On selection of kernel parametes in relevance vector machines for hydrologic applications

Recent advances in statistical learning theory have yielded tools that are improving our capabilities for analyzing large and complex datasets. Among such tools, relevance vector machines (RVMs) are finding increasing applications in hydrology because of (1) their excellent generalization properties, and (2) the probabilistic interpretation associated with this technique that yields prediction uncertainty. RVMs combine the strengths of kernel-based methods and Bayesian theory to establish relationships between a set of input vectors and a desired output. However, a bias–variance analysis of RVM estimates revealed that a careful selection of kernel parameters is of paramount importance for achieving good performance from RVMs. In this study, several analytic methods are presented for selection of kernel parameters. These methods rely on structural properties of the data rather than expensive re-sampling approaches commonly used in RVM applications. An analytical expression for prediction risk in leave-one-out cross validation is derived. For brevity, the effectiveness of the proposed methods is assessed first by data generated from the benchmark sinc function, followed by an example involving estimation of hydraulic conductivity values over a field based on observations. It is shown that a straightforward maximization of likelihood function can lead to misleading results. The proposed methods are found to yield robust estimates of parameters for kernel functions.     

3D Visualization of Landslide Deformation

The deformation of landslide is usually expressed in horizontal and vertical direction separately. Based on the X, Y, Z coor-dinates from a real monitoring project, a method to build three-dimensional(3D) model of landslide by constructing triangulated ir-regular network (TIN) and extruding contour lines is proposed. The almost imperceptible displacements of monitoring points are zoomed by two kind of exaggerating methods. The deformation process is replayed using 3D animation technique.     

GIS-based groundwater potential mapping using artificial neural network and support vector machine models: the case of Boryeong city in Korea

Groundwater productivity-potential (GPP) was analysed using the data mining models of an artificial neural network (ANN) and a support vector machine (SVM) in Boryeong city, Korea. The groundwater-productivity data with specific capacity (SPC) is strongly related to hydrogeological factors, and hence the relation may allow for groundwater potential mapping from hydrogeological factors through the ANN and SVM models. A back-propagation algorithm was used for the ANN model while a polynomial kernel was adopted for the SVM model. For the validation of the GPP maps generated from the ANN and SVM models, the area-under-the-curve analysis was performed using the SPC values of well data. The accuracies achieved from the ANN and SVM models are about 83.57 and 80.83%, respectively. It proves that the ANN and SVM models will be highly conducive to developing useful groundwater resources.     

GIS中面状要素矢量栅格化的面积误差分析

在探讨矢量栅格化的几种形式和产生误差原因的基础上,系统分析在面状要素矢量栅格化的误差分析方法研究方面所取得的进展,指出当前分析方法存在的主要问题,并提出利用结构性栅格数据进行面状要素矢量栅格化的误差分析方法,然后以土地利用数据的栅格化为例进行实证研究。研究表明,常规分析方法很大程度上低估了栅格化的误差,因为没有充分考虑栅格化误差存在空间上的"此消彼长"现象。建议采用结构性栅格数据来改善栅格化误差分析,对误差作基于格网的量化计算,并标记在单元格上。它能有效地区分"此消"与"彼长"导致的误差,使面状要素栅格化的误差分析更加全面、客观、准确,且容易进行误差的量化分析与表示。     

地球化学异常识别的两种机器学习算法之比较

在Sklearn的Python语言代码基础上,开发了基于孤独森林和一类支持向量机的多元地球化学异常识别方法程序。选择吉林省和龙地区为实验区,从1∶5万水系沉积物资料中提取地球化学异常。把实验区已知矿点的空间分布位置作为"地真"数据,绘制两种机器学习算法的ROC曲线并计算AUC值,用来对比两种方法的多元地球化学异常识别效果。研究结果表明:两种机器学习算法都能够有效识别多元地球化学异常,所提取的多元地球化学异常与已知矿点具有显著的空间关联性;孤独森林算法在数据处理耗时和多元地球化学异常识别效果方面略优于一类支持向量机。     

一种新的COSMIC大气折射率资料观测误差估计方法及在GRAPES全球三维变分同化中的应用

观测误差协方差是变分同化系统中决定分析及预报效果的关键参数之一,观测误差的估计精度直接影响变分同化分析和预报效果。分析了新息增量法(H-L法)估计全球定位系统无线电掩星这类观测点不固定资料的观测误差的适用条件,并利用1年的气象、电离层及气候星座观测系统(COSMIC)折射率资料,针对局地观测算子,估计了COSMIC折射率在南、北半球高、中、低6个纬度带四季的观测误差,分析了COSMIC折射率观测误差的纬度、高度和季节变化的特点,并将估计的折射率观测误差应用于GRAPES(Global/Regional Assimilation and Prediction Enhanced System)三维变分同化系统。结果表明,折射率观测误差随高度和纬度有明显变化;在中、高纬度带,折射率观测误差有显著的季节变化:夏季折射率的观测误差约为冬季2倍,春、秋两季折射率误差具有较好的南北对称性,冬、夏两季折射率观测误差南、北半球差异较大。与GRAPES原来使用的全球平均单一的折射率观测误差相比,在GRAPES全球三维变分同化系统中使用本研究估计的较高精度的随纬度变化的COSMIC折射率观测误差能够提高GRAPES全球变分同化系统的预报水平。     

An imaging condition for reverse time migration based on wavefield decomposition

Reverse Time Migration(RTM) is a high precision imaging method of seismic wavefield at present,but low-frequency noises severely affect its imaging results.Thus one of most important aspect of RTM is to select the proper noise suppression method.The wavefield characteristics of the Poynting vector are analyzed and the upgoing,downgoing,leftgoing and rightgoing waves are decomposed using the Poynting vector of the acoustic wave equation.The normalized wavefield decomposition cross-correlation imaging condition is used to suppress low-frequency noises in RTM and improve the imaging precision.Numerical experiments using the Mamousi velocity model are performed and the results demonstrate that the upgoing,downgoing,leftgoing and rightgoing waves are well decomposed using the Poynting vector.Compared with the normalized cross-correlation imaging and Laplacian filtering method,the results indicate that the low-frequency noises are well suppressed by using the normalized wavefield decomposition cross-correlation imaging condition.     

附加视线向量修正的卫星影像区域网平差

针对传统有理函数模型(RFM)区域网平差方法局限于姿态和轨道测量误差小、相机视场角小及影像交会角良好的情况,提出了附加视线向量修正的卫星影像区域网平差方法。首先利用影像附带的有理多项式系数(RPC)计算出像元视线向量,其次根据该视线向量恢复成像时刻虚拟位置和姿态信息,然后对恢复的虚拟位置和姿态构建误差补偿模型,最后通过最小二乘方法整体解算模型参数和连接点物方坐标。该方法从系统误差产生的原因构建补偿模型,可以规避传统区域网平差方法的近似假设和条件限制。通过对模拟数据以及多套测绘卫星和非测绘卫星数据进行试验的结果表明,该方法处理大姿态角误差、大视场角以及弱交会角等各种严苛条件下的卫星影像能达到比传统方法更好的效果。     

区域地壳水平运动模型的Euler-GABP神经网络构建

针对传统的欧拉(Euler)矢量法在具有复杂构造运动的区域地区内构建GPS速度场模型精度低的问题,该文提出了欧拉矢量法结合遗传算法(GA)和BP神经网络方法来构建区域GPS速度场。以云南区域1999—2015年间的GPS速度场为实验数据,分别使用欧拉矢量法、遗传神经网络(GABP)、欧拉矢量结合神经网络(Euler-BP)和欧拉矢量的遗传神经网络(Euler-GABP)法构建区域速度场模型。实验结果表明,相比其他3种方法,该文提出的Euler-GABP方法在构建区域GPS速度场模型是可行和有效的且精度高。     

���Իع�������С����ƽ��ģ�ͼ�����

??????????????г????е?????????????С???????????????????????С?????????????????????????????????????С????????????????????????????????????????????