We present a significantly-improved data-driven global weather forecasting framework using a deep convolutional neural network (CNN) to forecast several basic atmospheric variables on a global grid. New developments in this framework include an offline volume-conservative mapping to a cubed-sphere grid, improvements to the CNN architecture, and the minimization of the loss function over multiple steps in a prediction sequence. The cubed-sphere remapping minimizes the distortion on the cube faces on which convolution operations are performed and provides natural boundary conditions for padding in the CNN. Our improved model produces weather forecasts that are indefinitely stable and produce realistic weather patterns at lead times of several weeks and longer. For short- to medium-range forecasting, our model significantly outperforms persistence, climatology, and a coarse-resolution dynamical numerical weather prediction (NWP) model. Unsurprisingly, our forecasts are worse than those from a high-resolution state-of-the-art operational NWP system. Our data-driven model is able to learn to forecast complex surface temperature patterns from few input atmospheric state variables. On annual time scales, our model produces a realistic seasonal cycle driven solely by the prescribed variation in top-of-atmosphere solar forcing. Although it is currently less accurate than operational weather forecasting models, our data-driven CNN executes much faster than those models, suggesting that machine learning could prove to be a valuable tool for large-ensemble forecasting.
StretchCast: Global-Regional AI Weather Forecasting on Stretched Cubed-Sphere Mesh
Institute of Urban Meteorology, China Meteorological Administration (IUM, CMA)（中国气象局城市气象研究所）；LeadSee Team Beijing Meteorological Service（北京气象局LeadSee团队）
PEAR: Equal Area Weather Forecasting on the Sphere
VERSES AI；Department of Mathematical Sciences（数学科学系）；Chalmers University of Technology（楚姆勒斯技术大学）；University of Gothenburg（哥德堡大学）；Department of Computer Science and Engineering（计算机科学与工程系）；Department of Physics and Astronomy（物理与天文学系）；Recohere
Microsoft（微软）；Department of Atmospheric Sciences, University of Washington（华盛顿大学大气科学系）；Microsoft Research（微软研究院）
We present an ensemble prediction system using a Deep Learning Weather Prediction (DLWP) model that recursively predicts key atmospheric variables with six-hour time resolution. This model uses convolutional neural networks (CNNs) on a cubed sphere grid to produce global forecasts. The approach is computationally efficient, requiring just three minutes on a single GPU to produce a 320-member set of six-week forecasts at 1.4° resolution. Ensemble spread is primarily produced by randomizing the CNN training process to create a set of 32 DLWP models with slightly different learned weights. Althou…
A nonhydrostatic dynamical core has been developed by using the multi-moment finite volume method that ensures the rigorous numerical conservation. To represent the spherical geometry free of polar problems, the cubed-sphere grid is adopted. A fourth-order multi-moment discretization formulation is applied to the nonhydrostatic governing equations cast in local curvilinear coordinates on each patch of cubed sphere through a gnomonic projection. In vertical direction, the height-based terrain-following grid is used to represent the topography. To get around the CFL stability restriction imposed…
A nonhydrostatic dynamical core has been developed by using the multi-moment finite volume method that ensures the rigorous numerical conservation. To represent the spherical geometry free of polar problems, the cubed-sphere grid is adopted. A fourth-order multi-moment discretization formulation is applied to solve the governing equations cast in the local curvilinear coordinates on each patch of cubed sphere through a gnomonic projection. In vertical direction, the height-based terrain-following grid is used to deal with the topography and a conservative finite difference scheme is adopted fo…
A positivity-preserving conservative semi-Lagrangian transport model by multi-moment finite volume method has been developed on the cubed-sphere grid. In this paper, two kinds of moments, i.e. point values (PV moment) at cell boundaries and volume integrated average (VIA) value, are defined within a single cell. The PV moment is updated by a conventional semi-Lagrangian method, while the VIA moment is cast by the flux form formulation that assures the exact numerical conservation. Different from the spatial approximation used in CSL2 (conservative semi-Lagrangian scheme with second order polyn…
Learning to Advect: A Neural Semi-Lagrangian Architecture for Weather Forecasting
Recherche en prévision numérique atmosphérique, Environnement et Changement climatique Canada（环境与气候变化加拿大大气数值预报研究）；Department of Earth, Ocean and Atmospheric Sciences, University of British Columbia（不列颠哥伦比亚大学地球、海洋和大气科学系）；Department of Computer Science, University of Saskatchewan（萨斯喀彻温大学计算机科学系）；Department of Mechanical Engineering, Rochester Institute of Technology（罗切斯特理工学院机械工程系）
Indian Institute of Tropical Meteorology（印度热带气象研究所）；University of Texas at Austin（德克萨斯大学奥斯汀分校）；Indian Institute of Technology Bombay（孟买印度理工学院）；Queen Mary University of London（伦敦玛丽女王大学）
The formation of precipitation in state-of-the-art weather and climate models is an important process. The understanding of its relationship with other variables can lead to endless benefits, particularly for the world's monsoon regions dependent on rainfall as a support for livelihood. Various factors play a crucial role in the formation of rainfall, and those physical processes are leading to significant biases in the operational weather forecasts. We use the UNET architecture of a deep convolutional neural network with residual learning as a proof of concept to learn global data-driven mode…
This report summarizes the major progresses to develop the dynamic core for next-generation atmospherical model for both numerical weather prediction and climate simulation. The numerical framework is based on a general formulation, so-called multi-moment constrained finite volume (MCV) method, which is well-balanced among solution quality (accuracy and robustness), algorithmic simplicity, computational efficiency and flexibility for model configuration. A local high-order limiting projection is also devised to remove spurious oscillations and noises in numerical solutions, which allows the nu…
Google Research（谷歌研究院）；California Institute of Technology（加州理工学院）；University of Oklahoma（俄克拉荷马大学）
Heat waves are projected to increase in frequency and severity with global warming. Improved warning systems would help reduce the associated loss of lives, wildfires, power disruptions, and reduction in crop yields. In this work, we explore the potential for deep learning systems trained on historical data to forecast extreme heat on short, medium and subseasonal timescales. To this purpose, we train a set of neural weather models (NWMs) with convolutional architectures to forecast surface temperature anomalies globally, 1 to 28 days ahead, at $\sim200~\mathrm{km}$ resolution and on the cubed…
Jackson School of Geosciences（杰克逊地球科学学院）；The University of Texas at Austin（德克萨斯大学奥斯汀分校）；Indian Institute of Tropical Meteorology（印度热带气象研究所）；Ministry of Earth Sciences（地球科学部）；CIRES, University of Colorado Boulder（科罗拉多大学博尔德分校环境科学研究合作研究所）；NOAA/Physical Sciences Laboratory（美国国家海洋和大气管理局物理科学实验室）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）
Precipitation governs Earth's hydroclimate, and its daily spatiotemporal fluctuations have major socioeconomic effects. Advances in Numerical weather prediction (NWP) have been measured by the improvement of forecasts for various physical fields such as temperature and pressure; however, large biases exist in precipitation prediction. We augment the output of the well-known NWP model CFSv2 with deep learning to create a hybrid model that improves short-range global precipitation at 1-, 2-, and 3-day lead times. To hybridise, we address the sphericity of the global data by using modified DLWP-C…
Aircast-Mars: A Mars Foundation Model for Global Weather Forecasting with HEALPix-Aware Convolutions
The University of Texas at Austin（德克萨斯大学奥斯汀分校）；Western Kentucky University（西肯塔基大学）；Ashoka University（阿肖卡大学）；California Institute of Technology（加州理工学院）；NASA Jet Propulsion Laboratory（美国国家航空航天局喷气推进实验室）
