Tracking Local Sea Ice Using Distributed Acoustic Sensing And Machine Learning
Researchers from the University of New Mexico and Sandia National Laboratories recently created a way to track local sea ice using a telecommunications fiber optic cable, located in Alaska, combined with Machine Learning algorithms to analyze ground vibrations due to ocean waves. The research titled, “Tracking Local Sea Ice Extent in the Beaufort Sea Using Distributed Acoustic Sensing and Machine Learning” examines a fiber optic cable, positioned offshore of Oliktok Point, Alaska.
Utilizing emerging technology called distributed acoustic sensing (DAS), Andres Felipe Peña Castro, lead author, from the University of New Mexico and team detected seismic indications of sea ice coverage and strength. DAS data was used to detect when and where ground vibrations normally created by ocean waves were suppressed by the presence of sea ice. This method presents an avenue for monitoring sea ice, exhibiting enhanced precision in both spatial and temporal dimensions—measured in minutes and meters. This stands in contrast to satellite pictures, refreshed on a daily basis, which can encompass vast areas but provide lower local resolution.
The fiber optic cable, like cables used to transmit internet, is connected to an interrogator, which sends pulses of light into the cable and measures reflected energy. The team could only view 40 kilometers of the line because eventually the reflected energy becomes too weak. Fiber optic cables are essentially skinny wires made of glass and they efficiently transmit data as optical signals.
“We're not using it to transmit internet data, but rather to continuously measure vibrations in the cable. We input known pulses of light and see what’s reflected back to the interrogator and how that changes through time within small subsets of the cable,” Brandon Schmandt, UNM professor of Earth and Planetary sciences explained. “It's constantly querying the cable and recording at a high sample rate of a thousand times per second the changes in each 2-meter segment of the cable. Little changes in the signal can be translated into ground motion slightly changing the length of the cable, stretching or squeezing it by tiny fractions.”
With these two components researchers can measure deformation along the cable. Researchers at Sandia National Laboratories, Michael Baker and Robert Abbott, lead the project and collaborated with the seismic research team at UNM to explore different kinds of signals that might arise from the subsurface, ocean, ice, and atmosphere. The researchers set out to study how sources of ground vibration varied along the cable and wanted to explore automated ways to analyze the giant amount of DAS data.
“We didn't necessarily set out to do a sea ice tracking project. We knew there was a really rich data set for activity in the near coastal environment, and we expected sea ice among a variety of things could be in that story. But the sea ice part just popped out as one of its clearest capabilities,” explained Schmandt.
With over three terabytes of data collected each day, Peña Castro and team started using frequency-time representations of the data which are images showing what frequency content is strongest at different times and how it evolves. The team decided to start applying some machine learning algorithms to find patterns within the massive dataset.
“The most important result is that we show that this new technology can be used, at least in the Arctic, to track sea ice and complement satellite observations in the Arctic with locally higher spatial resolution”, Peña Castro stated. “Sea ice is one of the critical measurements scientists are using to study ongoing climate change and its impacts.”
The researchers found that the data seemed to showcase two dominant behaviors, and understanding which behaviors were active at which point on the cable started to highlight the importance of sea ice for extremely local seismic noise. The biggest control on the ground vibrations in that environment of the shallow sea was where there’s ice and where there isn't ice. The fiber optic line extends about 40 kilometers straight offshore from the North Slope of Alaska, sticking out into the shallow margin of the Beaufort Sea. As Peña Castro and Schmandt point out, Arctic sea ice changes drastically through the year and between years. It is important to track this activity to measure climate change and navigate the Arctic.
“We're very interested in monitoring change in the Arctic. The Poles are some of the most rapidly changing areas on Earth due to long-term warming. They also change massively through the year, from being covered by sea ice in the winter to a brief but important period of open water in the summer,” Schmandt said, “Some of the motivation is understanding long-term climate. But it can also be more immediately practical by helping address questions like: When can you safely navigate the Arctic Ocean coastline? The kind of signal that Andres found was that when there's strong enough ice, it dampens or mutes the ocean waves. When ice is strong enough to get rid of the ocean waves, that's probably the ice you don't want to drive a boat into. So, it's kind of a different definition of ice coverage than optically saying, does the ocean look white or blue? It could be very practical in terms of navigation in the Arctic.”
The importance of understanding where ice is and how it changes affects things like coastal erosion, the stability of the coastline in the Arctic, and the big picture of monitoring the climate in the Arctic.
This research was funded by Sandia’s Laboratory Directed Research and Development program. It provided UNM researchers a great opportunity to collaborate and explore cutting edge data from an extremely remote environment.
Source: The University of New Mexico