地理科学进展

地理科学进展 ›› 2013, Vol. 32 ›› Issue (4): 538-547.doi: 10.11820/dlkxjz.2013.04.006

• 水文过程与环境变化 • 上一篇    下一篇

分布式水文模型的并行计算研究进展

刘军志1,2, 朱阿兴1,3, 秦承志1, 陈腊娇4, 吴辉1,2, 江净超1,2   

  1. 1. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室,北京100101;
    2. 中国科学院大学,北京100049;
    3. 威斯康星大学麦迪逊分校地理系,麦迪逊53706,美国;
    4. 中国科学院遥感与数字地球研究所,北京100094
  • 收稿日期:2012-12-01 修回日期:2013-03-01 出版日期:2013-04-25 发布日期:2013-04-25
  • 通讯作者: 朱阿兴(1962-),男,浙江长兴人,博士,研究员,主要从事GIS/RS及流域系统科学研究。E-mail:axing@lreis.ac.cn
  • 作者简介:刘军志(1984-),男,山东海阳人,博士研究生,主要研究方向为高性能地学计算。E-mail:liujz@lreis.ac.cn
  • 基金资助:
    国家863 计划项目(2011AA120305);国家自然科学基金项目(41023010);科技部国际科技合作项目(2010DFB24140)。

Review on parallel computing of distributed hydrological models

LIU Junzhi1,2, ZHU A-Xing1,3, QIN Chengzhi1, CHEN Lajiao4, WU Hui1,2, JIANG Jingchao1,2   

  1. 1. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    2. University of Chinese Academy of Sciences, Beijing 100049, China;
    3. Department of Geography, University ofWisconsin-Madison, Madison WI 53706, USA;
    4. Institute of Remote Sensing and Digital Earth, CAS, Beijing 100094, China
  • Received:2012-12-01 Revised:2013-03-01 Online:2013-04-25 Published:2013-04-25

PDF (PC)

赞

被引次数

61 | 13

摘要： 大流域、高分辨率、多过程耦合的分布式水文模拟计算量巨大,传统串行计算技术不能满足其对计算能力的需求,因此需要借助于并行计算的支持。本文首先从空间、时间和子过程三个角度对分布式水文模型的可并行性进行了分析,指出空间分解的方式是分布式水文模型并行计算的首选方式,并从空间分解的角度对水文子过程计算方法和分布式水文模型进行了分类。然后对分布式水文模型的并行计算研究现状进行了总结。其中,在空间分解方式的并行计算方面,现有研究大多以子流域作为并行计算的基本调度单元;在时间角度的并行计算方面,有学者对时空域双重离散的并行计算方法进行了初步研究。最后,从并行算法设计、流域系统综合模拟的并行计算框架和支持并行计算的高性能数据读写方法3个方面讨论了当前存在的关键问题和未来的发展方向。

关键词: 并行计算, 分布式水文模型, 空间分解, 研究进展

Abstract: High resolution distributed hydrological simulations over large watersheds require very large amounts of computations, which cannot be provided by sequential computation techniques on which existing hydrological models were developed. So parallel computing of distributed hydrological models is needed. In this paper, we first analyzed the parallelizability of distributed hydrological models from three angles (spatial, temporal and sub-process) and pointed out that spatial domain decomposition is the preferred approach to parallel computing of distributed hydrological models. According to spatial relationships among simulation units, distributed hydrological models, as well as simulation methods for hydrological processes, are classified into different types. Then, current studies on parallel computing of distributed hydrological models were introduced. For most current studies on parallel computing using spatial domain decomposition methods, sub-basin was adopted as the basic scheduling unit for parallel computing. The temporal-spatial discretization method proved the feasibility of parallel computing utilizing parallelization from the temporal angle. Last, the key technologies and future research directions were discussed in the following aspects: 1) parallel algorithms; 2) parallel computing framework for integrated watershed system simulations; 3) high performance input/output for parallel computing of distributed hydrological models.

Key words: distributed hydrological model, parallel computing, review, spatial domain decomposition