认识气候和流域下垫面等变化条件下的流域水文响应特征，有效预测和预报流域水资源和洪水的变化，是“维持健康长江、实现人水和谐”的科学基础。传统的基于经验和河道流量观测资料建立起来的水文模型，难以适应变化条件下的降雨径流模拟。本论文采用基于物理机制的分布式水文模型，在模拟流域水文循环全过程的基础上，探讨了分布式水文模型在水资源评价、干旱评估和洪水预报中的应用。 论文首先分析了20世纪下半叶长江流域气温、降水和径流的变化趋势。结果发现：流域年均气温升高趋势显著且年际波动幅度增大；降水时空分布有改变的趋势，长江中下游夏季降水增加，而四川盆地秋季降水减少。 在流域面积约为100万km2的长江上游，利用DEM及相关遥感和地理信息来表现流域下垫面的空间差异性，构建了基于流域地形地貌特征的大尺度分布式水文模型。模型采用次网格参数化方法来处理大尺度网格内下垫面的非均一性，采用具有物理意义的水流运动方程来描述产汇流过程。通过模拟1961~2000年的天然径流过程，论文从径流、蒸散发和土壤含水量等分析了长江上游水资源的演变趋势。根据模拟结果，论文还分析了长江上游干旱的发展趋势，并构建了GBHM-PDSI干旱评估指标。结果显示，该指标综合反映水分亏缺和持续时间对干旱程度的影响，能够表现旱情在流域空间上的分布和发展变化。 在洪水预报应用方面，本论文着重探讨了提高三峡区间洪水预报能力的可能性。在三峡区间构建了尺度相对较小的分布式洪水预报模型，提出了利用自动遥测雨量站观测数据进行雷达测雨校正的方法。结果显示，分布式洪水预报模型与雷达测雨数据结合，可以减少洪水预报中的不确定性，提高预报精度。 面对三峡水库建成后水资源管理和防洪实时调度的要求，以上述两种尺度的水文模型为基础，结合水动力学模型，在长江上游初步构建了空间嵌套式水文模型，实现了上游来水预报、区间洪水预报和三峡库区洪水演进的整合。以空间嵌套式流域水文模型为核心，提出了构建流域数字水文模拟系统的设想，以适应大型流域水资源管理和防洪决策的多任务、多目标和多尺度要求。

The Yangzte River basin (also called the Changjiang River basin) is the largest river basin in China, which has complicated geomorphological features and diverse climatic patterns, and frequent natrual disaters caused by flood and drought. Understanding the hydrological response to climate change and human activities, assessment of the changes in water resources and the flood forecast are the key research topics for integrated watershed management with new anthropo-concept of health and harmonious river. Building on the geomorphology-based hydrological model -- GBHM, a large-scale distributed hydrological model has been established in the upper Yangtze River. Applications of the model for water resources assessment, drought severity evaluation and flood forecast have been addressed in this research.Using meteorological data from 154 weather stations together with river discharge from 26 hydrological gauges, it analyzed the spatio-temporal variation of temperature, precipitation and runoff during the last half century. The TFPW-MK statistical test has been applied to detect the significance of trends. Both runoff and precipitation showed a significant increasing trend in summer along the middle and lower reaches in the 1990s. However, a decreasing trend for runoff and precipitation was found in spring and autumn, particularly in the upper reaches.Then, a large scale distributed hydrological model has been established in the upper Yangtze River with area of 1 million km2. In this model, the study basin is divided into a discrete grid system of 10-km size, and each grid is represented by a number of geometrically-symmetrical hillslopes. The basin is divided into a number of sub-basins with the proper size and the river network is identified up to the main stream of the minimum sub-basin. The hydrological components, including runoff generation from the hillslopes and the flow routing in the river network are modelled using physically-based approaches. Taking the meteorological inputs, the model simulated river discharge during 1961~2000 at different locations in the river network, and temporal changes and spatial distributions of soil moisture and evapotranspiration, which provide an inside investigation into water resources in the study basin. Results showed that the ratio of seasonal runoff to annual one has a significant increasing trend in summer in the eastern Sichuan basin and the Three Gorges region in the 1990s, but a decreasing trend in autumn. This implies an increasing flood risk in summer and water shortage in autumn. Based on the hydrological simulation, a new drought index, GBHM-PDSI, was proposed. It was found that the new drought index has obvious advantages on describing the temporal change of drought severity and the spatial variation of dryness.In order to reduce the uncertainties of real time flood forecast in the Three Gorges region, the radar rainfall data has been used together with the distributed hydrological model. Results showed that in addition to the rain gauge network the weather radar can provide better spatial distribution of rainfall. By means of physically-based distributed hydrological model combining with radar rainfall data, it can capture adequately the spatial variation of rainstorm, and provide better flood forecast at real time.For meeting with the requirements of integrated operation of the Three Gorges reservoir, a new space-nested hydrological model with different spatial scales (present used 10km and 1km grid sizes) has been built. Inflow simulation from the upper streams of the reservior, inflow simulation from the inter-basin of the Three Gorges and the flood routing in the reservoir were coupled into one model system. The preliminary results showed that this model could sucessfuly simulate the dynamic change in reservoir water level and provide a useful tool for the reservoir operation. Using this model as the core, a Digital Hydrology Simulation System for the Yangtze River basin has been proposed as the tool for the integrated river management in future.