摘要
针对分布式水文模型的率定过程海量计算难题,本研究提出了基于Hadoop和Redis集群的泛化似然不确定估计(GLUE)率定算法——HR-GLUE。该方法通过Redis缓存模型输入,利用MapReduce算法实现的GLUE率定方法并行计算。研究以典型分布式水文模型——SWAT(Soil and Water Assessment Tool)的并行率定为例对该方法的计算效率和效果进行了验证。结果表明HR-GLUE可以显著堤高模型的率定速度,在14个作业节点的Hadoop集群满负荷工作时,可将模型的速度提高28.9倍,且利用其速度优势,可获得更优的率定效果。
In order to solve the prohibitively large amount of computational demands during a calibration process of a distributed hydrological model,this study proposed a parallel generic likelihood uncertainty estimation method(namely,the HR-GLUE)based on clusters of Hadoop and Redis.This method is first cache model inputs in a Redis cluster and implement GLUE method in parallel with the MapReduce algorithm.The performance of the proposed HR-GLUE was estimated with widely employed hydrological model,the Soil and Water Assessment Tool(SWAT).Results show that HRGLUE can greatly speed up the calibration process.With 14 task nodes at full capacity,the HRGLUE can achieve 28.9 times speedup and gain more accurate modeling results.
作者
陈平
张德健
何原荣
CHEN Ping;ZHANG De-jian;HE Yuan-rong(a.College of Computer and Information Engineering,Xiamen University of Technology,Xiamen 361024,China;Big Data Institute of Digital Natural Disaster Monitoring in Fujian,Xiamen University of Technology,Xiamen 361024,China)
出处
《南宁师范大学学报（自然科学版）》
2019年第4期50-56,共7页
Journal of Nanning Normal University：Natural Science Edition
基金
福建省自然科学基金面上项目(2018J01481)
厦门市科技计划产学研项目(3502Z20183056)
厦门理工学院引进人才科研启动项目(YKJ16017R)
