【作者】 章大全；

【导师】 丑纪范； 封国林；

【作者基本信息】 兰州大学 ， 气象学， 2011， 博士

【摘要】 本文从干旱的成因和主要影响因子,监测检测技术手段,极端干旱的可预测性以及年代际旱涝趋势转折预测的角度对我国干旱发生发展规律进行研究。首先根据中国气象局1958—2007年194站的温度、降水和Palmer干旱指数(Palmer Drought Severity Index, PDSI)均一化数据库,构建统计模型量化研究温度和降水变化在干旱成因中所占的比重,尤其讨论了中国8个气候区域温度和降水变化在干旱形成中的特征。结果表明：受全球增温趋势的影响,未来中国北方地区的干旱化趋势仍将继续。未来五年内干旱发生概率较大的地区主要分布在中国华北且有持续向南扩张的趋势,一直延伸到江淮地区。在全球变暖的背景下,降水减少仍然是中国东部干旱形成的主要因素。相对于南方地区,中国华北、东北及西北东部等干旱与半干旱区的干旱化进程对温度比降水变化更为敏感。根据我国1960年以来569站的Palmer干旱指数月资料,将全国站点资料格点化后,依据极端干旱发生格点的空间邻近关系确定每次干旱事件的影响范围,并在时间上对每次极端干旱过程进行跟踪监测。分季节对近50年极端干旱发生面积、频次、强度等进行分析,并研究了中国极端干旱的空间影响范围和时间持续性的关系,结果显示极端干旱的持续时间和空间影响范围均呈现幂率分布。基于百分位定义,研究了Lorenz系统极端事件的前兆信号分布特征及可预测性,利用前兆信号及接收器算子特征曲线(Receiver Operator Characteristic curve,ROC)对中国旱涝极端事件可预测性进行研究,得到了极端事件的可预测效果与序列的自相关性,极端事件的极端程度之间的关系。研究表明,无论是Lorenz系统分量序列还是Palmer干旱指数序列,存在极端事件的可预测性随极端程度增强而增加的现象。同时对各区域极端旱涝事件前兆信号及可预测性进行分析,为中国旱涝极端事件的预测提供一定科学依据。分别从概率分布、小波分析以及趋势变化三个角度对中国近50年旱涝转折时空分布特征进行分析,研究了年代际尺度主要海气系统因子对我国旱涝时空格局演变的影响,发现在华南及长江中下游地区,太平洋年代际涛动(Pacific Decadal Oscillation. PDO)和旱涝指数的相关关系存在半年时间尺度的周期振荡。同时PDO与我国北方地区旱涝变化的负相关关系并不局限于华北,而是包括西北东部、东北南部、黄淮在内的广大北方地区。结合已有的我国旱涝不同时空尺度演变规律研究成果,直接针对干旱长期发展趋势中的转折点构建非线性预测模型。在此基础上,利用中国近50年Palmer干旱指数资料,对中国各区域年代际尺度的旱涝时空变化及其可能出现的趋势转折进行预测试验分析。结果显示,未来五年,东北和华北北部的干旱化趋势仍将持续,西北地区东部的干旱将得到缓解。为了克服年代际尺度预测中转折点样本量不足的困难,基于状态空间重构理论和嵌入定理,在转折点预测模型中引入时空序列预测方法。并利用中国旱涝指数资料,对中国区域年代际尺度的旱涝及可能转折进行预测试验分析。回报结果表明,时空序列方法一定程度上改进了转折点预测效果。更多还原

【Abstract】 From the angle of drought’s key impact factors, objective method of detection and predictability of extreme drought events, and its decadal variations we study the spatio-temporal characteristics of drought’s formation and variation in China. Based on homogeneous temperature, precipitation and the Palmer Drought Severity Index (PDSI) data between 1958 and 2007 of 194 stations in China, quantitatively measured the impact of global warming and precipitation variations on the formation of drought in China through a statistical model. Under the background of global warming, the trend of drought in north China still persists. The most probable region of drought in the next five years shows a tendendy of expansion, especially in the direction of south. Decreased precipitation is still the key factors in drought formation for most regions. However, in North, Northeast and eastern Northwest China, global warming plays a bigger role in drought’s formation.During the period of late 1990s to early of the 21st century, the percentage of severe drought (PDSI<-3.0) is unprecedented under the background of global warming and the variation of the distributions of area coverage of droughts in different seasons is uneven. The distribution of spatial extent and durations of severe droughts follows a power-law, e.g. the spatio-temporal characteristics of severe droughts appear to be scale free. In general the area of severe droughts is closely correlated with its duration, while the deviations of the linear fit of correlation is also high.We investigate precursors and predictability of extreme events of Lorenz system with different percentile, the precursors and predictability of PDSI extremes in China are analyzed, and obtained how the quality of the predictions depends on the size of the extreme events and on the correlation strength. Results indicate that for extreme events of Lorenz system and PDSI series, the more extreme, the better predictable. Moreover, we analyzed precursors and predictability of eight regions in China and provide certain scientific basis for flood and drought predictions.Furthermore, we studied the key factors of climate system on the decadal scale and its influence on the variation of characteristics of drought over China. Results indicate that in the middle and lower reaches of Yangtze River and South China, the coirrelation coefficient between Pacific Decadal Oscillation (PDO) and PDSI exhibit periodic variation of half year scale. Meanwhile, the relationship of inverse correlation between PDO and northern China is not only limited to North China, but also include eastern Northwest, southern Northeast and Huang-huai.In order to predict the turning points in long-term trends of droughts over China it is proposed to construct nonlinear dynamics model only from turning points’ magnitudes and intervals. Utilizing PDSI datasets of 569 stations in China in last 50 years, we obtained the possible long-term upward or downward trends and turning points in droughts of the next five years. Results indicate that the aridication of North and Northeast China will be continued while the droughts of eastern Northwest China would be eased.Aiming at overcoming the shortage of sample volume in the prediction of turning points, basic ideas of the spatio-temporal series analysis method are introduced, and the prediction roles of this method are given. Furthermore, we perform hindcast test on the modified predictive model of turning points. The preliminary results show that spatio-temporal series analysis can efficiently improve the ergodicity to enhance the predictive skill of turning points model effectively compared with the single point time series analysis.更多还原