Forecasting the western Pacific subtropical high index during typhoon activity using a hybrid deep learning model

Seasonal location and intensity changes in the western Pacific subtropical high(WPSH) are important factors dominating the synoptic weather and the distribution and magnitude of precipitation in the rain belt over East Asia. Therefore, this article delves into the forecast of the western Pacific subtropical high index during typhoon activity by adopting a hybrid deep learning model. Firstly, the predictors, which are the inputs of the model, are analysed based on three characteristics: the first...     

Seeing the local in the global: Political ecologies, world-systems, and the question of scale

Scale, as concept, has featured prominently in political ecology and remains, even if implicitly, a crucial point of analytical reference. Recent studies, drawing from both human geography and ecology, have sought to demonstrate how scales, rather than pre-existing ontologically, are both socially and environmentally produced. Given the different scales through which social and environmental processes occur, the study of society-environment relations can be improved by analysing varying scalar configurations of interaction. This recent and promising methodological corrective would greatly benefit from a dialogue with world-systems approaches, which integrate diverse scale-producing processes and to some extent overlap in scope with political ecology. World-systems perspectives, by focusing on the long-term systemic character of people-environment relations, effectively connect micro- to macro-scale social and ecological processes and explain long-term internal dynamics and interrelations of systems at different scales. Conversely, world-systems approaches could learn much from political ecologists’ consideration of nonhuman processes into understandings of scale and society-environment relations, which has a long tradition in geography, as well as from the more context-sensitive analytical framework brought to those understandings. Case studies are discussed to demonstrate not only how these two perspectives could be integrated, but also how explanations of environmental change can be thereby improved. Combining the two approaches provides the basis for a more ecologically oriented world-systems paradigm and, in political ecology, for greater sensitivity to socially large-scale systemic processes and, given the originally anti-capitalist underpinnings of both paradigms, for more political coherence.     

General seismic wave and phase detection software driven by deep learning

We developed an automatic seismic wave and phase detection software based on PhaseNet, an efficient and highly generalized deep learning neural network for P- and S-wave phase picking. The software organically combines multiple modules including application terminal interface, docker container, data visualization, SSH protocol data transmission and other auxiliary modules. Characterized by a series of technologically powerful functions, the software is highly convenient for all users. To obtain the P- and S-wave picks, one only needs to prepare three-component seismic data as input and customize some parameters in the interface. In particular, the software can automatically identify complex waveforms (i.e. continuous or truncated waves) and support multiple types of input data such as SAC, MSEED, NumPy array, etc. A test on the dataset of the Wenchuan aftershocks shows the generalization ability and detection accuracy of the software. The software is expected to increase the efficiency and subjectivity in the manual processing of large amounts of seismic data, thereby providing convenience to regional network monitoring staffs and researchers in the study of Earth's interior.     

IEU-Net高分辨率遥感影像房屋建筑物提取

房屋建筑物作为人类活动的主要场所,快速准确地将其从高分辨率遥感影像中提取出来,对促进遥感信息在防灾减灾、城镇管理等方面的应用具有重要意义。本文基于深度学习,提出了高分辨率遥感影像房屋建筑物像素级精确提取方法。首先,针对样本图像边缘像素特征不足现象,以U-Net模型为基础提出IEU-Net模型,设计了全新的忽略边缘交叉熵函数IELoss并将其作为损失函数,另外添加Dropout和BN层在避免过拟合的同时提高模型训练速度和鲁棒性。其次,为解决模型特征丰富度有限的问题,引入形态学建筑物指数MBI,与遥感影像RGB波段一同参与到模型的分类过程。最后,在模型预测时与IELoss相对应采用忽略边缘预测策略从而获得最佳建筑物提取结果。实验对比分析表明：本文方法能有效克服样本边缘像素特征不足问题并抑制道路、建筑物阴影对结果的影响,提升高分辨率遥感影像中房屋建筑物的提取精度。     

深度学习遥感图像标注系统研究

随着深度学习的发展,遥感影像处理技术也从传统机器学习算法向深度学习转变,然而,用于遥感图像的训练数据集却十分稀少,且数据标注困难。本文将GIS技术与图像标注技术相结合,基于Flask Web框架设计一个可用于海量遥感数据的标注系统。该系统可用于海量遥感数据的数据框标注、数据类别标注,以及目标关键点标注,同时能将标注数据导出为深度学习训练最常用的COCO数据集和VOC2007两种格式。     

联合卷积神经网络与集成学习的遥感影像场景分类

针对人工设计的中、低层特征难以实现复杂场景影像的高精度分类以及卷积神经网络依赖大量训练数据等问题，结合迁移学习与集成学习，提出了一种联合卷积神经网络与集成学习的遥感影像场景分类算法。首先基于迁移学习的思想，利用在自然影像数据集上训练好的多个深层卷积神经网络模型作为特征提取器，提取图像多个高度抽象的语义特征；然后构建由Logistic回归和支持向量机组成的Stacking集成模型，对同一图像的多个特征分别训练Logistic模型，将预测概率结果融合构建概率特征；最后利用支持向量机对概率特征训练和预测，得到场景影像的分类结果。利用UCMerced_LandUse和NWPU-RESISC 45两种不同规模的遥感影像数据集进行试验，即使在只有10%的数据作为训练样本情况下，本文方法能够分别达到90.74%和87.21%的分类精度。     

多尺度字典学习的高光谱图像压缩算法

结合小波变换及字典学习提出了一种针对高光谱图像的压缩算法。该算法首先通过小波变换构建多尺度样本集,在小波域使用K-均值奇异值分解(K-SVD)方法学习得到原子尺寸不同的多尺度字典,然后在稀疏表示的过程中,定义一个原子使用频次筛选因子,通过统计局部最优波段稀疏表示时原子使用情况,结合筛选因子对字典原子进行优化筛选,使用精简后的字典对其余波段进行稀疏求解,最后针对不同尺度的表示系数采用自适应的量化编码。实验结果表明,与目前常用的3D-SPIHT和其他的多尺度字典学习算法相比,本文算法在中低比特率下,具有更好的重建性能。     

基于机器学习的地震动特征提取与模拟以2021年云南漾濞6.4级地震为例

随着地震动数据数量的增长和质量的提高,将基于数据驱动的机器学习方法应用到地震动模拟中有重要意义。以2021年5月21日云南漾濞M_S6.4地震为例,利用主成分析方法从前震及余震地震动记录中提取特征母波时程,将地震动三要素作为模拟误差约束,在求解母波的线性组合系数时使用多目标优化算法寻优,最终找到帕累托最优解作为模拟目标台站记录时的组合系数,得到模拟地震动时程。结果表明:主成分析法在对实际地震动记录进行特征提取后,得到的特征母波时程可以在一定程度上保留原始数据的主要信息;考虑幅值、频谱和持时这三要素的角度去控制模拟误差,可以使得模拟的地震动时程更加接近真实记录。提出的基于特征提取的地震动模拟方法可以为基于小震数据合成大震地震动提供参考。     

区域土地资源配置与社会经济发展

文章在阐述土地资源配置与区域社会经济发展的辩证关系的基础上,分别从土地资源供给约束、资源禀赋和利用结构方面探讨了土地资源配置对区域社会经济发展的影响与制约,并对实现十六大提出的全面建设小康社会目标,从土地资源配置角度提出对策与建议。     

明清时期南阳盆地城(集)镇职能组合结构探析

明清时期,南阳盆地由于人口的增长、农业经济和交通的发展,城(集)镇经济也逐步发展和繁荣起来,不同类型的城(集)镇职能组合结构开始形成。在原有的以政治、军事职能为主的城(集)镇基础上,出现了一批手工业集镇、商业集镇和交通集镇。其城(集)镇功能的消长表现在:以政治、军事职能为主的城(集)镇发展相对缓慢,以经济、商业职能为主的集镇得到迅速发展,商业集镇数量增加、人口增长、规模扩大、开市日期增多、集市广设店铺,出现了繁盛的局面。