Abstract

Polar science studies many aspects of the north and south poles, and glaciers. In order to better understand the internal structure of ice sheets and how they are impacted by climate change, ice penetrating radar is used. This radar captures the layers inside the ice sheet which can then be used to understand layer behavior and flow inside the glacier. Layers can be picked out of these radargrams and then visualized to give an indication of the internal topology. We explore two areas in which these radargrams are used, namely analysis of machine-picked layers and immersive visualization of these layers. Analysis of machine picked layers is achieved through the development of quality metrics which are independent of number of layers. We present these metrics and analysis of their utility. Immersion is accomplished through WebXR, inspired by recent research in both polar science and other scientific fields. We present the system, its controls and limitations, and evaluations to be performed on this system. 

Committee Members

Professor Don Engel (Co-chair), Professor Rebecca Williams (Co-chair), Professor Tejas Gokhale

Image Note: In the left panel, Naomi Tack is defending her dissertation. In the right panel, Naomi stands with her iHARP mentors, (left to right: Dr. Becca Williams, Naomi Tack, and Dr. Don Engel).

iHARP’s Emam Hossain, UMBC PhD candidate, presented his paper "Incorporating Causality with Deep Learning in Predicting Short-term and Seasonal Sea Ice" at the American Meteorological Society’s 23rd Conference on Artificial Intelligence for Environmental Science, 104th AMS Annual Meeting, 2024.  Emam’s research highlights the importance of Arctic sea ice (ASI) in controlling global warming and the concerning trend of its decline. It underscores the need for accurate forecasting to address various socio-economic issues. While current correlation-based machine learning (ML) models have limited capabilities, causal models offer promise by considering cause-and-effect relationships between atmospheric variables. The study aims to predict ASI using causal features identified by Granger Causality (GC) and PCMCI+ algorithms. Results show that models trained on causal features outperform those trained on correlated features. The study seeks to develop a deep learning model incorporating causality for both short-term and long-term sea ice prediction, with potential for further improvement by considering exact time lags identified by the causal algorithm.

We congratulate Emam for presenting and representing at AMS to share his research and knowledge on data science, polar science, and data analysis!

To learn more about iHARP, please visit us at iharp.umbc.edu

iHARP is proud to announce Chhaya Kulkarni, UMBC PhD Candidate, will be a part of the GESTAR Fellowship! Through this fellowship, she will gain experience working with NASA's Global Modeling and Analysis Office (GMAO) under the mentorship of Dr. Nikki Prive. NASA's GMAO is responsible for developing and distributing a variety of products that either incorporate NASA's satellite observations, aid in the development of new missions, or assist with satellite missions and field activities. 

Through the simulated data from the observational systems, Chhaya will apply and evaluate her model with the help of her skills in data analysis, visualization, and storytelling. By utilizing the skills she has gained from this fellowship, Chhaya plans to improve her research skills, learn about new cutting-edge technologies so that she may produce high-quality papers, and use this research to further integrate data science and polar science in future research.