Introducing Bedartha Goswami, new Section Editor for ML/AI at PLOS Climate

We are delighted to introduce Bedartha Goswami (IISER Pune, India) as Section Editor for PLOS Climate‘s new section on Machine Learning and Artificial Intelligence (ML/AI). In this blog post, we hear about Bedartha’s background and his aspirations for this section of the journal.

Could you tell us about your research background and your current work?

Throughout my research career, I have been developing data-driven approaches to analyse high dimensional weather and climate data sets from the current observational record, as well as from paleoclimate sources. Most of my work so far has been to use ideas from the fields of statistical machine learning and nonlinear data analysis to discover non-trivial features in weather and climate data, in the hope that they might give insights into developing predictive models for weather patterns.

In the last five years, I have been focusing on developing deep learning weather models for predicting seasonal to inter-annual phenomena such as the El Niño Southern Oscillation, and more recently, for short- to medium-range weather prediction. Since moving back to India in 2024, I am working on developing deep learning weather models that are frugal, open access, and high performing over South Asia.

Why did you decide to join PLOS Climate as a Section Editor? What excites you about the new ML/AI section of the journal?

The philosophy of PLOS journals to enable transparent and freely accessible scientific research is one that I agree with personally. In particular, PLOS Climate‘s focus on rigor and interdisciplinarity, and its push to highlight regional issues related to climate and weather patterns, especially in the context of climate change appeals to me greatly. Being born and raised in a Tier 2 city in India, I am convinced of the importance of showcasing high quality research on climate problems from the Global South to a worldwide audience. As a Section Editor at PLOS Climate, I will have a unique opportunity to help chart a more “region conscious” research agenda at the journal, a task which is nevertheless going to be challenging.

The ML/AI section of the journal, of course, aligns perfectly with the kind of research that I do and the kind of science that inspires me. I think the interface of ML/AI and weather/climate is buzzing with activity at the moment, fueled not in the least by the phenomenal AI weather models that have been released in the past five years. The field is moving at breakneck pace, with new models coming out almost every month. To be part of this revolutionary scientific moment, and to be able to foster research within this area is something that I am really looking forward to.

What kinds of submissions would you like to see to the ML/AI section of PLOS Climate?

Since ML/AI for weather/climate is a nascent field, there is quite some work left to be done in terms of methodological development, and thus new models are always welcome, but they need to present a significant and novel advancement over existing ones, in order to justify their usage. The advancement need not be one based on performance alone but something that addresses auxiliary issues related to real-world impact, such as those related to uncertainty, accessibility, frugality, and sustainable deployment.

Additionally, given that there are already numerous studies that have applied ML/AI to various aspects of weather and climate modeling, and in novel and interesting ways, we need studies that present use cases that ground these methods in real-world scenarios. In this regard, I would like to see more work, for instance, on how existing methods perform for regional weather and climate modeling, for extremes, for well-known seasonal weather patterns such as the Madden Julian Oscillation.

And as a last point, we need more focus on interpretability, and how ML/AI models and methods can contribute to evidence-based policy decisions. 

Ready to submit your work to PLOS Climate? Follow our step-by-step guide to the submission process.