基于无导数优化方法的数值模式误差估计

初始场误差和模式误差是制约数值预报准确率的两个关键因素,本文主要考虑利用历史观测资料实现时空演变的模式误差的估计问题.通过把模式误差综合考虑成为准确模式中的未知项,把历史资料看作是带有未知项的准确模式的特解,构造了求解时空演变的模式误差项的反问题及其最优控制问题.给出了一个解决最优控制问题的无导数优化方法,该方法的优点是不需要建立原数值模式的切线性模式与伴随模式,它只需在增加一个外强迫项的基础上运行原数值模式即可实现模式误差项的最优估计.关于Burgers方程的算例表明,无论模式的初始状态是否准确已知,无导数优化方法都能有效解决时空演变的模式误差的最优估计问题,它为实际业务模式利用历史数据提取模式误差信息并显著地改进预报效果提供了一种方便可行的数值方法与理论依据.     

吸收散射性三维矩形介质内辐射源项的反问题

提出了一种由边界出射辐射强度反演吸收散射性三维矩形介质内辐射源项分布的方法。该方法是在辐射传递方程离散坐标近似的基础上,用求目标函数极小值的共轭梯度法进行反演计算。通过对介质辐射特性、光学厚度等参数对反演精度影响的分析,结果表明,即使存在测量误差,本文所提出的方法可较精确地反演辐射源项。     

Computation of the analysis error covariance in variational data assimilation problems with nonlinear dynamics

The problem of variational data assimilation for a nonlinear evolution model is formulated as an optimal control problem to find the initial condition function. The data contain errors (observation and background errors), hence there will be errors in the optimal solution. For mildly nonlinear dynamics, the covariance matrix of the optimal solution error can often be approximated by the inverse Hessian of the cost functional. Here we focus on highly nonlinear dynamics, in which case this approximation may not be valid. The equation relating the optimal solution error and the errors of the input data is used to construct an approximation of the optimal solution error covariance. Two new methods for computing this covariance are presented: the fully nonlinear ensemble method with sampling error compensation and the ‘effective inverse Hessian’ method. The second method relies on the efficient computation of the inverse Hessian by the quasi-Newton BFGS method with preconditioning. Numerical examples are presented for the model governed by Burgers equation with a nonlinear viscous term.     

Inverse radiation problem of temperature distribution in one-dimensional isotropically scattering participating slab with variable refractive index

In this paper, an inverse analysis is performed for estimation of source term distribution from the measured exit radiation intensities at the boundary surfaces in a one-dimensional absorbing, emitting and isotropically scattering medium between two parallel plates with variable refractive index. The variation of refractive index is assumed to be linear. The radiative transfer equation is solved by the constant quadrature discrete ordinate method. The inverse problem is formulated as an optimization problem for minimizing an objective function which is expressed as the sum of square deviations between measured and estimated exit radiation intensities at boundary surfaces. The conjugate gradient method is used to solve the inverse problem through an iterative procedure. The effects of various variables on source estimation are investigated such as type of source function, errors in the measured data and system parameters, gradient of refractive index across the medium, optical thickness, single scattering albedo and boundary emissivities. The results show that in the case of noisy input data, variation of system parameters may affect the inverse solution, especially at high error values in the measured data. The error in measured data plays more important role than the error in radiative system parameters except the refractive index distribution; however the accuracy of source estimation is very sensitive toward error in refractive index distribution. Therefore, refractive index distribution and measured exit intensities should be measured accurately with a limited error bound, in order to have an accurate estimation of source term in a graded index medium.     

On optimal solution error covariances in variational data assimilation problems

The problem of variational data assimilation for a nonlinear evolution model is formulated as an optimal control problem to find unknown parameters such as distributed model coefficients or boundary conditions. The equation for the optimal solution error is derived through the errors of the input data (background and observation errors), and the optimal solution error covariance operator through the input data error covariance operators, respectively. The quasi-Newton BFGS algorithm is adapted to construct the covariance matrix of the optimal solution error using the inverse Hessian of an auxiliary data assimilation problem based on the tangent linear model constraints. Preconditioning is applied to reduce the number of iterations required by the BFGS algorithm to build a quasi-Newton approximation of the inverse Hessian. Numerical examples are presented for the one-dimensional convection–diffusion model.     

Upper bound of errors in solving the inverse problem of identifying a voice source

The paper considers the inverse problem of finding the shape of a voice-source pulse from a specified segment of a speech signal using a special mathematical model that relates these quantities. A variational method for solving the formulated inverse problem for two new parametric classes of sources is proposed: a piecewise-linear source and an A-source. The error in the obtained approximate solutions of the inverse problem is considered, and a technique to numerically estimate this error is proposed, which is based on the theory of a posteriori estimates of the accuracy in solving ill-posed problems. A computer study of the adequacy of the proposed models of sources, and a study of the a posteriori estimates of the accuracy in solving inverse problems for such sources were performed using various types of voice signals. Numerical experiments for speech signals showed satisfactory properties of such a posteriori estimates, which represent the upper bounds of possible errors in solving the inverse problem. The estimate of the most probable error in determining the source-pulse shapes for the investigated speech material is on average ~7%. It is noted that the a posteriori accuracy estimates can be used as a criterion for the quality of determining the voice-source pulse shape in the speaker-identification problem.     

A new inverse approach for the equivalent gray radiative property of a non-gray medium using a modified zonal method and the complex-variable-differentiation method

Non-gray radiative properties of an absorbing, emitting, non-gray participating medium significantly increase the difficulty of solving the radiative transfer equation. This paper presents a new inverse approach for the equivalent gray radiative property of a non-gray medium. In this approach, the unknown equivalent gray radiative properties are treated as the optimization variables, and the errors to be minimized are the differences between the calculated temperatures and the measured ones. The measured data are simulated by solving the direct problem, in which a modified zonal method together with the Edwards exponential wide-band model is employed. In the inverse problem, the sensitivity coefficients are first calculated by the complex-variable-differentiation method, and then the least-square method and the Newton-Raphson iterative method are employed to minimize the target function. The effectiveness and efficiency of the inverse problem are demonstrated in an example, and another case is given to show the accuracy and potential of the proposed algorithm. The effects of the measurement error and the number of measurement points on the accuracy of the inverse analysis are also investigated in detail.     

Analysis of the accuracy of the inverse problem solution for a differential heterodyne microscope as applied to rectangular plasmonic waveguides

The existence, uniqueness, and stability of the inverse problem solution for a scanning differential heterodyne microscope as applied to rectangular plasmonic waveguides have been analyzed. The consideration is based on an algorithm using a trial-and-error method that we proposed previously to characterize plasmonic waveguides with a triangular profile. The error of the inverse problem (IP) solution is calculated as dependent on the initial data and with allowance for their errors. Instability domains are found for the IP solution, where the solution error sharply increases. It is shown that the instability domains can be eliminated and the accuracy of the IP solution can be significantly improved in the entire range of initial data by taking initial data in the form of two phase responses of the microscope at different wavelengths.     

Numerical determination of the density profile of an inhomogeneous membrane: solution of the inverse vibration problem

Given one or more vibrational modes of a membrane, the free vibration equation can be applied to infer the mass surface density. This paper considers determining the surface density of an inhomogeneous membrane from digitized holographic projections (interferograms) of the modeshapes. Spatially discrete numerical models of the membrane surface are presented, which can be used to solve both forward and inverse vibration problems. The accuracy of the discrete models is examined for exactly solvable free vibration problems involving inhomogeneous membranes. For the solution of the inverse problem, error estimates are given for the mass surface density deduced from modeshape interferograms. The practicability of the method is investigated using simulated experimental data for membranes with composite and continuously inhomogeneous density profiles. Strategies are discussed for reducing errors in the reconstructed densities.     

Solution of the inverse radiative load problem in a two-dimensional system

An inverse radiation analysis is presented for estimating the temperature and the heat load distributions of the heating surface from the temperature and the heat flux measurements of the heated object. The Monte Carlo method is employed to solve the direct radiation problem. The inverse radiation problem is solved using the conjugate gradient and singular value decomposition methods. The measured data are simulated by adding random errors to the exact solution of the direct problem. The effects of the measurement errors on the accuracy of the inverse analysis are investigated. The study shows that the heat load distribution of the heating surface can be estimated accurately for the exact and noisy data. And the conjugate gradient method is better than the singular value decomposition method since the former can obtain more accurate results if the measurement errors are the same.     

Exploiting dynamical coherence: A geometric approach to parameter estimation in nonlinear models

Parameter estimation in nonlinear models is a common task, and one for which there is no general solution at present. In the case of linear models, the distribution of forecast errors provides a reliable guide to parameter estimation, but in nonlinear models the facts that predictability may vary with location in state space, and that the distribution of forecast errors is expected not to be Normal, means that parameter estimation based on least squares methods will result in systematic errors. A new approach to parameter estimation is presented which focuses on the geometry of trajectories of the model rather than the distribution of distances between model forecast and the observation at a given lead time. Specifically, we test a number of candidate trajectories to determine the duration for which they can shadow the observations, rather than evaluating a forecast error statistic at any specific lead time(s). This yields insights into both the parameters of the dynamical model and those of the observational noise model. The advances reported here are made possible by extracting more information from the dynamical equations, and thus improving the balance between information gleaned from the structural form of the equations and that from the observations. The technique is illustrated for both flows and maps, applied in 2-, 3-, and 8-dimensional dynamical systems, and shown to be effective in a case of incomplete observation where some components of the state are not observed at all. While the demonstration of effectiveness is strong, there remain fundamental challenges in the problem of estimating model parameters when the system that generated the observations is not a member of the model class. Parameter estimation appears ill defined in this case.     

遗传算法结合正则化方法反演海洋大气波导

针对正则化方法在解决实际反演问题时既能克服问题的不适定性又可以很大程度上抑制噪声和误差的传播, 本文提出了利用遗传算法结合正则化方法的新算法, 在遗传算法适应度函数中引入正则化项来反演波导参数; 然后对算法进行仿真试验, 结果表明新算法与传统遗传算法相比具有较高的反演精度, 并指出当噪声误差小于10%时, 算法具有较强的“去噪”性能; 最后利用机载雷达在Wallops岛探测的海表面处局部回波资料进行反演试验, 将反演结果与实测大气折射率廓线进行比较, 说明该算法的有效性. 新方法为海洋大气波导反演研究提供了一种新思路.     

Accuracy in determining voice source parameters

The paper addresses the accuracy of an approximate solution to the inverse problem of retrieving the shape of a voice source from a speech signal for a known signal-to-noise ratio (SNR). It is shown that if the source is found as a function of time with the A.N. Tikhonov regularization method, the accuracy of the found approximation is worse than the accuracy of speech signal recording by an order of magnitude. In contrast, adequate parameterization of the source ensures approximate solution accuracy comparable with the accuracy of the problem data. A corresponding algorithm is considered. On the basis of linear (in terms of data errors) estimates of approximate parametric solution accuracy, parametric models with the best accuracy can be chosen. This comparison has been carried out for the known voice source models, i.e., model [17] and the LF model [18]. The advantages of the latter are shown. Thus, for SNR = 40 dB, the relative accuracy of an approximate solution found with this algorithm is about 1% for the LF model and about 2% for model [17] as compared to an accuracy of 7–8% in the regularization method. The role of accuracy estimates found in speaker identification problems is discussed.     

Impact of nonlinearities and model error on pseudo-inverse calculations

Pseudo-inverse calculations have been made within the operational and research meteorological communities to identify components of the error in the initial state that are responsible for a significant portion of the forecast error. These calculations are based on the assumptions of a perfect model and linear perturbation growth, conditions not realizable in operational forecasting. In this study, the impact of nonlinearities and model error on pseudo-inverse calculations is investigated within an idealized framework using a simple atmospheric model. Forecasts are run within the perfect and imperfect model frameworks, with initial errors of varying sizes. Model error is introduced by changing the model dissipation terms. It is found that for pseudo-inverses composed of a small subset of the leading singular vectors (SVs), the nonlinear forecast correction is often better than the expected theoretical correction, indicating the suppression of error growth both inside and outside the linear pseudo-inverse subspace. As the size of the pseudo-inverse is increased, the nonlinear forecast correction starts to degrade. This forecast degradation coincides with a degradation in the analysis correction. It is possible to improve the forecast by degrading the analysis in the presence of model error, especially when the initial error is very small. However, for initial errors of reasonable magnitude, this is unlikely to happen in instances when the nonlinear forecast correction is better than the theoretical correction. Just as improving the initial state may suppress errors outside of the linear SV subspace, degrading it may likewise increase errors outside the SV subspace. This suggests that the size of the nonlinear correction relative to the expected theoretical correction may be useful in determining when pseudo-inverse perturbations are likely to have improved the analyses.     

小尺寸平面源的误差精确分析

基于Helmholtz方程的严格远场解和源的“误差面积”描述方法,讨论了大发散角光辐射远场振幅和相位对于源包络扰动的依赖关系,讨论了由于源振幅扰动而产生的远场误差,建立了一种恰当的描述远场误差的方法.模拟实验表明,在“相对误差面积”的基础上,可以结合利用“离轴距离”来精确描述源误差和分析远场误差,利用“判据点”可以判断离轴误差的存在.     

Impact of aerosols on the forecast accuracy of solar irradiance calculated by a numerical weather prediction model

The impact of aerosols on the forecast accuracy of solar irradiance calculated by a fine-scale, one day-ahead, and operational numerical weather prediction model (NWP) is investigated in this study. In order to investigate the impact of aerosols only, the clear sky period is chosen, which is defined as when there are no clouds in the observation data and in the forecast data at the same time. The evaluation of the forecast accuracy of the solar irradiance is done at a single observation point that is sometimes affected by aerosol events. The analysis period is one year from April 2010 to March 2011. During the clear sky period, the root mean square errors (RMSE) of the global horizontal irradiance (GHI), direct normal irradiance (DNI), and diffuse horizontal irradiance (DHI) are 40.0?W?m^?2, 84.0?Wm^?2, and 47.9?W?m^?2, respectively. During one extreme event, the RMSEs of the GHI, DNI, and DHI are 70.1?W?m^?2, 211.6?W?m^?2, and 141.7?W?m^?2, respectively. It is revealed that the extreme events were caused by aerosols such as dust or haze. In order to investigate the impact of the aerosols, the sensitivity experiments of the aerosol optical depth (AOD) for the extreme events are executed. The best result is obtained by changing the AOD to 2.5 times the original AOD. This changed AOD is consistent with the satellite observation. Thus, it is our conclusion that an accurate aerosol forecast is important for the forecast accuracy of the solar irradiance.     

含有误差校正的小波神经网络交通流量预测

交通流量的准确预测对于高速路管理者进行决策至关重要。建立了小波神经网络(WNN)交通流量预测模型,并通过预测训练误差和测试误差校正预测结果来提高预测精度。首先构建WNN模型对交通流量进行初步预测,然后利用经验模态分解(EMD)和WNN模型对训练误差和测试误差进行预测。分别用训练误差预测值、测试误差预测值和两种误差预测值的加权对流量初步预测结果进行修正得到最终预测值。采用四川省成灌高速路交通流量数据进行了仿真对比实验,仿真结果表明含有误差校正的小波神经网络模型能有效提高交通流量预测精度,并且利用两种误差加权修正模型的预测精度高于利用测试误差的修正模型和利用训练误差的修正模型。     

Recovering doping profiles in semiconductor devices with the Boltzmann–Poisson model

We investigate numerically an inverse problem related to the Boltzmann–Poisson system of equations for transport of electrons in semiconductor devices. The objective of the (ill-posed) inverse problem is to recover the doping profile of a device, presented as a source function in the mathematical model, from its current–voltage characteristics. To reduce the degree of ill-posedness of the inverse problem, we proposed to parameterize the unknown doping profile function to limit the number of unknowns in the inverse problem. We showed by numerical examples that the reconstruction of a few low moments of the doping profile is possible when relatively accurate time-dependent or time-independent measurements are available, even though the later reconstruction is less accurate than the former. We also compare reconstructions from the Boltzmann–Poisson (BP) model to those from the classical drift–diffusion-Poisson (DDP) model, assuming that measurements are generated with the BP model. We show that the two type of reconstructions can be significantly different in regimes where drift–diffusion-Poisson equation fails to model the physics accurately. However, when noise presented in measured data is high, no difference in the reconstructions can be observed.     

Inverse radiation problem in one-dimensional slab by time-resolved reflected and transmitted signals

A time-domain inverse approach is proposed for estimating the distribution of absorbing and scattering coefficients in one-dimensional inhomogeneous media. The temporal reflected and transmitted signals are detected when an ultra-short pulse irradiates on the boundary of semi-transparent scattering media. Forward computation and inverse algorithm employ the least-squares finite element method and conjugate gradient method, respectively. As the prevalent diffusion approximation is not employed in our model, the present approach can be extended to more comprehensive application. The investigation about detected signals indicates that the reflected signals play a significant role in reconstructing optical properties; the signals in early sampling time are more important than those at long-time logarithm slope, and so, more attention should be paid to the early signals in the solution of inverse radiation problem. Three different inverse radiation problems are investigated to show the ability of the present approach to deal with the two-layer, three-layer and continuous inhomogeneous media. The effect of measured errors on the accuracy of reconstruction is investigated by adding artificial random errors. The results indicate that accurate reconstruction depends on not only precise numerical simulation but also quality of detected data.     

基于Lorenz模型的集合预报与单一预报的比较研究

根据非线性局部Lyapunov向量方法和增长模繁殖方法,选取Lorenz63模型和Lorenz96模型的不同状态为例,对集合预报与单一预报的预报技巧开展了对比研究.结果表明:与单一预报比较,集合预报的均方根误差和型异常相关有明显改善,随预报时间推移,改善效果越显著,且集合平均优于单一预报的实验个例数逐渐增多.就概率分布(f)而言,单一预报状态的f与真实状态基本一致,不随时间变化;而集合平均预报状态的f则随时间呈现出值域变窄、峰值变大的特点.表明随预报时间的延长,单一预报状态为混沌吸引子上的随机状态,而集合平均预报状态为吸引子子集上的随机状态,这可能是集合平均误差小于单一预报的原因.