Toward Global Earthquake Early Warning with the MyShake Smartphone Seismic Network, Part 1: Simulation Platform and Detection Algorithm

Cite this article as Kong, Q., R. MartinShort, and R. M. Allen (2020). Toward Global Earthquake Early Warning with the MyShake Smartphone Seismic Network, Part 1: Simulation Platform and Detection Algorithm, Seismol. Res. Lett. 91, 2206–2217, doi: 10.1785/0220190177. Supplemental Material The MyShake project aims to build a global smartphone seismic network to facilitate large-scale earthquake early warning and other applications by leveraging the power of crowdsourcing. The MyShake mobile application first detects earthquake shaking on a single phone. The earthquake is then confirmed on the MyShake servers using a “network detection” algorithm that is activated by multiple single-phone detections. In this part one of the two article series, we present a simulation platform and a network detection algorithm to test earthquake scenarios at various locations around the world. The proposed network detection algorithm is built on the classic density-based spatial clustering of applications with noise spatial clustering algorithm, with modifications to take temporal characteristics into account and the association of new triggers. We test our network detection algorithm using real data recorded by MyShake users during the 4 January 2018 M 4.4 Berkeley and the 10 June 2016 M 5.2 Borrego Springs earthquakes to demonstrate the system’s utility. In order to test the entire detection procedure and to understand the first order performance of MyShake in various locations around the world representing different population and tectonic characteristics, we then present a software platform that can simulate earthquake triggers in hypothetical MyShake networks. Part two of this paper series explores our MyShake early warning simulation performance in selected regions around the world. Introduction Earthquake early warning (EEW) is a technology that uses networks of seismometers to quickly determine the location and magnitude of an earthquake after it has begun and issues warnings to regions anticipated to experience shaking (e.g., Kanamori, 2007; Allen et al., 2009; Allen and Melgar, 2019). Such alerts are typically sent within seconds of the earthquake origin time and can provide up to several minutes of warning depending on the geometry of the monitoring network and the distance between the event and population centers (Allen, 2011, 2013). During this warning time, actions can be taken by individuals and organizations that could potentially save lives and mitigate damage (Strauss and Allen, 2016). To be effective, EEW requires the existence of a dense seismic network that has the capability of real-time monitoring of potential earthquake signals. The closer the instruments are to the epicenter, the faster the detection, and hence the larger the warning times can be. EEW has been mainly developed using traditional seismic and geodetic networks, which are costly to operate and only exist within a small number of countries (Allen and Melgar, 2019). Much of the global population at high risk from earthquake damage thus currently is not benefitting from EEW. Many alternative, cheaper, nontraditional networks have been proposed, including microelectromechanical system accelerometers installed in buildings, Universal Serial Bus (USB) accelerometers attached to personal computers or other low-cost sensory equipment such as the Quake Catcher network, community seismic network, P-alert, and Raspberry Shake (Cochran et al., 2009; Luetgert et al., 2009; Chung et al., 2011; Clayton et al., 2015; Wu, 2015; Wu et al., 2016; Nugent, 2018). Although promising, these ideas suffer from the same disadvantages as traditional networks in that they require physical installation and maintenance by the network 1. Berkeley Seismology Laboratory, University of California, Berkeley, Berkeley, California, U.S.A. *Corresponding author: kongqk@berkeley.edu © Seismological Society of America 2206 Seismological Research Letters www.srl-online.org • Volume 91 • Number 4 • July 2020 Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/91/4/2206/5082113/srl-2019177.1.pdf by rallen on 29 July 2020 operators, which hampers the sustainability and expandability of the EEW system, especially in remote regions. Recent advances in mobile accelerometer technology mean that smartphones are becoming a viable alternative to fixed seismometers as the primary sensing instruments for EEW (Faulkner et al., 2011; Dashti et al., 2012; Finazzi, 2016; Kong, Allen, Schreier, and Kwon, 2016). Furthermore, there is also interest in the development of the smartphone networks that use Global Positioning System and users’ mobile application launching times to detect earthquakes (Minson et al., 2015; Bossu et al., 2018; Steed et al., 2019). There are many advantages of using smartphone networks for this application: The devices are globally ubiquitous, even in regions without traditional earthquake monitoring. Because the hardware is maintained by the users, the only requirement for the network operators is to develop and market a software application that can be made accessible via the Google Play or iOS store, and then to maintain a cloud server to collect data. This makes the network easier to maintain and grow. However, the use of smartphones for EEW is not without its challenges. Namely, the detection software must be capable of reliably distinguishing between earthquake shaking and all other vibrations that the device might experience. Furthermore, the noise floor of mobile accelerometers is significantly higher than that of traditional seismometers, the extent of coupling between the smartphone and the ground may be poor, and the recording of earthquakes is not a priority for users. MyShake is a smartphone application developed at the University of California, Berkeley Seismology Lab to monitor smartphone accelerometer data and detect earthquakes. It uses an artificial neural network (ANN) trained on examples of earthquake and nonearthquake waveforms and is able to successfully distinguish earthquake motions from human activityrelated motion recorded by the phone (Kong, Allen, Schreier, and Kwon, 2016; Kong, Inbal, et al., 2019). The MyShake application monitors the accelerometer on the device and sends real-time messages containing time, location, and ground acceleration data to a server when earthquake-like motions are detected. Kong, Allen, Schreier, and Kwon (2016) and Kong, Inbal, et al. (2019) should be consulted for a complete description of the MyShake application and its operation. Since the app’s first public release in February 2016, MyShake phones have successfully recorded over 900 earthquakes worldwide; the app has approximately 300,000 downloads and 40,000 active users, with approximately 6000 devices making data contributions daily. The data recorded by MyShake has potential uses for various applications such as mapping ground motion (Kong, Allen, and Schreier, 2016), routine seismic operation (Kong, Patel, et al., 2019), building health monitoring (Kong et al., 2018), and dense array detection (Inbal et al., 2019). EEW is also a goal of this global smartphone seismic network. In regions where there are no traditional seismic networks or early warning capabilities, MyShake could work as a standalone system to detect earthquakes and issue warnings to the public. Furthermore, in regions where traditional EEW does exist, MyShake could provide additional data and serve as a platform to deliver the alerts from traditional EEW systems. In October 2019, for example, the MyShake mobile application started to deliver EEW warnings in California from the statewide ShakeAlert EEW system, which uses a traditional seismic network (Strauss et al., 2020). The use of MyShake as platform to deliver EEW alerts produced by traditional seismic networks is beyond the scope of this article; here, we provide an analysis of the capabilities of MyShake smartphone network alone. Because of the fact that the current MyShake network is relatively sparse, especially outside the United States, the potential for MyShake networks to contribute to EEW has not been systematically assessed beyond a handful of basic simulations. Such systematic assessment is vital before MyShake can begin to issue public early warnings. The usefulness of MyShake networks for early warning will vary from region to region, depending on a wide range of factors such as the distance between population centers and active faults, the density and distribution of MyShake users and the origin time, and the magnitude of the earthquake. Quantification of these factors will allow the MyShake development team to identify regions of the world where EEW with MyShake would be feasible and most beneficial, the minimum number or density of users required for accurate rapid detections, and the likely warning times that could be issued to major population centers in the event of large earthquakes. The purpose of this article is to describe a simulation platform that can be used to understand MyShake EEW performance under the condition that we have a sufficiently dense network of users, for example, 0.1% of the population. Our network detection algorithm is designed to detect earthquakes by clustering triggers from phones in both time and space. This first part of our two-article series describes a simulation platform and a network detection algorithm that have been built to understand the performance of MyShake networks. The platform, built on top of MyShake observations with the aid of a simple physics model and a series of machine learning algorithms, can be used to test and understand the whole MyShake workflow from individual phone triggers to the final detection of the earthquake and estimation of the alerting area. It can simulate the trigger times and ground acceleration values that might be expected from hypothetical MyShake networks responding to given input events and population densities. The locations, time