地震资料波阻抗多尺度融合反演

本文依据最佳约束反演策略,应用波动方程多尺度反演方法反演中频段波阻抗.利用神经网络技术实现对中频段波阻抗的标定及高、中、低频波阻抗的融合,避免了因高、中、低频波阻抗频段重叠及中频段波阻抗精度低所造成的反演结果的误差.这种方法具有对初始模型依赖性低、收敛速度快、反演结果稳定、抗噪能力强等特点.     

井间地震层析成像的现状与进展

综述了井间地震层析成像研究的现状，给出了小波变换域井间地震层析成像方法的最新进展，并对井间地震层析成像研究给予展望。     

Analyses of geographical observations in the Heihe River Basin: Perspectives from complexity theory

Journal of Geographical Sciences - Since 2005, dozens of geographical observational stations have been established in the Heihe River Basin (HRB), and by now a large amount of meteorological,...     

基于MAEU-CNN的高分辨率遥感影像建筑物提取

从高空间分辨率图像（HSRI）中提取建筑物信息在遥感应用领域具有重要意义。然而,由于遥感影像中的建筑物尺度变化大、背景复杂和外观变化大等因素,从HSRI中自动提取建筑物仍然是一项具有挑战性的任务。特别是从影像中同时提取小型建筑物群和具有精确边界的大型建筑物时,难度更大。为解决这些问题,本文提出了一种端到端的编码器-解码器神经网络模型,用于从HSRI中自动提取建筑物。所设计的网络称为MAEU-CNN(Multiscale Feature Enhanced U-shaped CNN with Attention Block and Edge Constraint)。首先,在设计的网络编码部分加入多尺度特征融合（MFF）模块,使网络能够更好地聚集多个尺度特征。然后,在编码器和解码器部分之间添加了多尺度特征增强模块（MFEF）,以获得不同尺寸的感受野,用于获取更多的多尺度上下文信息。在跳跃连接部分引入双重注意机制,自适应地选择具有代表性的特征图用于提取建筑物。最后,为了进一步解决MAEU-CNN中由于池化及卷积操作导致的分割结果边界模糊的问题,引入多任务学习机制,将建筑物的边界几何信息融入网络...     

Regional watershed characterization and classification with river network analyses

In order to understand and manage a hydrological region, one usually needs to comprehensively characterize the watersheds (basins) and their river networks. This usually and primarily involves analysis of hydrological and geomorphological properties of the watershed derived from the digital terrain model (DTM), but this approach neglects the information content of the associated river networks. In this study, we used a combination of traditional DTM and original river network‐related indices to the watersheds of an understudied region, Haiti. We also used Monte Carlo simulations to estimate index confidence levels of these indices. Compared to commonly used indices, the network indices provided valuable information that could then be used in statistical analyses as a way to identify the dominant features of the country's watershed morphology. On this basis, we identified four watershed groups in Haiti: (i) high, elongated watersheds, (ii) lowlands, with sinuous networks and relatively slow runoff, (iii) high watersheds with dendritic networks, and (iv) lowlands with high downstream–upstream contrast in properties and rapid runoffs. We argue that river network features provide complementary information in terms of flow topology, highly relevant to characterize contrasting relief countries, such as Haiti. Hence, more exhaustive characterization of watersheds would predictably benefit from the approach outlined in this paper. Copyright © 2017 John Wiley & Sons, Ltd.     

Analysing the influences of ENSO and PDO on water discharge from the Yangtze River into the sea

The El Niño Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) are two important climate oscillations that affect hydrological processes at global and regional scales. However, few studies have attempted to identify their single and combined influences on water discharge variability at multiple timescales. In this study, we examine temporal variation in water discharge from the Yangtze River into the sea and explore the influence of the ENSO and the PDO on multiscale variations in water discharge over the last century. The results of the wavelet transform analysis of the water discharge series show significant periodic variations at the interannual timescale of 2 to 8 years and the decadal timescale of 15 to 17 years. Water discharge tended to be higher during the La Niña–PDO cold phase and lower during the El Niño–PDO warm phase. The results of the cross wavelet spectrum and wavelet coherence analyses confirm the relationship between the interannual (i.e., 2 to 8 years) and decadal (i.e., 15 to 17 years) periodicities in water discharge with the ENSO and the PDO, respectively. As an important large‐scale climate background, the PDO can modulate the influence of the ENSO on water discharge variability. In general, the warm PDO enhances the influence of El Niño events, and the cold PDO enhances the influence of La Niña events. Our study is helpful in understanding the influencing mechanism of climate change on hydrological processes and provides an important scientific guideline for water resource prediction and management.     

A multiscale method for distributed parameter estimation with application to reservoir history matching

A method for multiscale parameter estimation with application to reservoir history matching is presented. Starting from a given fine-scale model, coarser models are generated using a global upscaling technique where the coarse models are tuned to match the solution of the fine model. Conditioning to dynamic data is done by history-matching the coarse model. Using consistently the same resolution both for the forward and inverse problems, this model is successively refined using a combination of downscaling and history matching until model-matching dynamic data are obtained at the finest scale. Large-scale corrections are obtained using fast models, which, combined with a downscaling procedure, provide a better initial model for the final adjustment on the fine scale. The result is thus a series of models with different resolution, all matching history as good as possible with this grid. Numerical examples show that this method may significantly reduce the computational effort and/or improve the quality of the solution when achieving a fine-scale match as compared to history-matching directly on the fine scale.     

Multiscale tomography of buried magnetic structures: its use in the localization and characterization of archaeological structures

We have previously developed a method for characterizing and localizing 'homogeneous' buried sources, from the measure of potential anomalies at a fixed height above ground (magnetic, electric and gravity). This method is based on potential theory and uses the properties of the Poisson kernel (real by definition) and the continuous wavelet theory. Here, we relax the assumption on sources and introduce a method that we call the 'multiscale tomography'. Our approach is based on the harmonic extension of the observed magnetic field to produce a complex source by use of a complex Poisson kernel solution of the Laplace equation for complex potential field. A phase and modulus are defined. We show that the phase provides additional information on the total magnetic inclination and the structure of sources, while the modulus allows us to characterize its spatial location, depth and 'effective degree'. This method is compared to the 'complex dipolar tomography', extension of the Patella method that we previously developed. We applied both methods and a classical electrical resistivity tomography to detect and localize buried archaeological structures like antique ovens from magnetic measurements on the Fox-Amphoux site (France). The estimates are then compared with the results of excavations.     

面向类型特征的自适应阈值遥感影像变化检测

为了实现两个不同年份单时相遥感影像之间的土地覆盖变化检测，提出了一种基于土地覆盖类型特征自适应确定阈值的遥感影像变化检测方法。以2015年土地覆盖数据为基础，综合2013年和2015年Landsat 8-OLI影像数据，首先，采用时相不变点群法TIC（Temporally Invariant Cluster）保证了两期影像辐射水平的一致性。其次，对两期影像进行多尺度分割，并在各级尺度下构建分割对象的变化向量。然后，采用最大类间方差的方法分别进行单一变化阈值变化检测以及基于土地覆盖类型的多阈值变化检测分析，并利用目视解译样点进行精度验证与评价。结果表明：（1）单一阈值变化检测结果的总体精度为79.6%，Kappa系数为0.601，多阈值变化检测结果的总体精度为87.2%，Kappa系数为0.741，多阈值变化检测具有更高的精度。（2）进一步逐土地覆盖类型精度评价可知，多阈值变化检测能在一定程度上减弱物候期的影响，具有更高的稳定性。该研究以土地覆盖数据为底图，逐类别的选取变化检测阈值，提高了变化区域检测的精度，在大范围高效更新土地覆盖数据的应用中具有一定的参考价值。