Atmospheric and Oceanic Physics
- [1] arXiv:2405.13043 [pdf, ps, other]
-
Title: Towards Specialized Supercomputers for Climate Sciences: Computational Requirements of the Icosahedral Nonhydrostatic Weather and Climate ModelTorsten Hoefler, Alexandru Calotoiu, Anurag Dipankar, Thomas Schulthess, Xavier Lapillonne, Oliver FuhrerSubjects: Atmospheric and Oceanic Physics (physics.ao-ph); Hardware Architecture (cs.AR); Distributed, Parallel, and Cluster Computing (cs.DC); Computational Physics (physics.comp-ph)
We discuss the computational challenges and requirements for high-resolution climate simulations using the Icosahedral Nonhydrostatic Weather and Climate Model (ICON). We define a detailed requirements model for ICON which emphasizes the need for specialized supercomputers to accurately predict climate change impacts and extreme weather events. Based on the requirements model, we outline computational demands for km-scale simulations, and suggests machine learning techniques to enhance model accuracy and efficiency. Our findings aim to guide the design of future supercomputers for advanced climate science.
- [2] arXiv:2405.13063 [pdf, ps, html, other]
-
Title: Aurora: A Foundation Model of the AtmosphereCristian Bodnar, Wessel P. Bruinsma, Ana Lucic, Megan Stanley, Johannes Brandstetter, Patrick Garvan, Maik Riechert, Jonathan Weyn, Haiyu Dong, Anna Vaughan, Jayesh K. Gupta, Kit Tambiratnam, Alex Archibald, Elizabeth Heider, Max Welling, Richard E. Turner, Paris PerdikarisSubjects: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
Deep learning foundation models are revolutionizing many facets of science by leveraging vast amounts of data to learn general-purpose representations that can be adapted to tackle diverse downstream tasks. Foundation models hold the promise to also transform our ability to model our planet and its subsystems by exploiting the vast expanse of Earth system data. Here we introduce Aurora, a large-scale foundation model of the atmosphere trained on over a million hours of diverse weather and climate data. Aurora leverages the strengths of the foundation modelling approach to produce operational forecasts for a wide variety of atmospheric prediction problems, including those with limited training data, heterogeneous variables, and extreme events. In under a minute, Aurora produces 5-day global air pollution predictions and 10-day high-resolution weather forecasts that outperform state-of-the-art classical simulation tools and the best specialized deep learning models. Taken together, these results indicate that foundation models can transform environmental forecasting.
- [3] arXiv:2405.13334 [pdf, ps, other]
-
Title: Enhanced continuous aerosol optical depth (AOD) estimation using geostationary satellite data: focusing on nighttime AOD over East AsiaComments: 38 pages (including references and supplementary); 3 tables, 9 figures; 2 supplementary tables, 4 supplementary figuresSubjects: Atmospheric and Oceanic Physics (physics.ao-ph)
Continuous aerosol monitoring in East Asia is essential due to the massive aerosol emissions from natural and anthropogenic sources. Geostationary satellites enable continuous aerosol monitoring; however, the observation is limited to the daytime. This study proposed machine learning-based models to estimate daytime and nighttime aerosol optical depth (AOD) in East Asia using a geostationary satellite, Geo-KOMPSAT-2A (GK-2A). The input variables for the machine learning models include the brightness temperature (BT) and top-of-atmosphere (TOA) reflectance from GK-2A, meteorological and geographical data, and auxiliary variables. The two models that used different combinations of GK-2A variables were proposed and compared: the all-day BT model, which estimates AOD during both day and night using BT variables, and the daytime TOA model, which estimates AOD during the day using TOA reflectance variables as well. The estimated AODs by the models were validated with ground-based AOD data from the Aerosol Robotic Network (AERONET) by 10-fold cross-validation and hold-out validation methods. The performance of the daytime TOA model was slightly higher than the all-day BT model during the day (R2 = 0.80-0.82, root mean square error (RMSE) = 0.107-0.116 for the all-day BT model, R2 = 0.83, RMSE = 0.098 for the daytime TOA model). The SHapley Additive exPlanations (SHAP) analysis showed that total precipitable water content and seasonality contributed the most for both proposed models. BT differences and TOA reflectance variables were identified as the next most contributing variables for the all-day BT and daytime TOA models. The spatiotemporal distributions of estimated AODs from the proposed models show similar patterns compared with other AOD products. A time series comparison at a test station demonstrated that the estimated AOD of the proposed models was consistent with the AERONET AOD.
- [4] arXiv:2405.13988 [pdf, ps, html, other]
-
Title: Melancholia States of the Atlantic Meridional Overturning CirculationSubjects: Atmospheric and Oceanic Physics (physics.ao-ph); Chaotic Dynamics (nlin.CD)
The Atlantic Meridional Overturning Circulation (AMOC) is a much studied component of the climate system, because its suspected multistability leads to tipping behaviour with large regional and global climatic impacts. In this paper we investigate the global stability properties of the system using an ocean general circulation model. We construct an unstable AMOC state, i.e., an unstable solution of the flow that resides between the stable regimes of a vigorous and collapsed AMOC. Such a solution, also known as a Melancholia or edge state, is a dynamical saddle embedded in the boundary separating the competing basins of attraction. It is physically relevant since it lies on the most probable path of a noise-induced transition between the two stable regimes, and because tipping occurs when one of the attractors and the Melancholia state collide. Its properties may thus give hints towards physical mechanisms and predictability of the critical transition. We find that while the AMOC Melancholia state as viewed from its upper ocean properties lies between the vigorous and collapsed regimes, it is characterized by an Atlantic deep ocean that is fresher and colder compared to both stable regimes. The Melancholia state has higher dynamic enthalpy than either stable state, representing a state of higher potential energy that is in good agreement with the dynamical landscape view on metastability.
- [5] arXiv:2405.14204 [pdf, ps, html, other]
-
Title: Multi-instrument analysis of L-band amplitude scintillation observed over the Eastern Arabian PeninsulaSubjects: Atmospheric and Oceanic Physics (physics.ao-ph); Signal Processing (eess.SP); Space Physics (physics.space-ph)
This study investigates the spatial and temporal characteristics of L1 amplitude scintillation-causing ionospheric irregularities over the Eastern Arabian Peninsula during the ascending phase of solar cycle 25 (years 2020--2023). The temporal occurrences of weak and strong scintillation were separated by sunset, with weak scintillation observed predominantly pre-sunset during the winter solstice and strong scintillation observed mainly post-sunset during the autumnal equinox. Strong scintillation was much more pronounced in 2023 compared to the other three years, indicating a strong influence of solar activity. Spatially, weak-scintillation-causing irregularities exhibited a wide distribution in azimuth and elevation, while strong-scintillation-causing irregularities were concentrated southwards. The combined analysis of S4 and rate of total electron content index (ROTI) suggested that small-scale ionospheric irregularities were present in both pre- and post-sunset periods, while large-scale irregularities were only seen during the post-sunset period. Furthermore, the presence of southward traveling ionospheric disturbances (TIDs) during the 2023 autumnal equinox was confirmed with the total electron content anomaly ($\Delta\text{TEC}$), while the Ionospheric Bubble Index (IBI) provided by the Swarm mission was unable to confirm the presence of equatorial plasma bubbles during the same period. Observations from the FORMOSAT-7/COSMIC-2 mission indicated that strong-scintillation-causing irregularities were more prevalent under the F2-layer peak, while the weak-scintillation-causing irregularities were mostly observed at the E-layer, F2-layer, and above the F2-layer.
New submissions for Friday, 24 May 2024 (showing 5 of 5 entries )
- [6] arXiv:2405.13180 (cross-list from eess.SP) [pdf, ps, html, other]
-
Title: Data Assimilation with Machine Learning Surrogate Models: A Case Study with FourCastNetSubjects: Signal Processing (eess.SP); Machine Learning (cs.LG); Chaotic Dynamics (nlin.CD); Atmospheric and Oceanic Physics (physics.ao-ph); Applications (stat.AP)
Modern data-driven surrogate models for weather forecasting provide accurate short-term predictions but inaccurate and nonphysical long-term forecasts. This paper investigates online weather prediction using machine learning surrogates supplemented with partial and noisy observations. We empirically demonstrate and theoretically justify that, despite the long-time instability of the surrogates and the sparsity of the observations, filtering estimates can remain accurate in the long-time horizon. As a case study, we integrate FourCastNet, a state-of-the-art weather surrogate model, within a variational data assimilation framework using partial, noisy ERA5 data. Our results show that filtering estimates remain accurate over a year-long assimilation window and provide effective initial conditions for forecasting tasks, including extreme event prediction.
- [7] arXiv:2405.13711 (cross-list from cs.LG) [pdf, ps, html, other]
-
Title: VAE-Var: Variational-Autoencoder-Enhanced Variational AssimilationSubjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Dynamical Systems (math.DS); Atmospheric and Oceanic Physics (physics.ao-ph)
Data assimilation refers to a set of algorithms designed to compute the optimal estimate of a system's state by refining the prior prediction (known as background states) using observed data. Variational assimilation methods rely on the maximum likelihood approach to formulate a variational cost, with the optimal state estimate derived by minimizing this cost. Although traditional variational methods have achieved great success and have been widely used in many numerical weather prediction centers, they generally assume Gaussian errors in the background states, which limits the accuracy of these algorithms due to the inherent inaccuracies of this assumption. In this paper, we introduce VAE-Var, a novel variational algorithm that leverages a variational autoencoder (VAE) to model a non-Gaussian estimate of the background error distribution. We theoretically derive the variational cost under the VAE estimation and present the general formulation of VAE-Var; we implement VAE-Var on low-dimensional chaotic systems and demonstrate through experimental results that VAE-Var consistently outperforms traditional variational assimilation methods in terms of accuracy across various observational settings.
- [8] arXiv:2405.14693 (cross-list from astro-ph.EP) [pdf, ps, html, other]
-
Title: Interpolation and synthesis of sparse samples in exoplanet atmospheric modelingComments: Accepted for publication in the Planetary Science JournalSubjects: Earth and Planetary Astrophysics (astro-ph.EP); Instrumentation and Methods for Astrophysics (astro-ph.IM); Atmospheric and Oceanic Physics (physics.ao-ph); Geophysics (physics.geo-ph)
This paper highlights methods from geostatistics that are relevant to the interpretation, intercomparison, and synthesis of atmospheric model data, with a specific application to exoplanet atmospheric modeling. Climate models are increasingly used to study theoretical and observational properties of exoplanets, which include a hierarchy of models ranging from fast and idealized models to those that are slower but more comprehensive. Exploring large parameter spaces with computationally-expensive models can be accomplished with sparse sampling techniques, but analyzing such sparse samples can pose challenges for conventional interpolation functions. Ordinary kriging is a statistical method for describing the spatial distribution of a data set in terms of the variogram function, which can be used to interpolate sparse samples across any number of dimensions. Variograms themselves may also be useful diagnostic tools for describing the spatial distribution of model data in exoplanet atmospheric model intercomparison projects. Universal kriging is another method that can synthesize data calculated by models of different complexity, which can be used to combine sparse samples of data from slow models with larger samples of data from fast models. Ordinary and universal kriging can also provide a way to synthesize model predictions with sparse samples of exoplanet observations and may have other applications in exoplanet science.
Cross submissions for Friday, 24 May 2024 (showing 3 of 3 entries )
- [9] arXiv:2312.03861 (replaced) [pdf, ps, html, other]
-
Title: Data-driven analysis of annual rain distributionsComments: 26 pages, 18 figuresJournal-ref: Phys. Rev. Research 6, 023187 (2024)Subjects: Atmospheric and Oceanic Physics (physics.ao-ph); Statistical Mechanics (cond-mat.stat-mech); Data Analysis, Statistics and Probability (physics.data-an)
Rainfall is an important component of the climate system and its statistical properties are vital for prediction purposes. In this study, we have developed a statistical method for constructing the distribution of annual precipitation. The method is based on the convolution of the measured monthly rainfall distributions and does not depend on any presumed annual rainfall distribution. Using a simple statistical model, we demonstrate that our approach allows for a better prediction of extremely dry or wet years with a recurrence time several times longer than the original time series. The method that has been proposed can be utilized for other climate variables as well.