Monitoring and forecasting the state of Svalbard’s cryosphere using a digital twin (SvalbardDT)

In this contribution, we will demonstrate the first prototype of the Svalbard cryosphere digital twin (SvalbardDT v1; https://svalbarddt.org/), which is a Digital Twin Component of ESAs Digital Twin Earth (DTE). An environmental digital twin observes a physical entity (i.e. the cryosphere), fuses multi-modal observations together, then generates what if scenarios to improve human decision-making. In this way, there are two-way flows of information (i.e. data) between the physical and digital assets. In Svalbard, where rates of warming are six times faster than the global average, digital twins have the potential to be used by researchers to analyse trends in the cryosphere, by local communities to improve navigation, and by policy-makers to improve decision-making. Furthermore, Svalbard is considered a ‘super site’ of in situ observations in a pan-Arctic context owing to the permanent infrastructure and long history of international scientific collaboration on the archipelago, hence it is a prime location to develop Arctic digital twin technology. Here, we will provide a demonstration of the capabilities of SvalbardDT version 1 as a new tool for monitoring Svalbard’s cryosphere in the 21^st century (2010-present), the underlying architecture, and the next steps towards full operationalisation.

SvalbardDT represents a digital twin of Svalbard’s cryosphere covering glaciers, snow and sea ice observed through a combination of Earth Observation data sets and reanalysis data products. These data products are multi-modal i.e. they are collected at different resolutions, scales and time/spatial periods. SvalbardDT fine-tunes a deep learning foundational model to ingest the relevant data products and harmonise them into a 4D data cube describing the data set variable, x-dimension, y-dimension, and its changes over time. This produces weekly to monthly data cubes describing ~21 parameters. SvalbardDT focuses on the application of two case studies which are accessed through an online dashboard: (1) exploring the current state of cryosphere conditions in Svalbard to simulate terrestrial and marine navigation routes across the archipelago; and (2) forecast extreme Rain on Snow and Ice (ROSI) events using an AI foundation model. The associated toolboxes help to improve our understanding of the interconnecting processes shaping glaciers and ice caps across Svalbard.