机器学习在地震预测中的应用进展

机器学习(Machine Learning,ML),特别是深度学习(Deep Learning,DL),在最近几年发展迅速,在数据挖掘、计算机视觉、自然语言处理、数据特征提取和预测等方面的应用中取得了令人振奋的进展.地震预测是复杂、涉及面广、不成熟而且充满争议的科学问题;其发展受到尚不清楚的地震机理和孕震结构、不完备的...     

Assessment and zoning of eco-environmental sensitivity for a typical developing province in China

Irrational exploitation and utility of resources and extensive economic growth have become the primary causes of deterioration of ecological environment in many provinces in China. Ecological environment deterioration necessitates ecological zoning and planning in order to regulate human activities and guide the sustainable use of resources and ecological services. Eco-environmental sensitivity assessment is a basis upon which the establishment of rational zoning and planning rests. This research focused on evaluating the eco-environmental sensitivity in Anhui province, a typical province confronted with contradiction between socio-economic development and resource restriction. A multifactor assessment approach to zoning the eco-environmental sensitivity for the province in complex ecological environment and physical geographic condition was presented by incorporating remote sensing, Geographic information system, and observation and statistical data. Based on these, to support landscape and urban planning and design, rational deployment of resources and arrangement of industry, and promote sustainable development of ecology, society and economy, the recommendations for ecological construction and regional development were proposed for different zones of principal functions, and an ecological security pattern was constructed from one ecological functional core, four ecological functional zones, six ecological functional belts, three types of ecological functional corridors, and some ecological functional patches.     

深度学习在地震定位中的应用综述

近年来,深度学习的发展给科研人员开辟了地震定位研究的新思路,科研人员将深度学习技术应用于地震定位并取得了较好的效果。文章首先介绍根据神经网络的编码与解码对深度神经网络的分类,然后对深度学习的基本流程进行总结,最后对深度学习中广泛应用于地震定位的方法进行综述,总结不同方法的特点和实际应用情况。结果表明：深度学习方法能够实现地震事件的自动定位,且定位的精度较高,缩短了地震定位所需时间,在处理地震大数据方面也具有明显优势,能够克服目前传统地球物理方法在地震定位方面的一些不足之处。相信随着深度学习技术的进一步发展,必将更为广泛地应用于地震定位研究中。     

Integrative ecology in the era of big data——From observation to prediction

Most ecological and environmental issues faced by human society can only be solved at the ecosystem, watershed,regional and even global scale. Thus, ecological research is developing rapidly towards macro-scale studies. With the rapid development of observational networks and information technology, the spaceborne-aircraft-ground based observation system is becoming an important feature of ecosystem monitoring in the new era. With the gradual formation of the global new-generation observational systems and the rapid expansion of massive multi-source heterogeneous data, ecology has entered the era of big data, big science, and big theory. How to integrate ecological big data, discover valuable ecological laws and mechanisms, and further expand them to solve eco-environmental issues that closely relate to human development are the major opportunities and challenges in this field. In this paper, we systematically summarized the research progresses in ecological big data, reviewed the opportunity and demand of integrative ecology, and further discussed the main approaches of ecological big data integration by using meta-analysis, data mining, and data-model fusion. Finally, we proposed the prospects and research directions of integrative ecology and suggested that future researches need to integrate big data into land models so as to improve the accuracy of ecological forecasting. It can be foreseen that under the background of global change and the rapid development of big data in the future, integrative ecology will be extensively applied and developed to serve the sustainable development of human society.     

基于深度学习的重力异常与重力梯度异常联合反演

高效高精度的反演算法在重力大数据时代背景下显得尤为重要,受深度学习卓越的非线性映射能力的启发,本文提出了一种基于深度学习的重力异常及重力梯度异常的联合反演方法.文中首先提出了一种基于网格点几何格架的重力异常及重力梯度异常的空间域快速正演算法,这为本文深度学习反演算法的实现奠定了基础;其次对大量的不同密度模型进行正演计算获得样本数据集;然后设计了一种端到端的深度学习网络结构(GraInvNet),再利用样本数据对该网络结构进行训练;最后进行反演预测.组合模型试验表明,多维度数据联合反演相比单一分量反演其结果更“聚焦”,且与模型边界高度吻合,并且对于复杂模型的姿态与物性预测具有极为显著的优势,以及对于含噪声数据的反演,其质量也不会降低;Vinton岩丘实测重力数据也验证了文中方法的有效性;从而证明了深度学习在重力数据的高效高精度反演方面具有的巨大潜力.     

基于深度学习网络的地震地质灾害识别研究

遥感影像识别方法是破坏性地震震后地质灾害快速、准确获取的重要方法之一,传统的遥感影像识别方法主要以人工目视识别方法和半自动识别方法为主,需投入大量的人力和时间。针对破坏性地震震后地质灾害解译时间长、投入人力多等问题,以2017年8月8日四川九寨沟7.0级地震震后高分辨率无人机遥感影像为研究样本,提出基于深度学习网络的地震地质灾害识别方法。首先结合震后遥感影像解译资料和现场调查资料,提取九寨沟地震地质灾害无人机遥感影像特征,并构建研究区地震地质灾害解译指标和分类数据集;然后采用DeepLabv3+网络结构及softmax损失函数,建立基于深度学习网络的地震地质灾害遥感影像图像语义分割模型方法;最后采用半监督学习方法进行结果验证。研究结果表明,基于深度学习网络的地震地质灾害识别方法可有效识别九寨沟地震地质灾害分布信息,整体分类识别准确率为94.22%,F₁分数值为0.77,结果具有较好的一致性和准确性,可提升地震现场灾情获取和重点地震隐患识别等工作效率及服务能力。     

基于深度学习的低剂量CT成像算法研究进展

计算机断层扫描成像(CT)技术具有成像速度快分辨率高的优点,广泛应用于医学临床诊断中.然而,提高剂量辐射会引发人体组织器官受损,降低剂量又会造成成像质量严重下降.为解决上述矛盾,在确保成像质量满足临床诊断需求的条件下,研究如何最大程度地降低X射线辐射对人体造成的伤害,己成为低剂量CT成像技术的研究热点.近年来,在人工智...     

基于特征学习的低剂量CT成像算法研究进展

随着CT（computed tomography）技术在临床中的大量应用,其辐射伤害问题也越来越受到人们的关注。与此同时,高性能低剂量的成像也已经成为近年来CT研究领域中的重要研究方向。随着学习型算法的提出及广泛应用,为低剂量CT成像算法的发展带来了新的方向。在影像大数据环境下,基于特征学习方法的低剂量CT成像有着更广阔的发展空间。本文将从稀疏表示和深度学习两个方面,介绍一些国内外应用于改善CT成像质量的相关技术,包括CT成像技术的发展趋势,特征学习相关算法的研究现状,提高低剂量CT扫描成像质量的相关方案等。本文对近年来在低剂量CT成像及特种学习算法等领域的研究成果进行了介绍,并进行相关总结和分析。      

基于电磁卫星的闪电哨声波智能检测算法的研究进展

闪电哨声波作为探索空间物理环境的重要媒介,淹没在海量的电磁卫星数据中.近年来随着计算机视觉和深度学习等人工智能技术的发展,从电磁卫星的存档数据中自动检测闪电哨声波的算法取得了一定的效果.本文对近年来闪电哨声波智能检测算法的文献进行了整理和总结.首先,阐述闪电哨声波在电磁卫星数据中呈现的时频特征和类型;然后,介绍了闪电哨声波智能检测算法的流程并从闪电哨声波的特征提取、分类和定位三个方面对主要的智能检测算法进行归纳、总结和评述;其次,简述了闪电哨声波智能检测模型的评价指标;接着,在张衡一号(ZH-1)卫星的磁场数据上对三种典型的闪电哨声波智能检测算法进行复现,并对三种算法的优缺点进行了较深入的分析;最后,对基于电磁卫星的闪电哨声波智能检测的研究领域进行总结和展望.     

基于深度能量模型的低剂量CT重建

降低计算机断层扫描（CT）的剂量对于降低临床应用中的辐射风险至关重要,深度学习的快速发展和广泛应用为低剂量CT成像算法的发展带来了新的方向。与大多数受益于手动设计的先验函数或有监督学习方案的现有先验驱动算法不同,本文使用基于深度能量模型来学习正常剂量CT的先验知识,然后在迭代重建阶段,将数据一致性作为条件项集成到低剂量CT的迭代生成模型中,通过郎之万动力学迭代更新训练的先验,实现低剂量CT重建。实验比较,证明所提方法的降噪和细节保留能力优良。      

长江上游生态保护与经济发展综合改革方略研究

依托长江黄金水道推动长江经济带发展是国家重大战略举措,如何解读"坚持生态优先、绿色发展"、"共抓大保护、不搞大开发"的理念,是长江经济带发展需要破解的重大命题.本文针对长江经济带发展的难点地区——长江上游地区,提出设立"长江上游经济带经济体制和生态文明体制综合改革试验区"的设想,认为应着眼于体制变革和机制创新,以依托黄金水道推动长江经济带发展为目标,以梯级水电开发为抓手,以深化区域综合配套改革为根本动力,给予地方政府在资源、环境、经济社会发展等领域中更大的自主权.本文分析了设置试验区的战略意义和定位,提出了该试验区的重点创新内容.     

云南高原湖泊的开发与保护

介绍了云南高原湖泊的形成演化,重点论述其在云南社会、经济发展中的地位,分析了湖泊不合理开发带来的不良生态后果,并提出了一些建议。     

Total consumption controlled water allocation management for multiple sources and users with inexact fuzzy chance-constrained programming: a case study of Tianjin,China

Environmental and ecological issues caused by water resources crisis have brought enormous challenges to the sustainable development of water-deficient area. Water resources allocation management balancing the relationship between the social-economic development and the ecological environment has become a hot topic in recent years. In this paper, an inexact fuzzy chance-constrained programming (IFCCP) approach is proposed for regional water resource allocation optimization with the aim of promoting the harmonious development of the social economic and the ecological environment, improving water utilization efficiency, and realizing water resources consumption control under uncertainties. The method is incorporated with interval parameter programming, fuzzy programming, and chance-constrained programming, for handling system uncertainties and balancing the optimal objectives with the risk of violating system constraints. Under this framework, an IFCCP model for water resources allocation management was successfully formulated and applied to a typical water-deficit area, Tianjin, China, for obtaining a better water resources plan among multiple users under resources and environmental limitation. Different total water consumption control policies are designed for assessing regional water allocation schemes. The results indicated that the gap of supply and demand will only be solved by foreign water, the transferred water from Luan River and Changjiang River would still be the main supplier in planning horizon. Moreover, the strict total water consumption control policy would guarantee the water requirement of ecological environment, lead to changes in the structure of water supply, actively guide on water conservation, and promote the large-scale utilization of desalted water and recycle water.     

太湖与长三角区域一体化发展:地位、挑战与对策

太湖及太湖流域在长三角自然和经济地理空间具有举足轻重的地位,更是长三角区域一体化发展国家战略实施的关键地区.在当前区域一体化高质量发展背景下,太湖治理面临新的形势和任务,也面临一系列新的挑战,着重体现在水环境治理形势更加复杂严峻和水资源供给压力不断提升两方面.传统水陆分割治理难以解决湖泊问题,行政区治理难以适应区域一体化发展要求,水资源共享和付费机制尚未建立等问题,对长三角区域一体化进程也构成了较大阻滞.太湖水问题绝不局限于太湖本身,需要从湖泊—流域系统视角构建资源、环境、生态、社会与经济多要素协调统筹治理的新体系,考虑水陆空间协同治理与开发和不同次区域经济、社会、自然效益的均衡,以推动长三角区域一体化高质量发展.基于此,研究提出新背景下的太湖及流域治理思路:推动太湖水资源生态环境之间以及与流域经济社会的联动,撬动长三角生态环境一体化的思维创新;探索关键卡脖子技术研发和流域管理综合集成的科学体系,推动长三角资源环境领域的创新一体化发展;探索太湖流域跨行政区域的水生态环境共治、水资源共享和绿色发展机制,牵动一体化协调制度的改革创新.     

Measurement of ecological capital of Chinese terrestrial ecosystem based on remote sensing

The biosphere of the Earth is the material founda-tion of human survival and development.With the increasing problems of the resources,the environment and population,most countries in the world have paid special attention to the measurement and evaluation of ecological capital,which refers to the sum of the direct natural resources value and the indirect ecosystem services value.Ecological capital depends on not only the quality of the national ecological environment,butalso the social recogni…     

Elastic impedance based facies classification using support vector machine and deep learning

Machine learning methods including support-vector-machine and deep learning are applied to facies classification problems using elastic impedances acquired from a Paleocene oil discovery in the UK Central North Sea. Both of the supervised learning approaches showed similar accuracy when predicting facies after the optimization of hyperparameters derived from well data. However, the results obtained by deep learning provided better correlation with available wells and more precise decision boundaries in cross-plot space when compared to the support-vector-machine approach. Results from the support-vector-machine and deep learning classifications are compared against a simplified linear projection based classification and a Bayes-based approach. Differences between the various facies classification methods are connected by not only their methodological differences but also human interactions connected to the selection of machine learning parameters. Despite the observed differences, machine learning applications, such as deep learning, have the potential to become standardized in the industry for the interpretation of amplitude versus offset cross-plot problems, thus providing an automated facies classification approach.     

一种综合地学数据的矿产资源潜力制图方法

研究适应信息化时代特征的矿产资源潜力制图新技术、新方法对推动矿产资源评价理论与技术的发展具有重要的意义.笔者把GIS技术、图像分类算法和空间统计学理论进行有机集成,在空间统计学的空间结构分析技术和遥感图像纹理分类算法的基础上,提出了一种以综合地学数据（地质、地球物理、地球化学和遥感图像数据等）为基本数据源的矿产资源潜力自动制图方法.该方法的技术流程为：①数据准备,即对地球物理和地球化学勘探数据进行预处理,生成一个物化遥综合图像文件;②图像空间结构性分析和纹理图像生成,以综合地学图像为研究对象,用空间统计学的结构分析技术研究地学数据综合图像的空间结构性,生成纹理图像;③纹理图像多元分类,用实验变差函数纹理分类方法对研究区进行多元分类,生成分类专题图;④分类后处理,用叠置分析修正空间分类结果,生成区域矿产资源潜力分布图.     

Macrobenthic community for assessment of estuarine health in tropical areas (Northeast, Brazil): Review of macrofauna classification in ecological groups and application of AZTI Marine Biotic Index

The aim of the present study was to investigate the ecological quality of tropical estuaries on the northeastern coast of Brazil using the AMBI. Macrofauna classification based on ecological groups was reviewed using the Indicator Value (IndVal) coefficient. The results indicate that the ecosystems exhibit some level of disturbance. Most sites are situated between slightly-moderately disturbed boundaries due to the higher proportion of Nematoda (assigned here as Ecological Group I) and of Oligochaeta and Tubificidae (both classified as Ecological Group V). The AMBI proved efficient in evaluating environmental status, although the applicability of this index requires adjustments regarding some species in ecological groups. The present study also highlights the merits of the IndVal method for examining the assignments of species/taxa to an ecological group and demonstrates the validity of this coefficient is an assessment tool. Moreover, the complementary use of different methods is recommended for the assessment of ecosystem quality.     

湖泊藻型富营养化控制——技术、理论及应用

湖泊富营养化防治走过了从控制营养盐、直接除藻、到生物调控、生态工程及生态恢复等艰难历程,各国为此投入了巨额资金、然而收效甚微,富营养化依然是全球性重大水环境问题。回顾和分析富营养化湖泊治理研究与实践的成功经验与失败教训,无疑将有助于采取更切实可行的技术有效控制湖泊富营养化。综观全球营养养化治理研究成果,不难看出,富营养化是一个典型的生态问题,生态问题只有用生态学方法解决。在全湖性富营养化难以快速根治的情况下,如何集中技术优势和有限财力,优先解决对人类生活影响较大的局部水域富营养化问题,逐步修复受损的湖泊生态系统,提高水体自净能力、改善水体环境质量并建立湖泊健康生态系统。     

Biogeography of the Oceans: a Review of Development of Knowledge of Currents,Fronts and Regional Boundaries from Sailing Ships in the Sixteenth Century to Satellite Remote Sensing

The development of knowledge of global biogeography of the oceans from sixteenthcentury European voyages of exploration to present-day use of satellite remote sensing is reviewed in three parts; the pre-satellite era (1513–1977), the satellite era leading to a first global synthesis (1978–1998), and more recent studies since 1998. The Gulf Stream was first identified as a strong open-ocean feature in 1513 and by the eighteenth century, regular transatlantic voyages by sailing ships had established the general patterns of winds and circulation, enabling optimisation of passage times. Differences in water temperature, water colour and species of animals were recognised as important cues for navigation. Systematic collection of information from ships’ logs enabled Maury (The Physical Geography of the Sea Harper and Bros. New York 1855) to produce a chart of prevailing winds across the entire world’s oceans, and by the early twentieth century the global surface ocean circulation that defines the major biogeographic regions was well-known. This information was further supplemented by data from large-scale plankton surveys. The launch of the Coastal Zone Color Scanner, specifically designed to study living marine resources on board the Nimbus 7 polar orbiting satellite in 1978, marked the advent of the satellite era. Over subsequent decades, correlation of satellite-derived sea surface temperature and chlorophyll data with in situ measurements enabled Longhurst (Ecological Geography of the Sea. Academic Press, New York 1998) to divide the global ocean into 51 ecological provinces with Polar, Westerly Wind, Trade Wind and Coastal Biomes clearly recognisable from earlier subdivisions of the oceans. Satellite imagery with semi-synoptic images of large areas of the oceans greatly aided definition of boundaries between provinces. However, ocean boundaries are dynamic, varying from season to season and year to year. More recent work has focused on the study of variability of currents, fronts and eddies, which are often the focus of high biological productivity. Direct tracking of animals using satellite-based systems has helped resolve the biological function of such features and indeed animals instrumented in this way have helped the study of such features in three dimensions, including depths beyond the reach of conventional satellite remote sensing. Patterns of surface productivity detected by satellite remote sensing are reflected in deep sea life on the sea floor at abyssal depths >3,000 m. Satellite remote sensing has played a major role in overcoming the problems of large spatial scales and variability in ocean dynamics and is now an essential tool for monitoring global change.