View this email in your browser

Follow Us

The 2024 SASIP annual meeting was hosted in last June by the University of Bologna, Italy. This event gathered 60 participants from all the institutions of the SASIP consortium as well as external speakers included members from the UCLouvain, NASA GISS and CMCC. It was an excellent opportunity to share latest research and inspiring discussions about sea ice!

Recordings of the talks are available on our YouTube channel.

Science Spotlight

Towards diffusion models for large-scale sea-ice modelling

In this new article accepted at the ICML 2024 workshop on “Machine Learning for Earth System Modelling”, Tobias Finn et al. show that the popular latent diffusion models reintroduce problems with smoothing into the data generation while they allow to easily incorporate physical bounds.

Tobias Finn will present this work in a poster session on July 26 at the ICML 2024 workshop in Vienna, Austria.

Figure 1. Samples generated with a diffusion model in data space (a–d) and a latent diffusion model with censoring (e–h). The thickness (a, e) and concentration (b, f) are directly generated, while the speed (c, g) and deformation (d, h) are derived from the velocities. Note, for visualisation purpose, only the central Arctic is shown while the whole Arctic is modelled. The remaining noise might be caused by using only 20 integration steps with a Heun sampler

Science Updates

Introducing NEDAS: a Light-weight and Scalable Python Solution for Ensemble Data Assimilation

A Python software, NEDAS, is introduced to provide a rapid testing environment for ensemble data assimilation algorithms in real models, such as the neXtSIM. This new preprint Submitted by Yue Ying to the Journal of Advances in Modeling Earth System demonstrates the computational efficiency of the software and algorithmic flexibility due to its modular design.

Figure 2. Memory layout for state variables (square pixels) and observations (circles) using (a) batch and (b) serial assimilation strategies. The colors of each state/observation represent the processor (p) that stores the data in its local memory.

BBM rheology successfully running in NEMO4.2

As part of WP5, BBM rheology is now successfully running in NEMO4.2 at a global scale with a 1/4° resolution. The results are consistent with the "standard" AEVP rheology but exhibit significantly more cracks and sea ice openings in both the Arctic and Antarctic. Current efforts focus on refining the reference simulation by tuning ocean and sea ice parameters to increase confidence in the comparison. Stay tuned for exciting results as BBM rheology will be integrated into the fully coupled NEMO4.2/ARPEGE model in the coming months!

Comparison of simulated sea ice concentration in % between aEVP (left panel) and BBM (right panel) rheology for the Arctic Ocean (top row) and Antarctic (bottom row) the month of September. Reference: Romain Bourdallé-Badie from Mercator Ocean International, France.  

News

🧑‍🏫 ICCS 2024 workshop on "Machine Learning and Data Assimilation for Dynamical Systems” on July 3 in Malaga, Spain:

• Ivo Pasmans will talk about the Ensemble Kalman filter in latent space using a variational auto encoder pair.

  
  

• Simon Driscoll will present his work on Emulating melt pond on sea ice using neural networks

👨‍💻SASIP June webinar "The nextSIM-DG sea ice model: rewriting, refactoring and reusing" by Tim Spain is available on our Youtube Channel!

The next webinar will be in September. More information to follow, stay tuned!

Vacancies 👥

A postdoc position at CECI/CERFACS in Toulouse, France, to assess the climate relevance of integrating a newly developed sea ice model based on Brittle Bingham-Maxwell (BBM) rheology into a fully coupled climate model, as part of WP5.

👉 See more information about this job offer.

Application deadline is September 3!