Proceedings of the 12th International INQUA meeting on paleoseismology, active tectonic and archaeoseismology

We analysed surface rupture traces and displacement data points of normal and reverse earthquakes in the SURE 2.0 database, for a total of 18 normal (Mw from 5.6 to 7.5) and 16 reverse (Mw from 4.9 to 7.9) earthquakes. Logistic regression models are used to evaluate the likelihood of DR occurrence. Factors such as earthquake magnitude, proximity to the principal fault, faulting style, Ranking of DRs, and footwall vs hanging wall location of DR are considered as predictive variables. Furthermore, a fault displacement model is developed to estimate the median throw on DRs based on these parameters. R E S U LT S A N D A P P L I C A B I L I T Y The probability of occurrence of a DR at a given distance to PF P(DRocc) is obtained by applying a multinomial logistic model, and is given by the equation: Fig. 1: Diagram illustrating the rupture ranking for normal faults according to the SURE 2.0 database (Nurminen et al., 2022). Logistic regression models are used to evaluate the likelihood of DR occurrence. Factors such as earthquake magnitude, proximity to the principal fault, faulting style, Ranking of DRs, and footwall vs hanging wall location of DR are considered as predictive variables. Furthermore, a fault displacement model is developed to estimate the median throw on DRs based on these parameters. RESULTS AND APPLICABILITY The probability of occurrence of a DR at a given distance to PF P(DR occ ) is obtained by applying a multinomial logistic model, and is given by the equation: ( !"" ) = 1 − ( ) 1 + ( ) where = + + + where Mw and Dist are the earthquake magnitude and distance from PF , HF indicates the hanging wall ( HF = 0) or footwall ( HF = 1) location of DR, and parameters a , b , c , and d depend on style of faulting, DR Rank and site dimension. For computing the P(DR occ ) of a Rank 2 DR along the dimension of a site of interest parallel to the PF strike (e.g., a site shorter than the entire length of the PF), we updated the Monte Carlo methodology proposed in Nurminen et al. (2020) by implementing the slicing window method for considering different widths. faulting (SoF=1 for normal earthquake, SoF=0 for reverse earthquake), location of DR with respect to PF (HWFW = 0 for hanging wall, HWFW = 1 for footwall); and Rank DR-to-Rank PF combination. Fault displacement hazard analysis requires to collect detailed geologic information within a certain distance to the site of interest (e.g., IAEA, 2022). But the degree of knowledge that can be collected varies from site to site and may evolve with time. Thanks to the different logistic and fault displacement models for DRs developed here, a PFDHA decision process can be i plemented considering the level of geologic knowledge available. The idea is that, to manage the epistemic uncertainties, the level of geologic background of the area under investigation should guide the applicant in choosing the most appropriate regressions, or combination of regressions, to be applied. To illustrate this PFDHA decision process, consider a site located at a certain distance from a well-located PF trace (Fig. 2). In this example, attention is focused on the application of the new regressions here developed, ignoring the probability of surface rupturing and the annual rate of occurrence of the earthquakes. Three different situations can be considered: Case 1: a pre-existing fault (potential rupture of Rank 1.5) cannot be excluded under the site or at a close distance to the site because it has not been studied. In this case the probability of its reactivation as DR should be considered. The regressions for probability of occurrence and displacement exceedance to be used are then: i) the occurrence Rank 2 DRs associated with PF rupturing (Rank 1); ii) Rank 2 associated with Rank 1.5; and Rank 1.5 associated with Rank 1. In this case, the hazard is dominated by DR Rank 1.5. The total hazard is given as a sum of combinations. The same applies when a pre-existing fault is proved to exist beneath the site (i.e., it has been studied and located). Case 2: following detailed investigations, a pre- existing fault (Rank 1.5) is identified at a distance from the site and excluded under the site. Then, only the where Fig. 1: Diagram illustrating the rupture ranking for normal faults according to the SURE 2.0 database (Nurminen et al., 2022). Logistic regression models are used to evaluate the likelihood of DR occurrence. Factors such as earthquake m gnitude, proximity to the principal fault, fau ting style, Ranking of DRs, and footwall vs hanging w ll location of DR are consider d as predictive ariables. Further ore, a fault splacement model is d veloped to estimat the median throw on DRs based on these parameters. RESULTS AND APPLICABILITY The probability of ccurrence of a DR at a given distance to PF P(DR occ ) is obtained by applying a multi omial logistic model, and is given by the equation: ( !"" ) = 1 − ) 1 + ( ) where = + + + wher Mw and Dist are the earthquake magnitude and distance from PF , HF indicates the h ging wall ( HF = 0) or footwall ( HF = 1) location of DR, and p rameters a , b , c , and d depend on style of faulting, DR Ra k and site dimension. For computing the P(DR occ ) of Rank 2 DR along th dimension f a site of int rest parallel to the PF strike (e.g., a si e shorter th n the entire ength f th PF), we updated he Monte Carlo m thodology proposed in Nurminen et al. (2020) by implementing the slicing window method for considering different widths. indicator variables, introduced to specify style-o faulting (SoF=1 for normal earthquake, SoF=0 f reverse earthquake), loc tion of DR with espect to P (HWFW = 0 f r hanging wall, HWFW = 1 for footwal and Rank DR-to-Rank PF combination. F ult displaceme t hazard analysis requires to coll detailed geologic information within a cer ain distan to the site of interest ( .g., IAEA, 2022). But the d gre of k owledge that can be collected varies from site s te and may evolve with ti . Thanks to the differe logistic and f ult displac ment m d l for DR developed here, a PFDHA decision process can b impl m nted consid ring the level of olog knowledge avail ble. The id a is that, to manage th epistemic uncert inties, the l vel of ge log background of he area u der inve tigation shou uide the appli ant i ch osing the most appr pria regressi ns, or combination of r gressions, to b appli d. To il ustrate this PFDHA ecision proces consider a site located at a certain distance from well-located PF trace (Fig. 2). In is example, attentio is focus d on the application of the new regressio here developed, ig oring he pr bability of surfa rupt ring and th annual rate of occurrence of th earthquakes. Three different situations can b considered: Case 1: a pr - isting fault (potential rupture of Ran 1.5) annot be xcluded under the sit or at a clo dist nce o the site ecause it has no been studied. this case the p obability of it reactivati n as D should be consi ered. Th r gressions for proba ili of occurr ce and displa me t exceedance to b used a e then: i) the occurrence Rank 2 DRs associate with PF rupturi g (Rank 1); ii) Rank 2 associated wi Rank 1.5; nd Rank 1.5 associated with Rank 1 In th c se, the hazard is dominated by DR R nk 1 5. Th total hazard is given as a sum o combinations. Th same applies when a pre-existing fault is proved exist beneath the site (i.e., it has been studied an located). Case 2: following detailed investigations, a pr xis ing fault (Rank 1.5) is identified at a dista ce fro the site and excluded under the site. Then, only th where Mw and Dist are the earthquake magnitude and distance from PF, HF indicates the hanging wall (HF = 0) or footwall (HF = 1) location of DR, and parameters a, b, c, and d depend on style of faulting, DR Rank and site dimensio . F r computing the P(DRocc) of a R k 2 DR al g the dimension of a site of interest parallel to the PF strik (e.g., a site shorter than the entire length of the PF), we updat d the Monte Carlo methodology proposed in Nurminen et al. (2020) by implementing the slicing window method for considering different widths. The expression adopted for evaluating the probability of exceedance of the vertical component of slip (throw) on a DR is: seismology, Active Tectonics and Archaeoseismology (PATA), -11 th , 2024, Los Andes, Chile ent the om 7.9) mal al., The expression adopted for evaluating the probability of exceedance of the vertical component of slip (throw) on a DR is: ( ) = + ( ( )) + ( ( ℎ )) + ( ) + ( ) + ( ) + ( ) where a is the intercept, s , throwPFmean , and Mw are, respectively, the distance, the mean throw on PF derived from a distance-dependent approach, and moment magnitude. SoF , HWFW , and COMB are indicator variables, introduced to specify style-of- faulting (SoF=1 for normal earthquake, SoF=0 for reverse earthquake), location of DR with respect to PF (HWFW = 0 for hanging wall, HWFW = 1 for footwall); and Rank DR-to-Rank PF combination. Fault displacement hazard analysis requires to collect detailed geologic information within a certain distance to the site of interest (e.g., IAEA, 2022). But the degree of knowledge that can be collected varies from site to site and may evolve with time. Thanks to the different logistic and fault displacement models for DRs where a is the intercept, s, throwPFmean, andMw are, respectively, the distance, the mean throw on PF derived from a distance- dependent approach, and moment magnitude. SoF, HWFW, and COMB are indicator variables, introduced to specify style- offaulting (SoF=1 for normal earthquake, SoF=0 for reverse earthquake), location of DR with respect to PF (HWFW = 0 for hanging wall, HWFW = 1 for footwall); and Rank DR-to-Rank PF combination.