卫星和地面站资料联合同化及其对陆面模式积雪过程模拟的影响

Snow plays a critical role in global and regional water and energy cycle due to its special hydrothermal property. Snow is mostly distributed in cold area with complex terrain and its spatial distribution varies rapidly, the mechanism of snow process is still not clear currently. This is a severe constrain on the precision of regional numerical weather forecast and streamflow prediction, which places a threat on people's life and property security. Limited by the uncertainty in meteorological forcing data and physical parameterization, snow process was identified as a major weakness in land surface models. At the same time, surface station observations and satellite data contain rich information. Assimilation of multi-source snow data into land surface model is a development trend. However, as the surface station observations and satellite snow data have different variables and are inconsistent in spatial and temporal scale and they both have their limitations and uncertainty, the research of assimilating them simultaneously is rare currently. In this project, we propose novel cross validation and joint assimilation scheme to combine the strength of physical parameterization of snow process in advanced land surface model and multi-source snow data and analyze the law and mechanism regarding spatio-temporal evolution of snow process in depth, which could provide reasonable scientific guide for improving the parameterization scheme of snow process in numerical weather models and make preparation for the successfully holding of 2022 Beijing-Zhangjiakou Winter Olympic Games.

积雪因其特殊的水热属性，对全球和区域的水热循环有显著的影响。积雪多分布于观测站点稀疏、地形复杂的高寒地带，且空间分布变化迅速，目前对积雪过程的反馈机理的认识尚不清晰，这严重地制约了区域数值天气预报和水文预报的精度，对人们的生命和财产安全构成了威胁。受限于驱动数据和模型物理结构的不确定性，陆面模式对积雪过程的模拟是一个突出的弱点，与此同时，站点观测和卫星积雪资料含有丰富的信息，陆面模式的多源积雪资料同化是一个研究趋势。然而，由于站点和卫星积雪资料的观测变量和时空尺度不统一且存在各自的局限性和不确定性，目前将二者进行联合同化的研究较少。本项目提出新颖的交叉验证和联合同化方案，将先进的陆面模式的积雪物理参数化方案和多源积雪观测的优势结合起来，深入分析积雪的时空演化规律和机理。该研究有助于为数值天气模式中的积雪参数化方案的改进提供合理的科学依据，为2022年北京-张家口冬奥会的成功举办做准备。

积雪在局地和全球水热循环中占有重要地位，且积雪多分布于地形复杂、观测站点较稀疏的高寒地区，且空间分布变化迅速，陆面模式对积雪过程的模拟是一个突出的弱点。本研究采用集合卡尔曼滤波（EnKF）同化方法将融合的积雪资料同化到多物理过程陆面模式（Noah-MP）中，从而将陆面模式物理参数化方案和多源积雪资料的优势结合起来，提升了积雪模拟的精度。积雪资料融合方面，采用薄板平滑样条模型融合了站点观测积雪深度资料和卫星积雪资料，充分利用了站点稀疏和密集地区两种资料的优势，并基于薄板平滑样条模型定量化估计了融合积雪资料的不确定性。本研究通过提升驱动资料的精度以及更新静态资料提升了陆面要素的模拟精度，并基于方差分析提出了定量评估驱动资料、物理参数化方案和物理参数的不确定性的方法，该方法可以应用在其他区域的模式模拟研究中。采用MODIS卫星反照率资料和地面站地表温度资料的检验结果显示，同化后模式模拟的反照率和地表温度的精度也有所提升，这表明积雪资料同化改善了模式模拟的水分和能量循环的精度，这对于区域数值天气预报精度的提升有积极的意义。目前本研究的积雪同化的工作已经在区域数值天气预报模式中进行了应用检验，有一些初步的结果。