Overview and Background
The drainage of supraglacial lakes plays an important role in modulating ice velocity, and thus the mass balance, of the Greenland Ice Sheet. To date, research has primarily focused on drainage events and their impacts on ice motion during the summer melt season, but recent research has shown that drainage events during winter can also affect ice dynamics. However, current approaches are limited to a single season and/or 1–2 satellite sensors, limiting observations when year-round and multi-sensor monitoring are required to fully understand lake processes and their impacts. This PhD will utilise petabytes of available Earth Observation data and exascale computing to perform ice-sheet–scale analysis of year-round supraglacial lake drainage events on the Greenland Ice Sheet and produce scalable workflows that can be used to assess the impact of lake drainage events in other glaciological environments.
Methodology and Objectives
This PhD project will advance capabilities in the detection of supraglacial lake drainage events on the Greenland Ice Sheet (GrIS) and assess the impact of these drainage events on ice dynamics. The PhD will commence with two exploratory projects (~6 months each), providing complementary experience in deploying exascale compute and performing big-data analysis.
Teaser Project 1: Near-real-time, automated, multi-sensor, year-round monitoring of supraglacial lake drainage
Current assessments of supraglacial lake drainage on the GrIS are largely restricted to a single season and/or satellite sensor, particularly during winter, where existing approaches remain constrained to low-volume pipelines that analyse single orbits and provide limited temporal and spatial coverage. However, the plethora of remotely sensed imagery now available provides the opportunity to detect supraglacial lake drainage events at high temporal and spatial resolution across the entire ice sheet in near real-time through exascale big-data analysis. This project will scale a SAR-based methodology for supraglacial lake drainage detection to ice-sheet-wide monitoring, using a high-data-volume approach to achieve near-daily temporal resolution by leveraging all available orbits and both C- and L-band SAR. The project will exploit access to exascale compute to deploy GPU-accelerated machine learning methods (e.g., convnets or Unets), able to extract spatiotemporal patterns from large and complex volumes of multi-frequency inputs, to support robust, scalable detection of drainage events across diverse glaciological settings. Validation and training will draw on timestamped ArcticDEM strips and ICESat-2 altimetry, ensuring reliable accuracy assessments at ice-sheet scale, using methods for (cross-) validation across space and time (e.g., Otto et al. 2024).
Teaser Project 2: Near-real-time evaluation of the impact of supraglacial lake drainage on the GrIS
As supraglacial lakes form increasingly farther inland under increasing atmospheric temperatures, the year-round impact of their drainage events on ice dynamics is not yet well understood. Recent research has shown that winter drainage events are numerous, often occur as “cascade events”, and can result in short-term increases in ice velocity (Dean et al., under review). However, a systematic, year-round analysis of the impact of supraglacial lake drainage events on ice dynamics across the GrIS has not yet been undertaken. This project will set up a pipeline on a sub-region of the ice sheet to analyse the impact of drainage events on ice dynamics in real-time, allowing later scaling up to an ice-sheet-scale. Access to exascale compute will enable comparison of the database of drainage events created for Project 1 with climate and glaciological data (e.g., temperature, precipitation, surface energy balance, ice surface velocity, ice thickness, and bed elevation). By applying scalable statistical methods like changepoint analysis (offline detection), statistical process monitoring (online surveillance), and anomaly detection using deep learning methods, the impact of lake drainage events will be evaluated in real-time and assessed over a range of timescales. Additionally, the trained DNNs from Project 1 can be used for dimensionality reduction and process monitoring based on data depths, which can indicate further sources/reasons for detected changes (Malinovskaya et al. 2024).
Long-term pathway and objectives
At the end of Year 1, the student will select a pathway for further development. If Project 1 is chosen, the PhD will focus on leveraging additional compute resources and machine learning methods to scale the analysis to include additional, larger, and multi-modal data sources in our drainage detection methodology (such as optical to enhance summer detection) and deploying our method over other regions, such as Antarctic ice shelves, where real-time monitoring of supraglacial lake drainage events and cascades could be a useful precursor for forecasting ice-shelf disintegration (Banwell et al., 2013). If Project 2 is chosen, the PhD will focus on exploiting access to exascale compute and machine learning methods to scale the analysis to an ice-sheet-wide scale, enabling near-real-time assessment of the impact of supraglacial lake drainage events on the GrIS and potentially other glaciological environments.
For either pathway, the aims of the PhD are to:
- Advance understanding of year-round supraglacial lake drainage events on the Greenland Ice Sheet.
- Exploit exascale compute resources and machine learning to enable real-time detection of supraglacial lake drainage events at an unprecedented scale and assess their impacts.
- Produce scalable workflows that can be applied to supraglacial lake drainage events in other glaciological regions.
References & Further Reading
Banwell, A. et al., (2013), Breakup of the Larsen B Ice Shelf triggered by chain reaction drainage of supraglacial lakes, Geophysical Research Letters, 40, 22, 5872-5876, doi.org/10.1002/2013GL057694
Christoffersen, P. et al., (2018), Cascading lake drainage on the Greenland Ice Sheet triggered by tensile shock and fracture, Nature Communications, 9, 1064, doi.org/10.1038/s41467-018-03420-8
Dean, C. et al. (under review), A decade of winter supraglacial lake drainage across Northeast Greenland using C-band SAR, The Cryosphere Discussions
Dunmire, D. et al., (2025), Greenland Ice Sheet wide supraglacial lake evolution and dynamics: insights from the 2018 and 2019 melt seasons, Earth and Space Science, 12, 2, doi.org/10.1029/2024EA003793
Leeson, A. et al., (2015), Supraglacial lakes on the Greenland ice sheet advance inland under warming climate, Nature Climate Change, 5, 51–55, doi.org/10.1038/nclimate2463
Malinovskaya, A., Mozharovskyi, P., & Otto, P. (2024). Statistical process monitoring of artificial neural networks. Technometrics, 66(1), 104-117, doi.org/10.1080/00401706.2023.2239886
Miles, K., et al., (2017), Toward monitoring surface and subsurface lakes on the Greenland Ice Sheet using Sentinel-1 SAR and Landsat-8 OLI imagery, Frontiers in Earth Science, 5, doi.org/10.3389/feart.2017.00058
Otto, P., Fassò, A., & Maranzano, P. (2024). A review of regularised estimation methods and cross-validation in spatiotemporal statistics. Statistic Surveys, 18, 299-340, doi.org/10.1214/24-SS150
