18. Using data science to extract submesoscale features from high-resolution aerial remote sensing data

 

Daniel Carlson

We use high resolution aerial remote sensing data (LWIR and hyperspectral imagery) from aircraft to study submesoscale to ocean boundary layer-scale ocean dynamics. When an eddy or front is detected in the aerial imagery, the data are used to direct ships and robotic platforms to the feature to carry out intensive in situ sampling. This observational strategy results in high-resolution measurements of sea surface temperature, ocean color, density, and velocities to provide unique, data-rich datasets of transient ocean features.

What is the data science project’s research question? Intensive aerial observations of ocean dynamics produce more data than human researchers can process. Therefore, this project attempts to answer the following question: Can data science methods, specifically unsupervised algorithms, be used to reliably and quickly extract submesoscale flow features from high-resolution remote sensing data?

What data will be worked on? High resolution (~3-5 m; 50 Hz) aerial observations of sea surface temperature acquired by long-wave infrared cameras and possibly also hyperspectral imagery from campaigns in the Baltic Sea and the Atlantic Ocean. The data are georectified and are analysis-ready.

What tasks will this project involve?  The project will require the development of algorithms that can efficiently process large amounts of data to identify coherent ocean features, like fronts and eddies, and extract relevant information (velocity, vorticity, convergence/divergence, etc.). In some cases, in situ data are available for validation.

What makes this project interesting to work on?  Ocean submesoscales (0.1 – 10 km) take the form of sharp filaments, fronts, and eddies (vortices), which can result in relatively strong vertical velocities. Submesoscale features, therefore, impact the dispersion of buoyant materials and the distribution of tracers like heat and momentum in the mixed layer. Most of our knowledge of submesoscale dynamics is based on numerical simulations and observational techniques are only just beginning to “catch up.” The project will allow applicant(s) to work on with some of the most advanced observational datasets acquired to date. The applicant(s) will develop data science techniques that will lead to efficient extraction and visualization of flow features from large datasets to aid in our collective understanding of submesoscale ocean dynamics. .

What is the expected outcome? Contribution to research paper, Contribution to software development

What infrastructure, programs and tools will be used? Can they be used remotely?  Existing analysis tools were developed in Python and Matlab. High performance computing facilities are available. Remote access is available.

What skills are necessary for this project? Data analytics / statistics, Scientific computation, Data mining / Machine learning, Deep learning, Visualization, Geographic Information Systems, High-performance computing

Is the data open source? The data will be open source before the end of 2021

Interested candidates should be at Phd level . Daniel Carlson  is looking for 2 visiting scientists , working on the project together with the team.