26 April 2025 in Journal of Geophysical Research: Earth Surface
The advent of machine learning has significantly improved the accuracy of identifying mass movements through the seismic waves they generate, making it possible to implement real‐time early warning systems for debris flows. However, we lack a profound understanding of the effective seismic features and the limitations of different machine learning models. In this work, we investigate eighty seismic features and three machine learning models for single‐station‐based binary debris flow classification and multi‐station‐based warning tasks. These seismic features, derived from physical and statistical knowledge of impact sources, are grouped into five sets: Benford's law, waveform, spectra, spectrogram, and network. The machine learning models belong to two families: two ensemble models, Random Forest and eXtreme Gradient Boosting (XGBoost); one recurrent neural network model, Long Short‐Term Memory (LSTM). We analyzed feature importance from the ensemble models and found that the number and even the types of seismic features are not critical for training an effective binary classifier for debris flow. When using models designed to capture patterns in sequential data rather than focusing on information only in one given window, using the LSTM does not significantly improve the performance of binary debris flow classification task over Random Forest and XGBoost. For the multi‐station‐based debris flow warning task, the LSTM model predicts debris flow probability more consistently and provides longer warning times. Our proposed framework simplifies machine learning‐driven debris flow classification and lays the foundation for affordable seismic signal‐driven early warning using a sparse seismic network.
12 September 2024 in Journal of Geophysical Research: Earth Surface
Seismic instruments placed outside of spatially extensive hazard zones can be used to rapidly sense a range of mass movements. However, it remains challenging to automatically detect specific events of interest. Benford's law, which states that the first non‐zero digit of given data sets follows a specific probability distribution, can provide a computationally cheap approach to identifying anomalies in large data sets and potentially be used for event detection. Here, we select vertical component seismograms to derive the first digit distribution. The seismic signals generated by debris flows follow Benford's law, while those generated by ambient noise do not. We propose the physical and mathematical explanations for the occurrence of Benford's law in debris flows. Our finding of limited seismic data from landslides, lahars, bedload transports, and glacial lake outburst floods indicates that these events may follow Benford's Law, whereas rockfalls do not. Focusing on debris flows in the Illgraben, Switzerland, our Benford's law‐based detector is comparable to an existing random forest model that was trained on 70 features and six seismic stations. Achieving a similar result based on Benford's law requires only 12 features and single station data. We suggest that Benford's law is a computationally cheap, novel technique that offers an alternative for event recognition and potentially for real‐time warnings.
23 July 2024 in Earth and Space Science
Bedload transport is a natural process that strongly affects the Earth's surface system. An important component of quantifying bedload transport flux and establishing early warning systems is the identification of the onset of bedload motion. Bedload transport can be monitored with high temporal resolution using passive acoustic methods, for example, hydrophones. Yet, an efficient method for identifying the onset of bedload transport from long‐term continuous acoustic data is still lacking. Benford's Law defines a probability distribution of the first‐digit of data sets and has been used to identify anomalies. Here, we apply Benford's law to continuous acoustic recordings from Baiyang hydrometric station, a tributary of Liwu River, Taroko National Park, Taiwan at the frequency of 32 kHz from stationary hydrophones deployed for 3 years since 2019. We construct a workflow to parse sound combinations of bedload transportation and analyze them in the context of hydrometric sensing constraining the onset, and recession of bedload transport. We identified three separate sound classes in the data related to the noise produced by the motion of pebbles, water flow, and air. We identify two bedload transport events that lasted 17 and 45 hr, respectively, covering about 0.35% of the total recorded time. The workflow could be transferred to other different catchments, events, or data sets. Due to the influence of instrument and background noise on the regularity of the residuals of the first‐digit, we recommend identifying the first‐digit distribution of the background noise and ruling it out before implementing this workflow.
18 February 2026 in Journal of Geophysical Research: Earth Surface