Updated: Jul 13
The CRISiSLab team’s earthquake early warning technical paper, entitled “Saving Precious Seconds”, was recently published in the Informatics Journal. For all the technical details, read our paper in the MDPI Informatics Journal. It is open access!
In our paper, we proposed to save a few seconds on earthquake early warning (EEW) through a new decentralised architecture! Most EEW systems use centralised servers, which means when sensors detect something, they will have to go through a centralised server to do the computations before issuing a warning.
For our approach, we took the central server out of the picture! We got the sensors to directly talk to each other and do all the essential computations! We hypothesised that without a central server, it would save us a critical few seconds to give people more warning.
Illustrating different architectures and their layers: (a) centralised architecture, (b) typical decentralised architecture, and (c) proposed decentralised architecture.
What’s unique about our EEW system is that we used low-cost sensors, called Raspberry Shakes, installed in volunteers’ homes. To be able to do this, we needed to decide on some technologies to use: SD-WAN, hole-punching mechanism, and PLUM algorithm. I’ll explain.
We needed first to establish secure communication links between the installed sensors. We chose SD-WAN: a virtual Wide Area Network that allows us to leverage any combination of transport services – including LTE and broadband – to connect the sensors to each other securely. SD-WAN is cool because if there is an outage with one sensor, the network can “self-heal” and quickly reconnect with the remaining online sensors!
We also employed a hole-punching mechanism. Hole punching is a technique in computer networking that allows direct connection between two parties even if they have firewalls. Skype uses this mechanism. This is advantageous for our network, as it allows the sensors to talk to each other with fewer barriers.
We used the PLUM algorithm to monitor ground motions. The PLUM algorithm is faster than alternatives because it does not need to locate and estimate the earthquake’s magnitude. Sensors only need to observe ground motion data to warn the sensors further afield of shaking.
Having these details for our decentralised network, we ran 60 earthquake simulations. We used six different rupture scenarios and used our experimental network in Wellington, New Zealand. We tested and compared results between the centralised and the decentralised approach.
Our study results show that using a decentralised approach is much faster than the centralised ones! It can save up to 2.09 seconds. It doesn’t sound much, but that is a lot in the context of earthquake early warning! It is “saving precious seconds”
Thank you to EQC and Massey University for funding this research.