UC Berkeley

Earthquakes are global hazards that account for many deaths and economic losses each year. With the development of seismology and technology, we have denser and denser sensors to monitor the earthquakes and provide valuable dataset to understand the processes of the plate tectonics, earthquake rupture processes, etc. Many useful applications built on top of these sensors including earthquake early warning systems that have the great potential to reduce the earthquake hazards for human civilizations. From the seismological point of view, the seismologists have been trying different ways to increase the density of the sensor network to monitor the earthquakes. From high-quality broadband seismic stations, to low-cost microcontroller devices, the instruments used in seismology are crossing the whole spectrum. In this dissertation, I report our progress on building a global seismic network using the consumer smartphones. The goal is to use the power of crowdsourcing devices to setup a scalable seismic network to compliment the existing high-quality seismic network, especially provide useful data at places where they cannot afford the high-quality instruments.

This thesis starts with the design of the methodology and experiments we did before the building of this global network, including the noise floor tests, shake table tests, and the design of the artificial neural network to distinguish the earthquake signals recorded on the phone from the daily human activities. With the earthquake early warning application in mind, these form the basis and the blue prints to build the network. In Feb 12th 2016, MyShake application and the whole system released to the public. Within very short time MyShake users cover 6 continent and starting to provide the shaking data related to the earthquakes. The initial observations from this network validate the initial design and concepts, at the same time it shows great potential to use the recorded data to do routine seismological applications. The two types of data from MyShake, i.e. real-time trigger messages and the 5-min long 3-component waveforms have different applications. The real-time trigger messages enable MyShake network to be used as a stand-alone earthquake early warning system, including estimate the initial location, magnitude and origin time of the earthquake. On the other hand, with the waveforms we recorded from the smartphones, we could refine these earthquake parameters at better accuracy. The comparison of the estimated locations, origin times and magnitudes from the MyShake recordings with those from the catalog shows the data from these consumer devices are useful to quantify the earthquakes. This will be really useful by providing extra data at places where no or few seismic stations near the earthquake but with a large population. This thesis also talks about the potential application of using MyShake to conduct structural health monitoring of buildings in the future. The shaker test we did proves the sensors in the phones could be used to extract the fundamental frequencies from the shaking of the buildings.

The success of building this global smartphone seismic network and the initial analysis we conducted using the data recorded on it provide the community an exciting way to monitor earthquakes, though there are still many challenges and limitations need to be addressed. The last part of this thesis talks about the pathway forward from our experience with this network. Besides, MyShake project is an example of the combination of data science and earth science, I will end the thesis with a discussion of my thinking of how to take the full advantage of both sides.