Fundamental and higher-mode Rayleigh wave characteristicsof ambient seismic noise in New ZealandLaura A. Brooks,1John Townend,1Peter Gerstoft,2Stephen Bannister,3and Lionel Carter4Received 6 August 2009; revised 19 October 2009; accepted 26 October 2009; published 2 December 2009.[1] In order to use ambient seismic noise for mappingEarth’s structure, it is important to understand the spatio-temporal characteristics of the noise field. This study usesdata collected during four austral winter months of 2002 toinvestigate New Zealand’s ambient seismic noise field in thedouble-ocean-wave-frequency range (0.1–0.3 Hz). It isshown via beamforming analysis that there are two distinctdispersive waves in the data. These waves can be separated.Their estimated phase velocities (2.5–2 and 4– 3 km/s inthe frequency range 0.14–0.25 Hz) match well withfundamental and higher-mode Rayleigh dispersion curves.Studies of double-wave-frequency microseisms elsewheregenerally show the Rayleigh noise fields to be dominated byfundamental mode waves. The reason why higher-modesignals are observed here may reflect a combination oflong ocean wave periods, large waveheights, the directdeep water approach to narrow continental margins, andthe proximity of the seism ograph array to the source regions.Citation: Brooks, L. A., J. Townend, P. Gerstoft, S. Bannister,and L. Carter (2009), Fundamental and higher-mode Rayleighwave characteristics of ambient seismic noise in New Zealand,Geophys. Res. Lett., 36, L23303, doi:10.1029/2009GL040434.1. Introduction[2] The ambient seismic noise field is largely dominatedby signals with frequencies of <1 Hz that correspond toRayleigh waves produced by nonlinear ocean waveprocesses [Bromirski and Duennebier,2002].Themicro-seism spectrum typically exhibits a small peak at 0.06–0.07 Hzand a larger peak at 0.12–0.15 Hz, termed the single frequency(SF) and double frequency (DF) microseism peaks, respectively[Webb,1998].TheSFpeakisgenerallythoughttobegeneratedby direct ocean wave-induced pressure fluctuations at the seafloor, the amplitudes of which decrease with ocean depth[Bromirski and Duennebier , 2002]. The DF peak occursdue to non-linear interaction of opposing wavefields of similarwavenumbers [Longuet-Higgins, 1950], which creates anexcitation pulse at twice the ocean wave frequency. Thispropagates almost unattenuated to the seafloor and couples intoaRayleighwave.Theminimalattenuationinthewatercolumnmeans that DF signals represent both shallow and deep waterphenomena, though DF signals recorded on land are usuallydominated by shallow water excitation [Tanimoto,2007].[3] New Zealand’s geographic isolation and !15,000 km-long coastline expose it to a particularly energetic ocean,producing a high-amplitude seismic noise field [Pickrilland Mitchell, 1979; Gorman et al., 2003], and severalNew Zealand noise studies relating ambient noise spectra/amplitudes to wave spectra/heights have been conductedpreviously [Kibb lewhite and Ewans,1985;Tindle andMurphy, 1999]. The energetic noise field makes the NewZealan d regio n a suita ble locati on for usi ng ambientseismic noise for seismic tomography studies [Lin et al.,2007]; however, the long coastline and complex oceanregime that generate this noise fiel d also result in morecomplex spatial and temporal noise distributions than maybe observed in many other regions (e.g., continental USA).To further increase the accuracy of geophysical estimatesusing ambient noise in New Zealand, a greater understand-ing of the spatio-temporal noise characteristics is needed(Y. Behr et al., Shear-velocity structure of the NorthlandPeninsula, New Zealand, inferred from ambient noise corre-lations, submitted to Journal of Geophysical Research, 2009).[4] Results from beamforming of microseismic data usingseismic arrays have been reported elsewhere [e.g., Haubrichand McCamy, 1969; Chevrot et al., 2007; Gerstoft et al.,2008], but not in New Zealand. Here we employ frequency-domain beamforming of vertical-co mpone nt data from aseismograph array located in the Taranaki region, westernNorth Island (see Figure 1), to determine the locations ofmicroseism generation. We compare our results to shallowwater wave heights (supplied by the National Institute ofWater and Atmospheric Research from their wave predictionmodel, NIWAM). Two distinctive dispersive waves, namelyfundamental and higher-mode Rayleigh waves, are observedin the beamformed data. This result differs from those ofother studies of double-wave-frequency microseisms, whichhave shown the noise field to be dominated by a single mode,namely fundamental mode Rayleigh waves [Lacoss et al.,1969; Tanimoto and Alvizuri, 2006]. The dominant sourceregions of the two signals we observe are interpreted withreference to New Zealand’s oceanographic conditions.2. Data Processing[5] The seismic data analyzed here were recorded in theaustral winter months of May–August, 2002, on the vertical-components of a 61-el ement broadband three-componentseismograph array located in the Taranaki region (Figure 1)[Sherburn and White, 2005]. Useful data were obtained from50–58 seismographs at any one time.[6] Day-long seismic traces recorded at 5 Hz were band-pass filtered to 0.02 –0.4 Hz and downsampled to 1 Hz. TheGEOPHYSICAL RESEARCH LETTERS, VOL. 36, L23303, doi:10.1029/2009GL040434, 2009ClickHereforFullArticle1School of Geography, Environment and Earth Sciences, VictoriaUniversity of Wellington, Wellington, New Zealand.2Scripps Institution of Oceanography, University of California, SanDiego, La Jolla, California, USA.3GNS Science, Lower Hutt, New Zealand.4Antarctic Research Centre, Victoria University of Wellington, Wellington,New Zealand.Copyright 2009 by the American Geophysical Union.0094-8276/09/2009GL040434$05.00L23303 1 of 5data were separated into 3-hour long segments and clipped tohalf their standard deviation. The data were split into 1024-slong time series and Fourier transformed, giving a collectionof short-time transforms across the array. In order to retainonly the phase, the signals within each frequency band werenormalized by their amplitudes. The frequency and timedomain normalizations reduce the effect of episodic pro-cesses [Gerstoft et al., 2008]. All stations had the samenominal response and hence station response correctionswere not made.[7]Planewavefrequencydomainbeamformingwasimplemented following the methodology of Johnson andDudgeon [1993]:[8] 1. We first calculated the vec