带有变量中误差非正态模型中的Box—Cox变换

Ｂｏｘ－Ｃｏｘ变换主要用于得到近似正态分布的随机变量，线性模型中的响应（因）变量Ｙ（或经过变换的Ｙ）通常建立在精确测量无误差的独立变量集上，本文研究独立变量中的测量误差（即变量中误差）对Ｂｏｘ－Ｃｏｘ变换中参数估计影响，研究结果表明，忽略变量中的误差将产生变换参数估计的偏离，该结论由分析论证和随机模拟得以支持。     

三维变分同化中多变量平衡约束设计

在现代变分同化系统中,背景场误差协方差起着决定观测信息的空间分布特征、匹配不同变量间的关系和保证分析增量平衡的作用。基于NMC(National Meteorology Centre)方法,设计了一个新的多变量平衡约束算子:在物理变换中,构建相对湿度和其余控制变量间的平衡约束算子;同时分别采用经验正交函数方法和递归滤波器来模拟控制变量(ψ,χ_u,T_u,h_ru,p_su)^T的垂直误差协方差和水平误差协方差。利用2009年6月2日到8月9日间WRF模式的预报差值场,对新的背景场误差协方差进行模拟分析。单点观测试验表明,新的背景场误差协方差实现了观测信息在干湿变量之间的传递,而且相对湿度具有与温度相似的增量场分布。     

矩阵论中的图论匹配法

G.Birkhoff用代数的方法证明了如果一个矩阵是双随机矩阵,则它能表示成置换矩阵的凸线性组合.设G是具有两分类(X,Y)的二部图,则G中含有饱和X中的所有顶点的匹配M的充分必要条件为:对∀S⊆X,有d_G(S)≥|S|.文章借助上述二部图的匹配思想,给出这一结论的图论证明.     

基于卷积神经网络的长江流域夏季日最高温度延伸期预报方法研究

长江流域是我国夏季高温热浪灾害的多发区之一,该地区日最高温度(T_max)具有显著的低频(10～30 d和30～60 d周期)变化特征,超前-滞后相关分析和气温方程诊断的结果显示,影响长江流域T_max低频变化的大尺度环流/对流信号包含:自欧亚大陆东移南下的低频波列,自东北亚向西南方向传播的异常环流,以及由西太平洋向东亚传播的低频对流;这些低频对流/环流活动通过改变辐射加热过程及绝热过程,导致长江流域T_max的低频变化。为了客观且有效地辨识和捕捉这些先兆信号,并考虑长江流域T_max与大尺度因子间的非线性作用,本文采用机器学习方法中的卷积神经网络(Convolutional Neural Network,CNN)对大量历史数据进行训练,并构建了长江流域T_max的延伸期预报模型。在独立预报阶段,CNN预报模型对长江流域区域平均T_max的预报时效达30 d,提前5～30 d预报的T_max与观测T_max的时间相关系数介于0.63～0.70(通过99％置信度的显著性检验),量级偏差(均方根误差)小于1个标准差,显示出CNN在延伸期灾害天气预报的应用潜力。     

GSI-3DVAR背景误差协方差水平相关特征尺度对地面自动气象站资料同化的适应性研究

基于华东地区3 km分辨率WRF (Weather Research and Forecasting) 模式和高密度地面自动气象站(AWS)观测,研究GSI-3DVAR同化系统的R_HZSCL对AWS观测的地面温度和风观测同化的敏感性。结果表明:运用GSI-3DVAR同化地面AWS观测时,R_HZSCL的取值较为敏感;选取合适的R_HZSCL能有效改进地面分析场精度,相较于背景场地面温度和地面矢量风差(VWD) RMSE均可减小35％以上。当R_HZSCL过大会导致温度高、低值中心的影响范围过大,风分析场较为平滑,无法反映出中小尺度环流结构。但R_HZSCL过小则会使得温度分析场增加误差,并导致风分析场出现虚假大风。观测密度稀疏化的敏感性试验结果表明,地面温度场及风场所适应的最优R_HZSCL皆随着观测密度的增大而相应减小。     

暴雨发生前AIRS卫星资料适用性检验研究

利用AIRS卫星产品中的气温和水汽资料,计算出K指数(I_K)和沙氏指数(I_S)这两种大气不稳定指数。对暴雨发生前6 h左右这两种大气不稳定指数进行统计分析。统计结果表明:在暴雨发生前6 h左右,80%左右的暴雨发生在I_K >27.5℃或I_S <3℃的情况下。由于I_K的分布与暴雨发生的频率基本呈现出较为明显的递增变化,因此I_K与I_S相比能更好地反映暴雨天气的发生。为了更好地描述暴雨发生前的大气不稳定特征,将AIRS计算的I_K和I_S做了适当的结合,得到KS指数(I_KS)并将其运用到一次暴雨个例进行验证,从验证效果来看:暴雨发生的区域在6 h前基本都处在了I_KS较高的情况下,I_KS对暴雨具有一定的指示意义。通过对暴雨区域像元中AIRS反演的气温和水汽误差分析中可以得出:AIRS计算的I_K和I_S误差主要由AIRS在有效云量较高时850hPa高度上反演的气温以及700hPa和850hPa高度上反演的水汽的误差导致的。     

不确定奇异系统的时滞相关鲁棒H∞滤波器设计

为了设计出全阶的滤波器,同时使得滤波误差动态系统是正则的、无脉冲的,并且满足一定的H_∞性能指标,研究了连续奇异系统的全阶H_∞滤波器的设计问题.利用Lyapunov-Krasovskii泛函及二次型的积分不等式方法获得了滤波误差动态系统的H_∞性能时滞相关的判据,给出了奇异系统的鲁棒H_∞滤波器存在的时滞依赖的充分条件.最后的数值例子说明了该方法的有效性.     

新型高效吸水材料(γ-PGA)的农业应用研究初报

对由南京化工大学制药与生命科学学院新研制的生物聚合高效吸水材料——γ-聚谷氨酸(γ-PGA)在农业上的应用前景进行了初步研究。试验结果表明,γ-PGA两次重复的平均吸水率可达到1108.4倍,比目前市售的聚丙烯酸盐类吸水树脂高2.85倍。γ-PGA对土壤水分的吸收率为30～80倍。γ-PGA的水浸液在土壤中有较强的保水力和较理想的释放效果,具有明显的抗旱促苗效应。用不同浓度PEG(6000)试剂(聚乙二醇)模拟土壤吸水力的种子发芽试验表明,γ-PGA有较强的吸水和保水能力,可明显提高小麦和玉米的发芽率。γ-PGA的直接拌种试验和γ-PGA水浸液的鲜切花保鲜试验对提高植物种子发芽率和延长保鲜期都有显著效果。为进一步发挥γ-PGA在农业上的应用价值,需要进一步开展植物-土壤-γ-PGA的水分平衡模型研究。     

HⅢ(2n)中元素的一种标准化表示

给出HⅢ(2n)中元素的一种标准化表示,HⅢ(2n)={Z∈C^2n×2n|1/2(Z+Z'')>0,ZJ=JZ''},J=0 I_n-I_n 0,1/2(Z+Z)>0表示矩阵1/2(Z+z'')是正定的。     

具有遗忘时滞的静态神经网络的H∞状态估计

本文研究了具有遗忘时滞的静态神经网络的H_∞状态估计问题.首先降低了时变时滞可微的条件,然后通过构造合适的Lyapunov-Krasovskii泛函,设计保H_∞性能的状态估计器,使得误差系统实现全局渐近稳定.最后,借助Matlab中线性矩阵不等式工具箱进行数值仿真,验证了结论的有效性.     

Detection Time for Plausible Changes in Annual Precipitation, Evapotranspiration, and Streamflow in Three Mississippi River Sub-Basins

We use diagnostic studies of off-line variable infiltration capacity (VIC) model simulations of terrestrial water budgets and 21st-century climate change simulations using the parallel climate model (PCM) to estimate the time required to detect predicted changes in annual precipitation (P), evapotranspiration (E), and discharge (Q) in three sub-basins of the Mississippi River Basin. Time series lengths on the order of 50–350 years are required to detect plausible P, E, and Q trends in the Missouri, Ohio, and Upper Mississippi River basins. Approximately 80–160, 50, and 140–350 years, respectively, are needed to detect the predicted P, E, and Q trends with a high degree of statistical confidence. These detection time estimates are based on conservative statistical criteria (α = 0.05 and β = 0.10) associated with low probability of both detecting a trend when it is not occurring (Type I error) and not detecting a trend when it is occurring (Type II error). The long detection times suggest that global-warming-induced changes in annual basin-wide hydro-climatic variables that may already be occurring in the three basins probably cannot yet be detected at this level of confidence. Furthermore, changes for some variables that may occur within the 21st century might not be detectable for many decades or until the following century – this may or may not be the case for individual recording station data. The long detection times for streamflow result from comparatively low signal-to-noise ratios in the annual time series. Finally, initial estimates suggest that faster detection of acceleration in the hydrological cycle may be possible using seasonal time series of appropriate hydro-climatic variables, rather than annual time series.     

Radar Quantitative Precipitation Estimation Based on the Gated Recurrent Unit Neural Network and Echo-Top Data

The Gated Recurrent Unit(GRU) neural network has great potential in estimating and predicting a variable. In addition to radar reflectivity(Z), radar echo-top height(ET) is also a good indicator of rainfall rate(R). In this study, we propose a new method, GRU＿Z-ET, by introducing Z and ET as two independent variables into the GRU neural network to conduct the quantitative single-polarization radar precipitation estimation. The performance of GRU＿Z-ET is compared with that of the other three meth...     

Outlier Problem in Evaluating Similarity Functions in the Stable Atmospheric Boundary Layer

The gradient-based similarity approach removes turbulent fluxes as governing parameters and replaces them with vertical gradients of mean wind speed and potential temperature. As a result, the gradient Richardson number, Ri, appears as a stability parameter instead of the Monin–Obukhov stability parameter z/L (L is the Obukhov length). The gradient-based scaling is more appropriate for moderate and very stable conditions when the gradients are large and their errors are relatively small whereas z/L becomes ambiguous in these conditions because turbulent fluxes are small. However, the gradient-based formulation is faced with a problem related to the influence of Ri outliers: outliers with high values of Ri can exist in conditions that are really near-neutral. These outliers are mapped into the very stable range in plots in which Ri is the independent variable and may lead to spurious dependencies for bin-averaged data (spurious bin-averaging). This effect is quite large for functions that are steep for the gradient-based scaling. The present study uses the Surface Heat Budget of the Arctic Ocean (SHEBA) data to examine the problem and proposes two methods, conditional analysis and independent binning, to limit the influence of outliers on bin-averaging. A disadvantage of the conditional analysis is associated with eliminating outliers based on criteria that could be considered as subjective. The independent bin-averaging method does not have this disadvantage, but the scatter of the bin-averaged points is higher than for the conditional analysis, rendering data analysis and interpretation difficult.     

系统辨识(5):迭代搜索原理与辨识方法

递推辨识与迭代辨识构成了两类重要的参数估计方法.递推辨识的递推变量与时间有关,因而可以用于在线估计系统参数;迭代辨识的迭代变量是自然数,与客观世界的时间无关,通常用于离线估计系统参数.基于辅助模型辨识思想、多新息辨识理论、递阶辨识原理、耦合辨识概念等辨识方法都可以用递推算法和迭代算法实现.迭代方法渊源很早,如求解矩阵方程Ax=b的雅可比迭代、高斯-赛德尔迭代等.迭代辨识方法主要使用梯度搜索、最小二乘搜索、牛顿搜索原理来实现.为此主要研究了CARMA系统和Box-Jenkins系统的最小二乘迭代辨识方法与梯度迭代辨识方法.这些方法也可推广到其他所有方程误差类系统和输出误差类系统,以及非线性系统.迭代辨识方法通常用于有限量测数据的系统辨识,其收敛性证明是辨识领域极具挑战性的研究课题.     

Analysis of the Systematic Errors Found in the Kipp &; Zonen Large-Aperture Scintillometer

Studies have shown a systematic error in the Kipp & Zonen large-aperture scintillometer (K&ZLAS) measurements of the sensible heat flux, H. We improved on these studies and compared four K&ZLASs with a Wageningen large-aperture scintillometer at the Chilbolton Observatory. The scintillometers were installed such that their footprints were the same and independent flux measurements were made along the measurement path. This allowed us to compare H and the direct scintillometer output, the refractive index structure parameter, C_n²{C_{n}^{2}} . Furthermore, spectral analysis was performed on the raw scintillometer signal to investigate the characteristics of the error. Firstly, correlation coefficients ≥ 0.99 confirm the robustness of the scintillometer method, and secondly we discovered two systematic errors: the low-C_n²{C_{n}^{2}} error and the high-C_n²{C_{n}^{2}} error. The low-C_n²{C_{n}^{2}} error is a non-linear error that is caused by high-frequency noise, and we suspect the error to be caused by the calibration circuit in the receiver. It varies between each K&ZLAS, is significant for H ≤ 50 W m⁻², and we propose a solution to remove this error using the demodulated signal. The high-C_n²{C_{n}^{2}} error identified by us is the systematic error found in previous studies. We suspect this error to be caused by poor focal alignment of the receiver detector and the transmitter light-emitting diode that causes ineffective use of the Fresnel lens in the current Kipp & Zonen design. It varies between each K&ZLAS (35% up to 240%) and can only be removed by comparing with a reference scintillometer in the field.     

Retrieving daily global solar radiation from routine climate variables

Solar radiation is an important variable for studies related to solar energy applications, meteorology, climatology, hydrology, and agricultural meteorology. However, solar radiation is not routinely measured at meteorological stations; therefore, it is often required to estimate it using other techniques such as retrieving from satellite data or estimating using other geophysical variables. Over the years, many models have been developed to estimate solar radiation from other geophysical variables such as temperature, rainfall, and sunshine duration. The aim of this study was to evaluate six of these models using data measured at four independent worldwide networks. The dataset included 13 stations from Australia, 25 stations from Germany, 12 stations from Saudi Arabia, and 48 stations from the USA. The models require either sunshine duration hours (Ångstrom) or daily range of air temperature (Bristow and Campbell, Donatelli and Bellocchi, Donatelli and Campbell, Hargreaves, and Hargreaves and Samani) as input. According to the statistical parameters, Ångstrom and Bristow and Campbell indicated a better performance than the other models. The bias and root mean square error for the Ångstrom model were less than 0.25 MJ m² day^?1 and 2.25 MJ m² day^?1, respectively, and the correlation coefficient was always greater than 95 %. Statistical analysis using Student’s t test indicated that the residuals for Ångstrom, Bristow and Campbell, Hargreaves, and Hargreaves and Samani are not statistically significant at the 5 % level. In other words, the estimated values by these models are statistically consistent with the measured data. Overall, given the simplicity and performance, the Ångstrom model is the best choice for estimating solar radiation when sunshine duration measurements are available; otherwise, Bristow and Campbell can be used to estimate solar radiation using daily range of air temperature.     

CCA与SVD分析方法比较研究

文中从理论分析、方法比较以及实例计算几方面,对目前气象资料处理分析中常用的CCA与SVD分析方法进行了比较,结果表明(1) 对同样的变量组X,Y分别用CCA和SVD方法进行相关分析,得到了完全不同的分析结果,CCA所得到的相关就是原始变量组之间的相关;SVD所得到的相关是组合变量L,M5"BZ〗间的相关而不是原始变量组X,Y之间的相关.理论分析和实例计算都表明,两种方法分析得到的最大相关有非常显著的差别,CCA明显比SVD要大得多,且CCA收敛快而SVD收敛慢.所以SVD不能有效地提取两组变量或两个变量场之间相关关系的主要特征,只有CCA才能最大限度地提取它们之间相关关系的主要特征.(2) CCA所得出的两变量组的变量是独立正交变量,所以通过分析CCA组合变量间的相关来表示原变量组间的相关关系是有意义的.而SVD所得到的两变量组的变量不具有独立性和正交性,信息提供重复,存在共线性,所以通过分析SVD组合变量来表示两变量组的相关关系没有CCA方法有意义.(3)CCA是在考虑了各个变量场自身变化的情况下来分解两个变量场间的关系;而SVD是在没有考虑各个变量场自身变化的情况下来分解两个变量场间的协方差关系.很显然,CCA比SVD更全面、更完整和更准确.(4)凡用SVD方法分析得到的结论,由于总可以重新应用CCA方法找到相关更好的不同结果,所以有值得进一步深讨的必要.     

Footprint estimation for scalar flux measurements in the atmospheric surface layer

The flux footprint is the contribution, per unit emission, of each element of a surface area source to the vertical scalar flux measured at height z _m ; it is equal to the vertical flux from a unit surface point source. The dependence of the flux footprint on crosswind location is shown to be identical to the crosswind concentration distribution for a unit surface point source; an analytic dispersion model is used to estimate the crosswind-integrated flux footprint. Based on the analytic dispersion model, a normalized crosswind-integrated footprint is proposed that principally depends on the single variable z/z _m , where z is a measure of vertical dispersion from a surface source. The explicit dependence of the crosswind-integrated flux footprint on downwind distance, thermal stability and surface roughness is contained in the dependence of z on these variables. By also calculating the flux footprint with a Lagrangian stochastic dispersion model, it is shown that the normalized flux footprint is insensitive to the analytic model assumption of a self-similar vertical concentration profile.The National Center for Atmospheric Research is funded by the National Science Foundation.     

The Albedo of Vegetated Land Surfaces: Systems Analysisand Mathematical Modeling

Summary The albedo of vegetated land surfaces (surface albedo) is a key factor in climate modeling and in mechanistic accounting of many ecological processes. This paper proposes a testable numerical equation for the analysis and projection of surface albedo. Conceptualized as the manifestation of a canopy elements-determined basic property after modifications by temporal and spatial circumstances, surface albedo was depicted analytically in relation to 11 driving variables (leaf size, leaf life span, relative leaf age, canopy leaf cover, relative stem cover, vegetation height, stress-calendar day, drought indicator, optical air mass, station atmospheric pressure, snow cover). With peripheral algorithms developed to derive all but two of those variables, surface albedo was linked ultimately to eight rudimentary factors (calendar day, latitude, elevation, vegetation height, dominant plant species, monthly air temperature, monthly precipitation, snow cover). The analytical framework, and then its coefficient values, for surface albedo were generally supported by a series of statistical evaluations in terms of: (i) the equation’s ability to capture, by regression fitting, the variation in the surface albedo of 26 forests (135 data points) distributed around the world; (ii) the quantitative significance of individual driving variables; (iii) the randomness of residual or error distributions; (iv) the performance of the forests-fitted equation in extrapolative prediction of surface albedo against independent data for 8 deforested sites (93 data points) and for 3 types of vegetation (7 data points) at the Arctic treeline. Compared to the data, the fitted or projected albedo values had a margin of error generally within ±10%. The individual coefficient values and component functions of the final equation were consistent with their supposed mechanistic underpinnings, based on independent information from the literature. The equation shed new insight into the quantitative behavior of surface albedo, and upon further validation, should be useful for modeling surface albedo as a key land surface-atmosphere feedback link that varies and interacts with climate and vegetation. Received August 18, 1997     

Monthly precipitation mapping of the Iberian Peninsula using spatial interpolation tools implemented in a Geographic Information System

Summary In this study, spatial interpolation techniques have been applied to develop an objective climatic cartography of precipitation in the Iberian Peninsula (583,551 km²). The resulting maps have a 200 m spatial resolution and a monthly temporal resolution. Multiple regression, combined with a residual correction method, has been used to interpolate the observed data collected from the meteorological stations. This method is attractive as it takes into account geographic information (independent variables) to interpolate the climatic data (dependent variable). Several models have been developed using different independent variables, applying several interpolation techniques and grouping the observed data into different subsets (drainage basin models) or into a single set (global model). Each map is provided with its associated accuracy, which is obtained through a simple regression between independent observed data and predicted values. This validation has shown that the most accurate results are obtained when using the global model with multiple regression mixed with the splines interpolation of the residuals. In this optimum case, the average R ² (mean of all the months) is 0.85. The entire process has been implemented in a GIS (Geographic Information System) which has greatly facilitated the filtering, querying, mapping and distributing of the final cartography.