A research team developed a novel deep learning model that made it possible to create a high-resolution dataset of 149 marine-terminating glaciers on Svalbard to track ice loss at their calving fronts.
Svalbard is one of the regions that is warming many times faster than the rest of the world. The fast-rising temperatures are particularly affecting the glaciers, which are rapidly losing mass. However, the mechanisms behind the ice loss have not yet been properly understood, especially the calving dynamics of marine-terminating glaciers.
In Earth System Science Data on February 20, a research team from the University of Bristol and the Technical University of Munich published a unique dataset comprising almost 125,000 satellite images and visualizing the positions of the calving fronts of 149 tidewater glaciers on Svalbard in the period from 1985 to 2023 – an important tool for researchers to answer the open questions about glacier calving.
In a press release from the EU project Arctic PASSION, Dr. Tian Li, researcher at the Bristol Glaciology Centre and lead author of the study, says that “this dataset can be used to improve the mass balance assessments for Svalbard tidewater glaciers. Additionally, it enables the exploration of the drivers and processes controlling glacier calving. This is crucial for understanding the calving dynamics, a key indicator of how glaciers respond to climate change.”
Based on the extensive satellite data catalog, the researchers used an automated process with deep learning and neural networks to determine the seasonal and annual fluctuations in the calving front with an average temporal resolution of just four days.
Apart from a handful of glaciers on Nordaustlandet – the second largest island of Svalbard, located in the northeast of Spitsbergen – all the glaciers studied are retreating, with accelerated mass loss in recent decades.
In addition, the research team also identified so-called “surging events”, in which a glacier flows faster within a short period of time and thus more ice is lost from its front.
In earlier studies, ice loss at the glacier fronts was usually not taken into account because there was simply a lack of data. The newly developed data set now makes it possible to better understand and forecast future glacier retreat in the Arctic, which is crucial for predicting future sea level rise. The research team plans to apply the new method to all marine-terminating glaciers in the Arctic.
The research team developed the novel and state-of-the-art data set as part of the EU project Arctic PASSION, which aims to create a pan-Arctic observation system for the most important climate variables of the Arctic cryosphere.
Julia Hager, Polar Journal AG
More about this topic:
