Numerical weather prediction and climate models require continuous adaptation to take advantage of advances in high-performance computing hardware. This paper presents the port of the ICON model to GPUs using OpenACC compiler directives for numerical weather prediction applications. In the context of an end-to-end operational forecast application, we adopted a full-port strategy: the entire workflow, from physical parameterizations to data assimilation, was analyzed and ported to GPUs as needed. Performance tuning and mixed-precision optimization yield a 5.5× speed-up compared to the CPU baseline in a socket-to-socket comparison. The ported ICON model meets strict requirements for time-to-solution and meteorological quality, in order for MeteoSwiss to be the first national weather service to run ICON operationally on GPUs with its ICON-CH1-EPS and ICON-CH2-EPS ensemble forecasting systems. We discuss key performance strategies, operational challenges, and the broader implications of transitioning community models to GPU-based platforms.
Decades of orbital missions have produced multi-modal remote sensing data for the Moon, spanning optical imagery, spectroscopy, thermal emission, radar, gravity, and elemental composition. Yet these datasets remain fragmented across archives, and no benchmark exists for evaluating machine learning on lunar data. We introduce Moonstone, the first multi-modal foundation model benchmark for lunar remote sensing. Our contributions are: (1) a 28-channel, 128 pixels-per-degree ( ∼ 237 m) global lunar pretraining dataset from seven instrument families across five missions, (2) MG-MAE, a modality-grouped masked autoencoder with per-group convolutional tokenizers, a shared Vision Transformer encoder, attention masking for missing modalities, coverage-adaptive masking for heterogeneous spatial coverage, and spectral continuity regularization for physically plausible reconstructions, and (3) a benchmark of six downstream tasks covering classification, regression, and segmentation. MG-MAE pretrained features outperform scratch baselines on all tasks and surpass both ImageNet-pretrained and vanilla MAE baselines by large margins. We release the pretraining dataset, code, and the benchmark suite (Data: https://huggingface.co/datasets/ayushprd/Moonstone Code: https://github.com/ayushprd/Moonstone ).
The RAPSODI (Radiosonde Atmospheric Profiles from Ship and island platforms during ORCESTRA, collected to Decipher the ITCZ) radiosonde dataset was collected during the ORCESTRA field campaign in August and September 2024. It is designed to investigate the mechanisms linking mesoscale tropical convection to tropical waves and to air-sea heat and moisture exchanges that regulate convection and tropical cyclone formation. The campaign began at the Instituto Nacional de Meteorologia e Geofísica (INMG) on Sal in the Cape Verde Islands, continued with ship-based observations aboard the German research vessel R/V Meteor during an Atlantic transect, and concluded at the Barbados Cloud Observatory (BCO) in the eastern Caribbean. Over the 52 d campaign, a total of 624 radiosondes were launched at high temporal frequency (typically every three hours), capturing high-resolution vertical profiles of temperature, humidity, pressure, and winds from three complementary platforms. The dataset encompasses raw, quality-controlled, and vertically gridded data, is detailed in this paper and offers a valuable resource for investigating the atmospheric structure and processes shaping tropical convection and the intertropical convergence zone (ITCZ). The datasets generated in this study include raw radiosonde measurements (Level 0), oscillating and merged radiosonde profiles (Level 1), and vertically gridded profiles (Level 2), which are publicly available via the ORCESTRA data portal and DOI-referenced archives (Winkler et al., 2025a; https://ipfs.io/ipns/latest.orcestra-campaign.org/raw/BCO/radiosondes/, Winkler et al., 2025b; https://ipfs.io/ipns/latest.orcestra-campaign.org/raw/INMG/radiosondes/, Winkler et al., 2025c; https://ipfs.io/ipns/latest.orcestra-campaign.org/raw/METEOR/radiosondes/, Winkler et al., 2025d; https://doi.org/10.82246/BAFYBEIHXRAJOJUQZYX65QSO7AMA6NGVREETKDW3HQZX3SDZFB7LCMG6VAQ, Winkler et al., 2026; https://doi.org/10.82246/BAFYBEIA34AUWYVBH2RQ7CN7AGUZZ7PULQ2KRDDDIEESM6KPYSI, Winkler and Rixen, 2026a; https://doi.org/10.82246/BAFYBEID7CNW62ZMZFGXCVC6Q6FA267A7IVK2W, Winkler and Rixen, 2026b).
