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.

As high-performance computing (HPC) platforms continue to scale up, communication costs have become a critical bottleneck affecting overall application performance. An effective strategy to overcome this limitation is to overlap communication with computation. The Message Passing Interface (MPI), as the de facto standard for communication in HPC, provides non-blocking communication primitives that make such overlapping feasible. By enabling asynchronous communication, non-blocking operations reduce idle time of cores caused by data transfer delays, thereby improving resource utilization. Overlapping communication with computation is particularly important for enhancing the performance of large-scale scientific applications, such as numerical simulations, climate modeling, and other data-intensive tasks. However, achieving efficient overlapping is non-trivial and depends not only on advances in hardware technologies such as Remote Direct Memory Access (RDMA), but also on well-designed and optimized MPI implementations. This paper presents a comprehensive survey on the principles of MPI non-blocking communication, the core techniques for achieving computation–communication overlap, and some representative applications in scientific computing. Alongside the survey, we include a preliminary experimental study evaluating the effectiveness of asynchronous progress mechanism on modern HPC platforms to support the development of parallel programs for HPC researchers and practitioners.

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.

The Fire Dynamics Simulator (FDS) is widely used for fire simulation but faces scalability challenges due to its limited grid partitioning capabilities. To address this issue, we propose Para-FDS, a scalable multilevel parallel scheme for accelerating FDS targeted at multicore architectures. Para-FDS integrates three key optimizations: (1) an adaptive grid partitioning algorithm to enhance scalability, (2) a communication localization optimization approach to reduce overhead between computing nodes, and (3) a NUMA-aware process mapping strategy to improve core utilization within Non-Uniform Memory Access (NUMA) architecture. Implemented on the Tianhe next-generation supercomputer, Para-FDS achieves a speedup of up to 214× on a practical example. It further reduces communication overhead by up to 38

Numerical weather prediction (NWP) is the core technology for weather forecast and disaster prevention and mitigation. The research and operational applications of NWP have always been highly valued in China, and have achieved great progress with an appreciable international influence in the theories, algorithms, and operational system developments. This paper first summarizes the scientific and technological evolution of NWP in China, and then focuses on the current status and recent updates of the two homemade global NWP systems: GRAPES (Global/Regional Assimilation and PrEdiction System) and YHGSM (YinHe Global Spectral Model). (1) GRAPES possesses both deterministic and ensemble forecast systems, with global (regional) model versions running on 12–50-km (3–10-km) resolutions. Significant improvements have been made on its dynamic core, four-dimensional variational (4D-Var) assimilation, satellite and radar data assimilation, ensemble forecast, and cloud microphysics schemes, and so on. It is capable to perform subseasonal to seasonal forecast and has incorporated an atmospheric chemistry model, typhoon numerical forecast model, and ocean wave model. (2) YHGSM continues to follow the development route of spectral models, featured prominently with a dry-mass conserved spectral dynamical core, ensemble 4D-Var assimilation, coupled ocean–land–atmosphere ensemble forecast, and the medium-term and monthly-extended global high-resolution forecast as the baseline. These NWP systems autonomouly developed by the China Meteorological Administration and the national defense insitution benefit from long-term adherence to the national science and technology development strategies and close research to operation practices.

Atmospheric General Circulation Model (AGCM) is widely used for predicting climate change and extreme weather. AGCM has been scaled up to thousands of cores on HPC platforms to provide timely forecasts. However, the communication overhead significantly impacts the performance and scalability of AGCM when using a two-dimensional domain decomposition algorithm in parallel. To tackle this issue, we introduce Pipe-AGCM, a fine-grain pipelining scheme to reduce communication overhead. We divide latitudes into groups to overlap communication from one group with the computation of another. By adjusting the group count, we can establish a pipelining scheme, optimizing AGCM performance. Pipe-AGCM has demonstrated efficiency and scalability in experiments on HPC platforms. The pipelining scheme reduced communication time and wall clock time by 78.02

The phenomenon of seasonal hysteresis between urban heat islands (UHIs) and background surface temperature was reported in many studies, but a unifying explanation for the causes of its magnitude, shape and looping direction remained less clear, and the impacts of urban landscapes and background climate on such hysteresis were also the subject of enquiry. Here we performed a systematic study based on a long-term offline simulation for surface energy budgets, surface air and skin temperature of the urban-biosphere system in East China. It was carried out with an up-to-date land surface model - the CLM5-LCZs (the Community Land Model version 5 (CLM5) coupled with a newly developed urban canopy model representing diverse urban landscapes with the local climate zones classification scheme). To isolate the causes of UHIs hysteresis, a simplified framework was constructed by combining a mathematical representation of grid/subgrid model outputs, the two-resistance mechanism method and the time-lagged cross correlation analysis. The results exhibited distinct hysteresis patterns of UHIs such as twisted, concave-down, concave-up and convex-up curve across diverse background climates and urban landscapes. The basic shapes of hysteresis depict a gradual transition from concave/convex-up to concave/convex-down by characterized by a larger winter UHIs from a wetter and warmer climate to a drier and colder climate. Magnitude and seasonality of surface UHIs are separately dominated by the urban-rural contrast of evapotranspiration and heat storage at daytime and nighttime, whereas across-climate variations of daytime and nighttime surface UHIs hysteresis are regulated by the phase shift of convection efficiency and anthropogenic heat. Urban landscapes affect the hysteresis mainly by altering the amplitude rather than the shape of looping curves. Our work could provide a fuller picture of how to mitigate urban warming considering inter-seasonal tradeoffs over a broader region. The seasonality of urban heat islands (UHIs) has been investigated toward its characteristics and controls, of which analyses were conducted on several cities and left intra-urban variations out of account. It was essential to understand whether the local seasonality-arguments can be extended to broader geographic regions or lots of cities with different landscapes. In this study, an updated urban climate model describing intra-urban heterogeneity better was utilized to simulate surface energy fluxes and UHIs. Based on the subgrid outputs of this model and the hypothesis of phase shift mechanism, we presented a simplified framework to isolate the contribution of time lags of different biophysical factors to the magnitude, shape and looping direction of UHIs hysteresis. These hysteretic characteristics are varied with background climates and urban landscapes. The time lags of evaporative cooling and heat storage corresponding to background surface temperature between urban and rural areas account for different hysteresis patterns at daytime and nighttime. Hysteretic typology for different urban classes or cities in diverse background climates is regulated by the time lead/lag of other secondary biophysical factors. The results could serve as a valuable reference for coping with the growing heat risk in cities. The seasonality of surface and canopy urban heat island exhibited hysteresis-like curves, whose looping patterns vary across background climate and urban features A simplified conceptual model to explain hysteretic behaviors was constructed based on a second-order Fourier representation, the attribution method and a newly developed land surface model The urban-biosphere time lags of evaporative cooling rates and storage heat, dominated such hysteresis at daytime and nighttime respectively

Knowledge graph reasoning can deduce new facts and relationships, which is an important research direction of knowledge graphs. Most of the existing methods are based on end-to-end reasoning which cannot effectively use the knowledge graph, so consequently the performance of the method still needs to be improved. Therefore, we combine causal inference with reinforcement learning and propose a new framework for knowledge graph reasoning. By combining the counterfactual method in causal inference, our method can obtain more information as prior knowledge and integrate it into the control strategy in the reinforcement model. The proposed method mainly includes the steps of relationship importance identification, reinforcement learning framework design, policy network design, and the training and testing of the causal reinforcement learning model. Specifically, a prior knowledge table is first constructed to indicate which relationship is more important for the problem to be queried; secondly, designing state space, optimization, action space, state transition and reward, respectively, is undertaken; then, the standard value is set and compared with the weight value of each candidate edge, and an action strategy is selected according to the comparison result through prior knowledge or neural network; finally, the parameters of the reinforcement learning model are determined through training and testing. We used four datasets to compare our method to the baseline method and conducted ablation experiments. On dataset NELL-995 and FB15k-237, the experimental results show that the MAP scores of our method are 87.8 and 45.2, and the optimal performance is achieved.

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.

In the semi-Lagrangian interpolation scheme of Yin-he Global Spectral model (YHGSM), communication needs to be performed before interpolation, resulting in significant communication overhead. To solve this problem, we propose an optimized scheme that overlaps computation with communication based on grouping levels. The scheme divides the vertical levels into three groups and overlaps the computation of one group with communication of another. Experimental results show that our scheme can reduce the running time for the semi-Lagrangian scheme by 12.5% and effectively reduce the communication overhead of the model, improving the efficiency of YHGSM.

In complex systems constrained by multiple factors, it is very important to identify the key influencing factors for mastering the evolution and development law of a system and for obtaining scientific decision-making suggestions or schemes. At present, the method based on experimental simulation is limited by the difficulty of system model construction; DEMATEL (Factual Decision Trial and Evaluation Laboratory) is inevitably influenced by subjective factors. In view of this, we propose a novel model based on heuristic causal inference. By combining the network analysis in complex network science, the model defines the global/local causal pathway and the causal pathway’s length in the causal network and takes the causal pathway contribution degree as an indicator to measure the approximate causal effects. The model includes steps such as causal network learning, causal pathway contribution degree calculation, and key influencing factor identification. The model uses the Fast Causal Inference (FCI) algorithm with prior knowledge to learn the global causal network of the complex system and uses the heuristic causal inference to calculate the causal pathway contribution degree. The heuristic method draws on the idea of complex network topology analysis and measures the influence degree between variables by the number and distance of causal pathways. The key influencing factors are finally identified according to the causal pathway contribution degree. Based on the SECOM dataset, we carried out simulation experiments and demonstrated the feasibility and effectiveness of the proposed method.

The basic orographic dataset in numerical weather prediction (NWP) is critical for accurate weather and climate predictions. However, the original high-resolution orographic elevation data cannot be used directly in NWP models because of the presence of some scale features that are detrimental to NWP. Therefore, high-resolution orographic elevation data requires some pre-processing operations to filter out harmful information before it can be used to NWP. The global NWP spectral models have the advantages of high calculation accuracy and fast convergence speed. However, due to the large amount of computation and long computation time in the NWP spectral model, most of the research and development on orographic datasets for NWP is aimed at grid-point models. Aiming at the autonomously construction, this study uses the earth observation data from two sources to make a basic high-precision orographic dataset suitable for the spectral model. These methods are useful for producing orographic datasets in the global NWP spectral model.

The prediction of chaotic time series systems has remained a challenging problem in recent decades. A hybrid method using Hankel Alternative View Of Koopman (HAVOK) analysis and machine learning (HAVOK-ML) is developed to predict chaotic time series. HAVOK-ML simulates the time series by reconstructing a closed linear model so as to achieve the purpose of prediction. It decomposes chaotic dynamics into intermittently forced linear systems by HAVOK analysis and estimates the external intermittently forcing term using machine learning. The prediction performance evaluations confirm that the proposed method has superior forecasting skills compared with existing prediction methods.

The Yin-He global spectral model (YHGSM), embodies a parallel semi-Lagrangian solver and has two schemes implemented: maximum wind speed scheme and on-demand communication scheme. Maximum wind speed communication adopts a single and fixed data structure, which has a large communication overhead. Although the overhead of on-demand communication is reduced, it is still pretty huge. In this paper, a novel adaptable approach is proposed in which a monthly maximum wind speed is used in the YHGSM. This approach reduces the difference between the actual wind speed and the maximum wind speed used in the model; in turn, the communication overhead in the trajectory computation is further reduced. Experiments show that in the maximum wind speed scheme and on-demand schemes, the communication overheads with the adaptive maximum wind speed are significantly reduced. In addition, in a ten-day forecast with the on-demand communication scheme, the total overhead for the semi-Lagrangian computing and the total parallel execution time are also both reduced, and the reduction ratio increases as the number of nodes increases.

All kinds of meteorological elements near the airport directly affect the dissipation process of fog, which also affects the visibility. The correlation analysis of temperature, humidity, dew point temperature, air pressure, wind speed and other meteorological factors with visibility Runway Visual Range (RVR)and Meteorological Optical Visual Range (MOR)was carried out by SPSS. It was found that the four variables, TEMP, RH, WS2A and PAINS, had great influence. In this paper, the above four variables are selected as inputs to nonlinear fit the RVR value of visibility. Combined with Radial Basis Function (RBF), it is found that SVM-RBF model has better generalization ability and fitting degree.

The Yi‐he global spectral model (YHGSM) is a numerical weather prediction (NWP) model developed by the National University of Defense Technology, China. The YHGSM has two optimization techniques that have recently been adopted for the on‐demand communication version of the semi‐Lagrangian scheme: one‐sided communication and an adaptive strategy for the maximum wind speed. However, the limitation of the original model is that the computation and communication cannot be performed simultaneously, resulting in a waste of resources. This article addresses this limitation by dividing the array to be interpolated into three independent subarrays. The communication and interpolation operations for each subarray are rearranged into three groups. This allows the overlap of the communication with computation between different groups. In addition, by adjusting the synchronization operations, the degree of overlap is further increased. In the optimized scheme, the per‐step overhead of the interpolation phase is greatly reduced, and its speedup decreases first and increases afterwards with the increase of the number of cores. Therefore, when the number of cores is large, the benefits from overlapping the communication with computation will be more substantial.

The covariance matrix estimated from the ensemble data assimilation always suffers from filter collapse because of the spurious correlations induced by the finite ensemble size. The localization technique is applied to ameliorate this issue, which has been suggested to be effective. In this paper, an adaptive scheme for Schur product covariance localization is proposed, which is easy and efficient to implement in the ensemble data assimilation frameworks. A Gaussian-shaped taper function is selected as the localization taper function for the Schur product in the adaptive localization scheme, and the localization radius is obtained adaptively through a certain criterion of correlations with the background ensembles. An idealized Lorenz96 model with an ensemble Kalman filter is firstly examined, showing that the adaptive localization scheme helps to significantly reduce the spurious correlations in the small ensemble with low computational cost and provides accurate covariances that are similar to those derived from a much larger ensemble. The investigations of adaptive localization radius reveal that the optimal radius is model-parameter-dependent, vertical-level-dependent and nearly flow-dependent with weather scenarios in a realistic model; for example, the radius of model parameter zonal wind is generally larger than that of temperature. The adaptivity of the localization scheme is also illustrated in the ensemble framework and shows that the adaptive scheme has a positive effect on the assimilated analysis as the well-tuned localization.

针对有限元分析的计算问题,在现有采用全局通信方案的简单并行算法基础上,对其所涉核心算法,采用稀疏数据结构与局部通信进行并行算法优化设计,有效减少了通信所涉及的处理器个数与通信量.同时,通过采用非阻塞通信,并将与通信无关计算进行分离与前置的方法,进行计算与通信重叠,以有效隐藏通信开销的影响.实验结果表明,优化所得算法相比现有算法具有明显改进,特别是对稀疏矩阵稠密向量乘与单元贡献装配,改进很大.同时,随着任务个数的增加,改进效果越明显.

Studying the influences of meteorological factors on air pollutants (including SO2, NO2, O3 and PM10) and quantifying the relationships between them should enable further investigation on the variation of air pollutants in a region, especially their temporal-spatial distribution characteristics. We combined observed air pollutants data from ground-based monitoring sub-stations and relevant meteorological data in the Pearl River Delta region to investigate relationships between air pollutants and various meteorological factors, and then established regression models for each sub-station through principal component regression (PCR). According to these regression equations, the seasonal average temporal-spatial distribution of air pollutants is determined by inversion of the distance priority principle, and it provides strong support for the study of the spatial distribution and temporal variation of air pollutants in this region. Our conclusions are as follows: Different air pollutants have different spatial distribution characteristics with significant seasonal changes, and the concentration of each air pollutant in winter was higher than that in summer; PM10 levels in this region were higher than SO2, NO2 and O3.