共查询到20条相似文献，搜索用时 93 毫秒
1.
海洋数值模式当前已经成为海洋研究和预测的核心工具,其高分辨率。多物理过程的发展趋势对数值模式计算速度也提出了越来越高的要求。为了使得海洋数值模式更好的适应计算机基础架构,充分发挥现代化计算机体系的计算特点,提高计算效率,本文提出了一种简单易行且有效的代码现代化优化方案,并以MASNUM海浪模式为例进行了测试。首先利用诊断工具Intel Vtune Amplifier XE和Intel Trace Analyzer Collector,对模式的性能和负载均衡性进行了分析;之后,针对热点函数,在单节点上制定了4个优化步骤,包括编译器选项优化,串行和标量优化,向量化和OpenMP并行优化。结果表明,经过优化后,单节点内模式的计算速度可以提高1.95倍,多节点的模式强扩展性呈线性。这表明本文提出的代码现代化方案是一种行之有效的优化方法。 相似文献
2.
海洋数值模式是定量描述海洋物理现象及其变化的数值模型,也是海洋与气候研究、预测的核心工具。随着海洋观测的不断投入与积累、对海洋认识的不断深入,特别是在高性能计算技术的支撑下,海洋数值模式已有了长足进步,正朝着高分辨率和多物理过程的方向发展。随着分辨率的提高、物理过程的细化,海洋数值模式的发展面临着多个方面的挑战。当前,海洋数据数量和种类不断增多,同时超级计算机、高性能计算和深度学习等技术的快速发展,为海洋数值模式的突破提供了机遇与挑战。本研究回顾了海洋数值模式的发展现状,梳理和分析了其发展中遇到的大规模高效并行计算和参数优化这两个关键问题,探讨和展望了当前海量数据驱动下海洋数值模式的发展趋势。提出计算负载均衡、计算与I/O重叠的并行流水线设计以及降低全局交换的算法改进是当前突破高分辨率海洋模式大规模高效并行效率的关键。从海洋科学、高性能计算以及深度学习深度交叉融合的角度,提出了实现海洋科学与深度学习相结合的6个途径,在此基础上,探讨了基于深度学习的参数化优化可能实现的途径。 相似文献
3.
南海上层海洋环流两层半模式的数值模拟Ⅰ.闭边界海盆季节性环流 总被引：10，自引：4，他引：10
本文应用一个包含动力学和热力学的热带海洋两层半模式,模拟南海上层海洋闭边界条件下的季节性环流.模式海洋由不同密度的上混合层、季节性温跃层和静止的深水层组成.模式考虑挟卷(卷入entrainment和卷出detrainment)引起海洋上下活动层间的质量、动量与热量的交换,在海面月平均气候风场动力强迫和通过海面热通量的热力影响下,计算了封闭海盆假定下的南海上层海洋的季节环流.用数值试验的方法讨论了非线性效应、摩擦阻尼、侧向混合对大尺度环流的影响,并得出有关模式稳定性的结论.模拟结果与南海海洋实测和动力诊断的环流趋势吻合较好,显示了模式对南海海盆尺度的环流系统有较好的模拟能力. 相似文献
4.
5.
6.
7.
中国近海环流数值模拟研究综述 总被引：1，自引：0，他引：1
按照研究时间由早期到现代,研究方法由简单到复杂的顺序,综述了近几十年来海洋数值模拟工作在中国近海所取得的研究成果.目前,通过数值模式已经能够基本合理地模拟中国近海的海洋环流,黑潮、台湾暖流、黄海暖流等的路径、流量,对马暖流源区等问题的特征在数值模拟结果中也逐渐清晰.通过分析已有的工作,认为还存在多项关键技术问题,并受对中国近海环流认知程度的限制,中国近海环流的数值模拟研究还需要进一步的发展和完善.在未来的数值模拟研究中,通过变网格或多重网格嵌套技术,实现中国近海高分辨率的数值模式;应用更加符合实际情况的海洋参数化方案,发展风-浪-流-潮耦合模式;结合资料同化技术,发展预报模式等都应是中国近海环流数值模拟研究的重点发展方向. 相似文献
8.
使用序列化和不规则矩形划分的方法,开发了基于消息传递接口(MPI)的环流数值模式并行算法,并通过邻近点交换策略(NPES)进一步提升了计算效率。在HP C7000刀片系统上设计了两个实验,数值结果表明,使用了NPES的并行版本(PVN)的计算效率要高于原并行版本(PV)。在第二个实验中当进程数为100时PVN的计算效率仍可达到0.9以上,然而此时PV的计算效率已迅速下降至0.39。并用环流模式的PVN版本对一个高分辨率的区域进行了模拟,效果较好。该算法具备普适性,可用于其它环流模式的高效并行。 相似文献
9.
为研究大尺度海洋环流和一些大尺度海洋物理过程,我们建立了一个无海底地形的三层斜压海洋环流数值模式。模式方程采用了静力近似和Boussinesq近似,在上边界采用了钢盖近似。同时,还做了两个数值试验:第一个模拟试验,海洋从静止状态开始,以4个季节的气候平均风应力和海表大气季节平均温度为强迫条件,积分区城选择45℃以北的太平洋,数值积分了5个模式年,这时海洋的上层环流和温度的年变化趋于平稳。计算结果再现了太平洋气候平均状态的海温分布和主要海流的大尺度特征以及它们的年变化特征。如西部强化流、赤道潜流等。第二个数值试验,在第一个数值模拟的基础上给定赤道西太平洋一个风应力异常区,数值积分了几个月。从计算结果中可看到类似E1 Ni■o的现象,并且初步地分析了赤道波动在E1 Ni■o形成过程中的作用。 相似文献
10.
分别用Ekman有限深海风漂流计算公式,普林斯顿海洋模式（POM）数值计算了南黄海风漂流。通过对这两种方法的计算结果的对比分析得出：在水深较小时,模式计算的风漂流场与公式计算结果较为接近;而在水深较大时公式计算所得风漂流场与实际流场相差较大,模式计算的风漂流场则基本再现了夏季南黄海的环流结构;风速越大,两种计算方法所得的流场之间的差别越大。同时从数值试验的角度还得到夏季由于风速普遍偏小,风场对环流结构的影响都相对较小,热盐效应是控制夏季环流的主导因素。但当风速增大到一定程度时,环流结构会发生根本改变。 相似文献
11.
在大洋环流模式中，铅直混合的参数化方法起着关键性的作用。将大洋细结构混合参数化方法首次应用于世界大洋环流模式中。使用中科院大气所(LASG)发展的20层世界大洋环流模式(OGCM)ML20，月平均风场作为强迫场，利用ML20模式在稳定初始状态下运行300a后的计算结果作为本实验进行数值模拟的初始场。该参数化方法对世界大洋环流模式的影响主要表现为：永久性温度跃层的厚度明显变薄；对深层水和底层水的模拟有改进；对南极中层水的模拟比较成功；但是对赤道海区的模拟结果欠佳。 相似文献
12.
在大洋环流模型中对应于惯性内波破碎的垂直混合的一个参数化方案 总被引：1，自引：0，他引：1
Based on the theoretical spectral model of inertial internal wave breaking(fine structure) proposed previously, in which the effects of the horizontal Coriolis frequency component f-tilde on a potential isopycnal are taken into account, a parameterization scheme of vertical mixing in the stably stratified interior below the surface mixed layer in the ocean general circulation model(OGCM) is put forward preliminarily in this paper. Besides turbulence, the impact of sub-mesoscale oceanic processes(including inertial internal wave breaking product) on oceanic interior mixing is emphasized. We suggest that adding the inertial internal wave breaking mixing scheme(F-scheme for short) put forward in this paper to the turbulence mixing scheme of Canuto et al.( T-scheme for short) in the OGCM, except the region from 15°S to 15°N. The numerical results of F-scheme by using WOA09 data and an OGCM(LICOM, LASG/IAP climate system ocean model) over the global ocean are given. A notable improvement in the simulation of salinity and temperature over the global ocean is attained by using T-scheme adding F-scheme, especially in the mid- and high-latitude regions in the simulation of the intermediate water and deep water. We conjecture that the inertial internal wave breaking mixing and inertial forcing of wind might be one of important mechanisms maintaining the ventilation process. The modeling strength of the Atlantic meridional overturning circulation(AMOC) by using T-scheme adding F-scheme may be more reasonable than that by using T-scheme alone, though the physical processes need to be further studied, and the overflow parameterization needs to be incorporated. A shortcoming in F-scheme is that in this paper the error of simulated salinity and temperature by using T-scheme adding F-scheme is larger than that by using T-scheme alone in the subsurface layer. 相似文献
13.
Applications of adjoint data assimilation, which is designed to bring an ocean circulation model into consistency with ocean
observations, are computationally demanding. To improve the convergence rate of an optimization, reduced-order optimization
methods that reduce the size of the control vector by projecting it onto a limited number of basis functions were suggested.
In this paper, we show that such order reduction can indeed speed up the initial convergence rate of an assimilation effort
in the eastern subtropical North Atlantic using in situ and satellite data as constraints. However, an improved performance of the optimization was only obtained with a hybrid approach
where the optimization is started in a reduced subspace but is continued subsequently using the full control space. In such
an experiment about 50% of the computational cost can be saved as compared to the optimization in the full control space.
Although several order-reduction approaches seem feasible, the best result was obtained by projecting the control vector onto
Empirical Orthogonal Functions (EOFs) computed from a set of adjusted control vectors estimated previously from an optimization
using the same model configuration. 相似文献
14.
韩忠南 《海洋学报(英文版)》1987,6(4):537-546
A computational method for the ocean ship routing based on optimization theory is presented in this paper. In the computational programme "breaking circle method" is used, by which a principal contradictory line can be obtained in the operations research. The automatic selection of calculating network and the initial values have been realized. The quantitative estimation for the change of navigating speed under the different conditions of ocean wave and current has been done. The objectivly quantitative computational method is provided for ocean ship routing and forecast service following ship tracks. The calculating examples and practical experiments of ships have proved that this computational method is highly effective. 相似文献
15.
The response time of the large-scale oceanic circulation due to freshwater perturbations is investigated with models of different complexity. A three-dimensional ocean general circulation model (OGCM) and a zonally averaged ocean model are employed. In order to distinguish advection and diffusion from fast baroclinic processes (e.g. waves in the OGCM) a color tracer is injected at the same time and location as the freshwater discharges. In spite of the inability of the zonally averaged model to represent wave processes in a realistic way similarities with the OGCM are found for the propagation patterns of density anomalies and of color tracer. In the OGCM as well as in the zonally averaged model, density anomalies propagate faster than anomalies of a passive color tracer in the case of vertical density stratification. The progression of density anomalies leads to changes of the oceanic circulation, and both oceanic models exhibit circulation changes in areas distant from the discharge places long before the passively entrained color tracer has reached these regions. The fact that a zonally averaged model simulates baroclinic processes faster than advection even if internal gravity waves are not represented due to neglected acceleration terms, is clarified with a conceptual box model. 相似文献
16.
Ayumi Fujisaki Hajime Yamaguchi Fengjun Duan Genki Sagawa 《Journal of Oceanography》2007,63(5):775-790
In this study, a numerical model of 7-day forecast of sea ice produced by the Japan Meteorological Agency was improved by
the following approaches. First, a new ice dynamic model was introduced: the distributed mass/discrete floe model. The model
takes account of discrete characteristics of ice floes and well simulates the ice edge location at low computational cost.
Secondly, the grid size was reduced to 5 × 5 km for the future high resolution forecasts. Next, the sea surface current data
was examined because it significantly influences sea ice movement. We applied two new datasets of HINO and Okhotsk Ocean General
Circulation Model (Okhotsk OGCM), which are estimated by numerical simulations, for the 7-day forecast of sea ice. Ice southward
speed in January and the whorl formations in February and March were well reproduced with Okhotsk OGCM datasets. Finally,
the ocean heat flux at the ice-ocean interface was refined. As a result, we achieved an ice edge error reduction from 30.8
km to 23.5 km. 相似文献
17.
国家气候中心全球海洋资料四维同化系统在热带太平洋的结果初步分析 总被引：6，自引：0，他引：6
介绍了国家气候中心(NCC)“全球海洋资料四维同化系统”(简称NCC-GODAS).该系统包含观测资料预处理系统、插值分析系统和所应用的动力模式.插值分析系统采用四维同化技术方案,在时间上设置一个4周的窗口,将此窗口之内的观测资料以一定的权重插入插值分析系统,在空间上采用三维变分方案.海洋动力模式为“九五”期间LASG研制的L30T63OGCM 1 0版本.文中考察了该系统从1982年到2003年3月在热带太平洋的部分同化分析结果,并与NCEP的再分析资料和EMC/NCEP的太平洋区域海洋资料同化系统的结果进行了对比分析.结果显示,该系统的同化结果(如SST,SSTA,Nino指数,次表层海温变化等)与NCEP的同期同化结果具有很好的一致性.同时,该系统得出的海洋同化资料气候场与海洋模式气候场相比,有显著的改进.表明该系统具有较好的同化能力,其同化结果可为海-气耦合模式进行季节和跨季节的气候预测业务提供可信的海洋初始场资料,同时还可以为相关研究提供海洋分析场. 相似文献
18.
The direct disposal of CO2 into the ocean interior represents a possible means to help mitigate rising levels of atmospheric CO2. Here, we use three different versions of an ocean general circulation model (OGCM) to simulate the direct injection of liquid CO2 near Tokyo. Our results confirm that direct injection can sequester large amounts of CO2 from the atmosphere when disposal is made at sufficient depth (80–100% of the carbon injected at 3000 m remains in the ocean after 500 years) but show that the calculated efficiency is rather sensitive to the choice of physical model. Moreover, we show, for the first time in an OGCM and under a reasonable set of economic assumptions, that sequestration effectiveness is quite high for even shallow injections. However, the severe acidification that accompanies injection and the impossibility of effectively monitoring injected plumes argue against the large-scale viability of this technology. Our coarse-grid models show that injection at the rate of 0.1 Pg-C/yr lowers pH near the site of injection by as much as 0.7–1.0 pH-unit. Such pH anomalies would be much larger in more finely gridded models (and in reality) and have potential to severely harm deep-sea organisms. We also show that, after several hundred years, one would effectively need to survey the entire ocean in order to accurately verify the inventory of injected carbon. These results suggest that while retention may be sufficient to justify disposal costs, other practical problems will limit or at best delay widespread deployment of this technology. 相似文献
19.
Non-linear evolution of uni-directional focussed wave-groups on a deep water: A comparison of models
Up until the point at which ocean waves break, their dynamics are generally assumed to be accurately modelled by potential flow theory. For practical and computational reasons it is often useful to approximate the full potential flow solution with bandwidth and amplitude limited equations. A approximation used for waves on deep water is the Broad-banded Modified Non-linear Schrödinger equation (also known as the modified Dysthe equation). In this paper we compare this approximate model with potential flow simulations of focussing uni-directional wave-groups. We find that for moderate non-linearity the approximate model predicts very similar changes to the potential flow model. However, one of the dominant non-linear changes to the wave-group is a localised increase in the bandwidth and contraction in physical length, and beyond a certain point the approximate model fails to accurately reproduce this causing other elements, such as the maximum wave amplitude, to be poorly modelled. This modelling inaccuracy occurs in cases where, based on the initial conditions of the simulation, the approximate model would be expected to be accurate. 相似文献
20.
An alternative ‘compressed’ coordinate system for representing finite difference grids in ocean models is presented and compared with the traditional Cartesian method. The alternative method represents any arbitrary three-dimensional domain as a one-dimensional vector. Advantages of using this method include the exclusion of all land ‘dry’ points from the computational grid, leading to savings in the memory used and increased execution speed where domains have a low ratio of wet to total computational cells. The Cartesian and compressed models are applied to a variety of domains to assess the performance of each in terms of memory use and computational speed. 相似文献