【作者】 成静清；

【导师】 宋松柏；

【作者基本信息】 西北农林科技大学 ， 水文学及水资源， 2010， 硕士

【摘要】 由于受到气候变化和人类活动的频繁影响,流域(或区域)内的下垫面条件发生了较大变化,破坏了天然水文序列的一致性,使其不能直接采用传统的水文频率计算方法进行计算(谢平等2005a)。本文以陕北及关中地区的非一致性年径流资料为例,采用三种不同的方法进行了频率计算研究,以期为该地区水利工程的规划、设计、施工和运行管理提供水文依据。主要研究内容和结论如下:(1)水文变异综合诊断。采用水文变异综合诊断方法对研究区15个水文测站的年径流序列进行诊断。初步诊断包括过程线法、累积曲线斜率差异幅度分析法和Hurst系数法。详细诊断包括趋势详细诊断和跳跃详细诊断:对于趋势变异,采用线性趋势相关系数检验法、斯皮尔曼(Spearman)秩次相关检验法和肯德尔(Kendall)秩次相关检验法进行判断;对于跳跃变异,采用Lee-Heghinian法、有序聚类法、最优信息二分割法、R/S法、Brown-Forsythe法、滑动F法,滑动T法、滑动秩和法、滑动游程法和Mann-Kendall法进行判断。综合诊断包括趋势综合和跳跃综合,并根据效率系数确定序列变异形式。结合实际水文调查分析,对变异形式和结论进行确认,进而得到最可能的变异诊断结果。经诊断,研究区15个水文测站中,有7个测站的的年径流序列发生变异,其中属于趋势变异的测站为赵石窑站;属于跳跃变异的测站和变异年份分别为:神木站1979年、枣园站1996年、黄陵站1985年、林家村站1985年、张家山站1985年和状头站1985年。(2)基于随机方法的非一致性年径流频率计算。根据水文变异综合诊断的结果,对赵石窑站采用基于趋势分析的非一致性年径流序列频率计算方法进行计算;对神木站、枣园站、黄陵站、林家村站、张家山站和状头站采用基于跳跃分析的非一致性年径流序列频率计算方法进行计算,得到了过去、现在和未来条件下满足P-Ⅲ型分布的统计参数。通过对7个变异测站的统计参数以及频率曲线的分析可知, 7个测站的年径流量在未来条件下存在不同程度的减少,也就是说,其年径流序列的总体变化趋势是逐年减小的。(3)基于条件概率的非一致性年径流序列频率计算。对基于条件概率的非一致性洪水频率计算方法进行转换和改进,得到研究区基于条件概率的非一致性年径流频率计算方法:首先,根据水文变异综合诊断结果,对于非一致性年径流序列,按其变异点的位置将序列进行划分。其次,对概率权重和条件概率进行估计。本文对条件概率的估计假设了两种情况,即条件概率服从对数正态分布或服从P-Ⅲ分布。最后,采用ABS、OLS和WLS准则对频率计算结果进行拟合优度检验。从研究区7个变异测站的频率曲线拟合优度情况来看,基于条件概率的非一致性年径流频率计算方法能满足参数估计精度的要求。(4)基于混合分布的非一致性年径流序列频率计算。对基于混合分布的非一致性洪水频率计算方法进行转换和改进,得到研究区基于混合分布的非一致性年径流频率计算方法,并将其应用到研究区6个跳跃变异测站的年径流资料上。首先,假设非一致性年径流水文序列的变异点前后分别服从于两个分布,全序列服从于由这两个分布组成的混合分布,本文采用的两对数正态混合分布和两P-Ⅲ混合分布。其次,采用模拟退火算法进行参数估计。最后,采用ABS、OLS和WLS准则对频率计算结果进行拟合优度检验。基于混合分布的非一致性年径流频率分析计算过程,包括统计分析和成因分析,使其较之传统的频率计算方法更具合理性;其次,混合分布的子分布可以根据拟合优度的要求来选取,这奠定了其方法上的灵活性;再次,采用模拟退火算法估计混合分布参数,可以满足各种分布参数求解的要求,不受参数个数的限制;最后,从拟合优度来看,混合分布模型优于条件概率模型。更多还原

【Abstract】 Due to the frequent human activities and the climate change, partial climate and underlying surface conditions were changed considerablly. All this change destroyed the consistence of natural hydrological data, which made it cannot use traditional frequency analysis methods. This research used three methods to calculate the frequency of inconsistent annual runoff serise in Shanbei and Guanzhong region, in hopes to provide hydrological design values for these regions. The main contents and the conclusions are as follows:(1) Comprehensive hydrology change point analysis. Adopt the comprehensive hydrology change point analysis method to determine the change-point of inconsistent annual runoff in the regions. Preliminary diagnosis includes the process line method, the differences of cumulative slope method and Hurst coefficient method. Detailed diagnosis includes trend detailed diagnosis and jump point for diagnosis of detailed. Three methods for trend detailed diagnosis were enployed: correlation coefficient method, Spearman correlation test and Kendall rank correlation test. Using 11 methods for jump detailed diagnosis: Lee-Heghinian method, orderly clustering method, two segmentation method of optimal information, R/S method, Brown-Forsythe method, sliding F method, sliding T method, sliding rank method, sliding run method and Mann-Kendall method. Comprehensive diagnosis includes trend integrated diagnosis and jump integrated diagnosis, and the change-point can be determined according to efficiency coefficients and practical survey analysis. Through analysis, there are 7 stations’annual runoff series hydrological changed: Zhaoshiyao belong to trend type, and the rest of the annual runoff series belong to jump type, and the change point is Shenmu 1979, Zaoyuan 1996, Huangling 1985, Linjiacun 1985, Zhangjiashan 1985 and Zhuangtou 1985.(2) For this 7 variation stations, using the inconsistent annual runoff frequency calculation method which based on random analysis, we can gain the statistical parameters of P-Ⅲdistribution in the past, present and future. And the results indicated that the annual runoff series reduced year by year.(3) Frequency calculation of inconsistent annual runoff series based on conditional probability. Firstly, according to change-point, classified inconsistent annual runoff series as two parts. Secondly, eatimate the probability weight and conditional probability(in this paper probability include LN(2)and P-Ⅲ). Finally, by use ABS, OLS and WLS standard test the goodness-of-fit, we can find the frequency curve fitting well. (4) Frequency calculation of inconsistent annual runoff series based on mixed distribution. Firstly, according to change-point, clasiified inconsistent annual runoff series as two parts. Assumpt this two series follow differently distribute, and the whole series follows mixed distribution(mixed-two lognormal distribution and mixed-two P-Ⅲdistribution). Secondly, use the simulated annealing algorithm estimate the parameters of mixed distribution. Finally, ABS, OLS and WLS standard were tested the goodness-of-fit, we can find the frequency curve fitting well.The mixed distribution contains statistic analysisy and casue-effect analysis, is more rational and flexibal than traditonal methods. Acorrding to calculation results, the mixed distributiom model is better than to conditional probability model.更多还原