Interview: "I am developing a physics-informed machine learning model to improve our estimates of ocean temperature and salinity in the Southern Ocean"

In this interview, iC3 postdoc Vijay Prakash Kondeti explains how he uses physics-informed machine learning, satellite observations and field data to investigate Southern Ocean heat transport, Antarctic ice-sheet melting, and future sea-level changes.

Can you briefly describe yourself and the focus of your research?

My name is Vijay Prakash Kondeti. I am an MSCA Postdoctoral Fellow at the iC3 Polar Research Hub, working in the Oceanography Section at the Norwegian Polar Institute.

My current research focuses on ocean–ice interactions, using satellite observations, field measurements, and machine learning. More specifically, I am developing a physics-informed machine learning model to improve our estimates of ocean temperature, salinity, and current fields in the Southern Ocean, particularly in the circum-Antarctic region.

My main aim is to improve our understanding of the processes through which ocean heat reaches Antarctica and contributes to ice-sheet melting, as well as to support future projections of the Antarctic Ice Sheet’s contribution to global sea-level rise.

Credit: Phyo Wai Thaw.

Why is the Southern Ocean so important for climate research?

The Southern Ocean connects the global ocean with Antarctica and plays a central role in transporting and storing heat. This is especially important around Antarctica because warm water masses-particularly Circumpolar Deep Water-can cross the Antarctic continental shelf, enter cavities beneath floating ice shelves, and drive basal melting.

Understanding when, where, why, and how Circumpolar Deep Water reaches the Antarctic margin is therefore essential for improving projections of Antarctic Ice Sheet mass loss and its contribution to future global sea-level rise.

How does your model work?

The model will be trained using field observations from Southern Ocean moorings and ocean-profile networks, combined with satellite and atmospheric data. The objective is to use widely available surface and atmospheric observations to improve estimates of subsurface ocean conditions across the Southern Ocean, including regions where direct measurements are sparse.

Unlike a purely data-driven approach, the model incorporates physical equations and oceanographic constraints into the learning process. This means that its predictions must remain consistent with known physical principles, helping to reduce unrealistic results.

The model is therefore designed to provide subsurface temperature, salinity, and current estimates that are both accurate and physically consistent.

Will the research include future projections?

First, I aim to generate hindcast simulations spanning approximately three to four decades and use them to analyze the variability of Circumpolar Deep Water. I want to understand how its behaviour differs across Antarctic shelf regimes and seasons, and how it responds to environmental forcing.

Once we have a clearer understanding of these processes, we can use future climate scenarios to assess how changes in atmospheric temperature, winds, sea ice, and ocean circulation may affect ocean heat delivery to Antarctic ice shelves. This will help evaluate potential future changes in basal melting, Antarctic Ice Sheet mass loss, and global sea-level rise.

What is the biggest scientific challenge in your work?

The main challenge is the limited availability of field observations. Satellite-based ocean studies and machine-learning models depend on high-quality in situ measurements for training, validation, and uncertainty assessment.

In both the Arctic and Antarctic, collecting these observations is difficult because of the remoteness of the region, extensive sea-ice cover, harsh weather, winter darkness, and the logistical cost of ship-based expeditions. As a result, the Southern Ocean remains sparsely sampled, particularly during winter and in sea-ice-covered regions.

A further challenge is the limited spatial resolution of many satellite products to resolve small-scale processes around Antarctica - such as cross-shelf exchange, coastal currents, eddies, and the ocean heatpathways.

What would you do with an unlimited research budget?

Comprehensive observations are essential for understanding polar climate processes because the atmosphere, ocean, sea ice, and ice sheet are closely interconnected. With an unlimited research budget, I would establish integrated, long-term observing sites around Antarctica. Each site would simultaneously measure ocean conditions, atmospheric variables, sea-ice mass balance, ice-sheet mass balance, and basal melt rates.

What do you enjoy doing outside science?

I consider myself a thalassophile - a lover of the sea. Wherever I travel, I enjoy finding a quiet beach with clean water, gentle waves, and a peaceful atmosphere.

In my free time, I also enjoy spending time with friends and cooking together.

Can you share a book or movie that has significantly influenced you? 

One film that has stayed with me is the Indian movie Three Idiots. It highlights the importance of conceptual and application-oriented learning.

I particularly value its message that education should encourage curiosity, independent thinking, and meaningful research.

What personal belief guides your approach to life and work?

My personal motto is: “God may not give what we want, but He always provides what we need.”

Vijay Prakash Kondeti is an MSCA postdoc working at the iC3 Polar Research Hub, based in the Oceanography Section at the Norwegian Polar Institute, Tromsø. He is part of iC3 Research Unit 5: Future climate and oceans and iT1: Novel technologies. To find out more about his work, please check out his list of publications, or contact him by email or via LinkedIn.