Physically-Scaled Machine Learning for Tropical Cyclone Intensity Prediction

• Incorporating physical knowledge can improve machine learning predictions and ground them in physical reality. I was a visiting scientist and am an ongoing collaborator with the Data-Driven Atmospheric and Water Dynamics Group at the Université de Lausanne. Our ongoing work is focused on developing a physics-based scaling for tropical cyclone intensity prediction. I also help with other projects in the group, including causally-informed feature selection for tropical cyclone prediction.
  Related material: AI for Tropical Meteorology: Challenges and Opportunities, AI for Good Seminar Series.

Machine Learning and Tropical Cyclone Forecasting

• Much of my work in the last few years has focused on machine learning and tropical cyclone forecasting. Some of my projects include machine learning and tropical cyclone intensity forecasting; using AI-generated simulated passive microwave data to examine tropical cyclone structure; machine learning and tropical cyclogenesis; and using machine learning to estimate tropical cyclone track and intensity forecast uncertainty for realtime TC forecasting operations.
  Related material: my AMS Tropical talk about TC structure and simulated passive microwave data; Mark DeMaria's AMS Tropical talk about machine learning estimates of TC track and intensity; 2022 Hollings Scholar Marshall Baldwin's talk about machine learning and tropical cyclogenesis; AI for Tropical Meteorology: Challenges and Opportunities, AI for Good Seminar Series.

Uncertainty Quantification for AI Methods in Geosciences

• Neural networks are becoming a popular analysis and prediction tool for many geoscience applications. However, in addition to making a prediction of the mean state, many geoscientists also want to be able to estimate the uncertainty of their predictions, in order to estimate how confident we can be in our predictions, as well as to better understand the full range of possible outcomes. We compared several common uncertainty quantification methods and their performance in geoscience tasks, such as making predictions from satellite imagery, to try to understand the pros and cons of various uncertainty quantification methods in geoscience applications.
  Related material: uncertainty quantification for neural networks paper, and code; Mark DeMaria's AMS Tropical talk about machine learning estimates of TC track and intensity.

AI and Ethics

• As AI explodes in popularity for geoscience prediction, it raises a host of questions about ethics, responsibility, and bias. We broke down different forms of bias for AI applications in earth sciences, including data biases, statistical and computational biases, human biases, and systemic and structural biases. I am especially interested in better understanding data biases and how they affect AI modeling results, particularly selection bias and processing bias, and the intersection of data biases and statistical/computational biases.
  Related material: McGovern et al. (2024).

AI and Education

• I have been involved in several efforts aimed at providing AI education for earth science topics. Currently, I am part of a team partnered with the NOAA Center for Artificial Intelligence (NCAI) that is developing a learning journey for tropical cyclones and AI. I have also served as an instructor at a summer school focused on trustworthy AI for environmental sciences, and a mentor for the ClimateChangeAI summer school.
  Related material: NCAI learning journey; McGovern et al. (2023); ClimateChangeAI summer school presentation from the "Tackling Climate Change with Machine Learning" Workshop at NeurIPS 2022.

Causal Discovery and Atmospheric and Climate Science

• As part of my Ph.D., I worked on making causality-based tools more accessible to atmospheric and climate scientists. I worked closely with colleagues in electrical and computer engineering to refine and apply causality-based methods and algorithms to climate science problems. This collaboration led to a subsequent collaboration with Professor Alex O. Gonzales of Woods Hole Oceanographic Institute.
  Related material: My recent talk about causality and climate science at the Aspen Global Change Institute Workshop on Earth System Modeling with Machine Learning and Big Data, and the subsequent paper; Gonzalez et al. (2022); Samarasinghe et al. (2019).

Predicting Extreme Sea Ice Loss on Day-to-Day Timescales

• For my postdoc at the University of Washington, I worked on forecasting short-term predictions of sea ice loss, particularly extreme sea ice loss, and sea ice loss related to Arctic cylcones. I am not currently working on sea ice prediction, but this project was a great way to explore my interests predicting extremes, subseasonal-to-seasonal forecasting, and polar climate.
  Related material: McGraw et al. (2022); Clancy et al. (2021).

Ph.D. Research

• During my Ph.D., I was able to explore a number of topics related to large scale atmospheric dynamics, climate dynamics, and statistical modeling. Much of my Ph.D. work was focused on understanding the relationships between Arctic and midlatitude weather and climate using a variety of statistical and causal-discovery based methods. In the meantime, I got to explore midlatitude moisture transport into the Arctic; midlatitude jet variability and climate change; and volcanic eruptions and atmospheric circulation.
  Related material: McGraw and Barnes (2020), Samarasinghe et al. (2019), McGraw and Barnes (2018); McGraw et al. (2020); Woollings et al. (2017), McGraw and Barnes (2016); McGraw et al. (2016).