Article
Meteorology & Atmospheric Sciences
Stephan Rasp, Nils Thuerey
Summary: The traditional numerical weather prediction is based on discretizing the dynamical and physical equations of the atmosphere, while the rise of deep learning has brought increased interest in purely data-driven medium-range weather forecasting. Through the WeatherBench benchmark challenge, we trained a deep residual convolutional neural network to predict geopotential, temperature, and precipitation up to 5 days ahead, resulting in forecasts outperforming previous submissions and comparable to a physical baseline. Analysis shows the model has learned physically reasonable correlations, and scaling experiments have been conducted to estimate the potential skill of data-driven approaches at higher resolutions.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2021)
Article
Environmental Sciences
Bogdan Bochenek, Zbigniew Ustrnul
Summary: In this paper, an analysis of the most relevant scientific articles on machine learning methods in the field of climate and numerical weather prediction was conducted. The common topics of interest and the most frequently examined meteorological fields, methods, and countries were identified through the analysis of abstracts. The authors predicted that machine learning methods will play a key role in future weather forecasting based on critical reviews of the literature.
Article
Environmental Sciences
Alexandra-Ioana Albu, Gabriela Czibula, Andrei Mihai, Istvan Gergely Czibula, Sorin Burcea, Abdelkader Mezghani
Summary: This paper proposes a convolutional network model called NeXtNow for weather forecasting, which achieves more accurate predictions by incorporating multiple past radar measurements. Compared to previous models, NeXtNow shows significant improvements in critical success index and root mean square error.
Article
Computer Science, Artificial Intelligence
Alex Bihlo
Summary: Using conditional deep convolutional generative adversarial networks, the study successfully predicts geopotential height and two-meter temperature over Europe, but fails to accurately predict total precipitation. The use of Monte-Carlo dropout helps improve the forecasting model's skill by quantifying uncertainty in current weather forecasts.
Article
Environmental Sciences
Yeji Choi, Keumgang Cha, Minyoung Back, Hyunguk Choi, Taegyun Jeon
Summary: Quantitative precipitation prediction is crucial for managing water-related disasters, and recent advances in data-driven approaches using deep learning techniques have shown improved performance. The RAIN-F+ dataset is introduced for rainfall prediction, showing better results than radar-only data over time, with limitations in predicting heavy rainfall. This highlights the importance of multi-modal information for precipitation nowcasting using deep learning.
Article
Multidisciplinary Sciences
M. G. Schultz, C. Betancourt, B. Gong, F. Kleinert, M. Langguth, L. H. Leufen, A. Mozaffari, S. Stadtler
Summary: The recent hype around artificial intelligence has renewed interest in applying successful deep learning methods in the field of meteorology. Evidence suggests that better weather forecasts can be achieved with the introduction of big data mining and neural networks. However, fundamental breakthroughs are needed before numerical weather models can be completely replaced by DL approaches.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2021)
Article
Green & Sustainable Science & Technology
Serkan Ozdemir, Sevgi Ozkan Yildirim
Summary: In recent years, intensive water use and global climate change have caused fluctuations in freshwater lake levels, water quality, and water ecosystem balance. Deep learning models, such as artificial neural networks and recurrent neural networks, can provide fast and reliable predictions of lake water levels. The Long Short-Term Memory model has shown the best performance in predicting water levels for up to 60 days.
Article
Meteorology & Atmospheric Sciences
Philipp Hess, Niklas Boers
Summary: The accurate prediction of heavy rainfall events remains challenging for numerical weather prediction models. In this study, a U-Net-based deep neural network is used to learn heavy rainfall events from a NWP ensemble. A frequency-based weighting method is proposed to enable the learning of heavy rainfall events in the distributions' tails. Applying this method in post-processing improves the forecast skill of heavy rainfall events.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Geosciences, Multidisciplinary
Xiaodong Chen, L. Ruby Leung, Ning Sun
Summary: This study introduces a new method called Weather Anomaly Clustering (WAC-hydro) for predicting both precipitation and temperature, which helps link large-scale climate conditions to regional hydroclimate conditions. By identifying 12 clusters of daily weather anomaly modes in the US Pacific Northwest Puget Sound region, this method provides insights into the flood mechanisms and their connections to climate variability modes.
GEOPHYSICAL RESEARCH LETTERS
(2023)
Article
Meteorology & Atmospheric Sciences
Andreas F. Prein, Janice Coen, Abby Jaye
Summary: Five of the largest wildfires in California occurred in 2020, with the largest complex surpassing the previous record by over 100%. Previous studies focused on human activities and atmospheric thermodynamics, but the impact of changing atmospheric dynamics remains largely unknown. This study identifies different weather types associated with historically large burned areas in California and suggests that climate change is likely to reduce the frequency of certain extreme weather events while increasing the risk of catastrophic fires in the future.
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES
(2022)
Article
Transportation Science & Technology
Yutian Pang, Xinyu Zhao, Hao Yan, Yongming Liu
Summary: Trajectory prediction is crucial for the national air transportation system, and reliable models need to account for uncertainties. This study focuses on the impact of environmental factors and proposes an advanced Bayesian Deep Learning method for aircraft trajectory prediction. Results show a significant reduction in prediction variance compared to existing methods, validating the effectiveness of the proposed approach.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2021)
Article
Construction & Building Technology
Jung Min Han, Yu Qian Ang, Ali Malkawi, Holly W. Samuelson
Summary: This study introduces a methodology using recurrent neural networks to generate synthetic localized weather data, which significantly improves accuracy of local climate conditions. The method can be applied to various built environment applications, enhancing the accuracies of building energy models.
BUILDING AND ENVIRONMENT
(2021)
Article
Astronomy & Astrophysics
Maria J. Molina, David John Gagne, Andreas F. Prein
Summary: This case study investigates the ability of deep learning methods to classify thunderstorms in a future climate based on training data from the present-day climate. The results show that a convolutional neural network performed well in both present and future climates, learning physical characteristics of thunderstorms and environments. This study demonstrates that deep learning can generalize to future climate conditions and exhibit robustness with hyperparameter tuning in certain applications.
EARTH AND SPACE SCIENCE
(2021)
Article
Multidisciplinary Sciences
Saahil Shenoy, Dimitry Gorinevsky, Kevin E. Trenberth, Steven Chu
Summary: By analyzing temperature and rainfall data, we found that there is an increasing trend in extreme climate events in the continental United States over the past 41 years. The risk of high-temperature events has increased 2.1-fold, with a 2.6-fold increase from July to October. On the other hand, the risk of high rainfall extremes has increased 1.4-fold in December and January but decreased by 22% during the spring and summer months.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2022)
Article
Computer Science, Information Systems
Rebeca Quintero Gonzalez, Jamal Jokar Arsanjani
Summary: This study used three machine learning algorithms to predict future changes in groundwater levels in Denmark based on climate change scenarios. The random forest (RF) model outperformed the other two models, showing a slight increase in water table levels in the future, especially during winter. The developed approach and models can be applied to other areas to improve prevention and adaptation plans for future climate change scenarios.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2021)
Article
Meteorology & Atmospheric Sciences
Joaquin E. Blanco, Rodrigo Caballero, George Datseris, Bjorn Stevens, Sandrine Bony, Or Hadas, Yohai Kaspi
Summary: Recent research indicates that the reflection of sunlight is almost equal in the Northern and Southern Hemispheres due to compensating asymmetries in cloud albedo. The study investigates the causes of this asymmetry and finds that it is mainly due to clouds in oceanic regions. These findings are crucial for understanding global cloud feedbacks and developing a theory for planetary albedo and its symmetry.
JOURNAL OF CLIMATE
(2023)
Article
Multidisciplinary Sciences
Or Hadas, George Datseris, Joaquin Blancoc, Sandrine Bony, Rodrigo Caballero, Bjorn Stevens, Yohai Kaspi
Summary: Clouds have a significant impact on Earth's climate, particularly midlatitude clouds which play a crucial role in shaping Earth's albedo. This study explores the relationship between baroclinic activity and cloud-albedo, and its connection to existing hemispheric albedo symmetry. The research reveals a strong correlation between baroclinic activity and cloud-albedo, explaining how cloud-albedo increases with intensity. It also demonstrates that the difference in cloud-albedo between hemispheres can be explained by the disparity in the population of cyclones and anticyclones.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2023)
Article
Meteorology & Atmospheric Sciences
Traute Crueger, Hauke Schmidt, Bjorn Stevens
Summary: Earth's planetary albedo exhibits hemispheric symmetry, and this study evaluates how CMIP models symmetrize the hemispheric clear-sky albedo asymmetry and the role of clouds in this process. The results show that CMIP multimodel means are similar to CERES, with clouds compensating reference asymmetries by a certain percentage. The study also highlights the contributions of tropical clouds and extratropical storm track regions in compensating hemispheric clear-sky asymmetries.
JOURNAL OF CLIMATE
(2023)
Article
Meteorology & Atmospheric Sciences
Anna Lea Albright, Bjorn Stevens, Sandrine Bony, Raphaela Vogel
Summary: Using observations from the EUREC4A campaign, a new conceptual picture of the trade wind transition layer is proposed. Cloud formation mainly occurs within the transition layer, and the cloud-top height distribution is bimodal. The life cycle of the first cloud population maintains the transition-layer structure through a condensation-evaporation mechanism.
JOURNAL OF THE ATMOSPHERIC SCIENCES
(2023)
Article
Meteorology & Atmospheric Sciences
Theresa Lang, Ann Kristin Naumann, Stefan A. Buehler, Bjorn Stevens, Hauke Schmidt, Franziska Aemisegger
Summary: We conducted eight 45-day experiments using a global storm-resolving model (GSRM) to test the sensitivity of relative humidity (R) in the tropics to changes in model resolution and parameterizations. The perturbations we applied to the model resulted in tropical mean R changes ranging from 0.5% to 8% in the mid troposphere. Changes in parameterizations had the largest impact on R, indicating that model physics are a major source of humidity spread across GSRMs.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Editorial Material
Geosciences, Multidisciplinary
Susan Trumbore, Ana Barros, Thorsten Becker, M. Bayani Cardenas, Eric Davidson, Nicolas Gruber, Eileen Hofmann, Mary Hudson, Tissa Illangasekare, Sarah Kang, Alberto Montanari, Marcos Moreno, Francis Nimmo, Larry Paxton, Vincent J. M. Salters, David Schimel, Bjorn Stevens, Hang Su, Donald Wuebbles, Peter Zeitler, Binzheng Zhang
Summary: The editorial board of AGU Advances expresses gratitude to the individuals who reviewed manuscripts for the journal in 2022. Thanks to the 131 reviewers who contributed to AGU Advances in 2022.
Article
Geosciences, Multidisciplinary
Geet George, Bjorn Stevens, Sandrine Bony, Raphaela Vogel, Ann Kristin Naumann
Summary: Understanding the drivers of cloud organization is crucial for accurately estimating cloud feedbacks and their contribution to climate warming. Shallow mesoscale circulations, which have not been observed before, are found to exist and play an important role in cloud organization. These circulations are associated with large variability in mesoscale vertical velocity and amplify moisture variance at the cloud base. The ubiquity of these circulations suggests their integral role in determining how clouds respond to climate change.
Article
Meteorology & Atmospheric Sciences
Andre Ehrlich, Martin Zoeger, Andreas Giez, Vladyslav Nenakhov, Christian Mallaun, Rolf Maser, Timo Roeschenthaler, Anna E. Luebke, Kevin Wolf, Bjorn Stevens, Manfred Wendisch
Summary: The HALO research aircraft's instrumentation is expanded by the BACARDI to measure radiative energy budget. Two sets of pyranometers and pyrgeometers are installed to measure solar and thermal-infrared irradiances. The BACARDI measurements show that dynamic thermal effects can be corrected by parameterizing the rate of change of radiometer sensor temperatures. The measurements also demonstrate the reliability of common geometric attitude correction of solar downward irradiance.
ATMOSPHERIC MEASUREMENT TECHNIQUES
(2023)
Article
Meteorology & Atmospheric Sciences
L. Paccini, B. Stevens
Summary: This study investigates the representation of precipitation in the Amazon basin through explicit convection and its relation to organized systems. Ensemble simulations of the ICON-NWP model at different resolutions with explicit and parameterized convection are used, alongside satellite data. The improvements in precipitation representation by explicit convection are shown to be in intensity distribution and spatial distribution. The well-simulated precipitation features in the Amazon are predominantly driven by the distribution of organized convective systems. However, further research is needed to address unresolved processes that affect the representation of precipitation systems in the Amazon.
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES
(2023)
Article
Geosciences, Multidisciplinary
Adriana Bailey, Franziska Aemisegger, Leonie Villiger, Sebastian A. Los, Gilles Reverdin, Estefania Quinones Melendez, Claudia Acquistapace, Dariusz B. Baranowski, Tobias Bock, Sandrine Bony, Tobias Bordsdorff, Derek Coffman, Simon P. de Szoeke, Christopher J. Diekmann, Marina Duetsch, Benjamin Ertl, Joseph Galewsky, Dean Henze, Przemyslaw Makuch, David Noone, Patricia K. Quinn, Michael Roesch, Andreas Schneider, Matthias Schneider, Sabrina Speich, Bjorn Stevens, Elizabeth J. Thompson
Summary: This paper describes the EUREC(4)A isotopic in situ data collection and guides readers to complementary remotely sensed water vapor isotope ratios. All the field data have been publicly available, even with known biases, to promote dialogue around improving water isotope measurement strategies for the future. The high-quality data create unprecedented opportunities to close water isotopic budgets and evaluate water fluxes and their influence on cloudiness in the trade-wind environment.
EARTH SYSTEM SCIENCE DATA
(2023)
Article
Geosciences, Multidisciplinary
Cathy Hohenegger, Peter Korn, Leonidas Linardakis, Rene Redler, Reiner Schnur, Panagiotis Adamidis, Jiawei Bao, Swantje Bastin, Milad Behravesh, Martin Bergemann, Joachim Biercamp, Hendryk Bockelmann, Renate Brokopf, Nils Brueggemann, Lucas Casaroli, Fatemeh Chegini, George Datseris, Monika Esch, Geet George, Marco Giorgetta, Oliver Gutjahr, Helmuth Haak, Moritz Hanke, Tatiana Ilyina, Thomas Jahns, Johann Jungclaus, Marcel Kern, Daniel Klocke, Lukas Kluft, Tobias Koelling, Luis Kornblueh, Sergey Kosukhin, Clarissa Kroll, Junhong Lee, Thorsten Mauritsen, Carolin Mehlmann, Theresa Mieslinger, Ann Kristin Naumann, Laura Paccini, Angel Peinado, Divya Sri Praturi, Dian Putrasahan, Sebastian Rast, Thomas Riddick, Niklas Roeber, Hauke Schmidt, Uwe Schulzweida, Florian Schuette, Hans Segura, Radomyra Shevchenko, Vikram Singh, Mia Specht, Claudia Christine Stephan, Jin-Song von Storch, Raphaela Vogel, Christian Wengel, Marius Winkler, Florian Ziemen, Jochem Marotzke, Bjorn Stevens
Summary: In this paper, the authors introduce a new model configuration called ICON-Sapphire, which allows representation of the Earth system with a grid spacing of 10 km and finer. Through various simulations, they demonstrate that ICON-Sapphire is capable of resolving different scales of phenomena in the atmosphere and ocean. The model shows promising results in accurately capturing basic features of the climate system and enables multi-decadal global simulations including interactive carbon.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2023)
Editorial Material
Water Resources
Julia Slingo, Paul Bates, Peter Bauer, Stephen Belcher, Tim Palmer, Graeme Stephens, Bjorn Stevens, Thomas F. Stocker, Georg Teutsch
HYDROLOGIE UND WASSERBEWIRTSCHAFTUNG
(2022)
Article
Geosciences, Multidisciplinary
Marco A. Giorgetta, William Sawyer, Xavier Lapillonne, Panagiotis Adamidis, Dmitry Alexeev, Valentin Clement, Remo Dietlicher, Jan Frederik Engels, Monika Esch, Henning Franke, Claudia Frauen, Walter M. Hannah, Benjamin R. Hillman, Luis Kornblueh, Philippe Marti, Matthew R. Norman, Robert Pincus, Sebastian Rast, Daniel Reinert, Reiner Schnur, Uwe Schulzweida, Bjorn Stevens
Summary: Classical numerical models for the global atmosphere are typically developed for CPU architectures, hindering their scientific applications on top-performing supercomputers with GPU architectures. The development of a GPU-enabled version of the ICON atmosphere model enables research on the quasi-biennial oscillation (QBO) and allows for global experiments at high resolutions on modern supercomputers.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2022)