Skip to content

4th Underwater Acoustics Hackathon – Dstl

Challenge #3

Categorisation of underwater acoustic environments

Context

The sound speed profile describes the speed of sound at different depths in the ocean. The profile is an important factor in predicting how sound propagates in the ocean, which has applications in fishing, oil, search & rescue, marine-life preservation, defence and more. Sound speed profiles vary spatially and temporally over different length and time scales making it complex, difficult, and computationally costly to model acoustic propagation at all points in a region of interest in the ocean. Additionally, most measurements of the sound speed profile are made at a single location using vertical profilers, meaning accurate characterisations of the variation in the sound speed in the vicinity are vital in sound propagation problems.

In the ocean, however, there are patterns and structures in the sound speed profiles that are tied to time and location in the water. For example, the temperature of the surface of the ocean in the North Sea is higher in the summer than the winter. If instead, sound speed profiles are categorised by their structural similarity and linked to parameters such as time (seasonal and diurnal) and geographic region, it should be possible to determine the most likely sound speeds for each category and identify global and local variations of the sound speed profiles using these categories.

Objectives

Participants are invited to explore methods for categorising / grouping sound speed profiles, relating these groups back to user-defined parameters.

Possible ways to group and sort sound speed profiles could be (but not limited to):

  • Structural similarity of the sound speed profile
  • Time of the day / year
  • Location
  • Water depth, numeric or categorical (shallow water, shelf, basin etc)
  • Underwater environment properties (temperature, biologics)
  • Vicinity to geographical features

From these groups, participants are invited to describe any local or global variations in the sound speed profiles. For example, if a category is bounded by a geographical location, what is the variation in the structure of the sound speed profile within this category.

As a scientific research organisation, Dstl are interested in an analysis of the grouping. This might include addressing the following questions (and others the participants produce):

  • How are the numbers of groups decided?
  • What are the dominant features in each group?
  • Do the groups cover equal total areas?
  • Are there significant features that dictate boundaries between groups?
  • What is the likelihood of occurrence of different groups?
  • Can the groups be used to identify profiles that significantly deviate from a particular cluster and therefore produce unusual acoustic behaviour?
  • Do the groups evolve monthly or seasonally and how?

This task invites multiple approaches, any of which are of interest. For example, machine-learning techniques (such as clustering) might be one approach, another might be an empirical study of different sound speed profiles.

You will need to bring a laptop with internet connectivity and any software you would like to use for the problem (MATLAB, Python, R, etc).

Resources

https://www.ncei.noaa.gov/products/world-ocean-atlas – World Ocean Atlas. Database of worldwide historic (physics calculation) temperature, salinity and more with location and time averages (5 degree point map and seasonal).

https://data.marine.copernicus.eu/products?pk_vid=bbf9ca7d9cb50b1b1777641498753b3e – Copernicus marine data – physics reanalysis. Similar data to the world ocean atlas. The Copernicus marine data website provides additional oceanographic data that may be interesting to include in categorisation if applicable.

http://resource.npl.co.uk/acoustics/techguides/soundseawater/underlying-phys.html – Useful reference for calculation of sound speed in the ocean from temperature, salinity and depth.

https://www.gebco.net/data-products/gridded-bathymetry-data – Global ocean depth (bathymetry) data. Users can download the full world dataset or defined regions.