Historical tsunami earthquakes in the Southwest Pacific: an extension to Δ > 80° of the energy-to-moment parameter Θ

Geophysical Journal International Geophys. J. Int. (2017) 210, 852–873 Advance Access publication 2017 May 16 GJI Seismology doi: 10.1093/gji/ggx197 Historical tsunami earthquakes in the Southwest Pacific: an extension to  > 80◦ of the energy-to-moment parameter  Emile A. Okal and Nooshin Saloor Department of Earth & Planetary Sciences, Northwestern University, Evanston, IL 60208, USA. E-mail: Accepted 2017 May 8. Received 2017 May 1; in original form 2016 October 18 Key words: New Zealand; Pacific Ocean; Earthquake source observations; Tsunami warning. (assuming the earthquake remains shallower than 80 km), and defined a slowness parameter: 1 I N T RO D U C T I O N This paper examines quantitatively three historical ‘tsunami earthquakes’ in the Southwest Pacific. We recall that this class of events was defined by Kanamori (1972) as earthquakes whose tsunamis are significantly larger than expected from their seismic magnitudes, especially classical ones; charter examples included the famous 1896 Sanriku and 1946 Aleutian earthquakes. Such events obviously pose enormous challenges, since tsunami warning remains largely based on an assessment of the parent earthquake (e.g. Okal 2008). While several models have been proposed to explain the occurrence of tsunami earthquakes in various environments (Fukao 1979; Tanioka et al. 1997; Bilek & Lay 2002), the systematics of their occurrence at any given subduction zone remain elusive. In this context, and because tsunami earthquakes are relatively rare, it is crucial to investigate as quantitatively as possible those events predating the development of digital networks. In the present paper, we extend to distances  > 80◦ the computation of the Energy-to-Moment parameter , introduced by Newman & Okal (1998) and used as a robust discriminant to characterize source slowness, notably during tsunami earthquakes. Following the work of Boatwright & Choy (1986), Newman & Okal (1998) developed an estimate EE of the seismic energy radiated by an earthquake into its teleseismic body waves, not requiring the precise knowledge of focal mechanism and source depth 852  C  = log10 EE , M0 (1) where M0 is the seismic moment of the source. Under seismic scaling laws,  should remain constant, its theoretical value being −4.90, but earthquakes featuring an anomalous source spectrum can have excessive or deficient  values, by as much as 2 logarithmic units, the latter being the case of tsunami earthquakes. In their original study, Newman & Okal (1998) had shown that three tsunami earthquakes (Nicaragua, 1992; Java, 1994; and Chimbote, Peru, 1996) postdating Kanamori’s (1972) study all featured deficient values of , in the −5.8 to −6.3 range. The computation of  was later implemented as part of routine procedures at a number of tsunami warning centres (e.g. Weinstein & Okal 2005). In order to allow a proper, theoretically justifiable, implementation of a distance correction into the  algorithm, Newman & Okal (1998) originally restricted its use to the window 25◦ <  < 90◦ . In later studies (e.g. Okal & Newman 2001; Weinstein & Okal 2005; Okal 2013), we used a narrower range of distances (35◦ <  < 80◦ ), made possible by the abundance of digital stations deployed in recent years. At shorter distances, this guards against the effects of the triplications resulting from mantle discontinuities, and at greater ones, against complexities due to reflections such as PcP, and more The Authors 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society. SUMMARY We extend to distances beyond 80◦ the computation of the energy-to-moment slowness parameter  introduced by Newman and Okal, by defining a regional empirical correction based on recordings at distant stations for events otherwise routinely studied. In turn, this procedure allows the study of earthquakes in a similar source-station geometry, but for which the only available data are located beyond the original distance threshold, notably in the case of historical earthquakes predating the development of dense networks of short-period seismometers. This methodology is applied to the twin 1947 earthquakes off the Hikurangi coast of New Zealand for which we confirm slowness parameters characteristic of tsunami earthquakes. In addition, we identify as such the large aftershock of 1934 July 21 in the Santa Cruz Islands, which took place in the immediate vicinity of the more recent 2013 shock, which also qualifies as a tsunami earthquake. In that subduction zone, the systematic compilation of  for both recent and pre-digital events shows a diversity in slowness correlating with local tectonic regimes controlled by the subduction of fossil structures. Our methodology is also well adapted to the case of analogue records of large earthquakes for which short-period seismograms at conventional distances are often off-scale. Tsunami earthquakes in the Southwest Pacific 25 MAR 1947 TUCSON (b) 17 MAY 1947 PASADENA (c) 21 JUL 1934 PASADENA Figure 1. Short-period P-wave seismograms used in this study. Time marks are minutes, uncorrected for clock errors. The durations of the seismograms are 106 s (a), 179 s (b), and 90 s (c). On (a), the high-frequency signal recorded half-an-hour later is a local shock, unrelated to the New Zealand earthquake. generally to the interaction of the generalized P wave with the D boundary layer, known to feature considerable lateral heterogeneity (e.g. Garnero & Helmberger 1996), even before the initiation of genuine diffraction by the core–mantle boundary around 102◦ . In a previous contribution, Ebeling & Okal (2012) used large digital data sets to define an empirical correction allowing the extension of  to distances as short as 5◦ ; in the context of tsunami warning in the regional field, these authors were motivated by the desire to obtain information on potential source slowness as soon as possible following the event, and hence from stations located as close as possible to the source. Our motivation in the present paper is different. We have shown in a number of previous studies that the  concept can be successfully applied to historical events, helping define or confirm the anomalous behaviour of both slow tsunami earthquakes such as the Mexican aftershock of 1932 June 22 ( = −6.18) and the Aleutian event of 1946 April 1 ( = −7.03) (López & Okal 2006; Okal & Borrero 2011), and fast, ‘snappy’ events, such as the Chillán shock of 1939 January 25 ( = −4.04) and the great Showa Sanriku earthquake of 1933 March 2 ( = −4.24) (Okal & Kirby 2002; Okal et al. 2016). However, such investigations must rely on short-period records of body-wave arrivals offering adequate and documented response in the relevant frequency range (typically 0.1 to 2 Hz). While torsion seismometers (Anderson & Wood 1925) can occasionally provide adequate records for historical events, those instruments were typically low-gain, and they (...truncated)