地理科学进展 ›› 2022, Vol. 41 ›› Issue (4): 731-740.doi: 10.18306/dlkxjz.2022.04.016

• 研究综述 • 上一篇    

高性能并行分布式水文模型研究进展

叶翔宇1,2(), 李强1,3, 郭禹含1,2, 梁廖逢1,2, 王中根1,4,*()   

  1. 1.中国科学院地理科学与资源研究所陆地水循环及地表过程重点实验室,北京 100101
    2.中国科学院大学资源与环境学院,北京 100049
    3.青岛大学,山东 青岛 266071
    4.应急管理部国家自然灾害防治研究院,北京 100085
  • 收稿日期:2021-08-09 修回日期:2021-09-27 出版日期:2022-04-28 发布日期:2022-06-28
  • 通讯作者: *王中根(1973— ),男,河南潢川人,研究员,主要从事水文水资源研究。E-mail: wangzg@igsnrr.ac.cn
  • 作者简介:叶翔宇(1996— ),男,浙江松阳人,博士生,研究方向为高性能水文模型、水文模型参数不确定性分析、山洪风险评估等。E-mail: yexy.18b@igsnrr.ac.cn
  • 基金资助:
    第二次青藏高原综合科学考察研究项目(2019QZKK0903);国家重点研发计划项目(2017YFB0203101)

Progress of research on high-performance parallel distributed hydrological model

YE Xiangyu1,2(), LI Qiang1,3, GUO Yuhan1,2, LIANG Liaofeng1,2, WANG Zhonggen1,4,*()   

  1. 1. Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
    2. College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China
    3. Qingdao University, Qingdao 266071, Shandong, China
    4. National Institute of Natural Hazards, Ministry of Emergency Management of the People's Republic of China, Beijing 100085, China
  • Received:2021-08-09 Revised:2021-09-27 Online:2022-04-28 Published:2022-06-28
  • Supported by:
    The Second Tibetan Plateau Scientific Expedition and Research Program(2019QZKK0903);National Key Research and Development Program of China(2017YFB0203101)

摘要:

传统分布式水文模型采用串行计算模式,其计算能力无法满足大规模水文精细化、多要素、多过程耦合模拟的需求,亟需并行计算的支持。进入21世纪后,计算机技术的飞速发展和并行环境的逐步完善,为分布式水文模型并行计算提供了软硬件支撑。论文从并行环境、并行算法2个方面对已有研究进行总结概括,分析了不同并行环境和并行算法的优势与不足,并提出提高模型并行效率的手段,即合理分配进程/线程数缩减通信开销,采用混合并行环境增强模型可扩展性,空间或时空离散化提高模型的可并行性及动态分配计算任务、平衡工作负载等。最后,论文对高性能并行分布式模型的未来研究方向进行展望。

关键词: 分布式水文模型, 并行计算, 高性能水文模型, 并行环境, 并行算法

Abstract:

Traditional distributed hydrological models adopt the serial computing mode and the computing power cannot meet the requirements of large-scale refined, multi-element, and multi-process coupling hydrological simulation, thus the support of parallel computing is urgently needed. In the 21st century, the rapid development of computer technology and the gradual improvement of the parallel environment have provided hardware and software support for the parallel computing of distributed hydrological models. This article summarized the existing research from the two aspects of parallel environment and parallel algorithm, analyzed the advantages and disadvantages of different parallel environments and parallel algorithms, and proposed several methods to improve the parallel efficiency of the models, such as rationally allocating the number of processes/threads to reduce communication overhead, adopting a hybrid parallel environment to enhance model scalability, spatial or spatiotemporal discretization to improve model parallelism, and dynamically allocating computing tasks to balance workloads, and so on. Finally, this article examined future research directions of high-performance parallel distributed models.

Key words: distributed hydrological models, parallel computing, high-performance hydrological model, parallel environment, parallel algorithm