3D Cloud reconstruction through geospatially-aware Masked Autoencoders

(2025)

Authors:

Stella Girtsou, Emiliano Diaz Salas-Porras, Lilli Freischem, Joppe Massant, Kyriaki-Margarita Bintsi, Guiseppe Castiglione, William Jones, Michael Eisinger, Emmanuel Johnson, Anna Jungbluth

Statistical constraints on climate model parameters using a scalable cloud-based inference framework – CORRIGENDUM

Environmental Data Science Cambridge University Press (CUP) 4 (2025)

Authors:

James Carzon, Bruno Abreu, Leighton Regayre, Kenneth Carslaw, Lucia Deaconu, Philip Stier, Hamish Gordon, Mikael Kuusela

3D Cloud reconstruction through geospatially-aware Masked Autoencoders

Workshop paper at “Machine Learning and the Physical Sciences”, NeurIPS (2024)

Authors:

Stella Girtsou, Emiliano Diaz Salas-Porras, Lilli J Freischem, Joppe Massant, Kyriaki-Margarita Bintsi, Guiseppe Castiglione, William Jones, Michael Eisinger, Emmanuel Johnson, Anna Jungbluth

Abstract:

Clouds play a key role in Earth's radiation balance with complex effects that introduce large uncertainties into climate models. Real-time 3D cloud data is essential for improving climate predictions. This study leverages geostationary imagery from MSG/SEVIRI and radar reflectivity measurements of cloud profiles from CloudSat/CPR to reconstruct 3D cloud structures. We first apply self-supervised learning (SSL) methods-Masked Autoencoders (MAE) and geospatially-aware SatMAE on unlabelled MSG images, and then fine-tune our models on matched image-profile pairs. Our approach outperforms state-of-the-art methods like U-Nets, and our geospatial encoding further improves prediction results, demonstrating the potential of SSL for cloud reconstruction.

Weak liquid water path response in ship tracks

Atmospheric Chemistry and Physics European Geosciences Union 24:23 (2024) 13269-13283

Authors:

Anna Tippett, Edward Gryspeerdt, Peter Manshausen, Philip Stier, Tristan WP Smith

Abstract:

The assessment of aerosol–cloud interactions remains a major source of uncertainty in understanding climate change, partly due to the difficulty in making accurate observations of aerosol impacts on clouds. Ships can release large numbers of aerosols that serve as cloud condensation nuclei, which can create artificially brightened clouds known as ship tracks. These aerosol emissions offer a “natural”, or “opportunistic”, experiment to explore aerosol effects on clouds, while also disentangling meteorological influences. Utilizing ship positions and reanalysis wind fields, we predict ship track locations, colocating them with satellite data to depict the temporal evolution of cloud properties after an aerosol perturbation. Repeating our analysis for a null experiment does not necessarily recover zero signal as expected; instead, it reveals subtleties between different null-experiment methodologies. This study uncovers a systematic bias in prior ship track research, due to the assumption that background gradients will, on average, be linear. We correct for this bias, which is linked to the correlation between wind fields and cloud properties, to reveal the true ship track response.

We find that, once this bias is corrected for, the liquid water path (LWP) response after an aerosol perturbation is weak on average. This has important implications for estimates of radiative forcings due to LWP adjustments, as previous responses in unstable cases were overestimated. A noticeable LWP response is only recovered in specific cases, such as marine stratocumulus clouds, where a positive LWP response is found in precipitating or clean clouds. This work highlights subtleties in the analysis of isolated opportunistic experiments, reconciling differences in the LWP response to aerosols reported in previous studies.

Glaciation of liquid clouds, snowfall and reduced cloud cover at industrial aerosol hot spots

Science American Association for the Advancement of Science 386:6723 (2024) 756-762

Authors:

Velle Toll, Jorma Rahu, Hannes Keernik, Heido Trofimov, Tanel Voormansik, Peter Manshausen, Emma Hung, Daniel Michelson, Matthew Christensen, Piia Post, Heikki Junninen, Benjamin J Murray, Ulrike Lohmann, Duncan Watson-Parris, Philip Stier, Norman Donaldson, Trude Storelvmo, Markku Kulmala, Nicolas Bellouin

Abstract:

The ability of anthropogenic aerosols to freeze supercooled cloud droplets remains debated. In this work, we present observational evidence for the glaciation of supercooled liquid-water clouds at industrial aerosol hot spots at temperatures between −10° and −24°C. Compared with the nearby liquid-water clouds, shortwave reflectance was reduced by 14% and longwave radiance was increased by 4% in the glaciation-affected regions. There was an 8% reduction in cloud cover and an 18% reduction in cloud optical thickness. Additionally, daily glaciation-induced snowfall accumulations reached 15 millimeters. Glaciation events downwind of industrial aerosol hot spots indicate that anthropogenic aerosols likely serve as ice-nucleating particles. However, rare glaciation events downwind of nuclear power plants indicate that factors other than aerosol emissions may also play a role in the observed glaciation events.