Modelling basal melt of North Greenland ice shelves using machine learning

6 months from mid-February 2023

In Greenland and Antarctica, glaciers flow from the interior of the continent to the ocean where they begin to float to form ice shelves. These extensions are key elements of the polar ice sheets because they regulate the flow of ice coming from the interior. International intercomparison experiments [1] have shown that a disappearance of the Antarctic ice shelves would lead to a sea level rise of several meters. Collapses of ice shelves have already been observed with the Larsens in West Antarctica, and in Greenland (Zachariae Isstrøm ice shelves). In each of these cases, the ice discharge from the glaciers was multiplied by a factor >3 [2][3], hence increasing their sea level contributions. Predicting the evolution of ice shelves is therefore a crucial issue if we want to reduce uncertainties in the future evolution of sea level and anticipate the effect of climate warming. Representing ice shelves and their evolution in numerical models is challenging because it requires to constrain a set of complex processes at the ice-ocean interface [4] . The use of machine learning models has made tremendous progress in glaciology and oceanography. Despite the difficulties of interpretation of these models, their power lies in the possibility of freeing themselves from the limiting hypotheses of parameterizations, while capturing highly non-linear processes [5]. Despite their strong potential, these approaches are largely under-used for modeling ice shelf weakening processes and require learning based on a relatively large number of observations before projections can be made. Thus, the implementation of these methodologies is all the more justified with the arrival of increasingly massive amounts of observations from space-based remote sensing [6][7].

The aim of this internship will be to model the basal melt of the last ice shelves present in North Greenland using machine learning. The student will investigate the spatio-temporal variability of the basal melting under the ice shelves, from ocean temperature and salinity obtained via in-situ data, re-analysis model, and from the bathymetry of the sea floor and surface runoff. We will evaluate several approaches, starting with standard supervised machine learning methods (e.g. Random Forest, XGBoost), which will use observations of the ice shelves as input. We will also consider the use of a convolutional neural network (CNN), a type of model that is particularly well-suited to simulating data with strong spatial dependencies.

Requirements :

MSc student in computer science or earth sciences
EU nationality
Good programming skills in Python
Knowledge in machine learning (e.g. sklearn, Tensorflow or Pytorch)
Interest in glaciology and/or climate science is a plus
We welcome students from diverse backgrounds and minorities

Internship details and application :

The internship is for a duration of a minimum of 6 months, paid at the standard fee established by French law (ca. 600€ net/month), and will be held at the IGE laboratory in Grenoble (France), supervised by Romain Millan (romi ign.ku.dk), Jordi Bolibar (j.bolibar uu.nl) and Jean Baptiste Barré (jean-baptiste.barre univ-grenoble-alpes.fr). Please contact us by email with a CV and a motivation letter if interested.

References :

[1] Sun, S., Pattyn, F., Simon, E., Albrecht, T., Cornford, S., Calov, R., . . . Zhang, T. (2020). Antarctic ice sheet response to sudden and sustained ice-shelf collapse (ABUMIP). Journal of Glaciology, 66(260), 891-904. doi:10.1017/jog.2020.67
[2] Rignot, E., Casassa, G., Gogineni, P., Krabill, W., Rivera, A., and Thomas, R. (2004), Accelerated ice discharge from the Antarctic Peninsula following the collapse of Larsen B ice shelf, Geophys. Res. Lett., 31, L18401, doi:10.1029/2004GL020697.
[3] Mouginot, J., Rignot, E., and Scheuchl, B. (2014), Sustained increase in ice discharge from the Amundsen Sea Embayment, West Antarctica, from 1973 to 2013, Geophys. Res. Lett., 41, 1576– 1584, doi:10.1002/2013GL059069.
[4] Jourdain, N. C., Asay-Davis, X., Hattermann, T., Straneo, F., Seroussi, H., Little, C. M., and Nowicki, S. : A protocol for calculating basal melt rates in the ISMIP6 Antarctic ice sheet projections, The Cryosphere, 14, 3111–3134, https://doi.org/10.5194/tc-14-3111-2020, 2020.
[5] Bolibar, J., Rabatel, A., Gouttevin, I., Zekollari, H., & Galiez, C. (2022). Nonlinear sensitivity of glacier mass balance to future climate change unveiled by deep learning. Nature Communications, 13(1), 409. https://doi.org/10.1038/s41467-022-28033-0
[6] Mouginot, J., Rignot, E., Bjørk, A. A., van den Broeke, M., Millan, R., Morlighem, M., et al. (2019). Forty-six years of Greenland Ice Sheet mass balance from 1972 to 2018. Proceedings of the National Academy of Sciences, 116(19), 9239–9244. https://doi.org/10.1073/pnas.1904242116
[7] Millan, R., Mouginot, J., Derkacheva, A., Rignot, E., Milillo, P., Ciraci, E., et al. (2022). Ongoing grounding line retreat and fracturing initiated at the Petermann Glacier ice shelf, Greenland, after 2016. The Cryosphere, 16(7), 3021–3031. https://doi.org/10.5194/tc-16-3021-2022