欧洲最大的冰帽­––冰岛瓦特纳冰帽，在未来21世纪可能随全球变暖而呈现出加速消融的态势。平流层硫酸盐气溶胶注入下的地球工程G4试验作为抑制全球变暖的潜在备选方案，一方面，可能通过抵消部分太阳辐射而降低冰岛气温，从而抑制瓦特纳冰帽消融。但另一方面，该试验也会增加北大西洋经向翻转环流（AMOC）强度，将更多热通量输送至冰岛周边区域。因此，G4试验能否缓解冰帽消融具有很大不确定性。

在第五次耦合模式相互比较计划（CMIP5）中，全球有多个地球系统模式参与了G4试验的气候态模拟，本研究首先对包含G4试验的4种地球系统模式（BNU-ESM、HadGEM2-ES、MIROC-ESM和MIROC-ESM-CHEM）输出的气候场进行统计降尺度，用以驱动表面物质能量平衡模型SEMIC，来模拟瓦特纳冰帽表面径流和表面物质平衡（SMB）在现代和未来的变化。再基于SMB、冰温以及地热通量等边界条件，采用冰盖动力学模式PISM，进一步估算冰帽体积、面积以及物质平衡变化，并探究了G4试验对冰帽产生的影响。本研究模拟的情景分别为CMIP5历史情景（1982–2005年）、典型浓度路径RCP4.5、RCP8.5情景（2006–2089年）以及地球工程G4情景（2020–2089年）。G4情景从2020年开始，在RCP4.5情景下，在赤道地区以5 Tg yr^-1的速率向平流层注入硫酸盐气溶胶，至2069年停止，但继续模拟至2089年，用以探究试验骤然停止后对气候产生的影响。本文的主要研究结果如下：

（1）硫酸盐气溶胶注入期间（2020–2069年），瓦特纳冰帽的近地面气温、地表下行长波辐射在所有情景下，都随时间呈上升趋势。相反，降雪呈下降趋势。而总降水基本保持不变。G4情景下的近地面气温、地表下行长波辐射比在RCP4.5情景下分别低0.4℃和2.4 W m^-2，证明地球工程G4试验能够抑制瓦特纳冰帽气温的不断上升。G4试验能够显著增加瓦特纳冰帽的降雪量，但并不会改变到达冰帽上的总降水量。

   （2）在2020–2069年间，SEMIC模型的模拟结果表明，G4情景下的瓦特纳冰帽表面径流要比在RCP4.5和RCP8.5情景下分别低6±6%和7±6%。冰帽SMB在G4、RCP4.5和RCP8.5情景下分别为-0.34±0.18 m
yr^-1、-0.56±0.06 m yr^-1和-0.66±0.04 m yr^-1。与RCP4.5情景对比，G4情景下径流量减少并且SMB增加，说明地球工程G4试验将有效抑制冰帽的持续消融。地球工程G4试验在2069年停止后的20年间，它对冰帽消融的抑制效应仍然存在。 

（3）PISM模式能够较好地重现瓦特纳冰帽在历史时期的初始状态。在模式的预热模拟（spin-up）中，冰帽大约在第1500年时达到稳态，即冰帽体积和面积不随时间改变。达到稳态后的冰帽体积和面积分别比历史观测值低0.1%和2.5%，但冰帽形态的空间分布差异明显，在冰帽边缘区，模拟值比观测值低50–150 m；而在冰帽内陆区域，模拟值比观测值高50–180 m。
   （4）1982–2089年间，瓦特纳冰帽体积在G4、RCP4.5和RCP8.5气候情景下分别减少了12±2%、16±4%和22±2%；面积分别减少8±2%、10±3%和14±1%。在2020–2069年间，冰帽物质平衡在G4情景下，分别比在RCP4.5和RCP8.5情景下高0.21±0.17 m yr^-1和0.33±0.22 m yr^-1。瓦特纳冰帽非SMB（动力学过程和地热引起的冰帽底部消融）导致的物质损失在所有气候情景下都为0.25 m yr^-1，且不随时间改变，证明冰帽非SMB对气候变暖的响应不敏感。对比前人研究G4试验对格陵兰冰盖的影响，G4试验对瓦特纳冰帽消融的减缓效率仅为格陵兰冰盖的63%。主要原因在于G4试验使得AMOC强度增加，为冰岛带来了更多的热量，在一定程度上削弱了G4试验对冰岛气温升高的抑制作用。

The largest ice cap in Europe, the Vatnaj?kull ice cap (VIC), is likely to melt at an accelerated rate in the 21st century associated with global warming. Geoengineering of G4 experiment by sulfate aerosol injection (SAI), as a potential alternative approach to mitigate global warming, may lower Icelandic temperatures by blocking solar radiation, thereby reducing melting of the VIC. However, the G4 experiment could enhance the Atlantic Meridional Overturning Circulation (AMOC), which would result in more heat flux being transported to the vicinity of Iceland. Therefore, whether G4 could reduce the ice cap melting remains large uncertainties.
There are several Earth System Models (ESMs) that simulate the G4 experiment in CMIP5 (Coupling Model Intercomparison Program). In this study, we use climate outputs from 4 available ESMs that with G4 experiment to simulate the evolution of VIC during historical and future periods. We statistically downscale the climate forcing from the ESMs we used, in order to drive surface energy and mass balance model SEMIC, to simulate the VIC surface runoff and surface mass balance (SMB). We also run the ice sheet model PISM with the physical boundaries of SMB, ice temperature, geothermal flux, etc., to estimate the changes in VIC volume, surface area, and mass balance, as well as to explore the response of VIC to the G4 experiment. We use the following scenarios: the CMIP5 historical scenarios (1982-2005), the Representative Concentration Pathway (RCP) 4.5 and 8.5 scenarios (2006–2089), and the geoengineering G4 scenario. The G4 scenario is based on the background of the RCP4.5 scenario, but specifies 5 Tg yr-1 of SO2 to be injected into the stratosphere during the period 2020–2069, and continues to run until 2089 with the RCP4.5 forcing, in order to see the termination effect brought by SAI. The main results of this thesis are as follows:
(1) During the SAI period (2020–2069), VIC near-surface air temperature and downward longwave radiation increase with time in all scenarios. In contrast, snowfall displays a downward trend, while precipitation shows no significant trend. Near-surface air temperature and downward longwave radiation under G4 are 0.4℃ and 2.4 W m-2 lower than those under RCP4.5, respectively. G4 increases VIC snowfall, but does not significantly change the amount of precipitation.
(2) During 2020–2069, simulations from SEMIC show that VIC surface runoff under G4 is 6±6% and 7±6% lower than that under RCP4.5 and RCP8.5 scenarios. VIC SMB under G4, RCP4.5 and RCP8.5 are -0.34±0.18 m yr-1, -0.56±0.06 m yr-1, and -0.66±0.04 m yr-1, respectively. Compared to RCP4.5, there are significant reductions in surface runoff under G4 associated with the increase in SMB, indicating that geoengineering could mitigate the melting of VIC. The mitigating effects still hold during the post-SAI period 2069–2089.
(3) The PISM model is capable of reproducing the initial state of the historical VIC geometry. In the spin-up simulations, the ice cap reaches the steady state (the ice cap volume and area remain constant over time) after about 2000 years. The equilibrium VIC volume and area are 0.1% and 2.5% less than measurements, respectively. However, spatial changes in ice cap thickness vary regionally, the equilibrium volume is 50–150 m lower than observation at the margin of VIC, while 50–180 m higher than inland.
(4) During the whole period 1982–2089, VIC volume under G4, RCP4.5 and RCP8.5 scenarios reduce by 12±2%, 16±4% and 22±2%, and area reduce by 8±2%, 10±3% and 14±1%, respectively. During the SAI period 2020–2089, VIC mass balance under G4 is 0.21±0.17 m yr-1 (0.33±0.22 m yr-1) lower than that under RCP4.5 (RCP8.5) scenario. Mass loss from the non-SMB (basal melt due to geothermal heat flux and ice dynamics) is about 0.25 m yr-1 under all scenarios and do not change over time, indicating that the non-SMB is insensitive to climate warming. Compare with existing studies of the Greenland ice sheet in response to the G4 experiment. The efficiency of SAI G4 over VIC is only 63% of that for Greenland ice sheet. This is because G4 experiment enhances the AMOC, thus bringing more heat flux to Iceland and offsetting the mitigating impact brought by G4.