quality displacement data and mapped fault traces available. All interpretation. For example, in areas like San Fernando (urban
the displacement measurements and fault orientation parameters area), many DRs are measured along the streets, thus rupture
reported by the first-hand authors were entered in the database as length ∼ street width, and single traces are drawn discontinuous
published. Although displacement data available is not uniform, under buildings and similar obstacles. Physical elements that may
all the published information was entered as it was published have caused bias in the data coverage are due to DR formation

Frontiers in Earth Science | www.frontiersin.org 2 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

TABLE 1 | Reverse earthquakes included in the database.

Earthquake Date Moment Kina SRLb MDvc MDnd Types References

(yyyymmdd) = EventID magnitude (km) (m) (m) of DRe
(Mw)

Calingiri, Australia 19700310 5.0 RL-L 3.3 0.4 1.2 2 1, 2

San Fernando, CA, United States 19710209 6.6 R-LL 16 0.76 2.5 2, 1.5, 21, 3 1, 3, 4
El Asnam, Algeria 19801010 7.1 R 31 5.0 6.5 2, 21, 22, 3 1, 5–7
Coalinga (Nunez), CA, United 19830611 5.4 R 3.3 0.50 1.0 2 1, 8
States
Marryat Creek, Australia 19860330 5.8 R-LL 13 0.9 1.1 2 1, 9–11
Tennant Creek, Australia 19880122 (3 events) 6.3 R 10.2 1.20 2.84 2 1, 12, 13
6.4 R-LL 6.7 1.10 2.60 2, 3
6.6 R 16 1.77 2.50 2
Spitak, Armenia 19881207 6.8 R-RL 25 1.6 2 2 1, 14, 15
Killari, India 19930929 6.2 R 5.5 0.6 1.2 2 16–18
Chi Chi, Taiwan 19990920 7.6 R-LL 72 9.8 11.65 2, 1.5, 3 19–32f
Kashmir, Pakistan 20051008 7.6 R 70 3.40 7.05 2, 21 33–35
Wenchuan, China 20080512 7.9 R-RL 312 6.9 13.0 2, 1.5, 3 36–48
Pukatja, Australia 20120323 5.4 R 1.6 0.5 1.0 2 49
Nagano, Japan 20141122 6.2 R 9.34 0.8 1.60 2 50–53
Petermann, Australia 20160520 6.1 R-LL 20 0.9 1.9 2 54
Le Teil, France 20191111 4.9 R 5 0.23 0.33 2 55
a
Kin. (kinematics): R, reverse; LL, left-lateral; RL, right-lateral.
b
SRL, surface rupture length (principal faulting).
c
MDv, maximum vertical displacement.
d
MDn, maximum net displacement.
e
Types of distributed rupturing: 2, simple; 1.5, primary distributed; 21, bending-moment; 22, flexural-slip; 3, sympathetic.
f
For Chi Chi, we used the data from Lin (2000), Chen et al. (2000), Huang et al. (2008), Kelson et al. (2001), Huang et al. (2008), Angelier et al. (2003), Kelson et al. (2003), Ota et al. (2007),
and Lee et al. (2003): Table 1, the points shown in maps Figures 4, 5. The points considered cover only the central-northern part of the rupture, but we do not expect a major bias in results
of this work as, according to the reference maps, most of the distributed rupturing took place in northern section.
References: 1, Wells and Coppersmith, 1994; 2, Gordon and Lewis, 1980; 3, U.S. Geological Survey Staff, 1971; 4, Kamb et al., 1971; 5, Yelding et al., 1981; 6, Philip and Meghraoui,
1983; 7, Meghraoui et al., 1988; 8, Rymer et al., 1983; 9, Fredrich et al., 1988; 10, Bowman and Barlow, 1991; 11, Machette et al., 1993; 12, McCaffrey, 1989; 13, Crone et al., 1992; 14,
Haessler et al., 1992; 15, Philip et al., 1992; 16, Lettis et al., 1997; 17, Seeber et al., 1996; 18, Rajendran et al., 1996; 19, Wesnousky, 2008; 20, Kelson et al., 2001; 21, Kelson et al., 2003;
22, Angelier et al., 2003; 23, Bilham and Yu, 2000; 24, Chen et al., 2000; 25, Huang et al., 2008; 26, Huang et al., 2000; 27, Lee et al., 2001; 28, Lee et al., 2003; 29, Lee et al., 2010; 30,
Lin, 2000; 31, Ota et al., 2007; 32, Central Geological Survey, MOEA, 2017; 33, Avouac et al., 2006; 34, Kaneda et al., 2008; 35, Kumahara and Nakata, 2007; 36, Xu et al., 2009; 37,
Liu-Zeng et al., 2009; 38, Liu-Zeng et al., 2012; 39, Yu et al., 2009; 40, Yu et al., 2010; 41, Zhou et al., 2010; 42, Zhang et al., 2013; 43, Chen et al., 2008; 44, Liu-Zeng et al., 2010; 45,
Wang et al., 2010; 46, Xu et al., 2008; 47, Zhang et al., 2012; 48, Zhang et al., 2010; 49, King et al., 2019; 50, Okada et al., 2015; 51, Ishimura et al., 2015; 52, Lin et al., 2015; 53, Ishimura
et al., 2019; 54, King et al., 2018; 55, Ritz et al., 2020.

being favored in correspondence to geometrical complexities segments per earthquake when considering only the simple DR,
connected to the PF or reflect the co-seismic patches at depth. in the hanging wall and footwall ranges, respectively, from 1 to
For addressing the issue of likely incompleteness of the DR 270 and from 1 to 41, the mean being 52 for the hanging wall and
mapping, we developed an approach that generalizes the 9 for the footwall, and the 16th to 84th percentiles are 7 and 109
probability of DR occurrence along the strike of PF by making for hanging wall and 1 and 15 for the footwall (Figures 2C,D).
no assumption of the completeness of the database; we call this The distributions in length and number of DRs and their spatial
approach “slicing” and it is explained more in detail in the distribution are based on the maps published by the field working
following chapter. Later, we balance this probability with the groups handled with the lowest level of interpretation possible.
ratio of the total length of DR with respect to the PF length for The shapefiles of the surface ruptures and displacement data
each earthquake. including the fault ranking of the 15 reverse earthquakes used in
Figure 2 shows the length distribution of the DR segments, this work are included in the Supplementary Table S1 and will be
and the number of DR segments per earthquake for hanging wall part of the next SURE database release. Data and results from this
and footwall separately. When considering “all types” DR, the work contribute also to the Fault Displacement Hazard Initiative
distribution of the rupture lengths in the hanging wall and (FDHI) working group (Sarmiento et al., 2019; https://www.
footwall ranges from 1 to 3,665 m and 1 to 2,155 m, risksciences.ucla.edu/nhr3/fdhi/home).
respectively, (Figures 2A,B); the mean length on hanging wall
is 82.2 and 184.1 m on the footwall, and the 16th to 84th
percentiles are 11.6–130.3 m on hanging wall, and METHODOLOGY FOR PROBABILISTIC
12.4–310.3 m on the footwall. Peaks in the simple DR length FAULT DISPLACEMENT HAZARD
distributions are in the range of 2–25 m, the mean being 66.5 m
for the hanging wall and 106.3 m for the footwall, and the 16th to
ANALYSIS
84th percentiles are 6.7 and 73.7 m for hanging wall, and 8.8 and PFDHA begins similarly to the traditional probabilistic seismic
181.7 m for the footwall. The distribution of the number of DR hazard assessment with seismic source identification, and

Frontiers in Earth Science | www.frontiersin.org 3 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 1 | Schematic illustration of fault ranking for reverse faults. Principal fault rupture (1) is the surface expression of the fault responsible for the earthquake, the
other fault types being various kind of off-fault rupturing. Primary distributed rupturing (1.5) refers to the pre-existing faults that are connected to the principal fault in
depth. These, however, rupture only together with the PF. Simple distributed rupturing (2) is the most general case of off-faulting, referring to the surface rupturing on
unpredictable locations (not pre-existing faulting, or hidden small pre-existing faults). Bending-moment (21) and flexural-slip (22) rupturing are both responses to
large scale folding. Sympathetic rupturing (3) occurs along a pre-existing fault that is triggered usually for rather discontinuous rupturing. Complex DR inspired by (a)
Tsauton back-thrust and (b) Tsauton frontal synthetic splay of Chi Chi 1999 rupture (Ota et al., 2007); (c) central zone (normal faults at extrados of folds in the hanging wall
of the main thrust), (d) northern zone (bedding plane slips in the sub-vertical limb of a footwall syncline), and (e) distant ruptures east of central zone of El Asnam 1980
rupture (Philip and Meghraoui, 1983). See Supplementary Figure S1 for examples of rupture ranking.

characterization of the earthquake frequency and distance the distance from the PF. Previous works (e.g., Moss and Ross,
distributions, but instead of ground motion distribution, 2011) have analyzed the PF faulting parameters for reverse
continues with calculating the conditional probability of faulting earthquakes, whereas in this work the emphasis is on
exceeding a displacement d (Youngs et al., 2003). Thus, for obtaining the probability distribution and expected displacement
obtaining this conditional probability of exceeding a certain of distributed rupturing. The methodology used for obtaining
displacement level at a given distance from the PF, we need these parameters is described in the following.
empirical attenuation relationships from the historical surface In the approaches based on gridding (Youngs et al., 2003;
rupturing earthquakes describing the probability of DR Petersen et al., 2011), the probability of DR occurrence, also
occurrence and the displacement distribution as a function of referred to as off-fault displacement to distinguish it from on-

Frontiers in Earth Science | www.frontiersin.org 4 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 2 | Number of distributed ruptures and associated DR lengths on hanging wall (A) and footwall (B) sides. For the sake of readability the DR lengths shown
in histogram are cut in first 500 m of length, the maximum DR length being 3,665 m on hanging wall, and on footwall 1,480 m when considering only simple type DR, and
2,156 m when considering all types of DR. The graphs below show the number of DRs per earthquake on hanging wall (C) and footwall (D). All types of DR plotted on
red, simple types of DR (only ranking 2) being a subset of “all” plotted on black.

fault rupturing for PF, is calculated for a given cell area located off (1) α(m) is the annual rate of occurrence of an earthquake with
the PF. The probability of a DR to pass through the site of interest magnitude m.
is a function of the distance from the PF trace and of the surface (2) P [D > 0 | m] is the probability of having surface rupture on the PF,
area occupied by the site of interest. The grid approach to evaluate given that an earthquake with magnitude m occurred on the fault.
the probability of having a nonzero displacement within the area (3) P [d > 0 | r, G] represents the probabilities of having a
of interest is based on the rate of the observed number of cells nonzero displacement at a distance r from the PF, as a
with at least one rupture over the entire number of cells at the function of a vector (G) that include magnitude and that
same distance from the PF. Logistic regression model was used by can additionally include the dimension of the site of interest
Youngs et al. (2003) to compute the conditional probability of and the expected total length of DRs.
distributed rupturing occurring at a point, using a functional
forma that incorporates the observed trends of decreasing In this paper we introduce the “slicing” method, more of which in the
frequency with increasing distance, increased density with following chapter, in which the term P [d > 0 | r, G] is divided in two
increasing magnitude and lower frequency and faster decrease terms: P [d > 0 | r, m] and P [d > 0 | r, dim, F]. The former represents the
in frequency with increasing distance on the footwall than on the probability of having a DR occurrence in a slice, being a function of the
hanging wall side. According to the PFDHA methodology distance from the PF, with no regards of the dimensions of the site of
(Youngs et al., 2003; Petersen et al., 2011), the rate (λ) at interest. The latter expresses the conditional probability of having a DR
which the displacement (d) on a site at a distance >0 from a within the site, given the occurrence of DR at the given distance. This
PF exceeds a specified level (x, with x > 0) is conducted by solving term is a function of the dimensions of the site (dim), and of the fraction
the equation: (F) of total DR length in respect to the length of the PF at the distance r,
and
λ(d ≥ x)  α(m) P[D > 0 | m]P[d > 0 | r, G]P[d >
 x | s, m, DN ], (1) (4) P [d >  x | s, m, DN] is the probability that the DR
displacement exceeds a given value (x), given r, m and DN.,
where where DN is the displacement expected at a particular position

Frontiers in Earth Science | www.frontiersin.org 5 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 3 | Principle of “slicing” in comparison to gridding. From the map view the occurrence of the distributed ruptures (blue lines) is analyzed in a cell (gridding) or
within a slice (slicing) parallel to the principal fault (red line). The approach used for obtaining the probability of DR occurrence in slicing is independent of the fault-normal
division into grids, and the completeness of the rupture tracing.

along the strike of the PF (to account for the variability in The frequency–distance distributions of DRs is computed as
rupture location and displacement on the PF) and s represents the sum of slices intercepting at least a partial DR segment,
the closest distance to that position from the DR. normalized to the total number of the events. When the data of all
earthquakes in the database are brought together, for each slice of
For the purpose of this work, we assume that an earthquake distance r we can have a value ranging from 0 (none of the
with magnitude m produces surface faulting on the PF, and we earthquakes has a rupture within the slice) to the total number of
focus on the third and fourth terms of Eq. 1 which require the events (all the earthquakes have at least a part of a DR segment
analyzing the spatial distribution of DR, deriving the most intercepting the slice). This count is divided by the number of
suitable measure of the displacement components along DR earthquakes to obtain the frequency. In Figure 5 we grouped the
from the data collected in the database, estimating the earthquakes according to the magnitude: 1) Mw 5 ≤ M < Mw 6
probability of DR and finally estimating the probability of (including Mw 4.9 Le Teil earthquake), 2) Mw 6 ≤ M < Mw 7, and
exceedance of a target level of displacement. 3) M ≥ Mw 7. Notice however, that the number of earthquakes per
class of magnitude is small: 5, 6, and 4, respectively. The list of
Measurements From the Database: measurements r, their length, and the ranking of the DR as well as
Distance Parameter r the ranking of the corresponding primary trace (1 or 1.5) are
We introduce here a new method called “slicing” for the analysis of given in Supplementary Table S3.
the spatial distribution of DR. In the “slicing” approach, the area off Unlike the previous approaches utilizing gridding (Youngs
the PF is divided in 10 m wide slices parallel to the PF strike. Slices et al., 2003; Petersen et al., 2011), the “slicing” method does
containing at least a part of a DR segment are counted considering not contain an assumption on the completeness of the database
the distance from the PF (Figure 3). In practice, we first calculate along the PF strike. We implicitly accept the very likely situation
the distance from DRs to the PFs by resampling the DRs and PF that not all the area is studied with the same precision in the field,
traces, filling in any gaps in coordinate data vectors greater than a as some parts can be hard if not impossible to reach. On the other
defined 10 m tolerance apart in either dimension. We then calculate hand, it is also likely that distributed rupturing does not occur
the minimum DRs to PF point-to-point distance. This point-to- homogenously along the PF strike due to the physical factors, such
point distance we name “r” (Figure 4A). According to the ranking as subterranean structures and material distribution, and the
introduced in Figure 1, we calculate the distance r for a DR of mechanics of the process. This will be accounted for when we
introduce in the analysis an estimate of the relative coverage of the
(1) rank 2 with respect to rupture traces of rank 1 or 1.5, DR mapping with respect to the PF length, as will be shown later.
whichever the closest;
(2) rank 1.5 with respect to rupture traces of rank 1; and Measurements From the Database: Vertical
(3) rank 21, 22, or 3 with respect to rupture traces of rank 1. Displacement and the Distance Parameter s
In situ measurements are performed and reported in different
The width and interval of the slice, as well as the resampling ways from one study to another. The field conditions
tolerance are chosen, so as to reflect the resolution of the original (visibility, accessibility, presence of displaced, and matching
maps used and still provide sufficient precision for the modeling. features) impact on the possibility to measure the

Frontiers in Earth Science | www.frontiersin.org 6 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 4 | The measurements obtained from the georeferenced maps and displacement data of each earthquake in the dataset distinguishing the principal fault
(PF; red line) and distributed ruptures (DR; thick blue line). (A) The definition of the distances “r” and “s,” and (B) ps is the nearest point of the PF to the vd. The DN
at ps is calculated considering VD1, VD2 and the corresponding distances. See text for details.

displacement, and the available slip components might not be the fault dip angle if not all the vectors are known. Fault
uniform. In some, but not all cases, it is possible to derive all attitude (strike, slip) is usually well documented along the PF,
the slip components from available data. However, this is not but the DR can have different orientations with respect to the
always possible, and therefore some blanks remain in the PF, and the assumptions of dip angles cannot be justified
database. For example, the net displacement (ND), which especially if not reported for the DR trace in consideration.
represents the best the total deformation caused by the Thus, obtaining rigorous ND based on reproducible
earthquake, can be calculated as a vector sum of the vertical mathematical methods is not equally correct when it comes
displacement (VD) and the two horizontal slip components to DR, as the DR slip parameters are rarely reported with a
(fault-parallel, and fault-normal slip), or obtained by utilizing high level of detail.

FIGURE 5 | Frequency of distributed rupturing as a function of the distance r from the PF for three classes of magnitude considering (A) all types of DR, (B) only
simple DR (ranking 2), or (C) only “Primary” distributed rupturing (ranking 1.5) on footwall (FW) and hanging wall (HW) sides.

Frontiers in Earth Science | www.frontiersin.org 7 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

Given that in the database the vertical slip component, VD for

PF, and vd for DR, is the measure most often available for the
surface ruptures of reverse faulting earthquakes, we decided to
use this measure off displacement as a reference metric for the
following analysis. This seems a reasonable choice as it is the most
representative component of dip slip ruptures. Some distributed
rupturing of the earthquakes considered in this work acted like
normal faults near the surface, vertical slip measured in the field
being opposite to the PF, but was considered here in absolute
values. The database contains 234 measures of vd, most of them
(53%) belonging to rank 2 DR, and 30% to 1.5 rank DR. The
further analysis is limited to the “simple” DR (rank 2), to avoid
biases due to folding-related or triggered DRs. Notice that the
vertical slip values of the “primary” DR (rank 1.5) will be used in
some cases later on for vd in the case where a rank 2 DR is closer
to a rank 1.5 DR than to the PF.
For the analysis of the spatial distribution of the vertical slip
measured along the distributed surface ruptures, we utilize the
nearest distance from the point of measurement to the
corresponding primary trace, calling this distance s
FIGURE 6 | Analysis of the spatial distribution of the vertical
(Figure 4A). The difference of s in respect to r is that s is displacement, vd, observations of simple DR (ranking 2) by earthquake in
computed exactly and exclusively for the point on DR with an respect to the distance from the principal fault on hanging wall (HW) and on
associated measure of vertical throw, whereas r is calculated by footwall (FW) sides. Dashed lines on FW and HW, in 200 and 350 m
resampling the coordinates of the DR segments. For a sufficiently distance respectively, indicate the zone in which the database contains the
largest variability of the values and can be assumed to be the most complete
dense resampling procedure, s is a sub-sample of r. respect to the total DR occurred in reality.
The displacement measured or expected at a particular
position along the strike of the PF (DN) could be obtained
from slip profiling, but composing a profile is not a trivial
procedure when faced with discontinuous fault traces and Probability of Distributed Rupturing
other geometrical complexities, such as overlapping branches In this section we analyze the probability of DR first in a general
and splays. Therefore, having the database of displacement points situation, and then at a site with specific size in a defined distance
on both DR and PF, we propose an alternative method in which r from the PF.
for each DR displacement point the parameter DN is derived from
the two nearest measured displacement points on PF (Figure 4B). Probability of a Distributed Rupturing Occurrence in a
DN is defined by the two nearest vertical displacement values VD Slice
along the PF, inversely weighted by their distance to the point of For the analysis of DR probability parameters (third term in Eq.
vd. DN is calculated for each point of vertical displacement along a 1), the distinction between PF and DR must be clearly defined.
DR as following: First of all, we establish a threshold distance within which all the
DN  (VD1 x2 + VD2 x1 )
(x1 + x2 ), (2) ruptures are considered to be part of the PF. In this work the
chosen distance is 5 m on both hanging wall and footwall sides,
where VD1 and VD2 are the nearest vertical displacement values defining a 10 m wide PF zone. Utilizing this type of buffer zone
along the corresponding primary trace (rank 1 or 1.5), and x1 and seems reasonable also in respect to the observations in the field,
x2 are the corresponding distances from the point of vd to the where the surface material may not result in one well-defined,
points of VD. In some cases, the nearest point used to calculate fully continuous principal fault trace. Indeed reverse faulting in
the s (ps in Figure 4B) is attributed to a segment different from the areas with loose cover material induce often a broad flexural
points of displacement along the PF used for calculating the DN. scarp, rather than a clear-cut scarp (e.g., Kelson et al., 2001; Yu
In Supplementary Table S4, we list the coordinates of the VD1, et al., 2010). All the little and discontinuous ruptures within a
VD2 and the end points of segments x1, x2, and s. distance inferior to 5 m to the PF were excluded from the further
In Figure 6, we show the vd of rank 2 DR with respect to their analysis of DR probabilities.
distance from the rank 1 and rank 1.5 ruptures. Figure 6 allows The “slicing” approach allows us to pose a simple question for
discussion of the spatial completeness of the vd with respect to the each earthquake: “What is the probability to have at least one
distance from the PF, in fact, for distances longer than few distributed surface rupture within a slice?” We then compute the
hundreds of meters both on the hanging wall and footwall, occurrence of DRs in a slice for all the 15 earthquakes in the
only large vd are mapped. A likely explanation is the difficulty database, which means we are assuming the occurrence being
of distinguishing DR of small dimension at large distances from independent from the position of the site along the PF strike.
the PF, and this must be kept in mind when using the Subsequently, we apply a multinomial logistic model, which
empirical data. allows the analysis of various uncategorized variables, and that

Frontiers in Earth Science | www.frontiersin.org 8 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

TABLE 2 | Coefficients for Eqs 4 and 6. at least a partial rank 2 DR and 0 where no rank 2 DR have been
Coefficient Footwall Hanging wall mapped. The global predictor matrix has 7.000 inputs, given by
15 earthquakes and 500 slices, which are centralized at distances
Equation 4 from 5 to 5,005 m from the PF. In our case, Eq. 3 is
a 8.5431 2.9179
b1 −1.5586 −0.5566 ln Pf
1 − Pf   a + b1 X1 + b2 X2 , (4)
b2 0.0099 0.0030
Equation 6 where Pf is the probability of an outcome being in category “at
a −5.1043 −4.2549
least a partial rank 2 DR” with respect to the reference category of
b1 −0.6483 −0.1514
c1 0.1983 0.4404
“no rank 2 DR,” and X1 and X2 are the earthquake magnitude and
d1 0.9461 0.5711 distance from the PF, respectively. The coefficients for Eq. 4 are
std 0.8812 0.9129 given in Table 2.
Figure 7 shows the predicted probabilities for the multinomial
logistic regression model with predictors, X1 and X2, and the
computes the relative risk of being in either the presence or coefficient estimates, b1 and b2, for hanging wall and footwall. The
absence of a DR from any given fault using a linear combination probability curves are shown for three magnitude Mw 5.5, Mw 6.5,
of predictor variables. Consequently, the probability of each and Mw 7.5. For a visual comparison, in Figure 6 we also show the
outcome is expressed as a nonlinear function of n predictor frequency–distance distributions for the three classes of
variables: magnitude: Mw 5 ≤ M < Mw 6, Mw 6 ≤ M < Mw 7, and M ≥
ln Pj
1 − Pj  a + b1 X1 + b2 X2 + bn Xn , (3) Mw 7. The number of the earthquakes is relatively small in each
class of magnitude, but we can see clear differences between the
where Pj is the probability of an outcome being in category j, and classes due to which the division is considered reasonable anyway.
n is the number of predictor variables. We focus on simple
distributed rupturing, ranking category 2. The data we Probability of Distributed Rupturing Occurrence at the
collected are organized into a two column matrix with the Site
magnitude of the causative events and the classes of distance Distributed rupturing does not occur homogenously and
from the PF. The response is a vector of 1/0, with 1 where we have continuously along the surface trace of the principal fault. To

FIGURE 7 | Logistic regression of the distributed rupturing in the database indicating the probability of observing at least a partial distributed rupture (DR) as a
function of distance from a principal fault (PF) on hanging wall (HW) and footwall (FW) sides. The squares show the response to the logistic regression of rupturing either
occurring (1) or not occurring (0) at a given distance. The points show the DR frequency in respect to the total number of events int each Mw class, and the lines indicate
the probability of at least a partial DR at a given distance.

Frontiers in Earth Science | www.frontiersin.org 9 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

expected length of the total DR will cover a fraction F of 0.03 of the

length of the PF on hanging wall, and F of 0.007 on footwall. For a
larger distance than 100 m, we assume F of 0.007 on hanging wall,
and F of 0.004 on footwall. The range of applicability of these values
in terms of distance from the PF is based on the reach of the
database, being 500 m on footwall, and 1,500 m on hanging wall.
Given the total length of DRs and the range of individual length,
we can derive the range of the expected number of DR. For example,
for a 30 km long PF, at a 25 m-wide site located at a distance greater
than 100 m from the PF, we expect the total length of the DR being
210 m (30 km × 0.007). Assuming that the individual length of a DR
ranges between the 16th and 84th percentiles of the observations,
respectively of 7 and 74 m, we calculated the number of expected
DR segments to be in the range of 3 (210/74) to 30 (210/7).
We then run 104 simulations, and for each simulation we

(a) extract the number (N) of ruptures between the minimum-

maximum number, by assuming a uniform distribution,
(b) deduce the length of each DR by dividing the total length of
FIGURE 8 | Fraction of DRs lengths to PF lengths, and power fit as a
function of distance from the PF: thicker fit to maximum value; thinner fit to “all”
DR by N,
values, and the fixed values of coefficient F and their applicability range. The (c) extract the N positions of the center of the surface ruptures
events in the database are grouped according to their magnitude as from the i sampled centers,
previously; Mw 5–6 plotted in red, Mw 6–7 in green, and Mw 7–8 in blue. The (d) build the surface ruptures, by distributing the length of
dashed lines show the validity ranges in which the various coefficients F listed individual DR symmetrically in respect to their center, and
in the table are valid; the subscript of the coefficient F indicates the side on
hanging wall (HW) or on footwall (FW) side, and order of the coefficient
(e) verify if at least a rupture is located inside the site.
counted from the PF (1 or 2). Only simple type of DR considered (rank 2 DR).
Figure 9 illustrates an example of this Monte Carlo approach. In
particular, Figure 9 is a zoom of a 30 km fault for a site with a
dimension parallel to the PF of 25 m, located at 500–525 m distance
estimate the probability of surface rupture occurrence at a site perpendicular to the PF. Note that the y-axis in Figure 9 serves to
located at a distance from the PF we adopted a Monte Carlo stack the simulations (only the first 100 of the total of 104 are shown).
approach. We discretized a space of length equal to the length of The PF is shown as a red line on the bottom of the figure, whereas
the PF in i sections of an arbitrary length of 1 m. Each section DRs are represented by the black lines parallel to the PF. To simplify
represents the possible center of a DR at a given distance off the the reading of the figure, the width of the site is represented by the
PF. Next, we propose to calculate the number of secondary DR of blue lines, projected along the y-axis. The number of favourable cases
given length considering the statistics of the database shown in (at least a partial rupture) is obtained by counting the number of at
Figure 2. Unfortunately, concerning the DR length statistics, the least one DR intersecting the site, i.e., how many times there is at least
database does not allow to infer any relation between number and a partial black brick between the blue dotted lines. The total cases are
dimension of the surface ruptures relative to the parameters of the the number of simulations.
PF, such as magnitude or fault length. The conditional probability of DR occurrence within the site
Concerning the total length of DRs expected for a given PF of interest P [d > 0 | r, dim, F], that is a function of the distance r
length, we use the ratio observed in Figure 8. For each slice, for each from the PF, of the dimension (dim) of the site, and of the length
earthquake, we computed the sum of the lengths of the DR traces, of the DR (that is the fraction F of the PF length), is then
and normalized this sum to the length of the PF; this measure being
P  number of favourable cases
total cases (5)
the ratio F. The gray curves in Figure 8 represent the fit of the
maximum ratio at each distance (thick curve above) and the fit of Because of the sampling interval of the distance from the DRs
the entire dataset (thin curve below). The rapid decay observed at a to the PF (10 m), we assume that the probability P represents the
distance of ∼100 m from the PF both on hanging wall and footwall probability for even a partial DR occurrence at a site with a
sides could be related to some physical reasons, but the bias in the dimension perpendicular to the PF of 10 m. This accuracy seems
data available does not allow to be confident in any particular adequate to most engineering solutions the modeling may serve.
extrapolation, and it could be, possibly, still related to the The code is available in the Supplementary Material of this paper.
completeness of the database. For this reason, we consider a
threshold distance of ∼100 m as a (soft) border between high to
low ratio of DR in respect to the length of PF. Instead of considering Probability of Exceedance of Displacement
either of the fit directly, we simplify the analysis by using two values Levels
of F at two ranges of distance from the PF based vaguely on the The probability of exceedance of a level of displacement is
maximum ratio. For a distance <100 m, we can assume that the obtained by analyzing the correlation between the amount of

Frontiers in Earth Science | www.frontiersin.org 10 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 9 | A site of 25 m width situated on hanging wall in 100–1,500 m distance from a 30 km long principal fault (PF) and the first 100 simulations of the
existence of distributed surface ruptures (DR) at the site. Each line on simulation plot represents a simulation of positioning DRs along the PF strike, and the vertical dotted
blue lines indicate the site of 25 m width in respect to the total length of the PF (red line).

displacement on DR vs. the distance from the PF, magnitude of Figures 10A,B show the natural logarithm of vd in respect to
the earthquakes and displacement on the PF. For this part, we use the natural logarithm of distance s from the PF, the natural
the distance parameter s, that indicates the distance from a point logarithm of DN and magnitude, respectively for hanging wall and
with a measured vertical surface slip along a DR to the nearest PF footwall. We can observe a weak correlation of vd with the
trace (Figure 4B). This distance is connected to the data point on distance both for hanging wall and footwall, and a linear
DR in consideration. correlation with DN and magnitude. A possible reason for the
Previous authors (Youngs et al., 2003; Petersen et al., 2011) weak correlation of vd with the distance from the PF is the short
chose the single value option to account for the magnitude of the distances we are analyzing, and we do not expect a rapid decay in
event. Normalization to the maximum displacement (MD) is a the first hundreds of meters. On the other hand, it is worth noting
commonly used approach, and it is easily available value for every that we are correlating vd with an epicentral distance from the
event. However, it does not consider the variability of the slip surface trace of the PF, and that we are ignoring possible
along the PF trace, which can be remarkable in large events with important parameters as, for example, the dip of the fault, the
long surface rupture length (SRL) breaching varying surface distribution of the co-seismic slip at depth, the structural
conditions. An alternative option, adopted in this work, is to complexities of the principal faults (e.g., bends, gaps). All the
estimate the distributed displacement values by the amount of slip distances in the database were measured from the georeferenced
at the corresponding location along the PF, i.e., at (or very near) maps, in a 2D map view. Though this ignores the changes in the
the other end of the s measurement line. This choice lets us topography and the fault dip, potentially biasing the actual
correlate the off-fault displacement to the estimated local slip distances between the fault planes in steep slope regions, we
along the PF and associated slip patches in depth. do not expect excessive mistakes in fault displacement hazard
For the systematic analysis of the database, only the equal type analyses when performed equally on a map basis. The statistical
of data could be confronted, thus the DN should be derived from processing of datasets is expected to smooth the potential errors
the similar slip parameters as the slip on DR considered. This by covering a wide range of geometrical arrangements between
approach considers only existing data; for the datapoints on DR topography and fault dip.
with reported vertical slip parameter the two nearest points along The functional form of the equation used for the regression is
PF with reported vertical displacement (VD) are considered for
ln(Y)  a + b1 (ln(s)) + c1 (ln(DN )) + d1 (m) , (6)
calculating the parameter of expected displacement (not
necessarily any two nearest points, if no VD available). This where Y is the median expectation of vd, a is the constant term, s,
procedure allows us to account for the variability in rupture DN, and m are, respectively, the epicentral distance, the
location and displacement along the PF. normalization factor derived from the vertical displacement of
To evaluate the expected mean value of vd for rank 2 DR, we the PF, and moment magnitude. The values of the coefficients a,
performed a regression analysis on vd and DN, magnitude and b1, c1 and d1, for hanging wall and footwall, and the residual
distance s from the PF. Basing on the analysis of the spatial standard deviations, are presented in Table 2. Supplementary
completeness of our database (Figure 6), we limited our analysis Figures S2A,B show the residual plots for the regression of Eq. 6
to a subset of data consisting of 48 vd mapped at distances of plotted to the distance from the PF. Even if the selected distances
5–350 m on the hanging wall and of 21 vd mapped at distances of both on hanging wall and on footwall represent the reach we
5–200 m on the footwall side, derived from 11 of the 15 assume the empirical dataset is the most complete, there is quite
earthquakes, with magnitudes from Mw 4.9 to Mw 7.9. large deviation in the vertical slip values.

Frontiers in Earth Science | www.frontiersin.org 11 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

FIGURE 10 | Relationship between vertical displacement ln (vd) and i) distance from the PF, ln (s), ii) ln DN, and iii) magnitude, Mw for (A) hanging wall and (B)
footwall.

We extrapolated the results from the analysis based on the Matlab codes used for this case study are provided in the
near PF data also to distances greater than 200 and 350 m on Supplementary Material.
footwall and hanging wall, respectively. Especially the large events
may result vd greater than the model would suggest, and in Suasselkä Post-Glacial Fault
general the empirical dataset contains more observations greater Northern Fennoscandia is characterized by large NE-SW
than the median, and greater the observations at short distances. orientated postglacial faults (PGF), which were generated in
This is somewhat counterintuitive but is likely due to the recent times (i.e., the last tens of thousands of years, during
collection of the data in the fields, and it is possible that a the Latest Pleistocene to Holocene) in response to NW-SE
certain distance from the PF only the DRs with largest vertical compressional stress (Lagerbäck and Sundh, 2008; Ojala et al.,
throw were observed and mapped. 2019b; Olesen et al., 2020). The PGFs of the area are mainly very
Finally, in Figure 11, we show the probability of exceeding of
vertical displacements at a site located at 500 m distance from the
PF. We model the probability using a 3 sigma truncated
distribution for two scenarios with Mw 7 and Mw 7.4, and
with the maximum vertical throw expected on the PF,
respectively of 2.3 and 3.85 m. The curves show the
probability for each scenario for the site located either on
hanging wall (solid lines), or on footwall (dashed lines).

CASE STUDY
In this work, we focused our attention on the development of
new approaches to PFDHA by analyzing the hazard related to
DR only, ignoring the probability of surface rupturing and slip
distribution along PF. To appreciate the value of this new
approach, we computed fault-displacement hazard curves for
two scenario events, assuming a surface rupturing on PF along
its total length. The probability of DR occurrence and
displacement levels are estimated here for a site located on FIGURE 11 | Curves for the probability of exceedance for given vertical
the hanging wall of our test case fault, at a distance from the PF displacement levels for hanging wall (HW; solid lines) and footwall (FW; dashed
lines) for Mw 7 (blue lines) and Mw 7.4 (red lines) at a site in 500 m distance from
greater than 100 m and for a characteristic earthquake scenario
the PF.
for which the Mw is derived from the total rupture length.

Frontiers in Earth Science | www.frontiersin.org 12 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

steep (45°–70°), and often associated with old pre-existing faults. until recent times (Sutinen et al., 2014). Radiocarbon datings
According to Craig et al. (2016), the reverse-faulting earthquakes indicate three seismic episodes in the area, the most recent being
causing the large observed scarps were a response, like the coeval 1–3 ky ago (Ojala et al., 2019b). In Suasselkä PGF complex, Ojala
rapid uplift, to the Weichselian ice cap melting: those large events, et al. (2019a) identified four fault systems composed of 37 isolated
with estimated magnitude from 7 up to 8 (see, e.g., Sutinen et al., segments, segment lengths varying between 150 and 7,500 m, and
2014), happened at a time of extensional strain rate field. Surface estimated moment magnitudes ranging from Mw 5.5 to 8.1 based
geodetical measurements (i.e., GNSS) provide velocity field that on the rupture length and coseismic displacement considering the
strongly suggests current extension and rapid uplift (6.5 mm/year fault segments separately, and together as an entire complex. The
in the area of concern, according to the National Land Survey of nearest instrumentally observed earthquake to the Suasselkä fault
Finland, 2018), whereas focal mechanisms show that complex took place in 2008 (Mw 0.9) (Ojala et al., 2019b) and, in a
compressional regime seems to be prevailing around the general consideration, the recent strike-slip to reverse seismic
Scandinavian PGFs (Heidbach et al., 2018). Also the recent activity has been very moderate in terms of magnitudes To
Mw 4.6 shallow earthquake that hit the Kiruna mine area in provide an insight on potential large earthquakes that may
Sweden (May 18, 2020), ∼25 km east of the Pärvie PGF, shows a occur on this fault, we estimate their moment magnitude
reverse focal mechanism as well on a steeply eastward dipping based on the PGF length and the scaling relationships. The
NNE-SSW striking fault. most recent studies of Suasselkä PGF complex based on the
Hence, we argue that the compressional environment that LiDAR data and detailed field studies of the area (see Ojala et al.,
generated the PGF scarps is still active in the region. This is 2019a; Ojala et al., 2019b) result in SRL estimates longer than the
consistent with the recent findings of the Suasselkä fault complex, previous assessments based mainly on visual inspection of the
the longest PGF situated in Northern Finland. The drill core area (e.g., Paananen, 1987). The total SRL calculated between the
sampling of the fault proved that the ancient deformation zone fault complex end points is 72 km; doing so, we assume that the
has reactivated various times in the past, and the alteration of the mapped segments are connected into a single crustal fault
Quaternary deposits indicates that the reactivation has continued at depth.

FIGURE 12 | Suasselkä post-glacial fault as mapped by Ojala et al. (2019a, 2019b). SE block is the upthrown side. The northern section (Scenario 1) consists of
Suaspalo, Suasoja and Nilimaa systems, the expected surface rupture length (SRL) for an earthquake rupturing the entire northern section is 37.6 km. The length of the
entire Suasselkä post-glacial fault complex is 72.4 km (Scenario 2).

Frontiers in Earth Science | www.frontiersin.org 13 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

Example of Probabilistic Fault 515 m). The probability of having a DR occurrence in at least one
Displacement Hazard Analysis: Scenarios 1 of the two r, is 1 – (1–0.408) (1–0.401), and it is equal to 0.645.
The probability of a rank 2 DR occurrence at the site is computed
and 2 via Monte Carlo simulations (Scenario 2 section). It is
In this work, the PFDHA modeling is applied to a hypothetical
independent from the distance r, and it is equal to 0.0182.
site of 10 m width × 20 m length (length  perpendicular to PF
Considering the size of the site under investigation (10 m
strike) situated on the hanging wall of the Suasselkä fault with the
width × 20 m length), the probability of DR occurrence is 1 –
closest side at the distance of 500 m from the PF surface trace. The
(1 – 0.0182)2 and is equal to 0.0361. The conditional probability
modeling is done for two possible scenarios: 1) rupturing of the
of exceedance is shown in Figure 13.
northern half of the complex (i.e., Suaspalo, Suasoja, and Nilimaa
faults), and 2) rupturing of the entire Suasselkä complex
Scenario 2
(Figure 12). We use the expected maximum vertical
The probabilities of having a rank 2 DR occurrence (Eq. 4) on the
displacement of the scenario event on Suasselkä fault for
hanging wall at distances r1 of 505 m and r2 of 515 m from the PF
transforming the attenuation parameters to the expected
are, respectively, 0.496 and 0.489. The probability of having a DR
displacement values on DR, and thus the results are applicable
occurrence in at least one of the two r, is 1 – (1–0.496) (1–0.489),
in the section of the PF trace where the maximum displacement
and it is equal to ∼0.7425. The probability of a rank 2 DR
can be expected to occur. We are not making specific assumptions
occurrence at the site, computed via Monte Carlo simulations,
for the location of the maximum displacement along the PF
independent from the distance r, is equal to 0.02. For the site of
strike. The magnitude is obtained taking an average of Wells and
interest (10 m width × 20 m length), the probability of DR
Coppersmith (1994) and Wesnousky (2008) equations for
occurrence at the site is 1 1 (1 – 0.02)2 and is equal to 0.0361.
magnitude vs. SRL. The Moss and Ross (2011) empirical
The conditional probability of exceedance is shown in Figure 13.
relationship is then used to calculate MD from magnitude,
and then its vertical component (MDv) from fault dip. For
scenario (1) the used Mw is 7.0, MD 2.37 m, and MDv 2.30 m
(dip 75°; Ojala et al., 2019b). For scenario (2) the used parameters
DISCUSSION
are Mw 7.4, MD 3.99 m, and MDv 3.85 m. The dataset used contains all the available surface rupture
information of reverse faulting earthquakes published in the
Scenario 1 literature. The challenge regarding the homogenization of
The probabilities of having a rank 2 DR occurrence (Eq. 4) within different types of the data in a single dataset is not new (Wells
the site of interest, extending from 500 to 520 m from the and Coppersmith, 1994). The first goal of such a study is to
Suasselkä fault, are calculated at distances r1 and r2 from the recognize the most accurate measure to consider for each
PF of 505 and 515 m, respectively, on the hanging wall. The earthquake. Displacement data used here covers 50 years of
obtained probabilities are 0.408 (r1  505 m) and 0.401 (r2  data collection and earthquakes in very different geological
contexts, the type of data available varies from event to
another, but also within an event depending on geologic and
geographic conditions. The resolution of the maps and data
provided by the field research groups have an impact on the
resolution of this study. When working with datasets published
by various study groups, the challenge lays also in bringing
together different data formats and fault traces mapped at
different scales. When it comes to the normalization
parameter for displacement, common parameters of maximum
displacement (MD) and average displacement (AD) along the PF
are not reported in a homogenous way in the reference literature,
especially the latter having more variety in its definition. The
integral-based average considering the displacement distribution
along the whole PF length would be the ideal factor describing the
movement along the total PF rupture, but it is rarely provided and
hardly obtained from the non-continuous data available. Here we
propose to estimate DN locally for each point in the database,
which seems to be reasonably close to the slip profile-based
approach but is remarkably faster to execute using the point-
FIGURE 13 | Conditional probability of exceeding a vertical
displacement, vd, on a distributed rupture at a distance r  510 m from the
based database of surface displacement.
principal fault (PF) on hanging wall side for two modeled scenarios on For a more reliable analysis, a standardized method for field
Suasselkä fault: scenario 1 of Mw 7 and maximum vertical displacement work could diminish the uncertainty of further calculations and
on PF (MDv) of 2.3 m (red), and scenario 2 of Mw 7.4 with a MDv of 3.85 m increase the accuracy of the future hazard analysis. Also, there is
(blue).
an evident need for harmonized nomenclature when it comes to

Frontiers in Earth Science | www.frontiersin.org 14 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

the fault displacement vectors and angles between them. Efforts CONCLUSIONS
have been made in the direction on harmonizing databases of
surface rupturing data and the analysis done here is in accordance Surface displacement data of 15 well-studied reverse earthquakes
with the SURE structure (Baize et al., 2019), thus the parameters were gathered in an uniform database and used to derive
used in modeling can be updated as data is added into the empirical parameters used for describing the probabilities and
database. Detailed field working of DR orientations and sub- expected levels of surface displacement in future events. The focus
surface structures will also help in more accurate ranking of of this work was dedicated to enhancing the methodology for
the DR. obtaining the parameters for distributed surface rupturing to be
The variety of features observed in DR has led us to an in- used in PFDHA on crustal reverse faulting tectonic regime.
depth analysis of their physical origin. This type of analysis has Ranking of the distributed rupturing according to their
not been done in previous studies likely due to the relatively complexity allows us to filter out the DR connected to the
simpler nature of DR associated with strike-slip or normal local structural setting in order to come up with parameters
faulting environments. The rank 2 DR in our database occur for modeling the simple, non-predictable DR that may take place
relatively close to the PF whereas rank >2 DR can occur at anywhere along the strike of the PF, and on any reverse faulting
much greater distances. The occurrence of the more complex earthquake where no structural complexities are expected.
type of DRs (rank > 2) depends largely on the local structural In this work, we introduce the “slicing” methodology for analyzing
geology, on the presence of pre-existing faults prone to be the spatial distribution of DR to account for completeness issues of the
triggered (sympathetic faults), or large scale folding of empirical database. We divide the area off the PF in slices parallel to the
hundreds of meters to kilometers in wavelength (possibly PF strike and consider the slices containing at least a partial DR
the origin of bending-moment or flexural-slip faulting). segment in order to generalize the probability and to avoid the bias
Following this analysis, we propose here to focus only on occurring from the variability in the field mapping conditions that may
the rank 2 DR, which implies that our probability models cause either over or under sampling.
can only be applied at short distances from PF. Vertical displacement data of various earthquakes are brought
Extrapolation to distances greater than ∼1 km from the PF together obtaining a new approach to estimate the vertical
should be considered with caution. Future studies will need to displacement on the principal fault at the position nearest to the
provide probability models that can account for rank >2 DRs distributed rupture along which the vertical throw parameter is
for those sites that are potentially affected by bending moments measured. The approach introduced here considers the alterations
or pre-existing structures. in the slip along the principal fault, and is derived directly from the
Completeness issues on the spatial extent of the compile point based database without slip profiling.
database have led us to propose an alternative approach to We combined empirical regressions and Monte Carlo
estimate the probability of DR at a given distance from the simulations for calculating the probability of DR at the site at
PF. The procedural enhancements introduced here in general a chosen distance from the PF, and to estimate the probability of
aim at reducing the need of user interpretation. For example, for exceedance of the chosen vertical displacement level. To show the
the DR distribution analysis by “slicing” is seen more developed methodology, we provide Matlab codes applied to
independent from the field work, and it does not require chosen scenario events along Suasselkä PGF.
analyzing the reason for the “blank” areas with no DRs on the
surface rupturing maps.
A second issue regards the data collection procedure which DATA AVAILABILITY STATEMENT
becomes increasingly biased for distances greater than 200 and
350 m (footwall, hanging wall) from the PF. Indeed, All datasets presented in this study are included in the article/
extrapolating the displacement prediction equation to greater Supplementary Material.
distances results in vd values lower than the values in the
database. This is somewhat counterintuitive but is likely due
to the collection of the data in the fields, and it is possible that a AUTHOR CONTRIBUTIONS
certain distance from the PF only the DRs with largest vertical
throw were observed and mapped. Those, the equation we FN compiled the database of reverse faulting earthquakes and
developed here are robust within few hundred meters from carried out the implementation. PB and SB provided the
the PF, and extrapolation beyond these ranges need to be geological expert opinion. FV, BP, AV, and OS performed the
carefully considered. numerical simulations. All authors discussed the results and
To better understand the physical parameters controlling vd, contributed to the final manuscript.
it is worth noting the existence of several other factors in
addition to the epicentral distance we are using here. For
example, in the future one could introduce the dip of the SUPPLEMENTARY MATERIAL
fault and the distribution of the co-seismic slip at depth that
can be found in the SRCMOD database (Mai and Thingbaijam, The Supplementary Material for this article can be found online
2014), or the structural complexities of the principal faults (e.g., at: https://www.frontiersin.org/articles/10.3389/feart.2020.581605/
bends, gaps). full#supplementary-material

Frontiers in Earth Science | www.frontiersin.org 15 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

REFERENCES associated with the 2014 Mw 6.2 Nagano earthquake using differential
Lidar. Bull. Seismol. Soc. Am. 109 (2), 780–796. doi:10.1785/0120180020
Kamb, B., Silver, L. T., Abrams, M. J., Carter, B. A., Jordan, T. H., and Minster, J. B.
Angelier, J., Lee, J.-C., Chu, H.-T., and Hu, J.-C. (2003). Reconstruction of fault slip (1971). Pattern of faulting and nature of fault movement in the San Fernando
of the September 21st, 1999, Taiwan earthquake in the asphalted surface of a car earthquake in the San Fernando, California, earthquake of February 9, 1971.
park, and co-seismic slip partitioning. J. Struct. Geol. 25, 345–350. doi:10.1016/ Washington, United States: US Geological Survey, Vol. 733, 41–54.
s0191-8141(02)00038-x Kaneda, H., Nakata, T., Tsutsumi, H., Kondo, H., Sugito, N., Awata, Y., et al. (2008).
Avouac, J.-P., Ayoub, F., Leprince, S., Konca, O., and Helmberger, D. V. (2006). Surface rupture of the 2005 Kashmir, Pakistan, Earthquake and its active tectonic
The 2006 Mw 7.6 Kashmir earthquake: sub-pixel correlation of ASTER images implications. Bull. Seismol. Soc. Am. 98, 521–557. doi:10.1785/0120070073
and seismic waveforms analysis. Earth Planet Sci. Lett. 249, 514–528. doi:10. Kelson, K. I., Kang, K. H., Page, W. D., Lee, C. T., and Cluff, L. S. (2001).
1016/j.epsl.2006.06.025 Representative styles of deformation along the Chelungpu Fault from the
Baize, S., Nurminen, F., Sarmiento, A., Dawson, T., Takao, M., Scotti, O., et al. 1999 Chi-Chi (Taiwan) earthquake: geomorphic characteristic and responses
(2019). A worldwide and unified database of surface ruptures (SURE) for fault of man-made structures. B.Seismol. Soc. Am. 91, 930–952. 10.1785/0120000741
displacement hazard analyses. Seismol Res. Lett. 91, 499–520. doi:10.1785/ Kelson, K. I., Koehler, R. D., Kang, K.-H., Bray, J. D., and Cluff, L. S. (2003). Surface
0220190144 2019 deformation produced by the 1999 Chichi (Taiwan) earthquake and
Bilham, R., and Yu, T.-T. (2000). The morphology of thrust faulting in the 21 interactions with built structures. Walnut Creek, CA: William Lettis and
September 1999, Chichi, Taiwan earthquake. J. Asian Earth Sci. 18, 351–367. Associates, Award No. 01HQ-GR-0122, 21.
doi:10.1016/s1367-9120(99)00071-1 King, T. R., Quigley, M., and Clark, D. (2019). Surface-rupturing historical
Boncio, P., Liberi, F., Caldarella, M., and Nurminen, F.-C. (2018). Width of surface earthquakes in Australia and their environmental effects: new insights from
rupture zone for thrust earthquakes: implications for earthquake fault zoning. Re-analyses of observational data. Geosciences 9, 408. doi:10.3390/
Nat. Hazards Earth Syst. Sci. 18, 241–256. doi:10.5194/nhess-18-241-2018 geosciences9100408
Bowman, J. R., and Barlow, B. C. (1991). Surveys of the fault scarp of the 1986 King, T. R., Quigley, M. C., and Clark, D. (2018). Earthquake environmental effects
marryat creek, south Australia, earthquake. Canberra, AU: [Australian] Bureau produced by the Mw 6.1, 20th May 2016 Petermann earthquake, Australia.
of mineral Resources, geology and Geophysics, BMR Record 1991/190, 12. Tectonophysics 747–748, 357–372. doi:10.1016/j.tecto.2018.10.010
Central Geological Survey, MOEA (2017). Available at: MOEA at http://gis. Kumahara, Y., and Nakata, T. (2007). Recognition of active faults generating the
moeacgs.gov.tw/gwh/gsb97-1/sys8/index.cfm (Accessed October, 2017). 2005 Pakistan earthquake based on interpretation of the CORONA satellite
Chen, G. H., Xu, X. W., Zheng, R. Z., Yu, G. H., Li, F., Li, C. X., et al. (2008). photographs. E - J. GEO. 2, 72–85 [in Japanese with English abstract]. doi:10.
Quantitative analysis of the co-seismic surface rupture of the 2008 Wenchuan 4157/ejgeo.2.72
earthquake, sichuan, China along the Beichuan Yingxiu fault. Seismol. Geol. 30, Lagerbäck, R., and Sundh, M. (2008). Early Holocene faulting and paleoseismicity
723–738 [in Chinese with English abstract]. in northern Sweden. Sweden: Geological Survey of Sweden. ISSN 1103-3371,
Chen, W. C., Chu, H. T., and Lai, T. C. (2000). Surface ruptures of the chi-chi ISBN 978-91-7158-859-3.
earthquake in the shihgang dam area, Special issue for the Chi-Chi earthquake, Lee, J. C., Chen, Y. G., Sieh, K., Mueller, K., Chen, W. S., Chu, H. T., et al. (2001). A
1999, Central geological survey, MOEA, Taipei, Taiwan. Spec. Publ. 12, 41–62 vertical exposure of the 1999 surface rupture of the Chelungpu Fault at Wufeng,
[in Chinese with English abstract]. Western Taiwan: structural and paleoseismic implications for an active thrust
Craig, T. J., Calais, E., Fleitout, L., Bollinger, L., and Scotti, O. (2016). Evidence for fault. Bull. Seismol. Soc. Am. 91 (5), 914–929. 10.1785/0120000742
the release of long-term tectonic strain stored in continental interiors through Lee, Y.-H., Hsieh, M.-L., Lu, S.-D., Shih, T.-S., Wu, W.-Y., Sugiyama, Y., et al. (2003).
intraplate earthquakes. Geophys. Res. Lett. 43, 6826. doi:10.1002/2016GL069359 Slip vectors of the surface rupture of the 1999 Chi-Chi earthquake, western
Crone, A. J., Machette, M. N., and Bowman, J. R. (1992). Geologic investigations of Taiwan. J. Struct. Geol. 25, 1917–1931. doi:10.1016/s0191-8141(03)00039-7
the 1988 Tennant Creek, Australia, earthquakes– implications for Lee, Y.-H., Wu, K.-C., Rau, R.-J., Chen, H.-C., Lo, W., and Cheng, K.-C. (2010).
paleoseismicity in stable continental regions. Denver, CO: US Geological Revealing coseismic displacements and the deformation zones of the 1999 Chi-
Survey, Bull. 2032-A, 51. Chi earthquake in the Tsaotung area, central Taiwan, using digital cadastral
Fredrich, J., McCaffrey, R., and Denham, D. (1988). Source parameters of seven data. J. Geophys. Res. 115, B03419. doi:10.1029/2009JB006397
large Australian earthquakes determined by body waveform inversion. Lettis, W. R., Wells, D. L., and Baldwin, J. N. (1997). Empirical observations
Geophys. J. Int. 95, 1–13. doi:10.1111/j.1365-246x.1988.tb00446.x regarding reverse earthquakes, blind thrust faults, and quaternary deformation:
Gordon, F. R., and Lewis, J. D. (1980). The Meckering and Calingiri earthquakes are blind thrust faults truly blind? Bull. Seismol. Soc. Am. 87, 1171–1198.
October 1968 and March 1970. B. Geol. Survey of Western Australia. Lin, A., Sano, M., Yan, B., and Wang, M. (2015). Co-seismic surface ruptures produced
126.169–188 by the 2014 Mw 6.2 Nagano earthquake, along the Itoigawa–Shizuoka tectonic line,
Haessler, H., Deschamps, A., Dufumier, H., Fuenzalida, H., and Cisternas, A. central Japan. Tectonophysics 656, 142–153. doi:10.1016/j.tecto.2015.06.018
(1992). The rupture process of the Armenian earthquake from broad-band Lin, W. H. (2000). On surface deformations from the Chi-Chi earthquake in the
teleseismic body wave records. Geophys. J. Int. 109, 151–161. doi:10.1111/j. Shihkang and Chutzekeng areas, special issue for the Chi-Chi earthquake, 1999,
1365-246x.1992.tb00085.x Central Geological Survey, MOEA, Taipei, Taiwan. Spec. Publ. 12, 1–17. [in
Heidbach, O., Rajabi, M., Cui, X., Fuchs, K., Müller, B., Reinecker, J., et al. (2018). Chinese with English abstract].
The world stress map database release 2016: crustal stress pattern across scales. Liu-Zeng, J., Sun, J., Wang, P., Hudnut, K. W., Ji, C., Zhang, Z., et al. (2012). Surface
Tectonophysics 744, 484–498. doi:10.1016/j.tecto.2018.07.007 ruptures on the transverse Xiaoyudong fault: a significant segment boundary
Huang, C., Chan, Y.-C., Hu, J.-C., Angelier, J., and Lee, J.-C. (2008). Detailed breached during the 2008 Wenchuan earthquake, China. Tectonophysics 580,
surface co-seismic displacement of the 1999 Chi-Chi earthquake in western 218–241. doi:10.1016/j.tecto.2012.09.024
Taiwan and implication of fault geometry in the shallow subsurface. J. Struct. Liu-Zeng, J., Sun, J., Zhang, Z., Wen, L., Xing, X., Hu, G., et al. (2010). Detailed mapping
Geol. 30, 1167–1176. doi:10.1016/j.jsg.2008.06.001 of surface rupture of the Wenchuan Ms 8.0 earthquake near Hongkou and
Huang, W. J., Chen, Z. Y., Liu, S. Y., Lin, Y. H., Lin, C. W., and Chang, H. C. (2000). seismotectonic implications. Quat. Sci. 30, 1–29 [in Chinese with English abstract].
Surface deformation models of the 1999 Chi–Chi earthquake between Liu-Zeng, J., Zhang, Z., Wen, L., Tapponnier, P., Sun, J., Xing, X., et al. (2009).
Tachiachi and Toupienkengchi, central Taiwan, Special Issue for the Chi- Co-seismic ruptures of the 12 May 2008, Ms 8.0 Wenchuan earthquake,
Chi Earthquake, 1999, Central Geological Survey, MOEA, Taipei, Taiwan, Sichuan: east–west crustal shortening on oblique, parallel thrusts along the
Special Publication, 12, pp. 63–87. (in Chinese with English abstract). eastern edge of Tibet. Earth Planet Sci. Lett. 286, 355–370. doi:10.1016/j.epsl.
Ishimura, D., Okada, S., Niwa, Y., and Toda, S. (2015). The surface rupture of the 22 2009.07.017
November 2014 Nagano-ken-hokubu earthquake (Mw 6.2), along the Kamishiro Machette, M. N., Crone, A. J., and Bowman, J. R. (1993). Geologic investigations of
fault, Japan, Active Fault Res. 43, 95–108. [in Japanese, with English abstract]. the 1986 Marry at Creek, Australia, earthquakes—implications for
Ishimura, D., Toda, S., Mukoyama, S., Homma, S. i., Yamaguchi, K., and paleoseismicity in stable continental regions. Denver, CO: US Geological
Takahashi, N. (2019). 3D surface displacement and surface ruptures Survey, Bull. 2032-B, 29.

Frontiers in Earth Science | www.frontiersin.org 16 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

Mai, P. M., and Thingbaijam, K. K. S. (2014). SRCMOD: an online database of Takao, M., Tsuchiyama, J., Masashi, A., and Testushi, K. (2013). Application of
finite-fault rupture models. Seismol Res. Lett. 85 (6), 1348–1357. doi:10.1785/ probabilistic fault displacement hazard analysis in Japan. Jpn. Assoc.
0220140077 Earthquake Eng. 13 (1), 17–36. doi:10.5610/jaee.13.17
McCaffrey, R. (1989). Teleseismic investigation of the January 22, 1988 Tennant U.S. Geological Survey Staff (1971). Surface faulting, in The San Fernando,
Creek, Australia, earthquakes. Geophys. Res. Lett. 16, 413–416. doi:10.1029/ California, earthquake of February 9, 1971. Washington, United States: US
gl016i005p00413 Geological Survey, Vol. 733, 55–76.
Meghraoui, M., Jaegy, R., Lammali, K., and Albarède, F. (1988). Late Holocene Wang, H., Ran, Y., Chen, L., Shi, X., Liu, R., and Gomez, F. (2010). Determination
earthquake sequences on the El Asnam (Algeria) thrust fault. Earth Planet Sci. of horizontal shortening and amount of reverse-faulting from trenching across
Lett. 90, 187–203. doi:10.1016/0012-821x(88)90100-8 the surface rupture of the 2008 Mw 7.9 Wenchuan earthquake, China,
Moss, R. E. S., and Ross, Z. E. (2011). Probabilistic fault displacement hazard Tectonophysics, 491, 10–20. doi:10.1016/j.tecto.2010.03.019
analysis for reverse faults. Bull. Seismol. Soc. Am. 101 (4), 1542–1553. doi:10. Wells, D. L., and Coppersmith, K. J. (1994). New empirical relationship among
1785/0120100248 magnitude, rupture length, rupture width, rupture area, and surface
Ojala, A. E. K., Mattila, J., Ruskeeniemi, T., Markovaara-Koivisto, M., Palmu, J.-P., displacement. Bull. Seismol. Soc. Am. 84 (4), 974–1002.
Nordbäck, N., et al. (2019a). Report 2019-1. Postglacial faults in Finland – a Wesnousky, S. G. (2008). Displacement and geometrical characteristics of
review of PGSdyn project results. Eurajoki, Finland: Posiva Oy, ISBN: 978-951- earthquake surface ruptures: issues and implications for seismic hazard
652-281-8, ISSN: 2343-4740. analysis and the earthquake rupture process. Bull. Seismol. Soc. Am. 98,
Ojala, A. E. K., Mattila, J., Ruskeeniemi, T., Palmu, J.-P., Nordbäck, N., Kuva, J., 1609–1632. doi:10.1785/0120070111
et al. (2019b). Postglacial reactivation of the Suasselkä PGF complex in SW Xu, X., Wen, X., Ye, J., Ma, B., Chen, J., Zhou, R., et al. (2008). The Ms 8.0
Finnish Lapland. Int. J. Earth Sci. 108, 1049–1065. doi:10.1007/s00531-019- Wenchuan earthquake surface ruptures and its seismogenic structure. Seismol.
01695-w Geol. 30, 597–629 [in Chinese with English abstract].
Okada, S., Ishimura, D., Niwa, Y., and Toda, S. (2015). The first surface rupturing Xu, X., Wen, X., Yu, G., Chen, G., Klinger, Y., Hubbard, J., et al. (2009). Co-seismic
earthquake in 20 years on a HERP active fault is not characteristic: the 2014 Mw reverse- and oblique-slip surface faulting generated by the 2008 Mw 7.9
6.2 Nagano event along the northern Itoigawa–Shizuoka tectonic line. Seismol Wenchuan earthquake, China. Geology 37, 515–518. doi:10.1130/g25462a.1
Res. Lett. 86, 1–14. doi:10.1785/0220150052 Yeats, R. S. (1986). Active faults related to folding in: active tectonics: impact on
Olesen, O., Olsen, L., Gibbons, S. J., Ruud, B. O., Høgaas, F., Johansen, T. A., et al. society. Washington, United States: The National Academics Press
(2020). Postglacial faulting in Norway—large magnitude earthquakes of late Washington, 280. doi:10.17226/624
Holocene age. CambridgeEngland: Cambridge University Press. Yelding, G., Jackson, J. A., King, G. C. P., Sinvhal, H., VitaFinzi, C., and Wood, R. M.
Ota, Y., Watanabe, M., Suzuki, Y., Yanagida, M., Miyawaki, A., and Sawa, H. (1981). Relations between surface deformation, fault geometry, seismicity, and
(2007). Style of the surface deformation by the 1999 Chichi earthquake at the rupture characteristics during the El Asnam (Algeria) earthquake of the 10 October
central segment of Chelungpu fault, Taiwan, with special reference to the 1980. Earth Planet Sci. Lett. 56, 287–304. doi:10.1016/0012-821X(81)90135-7
presence of the main and subsidiary faults and their progressive deformation in Youngs, R. R., Arabasz, W. J., Anderson, R. E., Ramelli, A. R., Ake, J. P.,
the Tsauton area. J. Asian Earth Sci. 31, 214–225. doi:10.1016/j.jseaes.2006.07. Slemmons, D. B., et al. (2003). A methodology for probabilistic fault
030 displacement hazard analysis (PFDHA). Earthq. Spectra. 19 (1), 191–219.
Paananen, M. (1987). Report YST. Venejärven, Ruostejärven, Suasseljän ja doi:10.1193/1.1542891
Pasmajärven postglasiaalisten siirrosten geofysikaalinen tutkimus. Espoo, Yu, G., Xu, X., Klinger, Y., Diao, G., Chen, G., Feng, X., et al. (2010). Fault-scarp
Finland: Geological Survey of Finland. features and cascading-rupture model for the Mw 7.9 Wenchuan earthquake,
Petersen, M. D., Dawson, T. E., Chen, R., Cao, T., Wills, C. J., Schwartz, D. P., et al. eastern Tibetan plateau, China. Bull. Seismol. Soc. Am. 100, 2590–2614. doi:10.
(2011). Fault displacement hazard for strike-slip faults. Bull. Seismol. Soc. Am. 1785/0120090255
101 (2), 805–825. doi:10.1785/0120100035 Yu, G.-H., Xu, X.-W., Chen, G.-H., Guo, T.-T., Tan, X.-B., Yang, H., et al. (2009).
Philip, H., and Meghraoui, M. (1983). Structural analysis and interpretation of the Relationship between the localization of surface ruptures and building damages
surface deformation of the El Asnam earthquake of October 10, 1980, Tectonics associated with the Wenchuan 8.0 earthquake. Chin. J. Geophys. 52, 1294–1311.
2, 17–49. doi:10.1029/tc002i001p00017 doi:10.1002/cjg2.1455
Philip, H., Rogozhin, E., Cisternas, A., Bousquet, J. C., Borisov, B., and Zhang, J. Y., Bo, J. S., Xu, G. D., and Huang, J. Y. (2012). Buildings setbacks research
Karakhanian, A. (1992). The Armenian earthquake of 1988 December 7: from surface-fault-rupture statistical analysis. Amministrare 204–208,
faulting and folding, neotectonics and palaeoseismicity. Geophys. J. Int. 110, 2410–2418. doi:10.4028/www.scientific.net/amm.204-208.2410
141–158. doi:10.1111/j.1365-246x.1992.tb00718.x Zhang, Y., Shi, J., Sun, P., Yang, W., Yao, X., Zhang, C., et al. (2013). Surface ruptures
Rajendran, C. P., Rajendran, K., Unnikrishnan, K. R., and John, B. (1996). induced by the Wenchuan earthquake: their influence widths and safety distances
Palaeoseismic indicators in the rupture zone of the 1993 Killari (Latur) for construction sites. Eng. Geol. 166, 245–254. doi:10.1016/j.enggeo.2013.09.010
earthquake. Curr. Sci. India. 70, 385–390. Available at: https://www.jstor. Zhang, Y. S., Sun, P., Shi, J. S., Yao, X., and Xiong, T. Y. (2010). Investigation of
org/stable/24097438 (Accessed March 10, 1996). rupture influenced zones and their corresponding safe distances for
Ritz, J. F., Baize, S., Ferry, M., Larroque, C., Audin, L., Delouis, B., et al. (2020). reconstruction after 5.12 Wenchuan earthquake. Eng. Geol. 18, 312–319 [in
Surface rupture and shallow fault reactivation during the 2019 Mw 4.9 Le Teil Chinese with English abstract].
earthquake, France. Commun. Earth Environ. 1, 10. Zhou, Q., Xu, X., Yu, G., Chen, X., He, H., and Yin, G. (2010). Width distribution of
Rymer, M. J., Kendrick, K. J., Lienkaemper, J. J., and Clark, M. M. (1990). the surface ruptures associated with the Wenchuan earthquake: implication for
“Surface rupture on the Nunez fault during the Coalinga earthquake the setback zone of the seismogenic faults in post-earthquake reconstruction.
sequence,” in The Coalinga, California, earthquake of May 2, 1983. Bull. Seismol. Soc. Am. 100, 2660–2668. doi:10.1785/0120090293
Editors M. J. Rymer and W. L. Ellsworth (Denver, CO: US Geological
Survey), 299–318, Prof. Paper 1487. Conflict of Interest: The authors declare that the research was conducted in the
Sarmiento, A., Abrahamson, N., Baize, S., Bozorgnia, Y., Chen, R., Coppersmith, K. absence of any commercial or financial relationships that could be construed as a
J., et al. (2019). “A new model database for next-generation fault displacement potential conflict of interest.
hazard analysis,” in SSA 2020 Annual Meeting, Albuquerque, New Mexico.
Seeber, L., Ekström, G., Jain, S. K., Murty, C. V. R., Chandak, N., and Copyright © 2020 Nurminen, Boncio, Visini, Pace, Valentini, Baize and Scotti. This
Armbruster, J. G. (1996). The 1993 Killari earthquake in central India: a is an open-access article distributed under the terms of the Creative Commons
new fault in Mesozoic basalt flows? J. Geophys. Res. 101, 8543–8560. doi:10. Attribution License (CC BY). The use, distribution or reproduction in other forums is
1029/95JB01865 permitted, provided the original author(s) and the copyright owner(s) are credited
Sutinen, R., Hyvönen, E., and Kukkonen, I. (2014). LiDAR detection of and that the original publication in this journal is cited, in accordance with accepted
paleolandslides in the vicinity of the Suasselkä postglacial fault, Finnish academic practice. No use, distribution or reproduction is permitted which does not
Lapland. Int. J. Appl. Earth Obs. Geoinf. 27, 91–99. doi:10.1016/j.jag.2013.05.004 comply with these terms.

Frontiers in Earth Science | www.frontiersin.org 17 October 2020 | Volume 8 | Article 581605

Nurminen et al. Probability of DR for Reverse Earthquakes

GLOSSARY PGF post-glacial fault

ps the nearest point on the PF trace from the point of vd on DR to which the
DN expected vertical displacement calculated for each vd from the two distance s is measured
nearest VD
r a distance vector from DR trace to the nearest PF trace
DR distributed (also called off-fault) rupturing: surface ruptures that are not
direct surface expressions of PF faulting s a distance vector from a point along DR with displacement to the nearest
PF trace
F fraction of total DR length in respect to the length of the PF of the
corresponding earthquake SRL surface rupture length
FDHI fault displacement hazard Initiative SURE a worldwide and unified database of surface ruptures
ND net displacement
VD vertical displacement on PF
PF principal (also called primary) fault rupturing; the fault along which the
seismic energy releases in the seismic event vd vertical displacement on DR
PFDHA probabilistic fault displacement hazard analysis VDL vertical displacement level

Frontiers in Earth Science | www.frontiersin.org 18 October 2020 | Volume 8 | Article 581605