Australia_National Seismic Hazard Model.pptx

Tells about the National Seismic hazard Map

1. The 2018 national seismic hazard assessment of Australia: Quantifying hazard changes and model uncertainties Authors: Trevor I Allen, Jonathan D Griffin, Mark Leonard, Dan J Clark and Hadi Ghasemi Presented By Ravindra Kumar Gupta

2. Summary •The NSHA18 is developed by Geoscience Australia in collaboration with the Australian seismology community, as stated by Allen et al. (2018b). •The incorporation of a diverse set of Seismic Source Models (SSMs) to thoroughly address the epistemic uncertainties inherent in national-scale seismic-source modelling is a fundamental novelty of the NSHA18 over previous versions. •19 different seismic-source characterizations for Australia and its nearby offshore territories. The NSHA18 used a range of model types, including background and regional seismicity, each of which contributed to the distinct evaluation of seismic hazard for different places. •These models were evaluated in a structured expert opinion workshop and synthesised inside a weighted logic-tree framework (Griffin et al., 2018).

3. • PGA values for the 1/500 Annual Exceedance Probability (AEP) (which is equivalent to 10% POE in 50 years) in Australia have been reduced by 72% on average when compared to values used in the Australian earthquake loading standard, AS 1170.4-2007 (Standards Australia, 2007). This remarkable drop in seismic hazard factors is due to a number of major variables. •Why this happened • First , there was a correction made to the estimates of local magnitudes (M L ) for earthquakes that occurred before 1990. • Second , these local magnitude values were converted to moment magnitudes (M w ), a more uniformly applicable scale. These changes have led to a lower predicted rate of such earthquakes, which in turn has contributed to the reduction in seismic hazard factors.

4. • Third, Increases in Gutenberg and Richter (1944) ‘b’ values have occurred, particularly in eastern Australia, as a result of M L to M w conversions, which reduce the frequency of infrequent large earthquakes relative to more often observed moderate magnitude earthquakes (Allen et al., 2018c). • Fourth, The use of modern ground-motion attenuation models that predict lower ground motions and faster attenuation of PGA and other spectral ordinates with increasing distance. •Forecasting seismic hazard in Stable Continental Regions (SCRs) is difficult due to long recurrence time of large earthquakes and the lack of detailed data required for model calibration, as opposed to Active Tectonic Regions (ATRs).

5. Seismotectonic setting of Australia 4. Eastern Australian Phanerozoic accretionary terranes (eastern region) 5. Eastern extended continental crust (eastern high lands) 6. Western extended continental crust 7. Passive margin extended continental crust (exterior region of country) 1. Precambrian Craton and non-reactivated Proterozoic crust (western and central region) 2. Reactivated Proterozoic crust (middle to western region) 3. Sprigg Orogen (bounded by passive margin extended continental crust)

6. Seismic design maps for Australia • The occurring of damaging earthquake in Australia is of low probability but if it would happen there will be high consequences due to infrastructure design, unplanned construction and low public perception of earthquake hazard. •The near complete destruction of township of Meckering in the south-west of Western Australia due to magnitude 6.5M earthquake in 1968; the initial seismic design requirement was introduced in 1979 by the Australian Standard AS 2121–1979 (Standards Australia, 1979). •Further, the next major earthquake loading followed by the deadly 1989 Newcastle earthquake with the AS 1170.4–1993 (McCue et al., 1993; Standards Australia 1993) and referenced by the recently published AS 1170.4–2007 (R2018) (Standards Australia, 2018).

7. • Approximately 70% of the buildings in the country were constructed before to the implementation of the AS 1170.4-1993 earthquake safety standards. This means that many buildings may not be robust enough to withstand earthquakes, posing a significant risk to Australians . •The origin of AS 1170.4–2007 (R2018) is taken from the PSHA of Gaull et al. (1990). This was the landmark research for that time and developed on the available data from the late 1980s and scientific understandings . •Later, McCue et al. (1993) have adjusted the Gaull et al. (1990) work through expert judgment for the design standard AS 1170.4–1993 . •This compiled report in 1991 was not a PSHA but revealed the joint understanding of seismic hazard in Australia at the time, and has guided engineering design since its publication. The 1991 map also by the Australia’s contribution to the Global Seismic Hazard Assessment Program (GSHAP) seismic hazard map.

8. Seismic Source Characterization • Despite these substantial efforts, the NSHA18's new seismic hazard model was not included in the AS 1170.4-2007 (R2018) Standard version. •This decision was made because the new model would have introduced significant modifications to the existing framework for seismic loading actions, changes that stakeholders may have believed were too large to implement without further consideration or additional data. • The NSHA18 built on previous attempts by employing a wide range of seismic-source models (SSMs) to account for uncertainty in earthquake modelling. It incorporated input from foreign contributors as well as Geoscience Australia, resulting in the development of 19 SSMs for Australia and its adjacent seas •Experts assessed these models, which describe how frequently earthquakes occur and how they rupture.

9. (i) National fault-source mode • The NSHA18 has first time introduces a National Fault-Source Model to show long term hazard posed by the known geological structure. This model includes 356 onshore faults and 23 offshore fauls and these are modeled as simplified planes and assigned a general dip direction and dip. •The faults dip are obtained preferentially from 1. seismic-reflection profiles 2. Surface geology and geomorphology 3. Using faults in similar neotectonic settings as a proxy •Slips rate are estimated preferentially from, 1. Displaced strata of known age 2. Surface expression combined with knowledge of landscape modification rates (erosion/deposition) 3. By the proxy of similar neotectonic settings

10. Source model logic tree *NSHA18: 2018 National Seismic Hazard Assessment; NSHM12: 2012 National Seismic Hazard Maps; NFSM: national fault-source model .

11. (ii) Source-model types for Australia • To avoid sharp changes in hazard predictions that can happen with single-model zones and mitigating the effect of these steep hazard gradients. The NSHA18 employed five distinct classes of SSMs to address this issue. • Background area-source models employ wide geographic zones where earthquakes have an equal chance of occurring anywhere. These are usually source models for the NSHA18 that have 20 or fewer area-source zones at the national level –The use of large background source models may need to be reconsidered for eastern Australia where seismicity has been comparatively stationary • Regional area-source models that presume the spatial distribution of seismicity is non uniform at the size of background source models and the historical seismicity distribution is valuable for forecasting future earthquake occurrence. On a national scale, these are often models with 30 or more area-source zones. –The regional source models (Figure 3b) forecast hazard intermediate to these model types and provide some balance to the discrepancies between the two SSM end-members

12. • Smoothed seismicity : Seismic models use a smoothing process to spread historical earthquake rates over an area, predicting future quakes (Frankel, 1995). These rely on the assumption that past earthquake locations are good indicators of future seismic activity. –Smoothed seismicity source models have demonstrated effectiveness in active tectonic regions (ATRs; for example, Helmstetter et al., 2007) where the historical observation period is more likely to have recorded the full seismic cycle. Although this is not the case in slowly deforming regions, Griffin et al. (2017) achieved similar predictive performance of smoothed seismicity models for Australia when compared with those for ATRs. •Seismotectonic models that combine a NFSM with regional source model (Clark et al., 2016). •Smoothed seismicity paired with a national fault-source model. These models integrate long-term geological data from the fault-source model with short-term data from the instrumental catalogue.

13. (iii) Northern Plate margin sources •Large earthquakes in eastern Indonesia and Papua New Guinea (PNG) have the potential to generate significant ground-shaking in northern Australia and may affect communities and infrastructure projects . •Due to the tectonic complexity of the region, and the availability of a number of recent geological, geodetic, and seismological studies, a revised SSM was developed to underpin both NSHA18 and the 2018 revision to the Australian Probabilistic Tsunami Hazard Assessment (PTHA18; Davies and Griffin, 2018 ). •A common plate margin SSM was appended to each of the SSMs developed for continental Australia . •This plate margin source model extends to a distance of more than 800 km from the mainland Australian coastline, based on the GMM integration distance recommended by the expert opinion panel (Griffin et al., 2018 ). •Fault-source models are defined for the major tectonic structures of the region (e.g. the Java Trench, the Timor Trough). Area-source models were then defined for regions not covered by the major fault sources.

14. Ground-motion characterization • Models to estimate the attenuation of PGA and peak ground velocity (PGV) were developed as part of the Gaull et al. (1990) national seismic hazard assessment. These models were based on mean isoseismal radii from well-documented Australian earthquakes (e.g. Everingham et al., 1982). •Due to lack of data at the time in Australia, people used attenuation relationships based on isoseismal radii can significantly overestimate median ground-motion intensities commonly used in modern hazard assessment for Australia. This model was based on the regional geology of Papua New Guinea (PNG). •During development of 1991 earthquake hazard map, it was general perception that Gaull et al. (1990) PGA attenuation relationships underestimated recorded accelerations for Australian earthquakes (McCue, 1993). Consequently, McCue et al. (1993) divided the PGV contours (in mm/s) developed by Gaull et al. (1990) by a factor of 750 to calibrate the PGA contours (in g). After calibration it was found that the effective PGA approximately 30-65% higher at the distance less than 100 Km.

15. Figure 4 . Compares the PGA attenuation with distance for a selection of GMMs as applied in the NSHA18. For comparison, the Gaull et al. (1990) GMMs are shown, together with the adjustments used to calibrate the AS 1170.4–1993 hazard contours (McCue et al., 1993). In general, the Gaull et al. (1990) GMMs attenuate at slower rates than modern GMMs. The southeastern Australian (SEA) model of Gaull et al. (1990), in particular, predicts very high ground-motions for moderate-to-large-magnitude earthquakes at all distances relative to most of the other non-cratonic GMMs used for the NSHA18.

16. • All SSMs were combined as input for the OpenQuake-engine (Allen et al., 2018a). The NSHA18 earthquake catalog (Allen et al., 2018c), homogenized to moment magnitude M w , is used to determine earthquake rates for all of the area and gridded seismicity source models for continental Australia. •‘a’ and ‘b’ values are estimated from GR relation (Gutenberg and Richter, 1944) using the Weichert (1980) maximum likelihood method. •The hazard curves are computed over a 15 km spaced grid for over 54000 sites across the e continent and surrounding region to allow for the generation of a national seismic hazard map for each ground-motion intensity measure. •The hazard curves were interpolated to return hazard grids at 10%, 9.52%, and 2% probability of exceedance in 50 years. These probabilities correspond to ground-motion return periods of 475 years (Figure 5), 500 years, and 2475 years, respectively.

17. •PGA values at the 10% probability of exceedance in 50-year level across Australia have decreased, on average, by 72% relative to the earthquake hazard factors provided for localities in the AS 1170.4–2007 (Standards Australia, 2007). Furthermore, the NSHA18 10% in 50-year PGA values are approximately half of those in the NSHM12 (Leonard et al., 2013, 2014), with an average decrease of 48% at AS 1170.4 localities.

18. Reason for changes of mean hazard • Use of a national fault-source model: New research that looks at the structure of the Earth's crust (seismotectonic SSMs) has shown that certain places, especially those near many young, active faults, might actually face a higher risk than previously thought. For example, cities like Adelaide and Canberra, which are near active faults, could see a bigger increase in their earthquake risk compared to places like Perth, which has fewer and less active faults nearby. Melbourne also has active faults nearby, but the increase in risk is smaller because of the distance from these faults to the city center. • Changes to Gutenberg–Richter b value: to increase the Gutenberg–Richter b value (Gutenberg and Richter, 1944), which in turn decreases the annualized rates of larger, potentially damaging events. • Use of modern GMMs: The final factor driving the reduction of calculated seismic hazard in Australia relative to the 1991 national map of McCue et al. (1993) is the use of modern GMMs

19. Adjustment of local magnitudes: Richter scale tends to overestimate M L relative to modern Australian magnitude formulae (Allen, 2010). M w is approximately 0.3 magnitude units lower than ML for moderate-to-large earthquakes (4.0<M<6.0). The M L corrections and the subsequent conversions to M approximately reduce the number of earthquakes exceeding magnitude 4.5 and 5.0, respectively, by 50% or more

20. Quantifying modeling uncertainty (i) Intrasource-model uncertainty : • For each SSM, earthquake occurrence rates from 15 MFD branches (five Mmax branches and three coupled a- and b-value branches) were collapsed onto a single MFD per source by calculating the weighted mean incremental rate for each magnitude bin in the distribution. •This preserves mean seismic hazard (see Fig. 11) and improves computational efficiency within the OpenQuake-engine. •Darwin demonstrates large withinmodel variability. This variability is likely driven by the large number of active seismic sources in the plate margin source model that affect the city and by the large variability between the selected GMMs used for these sources.

21. (ii) Intersource-model uncertainty : There is comparatively little epistemic uncertainty between the seismic models for Darwin (Figure 8b) suggesting the hazard is driven by the plate margin SSM (Griffin and Davies, 2018) and the variability among the GMM suite for these sources. For the eastern capital cities (e.g. Canberra, Melbourne, and Sydney), the background and regional source model CDFs tend to cluster in their respective model classes, with the regional models typically forecasting larger ground motions than the background models at the 10% probability of exceedance in 50-year level.

22. (iii) Total NSHA18 uncertainty: This section compares the total NSHA18 uncertainty and considers the weighted contributions of all sources of epistemic uncertainty, from the GMM logic tree to the variability among SSMs and their parameterization. Figure 9 shows the PGA ground-motion distributions for the full NSHA18 with a 10% exceedance probability in 50 years for Australian capital cities. Most of these CDFs generally show a log-normal shape, with the mean value relatively close to the median (or 50th percentile). The large tail at low fractiles is likely due to the use of background SSMs within the NSHA18, which do not consider the fault-source model. Background source models typically envelope broad regions and distribute earthquake rates, determined from the historical seismicity, uniformly across the zones. In regions with locally high rates of earthquakes, the use of background SSMs will tend to reduce earthquake occurrence rates relative to regional and smoothed seismicity models

23. Thank You