ML24355A119

From kanterella
Jump to navigation Jump to search
Kottke - Alternative Data Sources for Improving Seismic Hazard Assessments
ML24355A119
Person / Time
Site: Diablo Canyon  Pacific Gas & Electric icon.png
Issue date: 10/30/2024
From: Kottke A
Geosciences, Pacific Gas & Electric Co
To:
NRC/RES/DE
Thomas Weaver 301-415-2383
Shared Package
ML24355A102 List:
References
Download: ML24355A119 (21)
v • d • e


Text

Alternative Data Sources for Improving Seismic Hazard Assessments Albert Kottke, P.E., Ph.D.

Geosciences, Pacific Gas & Electric Co.

Quick intro to Diablo Canyon NPP

  • Seismic hazard dominated by four local fault sources and background areal zone
  • Detached blocks identified during Extreme Ground Motion Project 10/30/2024 DOE-NRC NPH 2

Fragile geologic features near Diablo Canyon 10/30/2024 DOE-NRC NPH 3

Rood et al. (2020) Study

  • Cosmogenic age dating used to constrain date of current configuration:
  • Ages range from 17 to 95 ky
  • Sampling show an incremental block removal process
  • Youngest cluster of exposure ages gives an average fragility age of 21 ky
  • Fragility estimated by Purvance et al.

(2008) approach:

  • 3D models based on photogrammetry
  • Fragility defined by a radius and toppling angle
  • Hazard defined by PGA and PGV/PGA 10/30/2024 DOE-NRC NPH 4

Rood et al. (2020) Conclusions

  • Features plotted in hazard space using Baker et al. (2013) method
  • Scales hazard considering probability of survival
  • 5% probability of survival assumed
  • Presence of feature DRW1 suggests hazard is conservative --

93% of the weighted hazard curves are inconsistent with its untoppled state 10/30/2024 DOE-NRC NPH 5

Ongoing work to improve estimates

  • Prior to using results, need to improve our understanding
  • Ongoing activities:
  • Instrumentation by UCSB
  • Christine Wittich (UNL) analytical feature-specific fragilities with laboratory calibrations
  • Ramon Arrowsmith(ASU) 3D modeling of features and automated data collection
  • Norm Abrahamson (UCB) inspection of hazard 10/30/2024 DOE-NRC NPH 6

Informing hazard with fragile feature existence Three approaches that could be used to update hazard based on existence of fragile features:

1. Reject hazard curves identified as being inconsistent and renormalized end-branch logic tree weights
2. Identify which logic tree branches are causing the inconsistency and re-evaluate the logic tree weights
3. Bayesian updating of logic tree weights All approaches assume that current logic tree is assumed to capture the true behavior 10/30/2024 DOE-NRC NPH 7

Evaluation of logic tree branches

  • For each node in the logic tree, check the fraction of realizations that are consistent with the PBR for each branch
  • Ratio of pass / total number
  • If fraction is are similar for the branches then this node does not contribute to the inconsistency 10/30/2024 DOE-NRC NPH 8

Feature survival implies larger characteristics magnitude along Hosgri Fault

Evaluation of logic tree branches

  • For each node in the logic tree, check the fraction of realizations that are consistent with the PBR for each branch
  • Ratio of pass / total number
  • If fraction is are similar for the branches then this node does not contribute to the inconsistency 10/30/2024 DOE-NRC NPH 9

Feature survival implies lower slip rate along Hosgri Fault

Evaluation of logic tree branches

  • For each node in the logic tree, check the fraction of realizations that are consistent with the PBR for each branch
  • Ratio of pass / total number
  • If fraction is are similar for the branches then this node does not contribute to the inconsistency 10/30/2024 DOE-NRC NPH 10 Shows that the GMM is a key cause of the inconsistency; only 10% of alternative models are consistent with the fragile features

Contaminated conclusions

  • Only branches with source characteristic models leading to the small hazard (e.g. smallest slip rates, largest Mmax) are selected
  • Evaluating logic tree branches is contaminated by the fact that the underlying ground motion model doesnt capture the true range of behavior
  • Using Bayesian updating will have this same issue; assumes that the correct model is within the set of alternatives
  • Need to refine ground motion models to better capture variation in behavior between sites 10/30/2024 DOE-NRC NPH 11

Other work on fragile features

  • PG&E and Arrowsmith planning on evaluating background models using fragile features in glaciated terrains 10/30/2024 DOE-NRC NPH 12

Regionalization of ground motion models

  • Ergodic ground motion models pool data from global datasets Required due to paucity of relevant data (large M, close distance)

Approach fails to capture repeatable local effects Process inflates unmodeled error (aleatory variability) because differences in locations are not modeled

  • Non-ergodic models adjust for repeatable local effects Increased model complexity with spatially varying coefficients Improved partitioning between modeled and unmodeled errors
  • Regionalization is not new (e.g., CENA and WNA, host-to-target corrections) but increased data has shown need to further refine models 10/30/2024 DOE-NRC NPH 13

Modeled and unmodeled components

  • Ergodic models represent the collection of alternative nonergodic models
  • Interested in the ground motion at a specific site - not the average over all sites
  • Use data to constrain nonergodic components:
  • Observations
  • Simulations 10/30/2024 DOE-NRC NPH 14 Plots of 7 samples; average computed by 250 samples

Shifting towards smaller magnitudes

  • Smaller magnitude earthquakes occur more frequently and can help characterize repeatable behaviors
  • Spectral acceleration:
  • Typically used by engineering for assessment and design
  • Adjustments depend on frequency content; nicely shown by Stafford et al. (2017)
  • Cant readily use observations at M 3 to inform behavior for M 7
  • Fourier amplitude spectrum (FAS):
  • Not directly usable for assessment and design
  • Adjustments are linear and independent of adjacent frequencies
  • Used for recent hazard studies and ground motion models being developed under PG&E supported NGA-West3
  • FAS permits leveraging data from weak motion to define non-ergodic effects 10/30/2024 DOE-NRC NPH 15

Example application of weak motion

  • Several PG&E dams instrumented:
  • Autonomous instrument record events PGA > 0.01 g
  • ~20 years of operation but only a few recordings
  • Lake Almanor
  • In 2023 free-field instrument didnt record event; battery failure
  • ~12 hours after event, deployed temporary broadband station
  • Recorded continuous data for 2 weeks and recorded 60 events >M 2 and 5 events >M 3
  • Not every site is going to be like Lake Almanor, but higher likelihood in observing weak motion 10/30/2024 DOE-NRC NPH 16

Distribution of recorded earthquakes 10/30/2024 DOE-NRC NPH 17

Empirical site response

  • Recordings processed using USGS developed PG&E supported GMProcess
  • Compute site terms using two different approaches:
  • Mixed-effects regression with Bayless and Abrahamson (2019) model
  • Markov-Chain Monte Carlo with a Stafford et al. (2022) point-source model
  • Compare site term with HVSR data 10/30/2024 DOE-NRC NPH 18 Dam response Fundamental mode of free-field Higher Mode of free-field; potentially influenced by dam

Response spectra adjustment 10/30/2024 DOE-NRC NPH 19 Compute site-adjusted response spectrum with random vibration theory (RVT)

Uncertainty is much lower than that estimated by Vs30

Conclusions

  • Fragile geologic features offer a unique opportunity to test the results of hazard studies and have already yielded results
  • Existing ergodic ground motion models dont provide site-specific hazard - without adjustment of median and error
  • Advancing site-specific hazard requires extending the range of usable data to smaller magnitudes and adoption of FAS-based ground motion models 10/30/2024 DOE-NRC NPH 20

Thank You Albert Kottke albert.kottke@pge.com 21

Retrieved from "https://kanterella.com/w/index.php?title=ML24355A119&oldid=5451002"