Water vapor has a significant impact on atmospheric circulations by moist processes. Subgrid-scale orographic drag (SOD) using effective moist stability can improve wind biases and change the location of precipitation. The southeastern coastal region of China, particularly northern Fujian and southern Zhejiang, is characterized by complex terrain and persistent biases in the simulation of summer precipitation. Using the Yin-He Global Spectral Model (YHGSM), this study evaluates the impact of a moist orographic gravity wave drag (moist-OGWD) parameterization on summer precipitation simulations in this region. One control experiment using the original dry OGWD scheme and one sensitivity experiment using the moist-OGWD scheme are conducted for the summers of 2022 and 2023. Results demonstrate that moist-OGWD significantly mitigates the model underestimation of precipitation over northern Fujian and southern Zhejiang. The underlying mechanism involves latent heat release associated with moisture, which reduces static stability in the lower troposphere and facilitates the breaking of orographic gravity waves in the lower atmosphere. Consequently, low-level gravity wave drag is substantially enhanced, while upper-level drag and blocking drag are reduced. Collectively, these changes induce an anomalous cyclonic circulation in the lower troposphere. This circulation anomaly strengthens water vapor transport from the ocean and enhances low-level convergence, and augments the regional net moisture budget, ultimately leading to increased precipitation. Plain language summary This study evaluates the impact of moist orographic gravity wave drag parameterization (moist-OGWD) on summer monsoon precipitation over southeastern coastal of China (part of the East Asian Summer Monsoon region). Simulations using the YHGSM model demonstrate that the moist-OGWD scheme can bring more precipitation than dry-OGWD scheme. The enhancement stems from moisture-induced latent heat release, which reduces lower-tropospheric static stability, facilitating earlier breaking of orographic gravity waves in the lower atmosphere and substantially strengthening low-level zonal gravity wave drag. This intensified drag triggers an anomalous cyclonic circulation over southeastern coastal of China, thereby promoting more realistic moisture convergence and precipitation.

The Yin-He Global Spectral Model (YHGSM) is one of the most representative numerical weather prediction (NWP) models in China and has already been operationally applied to global weather forecasting. At present, YHGSM has achieved satisfactory parallel performance on high-performance computing (HPC) platforms to meet the real-time requirements of operational forecasting. However, when using the parallel two-dimensional (2D) domain decomposition algorithm, communication overhead still significantly impacts the overall performance of YHGSM. To address this issue, we introduce a pipelined optimization scheme in the Inverse Legendre Transform stage, aiming to reduce communication overhead through computation–communication overlap. Specifically, computation and communication tasks are grouped along the vertical dimension, where data dependencies are relatively weak, enabling the communication of one group to be overlapped with the computation of another. Experiments conducted on HPC platforms demonstrate the effectiveness of this approach and highlight its potential for broader application in spectral models. Under the best-case configuration, the proposed optimization reduces communication time by up to 75

Multi-hop reasoning over knowledge graphs (KGs) is hampered by path dependency: erroneous entity or relation selections at early stages propagate through subsequent reasoning steps, causing cascading errors. Existing KG reasoning agents follow single-track approaches that commit to one reasoning path before considering alternatives. We propose KG-ART (Knowledge Graph Active Refinement Trajectory), a dual-track adversarial framework that shifts KG reasoning from post-hoc correction to online adversarial exploration. At each reasoning step, KG-ART generates actions from both a policy path and a counterfactual path. A KG Trajectory Critic evaluates both paths before execution, enabling real-time switching when a superior alternative is detected. The framework combines a KG toolbox, knowledge memory, and instruction-tuned LLM planner with adversarial prompts targeting relation selection, entity disambiguation, and constraint exploration. Experiments on WebQSP, ComplexWebQuestions, and GrailQA show that KG-ART achieves state-of-the-art performance with 5.3

With the proliferation of advanced vector instruction sets in modern CPUs, achieving performance portability for large Fortran-based scientific codes has become increasingly challenging. This paper focuses on CLOUDSC, a computationally intensive and data-access-complex cloud microphysics parameterization scheme in the Integrated Forecasting System (IFS) of ECMWF. We propose Auto-CLOUDSC, an auto-generation framework that optimizes CLOUDSC through the following approaches: (1) an auto-generator consisting of three modules-a function interface generator, a code structure analyzer, and an expression parser-to convert Fortran source code into vectorized instruction sets; (2) a physics-combine algorithm that applies loop fusion to reduce redundant memory accesses; and (3) a cache-aware algorithm that leverages cache tiling and data layout optimizations to enhance data reuse. Experiments demonstrate that the auto-generated code achieves a speedup of 1.3 to 2.1 times over the original Fortran baseline on the Phytium FT2000+ ARMv8 processor, and 1.9 to 3.4 times on the Intel Xeon Platinum 8358 processor.

From 19 to 21 July 2021, extreme rainfall occurred in Henan, China, with a single hour of 201.9 mm breaking all records in China. The orographic drag of the mountains around Henan is an important influencing factor. In addition, during this extreme rainfall event with abundant moisture, moisture significantly affects vertical air movement and atmospheric stability. As a result, the orographic gravity wave drag (OGWD) parameterization based on the “dry air” assumption becomes less effective. It is necessary to analyze the impact of moisture effects on the vertical propagation of gravity waves and how these affect this extreme rainfall event. This study simulates the “21.7” extreme rainfall event in Henan, respectively, using the dry and moist OGWD parameterization. The results show that the moisture effect reduces the buoyancy frequency, which will lead to an increase in the amplitude of the gravity wave and a decrease in the Richardson number. These are conducive to the breaking of gravity waves, which increases orographic drag in the troposphere in the OGWD parameterization. The change in OGWD results in a cyclonic circulation difference in northern Henan, which weakens the airflow moving toward the northwest mountainous areas and enhances the airflow moving northward. These changes in circulation significantly alter the water vapor budget in the heavy rainfall region, increasing the net water vapor budget by approximately 33%, thereby reducing the negative biases of average rainfall by about 14.59%. These indicate the moisture effect can significantly alter the vertical propagation of gravity waves and improve the intensity of heavy rainfall.

Abstract There have been significant wet biases in the simulation of precipitation over the Tibetan Plateau (TP) for a long time. One of the important reasons is that current numerical models cannot accurately describe the influence of small‐scale orography. In the orographic gravity wave drag (OGWD) parameterization used to describe small‐scale orographic effects, this study develops a high‐precision calculation method of sub‐grid mountain sharpness that varies with geographical areas to revise and optimize the current constant mountain sharpness parameter (results in a constant drag coefficient at the surface). Batch numerical experiments on precipitation over the TP are carried out using the Yin‐He Global Spectral Model. The results show that the newly calculated drag coefficient of revised OGWD scheme improves the simulation of precipitation and atmospheric circulation over the TP. The new mountain sharpness overall increases the OGWD over the TP on the westerly circulation. According to the vertical vorticity equation, changes in OGWD enhance the positive vorticity over the southern TP. Compared to the original scheme, the revised OGWD experiment exhibits a cyclonic circulation difference over the southern TP, which weakens the flow of water vapor into the TP. The water vapor budget of the main precipitation area over the western TP decreases by about 12.66%, thus reducing precipitation and wet biases over the western TP.

Numerical weather prediction is the cornerstone of modern weather forecasting, yet its operational implementation demands vast computational resources. While artificial intelligence (AI)-based forecasting models offer a computationally efficient alternative, these purely data-driven approaches often sacrifice physical consistency. Here, we bridge physics-based and AI-based models through a novel, efficient hybrid framework that integrates a low-resolution atmospheric dynamical core with a neural operator in the multigrid architecture. This framework achieves performance comparable to that of state-of-the-art medium-range global weather forecasting models, while incurring much lower training costs, and simultaneously enhances the physical consistency that black-box models often lack. Furthermore, our framework provides substantial flexibility in the choice of dynamical cores, since the training process of the neural network does not require gradient propagation through the dynamical core, which ensures scalability to a wide range of operational forecasting systems.

Accurately modeling the long-term evolution of complex physical systems governed by partial differential equations (PDEs) remains a central challenge in operator learning. In this work, we introduce the Physics-Guided Mamba Neural Operator (PGMNO), a framework specifically designed to capture long-range spatiotemporal dependencies. PGMNO overcomes the limitations of conventional single-step and Transformer-based operators by unifying linear multistep numerical methods with structured state space models (SSMs). To ensure temporal stability, the framework employs multistep temporal modeling in the forward pass, complemented by an implicit backward differentiation formula (BDF)-based scheme during training. Simultaneously, PGMNO leverages the efficient kernel integration properties of SSMs to achieve scalable and stable operator learning. We validate PGMNO on diverse PDE benchmarks. Experimental results demonstrate that PGMNO consistently outperforms state-of-the-art models in prediction accuracy, computational efficiency, and long-term stability. Additionally, the framework exhibits strong resolution-invariant extrapolation and generalization across varying spatial discretizations. By prioritizing clear methodological articulation, this work underscores the potential of unifying numerical integration with state space modeling to build robust surrogate solvers for PDE-governed dynamical systems.

With the rise of scalable vector extension instruction sets in ARM processor architectures, large Fortran-based scientific codes face challenges in performance portability. This paper focuses on CLOUDSC, a computationally intensive and data-access complex cloud microphysics parameterization scheme from the Integrated Forecasting System of ECMWF. We propose Auto-CLOUDSC, an auto-generation framework that optimizes CLOUDSC by the following methods, including (1) An auto-generator that contains three modules: function interface generator, code structure analyzer, and expression parser to convert Fortran to vectorization instruction sets. (2) A physics-combine algorithm that applies loop fusion to reduce redundant memory access. (3) A cache-aware algorithm that uses cache tiling and data layout optimization to improve data reuse. Experiments demonstrate that the auto-generated code achieves a speedup of 1.3 to 2.1 times over the original Fortran baseline on the Phytium FT2000+ ARMv8 processor.

Accurately calculating the momentum exchange between the atmosphere and the ocean is essential for global weather forecasting. However, current widely used parameterization schemes face challenges in accurately estimating sea surface roughness z0 $\left({z}_{0}\right)$. Deep learning (DL) technology offers new opportunities to address this problem. Here, we present a DL-based parameterization scheme for sea surface roughness, which we have integrated into a global forecasting system. The DL-based scheme demonstrates advantages over four widely used traditional schemes in offline and online tests. In the offline tests based on observational data, the DL-based scheme significantly improved root mean square errors (RMSEs) for the calculated z0 ${z}_{0}$, with a decrease of at least 50.00% $50.00\,\%$ compared to the four traditional schemes. In the online tests based on the coupling of the DL-based scheme with a global forecasting system, the DL-based scheme also shows advantages over four traditional schemes, especially in forecasting the geopotential height and the wind field for the Southern Hemisphere.

Subgrid-scale orographic gravity wave drag (OGWD) significantly influences atmospheric circulation and weather systems. However, Current OGWD schemes, based on the "dry air" assumption, struggle to meet high-precision simulation demands. This study uses the moist OGWD scheme that incorporates moisture effects in gravity wave surface stress and vertical propagation of waves to simulate the global summer circulation in 2023 and three recent heavy rainfall events in China. In this scheme, moist buoyancy frequency varies with moisture: it decreases with abundant moisture and increases with less moisture, compared to the original scheme. Results show that buoyancy frequency differences alter low-level blocking height and drag, directly affecting gravity wave surface stress. During the vertical propagation of gravity waves, reduced tropospheric buoyancy frequency increases wave amplitudes and reduces Richardson number in moist scheme, which enhances tropospheric wave breaking and reduces waves propagation to the stratosphere. This increases tropospheric OGWD and decreases stratospheric OGWD, improving positive biases of troposphere westerly winds and negative biases of stratosphere easterly winds in Northern Hemisphere (NH) mid-high latitude, as well as the bias in the stratospheric jet near the Antarctic. The moist OGWD scheme also improves simulations of three recent heavy rainfall cases. In the Henan extreme rainfall, moist buoyancy frequency decreases due to abundant water vapor. Increased tropospheric OGWD weaken circulation and moisture transport to western and northern mountainous areas, intensifying rainfall and improving underestimation. The moist OGWD scheme partially addresses the limitations of "dry air" assumption, improving atmospheric circulation and heavy rainfall simulations.

Data-driven approaches for global weather forecasting have shown great potential. However, conventional architectures of these models struggle with spherical distortions, leading to unstable autoregressive forecasts. Although methods such as spherical Fourier neural operator (SFNO) based on spherical harmonic convolution can alleviate these problems, they face the challenge of high computational cost. Here, we introduce a spherical multigrid neural operator (SMgNO) that integrates spherical harmonic convolution and low resolution SFNO in the multigrid framework, effectively alleviating data distortions while requiring few computational resources. Experiments for spherical shallow water equations and medium-range global weather forecasting demonstrate the effectiveness and robustness of SMgNO. For 500 hPa geopotential height with a 7 days lead time, SMgNO achieves a 9.31% and 6.83% improvement in anomaly correlation coefficient over IFS T42 and SFNO, respectively. Furthermore, SMgNO requires only 10% floating-point operations of SFNO for forward propagation and 30.90% less GPU memory than SFNO during training.

Traditional numerical methods face a trade-off between computational cost and accuracy when solving partial differential equations. Low-order solvers are fast but less accurate, while high-order solvers are accurate but much slower. To address this challenge, we propose a novel framework, the hybrid multigrid neural operator (HMgNO). The HMgNO couples a low-order numerical solver with a multigrid neural operator, and the neural operator is used to correct the low-order numerical solutions to obtain high-order accuracy at each fixed time step size. Thus, the HMgNO achieves accurate solutions while ensuring computational efficiency. Moreover, our framework supports multiple types of low-order numerical solvers, such as finite difference and spectral methods. Experiments on the Navier-Stokes, shallow-water, and diffusion-reaction equations demonstrate that the proposed framework achieves the lowest relative error and smallest spectral bias with few model parameters and fast inference speed.

Spectral models, due to their stability and efficiency, have become one of the most popular approaches for implementing numerical weather prediction systems. Given the complexity of these models, they often require the use of multi-node computing resources for parallel processing to meet the stringent real-time requirements. However, as the number of nodes increases, the efficiency of inter-node communication becomes a critical bottleneck. In the case of the Yin-He Global Spectral Model (YHGSM), developed by the National University of Defense Technology, communication overhead is very high during the Fourier transform section, which consists of the transform itself and the subsequent transposition from the z-μ decomposition to the z-m decomposition. To address this challenge, we introduce an optimized scheme that overlaps communication with computation. By grouping corresponding communication and computation tasks, this approach leverages non-blocking communication techniques within MPI, combined with the use of asynchronous communication progress threads. Our experimental results demonstrate that this scheme can reduce execution time by up to 30% compared to the non-overlapped version, thereby significantly hiding communication overhead and enhancing the efficiency of YHGSM.

Long-term stability stands as a crucial requirement in data-driven medium-range global weather forecasting. Spectral bias is recognized as the primary contributor to instabilities, as data-driven methods difficult to learn small-scale dynamics. In this paper, we reveal that the universal mechanism for these instabilities is not only related to spectral bias but also to distortions brought by processing spherical data using conventional convolution. These distortions lead to a rapid amplification of errors over successive long-term iterations, resulting in a significant decline in forecast accuracy. To address this issue, a universal neural operator called the Spherical Harmonic Neural Operator (SHNO) is introduced to improve long-term iterative forecasts. SHNO uses the spherical harmonic basis to mitigate distortions for spherical data and uses gated residual spectral attention (GRSA) to correct spectral bias caused by spurious correlations across different scales. The effectiveness and merit of the proposed method have been validated through its application for spherical Shallow Water Equations (SWEs) and medium-range global weather forecasting. Our findings highlight the benefits and potential of SHNO to improve the accuracy of long-term prediction.

In a typical non-hydrostatic spectral dynamic numerical weather prediction (NWP) kernel, all forecast variables are transformed between grid point and spectral spaces to compute their gradients and solve the implicit problem. This kernel requires numerous spectral transformations, which depend heavily on extensive global communication and significantly hinder parallel computing efficiency. This paper introduces an innovative non-hydrostatic spectral kernel that incorporates a finite-volume method within the spectral framework. We have developed a horizontal divergence (D)-based structure equation, allowing direct computation of most prognostic variables and their horizontal gradients at grid point space. By doing so, the need for spectral transformations is substantially decreased. Our experiments demonstrate that this new approach reduces the cost of spectral transformations by up to 40%, enhancing the overall model efficiency by 15%-22%. Additionally, a series of tests confirmed the accuracy and stability of this new solver. This paper proposes a new non-hydrostatic spectral solver by applying a finite-volume method (FVM) into the spectral framework, where a horizontal divergence (D)- based structure equation is constructed and most of the prognostic variables as well as their horizontal gradients are calculated in grid point space directly. The number of spectral transforms is dramatically reduced. In practice, the cost of spectral transform with the proposed solver is reduced up to 40%. image

The Yin‐He Global Spectral Model (YHGSM) is a dry‐mass conserving hydrostatic global spectral model, relying on spectral transforms to compute horizontal derivatives. We present an extension of YHGSM core named YHGSM‐FVM which uses a second‐order finite‐volume method (FVM) to compute the horizontal derivatives in grid‐point space instead of the spectral approach. With this approach, the computational efficiency of the spectral model is improved since part of the spectral transforms is superseded by FVM which only needs local data and the computational demand is lower. More importantly, YHGSM‐FVM is still a spectral model solving the Helmholtz equation directly in spectral space with a highly efficient semi‐implicit semi‐Lagrangian advection scheme. The comparisons between YHGSM‐FVM and YHGSM are conducted, and the results show that both models have comparable prediction skill, but YHGSM‐FVM outperforms YHGSM in computational efficiency.

The newly developed nonhydrostatic（NH） global spectral dynamical core is evaluated by using three-dimensional（3D） benchmark tests with/without moisture. This new dynamical core differs from the original Aladin-NH like one in the combined use of a dry-mass vertical coordinate and a new temperature variable, and thus, it inherently conserves the dry air mass and includes the mass sink effect associated with precipitation flux. Some 3D dry benchmark tests are first conducted, including steady state, dry baroclinic waves, mountain waves in non-sheared and sheared background flows, and a dry Held–Suarez test. The results from these test cases demonstrate that the present dynamical core is accurate and robust in applications on the sphere, especially for addressing the nonhydrostatic effects.Then, three additional moist test cases are conducted to further explore the improvement of the new dynamical core.Importantly, in contrast to the original Aladin-NH like one, the new dynamical core prefers to obtain simulated tropical cyclone with lower pressure, stronger wind speeds, and faster northward movement, which is much closer to the results from the Model for Prediction Across Scales（MPAS）, and it also enhances the updrafts and provides enhanced precipitation rate in the tropics, which partially compensates the inefficient vertical transport due to the absence of the deep convection parameterization in the moist Held–Suarez test, thus demonstrating its potential value for full-physics global NH numerical weather prediction application.

Quantitative precipitation nowcasting (QPN) is a highly challenging task in weather forecasting. The ability to provide precise, immediate, and detailed QPN products is necessary for a variety of situations, including storm warnings, air travel, and large gatherings. To address this challenge, this article proposes a new transformer lightweight physics-informed transformer (LPT)-QPN for QPN tasks, utilizing vertical cumulative liquid water content (VIL) products. This model adopts novel transformer modules to model the long-term evolution of precipitation and incorporates multihead squared attention (MHSA) to model its highly nonlinear relationships while reducing computational complexity. The results of experimental evaluations demonstrate the superiority of LPT-QPN when compared to existing state-of-the-art QPN models. In particular, the LPT-QPN model demonstrates greater accuracy for long lead time and in high-intensity areas, confirmed in both quantitative and qualitative evaluations. In addition, through three customized fine-tuning schemes, we are able to further improve the predictability of the LPT-QPN model for specific precipitation events. By incorporating the physical constraints of the convection-diffusion equation, our approach offers novel perspectives for future explorations that combine physical prior knowledge and deep-learning (DL) techniques.

研究生课程"并行算法"是在大规模科学与工程计算需求越来越大的情况下开设的,旨在让学生学会如何实现并行计算的方法,熟练掌握并行计算的实现过程.基于课程教学经验和实践,论述了侧重并行计算思维训练和理论实践相结合的教学理念,以及基于该理念与模块化方式,从基本概念、基本技术、具体算法到编程实践等层面对教学内容的优化设计,以提升学生的学习效果与综合素质.