Counting from one to nine to detect debris flows
A groundbreaking method using Benford’s law allows the detection of debris flows from seismic signals.
Mass movement processes, such as landslides, lahars, glacial lake outburst floods and debris flows, can have devastating impacts, including significant loss of life and costly damage to infrastructure and livelihoods. With the development of environmental seismology methods, continuous monitoring signals offer a new way to study mass movements and can be used in early warning systems. However, it remains challenging to automatically identify and detect specific events of interest from the large amount of data that are continuously collected.
Notwithstanding recent advances, complex information such as hazard detectors, numerous waveforms, and spectral and network features or parameters are required to be fed into the model to identify events, which is time-consuming and requires experience. Therefore, a convenient and computationally cheap approach to event detection needs to be developed before real-time warning systems can become a reality.
Benford’s law is a simple statistical law that states that the first non-zero digit of given datasets should follow a specific probability distribution, that is, that each non-zero digit is expected to appear a given number of times relative to the others in the dataset. Benford’s law has been successfully applied in earthquake detection. In their new study, Zhou et al. [2024] test the applicability of Benson’s law to detect debris flows. The authors calculate the first-digit distribution of seismic signals generated by debris flow events from the Illgraben, Switzerland. They find that seismic signals of debris flows follow Benford’s law during the run-out phase, while ambient noise does not. Furthermore, additional results from limited seismic data indicate that landslides, lahars, transport of gravel by rivers during floods, and glacial lake outburst floods may follow Benford’s law, while rockfalls may not.
The authors give us a computationally cheap, innovative technique that offers an alternative to the complex approaches currently available for the identification of debris flow events and potentially for real-time warning. Furthermore, features related to Benford’s law (e.g., first digit frequency of one to nine, the goodness of fit, and the power law exponent) can be further incorporated into the training of machine learning algorithms for a better detection of debris flow. This work has important implications for the building of effective debris flow early warning systems based on the use of easily measured seismic signals.