Direct assimilation of satellite radiance data in GRAPES variational assimilation system

Variational method is capable of dealing with observations that have a complicated nonlinear relation with model variables representative of the atmospheric state, and so make it possible to directly assimilate such measured variables as satellite radiance, which have a nonlinear relation with the model variables. Assimilation of any type of observations requires a corresponding observation operator, which establishes a specific mapping from the space of the model state to the space of observation. This paper presents in detail how the direct assimilation of real satellite radiance data is implemented in the GRAPES-3DVar analysis system. It focuses on all the components of the observation operator for direct assimilation of real satellite radiance data, including a spatial interpolation operator that transforms variables from model grid points to observation locations, a physical transformation from model variables to observed elements with different choices of model variables, and a data quality control. Assimilation experiments, using satellite radiances such as NOAA17 AMSU-A and AMSU-B （Advanced Microwave Sounding Unit）, are carried out with two different schemes. The results from these experiments can be physically understood and clearly reflect a rational effect of direct assimilation of satellite radiance data in GRAPES-3DVar analysis system.     

Scientific design and preliminary results of three-dimensional variational data assimilation system of GRAPES

The scientific design and preliminary results of the data assimilation component of the Global-Regional Prediction and Assimilation System （GRAPES） recently developed in China Meteorological Administration （CMA） are presented in this paper. This is a three-dimensional variational （3DVar） assimilation system set up on global and regional grid meshes favorable for direct assimilation of the space-based remote sensing data and matching the frame work of the prediction model GRAPES. The state variables are assumed to decompose balanced and unbalanced components. By introducing a simple transformation from the state variables to the control variables with a recursive or spectral filter, the convergence rate of iteration for minimization of the cost function in 3DVar is greatly accelerated. The definition of dynamical balance depends on the characteristic scale of the circulation considered. The ratio of the balanced to the unbalanced parts is controlled by the prescribed statistics of background errors. Idealized trials produce the same results as the analytic solution. The results of real data case studies show the capability of the system to improve analysis compared to the traditional schemes, Finally, further development of the system is discussed.     

Effect of four-dimensional variational data assimilation in case of nonlinear instability

The effect of four-dimensional variational data assimilation on the reduction of the forecast errors is investigated for both stable and unstable flows. Numerical results show that the effect is generally positive. Particularly,its effect is much more significant in the presence of nonlinear instability     

2010年7月8日英山特大暴雨过程的成因分析

利用GFS(0.5°×0.5°)6 h再分析资料、LAPS资料以及FY-2c卫星云图资料、武汉区域雷达拼图产品和自动站实时资料,对2010年7月8日发生在英山的特大暴雨过程进行了分析.结果表明:控制湖北地区的副高南退时,副高外围东移的短波槽引导北方冷空气下滑是此次过程的触发机制;低层切变、低空急流、露点锋和地面辐合线引起了低层辐合、中层辐散、低层θse锋生及中上层θse锋区下移,提供了有利的动力条件;750～1 000 hPa的高湿区是水汽条件;温度槽温度脊的配合、对流有效位能的持续增加和湿位涡上正下负的配置为此次过程提供了不稳定能量.     

利用改进的并行粒子群算法对变分资料同化的研究

近年来,为了提高同化精度和减少同化时间,粒子群算法(PSO)被引入到数值天气预报资料同化中来.粒子群算法虽然令同化精度有所提高,但同化时间仍然存在较大缺陷.基于此,首先设计了一种改进的并行粒子群算法(P2PSO),然后应用于含不连续"开关"过程的变分资料同化中,与时变双重压缩因子粒子群算法(PSOTVCF)和动态权重粒...     

一次区域性暴雨数值模拟及诊断分析

通过对陕西省2003年7月15日区域性暴雨过程进行分析发现,西太平洋副热带高压西侧的暖湿气流和新疆冷空气是这次暴雨的主要影响系统,充沛的水汽输送、能量的积聚和强烈的上升运动为暴雨的发生提供了充分的条件,700 hPa和850 hPa的低涡、切变线、地面倒槽是暴雨形成的触发机制.数值模拟结果表明:涡度场和散度场互耦;特强散度柱出现早于强涡度柱,而深厚的强气旋性涡柱则几乎与暴雨最强盛时期同时出现.     

海洋数值模拟中的伴随数据同化方法

从伴随方程的推导、伴随数值模型的构造以及伴随数据同化中的最优化方法3个方面对伴随数据同化基本理论的研究现状进行了较为系统的归纳和评述;从优化初始状态、优化模式中不确定参数和优化开边界条件等3方面,综述了伴随法在海洋数值模拟中的应用.最后讨论了伴随法在海洋数值模拟应用中存在的问题,指出,对于伴随法用于数值模型的参数优化,只能对其在物理意义范围内进行适当调整和估计,而不能背离物理背景使目标函数最优.     

SST data assimilation experiments using an adaptive variational method

An adaptive variational data assimilation method is proposed by Zhu and Kamachi^[1]. This method can adaptively adjust the model state without knowing explicitly the model error covariance matrix. The method enables very flexible ways to form some reduced order problems. A proper reduced order problem not only reduces computational burden but also leads to corrections that are more consistent with the model dynamics that trends to produce better forecast. These features make the adaptive variational method a good candidate for SST data assimilation because the model error of an ocean model is usually difficult to estimate. We applied this method to an SST data assimilation problem using the LOTUS data sets and an ocean mixed layer model (Mellor-Yamada level 2.5). Results of assimilation experiments showed good skill of improvement subsurface temperatures by assimilating surface observation alone.     

Forming proper ensemble forecast initial members with four-dimensional variational data assimilation method

A method has been presented to improve ensemble forecast by utilizing these initial members generated by four-dimensional variational data assimilation (4-D VDA), to conquer limitation of those initial members generated by Monte Carlo forecast (MCF) or lagged average forecast (LAF). This method possesses significant statistical characteristic of MCF, and by virtue of LAP that contains multi-time information and its initial members are harmonic with the dynamic model, six groups of numerical control and contrast experiments were performed with T42 spectral 4-D VDA system. The results show that in dekad range ensemble forecast with initial members generated by this method is prior to that by LAP. The anomaly correlation coefficient (ACC) of 500 hPa height in 10-d average by employing this method is 0.01-0.04 larger than by LAP, and root mean square error (RMSE) less than 0.2-0.4 dagpm.     

基于并行粒子群优化算法的变分资料同化

为了提升粒子群优化算法(PSO)应用到实际大气和海洋模式的资料同化时的计算时效性,针对一种最新提出的基于方向的粒子群优化算法(PSOBD),使用消息传递接口(MPI)和整体同步并行(BSP)计算模型,设计PSOBD的并行算法,较好地实现了BSP结构下PSOBD的全局通信操作。为检验并行化后的PSOBD的计算时效,将其用于潜水方程的资料同化并与基于串行的PSOBD的同化数值试验结果比较。大量的孪生同化试验结果的统计分析显示,并行化后的PSOBD与串行PSOBD一样能产生高质量的同化结果,且计算时效相对串行PSOBD提高了13倍以上。这一结果为PSOBD算法用于实际模式的四维变分资料同化(4D-Var)提供了依据。     

海洋变分同化背景场误差协方差矩阵构造技术

为改进背景场误差协方差矩阵构造技术,提高海洋资料变分同化效果,首先将背景场误差协方差矩阵分解为背景误差标准偏差矩阵、非平衡变量空间相关算子以及多变量间平衡算子3部分,然后利用广义扩散方程模拟空间相关算子,联合引入温盐约束关系、线性状态方程、海表高度诊断方程、静力平衡、地转平衡关系模拟多变量平衡,最后给出了一种适用于海洋资料变分同化的背景场误差协方差矩阵构造方案。单点海表温度观测模拟同化试验表明,该方案可实现观测信息在三维空间、多变量间的传播,充分体现了背景场误差协方差矩阵在同化中的作用。     

2010年云南“6.25”特大暴雨中尺度特征及成因的数值模拟分析

 使用1°×1°NCEP再分析资料、TBB黑体亮温资料和自动站逐时降水资料,应用MM5V3.6中尺度数值模式对2010年6月25日发生在云南曲靖马龙罕见的特大暴雨进行数值模拟,主要从动力结构和不稳定机制研究本次特大暴雨的中尺度特征及其成因.研究表明：降水发生期间中低层有较强的水汽辐合柱存在,强降水发生前中低层积累了较大的对流不稳定能量,近地层辐合线的形成诱发中低层辐合的产生,强烈的垂直上升运动促使对流不稳定能量迅速释放,触发强对流降水,高层辐散中低层辐合形成的“抽吸作用”以及利于垂直上升运动和降水的维持.地形敏感试验表明,近地层辐合线出现的位置与云南地形有很大的关系,从而影响强对流暴雨的落区.     

具有“on-off”过程的变分资料同化

为了验证伴随方法的有效性,利用一个含有"on-off"振荡的概念模式,研究了含物理参数化不连续性的变分资料同化中,代价函数关于初值梯度的存在性及伴随方法的可行性.结果表明:在解析情形中,尽管控制方程中出现"on-off" 振荡,代价函数关于初值的梯度除个别情形外总是存在的,且可由伴随模式的反向积分来表示.     

用四维变分法同化自动站降水资料

用自动站降水资料作了四维变分同化试验.试验表明,由于它的加入,增加了初始场中的中尺度信息,改进了中尺度数值模式MM5的预报,增强了模拟开始阶段的降水量, 改进了降水量的落区预报,减弱了模式开始阶段的"spin-up"现象.试验还表明,自动站降水资料的时间变化信息,在同化时也起重要作用.     

西北地区暴雨时空变化及异常年夏季环流特征

利用中国西北地区248个观测站1981-2010年逐日降水量资料、NCEP/NCAR再分析资料,对西北地区暴雨的时空变化及环流异常特征进行了分析.结果表明:暴雨主要出现在陕西、甘肃东南部、宁夏东南部,暴雨出现次数依次由东南向西北减少.暴雨的年际差异较大,年内主要集中在7-8月,两个月的暴雨日数占全年的64%,7月最多,8月次之,区域性暴雨7月下旬最多.西北地区大范围暴雨过程偏多年份夏季(7月和8月),西太平洋副热带高压偏强,西脊点伸到110?E附近,亚洲中高纬度多低值系统活动,低层(700 hPa)来自南海的偏东气流与青藏高原南侧的偏西气流合并为偏南气流向北伸展到西北地区东部,将充沛的水汽输送到这一区域,到达35?N附近与西北气流相遇;反之,西北地区大范围暴雨过程偏少年份夏季(7月和8月),西太平洋副热带高压偏弱,远离大陆,亚洲高纬度多高值系统活动,低层是一支偏北气流控制西北地区东部.     

环境大气中数据同化技术方法及应用

 从数据同化的角度,对卡尔曼滤波技术、元启发式算法、贝叶斯推导及逆扩散/轨迹模拟技术、非参数化回归等技术在污染环境大气模拟技术中的应用情况进行了总结分析。分析结果表明,经多种优化方法试用于解决环境监测数据的同化问题中,元启发式算法结合传统优化技术在有效解决环境监测的数据同化问题中具有良好应用前景。     

改进的粒子群算法在太阳能光伏发电资料同化中的应用研究

资料同化是目前太阳能光伏发电预测研究的一个关键和难点.近年来,遗传算法和粒子群算法等智能优化算法被引入到四维变分同化中.针对基于分子运动论的粒子群算法(MPSO)在处理大量数据时速度慢的不足,该文提出了并行分子运动论粒子群算法(PMPSO),并行计算的基本思想是将粒子群分成N个子集,每个子集交给一个线程控制,同时进行粒...     

WRF模式在云南一次强降水过程中的应用

利用WRF数值模式,针对云南省一次强降水过程进行的数值试验,得到几个有意义的结论:①WRF模式的三维变分同化系统能反映地转平衡等多变量之间的相互关系.②在此次强降水过程试验中,同化系统能把雷达反演风场中低空急流信息有效地引入模式初始场,并按照同化模型进行增量空间分配,有助于增强降水区的水汽输送,改善模式较强量级降水的预报漏报率.③多个时次雷达资料的同化对降水预报的改善更加突出,具有高时空分辨率特性的雷达资料同化在提高客观预报质量方面有着较好的应用前景.     

大亚湾海域流场数据同化的数值试验

用有限差分隐式方向交替(alternatingdirectionimplicit,ADI)法求解潮波方程,并进行数据同化的数值试验,与未经同化的ADI法计算值相比有较好的效果.结果表明,采用数据同化系数1 12×10-4s-1为最佳选择.