Large carbon dioxide emissions avoidance potential in improved commercial air transport efficiency

Communications Earth and Environment 7:1 (2026)

Authors:

S Gössling, M Klöwer, JC Leitão, S Hirsch, D Brockhagen, A Humpe

Abstract:

Aviation’s climate impact continues to grow, with little progress toward emission reductions aligned with global targets. While technological advances attract attention, operational efficiency across aircraft, airlines, airports, city pairs, and regions remains underexplored. Here we assess carbon dioxide efficiency for 27.5 million flights between 26,156 city pairs in 2023, using data from Airline Data, International Civil Aviation Organization, International Air Transport Association. Results show wide variation: 32–890 gram carbon dioxide per revenue passenger kilometres across routes and 60–360 gram carbon dioxide per revenue passenger kilometre across aircraft models. Efficiency differs by region and is lowest in Africa, Australia, and Norway, and highest in Brazil, India, and Southeast Asia. Operating all routes at their demonstrated optimum could cut emissions by 10.7%. A theoretical 50% reduction is possible with the most efficient aircraft, all-economy layouts, and 95% load factors. Efficiency-focused policy could swiftly reduce fuel use without limiting air transport capacity.

Enhancing maritime domain awareness using latent space representations

Artificial Intelligence for Security and Defence Applications IV SPIE (2026) 3-3

Authors:

Rikard Vinge, Miguel A Belenguer-Plomer, Isabelle Wittmann, Octavian Dumitru, Michele Lazzarini, Sergio Albani, Conrad M Albrecht

Abstract:

This work presents a novel application of Earth Observation (EO) embeddings for Maritime Domain Awareness (MDA) from the perspective of the geospatial foundation model <i>TerraMind</i>. By operating directly on compressed latent-space representations rather than on the original Sentinel-1 imagery, this approach directly addresses the data gravity bottleneck that increasingly constrains operational EO processing, in line with the objectives of the Embed2Scale project. In particular, linear and non-linear probing of compressed <i>embeddings</i>—latent space representations of synthetic aperture radar (SAR) imagery—is used to detect, localise, and characterise vessels from Sentinel-1 GRD images. Vessel detection achieves a ROC-AUC of 0.95, while vessel localisation has a median error of around 300 m for vessels longer than 100 m, and vessel AIS type-code classification achieves 62% accuracy for the majority classes. The latter, however, is highly correlated with vessel length. These results reveal both the potential and current limitations of Geospatial Foundation Models (GeoFMs) when applied to fine-grained MDA understanding in real-world scenarios. Additionally, a new maritime remote sensing dataset is released, pairing about 2,000 Sentinel-1 images with temporally aligned AIS signals from commercial vessels. Through manual inspection, each candidate sample was verified for bounding-box quality and for consistency between the AIS-reported vessel length and the observed vessel extent, cross-checking vessel identity and location against publicly available vessel records and, where available, photographs. The dataset was made ready for use with the <i>NeuCo-Bench</i> neural compression benchmark framework, publicly available on HuggingFace through https://huggingface.co/datasets/ embed2scale/S1-single-ship.

Steering Diffusion Models to Rare Events with Sequential Monte Carlo

(2026)

Authors:

Aavash Subedi, Tim Reichelt, Christopher Williams, Philip Stier, Yee Whye Teh, Saifuddin Syed

A neural network-based Universal Thermal Climate Index for reliable global thermal-stress classification across extreme weather

ArXiv 2609.35949 (2026)

Authors:

Bikem Pastine, Milan Klöwer, Tianning Tang, Sarah Wilson Kemsley, Louise Slater

Sensitivity of cloud structure and precipitation to cloud microphysics schemes in ICON and implications for global km-scale simulations

Geoscientific Model Development Copernicus GmbH 19:17 (2026) 8191-8212

Authors:

Maor Sela, Philipp Weiss, Philip Stier

Abstract:

<jats:p>Abstract. Cloud microphysics remains a major source of uncertainty in km-scale atmospheric models. While cloud-resolving models have advanced our understanding of cloud-climate interactions, their predictability remains limited. Most studies have examined either microphysics schemes or domain-size sensitivities, but their interactions are poorly understood. This study examines cloud structure and precipitation sensitivity to microphysics schemes and how they vary between regional and global configurations within a single, consistent modelling framework. We analyse three convection-permitting simulations over the Amazon: two regional runs employing single- and double-moment microphysics schemes and a global single-moment run, with all other configurations consistent. We find that cloud hydrometeor characteristics are sensitive to the microphysics scheme. Specifically, the double-moment scheme produces up to five times more graupel and twice as much rain, but half as much cloud water and one-fifth as much fog as the single-moment scheme. Despite these differences, precipitation, water vapour, and outgoing longwave radiation remain consistent across schemes, suggesting large-scale constraints primarily govern integrated quantities. Furthermore, domain configuration further amplifies sensitivities. The global simulation exhibits up to 150 % more fog and nearly double the cloud ice compared to the regional single-moment run, highlighting the role of large-scale circulation and lateral boundary conditions. These findings demonstrate that microphysics schemes primarily influence cloud processes, while the domain setup determines how these sensitivities manifest. Improved observational constraints and perturbed-parameter ensembles are therefore needed to evaluate model performance, assess the broader generalisability of these findings and separate tuning effects and structural uncertainty.</jats:p>