Context
We invite you to work with an exciting new technology; distributed fibre optic sensing. This amazing sensing system uses off the shelf fibre optic cables, like the ones we use for our internet and telecoms, and repurposes them to be vast arrays of microphones (or thermometers) measuring vibrations (and temperatures) every couple of meters over 10’s of kilometres.
We have access to the Ocean Observatories Initiative RAPID dataset where two seafloor cables located offshore Oregon recorded data over a 4 day period in 2021 (Wilcock et al., 2023). Over that time period we can observe wave and tidal effects, earthquakes, marine mammals (blue & fin whales), shipping and road traffic. The challenge is to integrate multiple datasets, separate the signals from the noise and improve detection and classification on these data.
In this hackathon, you will be provided with access to these seafloor cable data, and your task will be to develop algorithms that can process and analyze these phenomena.
Example:
Challenge Goals
This is a multi-faceted dataset that includes fibre optic Distributed Temperature Sensing (DTS), fibre optic Distributed Acoustic Sensing (DAS), AIS (vessel tracking data), bathymetry, earthquake catalogues and weather data. For this reason there are many different possibilities, we have suggested a few challenges below. You could tackle these sequentially or you could focus on just one. You are welcome to propose your own challenge as there are so many possibilities!
Challenge 1: Signal Conditioning
Whilst fibre optic data provides new acoustic sensing opportunities it does has challenges including a lower sensitivity. The first challenge is to improve the signal conditioning to improve data quality. In other words,
- What is the optimal processing flow for vessel noise, marine mammals, waves and earthquakes?
- These are vast datasets (26 Terabytes) so automated fast processing automated detection are needed.
How many vessels, whales and earthquakes can your team find !
Challenge 2: Information Extraction
Detecting and classifying an event is just the start of an information extraction process. Once we have identified a signal we want to extract more information. So, in the case of propeller noise we would want to know how fast the propeller is turning, how many blades does it have and where is the vessel? For marine mammals we might want to identify the species, it location and its depth. This challenge is to take the identified signals and try to extract as much information from it, such as
- Locating & tracking vessels/marine mammals
- Characterize ocean hydrodynamics tides/waves etc.
Challenge 3: Visualization of integrated datasets
This is a diverse dataset that would benefit from interactive data visualization tools. If you want to be inspired then take a look at the recent ‘Big Glass Mic’ exhibit at the Victoria & Albert Museum exhibit [link, Stamen Design]. This shows an interactive acoustic data visualization from ‘dark fibre’ around Stanford University’s campus using waterfall plots and spatial visualizations of the fibre.
Pre-requisites
We will provide a curated data subset and Jupyter notebooks to jumpstart the process. We will also suggest some libraries that you may wish to preinstall and test.
You will need to bring a laptop with appropriate computing tools e.g. Python / MATLAB etc., as well as any relevant libraries/toolboxes e.g. NumPy, Matplotlib, PyTorch, TensorFlow.
Background Material
- RAPID: Distributed Acoustic Sensing on the OOI’s Regional Cabled Array – Ocean Observatories Initiative. [Link]
- Victoria & Albert Museum Big Glass Mic Data Visualization [Link]
- Distributed acoustic sensing recordings of low-frequency whale calls and ship noise offshore Central Oregon: JASA Express Letters: Vol 3, No 2 (scitation.org) [pdf]
Online Resources
awesome-das: A curated list of DAS tools and resources on Github. [Link]
DAS4Whale https://github.com/leabouffaut/DAS4Whales
DASDae DASDAE · GitHub
Ocean Observatories Initiative RAPID dataset
Wikipedia: Distributed Acoustic Sensing [link]
Segment Anything Model [link]
Oriented Filtering [link]
Videos
Using DAS to observe nearshore waves and processes – Hannah Glover & Marcela Ifju [link]
Seafloor Fibre Optic Sensing Joint IRIS & DAS RCN Webinar [link]
References
Wilcock, W. and Ocean Observatories Initiative (2023). Rapid: A Community Test of Distributed Acoustic Sensing on the Ocean Observatories Initiative Regional Cabled Array [Data set]. Ocean Observatories Initiative. https://doi.org/10.58046/5J60-FJ89
Lior, I., A. Sladen, D. Rivet, J.-P. Ampuero, Y. Hello, C. Becerril, H. Martins, P. Lamare, C. Jestin, S. Tsagkli, C. Markou. 2021. On the Detection Capabilities of Underwater Distributed Acoustic Sensing. JGR 126 (3).
Bouffaut, L, et al, (2022) Eavesdropping at the speed of light: distributed acoustic sensing of Baleen whales in the arctic. Frontiers in Marine Science, 9. doi.org/10.3389/fmars.2022.901348
Waagaard, OH, et al, (2021) Real-time low noise distributed acoustic sensing in 171 km low loss fiber. OSA Continuum, 4(2), 688–701. doi.org/10.1364/OSAC.408761
Waagard, O. H. 2022. Listening across the oceans: Distributed acoustic sensing.
Ugalde, A. 2022. Noise levels and signals observed on submarine fibers in 1 the Canary Islands using DAS.
Taweesintananon, K. 2023. Distributed acoustic sensing of ocean-bottom seismo-acoustics and distant storms: A case study from Svalbard, Norway. Geophysics.
María R. Fernández-Ruiz, M. 2022. Seismic Monitoring With Distributed Acoustic Sensing From the Near-Surface to the Deep Oceans. Journal of Lightwave Technology 40 (5) 1453.
Rørstadbotnen RA, Eidsvik J, Bouffaut L,Landrø M, Potter J, Taweesintananon K, Johansen S, Storevik F, Jacobsen J,Schjelderup O, Wienecke S, Johansen TA,Ruud BO, Wuestefeld A and Oye V. 2023. Simultaneous tracking of multiple whales using two fiber-optic cables in the Arctic. Front. Mar. Sci. 10:1130898.doi: 10.3389/fmars.2023.1130898
Andreia Pereira, Danielle Harris, Peter Tyack, and Luis Matias. 2016. Lloyd’s mirror effect in fin whale calls and its use to infer the depth of vocalizing animals. Proc. Mtgs. Acoust. 27, 070002; doi: 10.1121/2.0000249
Hiroyuki Matsumoto, Eiichiro Araki, Toshinori Kimura, Gou Fujie, Kazuya Shiraishi,
Takashi Tonegawa, Koichiro Obana, Ryuta Arai, Yuka Kaiho, Yasuyuki Nakamura,
Takashi Yokobiki, Shuichi. 2021. Detection of hydroacoustic signals on a fiberæ‘optic submarine cable. Nature Scientific Reports. 11:2797
M. Karrenbach, R. Ellwood, V. Yartsev, E. Araki, T. Kimura, H. Matsumoto. 2020. Long-range DAS Data Acquisition on a Submarine Fiber-optic Cable. EAGE Workshop on Fiber Optic Sensing for Energy Applications in Asia Pacific.
Data Organization
The data from the North and South cables have been acquired by two separate companies; OptaSense and Silixa. OptaSense recorded data on both cables at different times with different recording parameters and used the HDF5 data format. Silixa recorded data on the South cable with the same recording parameters in the TDMS format.
For your convenience we have converted some of the data into the MatLab™ data format with a format naming convention of
Company_Cable_YYYYMMDDHHMMSS_GL_SPACING_SAMPLINGFREQUENCY.mat
where
Company = Company name (Either OPT for OptaSense or SLX for Silixa)
Cable = Cable name (either North or South)
YYYYMMDDHHMMSS = UTC Date and Time stamp
GL = Gauge Length in meters (related to fibre optic DAS)
SPACING = Spacing between adjacent channels in the data
SAMPLINGFREQUENCY = Sampling frequency in Hertz
So for example
is data recorded by OptaSense on the North cable on the 4th November 2021 at 02:02:51 UTC using a gauge length of 51m, with a 2m channel spacing at 200Hz.
The data examples are stored in the data folder with sub folders
Data can be converted from the original HDF5 and TDMS data formats using the Python Jupyter notebook
UKAN2023_ConvertToMatlabFormat_v01.ipynb
Code
We provide Python Jupyter notebooks (and MatLab™ scripts) to provide a quick start to the challenges;
These notebooks will read in data examples, implement minimal processing and display data plots.
We have converted some code from Python to Matlab™.