University of Washington

eScience News

Events & Seminars

  1. Using Data Science /
  2. Data Science Incubator /
  3. Incubator Project Archives /
  4. Investigating seasonal glacier fluctuations in northeast greenland using machine learning and explainable artificial intelligence
Multiyear (2010-2022) mean ice velocity (m/yr) at northernmost 79N and ZI. Evolving terminus positions are plotted from 1948 (purple) to 2019 (yellow), with a black line tracing 79N’s current grounding line location. (Right) Velocities observed at a downstream and slower upstream location are plotted for both 79N (purple) and ZI (green) glaciers. A gray bar highlights the period corresponding to ZI’s final ice tongue collapse. Note the enhanced seasonal amplitude at ZI relative to time series at 79N.

Investigating Seasonal Glacier Fluctuations in Northeast Greenland Using Machine Learning and Explainable Artificial Intelligence

Project Lead: Claire Jensen, UW Earth and Space Sciences

Data Science Lead: Scott Henderson

The Greenland Ice Sheet (GrIS) is losing mass due, in part, to the recent speedup of many of the outlet glaciers that line the ice sheet. Mass loss from outlet glaciers can be characterized by (1) retreat of the glacier’s frontal position, or terminus, via calving and ocean melting, and (2) speedup, characterized by an increase in ice velocity, leading to ice draining from the ice sheet more quickly. One of the most dramatic examples of outlet glacier change is observed at Zachariæ Isstrøm (ZI). ZI is one of three large glaciers that comprise the Northeast Greenland Ice Stream (NEGIS), a notable feature of fast-flowing ice that drains ~12% of the ice sheet by area. After its floating ice tongue collapsed in 2013, ZI underwent pronounced seasonal changes, causing ZI to lose more mass due to glacial retreat than any other outlet glacier in Greenland.

Multiple environmental and glacial processes likely contribute to the observed seasonality, and these relationships must be well constrained before future projections of seasonality can be made. These variables include (1) the summertime production and discharge of meltwater, which can impact glacier sliding speeds as it drains and alters pressure at the ice bed, and (2) proglacial mélange (the mixture of calved icebergs and sea ice that can form in glacier-fed fjords) which can impact the timing of glacier advance and retreat by buttressing the terminus. However, the relative impact of these variables on glacier flow and factors determining their impact remains unclear.

Understanding fluctuations in seasonality in relation to terminus changes and due to meltwater and mélange forcing is necessary for understanding ZI’s changing dynamics and predicting future mass loss. My work focuses on ZI and its NEGIS neighbor, Nioghalvfjerdsfjorden (79N), which is one of the few Greenland glaciers with a long (~80 km) floating ice tongue. I use machine learning and explainable artificial intelligence to quantify and compare the sensitivity of ZI and 79N, two glaciers in the same region with vastly different geometries and seasonalities. I aim to provide a deeper understanding of the evolving sensitivity and response of Greenland outlet glaciers to changing environmental conditions.