Analysis of Earthquake Hazard in Papua New Guinea

Analysis of Earthquake Hazard in Papua New Guinea Lawrence Anton Port Moresby Geophysical Observatory Outline Introduction Tectonics Seismology Earthquake hazard Discussions Concluding remarks Recommendations Top right: aerial view of CBD Port Moresby, National Capital Bottom: National Parliament

The aims of the study are to evaluate (and map) earthquake hazard in PNG at different scales, utilizing data and improved methods now available focusing on specific sites of population density and sites of important national industrial activity provide the basis for a much-needed revision of the PNG Earthquake Loading (Building) Code update results in previous studies Global tectonic plate configuration & kinematics

Regional tectonics and related elements PNG region is situated Within the collision zone of the India Australia, Pacific and Eurasian Plates Left-lateral shearing across New Guinea resulting from highly oblique collision (Gochioco et al., 2002) world s fastest continental shear zone (75-80 mm/yr), relative to northern Australia (McCaffrey, 1996) but accounts for only minimum release of seismic moment in the region Ontong Java Plateau dominating the Pacific Plate front thick massive oceanic crust caused break up of Melanesian Arc and reversal of subduction polarity Convergence is accommodated by thrusts and strikeslip faulting along frontal Highlands thrust belt; involved in mountain building (crustal shortening/thickening) spanning the central axis of New Guinea Island the central collisional belt; 300 km wide, 1300 km long & peaks of over 3 km Existence of minor plates within the collision zone (Wallace et al., 2004; this study) Many more being recognised/confirmed; eg. North Bismarck, New Guinea Highlands and Woodlark, amongst many others

TECTONICS Two mountain ranges dominate the region: Terrains of the Melanesian arc subjected to oblique collision in early Pliocene; still involved in active subduction in the Solomon Islands Highlands (Irian Jaya Fold Belt & Papuan Fold Belt) formed by two tectonic events Obduction of Papuan Ophiolites during Oligocene resulting in metamorphism of continental margin sediments Continental-arc collision during Pliocene causing intracontinental deformation Tectonic structure of New Guinea (Abers and McCaffrey, 1988; Abers, 1994)

Geological provinces of New Guinea (Davies, 1990) Geological provinces of SW Pacific region (Audrey-Charles, 1991)

Free-air gravity anomalies of OJP and New Guinea (Mann and Taira, 2004) GPS data from Tregoning et al. (1998)

Tectonic plates configuration of PNG (Ripper and Letz, 1991; 1993; PMGO) Seismicity of PNG based on PMGO earthquake catalogue

SEISMOLOGY Earthquake activity is a manifestation of and delineate Plate boundaries all possible types Subduction zones, continental and oceanic convergence Seafloor spreading centres and continental rifting Continental and oceanic transcurrent faulting Zones of crustal deformation, e.g. due to convergence Crustal fracture zones Volcanic centres Significant earthquakes are frequent, including six having significant tectonic effects; 1907, 1935, 1941, 1993, 2000, 2002, 2007 There may have been numerous others Seismicity of the PNG region based on ISC earthquake catalogue showing regional earthquake distribution Red circles denote shallow events (0-34 km) Yellow diamonds and blue triangles denote intermediate depth events (35-299 km) Green stars denote deep events (>300 km)

Significant earthquakes of PNG Great Earthquakes (Pacheco and Sykes, 1992; Engdahl et al., 1998 and ISC) Year Date Time Lat Long Depth Ms Moment Mw hr:min S E Km 10 20 Nm. 1906 14 Sep 16:04 7.00 149.00 S 7.4 12.70 8.0 1935 20 Sep 01:46 3.50 141.75 S 7.9 14.50 8.0 1971 14 Jul 06:11 5.50 153.90 53 7.8 12.00 8.0 1971 26 Jul 01:23 4.90 153.20 48 7.7 18.00 8.1 2000 16 Nov 04:54 3.98 152.17 33 8.2 12.40 8.0 2007 01 April 20:39 8.466 157.043 24 7.9 8.1 Significant earthquakes with tectonic effects (PMGO) Year Date Place Mw Effect. 1907 15 Dec North coast New Guinea 7.3 subsidence 1935 20 Sept Torricelli Mtn/N New Guinea 8.0 uplift 1941 13 Jan New Britain 7.0 horizontal displacement 1970 31 Oct Adelbert Range 7.0 Slumping - tsunami 1993 25 Oct Eastern New Guinea 7.1 vertical displacement 1998 17 Jul Northern New Guinea 7.1 submarine slumping - tsunami 2000 16 Nov Southern New Ireland 8.0 5.5m horizontal displacement 2002 08 Sept North coast New Guinea 7.8 30-40cm uplift tsunami 2007 01 April New Georgia Group, SI 8.1 1 metre uplift Coastal uplift during Eq of 1 April 07

Earthquake activity Amongst the most intense in the world Frequent magnitude 7 and above earthquakes about 2 per year High stress release; Low stress accumulation However, locations are poor; poor seismic network coverage Earthquakes smaller than magnitude 4 not located Locations have huge errors Active plate interaction Indicative of rapidly evolving of Plate boundaries Inter-slab and intra-slab activity Diverse earthquake wave propagation path Poor local seismograph coverage So most earthquakes are located using global seismograph data Locations and depths include very large errors Future great eqs are imminent & of concern Seismic instrumentation Proper monitoring started in the 1962 by USGS Local networks of soft and strong motion began about the same time; but have currently ceased due to lack of funding commitment Replacement networks an urgent need; For public information and early warning tsunami warning including For hazards mitigation and awareness risk management For proper hazard assessment for engineering use For proper identification of source zones as a result of improved location and magnitude determinations To improve earthquake catalogs which will benefit local and regional seismicity and tectonic studies; including hazard studies

EARTHQUAKE HAZARD After an understanding of the regional seismicity and tectonics having gained, the second stage for an earthquake hazard study is largely concerned with calculation of ground motion recurrence, representing the hazard due to shaking by seismic waves. other earthquake hazards such as liquefaction, surface rupture, landslides, and tsunami, are treated separately. Two main models required to compute ground motion recurrence are a seismotectonic model that specifies the assumed distribution of earthquake magnitude recurrence of earthquakes, and a ground motion model that specifies the expected ground motion from the earthquakes. Available data Compiled using databases established based on ISC, USGS, and sourced from many workers; of Hypocentres this study Intensities future Strong motion future component of the project Seismicity maps are now possible; therefore, Identification of seismic source zones made easier Hazard mapping was attempted in previous years Proper data and methods not available then Revise existing seismic hazard maps building code seismic zone

PNG Earthquake Loading Code The current PNG Earthquake (Building) Code does not reflect current knowledge of the tectonic structure of PNG When the review is completed, proper legislation is called for to include the revisions. Rapid national growth requires urgent legislation to ensure compliance by town planners and engineers. The Code was developed in the 1970s using data from abroad, as PNG didn t have the data then. PNG Statutory Instrument No.44 of 1971 (1971), Regulations made under the Building Ordinance 1971, documented by Papua New Guinea Government Printer, November, 1971 (plus Amendments). First revised in 1982 and documented by PNG National Standards Council (1983), Code of practice for general structural design and design loadings for buildings, Part 4, Earthquake loadings. Revisions were not taken onboard Method The hazard is represented by uniform probability response spectra computed using the 4-staged Cornell method 1. Develop seismotectonic map 2. Quantification of seismicity 3. Determine attenuation function 4. Earthquake hazard, and hazard mapping Based on the comprehensive PNG earthquake catalogue The seismotectonic model was developed, and ground motion recurrence for selected sites computed After several iterations of the model, earthquake hazard maps are produced

1. Develop seismotectonic map Divided the region in to seismotectonic zones based on: Existing earthquake catalogue; the reformat of which is based on ISC, PMGO, USGS databases Checked against regional geology, and geophysics, especially gravity and magnetics, Quaternary faults, topographic and geographic features There must be prior knowledge of the tectonics Models used influence hazard assessments Tectonics

Collision zone Seismotectonics uniform earthquake distribution Six layers of the model PNG1 A total of 120 source zones PNG region is classed as low to moderate resolution; Australia as low

2. Quantification of seismicity For each source zone (as well as faults future work) Based on available data distribution patterns defining activity in zones, including sources at depth Determine per zone: Rate of recurrence of earthquakes varying with magnitude (magnitude recurrence) Relative proportion of small to large earthquakes (bvalue) Maximum earthquake magnitude (Maximum Credible Earthquake); tectonic settings considered too! Example of source zone quantification: New Britain Arc A typical source zone, the New Britain Arc, is presented as an example of quantification of a source zone The earthquakes within the zone were extracted from the catalogue (using the MapInfo GIS system) The catalogue had previously been declustered independent mainshocks distinguished from dependent foreshocks and aftershocks the declustered listing was used for earthquake magnitude recurrence estimates

Figure shows magnitude-time plot for earthquakes in the New Britain Arc zone, and the catalogue completeness is estimated (considering seismograph coverage, and linearity of the earthquake magnitude recurrence plot) and represented by the blue line. The plot shows magnitude against time for the known declustered earthquakes, with dependent foreshocks and aftershocks removed. The earthquake magnitude recurrence plot for the New Britain Arc. This is a plot of Nx, the number of earthquakes per year equal to or larger than magnitude x, against magnitude x. Nx values indicating the number of earthquakes in entire source region per year. An alternative measure of earthquake activity is Ax, which is the number of earthquakes per year per 100 x 100 km. The Nx values depend on source region size so can't be compared (large zones have more earthquakes than small zones), but Ax values can be compared between zones. The gradient of this plot gives the b-value (a measure of the proportion of small to large earthquakes) for the zone, which in this case is a very high value of 1.386.

Table lists earthquake magnitude recurrence for the New Britain Arc zone. These estimate of earthquake magnitude recurrence for the New Britain Arc zone of model PNG1 used the earthquake catalogue to 2008-09-30. The zone covers 65,730 km2. The gradient is represented by beta = 3.19, which corresponds to b-value = 1.39. Events/year Events/year Ret Period Ret Period Whole zone /100x100km /100x100km (yr) for zone (years) No = 7952015.4 Ao = 1209800 0.00 0.000 N1 = 326796 A1 = 49718 0.00 0.000 N2 = 13430 A2 = 2043.21 0.00 0.000 N3 = 551.921 A3 = 83.968 0.01 0.002 N4 = 22.6818 A4 = 3.45075 0.29 0.044 N5 = 0.932132 A5 = 0.141812 7.05 1.073 N6 = 0.0383069 A6 = 0.00582792 171.59 26.105 N7 = 0.00157426 A7 = 0.0002395 4175.30 635.219 This process was repeated for every source zone for the seismotectonic model PNG1. At this stage the model PNG1 has a total of 120 zones in six depth ranges. 3. Determine attenuation function Gives earthquake ground motion as a function of magnitude, distance and other parameters No one attenuation relationship is possible for the entire region due to: Complex geology resulting in varying seismic energy propagation path Numerous crustal tectonic blocks Diverse tectonic structure Terrains and hilly topography As local data is not sufficient, relationship of similar tectonic environments was used Atkinson and Boore (2003) for subduction zones Chiou and Youngs (2008) for crustal, intra-plate zones Attenuation relationships to be developed in next stage of the project future work

4. Earthquake hazard Computed spectral ground motion recurrence (SGMR), integrating probabilities of motion from all earthquakes in space, magnitude and freq of motion using commercial software EZ-FRISK (McGuire, 1993) Computed SGMR at specific sites/points depending on complexity of seismotectonic and attenuation models; on bedrock as strong motion data lacking After several iterations of the model, earthquake hazard maps will be produced Repeating the process at many points on a grid covering the region, for better resolution Sites of Port Moresby, Lae, Kokopo, Kimbe, Buka, Madang, Wewak and Honiara have been attempted 1. Port Moresby Source zone contributions for Port Moresby ground motion. This plot gives contributions for peak ground acceleration.

PGA recurrence for Port Moresby. This is for bedrock motion, considering magnitudes 5.0 and higher. Figure shows that the high-frequency peak ground acceleration at Port Moresby is dominated by hazard contributions from shallow or crustal seismic zones using the attenuation of Chiou- Youngs (2008), but higher accelerations may be originating from the intra-slab zones, using the attenuation function of Atkinson-Boore (2003). Figure shows the uniform probability response spectra at Port Moresby for return periods of 475 years, 3,000 years and 10,000 years, for bedrock, and using magnitudes from 5.0. The peak ground acceleration is numerically equal to the response spectral acceleration at near zero period, for all the left-most points on each of the plots. The spectra for other return periods can be found by re-calculation, or approximated by interpolation.

Figure shows the deaggregation plot for 1.0 second period motion at a return period of 975 years for Port Moresby (corresponding to a relatively low amplitude of 0.07 g spectral acceleration). It shows three main source of hazard, including moderate magnitude nearby events, larger events at distances of 200 to 500 km, and some contribution from great earthquakes at distances of 500 km and beyond (cumulatively plotted at 500 km). Magnitude 6 earthquakes near Port Moresby occur infrequently, while the maximum credible magnitude of 7.3 has been determined, with a near-zero recurrence rate. 2. Kokopo Figure shows source zone contributions for Kokopo ground motion. This plot gives contributions for motion with spectral acceleration at a period of 1.0 seconds. Zones contributing most to the hazard are those closest to Kokopo and the most active are the zones of the New Britain Arc and New Britain Trench. These zones consequently contribute highest to the total hazard, or otherwise is spread amongst many other zones.

Note that for the total motion at the site, the Atkinson-Boore attenuation functions for subduction events are each weighted by 0.5, while the plot shows the ground motion recurrence for full weighting for each. Figure shows the strength of seismic zones of the subduction zone in the total hazard using the attenuation function by Atkinson and Boore (2003), especially in the lower and upper PGA. There was a reasonable contribution from zones non-subducting lithosphere using the attenuation function by Chiou and Youngs (2008), especially at the mid-range PGAs. Figure shows the response spectra for Kokopo. These uniform probability response spectra are for bedrock, using magnitudes from 5.0. Figure shows response curves for 475, 3,000 and 10,000 year return periods. Spectra for other return periods can be derived from re-calculation, or approximated by interpolation.

Figure shows the deaggregation for all source zones within 500 km, with associated hazard parameters. Two parts are shown to contribute earthquakes, from very local distance (0-25 km) and that from distance up to 150 km. No seismic zones is observed to contribute hazard at Kokopo from farther distances, even in adjoining subduction zones are observed. Maximum magnitude earthquake in Kokopo is 7.6 and maximum acceleration of 0.5g Figure magnitude-distance deaggregation for motion at Kokopo. Results for bedrock motion of period 1.0 seconds, considering magnitudes 5.0 and higher. 3. Lae Figure shows source zone contributions for Lae ground motion. This plot gives contributions for peak ground acceleration. It shows that the source contributions in Lae are dominated by the Huon seismotectonic zone within which the city is located, and the neighbouring Schrader and Adelbert zones, as well as Huon Peninsula at depth.

Figure shows PGA recurrence for Lae. Results for bedrock motion and peak ground acceleration, considering magnitudes 5.0 and higher. Figure shows that the high-frequency peak ground acceleration at Lae is totally dominated by contributions from nearby local earthquakes, determined using the Atkinson-Boore (2003) attenuation function. Figure shows response spectra for Lae. These uniform probability response spectra are for bedrock, using magnitudes from 5.0. The peak ground acceleration is numerically equal to the response spectral acceleration at near zero period, for all the left most points on each of the plots. The uniform probability response spectra for return periods of 475 years, 3000 years and 10,000 years are shown in the Figure. Spectra for other return periods can be found by recalculation.

For Lae, the deaggregation plot of 1.0 second for a return period of 975 years shows that most of the contributing earthquakes are originating locally, but the situation extends gradually and tapers off to a distance of about 210 km. Maximum acceleration of 0.35g and MCE of 7.1, at a mean distance of 71 km Outcomes The earthquake source zones are quantified using historical and recent seismicity data, and checked against geology, geophysics and geodesy (GPS deformation) during the current tectonic regime Attenuation functions that were derived using data from comparable tectonic environments were selected No check for consistency with the few existing PNG strong motion data, and isoseismal data A component of the next stage of the study Ground motion recurrence computations were then performed for selected locations

PGAs for 475 yr Return Period Sites PGA Max Cred Eq Port Moresby 0.07 7.3 Lae* 0.35 7.1 Kokopo 0.50 7.6 Madang 0.28 7.4 Wewak 0.24 6.9 Kimbe 0.32 7.6 Buka 0.36 7.7 Honiara 0.31 7.5 Mean dist (km) 254 71 60 66 41 83 91 64 CONCLUDING REMARKS PNG a geologically, seismically and geographically complex region Tectonic structure (debated) redefined and updated continuously Data and methods for earthquake hazard determination are now available; could have better data It was determined that earthquake hazard is significant As local strong motion data is not sufficient, that of similar tectonic setting could have been used data required to determine attenuation function There is a need for modern seismic equipment Replace the almost non-existing local network Monitoring, parameter determination and hazard assessment Build earthquake catalog for future hazard updates

Expected outcomes in the near future Following completion of the earthquake catalogue, computation of earthquake hazard was undertaken for representative locations within two years, and is anticipated that these will lead to revise hazard maps in the following two years This work will lead to the development of a set of modern earthquake hazard maps of the PNG region and the determination of earthquake response spectra at other selected sites The earthquake hazard maps and response spectra would form the basis for a revision of the PNG Earthquake (Building) Code RECOMMENDATIONS Things proposed to be achieved for the purpose of realising in full the value of the study. These include: (1) Determine earthquake hazard at other additional sites, to improve the resolution and therefore better hazard mapping (2) Improve on the seismotectonic model developed and hazard maps covering the whole geographic region (3) Immediate use of the hazard maps to facilitate the replacement of the existing earthquake building code (4) Replace seismic station network to improve data acquisition required which will in turn improve earthquake hazard analysis, and be able to sustain maintenance of the network (5) To improved hazard analysis; delineate active faults (6) Develop plans for the future updates of the hazard map (7) Acquire EZ-FRISK or similar tools for immediate use, and for future earthquake hazard updates

FUTURE WORK Will include: Determination of PNG earthquakes (epicenters, depths, magnitudes and mechanisms) using a local seismograph network to reduce the uncertainty and current scatter in epicentre and depth estimates; and hence allow delineation of active faults. Update and increase resolution of the seismotectonic model, particularly further reiteration of the analysis process to include more sites, and by computing the hazard contributions by specific active faults rather than assuming uniform area source zones. THANKS