A global approach to provide magnitude estimates for earthquake early warning alerts

Abstract

[1] We examine five different methods to estimate an earthquake's magnitude using only P wave data for use in earthquake early warning systems. We test two input parameters: the maximum predominant period of the P wave (τ p max ) and the displacement amplitude of the P wave's vertical component (Pd). We apply our algorithms to 174 earthquakes 3.0 < M < 8.0 from California and Japan that have also been used in previous calibration studies. We also apply them to 1992 0.2 < M < 5.7 earthquakes that were processed by the real-time Earthquake Alarm Systems in California. We find that τ p max does not scale with magnitude for small earthquakes (M < 3) and is less accurate for largemagnitude earthquakes than using Pd alone. We derive a global scaling relation between Pd and magnitude and conclude that this global relationship provides the most accurate and robust magnitude estimate. This relationship could be applied in earthquake source zones around the world. Citation: Kuyuk, H. S., and R. M. Allen (2013), A global approach to provide magnitude estimates for earthquake early warning alerts, Geophys. Res. Lett., 40,[6329][6330][6331][6332][6333]

How “good” are real‐time ground motion predictions from Earthquake Early Warning systems?

“…We use the regression relation of Kuyuk and Allen [] between peak‐observed displacement at individual stations, P d , i ( t ), to estimate a probability density function for magnitude P d , i ( t ) is the peak absolute displacement amplitude in centimeters measured over the first 4 s since the P wave arrival at the i th station or over whatever waveform segment is available at point in time t . If the direct S phase arrives within 4 s of the P onset, we only measure P d , i ( t ) up to the S arrival minus 0.5 s. We use σ 2 = 0.31 2 from Kuyuk and Allen [], and we assume all magnitudes to be moment magnitudes throughout this paper. is a probabilistic source/station distance estimate.…”

“…For each station at which the earthquake has been registered we make a magnitude estimate based on initial waveform observations. We use the regression relation of Kuyuk and Allen [2013] between peak-observed displacement at individual stations, P d,i (t), to estimate a probability density function for magnitude pm i ; t ð Þ:…”

“…Þþ5:39 þ Ν 0; σ 2 À Á : P d,i (t) is the peak absolute displacement amplitude in centimeters measured over the first 4 s since the P wave arrival at the ith station or over whatever waveform segment is available at point in time t. If the direct S phase arrives within 4 s of the P onset, we only measure P d,i (t) up to the S arrival minus 0.5 s. We use σ 2 = 0.31 2 from Kuyuk and Allen [2013], and we assume all magnitudes to be moment magnitudes throughout this paper. p log 10 b…”

Feasibility and Interest in Earthquake Early Warning Systems for Areas of Moderate Seismicity: Case Study for the Pyrenees

“…Thus, from a statistical standpoint, a look at the coefficient of determination R 2 from Table 2 directly shows that the most satisfactory proxies for the Pyrenees are, in order, P 10 d and P 10 v (which are very similar by nature and cannot be considered as independent parameters), the frequency parameter s c , and lastly the frequency parameter s max p . Moreover, it should be noted that, as underlined by KUYUK and ALLEN (2013b), the correlation observed in Fig. 6 between s max p and M \ 3 is possible because the high frequencies intrinsic to smaller earthquakes have been preserved in our analysis, while the 3 Hz low-pass filter initially proposed by ALLEN and KANAMORI (2003) for the calculation of s max p removes them.…”

“…When considering regional EEWS, this problem amounts to quickly assessing the magnitude of earthquakes from the first seconds of the waveforms. While many authors have demonstrated the ability to rapidly assess magnitudes [4 and up to 6 or 7 thanks to the first seconds of the P-waves, and they have established either regional (e.g., LANCIERI et al 2011) or global scaling relations (e.g., ZOLLO et al 2010), very few studies have considered lower magnitude events (e.g., HEIDARI et al 2013;KUYUK and ALLEN 2013b). However, while particular attention must be paid to the largest earthquakes (probably of the order of 6.5 in the Pyrenees), which are liable to produce the greatest damage, smaller earthquakes should also be considered when working in AMS so as to be certain to be able to distinguish them from larger earthquakes during the real-time analysis.…”