20th- and 21th-century climate changes in western China simulated with a hierarchy of global and regional climate models, including improved scenarios of human activity and reliable observational data for validation

AKSU TARIM-CLIM自2010年10月起由德国研究基金会（DFG）资助。此后，中国西部和中亚地区气候变化区域评估取得了重要进展。利用不同分辨率（0.5°和0.16°）的区域气候模式REMO进行了动态降尺度，与驱动的全球气候模式和再分析相比，模式偏差有所改善。这一成就使人们能够对亚洲山区未来的气候变化有一个相当详细的了解。另外，还开发了一种统计动态降尺度方法，并与REMO结果进行了比较。后一种方法首先分析影响区域的大尺度天气类型。对于每种天气类型的选定和代表性事件，使用区域气候模式CCLM在7 km（0.0625°）分辨率下进行动态降尺度。最后，利用全球气候模式模拟的天气类型频率将这些信息重新组合，以获得高分辨率的气候预估。总的来说，中国西部气候预估的动力和统计降尺度的附加价值可以被识别，模拟的气候数据现在将被应用于阿克苏塔里木项目组的影响模型，以评估天山山脉冰川质量平衡、冰川径流、永久冻土和积雪的变化。然而，仍然存在一些模式缺陷，需要进一步研究，以改善气候、冰川和水文模式之间的接口。为此，我们申请在AKSU TARIM-CLIM目前为期两年的项目阶段之后再增加一个项目年。在本项目期间，我们计划(1)开发和应用模型数据的统计后处理，以消除REMO和CCLM中的系统偏差，并将网格箱数据调整为单个冰川水平所需的当地气候信息。(2)中亚平原夏季明显偏暖是REMO最明显的大尺度缺陷之一。我们假设这种偏差是由于缺少灌溉的土地利用过程，因此，我们打算用REMO进行敏感性研究，系统地改变土地覆盖对灌溉的影响。这对中亚的整个水文循环具有特别的价值，因为灌溉土地表面的蒸发蒸腾是该地区水汽的重要来源，因此也是积雪的重要来源。(3)统计动力降尺度是区域气候变化评估中一种快速而有价值的选择。将开发一种新的方法，适用于2.8公里的空间尺度，更接近冰川的空间尺度，适用于日值和其他气象变量，特别是温度。此外，统计动力降尺度将通过使用来自不同gcm的变化的天气类型频率来扩展，以便更好地估计与目标区域气候预估相关的不确定性。

AKSU TARIM-CLIM is funded by the German Research Foundation (DFG) since October 2010. Since then important improvements in the regional assessment of climate change in western China and Central Asia have been achieved. Dynamical downscaling with the regional climate model REMO at different resolutions (0.5° and 0.16°) have been performed, revealing improvements in the model bias compared with the driving global climate model and re-analyses. This achievement enabled a rather detailed picture of future climate change in the mountainous region of Asia. Alternatively, a statistical-dynamical downscaling approach has been developed and compared with REMO results. The latter method first analyses the large-scale weather types affecting the region. For selected and representative episodes for each weather type, dynamical downscaling with the regional climate model CCLM at a resolution of 7 km (0.0625°) is performed. Finally, this information is recombined using the frequencies of weather types from global climate model simulations to obtain high resolution climate projections. In general, the added value of dynamical and statistical downscaling of climate projections in western China could be identified and the simulated climate data will now be applied to impact models within the AKSU TARIM project bundle, assessing changes in glacier mass balance, glacier runoff, permafrost and snow cover in the Tien Shan mountain range. Yet, some model deficiencies still exist which require further investigations in order to improve the interface between climate, glaciological and hydrological models. With this aim, we apply for an additional project year, following the current two-year project phase of AKSU TARIM-CLIM. During this project year we plan to (1) develop and apply a statistical postprocessing of the model data in order to remove systematic biases in REMO and CCLM and to adjust grid box data to local climate information required at level of the individual glaciers. (2) One of the most obvious larger-scale deficiencies in REMO consists of a remarkable summer-time warm bias over the Central Asian plain. We assume that this bias is due to the missing land-use process of irrigation and, hence, intend to carry out sensitivity studies with REMO, systematically modifying land-cover towards irrigation. This would be of particular value for the entire hydrological cycle over Central Asia, since evapo-transpiration from irrigated land surfaces represents a crucial source of water vapor and, thus, also snow accumulation in the region. (3) The statisticaldynamical downscaling has been found to be a quick and valuable option in regional climate change assessment. A new approach will be developed which applies to a spatial scale of 2.8 km, closer to the spatial scale of glaciers, to daily values and to additional meteorological variables, in particular temperature. Additionally, the statistical-dynamical downscaling will be extended by using changing weather type frequencies from different GCMs, in order to gain better estimation of the uncertainties associated with the climate projections for the target region.