As artificial intelligence (AI) has grown exponentially in both ability and accessibility in recent years, researchers are exploring how AI can help advance discoveries.

This is especially true for researchers using AI to model large-scale impacts due to a changing climate. For two days, led by assistant professor of data science and environmental sciences Antonios Mamalakis, the Climate Fellows at UVA’s Environmental Institute explored the topic in depth.

In recent decades, researchers have worked to leverage the advantages of theoretical physical models to understand climate dynamics. However, the coming era will be defined by scientists’ ability to harness the vast and growing amounts of data from ground-based observations, satellites, and simulations to better create realistic physical models based on what is already happening and use these to predict future scenarios.

AI offers a unique way to effectively analyze and draw insights from this data. In addition, an exciting frontier presented by the growth of AI is how “deep learning” captures the nonlinear, complex behaviors of climate systems.

Mamalakis is an experienced environmental data scientist and one of his projects, supported by the Institute, explores ways to develop AI models that take low-resolution climate maps as input and produce high-resolution counterparts that capture climatic structure. In addition to research, Mamalakis teaches several courses on the intersection of environmental science and data science, and together with his students, he recently won Best Paper in the Environment and Health category at the Systems and Information Engineering Design Symposium (SIEDS 2024).

Some of his papers have garnered international attention and have been highlighted by publishers. Examples include "A new interhemispheric teleconnection increases predictability of winter precipitation in southwestern US," published in Nature Communications (editors’ highlight); "Zonally contrasting shifts of the tropical rain belt in response to climate change," published in Nature Climate Change; and "Underestimated MJO variability in CMIP6 models," published in Geophysical Research Letters (editors’ highlight). 

“AI models, and specifically artificial neural networks, are fascinating tools that can capture highly complex system behavior, just by performing a large series of simple mathematical operations, like addition and multiplication” Mamalakis said. “It was so exciting to share AI insights with the talented climate researchers of the Institute and see their joy in realizing how conceptually simple, yet powerful, these tools are. Thinking ahead, I need to highlight that given the complexity and high dimensionality of climate systems, and the growing amounts of available data, AI is here to stay, as a tool in the geosciences. Thus, it is fundamental for me to train future generations of researchers in becoming fluent with these tools.”

The Institute asked two of the Climate Fellows, Vanessa Ines Cedeno and Swatah Borkotoky, to reflect on what they learned following the workshop with Mamalakis. 

Q: What was something interesting/surprising you learned in the workshop? 

A: Vanessa

It was interesting to learn about building, training, and predicting with an Artificial Neural Network (ANN) [a specific AI methodology]. The basic concepts are hard to grasp, but the workshop made it entertaining. It was effective to make the workshop hands-on by using tools like Interactive Python Notebooks, to learn about the fundamental unit of an ANN, the layers and activation functions.

A: Swatah

Antonios' approach in explaining the science behind AI with simple example problems was the most interesting and effective aspect of this workshop.

Q: Why do you think this topic is important? 

A: Vanessa

AI plays a crucial role in addressing important problems from different fields. ANNs are a powerful tool driving innovations and improvements across various industries. Innovative solutions and strategies can be developed to solve problems more effectively and efficiently.

A: Swatah

Given the ever-improving capacity and accuracy of Artificial Intelligence, it is important now more than ever to utilize this technology to solve some of the biggest challenges in science. One of which is climate change. As such this workshop was very useful in introducing these concepts, especially in the context of climate change.

Q: What challenges do you think AI will present for researchers in the future? 

A: Vanessa

AI will likely present several challenges for researchers in the future. Some of them are actively being taking into account in research, like ethical dilemmas related to data privacy, consent, and the implications of AI decisions. Also, it will be a significant challenge to ensure that AI systems are fair and free from biases. AI models have to be transparent and interpretable, in order for them to work in an interdisciplinary collaboration. 

A: Swatah

AI in its very essence is a bit of a “black box” [a mystery]. Hence, improving the understanding of the intricate mechanism of AI will be of immense importance and a challenging endeavor for researchers in the future.

Q: What benefits do you think AI could have for researchers in the future? 

A: Vanessa

AI could offer numerous benefits for researchers in the future, like enhancing the accuracy and speed of data interpretation. Also, facilitating interdisciplinary collaborations. With the use of predictive modeling, researchers can predict outcomes based on various variables, aiding in fields like climate science, epidemiology, and social sciences.     

A: Swatah

Some of the biggest problems in science are unsolvable by conventional wisdom due to the sheer scale of the problem. These problems however are more tractable using AI algorithms, as evidenced by recent winners of the prestigious Nobel prize. I hope to see similar if not bigger breakthroughs in climate science using AI.