THE SENSITIVITY INFLUENCE OF NUMERICAL MODEL INITIAL VALUES ON FOUR-DIMENSIONAL ASSIMILATION—STUDY BASED ON LORENZ SYSTEM

Numerical experiments of adjoint variational assimilation have been performed using the knownLorenz system.With the increase of sensitivity of model's initial values,it is more and more difficultto use the adjoint method to get the initial values which are consistent with the dynamics of the fore-cast model.Under some circumstances the algorithm completely fails.This shows that four-dimen-sional assimilation is related to the limit of predictability.On the other hand.with the increase ofmodel equation's error,the result of variational assimilation may become worse and worse so that theprediction has no meaning.But if the model parameters are corrected when variational assimilation ismade,the forecast results can be greatly improved based on Lorenz model.     

THE SENSITIVITY INFLUENCE OF NUMERICAL MODEL INITIAL VALUES ON FOUR-DIMENSIONAL ASSIMILATION-STUDY BASED ON LORENZ SYSTEM*

Numerical experiments of adjoint variational assimilation have been performed using the known Lorenz system.With the increase of sensitivity of model's initial values,it is more and more difficult to use the adjoint method to get the initial values which are consistent with the dynamics of the forecast model.Under some circumstances the algorithm completely fails.This shows that four-dimensional assimilation is related to the limit of predictability.On the other hand.with the increase of model equation's error,the result of variational assimilation may become worse and worse so that the prediction has no meaning.But if the model parameters are corrected when variational assimilation is made,the forecast results can be greatly improved based on Lorenz model.     

GRAPES全球四维变分同化系统极小化算法预调节

在进行多次外循环更新的增量分析框架下,前一次极小化迭代过程中产生的信息可提供给下一次极小化做预调节。该文在GRAPES全球四维变分同化系统中对极小化算法——L-BFGS算法实施了这种预调节,通过全观测的个例试验和批量试验进行评估,发现进行预调节后L-BFGS算法的收敛效率得到明显提高,而且在1个月的循环试验中表现十分稳定。该工作可以帮助GRAPES全球四维变分同化系统有效减少极小化的迭代次数,有利于满足业务化运行的时效要求。另外,间隔6 h和间隔24 h的两次4DVar分析对应的海森矩阵变化不大,因此,前一时刻极小化过程产生的信息提供给后一时刻的极小化进行预调节也有一定效果。     

数值模式初值的敏感性程度对四维同化的影响

用著名的Ｌｏｒｅｎｚ系统作了共轭变分同化的数值试验。发现随着模式对初值敏感性程度的增加，用这种方法得到和模式相协调的初始场愈来愈困难，直到某些情况下的完全失败。这表明四维同化和可预报期限是联系在一起的。另一方面，随着方程不精确程度的增加，变分同化的效果愈来愈差，直到所做的预报无任何意义可言。如果在做变分同化的同时对模式参数也进行反演，就可使得基于Ｌｏｒｅｎｚ系统所做的预报效果大大提高。     

一个用于气候模式的三维变分资料同化系统——方案设计及检验

介绍了一种适合格点模式的三维变分同化方案,并应用于中国科学院大气物理研究所的9层大气环流格点模式（IAP9L2°×2.5°-AGCM）中。同化方案考虑质量场和风场的动力约束关系,并通过变量变换将模式变量转化为互不相关的控制变量：流函数、非平衡速度势、非平衡位势和比湿;同时采用共轭梯度法求解代价函数避免了直接计算背景误差协方差矩阵逆的困难。通过理想试验、模式连续同化积分试验以及对东亚地区特殊寒潮天气的同化试验,结果表明：该系统能够正确反应多变量之间的相互关系,并且长时间运行稳定,更重要的是同化不仅可以明显改善直接参与同化的物理量,而且经过模式的动力调整也改善了未参与同化的物理量（如海平面气压、近地面气温、降水等）,这为大气的持续性信息存储于陆面物理量提供了可能,从而可为IAP9L2°×2.5°-AGCM提供陆—气相协调的初始场。     

Intelligent optimization algorithms to VDA of models with on/off parameterizations

Some variational data assimilation (VDA) problems of time- and space-discrete models with on/off parameterizations can be regarded as non-smooth optimization problems. Same as the sub-gradient type method, intelligent optimization algorithms, which are widely used in engineering optimization, can also be adopted in VDA in virtue of their no requirement of cost functions gradient (or sub-gradient) and their capability of global convergence. Two typical intelligent optimization algorithms, genetic algorithm (GA) and particle swarm optimization (PSO), are introduced to VDA of modified Lorenz equations with on-off parameterizations, then two VDA schemes are proposed, that is, GA based VDA (GA-VDA) and PSO based VDA (PSO-VDA). After revealing the advantage of GA and PSO over conventional adjoint methods in the ability of global searching at the existence of cost functions discontinuity induced by on-off switches, sensitivities of GA-VDA and PSO-VDA to population size, observational noise, model error and observational density are detailedly analyzed. Its shown that, in the context of modified Lorenz equations, with proper population size, GA-VDA and PSO-VDA can effectively estimate the global optimal solution, while PSO-VDA consumes much less computational time than GA-VDA with the same population size, and requires a much lower population size with nearly the same results, both methods are not very sensitive to observation noise and model error, while PSO-VDA shows a better performance with observational noise than GA-VDA. It is encouraging that both methods are not sensitive to observational density, especially PSO-VDA, using which almost the same perfect assimilation results can be obtained with comparatively sparse observations.     

The definition of mesoscale selective forecast error covariances for a limited area variational analysis

Summary ?The paper deals with an alternative formulation of the so-called NMC (National Meteorological Center, now National Centers for Environmental Prediction) statistics to compute the background error covariance matrix to be used in a mesoscale variational analysis. While the standard method uses differences of forecasts valid for the same time, but starting from different analysis times, the new formulation required the recomputation of the short-term forecast with the initial and lateral boundary data that come from the long-term run. In the frame of a limited-area model, this approach forces the error variances at large scales to decrease drastically, because those scales are controlled by the (constant data) lateral boundary coupling. As a result, the background cost function acts more scale selectively, with an emphasis on medium scales. The analysis increments obtained from the 3D-VAR system show that the analysis increments are sharper and more concentrated with the new formulation, both in single observation and in full observation experiments. This work is part of a wider project for building a variational assimilation system inside the ALADIN model. The complete system should concentrate on mesoscale features and it should not reanalyse those scales that were already treated by the global model (ARPEGE). Some difficulties and perspectives are drawn in the concluding discussion. Received February 12, 2001; revised July 24, 2001     

FY-3A卫星微波资料的集合变分混合同化试验

以2012年"北京7.21暴雨"为例,实现了集合变分混合同化方法对FY-3A的微波温度仪和微波湿度仪资料的直接同化,并与三维变分方法进行了比较。结果表明:虽然两种同化方法同化FY-3A微波资料都能改进降水模拟效果,但是与实况相比,集合变分混合同化方法改进效果更为明显,其能有效减少虚假强降水的模拟,改进强降水中心位置的模拟,SAL评分定量检验也同样表明,集合变分混合同化方法对暴雨的模拟效果要优于三维变分同化方法;无论是热力学变量还是动力学变量,集合变分同化得到的初始场均方根误差均显著小于三维变分同化的结果;两种方法同化FY-3A微波资料均能改变初始场中的各种物理量信息,但不同方法得到的同化增量大小和分布却有明显的差异:三维变分同化方法对初始场的调整区域和强度都要大于混合同化方法,且其同化增量表现出均匀和各向同性的分布特点;而利用集合信息的混合同化方法得到的同化增量分布表现为非均匀性和各向异性,具有"流依赖性"的特征,这使得初始场的分布更合理,有利于改善降水的模拟效果。     

GRAPES模式切线性垂直扩散方案的误差分析和改进

针对GRAPES四维变分同化系统的升级, 研究了GRAPES模式垂直扩散方案线性化问题。通过2005年8月7—27日21个个例的批量试验, 发现在GRAPES模式垂直扩散方案源代码的基础上逐句线性化得到的切线性垂直扩散方案即使能通过正确性测试试验, 在少数情况下也会存在很大误差。切线性模式计算的扰动气压场和扰动风场可能出现明显异常, 这种异常与垂直扩散方案中地表动量通量的强非线性有关。如果在切线性垂直扩散方案中忽略地表动量通量扰动, 既可以避免异常的出现, 又不影响其他正常时刻的计算精度。修改后的切线性垂直扩散方案能够在所有变量上一致地提高切线性模式的计算精度。     

An Ensemble-Based Three-Dimensional Variational Assimilation Method for Land Data Assimilation

Land surface models are often highly nonlinear with model physics that contain parameterized discontinuities. These model attributes severely limit the application of advanced variational data assimilation methods into land data assimilation. The ensemble Kalman filter （EnKF） has been widely employed for land data assimilation because of its simple conceptual formulation and relative ease of implementation. An updated ensemble-based three-dimensional variational assimilation （En3-DVar） method is proposed for land data assimilation This new method incorporates Monte Carlo sampling strategies into the 3-D variational data assimilation framework. The proper orthogonal decomposition （POD） technique is used to efficiently approximate a forecast ensemble produced by the Monte Carlo method in a 3-D space that uses a set of base vectors that span the ensemble. The data assimilation process is thus significantly simplified. Our assimilation experiments indicate that this new En3-DVar method considerably outperforms the EnKF method by increasing assimilation precision. Furthermore, computational costs for the new En3-DVar method are much lower than for the EnKF method.     

基于VDRAS的快速更新雷达四维变分分析系统

基于雷达资料快速更新四维变分同化 (RR4DVar) 技术和三维数值云模式,初步研发了一个针对对流尺度数值模拟的快速更新雷达四维变分分析系统。系统通过对京津冀6部多普勒天气雷达资料进行RR4DVar同化,并融合5 min自动气象站观测和中尺度数值模式结果,可快速分析得到12~18 min更新的低层大气三维动力、热力场的对流尺度结构特征。针对2009年7月22日发生在京津冀的一次强风暴个例,通过一系列敏感性试验,并利用局地加密资料进行检验对比,表明有效的雷达资料RR4DVar同化及自动气象站和中尺度模式资料融合方案、恰当的中尺度背景场设置和动力约束方法是获得合理结果的关键。研究也表明:恰当的系统配置能够模拟出与对流生消发展密切相关的近风暴环境特征,包括低层入流、垂直风切变、低层辐合上升和暖舌等,以及风暴自身形成的冷池、出流等与风暴演变密切相关的对流尺度结构。     

基于云依赖背景场误差协方差的雷达资料同化及对降雨预报影响研究

传统变分同化方法中使用各向同性和均质的背景场误差协方差,忽略了背景场误差协方差的天气系统依赖性,而在变分框架下引入集合流依赖的背景场误差协方差还需要额外的集合预报.为在变分同化中引入更合理的背景场误差协方差,通过引入云指数构建"云依赖"背景场误差协方差,提出了一种云依赖背景场误差协方差的同化方案,并应用于雷达等多源观测...     

用变分法修正黑河试验数据的初步研究

为了对黑河试验观测资料进行订正,利用一个一维边界层垂直扩散模式和以拟牛顿极小化算法为基础的最优化方法建立了一个变分同化系统,其中,目标函数相对于控制变量的梯度用扰动法求取.并利用模式资料和黑河试验若干时刻的观测资料进行了一些数值试验.试验证明,将初始场变量与模式误差系数同时作为控制变量加以调整,能够有效调整初始场观测资料.并且,用变分同化方法能够较线性插值(时间、空间)方法更有效、合理的修正黑河试验观测资料场的种种缺陷.     

变分同化方法在Lorenz系统中的简单应用研究

利用Lorenz模式作变分同化数值试验，通过对一个简单系统的讨论，介绍四维变分同化方法。对初值敏感性和观测点的个数及观测值作了对比试验，发现随着模式对初值敏感性的增加，同化效果会越来越差；观测点越少，观测值误差越大，这些都会影响同化效果，甚至导致同化失败。     

集合预报误差在GRAPES全球四维变分同化中的应用研究Ⅰ：局地化方案设计与动力平衡特征分析

通过引入流依赖的集合预报误差,使得同化分析与天气形势紧密相关,是改善初值分析质量的重要途径。文中在GRAPES（Global Regional Assimilation and PrEdiction System）全球四维变分资料同化（4DVar）中研究了如何有效应用集合预报误差,包括增加扩展控制变量时如何降低其计算消耗以及如何在局地化过程中保持不同变量之间的动力平衡。利用高斯分布的谱滤波实现水平局地化,利用垂直正交经验函数分解实现垂直局地化,并采用前8个主导特征模态来限制控制变量空间维数增加。引入20至180个集合样本,在水平二维局地化情形下,控制变量总数的增长可以限制在1.1—1.8倍,而在三维局地化情形下,控制变量总数的增长限制在1.7—7.1倍。对60个集合样本和1°水平分辨率内循环,4DVar引入扩展控制变量后墙钟时间增加了约30%。进一步,通过采用在非平衡分析变量上进行水平局地化,然后再将风压地转平衡关系重新叠加到非平衡分析变量上,使得分析更好地保持了风压平衡关系,初始场地面气压倾向变化减小。此外,虽然垂直局地化对分析平衡影响较大,但依靠目标函数中的数字滤波弱约束,分析变量之间仍能较好满足动力平衡关系。结果表明,GRAPES全球4DVar中发展的增加扩展控制变量、谱滤波实现水平局地化、非平衡分析变量进行水平局地化等有效应用集合预报误差的方法,适合集合样本数超过100个的情况,在分析质量改善的同时,4DVar系统的计算和存储消耗没有显著增加。      

The GRAPES Variational Bias Correction Scheme and Associated Preliminary Experiments

The variational assimilation theory is generally based on unbiased observations. In practice, however, almost all observations suffer from biases arising from observational instruments, radiative transfer operator, precondition of data, and so on. Therefore, a bias correction scheme is indispensable. The current scheme for radiance bias correction in the GRAPES 3DVar system is an offline scheme. It is actually a static correction for the radiance bias before the process of cost function minimization. In consideration of its effects on forecast results, this kind of scheme has some shortcomings. Thus, this study provides a variational bias correction (VarBC) scheme for the GRAPES 3DVar system following Dee’s idea. In the VarBC scheme, the observation operator is modified and a new control variable is defined by taking the predictor coefficients as the control parameters. According to the feature of the GRAPES-3DVAR, an incremental formulation is applied and the original bias correction scheme is maintained in the actual process of observations. The VarBC is designed to co-exist with the original scheme, because it is a dynamic revision to the observational operator on the basis of the old method, i.e., it adjusts the model state vector along with the control parameters to an unbiased state in the process of minimization and the assimilation system remains consistent with available information automatically. Preliminary experimental results show that the mean departures of background-minus-observation and analysis-minus-observation are reduced as expected. In a case study of the heavy rainfall that happened in South China on 11–13 June 2008, the 500-hPa geopotential height is better simulated using the analyzed field from the VarBC as the initial condition.     

GRAPES变分同化系统中动力平衡约束的统计求解

该文在GRAPES (Global/Regional Assimilation and PrEdiction System) 模式面三维变分 (3D_Var) 框架中引入了一种描述不同控制变量之间动力平衡约束的新方案。新方案采用统计得到的流函数和模式气压变量 ( π ) 之间的回归系数代替原方案中的线性平衡方程,来表达旋转风和质量场之间的平衡关系;采用流函数和势函数之间的回归系数,补充表达了原方案中所没有的旋转风和散度风之间的平衡关系。与原方案相比,新方案算法简单,避免了垂直方向的反复插值,减少了插值误差的引入。通过随机扰动试验和单点试验可以发现,在地转关系成立较好的区域,新方案中旋转风和质量场的耦合程度与原方案接近一致;而在地转关系不适用区域,新方案可以有效减小两者的耦合程度。此外,由于新方案中添加了旋转风和散度风之间的动力平衡约束,边界层的风场分析也更加接近大气真实状况。     

中尺度模式遗传算法同化系统及同化试验研究

设计了一个以MM5模式为基础的遗传算法同化系统,并对一次暴雨过程进行了实际降水的模拟,通过对比遗传同化和伴随同化的降水预报效果,对遗传算法同化系统的同化性能进行验证。试验结果表明,遗传算法与四维变分相结合的同化系统能有效地改善模式的初始场,使MM5模式要素预报和降水预报的准确率得到提高。     

区域四维变分资料同化的数值试验

针对中尺度数值预报模式预报误差的主要来源,尝试利用四维变分资料同化的方法来改善预报效果。在已建立的中尺度模式（ＭＭ４）四维变分资料同化系统基础上,进行了若干数值试验,通过比较同化前后的预报来检验同化的效果。这些试验中初始场、模式误差和侧边界条件被分别或同时作为控制变量来进行调整,主要探讨了模式误差和侧边界条件对同化及预报的影响,以及同时结合两者或三者的途径和方法。对两组个例分别进行的试验结果表明,区域中尺度模式预报误差除了来源于初始误差外,模式误差、侧边界条件也有不可忽视的作用。同化时应同时考虑初始场、模式误差和侧边界条件这三方面的共同作用,仅修正其中某一个或某两个会把由于其它方面造成的预报误差转嫁到它们之上,从而出现尽管目标函数下降很快而预报结果并没有相应改善的现象。     

A New Approach to Data Assimilation

A significant attempt to design a timesaving and efficient four-dimensional variational data assimilation (4DVar) has been made in this paper, and a new approach to data assimilation, which is noted as 'three-dimensional variational data assimilation of mapped observation (3DVM)' is proposed, based on the new concept of mapped observation and the new idea of backward 4DVar. Like the available 4DVar, 3DVM produces an optimal initial condition (IC) that is consistent with the prediction model due to the inclusion of model constraints and best fits the observations in the assimilation window through the model solution trajectory. Different from the 4DVar, the IC derived from 3DVM is located at the end of the assimilation window rather than at the beginning conventionally. This change greatly reduces the computing cost for the new approach, which is almost the same as that of the three-dimensional variational data assimilation (3DVar). Especially, such a change is able to improve assimilation accuracy because it does not need the tangential linear and adjoint approximations to calculate the gradient of cost function. Therefore, in numerical test, the new approach produces better IC than 4DVar does for 72-h simulation of TY9914 (Dan), by assimilating the three-dimensional fields of temperature and wind retrieved from the Advanced Microwave Sounding Unit-A (AMSU-A) observations. Meanwhile, it takes only 1/7 of the computing cost that the 4DVar requires for the same initialization with the same retrieved data.