【作者】 林盛吉；

【导师】 许月萍；

【作者基本信息】 浙江大学 ， 水文学及水资源， 2011， 硕士

【摘要】 随着人口增长、农业和工业用水需求量的增大以及全球环境的变化,干旱化趋势在进一步发展中。干旱已成为影响我国农业生产最为严重的自然灾害。干旱指数是干旱监测和预测的基础,近年来国内外学者已经研究了不少干旱评价的指标和方法,但尚未有相关工作研究气候变化下钱塘江流域的未来干旱情况,随着全球变暖,湿润地区干旱的预测和评估也同样引起了众人的关注。本文首先利用DEM提取钱塘江流域水流路径、数字水系和流域边界等特征,以此基础构建数字流域,用于后期流域水文过程的模拟和干旱时空分析等。论文中使用钱塘江流域23个气象站1951～2008年的月降水资料,利用Z指数和SPI通过求算给定时间尺度的累积概率,使其能够在多个时间尺度上进行计算,分析了钱塘江流域各站各季的干旱灾害情况、空间分布情况以及变化趋势情况,并着重分析了历史重大的干旱灾害的影响。文中利用支持向量机等方法进行气候变化下干旱的预测。本文利用1961～2000年NCEP再分析资料与钱塘江流域23个气象台站的历史降水观测资料,并利用主成分分析(PCA)与支持向量机(SVM)相结合的统计降尺度方法,建立大尺度气候预报因子与钱塘江流域各气象站点各月降水的统计降尺度模型；将主成分分析和SVM相结合的统计降尺度模型应用于三种全球气候模式HadCM3、Ccsm3、Echam5分别在A1B、A2、B1排放情景下的预报因子,拟合钱塘江流域23个站点的当前降水变化情景以及预测未来30年(2011～2040年)的降雨情况。干旱历时和干旱烈度是定量研究水文干旱事件的两个重要指标。利用解析的方法来推求干旱烈度概率分布或一定干旱历时对应的干旱烈度的条件概率分布还存在困难,目前最常用的方法是利用已有的理论分布来拟合干旱烈度的概率分布或条件概率分布,其拟合的好坏关键取决于对干旱烈度概率分布参数的估计。在假定干旱烈度服从GEV分布的基础上,利用线性距法对干旱烈度概率分布的参数做了较为精确的估计,同时计算了不同重现期的干旱烈度。重点分析了一定时期内极限干旱烈度概率分布以及钱塘江流域极限干旱烈度的期望值。结论表明：Z指数和SPI在计算结果上取得了很好的一致性,并较为准确的标识出了钱塘江流域历年的干旱情况,多时间尺度SPI更符合钱塘江流域的实际情况。由于九种气候变化预测结果各有优劣,特定的模式和情景评价干旱效果不是完全一致。总体而言HadCM3、Echam5这两种气候模式的趋势模拟效果优于Ccsm3模式。本研究成果对进行干旱预测和分析具有重要的借签意义,并且为湿润地区的干旱研究和监测提供了可靠的科学支持。更多还原

【Abstract】 With the increase of population and the agricultural and industrial demand for water and the change of global environment, the drought disaster is becoming so serious that it has caused the greatest losses of our agricultural production. The drought index is the basis of drought monitoring and prediction. In the recent years, many experts have made a lot of relative research on different indices and methods of drought evaluation. However, the research about prediction on drought degree of Qiantang River based on climate change prediction is scarce, and the prediction and assessment of drought disaster in humid areas is the front issue of hydrological research.In this thesis the DEM technology is used for extraction of features such as flow paths, the digital stream and watershed boundaries of Qiantang River, based on which digital watershed is built, the hydrological process is simulated and the analysis of drought is made.Monthly precipitation data during 1951 to 2008 from 23 weather stations in the Qiantang River Basin are collected for the research. The indices used include the Z index and the SPI index, by which the cumulative probability at a certain time scale is computed for further calculation at multiple time scales. The seasonal and dimensional drought conditions and the trend of the drought situation of different stations in the Qiantang River basin as well as the impact of major drought disasters in the history are analyzed.In this paper, the support vector machine is used for drought prediction under climate change. The NCEP reanalysis data and the precipitation data from 23 meteorological stations in Qiantang River basin during 1961 to 2000 are used. A statistical downscaling method of combining the principal component analysis (PCA) and support vector machine (SVM) is used for the establishment of statistical downscaling model between large-scale meteorological forecasting factors and the monthly precipitation of the meteorological stations in Qiantang River basin. This method combines the PCA and the SVM to calculate the forecasting factors of different global climate model HadCM3, Ccsm3, Echam5 respectively under A1B, A2, B1 emission scenarios and analyze the current and future rainfall condition in the next 30 years(2011-2040) of the Qiantang River basin. The Mann-Kendall nonparametric test method is also used for analyzing the long-term change trend in precipitation and the abrupt change times of drought.Drought duration, drought interval and drought severity are three important indicators in quantitative study of hydrologic drought. But, in present, there is some difficulties in determining drought probability distribution or drought severity through employing analytical method. At present, the most often used method is using assumed theoretical distribution to fit drought severity, and the quality of this method depends on parameter estimation of drought severity probability distribution. On the basis of the premise that drought severity subjects to GEV distribution, we calculate drought severity in different return periods using L-moment approach.The results show that although the value of Z-index and SPI consist with each other and account for the drought conditions of Qiantang River basin over the years reasonably, SPI with multiple time scales is more suitable for the actual situation of Qiantang River Basin. Nine different predicting results of drought conditions in Qiantang River Basin for next 30 years are obtained by applying the global climate models of HadCM3、Echam5 and Ccsm3 on the emission scenarios of A1B、A2 and B1. All the nine results have advantages and disadvantage, and as for which one to use, it depends on particular case. In general, trend modeling results of HadCM3 and Echam5 are better than Ccsm3, but when applied in A1B to model extreme drought, Hadcm3 has better effect than the other two.This study is a significant reference for further drought prediction and analysis, and provides scientific support for drought research and monitoring in humid region.更多还原