长江源(本研究指直门达水文站以上流域)地处青藏高原腹地，属于典型的高山冰冻圈区域，也是西南河流源区的代表性流域，对气候变化极为敏感。随着气候升温，长江源等西南河流源区积雪和冰川大幅减少，严重影响径流并威胁到区域水安全。认识长江源冰雪融水径流的形成和演变规律，对长江源等西南河流源区水资源利用和生态安全意义重大。但该区为典型的缺资料地区，地面观测资料匮乏，融水径流及总径流的监测、模拟和检验难度极大。为此，本研究结合多源遥感信息、数据同化及模型方法，提出了适用于长江源等缺资料江河源区积雪、冰川及径流的模拟和预估方法。研究取得了以下四个方面的成果：(1)提出了一种评估缺资料区积雪及融雪径流模拟的方法(积雪模型率定一致性评估)，并利用分布式水文模型CREST-Snow，对长江源2003?2014年的径流进行了模拟，该方法可量化全球其他类似区域积雪融水对总径流贡献；(2)采用参数和状态联合估计的粒子滤波算法，同化了高空间分辨率雪深遥感反演数据，改善了长江源积雪(降雪、雪水当量)模拟；(3)对比了以冰川整体物质平衡观测和冰川测点物质平衡观测作为率定参照的优缺点，并将基于?h参数法的冰川响应模块耦合至温度指数融雪模型中，实现了冰川后退/前进时几何形态(面积、体积)变化的模拟；(4) 将耦合?h参数法的冰川模型和径流模型CREST-Snow相集成，根据CMIP6五种气候模式输出的气象数据，预估了长江源2019?2100年总径流及冰雪融水径流，量化了气候变化对径流的影响。结果表明：长江源径流以降雨产流为主，但季节性融雪径流是长江源径流的重要来源，特别是春季(约占33%)；相较于传统率定法，粒子滤波算法可明显提高积雪空间分布的模拟效果(RMSE降低约15–30%)；在未来气候情景下，长江源冰川将持续强烈消融，在升温最高的SSP585情景下，长江源冰川在2008?2100年间冰川面积和体积将分别减少81%和91%，相比2008?2018年间多年均值，总径流、融雪径流和融冰径流在2090?2100年间多年均值将分别增加31%、减小44.4%和增加20.4%。本研究实现了缺资料高山区冰川形态、总径流及径流成分变化的可靠模拟和预估，为进一步认识气候变化对高山冰冻圈水文过程的影响提供了重要参考，为回答在气候变化影响下长江源等西南河流源区冰川积雪及其融水径流如何演变这一重要科学问题提供了有效方法。
The headwater region of the Yangtze River (HRYR), defined as the drainage area above the Zhimenda gauging station, is located in the hinterland of the Tibetan Plateau. The HRYR is a typical alpine cryospheric region; meanwhile, it is a typical river basin of Southwest China's headwater regions which are quite sensitive to climate change. Snowpack and glaciers over the HRYR and other headwaters of Southwest China are continuously decreasing under global warming, which inevitably lead to runoff changes and threaten regional water security. For the HRYR and headwaters across Southwest China, an improved understanding of meltwater runoff generation and evolution is of great value to the utilization of water resources and ecological integrity. However, these regions are poorly gauged, where in situ observations are extremely scarce. Monitoring, simulation, and validation of runoff and its components are therefore challenging. This study integrated multisource remote sensing information, data assimilation and modelling, and developed a suitable approach for the HRYR and other headwater regions of Southwest China, which can simulate and predict changes in snow, glacier, and runoff.This dissertation research was carried out with the following four contents: (1) developing a method for evaluating the simulation of snow and snowmelt runoff over poorly gauged basins (i.e., Evaluation based on Hydrological Consistency, EHC), and simulating HRYR's total runoff from 2003 to 2014 based on the distributed hydrological model CREST-Snow. EHC can be used as an effective method to quantify proportional contributions of snow meltwater to total runoff in other similar regions globally; (2) developing a dual state-parameter updating scheme based on the particle filter. High-spatial-resolution remotely sensed snow depths were assimilated into a snow melting model to improve snow simulation (particularly snowfall and snow water equivalent) over the HRYR. Snow distribution and evolution can be reliably simulated with the particle filter approach; (3) comparing advantages and limitations of using either glacier-wide mass balance observations or point-based glacier mass balance measurements as calibration references. By coupling an empirical glacier dynamic model (Δh-parameterization) with a temperature-index glacier melting model, glacier geometry change can be simulated when glacier is retreating/advancing; (4) projecting snow and glacier meltwater runoff and total runoff over the HRYR during 2019?2100 under three climate change scenarios by coupling the Δh-parameterization glacier melting model with CREST-Snow. Results show that streamflow of the HRYR is dominated by rainfall, but seasonal snowmelt runoff is an important component, especially during spring season (accounting for about 33% of the total runoff). Compared with traditional calibration methods, the particle filter approach obviously improves the spatial distribution of snow simulation (with RMSE decreased by about 15–30%). Under future climate scenarios, glaciers over the HRYR will continuously retreat. Under the SSP585 scenario with the highest air temperature increase, the area and volume of HRYR's glaciers will lose about 81% and 91% from 2008 to 2100. Compared with mean annual values during 2008?2018, mean annual total runoff, snowmelt runoff, and glacier melt runoff during 2090?2100 will increase by 31%, decrease by 44.4% and increase by 20.4%, respectively. This study can be used to reliably simulate and project changes in glacier geometry, total runoff, and runoff components over poorly gauged headwater regions. It can also serve as a basis for better understanding climate change impact on hydrologic processes over the alpine cryosphere. This study provides an effective approach of answering the scientific question as to how snow, glacier, and runoff evolve with time in the HRYR under climate change, which can also be applied to other poorly gauged headwater regions across Southwest China.
