1、利用贝叶斯模型平均（Bayesian model averaging method, BMA）技术，利用耦合模式比较计划第五阶段（Coupled Model Intercomparison Project Phase 5, CMIP5）中的21个大气环流模式（General Circulation Model, GCM）分析了21世纪中亚山区（Mountainous regions of Central Asia, MCA）的气候变化预测结果。研究结果表明：1）BMA优于简单的GCM结果集合，BMA得到的三个气候变量（即温度、降水、降雪）与观测值一致。2）到21世纪末（2070-2099），在RCP8.5模式下，气温将升高5.0 °C，降水将从1976 ~ 2005年的186 mm增加到197 mm，增加量为5.9%，而降雪将下降26.4%，从72 mm下降到53 mm。3）降水在北天山增加，在阿姆河地区减少；降雪在天山西部地区显著减少；平均降雪占降水的比例（S/P）从当前控制期（1976 ~ 2005）的0.58下降到21世纪末（2070 ~ 2099）的0.43（RCP8.5）。4）在春秋季节降雪对温度的敏感性较高，在冬季较低；高山边缘区降雪对温度最为敏感。研究区气候变化整体上表现为温度升高、降水增加和降雪量减少，这将导致潜在的洪水风险、固体水资源的流失和径流季节分配变化。
2、通过结合Morris和SDP（State-Dependent Parameter method）敏感性分析方法，应用SWAT（Soil and Water Assessment Tool）分布式水文模型分析了开都河流域水文过程，并量化了气象输入对水文模拟的贡献。结果如下：1）由于在高寒山区观测气象数据稀少，气象输入占了模型不确定性的比例很大（64%）；2）地下水过程是该流域最重要的水文过程；3）构建的流域水文模型在率定期和验证期日径流的Nash–Sutcliffe效率系数（NS）和R2都达到0.80以上，模拟效果很好；4）当不考虑气象要素随高程变化时，率定的水文模型效率系数只有NS = 0.47（考虑气象要素随高程变化时NS = 0.80），表明气象要素空间变化对分布式水文模型非常重要，尤其在气象资料稀少的山区更加显著。
3、比较分析了五种降水校正方法和三种温度校正方法在区域气候模式（Regional Climate Model, RCM）降尺度中的应用效果。降水校正方法包括线性缩放（Linear Scaling, LS）、局部强度缩放（Local Intensity Scaling, LOCI）、幂转换（Power Transformation, PT）、Gamma分布映射（Distribution Mapping, DM）和分位数映射（Quantile Mapping, QM）、温度校正方法包括LS、方差缩放（VARIance scaling, VARI）和Gaussian分布映射（DM）。结果表明：1）相对于观测数据，原始的RCM模拟存在严重偏差，导致了用原始RCM驱动的水文模拟存在严重偏差；2）所有的偏差校正方法均可有效提高模拟效果；3）对于降水而言，PT和QM方法在校正基于频率的指标（如标准偏差、百分位数值）方面表现最好；LOCI方法在校正基于时间序列的指标（如NS、R2）方面表现最好；4）对于温度而言，所有的校正方法校正效果均较好，各校正方法差别不大；5）对于模拟径流而言，降水校正方法比温度校正方法对径流模拟的影响更为显著。虽然这种降尺度方法是基于特定的RCM和水文模型应用于干旱区的特定流域，但是这种分析方法和一些结论可以应用到其他地区和模型中。
4、采用校正后的RegCM4.0预测的未来气候数据驱动率定好的水文模型，评估了21世纪开都河流域水文过程对气候变化的响应。结果表明：1）相对于1986 ~ 2005年，在RCP4.5和RCP8.5下，21世纪末开都河流域温度将升高2.2°C和4.6°C，降水将增加2% ~ 24%（干季显著增加、湿季变化很小）；2）相对于1986 ~ 2005年模拟的多年平均径流量（36.1 m3×108），21世纪径流相对变化-1% ~ 20%。未来蒸散发量将增加2% ~ 24%。利用简单气候变化（Simple Climate Change, SCC），即相对于1986 ~ 2005年温度绝对变化量设置为-1 ~ 6°C、降水相对变化设置为-20% ~ 60%，分析了开都河流域水文过程对气候变化的响应，发现径流几乎随降水增加呈线性增加，而径流对温度的响应则依赖于温度变化幅度，当温度增加大于2 °C时，径流显著降低。
|其他摘要||Global climate change has created social, economic and ecological problems, which have been a great concern for scientists, the public and government in recent decades. The arid and semi-arid regions are sensitive to climate change. The arid region in Central Asia is one of the most important arid areas in the world, where water originates from the mountains. Climate change impact on hydrological processes in this region has become a scientific hot spot. In this study, we investigated future climate change in the mountainous regions of Central Asia (MCA), studied the hydrological processes in a typical Kaidu River Basin, discussed and compared different bias correction methods in downscaling climate models and finally, analyzed the future hydrological processes in the Kaidu River Basin.
1. We estimated and predicted future climate change in the MCA based on an ensemble of 21 General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5) by using the Bayesian Model Averaging (BMA) technique. The results showed: (1) BMA outperforms the simple ensemble analysis and the BMA mean matches all three observed climate variables (i.e. temperature, precipitation, snowfall); (2) At the end of 21st century, generally, mean annual temperature will rise considerably by 5.0 °C, mean annual precipitation will increase by 5.9% from 186 mm to 197 mm, and mean annual snowfall will dramatically decrease by 26.4% from 72 mm to 53 mm under RCP8.5 compared to those in the control period (1976~2005); (3) Precipitation is increasing in the North Tianshan area, while it is decreasing in the Amu Darya region and snowfall shows a significant decrease in the western part of the Tianshan. The snowfall fraction (S/P) will also decrease from 0.58 to 0.43; (4) Snowfall shows a high sensitivity to temperature in autumn and spring of -25.9 ~ -1.5%/°C, while a low sensitivity during winter (about -8.6 ~ 3.8%/°C). In winter, the temperature has a positive impact on snowfall for 56% of grids. The climate change in the MCA is featured by an increasing temperature and precipitation but a decreasing snowfall, which poses a potential flood risk and which may cause a loss of solid water storage in the MCA and seasonal shifts in the runoff.
2. We analyzed the hydrological processes and quantified the contribution of the meteorological input to the model output by coupling the Morris method and the SDP method (State-Dependent Parameter method) and by applying a distributed hydrologic model of SWAT (Soil and Water Assessment Tool). (1) The meteorological input contributes up to 64 % of model uncertainty due to the scarcity of the observed meteorological data, especially in the alpine region; (2) The groundwater flow is the most important hydrological process in this watershed; (3) The model calibration is robust with the Nash–Sutcliffe coefficients (“NS”s) and “R2”s over 0.80 for both the calibration and the validation period (considering the length of the validation period is five times larger than the calibration period). The significance is obvious when compared to the simulation without considering the effect of spatial variation in the meteorological input (NS = 0.80 and NS = 0.47 for “with lapse rates” and “without lapse rates”, respectively). An accurate meteorological input is of great importance to the distributed hydrological model, especially in the mountainous regions.
3. We have compared five precipitation and three temperature correction methods in downscaling RCM (Regional Climate Model) simulations applied to the Kaidu River Basin. The implemented precipitation correction methods include linear scaling (LS), local intensity scaling (LOCI), power transformation (PT), distribution mapping (DM) and quantile mapping (QM), while the temperature correction methods include LS, variance scaling (VARI) and DM. (1) Streamflows are sensitive to precipitation, temperature and solar radiation but not to relative humidity and wind speed; (2) The raw RCM simulations are heavily biased from the observed meteorological data and their use for simulations results in large biases from the observed streamflow and all bias correction methods improved these simulations effectively; (3) Concerning precipitation: the PT and QM methods performed equally best in correcting the frequency-based indices (e.g. standard deviation, percentile values), while the LOCI method gained better results in terms of the time-series-based indices (e.g. NS, R2); (4) For temperature: all correction methods performed equally well in correcting the raw temperature; and (5) As to the simulated streamflow, the precipitation correction methods have a more significant influence than the temperature correction methods and the streamflow simulation performances are consistent with those of the corrected precipitation. The case study concerns an arid area in China (based on a specific RCM and hydrological model) but the methodology and some results can also be applied to other areas and models.
4. We assessed the impact of future climatic changes on hydrological processes in the Kaidu River Basin by using future climate data to force a well-calibrated hydrologic model. The future climatic data are bias-corrected RCM outputs for RCP4.5 and RCP8.5. (1) The temperature is likely to increase by 2.2 and 4.6 °C by the end of the 21st century under RCP4.5 and RCP8.5 respectively, while precipitation will increase by 2 ~ 24%, with a considerable rise in the dry season and a small change in the wet season; (2) The flow will change by -1 ~ 20%, while the evapotranspiration will increase by 2 ~ 24%. Also simple climate change (SCC) with an absolute temperature change of -1 ~ 6 °C and a relative precipitation change of -20 ~ 60%, was used to detect the responses of hydrological processes to climate change. Flow increases almost linearly with precipitation, while its response to temperature depends on temperature change magnitude and it decreases significantly for a temperature increase larger than 2 °C.
This dissertation evaluated the climate change impact on the hydrological processes synthetically. The innovations are as follows. Firstly, the future climate was predicted using BMA technique to incorporate information of different GCMs. Secondly, for the first time the contribution of meteorological inputs in hydrological modeling was quantified. Thirdly, different bias correction methods were compared in the Kaidu River Basin and the evaluation procedure could be used in other watersheds and models. However, every dissertation has its limitations, and this work was no exception. Firstly, the glacier processes were not included in the hydrological modeling, which could introduce some biases in the future predictions. Secondly, there may be great uncertainty in the hydrological modeling due to the scarce data.|