Utilizing Machine Learning to Gain a Better Understanding of Ocean Movement

Source: Journal of Advances in Modeling Earth Systems (JAMES)

Scientists who study the ocean, known as oceanographers, utilize satellites to observe the Earth and gather data on the height of the ocean’s surface. This data is valuable in mapping out the flow of ocean currents and comprehending their impact on heat distribution and climate change. The Surface Water and Ocean Topography (SWOT) satellite, set to launch in late 2022, will have the capability to capture finer details of sea surface heights at a scale of tens of kilometers, surpassing previous limitations of hundreds of kilometers.

The challenge lies in the fact that traditional approaches to interpreting ocean currents from sea surface heights are not effective at extremely detailed levels due to the complexities of physics. This is because closely examining the ocean also reveals waves below the surface which, while not impacting currents, can interfere with accurate measurements of sea surface height.

Now, Xiao et al. present a novel, machine learning method for using SWOT sea surface height data to estimate various aspects of current flow in the upper ocean. The method applies a computational approach inspired by human vision known as a convolutional neural network, which the research team trained on data from realistic simulations of sea surface heights and current dynamics.

The scientists showed that their method using a convolutional neural network can utilize precise measurements of sea surface heights to approximate certain aspects of current movement. This could lead to a better comprehension of how currents carry heat and carbon, potentially aiding in the prediction and understanding of climate change.

The authors of the study state that this initial success serves as evidence of the feasibility of the new approach, and additional investigations are necessary to improve its accuracy before it can be consistently applied to SWOT data.

For now, SWOT will be occupied with obtaining high-resolution pictures of not just the Earth’s oceans, but also nearly all bodies of surface water on a global scale, such as lakes, rivers, and reservoirs. (Journal of Advances in Modeling Earth Systems (JAMES), https://doi.org/10.1029/2023MS003709, 2023)

“Science Writer Sarah Stanley”

Reference: Stanley, S. (2023). Using machine learning to better understand ocean motion. Eos, 104, https://doi.org/10.1029/2023EO230394. Retrieved on 17 October 2023.

The text was written in 2023 by the authors and is licensed under CC BY-NC-ND 3.0.

Images are subject to copyright unless otherwise specified. It is prohibited to reuse them without the explicit permission of the copyright owner.