基于BP神经网络的单站总云量预报研究

云是能够对飞行活动产生影响、甚至危及飞行安全的气象要素之一.为了对北京单站总云量进行预报,利用人工神经网络的BP模型,针对不同时次进行关键因子的选取,尝试用多种预报因子的组合,建立了总云量预报模型.试验结果表明:所建立的模型具有较好的拟合、预报精度,并且没有出现"过拟合"现象,对新样本具有较好的泛化能力.     

支持向量机方法在单站降水预报中的应用探讨

将武汉天空云量预报的81个预报因子运用到该站中等以上强度的降水预报中,基于SVM方法进行了交叉验证和预报试验。结果表明用81个预报因子建立的5～9月和全样本的降水预报模型有较好稳定性、且对降水都有正的预报技巧。因此天空云量的预报因子可以用来做降水的预报因子,同时也证明了这些预报因子在天空云量和降水预报中是协调的。SVM方法为天空云量和降水的预报提供了客观参考依据。     

支持向量机分类方法在天空云量预报中的应用

以2001年5月1日至2004年12月31日逐日武汉市地面、高空观测资料及欧洲中心24小时预报场等资料为基础,构建了不同的训练样本集,基于支持向量机方法进行了大量多因子的随机交叉验证,从而筛选出了包含最佳预报因子的训练样本集和相应的核参数g,建立了武汉市天空云量的预报模型。交叉验证结果表明预报模型是稳定性的、且具有较好的预报能力和推广应用能力。预报试验和实时预报的结果都显示出SVM方法对天空云量有一定的预报能力。     

基于支持向量机的云量精细化预报研究

基于T639数值预报产品与地面气象观测资料,以环渤海地区兴城站为例,选取与云的形成密切相关的4类预报因子——水汽类、大气不稳定度类、大气上升运动类和天气系统强度类,以总云量、低云量为预报对象,运用支持向量机,选取最佳参数,建立兴城站云量的逐月、逐时次精细化预报模型。试预报结果表明:平均预报准确率总云量为71%,低云量为69%,预报准确率较逐步回归模型有所提高;在大部分月份、时次,试预报值的变化趋势与观测值一致,可以较好地反映实际阴晴变换和云量变化;基于支持向量机的回归模型对云量有较好的预报能力。     

最小二乘支持向量机在云量预报中的应用

基于2003-2006年逐年1、8月WRF区域数值预报产品和单站观测资料,采用最小二乘支持向量机回归方法,结合选取合适的参数和核函数,分别按月通过不同长度样本序列建立了台北和厦门站总云量和低云量短期释用预报模型,利用2007年1、8月样本资料对模型进行了预报和检验,并与神经网络方法进行了对比.结果表明:最小二乘支持向量机回归方法的预报效果要好于神经网络方法;两站不同长度样本的总云量和低云量预报模型,预报效果较好,其预报准确率不会因为训练样本的减少而降低.可见,最小二乘支持向量机回归在云量等气象要素释用预报方面,具有较好的应用前景.     

河南地区降水量和云量与地面太阳短波辐射的关系

对河南省2009~2018年地面太阳短波辐射量与降水量和云量的关系进行了分析,结果表明,地面太阳短波辐射量与降水量、云量具有显著的负相关关系,但在不同季节其相关关系差异较大。为了定量表征三者之间的关系,基于多年再分析资料,利用降水量和云量与地面太阳短波辐射量的关系对日辐射量进行拟合,并对拟合后的日辐射量进行检验。验证结果显示,此考虑季节性变化的回归模型对河南省地面太阳短波辐射有较好的拟合能力,可以作为基于气象预报的日辐射量短期变化评估基础,为光伏发电量的短期预报提供参考。     

云量的时间精细化预报研究-以榆中为例

利用2001年7月至2011年7月甘肃省榆中县地面测站的每日8次云量资料和同期NCEP每日4次等压面资料,由NCEP资料构造预报因子,以总云量和低云量为预报对象,分析预报因子和预报对象的相关性,采用逐步回归方法建立榆中县逐月8个时次的云量预报方程并进行回代;并利用2012年的资料检验预报方程的预报效果。结果表明：云量主要受整层湿度、垂直运动、不稳定能量、槽强度指数和700 hPa水汽通量散度影响,其中湿度状况和垂直运动是重要因素。建立的预报方程对总云量的预报效果比低云量好;总云量平均预报误差在2成左右,低云量平均预报误差在3成左右;预报值变化趋势可以部分地反映实际云量的变化趋势。     

湖北省天空云量的特征分析及其预报

利用１９９６～１９９８年的历史气象资料统计了湖北省天空云量,接着分析了天空云量与高空物理量的关系,在此基础上使用ＭＯＳ方法建立了天空云量的预报模型。经１９９９年９～１１月试验,其初步结果表明,该预报模型可较好地反映云量变化的趋势,对阴天和晴天之间转变的趋势预报较为正确,但其在应用时仍需加强订正工作。     

Logistic判别模型在强降水预报中的应用

利用Logistiv判别模型进行强降水预报,并设计3种方案进行对比分析。方案1直接使用14个影响因子进行判别预报,受因子共线性作用及噪音信号影响,虽然拟合效果较好,但预报效果明显下降。方案2对14个影响因子进行主成分分析,利用前6个主成分建模,虽然拟合效果较方案1降低,但由于消除了因子共线性作用以及噪音信号影响,预报效果较方案1提高。方案3运用Bootstrap抽样技术得到符干样本并建模计算模型参数,打乱了原有时间序列中的波动,仪保留平稳信息,拟合自由度进一步降低,导致拟合效果较方案案2下降,但预报效果却是3种方案中最好且最稳定的。在上述研究基础上,利用欧洲中心数值预报模式的预报场资料,建立基于Logistic判别模型的强降水客观预报系统,并在中央气象台业务运行。2013和2014年连续两年汛期预报检验结果表明,概模型对强降水预报的TS评分高于数值模式本身,具有一定的业务参考价值。     

天空云量预报及支持向量机和神经网络方法比较研究

使用支持向量机和人工神经网络两种方法,分别建立了天空云量的预报模型。利用2001年5月1日～2004年12月31日的武汉市地面、高空观测值及欧洲中心的24小时预报场等资料,通过按不同比例随机抽取样本进行交叉验证的方法,分析了SVM和ANN模型的预报能力和鲁棒性;然后再用全部样本资料建立预报模型,来预报2005年1月1日～5月31日武汉市天空云量。交叉验证和实例预报的结果显示：虽然SVM和ANN模型都表现了较好的预报能力,但SVM的预报能力高于ANN方法,且在计算速度上有ANN无法比拟的优势。     

On the Formulation of Lagrangian Stochastic Models of Scalar Dispersion Within Plant Canopies

Lagrangian stochastic models, quadratic in velocity and satisfying the well-mixed condition for two-dimensional Gaussian turbulence, are used to make predictions of scalar dispersion within a model plant canopy. The non-uniqueness associated with satisfaction of the well-mixed condition is shown to be non-trivial (i.e. different models produce different predictions for scalar dispersion). The best agreement between measured and predicted mean concentrations of scalars is shown to be obtained with a small sub-class of optimal models. This sub-class of optimal models includes Thomson's model (J. Fluid Mech. 180, 529–556, 1987), the simplest model that satisfies the well-mixed condition for Gaussian turbulence, but does not include two other models identified recently as being in optimal agreement with the measured spread of tracers in a neutral boundary layer. It is therefore demonstrated that such models are not universal, i.e. applicable to a wide range of flows without readjustment of model parameters. Predictions for scalar dispersion in the model plant canopy are also obtained using the model of Flesch and Wilson (Boundary-Layer Meteorol. 61, 349–374, 1992). It is shown that, when used with a Gaussian velocity distribution or a maximum-missing-information velocity distribution, which accounts for the measured skewness and kurtosis of velocity statistics, the agreement between predictions obtained using the model of Flesch and Wilson and measurements is as good as that obtained using Thomson's model.     

An Analytical Slab Model for the Growth of the Coastal Thermal Internal Boundary Layer Under Near-Neutral Onshore Flow Conditions

Analytical parameterisations of the thermal internal boundary-layer (TIBL) height based on the slab approach are widely used in coastal dispersion models. However, they tend to a singular behaviour when the stability of the onshore flow is close to neutral. Assuming that convective turbulence dominates mixing, we derive a more general analytical model that is valid for both stable and neutral onshore flows. The model is based on the existing framework for the slab approach but involving the Zilitinkevich correction (or the spin-up term). The height variation of the onshore flow lapse rate is accounted for in the model by including an initial TIBL height. An algebraic form of the model also includes the mechanical mixing contribution to the TIBL growth and is, therefore, suitable for use when the overland surface heat flux is small and friction velocity large. The new analytical model is tested with field measurements taken under near-neutral onshore flow conditions. The performance of the model is shown to be better than a commonly used TIBL parameterisation scheme.     

Flow and transport modeling in the sea-breeze. Part I: A modifiedE — ∈ model with a non-equilibrium level 2.5 closure

In order to overcome shortcomings of the conventionalE — model and to develop a more general model applicable to the variety of atmospheric conditions observed, for example, during a land-sea breeze cycle, a modifiedE — model is proposed. The model is a simplified form of an algebraic stress model including wall proximity effects of Gibson and Launder (1978). The proposed model is similar to the popular Mellor and Yamada (1982) level 2.5 model but does not employ a local equilibrium assumption in the algebraic equations for Reynolds stresses. The resulting model shows a wider realization region under unstable conditions than the Mellor and Yamada model.The modified model is compared herein to observations, higher order closure simulations and large eddy simulations under neutral, stable and convective conditions. Various dissipation rate equations were employed and compared to understand their performance with the modified model. The modifiedE — model reproduced the observed behavior well under all conditions except near the base of an elevated inversion layer under convective conditions. The ability of the model to describe flow dynamics under a wide range of atmospheric stabilities suggests that the model can be used to describe the complicated diurnal behavior of the land-sea breeze circulation.     

包含“开关”变量的MM5四维变分资料同化系统

数值模式中的湿物理过程对于模拟各种尺度的天气现象非常重要,但它会使得模式方程的状态变量不连续,这种不连续在模式程序中表现为“开关”变量。本文对包括Kuo降水参数化方案的非静力中尺度数值模式MM5的四维变分资料同化系统进行研究,利用对实际降水的模拟,比较详细地讨论了有关的“开关”变量以及切向线性化和伴随问题。结果表明：对于一个离散的数值模式,保持“开关”变量与基态一致,用伴随码方法构造的伴随模式计算的梯度值能够为最小化过程提供较好的下降方向;Kuo方案中对流每隔一个积分步的交替发生并不影响目标函数最小化的收敛速度;“开关”变量的存在也不影响将风、温度、气压和比湿结合起来同化对MM5降水预报准确性的提高。     

An application of the E-ε turbulence model for studying coupled air-sea boundary-layer structure

An E- turbulence model is used to study air-sea interaction characteristics and turbulence structure using a coupled model for air-sea boundary layers. The E- turbulence model consists of equations for the turbulent kinetic energy, the energy-dissipation, and for the turbulent exchange coefficient expressed in terms of turbulent kinetic energy and energy-dissipation. The energy-dissipation equations for the air-sea interface are solved analytically to obtain boundary conditions for energy-dissipation at the interface. The air-sea interaction and turbulence characteristics of the E- model are compared with those of the mixing-length model and with available observations.The simulations demonstrate that the air-sea interaction parameters obtained by the E- model agree well with observations. The numerical studies also show that the E- turbulence model with appropriate constants can give good results in modeling coupled air-sea boundary-layer flows.     

A three-dimensional model for calculating the concentration distribution in inhomogeneous turbulence

A new approach for calculating the concentration distribution in inhomogeneous turbulence is suggested. The model is a 3-D model, constrained to describe incompressible flow. The model requires a knowledge of the covariance matrix of the Eulerian velocities and the two-point third moments. The model is applied for three types of turbulent field: homogeneous isotropic turbulence, constant flux neutral boundary layer and free convective turbulence. The required Eulerian moments are calculated using the eddy model of the turbulent field. Concentration moments are calculated and results are compared to experimental data. Other model predictions which have no experimental support can be compared to measurements when available.     

Coupling an ocean wave model to an atmospheric general circulation model

A coupled model, consisting of an ocean wave model and an atmospheric general circulation model (AGCM), is integrated under permanent July conditions. The wave model is forced by the AGCM wind stress, whereas the wind waves modify the AGCM surface fluxes of momentum, sensible and latent heat. We investigate the following aspects of the coupled model: how realistic are the wave fields, how strong is the coupling, and how sensitive is the atmospheric circulation to the spatially and temporally varying wave field. The wave climatology of the coupled model compares favorably with observational data. The interaction between the two models is largest (although weak) in the storm track in the Southern Hemisphere. Young windsea, which is associated with enhanced surface fluxes is generated mostly in the equatorward frontal area of an individual cyclone. However, the enhancement of the surface fluxes is too small to significantly modify the climatological mean atmospheric circulation.This paper was presented at the Second International Conference on Modelling of Global Climate Variability, held in Hamburg 7–11 September 1992 under the auspices of the Max Planck Institute for Meteorology. Guest Editor for these papers is L. Dümenil     

Validation of the small-scale performance of a climate model

A canonical correlation analysis is used to study the simulation of the relationship between small-scale precipitation patterns near the Alps and large-scale flow patterns by the ECHAM (ECMWF model, HAMburg version) climate model. The analysis is performed on a monthly mean time scale for the winter months of ECHAM1/T21, ECHAM3/T21 (improved model physics) and ECHAM3/T42 (increased resolution) simulations. The obtained patterns are compared with an identical analysis of observational data. The coarse structures of the observed relationships seem to be reasonably well simulated by the ECHAM3/T42 model version despite the simple shape of the model Alps (hereafter the model Alps), while the results for the ECHAM1/T21 and ECHAM3/T21 are not as good. This appears to be not only due to the increased resolution but also to the improved model physics, since some indication of the relationship can be found in the simulation of the ECHAM3/T21, but not in the simulation of the ECHAM1/T21 model version.Paper presented in session OA19/ST15, Simulation of climates using comprehensive global models at the XIX General Assembly of the EGS, Grenoble, 25–29 April 1994     

The Adaptation of the Atmospheric Boundary Layer to a Change in Surface Roughness

The adaptation of the atmospheric boundary layer to a change in the underlying surface roughness is an interesting problem and hence much research, theoretical, experimental, and numerical, has been undertaken. Within the atmospheric boundary layer an accurate numerical model for the turbulent properties of the atmospheric boundary layer needs to be implemented if physically realistic results are to be obtained. Here, the adaptation of the atmospheric boundary layer to a change in surface roughness is investigated using a first-order turbulence closure model, a one-and-a-half-order turbulence closure model and a second-order turbulence closure model. Perturbations to the geostrophic wind and the pressure gradients are included and it is shown that the second-order turbulence closure model, namely the standard k - model, is inferior to a lower-order closure model if a modification to limit the turbulent eddy size within the atmospheric boundary layer is not included within the model.     

A Note on k - ε Modelling of Vegetation Canopy Air-Flows

The k - turbulence model is a standard of computational software packages for engineering, yet its application to canopy turbulence has not received comparable attention. This is probably due to the additional source (and/or sink) terms, whose parameterization remained uncertain. This model must include source terms for both turbulent kinetic energy (k) and the viscous dissipation rate (), to account for vegetation wake turbulence budget. In this note, we show how Kolmogorov's relation allows for an analytical solution to be calculated within the portion of a dense and homogeneous canopy where the mixing length does not vary. By substitution within model equations, this solution allows for a set of constraints on source term model coefficients to be derived.Those constraints should meet both Reynolds averaged Navier–Stokes equationsand large-eddy simulation sub-grid scale turbulence modelling requirements.Although originating from within a limited portion of the canopy, the predictedcoefficients values must be valid elsewhere in order to make the model capable of predicting the whole canopy-layer flow with a single set of constants.