Ricker子波核支持向量回归的Mercer条件拓展问题研究

Ricker子波核支持向量回归方法是消减地震勘探记录强随机噪声的新滤波技术.用判定支持向量允许核函数的Mercer条件探讨Ricker子波核函数的有效性,经过数值计算相应的核矩阵的最小特征值,发现在一个较大范围内存在极小的负值带,数量级为10-13~10-16,且在正值带中亦存在10-13~10-15数量级的量.考虑到正负极小量值的计算误差机理相近,认为判定核函数有效性的Mercer条件在工程应用时有可能适当放宽,即核矩阵不严格半正定,接近半正定亦可.为了将Ricker子波核支持向量回归滤波方法向实际应用发展,本文对不同的理论模型的Ricker子波核滤波和小波变换滤波、自适应维纳滤波这三种技术的效果进行了比较,包括时域波形、频域振幅谱、滤波前后的信噪比以及均方误差等方面.结果表明,Ricker子波核滤波方法优于另两种方法.为实际应用Ricker子波核滤波方法奠定基础.     

TLS估计性质的进一步讨论

通过对整体最小二乘估计进行分析,给出了未知参数估计及残差的统计性质。理论上得出它们均是有偏的,并且它们之间存在一定的线性关系。受系数矩阵误差影响,导致与传统最小二乘的相关结论有所区别。通过与最小二乘估计进行比较,证明了整体最小二乘估计为膨胀型有偏估计;同时得到了整体最小二乘估计的方差和均方误差的计算公式,严格论证了其值均大于最小二乘估计的方差和均方误差,因此当设计矩阵病态时整体最小二乘估计更易受病态性的影响。最后,对整体最小二乘估计与最小二乘估计之间的不同距离进行了比较,给出了选择的判定定理。     

Statistics for Modeling Heavy Tailed Distributions in Geology: Part I. Methodology

Many natural phenomena exhibit size distributions that are power laws or power law type distributions. Power laws are specific in the sense that they can exhibit extremely long or heavy tails. The largest event in a sample from such distribution usually dominates the underlying physical or generating process (floods, earthquakes, diamond sizes and values, incomes, insurance). Often, the practitioner is faced with the difficult problem of predicting values far beyond the highest sample value and designing his system either to profit from them, or to protect against extreme quantiles. In this paper, we present a novel approach to estimating such heavy tails. The estimation of tail characteristics such as the extreme value index, extreme quantiles, and percentiles (rare events) is shown to depend primarily on the number of extreme data that are used to model the tail. Because only the most extreme data are useful for studying tails, thresholds must be selected above which the data are modeled as power laws. The mean square error (MSE) is used to select such thresholds. A semiparametric bootstrap method is developed to study estimation bias and variance and to derive confidence limits. A simulation study is performed to assess the accuracy of these confidence limits. The overall methodology is applied to the Harvard Central Moment Tensor catalog of global earthquakes.     

CALIBRATION:REGRESSION MODELS FOR CONCENTRATION PROBLEMS

A brief review of the literature on point estimators in linear calibration problems is undertaken,Supportive evidence for the relative merits of the classical and inverse regression models,drawn in generalfrom the classical inferential and Bayesian approaches,is considered and the criteria for comparison ofthe estimators are discussed with respect to their suitability for certain classes of problems.Theperformance of the estimators is assessed with respect to determining the current value of ‘x’,thepercentage concentration of administered drug levels in blood in this example.No single‘best’methodof estimation appears to hold for all values of the unknown concentration when performance is assessedby criteria based on the mean square error (MSE).However,the inverse estimator would appear to besuperior to the classical for those values of unknown X close to x.     

Pollen-inferred vegetation and environmental changes in the central Tibetan Plateau since 8200 yr BP

The ecotone between alpine steppe and meadow in the central Tibetan Plateau is sensitive to climate changes. Here we used the pollen records from three lakes in this region to reconstruct the evolution of local vegetation and climate since 8200 cal. yr BP. The history of temperature and precipitation was reconstructed quantitatively with multi-bioclimatic indexes and a transfer function from pollen records. Results show that the steppe/meadow dominated during the period of 8200–6500 cal. yr BP, especially 8200–7200 cal. yr BP, indicating the central Tibetan Plateau was controlled by strong monsoon. The steppe dominated during the periods of 6000–4900, 4400–3900, and 2800–2400 cal. yr BP. The steppe decreased gradually and the meadow expanded during the period of 4900–4400 cal. yr BP. Three century-scale drought events occurred during 5800–4900, 4400–3900 and 2800 cal. yr BP, respectively. The first time when the regional climate shifted to the present level was at 6500 cal. yr BP in the central Plateau. Since 3000 cal. yr BP, the temperature and precipitation have decreased gradually to the present level. However, the cold climate between 700–300 cal. yr BP likely corresponds to the Little Ice Age. Supported by Chinese Academy of Sciences 100 Talents Project (Grant No. 29082762), National Natural Science Foundation of China (Grant Nos. 40671196, 40372085, 49371068, 49871078), and U.S. National Science Foundation (Grant Nos. ATM-9410491, ATM-008194)     

基于GEE和U-net模型的同震滑坡识别方法

地震滑坡解译是震后重建的重要基础工作,主要通过室内人工遥感解译和室外野外调查确定。地震滑坡相比其他地物来说更为复杂,很难通过简单指数识别。室内遥感解译通过滑坡后壁、侧壁和堆积等纹理特征进行识别,大面积同震滑坡解译工作往往耗费大量人力和物力,且耗时长,难以满足灾害应急需求。本研究利用U-net神经网络模型,结合Google Earth Engine（GEE）云平台和人工智能学习平台Tensorflow,以地震局解译的汶川滑坡作为样本数据,以震后30 m分辨率的Landsat影像、高程、坡度以及NDVI数据作为模型输入参数,自动识别并获取了汶川地震后的同震滑坡数据,同时比较了不同参数组合情况下U-net神经网络模型的分割识别精度。研究表明：① U-net模型可以用于以Landsat影像为基础数据的同震滑坡快速自动识别;② 随着高程、坡度以及NDVI等输入参数增加,模型分割精度在逐渐提高,但假阳性结果也会出现增多,震后滑坡影像+高程+坡度+NDVI的输入参数组合精度最高;③ 在细节上,模型在多参数组合的情况下,大型滑坡能够很好被识别,一些较小型的滑坡受制于影像分辨率的影响,分割精度较差。为了更好识别小型滑坡,后续研究可能需提高影像的分辨率。此外,GEE云平台大大提高了训练样本获取的效率,为科研人员快速进行基于神经网络与遥感数据的地物识别研究提供了条件。     

Strength failure of spatial reticulated structures under multi-support excitation

Under strong earthquakes, long-span spatial latticed structures may collapse due to dynamic instability or strength failure. The elasto-plastic dynamic behaviors of three spatial latticed structures, including two double-layer cylindrical shells and one spherical shell constructed for the 2008 Olympic Games in Beijing, were quantitatively examined under multi-support excitation (MSE) and uniform support excitation (USE). In the numerical analyses, several important parameters were investigated such as the peak acceleration and displacement responses at key joints, the number and distribution of plastic members, and the deformation of the shell at the moment of collapse. Analysis results reveal the features and the failure mechanism of the spatial latticed structures under MSE and USE. In both scenarios, the double-layer reticulated shell collapses in the "overflow" mode, and the collapse is governed by the number of invalid plastic members rather than the total number of plastic members, beginning with damage to some of the local regions near the supports. By comparing the numbers and distributions of the plastic members under MSE to those under USE, it was observed that the plastic members spread more sufficiently and the internal forces are more uniform under MSE, especially in cases of lower apparent velocities in soils. Due to the effects of pseudo-static displacement, the stresses in the members near the supports under MSE are higher than those under USE.     

基于小波变换的改进阈值函数自适应去噪方法

在多分辨分析小波阈值去噪方法的基础上[3],用一种改进的阈值函数和自适应阈值选取算法相结合的方法,克服了硬阈值法不连续性和软阈值法有偏差的缺点。该方法能自动跟踪噪声,不同尺度自适应采用不同的阈值,可有效去除每一尺度上的噪声,保留有用信号,提高信噪比。仿真实验和地震资料处理结果表明,该方法去噪效果明显,可在各类消除随机噪声的信号处理中发挥作用。     

莫尔圆法表征坐标中误差特性几何意义的探讨

阐述用方差协方差数据绘制莫尔圆，以及用莫尔圆表示点位误差椭圆各元素的方法。对用莫尔圆表征坐标中误差特性的几何意义作了探讨，取得了概念清楚、直观的效果。