Articles | Volume 17, issue 23
https://doi.org/10.5194/gmd-17-8873-2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/gmd-17-8873-2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Architectural insights into and training methodology optimization of Pangu-Weather
Deifilia To
CORRESPONDING AUTHOR
Scientific Computing Center (SCC), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Gholam Ali Hoshyaripour
CORRESPONDING AUTHOR
Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Markus Götz
CORRESPONDING AUTHOR
Scientific Computing Center (SCC), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Helmholtz AI, Karlsruhe, Germany
Achim Streit
CORRESPONDING AUTHOR
Scientific Computing Center (SCC), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Charlotte Debus
CORRESPONDING AUTHOR
Scientific Computing Center (SCC), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Related authors
No articles found.
Edgar Dolores-Tesillos, Olivia Martius, and Julian Quinting
Weather Clim. Dynam., 6, 471–487, https://doi.org/10.5194/wcd-6-471-2025, https://doi.org/10.5194/wcd-6-471-2025, 2025
Short summary
Short summary
An accurate representation of synoptic weather systems in climate models is required to estimate their societal and economic impacts under climate warming. Current climate models poorly represent the frequency of atmospheric blocking. Few studies have analysed the role of moist processes as a source of the bias of blocks. Here, we implement ELIAS2.0, a deep-learning tool, to validate the representation of moist processes in CMIP6 models and their link to the Euro-Atlantic blocking biases.
Marco Zanatta, Pia Bogert, Patrick Ginot, Yiwei Gong, Gholam Ali Hoshyaripour, Yaqiong Hu, Feng Jiang, Paolo Laj, Yanxia Li, Claudia Linke, Ottmar Möhler, Harald Saathoff, Martin Schnaiter, Nsikanabasi Silas Umo, Franziska Vogel, and Robert Wagner
Aerosol Research Discuss., https://doi.org/10.5194/ar-2025-12, https://doi.org/10.5194/ar-2025-12, 2025
Revised manuscript accepted for AR
Short summary
Short summary
Back carbon is an atmospheric pollutant from combustion, contributes to the Arctic warming. However, its properties change as it travels through the atmosphere, affecting its impact. We recreated Arctic transport conditions in a laboratory to study how black carbon evolves over time. Our findings show that temperature and altitude strongly influence its transformation, providing key insights for improving climate models and understanding Arctic pollution.
Lisa Janina Muth, Sascha Bierbauer, Corinna Hoose, Bernhard Vogel, Heike Vogel, and Gholam Ali Hoshyaripour
EGUsphere, https://doi.org/10.5194/egusphere-2025-402, https://doi.org/10.5194/egusphere-2025-402, 2025
Short summary
Short summary
Our study explores how intense wildfires created thunderstorm-like clouds that can affect weather and climate globally. Using simulations with high resolution, we found that fire heat and moisture help form these clouds, lifting particles high into the atmosphere. This process is crucial for understanding how fires impact the environment. Despite some differences with observational data, our findings align well over time, showing the importance of fire-induced heat in cloud formation.
Julia Bruckert, Simran Chopra, Richard Siddans, Charlotte Wedler, and Gholam Ali Hoshyaripour
EGUsphere, https://doi.org/10.5194/egusphere-2024-4062, https://doi.org/10.5194/egusphere-2024-4062, 2025
Short summary
Short summary
The 2022 Hunga eruption emitted about 150 Tg water vapor into the stratosphere. Here, we show that the water vapor injection not only accelerates SO2 oxidation and sulfate production but also increases the aging of ash (coating of ash by sulfate). Our study shows that aerosol aging alone does not explain the rapid loss of ash after the Hunga eruption as observed by satellite instruments. However, some ash might be masked in the observation due to the strong coating.
Alexandre Mass, Hendrik Andersen, Jan Cermak, Paola Formenti, Eva Pauli, and Julian Quinting
Atmos. Chem. Phys., 25, 491–510, https://doi.org/10.5194/acp-25-491-2025, https://doi.org/10.5194/acp-25-491-2025, 2025
Short summary
Short summary
This study investigates the interaction between smoke aerosols and fog and low clouds (FLCs) in the Namib Desert between June and October. Here, a satellite-based dataset of FLCs, reanalysis data and machine learning are used to systematically analyze FLC persistence under different aerosol loadings. Aerosol plumes are shown to modify local thermodynamics, which increase FLC persistence. But fully disentangling aerosol effects from meteorological ones remains a challenge.
Julian F. Quinting, Christian M. Grams, Edmund Kar-Man Chang, Stephan Pfahl, and Heini Wernli
Weather Clim. Dynam., 5, 65–85, https://doi.org/10.5194/wcd-5-65-2024, https://doi.org/10.5194/wcd-5-65-2024, 2024
Short summary
Short summary
Research in the last few decades has revealed that rapidly ascending airstreams in extratropical cyclones have an important effect on the evolution of downstream weather and predictability. In this study, we show that the occurrence of these airstreams over the North Pacific is modulated by tropical convection. Depending on the modulation, known atmospheric circulation patterns evolve quite differently, which may affect extended-range predictions in the Atlantic–European region.
Axel Seifert, Vanessa Bachmann, Florian Filipitsch, Jochen Förstner, Christian M. Grams, Gholam Ali Hoshyaripour, Julian Quinting, Anika Rohde, Heike Vogel, Annette Wagner, and Bernhard Vogel
Atmos. Chem. Phys., 23, 6409–6430, https://doi.org/10.5194/acp-23-6409-2023, https://doi.org/10.5194/acp-23-6409-2023, 2023
Short summary
Short summary
We investigate how mineral dust can lead to the formation of cirrus clouds. Dusty cirrus clouds lead to a reduction in solar radiation at the surface and, hence, a reduced photovoltaic power generation. Current weather prediction systems are not able to predict this interaction between mineral dust and cirrus clouds. We have developed a new physical description of the formation of dusty cirrus clouds. Overall we can show a considerable improvement in the forecast quality of clouds and radiation.
Patrick Ludwig, Florian Ehmele, Mário J. Franca, Susanna Mohr, Alberto Caldas-Alvarez, James E. Daniell, Uwe Ehret, Hendrik Feldmann, Marie Hundhausen, Peter Knippertz, Katharina Küpfer, Michael Kunz, Bernhard Mühr, Joaquim G. Pinto, Julian Quinting, Andreas M. Schäfer, Frank Seidel, and Christina Wisotzky
Nat. Hazards Earth Syst. Sci., 23, 1287–1311, https://doi.org/10.5194/nhess-23-1287-2023, https://doi.org/10.5194/nhess-23-1287-2023, 2023
Short summary
Short summary
Heavy precipitation in July 2021 led to widespread floods in western Germany and neighboring countries. The event was among the five heaviest precipitation events of the past 70 years in Germany, and the river discharges exceeded by far the statistical 100-year return values. Simulations of the event under future climate conditions revealed a strong and non-linear effect on flood peaks: for +2 K global warming, an 18 % increase in rainfall led to a 39 % increase of the flood peak in the Ahr river.
Susanna Mohr, Uwe Ehret, Michael Kunz, Patrick Ludwig, Alberto Caldas-Alvarez, James E. Daniell, Florian Ehmele, Hendrik Feldmann, Mário J. Franca, Christian Gattke, Marie Hundhausen, Peter Knippertz, Katharina Küpfer, Bernhard Mühr, Joaquim G. Pinto, Julian Quinting, Andreas M. Schäfer, Marc Scheibel, Frank Seidel, and Christina Wisotzky
Nat. Hazards Earth Syst. Sci., 23, 525–551, https://doi.org/10.5194/nhess-23-525-2023, https://doi.org/10.5194/nhess-23-525-2023, 2023
Short summary
Short summary
The flood event in July 2021 was one of the most severe disasters in Europe in the last half century. The objective of this two-part study is a multi-disciplinary assessment that examines the complex process interactions in different compartments, from meteorology to hydrological conditions to hydro-morphological processes to impacts on assets and environment. In addition, we address the question of what measures are possible to generate added value to early response management.
Ákos Horváth, James L. Carr, Dong L. Wu, Julia Bruckert, Gholam Ali Hoshyaripour, and Stefan A. Buehler
Atmos. Chem. Phys., 22, 12311–12330, https://doi.org/10.5194/acp-22-12311-2022, https://doi.org/10.5194/acp-22-12311-2022, 2022
Short summary
Short summary
We estimate plume heights for the April 2021 La Soufrière daytime eruptions using GOES-17 near-limb side views and GOES-16–MODIS stereo views. These geometric heights are then compared with brightness-temperature-based radiometric height estimates to characterize the biases of the latter. We also show that the side view method can be applied to infrared imagery and thus nighttime eruptions, albeit with larger uncertainty.
Julia Bruckert, Gholam Ali Hoshyaripour, Ákos Horváth, Lukas O. Muser, Fred J. Prata, Corinna Hoose, and Bernhard Vogel
Atmos. Chem. Phys., 22, 3535–3552, https://doi.org/10.5194/acp-22-3535-2022, https://doi.org/10.5194/acp-22-3535-2022, 2022
Short summary
Short summary
Volcanic emissions endanger aviation and public health and also influence weather and climate. Forecasting the volcanic-plume dispersion is therefore a critical yet sophisticated task. Here, we show that explicit treatment of volcanic-plume dynamics and eruption source parameters significantly improves volcanic-plume dispersion forecasts. We further demonstrate the lofting of the SO2 due to a heating of volcanic particles by sunlight with major implications for volcanic aerosol research.
Julian F. Quinting and Christian M. Grams
Geosci. Model Dev., 15, 715–730, https://doi.org/10.5194/gmd-15-715-2022, https://doi.org/10.5194/gmd-15-715-2022, 2022
Short summary
Short summary
Physical processes in weather systems importantly affect the midlatitude large-scale circulation. This study introduces an artificial-intelligence-based framework which allows the identification of an important weather system – the so-called warm conveyor belt (WCB) – at comparably low computational costs and from data at low spatial and temporal resolution. The framework thus newly enables the systematic investigation of WCBs in large data sets such as climate model projections.
Julian F. Quinting, Christian M. Grams, Annika Oertel, and Moritz Pickl
Geosci. Model Dev., 15, 731–744, https://doi.org/10.5194/gmd-15-731-2022, https://doi.org/10.5194/gmd-15-731-2022, 2022
Short summary
Short summary
This study applies novel artificial-intelligence-based models that allow the identification of one specific weather system which affects the midlatitude circulation. We show that the models yield similar results as their trajectory-based counterpart, which requires data at higher spatiotemporal resolution and is computationally more expensive. Overall, we aim to show how deep learning methods can be used efficiently to support process understanding of biases in weather prediction models.
Ákos Horváth, James L. Carr, Olga A. Girina, Dong L. Wu, Alexey A. Bril, Alexey A. Mazurov, Dmitry V. Melnikov, Gholam Ali Hoshyaripour, and Stefan A. Buehler
Atmos. Chem. Phys., 21, 12189–12206, https://doi.org/10.5194/acp-21-12189-2021, https://doi.org/10.5194/acp-21-12189-2021, 2021
Short summary
Short summary
We give a detailed description of a new technique to estimate the height of volcanic eruption columns from near-limb geostationary imagery. Such oblique angle observations offer spectacular side views of eruption columns protruding from the Earth ellipsoid and thereby facilitate a height-by-angle estimation method. Due to its purely geometric nature, the new technique is unaffected by the limitations of traditional brightness-temperature-based height retrievals.
Ákos Horváth, Olga A. Girina, James L. Carr, Dong L. Wu, Alexey A. Bril, Alexey A. Mazurov, Dmitry V. Melnikov, Gholam Ali Hoshyaripour, and Stefan A. Buehler
Atmos. Chem. Phys., 21, 12207–12226, https://doi.org/10.5194/acp-21-12207-2021, https://doi.org/10.5194/acp-21-12207-2021, 2021
Short summary
Short summary
We demonstrate the side view plume height estimation technique described in Part 1 on seven volcanic eruptions from 2019 and 2020, including the 2019 Raikoke eruption. We explore the strengths and limitations of the new technique in comparison to height estimation from brightness temperatures, stereo observations, and ground-based video footage.
Hengheng Zhang, Frank Wagner, Harald Saathoff, Heike Vogel, Gholam Ali Hoshyaripour, Vanessa Bachmann, Jochen Förstner, and Thomas Leisner
Atmos. Meas. Tech. Discuss., https://doi.org/10.5194/amt-2021-193, https://doi.org/10.5194/amt-2021-193, 2021
Revised manuscript not accepted
Short summary
Short summary
The evolution and the properties of Saharan dust plume were characterized by LIDARs, a sun photometer, and a regional transport model. Comparison between LIDAR measurements, sun photometer and ICON-ART predictions shows a good agreement for dust arrival time, dust layer height, and dust structure but also that the model overestimates the backscatter coefficients by a factor of (2.2 ± 0.16) and underestimate aerosol optical depth by a factor of (1.5 ± 0.11).
Assaf Hochman, Sebastian Scher, Julian Quinting, Joaquim G. Pinto, and Gabriele Messori
Earth Syst. Dynam., 12, 133–149, https://doi.org/10.5194/esd-12-133-2021, https://doi.org/10.5194/esd-12-133-2021, 2021
Short summary
Short summary
Skillful forecasts of extreme weather events have a major socioeconomic relevance. Here, we compare two approaches to diagnose the predictability of eastern Mediterranean heat waves: one based on recent developments in dynamical systems theory and one leveraging numerical ensemble weather forecasts. We conclude that the former can be a useful and cost-efficient complement to conventional numerical forecasts for understanding the dynamics of eastern Mediterranean heat waves.
Lukas O. Muser, Gholam Ali Hoshyaripour, Julia Bruckert, Ákos Horváth, Elizaveta Malinina, Sandra Wallis, Fred J. Prata, Alexei Rozanov, Christian von Savigny, Heike Vogel, and Bernhard Vogel
Atmos. Chem. Phys., 20, 15015–15036, https://doi.org/10.5194/acp-20-15015-2020, https://doi.org/10.5194/acp-20-15015-2020, 2020
Short summary
Short summary
Volcanic aerosols endanger aircraft and thus disrupt air travel globally. For aviation safety, it is vital to know the location and lifetime of such aerosols in the atmosphere. Here we show that the interaction of volcanic particles with each other eventually reduces their atmospheric lifetime. Moreover, we demonstrate that sunlight heats these particles, which lifts them several kilometers in the atmosphere. These findings support a more reliable forecast of volcanic aerosol dispersion.
Cited articles
Bodnar, C., Bruinsma, W., Lucic, A., Stanley, M., Brandstetter, J., Garvan, P., Riechert, M., Weyn, J., Dong, H., Vaughan, A., Gupta, J., Tambiratnam, K., Archibald, A., Heider, E., Welling, M., Turner, R., and Perdikaris, P.: Aurora: A Foundation Model of the Atmosphere, Tech. Rep., MSR-TR-2024-16, Microsoft Research AI for Science, https://www.microsoft.com/en-us/research/publication/aurora-a-foundation-model-of-the-atmosphere/ (last access: 4 December 2024), 2024. a
Chen, K., Han, T., Gong, J., Bai, L., Ling, F., Luo, J.-J., Chen, X., Ma, L., Zhang, T., Su, R., Ci, Y., Li, B., Yang, X., and Ouyang, W.: FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead, arXiv [preprint], https://doi.org/10.48550/arXiv.2304.02948, 2023a. a
Chen, L., Zhong, X., Zhang, F., Cheng, Y., Xu, Y., Qi, Y., and Li, H.: FuXi: a cascade machine learning forecasting system for 15-day global weather forecast, npj Climate and Atmospheric Science, 6, 190, https://doi.org/10.1038/s41612-023-00512-1, 2023b. a
Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz‐Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., De Chiara, G., Dahlgren, P., Dee, D., Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A., Haimberger, L., Healy, S., Hogan, R. J., Hólm, E., Janisková, M., Keeley, S., Laloyaux, P., Lopez, P., Lupu, C., Radnoti, G., de Rosnay, P., Rozum, I., Vamborg, F., Villaume, S., and Thépaut, J.-N.: The ERA5 global reanalysis, Q. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803, 2020. a
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on single levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS) [data set], https://doi.org/10.24381/cds.adbb2d47, 2023a. a
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., and Thépaut, J.-N.: ERA5 hourly data on pressure levels from 1940 to present, Copernicus Climate Change Service (C3S) Climate Data Store (CDS) [data set], https://doi.org/10.24381/cds.bd0915c6, 2023b. a
Lam, R., Sanchez-Gonzalez, A., Willson, M., Wirnsberger, P., Fortunato, M., Alet, F., Ravuri, S., Ewalds, T., Eaton-Rosen, Z., Hu, W., Merose, A., Hoyer, S., Holland, G., Vinyals, O., Stott, J., Pritzel, A., Mohamed, S., and Battaglia, P.: Learning skillful medium-range global weather forecasting, Science, 382, 1416–1421, https://doi.org/10.1126/science.adi2336, 2023. a
Lessig, C., Luise, I., Gong, B., Langguth, M., Stadtler, S., and Schultz, M.: AtmoRep: A stochastic model of atmosphere dynamics using large scale representation learning, arXiv [preprint], https://doi.org/10.48550/arXiv.2308.13280, 2023. a
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., and Guo, B.: Swin Transformer: Hierarchical Vision Transformer using Shifted Windows, 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada, 2021, 9992–10002, https://doi.org/10.1109/ICCV48922.2021.00986, 2021. a, b
Nguyen, T., Brandstetter, J., Kapoor, A., Gupta, J. K., and Grover, A.: ClimaX: A foundation model for weather and climate, in: Proceedings of the 40th International Conference on Machine Learning, edited by: Krause, A., Brunskill, E., Cho, K., Engelhardt, B., Sabato, S., and Scarlett, J., vol. 202 of Proceedings of Machine Learning Research, 25904–25938, PMLR, https://proceedings.mlr.press/v202/nguyen23a.html (last access: 10 December 2024), 2023. a
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S.: PyTorch: An Imperative Style, High-Performance Deep Learning Library, in: Advances in Neural Information Processing Systems, edited by: Wallach, H., Larochelle, H., Beygelzimer, A., d'Alché-Buc, F., Fox, E., and Garnett, R., vol. 32, Curran Associates, Inc., https://proceedings.neurips.cc/paper_files/paper/2019/file/bdbca288fee7f92f2bfa9f7012727740-Paper.pdf (last access: 10 December 2024), 2019. a
Rajbhandari, S., Rasley, J., Ruwase, O., and He, Y.: ZeRO: Memory optimizations Toward Training Trillion Parameter Models, in: SC20: International Conference for High Performance Computing, Networking, Storage and Analysis, 1–16, https://doi.org/10.1109/SC41405.2020.00024, 2020. a
Rasp, S., Hoyer, S., Merose, A., Langmore, I., Battaglia, P., Russel, T., Sanchez-Gonzalez, A., Yang, V., Carver, R., Agrawal, S., Chantry, M., Bouallegue, Z. B., Dueben, P., Bromberg, C., Sisk, J., Barrington, L., Bell, A., and Sha, F.: WeatherBench 2: A benchmark for the next generation of data-driven global weather models, arXiv [preprint], https://doi.org/10.48550/arXiv.2308.15560, 2023. a, b
To, D.: DeifiliaTo/PanguWeather: v1.0.2 (1.0.2), Zenodo [code], https://doi.org/10.5281/zenodo.12720775, 2024. a, b
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., and Polosukhin, I.: Attention is all you need, in: 31st Conference on Neural Information Processing Systems, Long Beach, CA, USA, 4 December 2017, https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf (last access: 10 December 2024), 2017. a
Willard, J. D., Harrington, P., Subramanian, S., Mahesh, A., O'Brien, T. A., and Collins, W. D.: Analyzing and Exploring Training Recipes for Large-Scale Transformer-Based Weather Prediction, in: 23rd Conference on Artificial Intelligence for Environmental Science, Baltimore, MD, USA, January 2024, https://doi.org/10.48550/arXiv.2404.19630, 2024. a
Short summary
Pangu-Weather is a breakthrough machine learning model in medium-range weather forecasting that considers 3D atmospheric information. We show that using a simpler 2D framework improves robustness, speeds up training, and reduces computational needs by 20 %–30 %. We introduce a training procedure that varies the importance of atmospheric variables over time to speed up training convergence. Decreasing computational demand increases the accessibility of training and working with the model.
Pangu-Weather is a breakthrough machine learning model in medium-range weather forecasting that...