怒江-萨尔温江（NSR）是东南亚最长的自由流动河流，在当地的生态保护和社会发展中发挥着重要作用。但该流域气候多样，地形复杂，观测资料匮乏，制约着该流域尤其是下游的水文研究。揭示该流域的水文特征及其对气候变化的响应，对流域水资源的合理利用和涉水灾害的有效应对有十分重要的意义。论文首先在NSR流域搭建了一个基于山坡产流机制的分布式水文模型GBHM-SW，利用ERA5再分析数据驱动模型重建了历史期的水文数据集（GBHM-ERA5）。基于实测径流的验证结果表明该数据能很好地重现径流量的时空分布特点，并显著优于其他公开可得的径流资料集。使用该数据揭示了流域下游的径流变化特征，发现在1981~2014年间：（1）下游的干湿变化呈现南北分异格局，北部变干，地表产流显著下降；南部变湿，地表产流不显著的上升；（2）下游出口处的径流量变化速率明显大于上游和中游，约以每年102万立方米的速率显著下降；（3）流域径流变化趋势的主导因素呈现季节分异性，在旱季主要由气温变化主导，在雨季主要受降水变化的影响。基于最新的CMIP6预估数据使用贝叶斯模型平均法（BMA）生成了空间分辨率为0.25°的月尺度集合数据集，包含1980~2100年SSP1-RCP2.6、SSP2-RCP4.5和SSP5-RCP8.5三种情景下的气温、短波辐射和降水数据。以ERA5为基准，发现BMA集合数据比简单模式平均法（SMA）生成的集合数据在NSR流域具有更好的适用性。利用BMA集合数据驱动GBHM预估NSR流域未来气候变化下的水文响应，发现未来近期(2019~2048年)该流域降水变化幅度较小，蒸散发增加，径流下降，水资源供给紧张；未来中期(2049~2078年)和远期(2079~2100年)降水增多、径流增大，水资源供给能力提高。空间上，上游未来气温增幅最大，蒸散发加剧，近期和中期径流明显降低；下游的水资源条件相对更好，未来7~8月下游径流量明显增加，缅甸掸邦西南部地表产流增幅最大，相应防汛压力将增大。本研究发现气象要素差异的影响在水量平衡模拟中逐步扩大：采用BMA和SMA方法生成CMIP6集合数据，它们预估的气象要素相对变化基本一致；它们驱动水文模型使用同一套参数模拟蒸散发，相对变化差值约为5%；而对径流，则相对变化的正负和量级均会出现不同，最高差别可达27%。该结果表明在评估气候变化下的NSR流域水文响应时不可忽视不同集合方法造成的影响。
The Nu-Salween River (NSR) is the longest free-flowing river in Southeast Asia and plays an irreplaceable role in local ecological protection and social development. However, the climate and topography of this basin are complicated and it is in lacks of observed and simulated data, which restricts hydrologic research in the basin, especially in the lower part. Analyzing the hydrological characteristics of the basin and revealing its response to climate change is of great importance for the rational use of water resources and effective response to water-related disasters.This study first developed a distributed hydrological model in the NSR basin named GBHM-SW based on the hillslope flow production mechanism and reconstructed the hydrological dataset (GBHM-ERA5) for the historical period using this model driven by ERA5 reanalysis data. The validation results against hydrological stations demonstrate that this reconstructed data can well reproduce the spatial and temporal distribution characteristics of streamflow and significantly outperforms other publicly available streamflow datasets. Based on this reconstructed data, the streamflow characteristics of the lower reaches of the basin were revealed, and it was found that during 1981~2014: (1) the surface runoff changes were different between north and south of the downstream, with the northern part showing a downward trend, while the south part upward; (2) The rate of decline in discharge at the outlet of the downstream was significantly higher than at the upstream and midstream, the discharge at the outlet of the downstream was decreased at a rate of about 1.02 million m3 per year; and (3) the dominant factors of surface runoff linear trends showed seasonal divergence, surface runoff trends in the dry season were mainly dominated by temperature changes while surface runoff trends in the rainy season were controlled by precipitation changes.A monthly-scale ensemble dataset with a spatial resolution of 0.25° was generated based on the latest CMIP6 data using Bayesian model averaging (BMA) mothed, containing temperature, shortwave radiation, and precipitation data from 1980 to 2100. The data for the future period includes three scenarios: SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5. The BMA has better applicability in the NSR basin compared with the simple model average (SMA) data.GBHM driven by BMA was used to predict the hydrological processes in the NSR basin under three climate change scenarios, it was found that the water resources situation of the basin will be severe in the near future (2019~2048), with small changes in precipitation, increased evapotranspiration, and decreased surface runoff. In the middle future (2049~2078) and the far future (2079~2100), the basin will experience increased precipitation, increased surface runoff, and increased water supply capacity. The water resources condition in the lower reaches will be relatively better, and the discharge of the downstream will increase significantly from July to August. The area of the downstream basin with the greatest increase in surface runoff is located in the southwestern Shan State of Myanmar, where cropland is widespread, and flood control pressure will increase. Unlike the downstream, the upstream will experience the largest increase in temperature, with increased evapotranspiration, and decreased surface runoff in the near future and middle future. The hydrological response characteristics in the northern part of the midstream are similar to those in the upstream, and the features of the southern part of the midstream are similar to those in the downstream.This study found that the effects of different meteorological data gradually expand in water balance simulations: the BMA and SMA methods are used to generate CMIP6 ensemble data, which predict essentially the same relative changes in meteorological elements; they drive the hydrological model to simulate evapotranspiration using the same set of parameters with a relative change difference of about 5%; while for streamflow, the signs and magnitudes of relative changes predicted by BMA and SMA are different, with the largest difference reaching 27%. The results suggest that the impacts of different ensemble methods cannot be ignored in the study of assessing the hydrological response of the NSR basin under climate change.
