Shen et al. (2026) Unsupervised Characterization of Rain‐Induced Seismic Noise in Urban Fiber‐Optic Networks Using Deep Embedded Clustering

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-01-01
• Authors: Junzhu Shen, Tieyuan Zhu
• DOI: 10.1029/2025wr041137

Research Groups

Specific research groups, labs, or departments are not explicitly mentioned in the provided abstract. The study was conducted using an array in State College, PA.

Short Summary

This study introduces a Deep Embedded Clustering (DEC) method to automatically detect and classify rain-induced seismic noise from Distributed Acoustic Sensing (DAS) data, demonstrating its ability to predict rainfall intensity and stormwater discharge duration.

Objective

• To introduce a data-driven method, Deep Embedded Clustering (DEC), to automatically detect and classify rain-induced noise from massive Distributed Acoustic Sensing (DAS) data, predicting the presence of moderate to heavy rain and the duration of stormwater discharge.

Study Configuration

• Spatial Scale: A 4.2 km-long underground fiber-optic array in State College, PA.
• Temporal Scale: Continuous DAS recordings from 2019 to 2021 (a 3-year period).

Methodology and Data

• Models used: Deep Embedded Clustering (DEC), Autoencoder (used within DEC for latent feature learning).
• Data sources: Continuous Distributed Acoustic Sensing (DAS) recordings from a preexisting telecommunication optical fiber array.

Main Results

• The Deep Embedded Clustering (DEC) model successfully detects and classifies rain-induced noise, predicting the presence of moderate to heavy rain and the duration of stormwater discharge.
• The model identified four distinct clusters corresponding to background noise, rain-induced noise of varying intensities, and stormwater discharge in sewers.
• Model-derived post-rain discharge durations align with synthetic hydrograph estimates, yielding a drainage system time of concentration of 21 minutes for the region.
• The workflow was successfully applied to two additional locations, demonstrating its potential for spatial monitoring.

Contributions

• Introduces a novel data-driven method (Deep Embedded Clustering) for automated detection and classification of rain-induced noise from Distributed Acoustic Sensing (DAS) data.
• Establishes a link between stormwater discharge processes and DAS-recorded signals, addressing the previous lack of a physical model.
• Provides a scalable solution combining machine learning and fiber-optic sensing for improved stormwater management in urban environments.
• Demonstrates the ability to infer drainage system characteristics, such as time of concentration, from DAS data.

Funding

Funding information is not provided in the abstract.

Citation

@article{Shen2026Unsupervised,
  author = {Shen, Junzhu and Zhu, Tieyuan},
  title = {Unsupervised Characterization of Rain‐Induced Seismic Noise in Urban Fiber‐Optic Networks Using Deep Embedded Clustering},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr041137},
  url = {https://doi.org/10.1029/2025wr041137}
}