Research

The topography, geology, and hydrology beneath the ice influences how glaciers behave. However, these conditions are often overlain by several kilometers of ice, making them difficult to observe. I use geophysical observations and machine learning techniques to study this region.

My research includes developing advanced geostatistical simulation methods and integrating geostatistics and other machine learning techniques with geophysical knowledge to characterize subglacial conditions. I am also interested in the role of topographic and geologic controls on ice sheet dynamics and evolution. My goal is to advance the use of machine learning methods for studying glaciers and make these tools accessible to the broader cryosphere community.

GlacierStats Software

GlacierStats is a collection of Python functions and demos specifically designed for stochastic methods in glaciology. In my own research, I have found that geostatistical tools designed for industry applications do not have the flexibility to address the unique combination of challenges in ice sheet problems.

These tools are part of our ongoing effort to develop and adapt geostatistical techniques and other machine learning methods to glaciology. Our goal is to make these methods more accessible.

​

Questions? Suggestions? Feel free to reach out!

​

Recent Publications

Stochastic modeling of subglacial topography exposes uncertainty in water routing at Jakobshavn Glacier

​

The topography beneath glaciers controls the flow of subglacial water. Traditional approaches for estimating the topography produce topography that is unrealistically smooth, which can bias hydrological models. In this project, we used stochastic methods to generate many realistically rough topographic realizations to use for hydrological modeling. We found that the water flow behavior is highly variable, and that traditional topographic estimation techniques cannot address topographic uncertainty in hydrologic models.

Antarctic Topographic Realizations and Geostatistical Modeling Used to Map Subglacial Lakes

​

We used machine learning techniques to predict the locations of subglacial lakes in Antarctica. Our analysis shows that hydrologically stable and active lakes are likely found in different types of environments. We modified a commonly used, unrealistically smooth topographic map of Antarctica to have realistic roughness, and found that this increases the expected number of lakes by an order of magnitude.

Photos