The Arctic region plays a vital role in the global climate system, it is strongly affected by climate change, and in turn one of its drivers. The Arctic is warming four times faster than the global average, transitioning into an entirely different climate than a few decades ago (Legg, 2021). Satellite data reveal that the September sea ice extent declined by ~13% per decade since 1979 causing major changes in the oceanic heat flux. Changes in the Arctic sea ice impact extreme weather and climate events beyond the Arctic region, favouring extreme Northern Hemisphere winters (Kretschmer et al., 2016) or wetter European summers (Screen, 2013). Arctic changes have a substantial socio-economical relevance (e.g. indigenous communities, shipping and tourism, fisheries), as well as a geopolitical dimension, given possible shipping routes and natural resources exploitation. Understanding the causes of these changes is thus of paramount importance yet substantial gaps still exist. Moreover, numerous studies have demonstrated the limitations of current-generation climate models in accurately representing essential aspects of polar climates, such as Arctic sea ice loss (Wang et al., 2016) or water mass changes (Ilicak et al., 2016).
24 months from 28/09/2023 to 27/09/2025
General aims
MELTED’s goal is to improve our understanding of ice-related phenomena in the Arctic region by integrating physics-based models with ML to build more skilful forecasts over different spatio-temporal scales. MELTED’s impact is not limited to scientific and theoretical aspects. The improvement in the forecasting ability is instrumental to supporting practical applications and opening new operational perspectives in the Arctic, such as the efficient management of new commercial shipping routes that will help reduce
global trade CO2 emissions.
CMCC role
WP0-WP1-WP3
Activities
1. Identification of Arctic dataset and climate linkage: to identify and synthesize existing and emerging in-situ and remote sensing datasets of the Arctic sea ice properties to boost process-level understanding of the Arctic climate system and identify possible drivers influencing its (thermo)dynamics variability.
2. Correcting sea-ice predictions with hybrid physical-data-centric DA-ML: to train an ML algorithm using the analysis increments in the ORAs, given as targets, and translate the learnt model error into a correction in prediction mode.
Expected results
1. Data provision for ML applications.
2. DA/ML methods to correct model error in sea-ice models
Partners
Fondazione CMCC
Università degli studi di Bologna