流域骤发性干旱形成机制与定量评估方法研究

Flash drought is an extreme climate phenomenon frequently occurs in different regions and most often in growing seasons. Different from the traditional creeping property of drought, flash drought has an extraordinarily rapid onset. This leads to existing drought evaluation methods with slow response and low sensitivity to the fast variation of moisture anomaly seem incompetent to capture the real status of flash drought. The objective of this program therefore, is to develop a new quantitative assessment method based on the formation mechanism of flash drought. Specifically, the abnormal conditions of meteorological factors during the initial drought stage will be detected by using routine hydro-meteorological observations, hydrological model simulations and remote sensing derived drought information. The relationships between flash drought trends and occurrence frequency, versus atmospheric circulation factors will be analyzed to explore the driving mechanism of climatic forces on flash drought. To search for the suitable variables for moisture description and optimal time scales for flash drought monitoring, the statistical relationships between key hydrological variables versus flash drought duration and variation rate of severity during drought development stage will be investigated. Then a functional relationship will be established to describe the non-linear structure among explanatory variables, and a standardized algorithm with a short time scale (weekly or a period of ten days) effect will be proposed. On this basis, a physically based drought assessment method which considers the formation mechanism and rapid-changing characteristics of flash drought will be developed. The expected progresses are of great significance for a deep understanding of drought evolution mechanism, which are promising to provide some scientific technical supports for drought monitoring and prevention.

骤发性干旱(简称骤旱)是近年来在全球不同地区尤其是作物生长季频繁发生的极端气候现象。区别于传统干旱缓慢、蠕变的特点，骤旱爆发异常迅速，现有干旱评估方法在识别旱情时容易出现反应迟缓、灵敏度低等缺陷，难以反映骤旱的真实状态。针对上述问题，本项目拟采用站点观测、水文模型模拟及遥感反演相结合的手段，诊断旱情形成初期区域气象因子的异常状态，分析骤旱演变趋势、发生频次与大气环流因子的成因联系，揭示触发骤旱的气候驱动机制；建立骤旱发展阶段关键水文要素变化特征与旱情持续时间、强度增减速率的统计关系，甄别能够有效捕捉骤旱演进过程的敏感解释变量和最佳时间尺度；在此基础上，建立描述解释变量非线性关联结构的函数关系，提出具有周、旬短时间尺度效应的标准化算法，进而研制能够考虑骤旱物理成因和快速变化特性的干旱评估方法。本项目对于深入认识干旱形成发展机理具有重要理论意义，可为科学开展流域旱灾监测与防治提供技术支撑。

干旱作为典型自然灾害，具有自然属性、灾害属性和社会属性多面特征。全球气候变化加速了水循环进程，改变了陆地水文情势的时空格局，同时也潜移默化地改变了干旱的内在属性，使其表现形式从缓慢、蠕变的特点衍生出历时短、强度高、速度快的异常现象，即为骤发干旱（简称骤旱）。骤旱的出现不仅打破了传统理解，同时在定义、机理上也提出了更高的需求。本项目了深入剖析干旱事件的形成过程及驱动机制，识别了干旱事件的形成发展过程，探明了流域骤发干旱形成初期气象异常条件的驱动机制，解译了旱情演进过程中蒸散发、土壤水的变化特征及互馈影响作用，构建了基于人工智能技术的骤旱定量评估方法。取得的研究成果主要体现在以下3个方面：①系统分析淮河流域和黄河流域干旱的历时、烈度、强度多维度特征，采用线性回归、Copula函数（非线性）开展干旱多维度特征频率分析，深入剖析干旱事件的形成过程及驱动机制，确定流域监测骤发干旱的最佳时间尺度；②以中国大陆为研究区，识别了骤发性干旱的形成发展过程，探明了流域骤旱形成初期气象异常条件的驱动机制，解译了旱情演进过程中蒸散发、土壤水的变化特征及互馈影响作用，发现了骤旱比传统缓慢干旱具有更强的气象驱动力，变化幅度约超过缓慢干旱0.5个标准差；③构建了综合表征流域水热通量动态变化的水分亏缺参数方案，提出了有效考虑土壤水分快速衰减（变化）特性的周时间尺度标准化过程，基于机器学习、深度学习等人工智能方法，研发了骤发干旱定量识别模拟方法。项目执行期间在国内外学术刊物发表论文21篇(其中SCI论文16篇)，申请发明专利4项（授权1项），软件著作权2项，10人获得博士硕士学位，获得教育部自然科学二等奖1项。