基于深度学习的镜下矿石矿物的智能识别实验研究

矿石矿物鉴定的智能化是智能地质学和智能矿床学的基础技术之一。计算机视觉技术和深度学习理论使矿石矿物鉴定的智能化成为可能。本研究基于深度学习系统Tensor Flow,以吉林夹皮沟金矿和河北石湖金矿的黄铁矿、黄铜矿、方铅矿、闪锌矿等硫化物矿物为例,设计有针对性的Unet卷积神经网络模型,有效自动提取矿相显微镜下矿石矿物的深层特征信息,实现镜下矿石矿物智能识别与分类。实验显示,模型在训练过程中,随着训练次数的增加,模型精度在不断增大,损失函数不断减小;经过3000个批处理之后,模型精度和损失函数基本趋于稳定。训练出的模型对测试集中的显微镜镜下矿石矿物照片的识别成功率均高于90%,说明实验所建立的模型,具有很好的图像特征提取能力,能完成镜下矿石矿物智能识别的任务。     

地质领域机器学习、深度学习及实现语言

地质大数据正在以指数形式增长。只有发展智能数据处理方法才有可能追上大数据的超常增长。机器学习是人工智能的核心,是使计算机具有智能的根本途径。机器学习已成为地质大数据研究的前沿热点,它将让地质大数据插上翅膀,并因此改变地质。机器学习是一个源于数据的模型的训练过程,最终给出一个面向某种性能度量的决策。深度学习是机器学习研究中的一个重要子类,它通过构建具有很多隐层的机器学习模型和海量的训练数据,来学习更有用的特征,从而最终提升分类或预测的准确性。卷积神经网络算法是最为常用的一种深度学习算法之一,它广泛用于图像识别和语音分析等。Python语言在科学领域的地位占据着越来越重要。其下的Scikit-Learn是一个机器学习相关的库,提供有数据预处理、分类、回归、聚类、预测、模型分析等算法。Keras是一个基于Theano/Tensorflow的深度学习库,可以应用来搭建简洁的人工神经网络。     

支持向量机、随机森林和人工神经网络机器学习算法在地球化学异常信息提取中的对比研究

地球化学勘查是通过发现异常、解释评价异常进行找矿的。因此,地球化学异常识别对矿产资源的定位、定量预测具有重要的的指示作用。在大数据时代的背景下,机器学习方法不要求数据满足正态分布的分布形式,且具有非线性以及泛化能力强等特点,因而逐渐地被应用于矿产资源的定量预测评价,如神经网络、支持向量机、贝叶斯网络、随机森林、受限玻尔兹曼机、极限学习机等。本文通过设计理论实验,可视化了不同算法,提出了不同机器学习方法在不同地区的地球化学异常信息提取中的效果存在不一致性的假设。在此基础上,以湖南香花岭锡多金属矿整装勘查区及甘肃合作金矿整装勘查区的地球化学异常提取为研究内容,将人工神经网络、随机森林以及支持向量机应用于研究区地球化学异常信息的提取与识别工作。在香花岭研究区,人工神经网络的结果较好,在合作研究区,随机森林的结果较好,从而验证了上述假设。通过生成两研究区的地球化学异常图,讨论了该方法在两研究区地球化学异常的地质意义和该方法的可靠性与实用性。此外,还完善了基于多种监督机器学习方法的地球化学异常信息提取流程,为软件开发提供了一定的理论依据。     

Development of Artificial Intelligence Technology in Weather Forecast

Numerical weather prediction, which is the major basis of current weather forecast, has some shortcomings, such as the understanding of the law of atmospheric motion, the assimilation and application of observation data, the expression of model physics, etc., leading to the forecast error of weather. The rapid development of artificial intelligence technology in recent years provides a new possibility for the advancement and innovation of weather forecast. In this paper, the background of the development of artificial intelligence, the current situation of the application of artificial intelligence technology to weather forecast and the future development trend are mainly described to account for this possibility. After that, the idea for development of weather forecast technology based on the integration of artificial intelligence and numerical forecast is put forward. Particularly, this study stresses that, in order to advance the AI algorithm of weather forecast in the future, it is requested to focus on the nonlinear and chaotic characteristics of atmospheric motion leading to the uncertainty of forecast. Starting from the essence of mathematics and physics, we need to realize the hybrid modeling of mathematics and physics, not only to establish the framework of input-output mapping, but also to provide solutions to the bottleneck problems of weather forecast.     

Landslide catastrophes and disaster risk reduction: A GIS framework for landslide prevention and management

A detailed and accurate inventory map of landslides is crucial for quantitative hazard assessment and land planning. Traditional methods relying on change detection and object-oriented approaches have been criticized for their dependence on expert knowledge and subjective factors. Recent advancements in high-resolution satellite imagery, coupled with the rapid development of artificial intelligence, particularly data-driven deep learning algorithms (DL) such as convolutional neural networks (CNN), have provided rich feature indicators for landslide mapping, overcoming previous limitations. In this review paper, 77 representative DL-based landslide detection methods applied in various environments over the past seven years were examined. We analyze the structures of different DL networks, discuss on five main application scenarios, and assess both the advancements and limitations of DL in geological hazard analysis. The results indicated that the increasing number of articles per year reflects growing interest in landslide mapping by artificial intelligence, with U-Net-based structures gaining prominence due to their flexibility in feature extraction and generalization. Finally, we explored the hindrances of DL in landslide hazard research based on the above research content. Challenges such as black-box operations and sample dependence persist, warranting further theoretical research and future application of DL in landslide detection.     

Deep slab subduction and dehydration and their geodynamic consequences: Evidence from seismology and mineral physics

We present new pieces of evidence from seismology and mineral physics for the existence of low-velocity zones in the deep part of the upper mantle wedge and the mantle transition zone that are caused by fluids from the deep subduction and deep dehydration of the Pacific and Philippine Sea slabs under western Pacific and East Asia. The Pacific slab is subducting beneath the Japan Islands and Japan Sea with intermediate-depth and deep earthquakes down to 600 km depth under the East Asia margin, and the slab becomes stagnant in the mantle transition zone under East China. The western edge of the stagnant Pacific slab is roughly coincident with the NE–SW Daxing'Anling-Taihangshan gravity lineament located west of Beijing, approximately 2000 km away from the Japan Trench. The upper mantle above the stagnant slab under East Asia forms a big mantle wedge (BMW). Corner flow in the BMW and deep slab dehydration may have caused asthenospheric upwelling, lithospheric thinning, continental rift systems, and intraplate volcanism in Northeast Asia. The Philippine Sea slab has subducted down to the mantle transition zone depth under Western Japan and Ryukyu back-arc, though the seismicity within the slab occurs only down to 200–300 km depths. Combining with the corner flow in the mantle wedge, deep dehydration of the subducting Pacific slab has affected the morphology of the subducting Philippine Sea slab and its seismicity under Southwest Japan. Slow anomalies are also found in the mantle under the subducting Pacific slab, which may represent small mantle plumes, or hot upwelling associated with the deep slab subduction. Slab dehydration may also take place after a continental plate subducts into the mantle.     

Landslide susceptibility mapping using machine learning algorithms and comparison of their performance at Abha Basin,Asir Region,Saudi Arabia

The current study aimed at evaluating the capabilities of seven advanced machine learning techniques(MLTs),including,Support Vector Machine(SVM),Random Forest(RF),Multivariate Adaptive Regression Spline(MARS),Artificial Neural Network(ANN),Quadratic Discriminant Analysis(QDA),Linear Discriminant Analysis(LDA),and Naive Bayes(NB),for landslide susceptibility modeling and comparison of their performances.Coupling machine learning algorithms with spatial data types for landslide susceptibility mapping is a vitally important issue.This study was carried out using GIS and R open source software at Abha Basin,Asir Region,Saudi Arabia.First,a total of 243 landslide locations were identified at Abha Basin to prepare the landslide inventory map using different data sources.All the landslide areas were randomly separated into two groups with a ratio of 70%for training and 30%for validating purposes.Twelve landslide-variables were generated for landslide susceptibility modeling,which include altitude,lithology,distance to faults,normalized difference vegetation index(NDVI),landuse/landcover(LULC),distance to roads,slope angle,distance to streams,profile curvature,plan curvature,slope length(LS),and slope-aspect.The area under curve(AUC-ROC)approach has been applied to evaluate,validate,and compare the MLTs performance.The results indicated that AUC values for seven MLTs range from 89.0%for QDA to 95.1%for RF.Our findings showed that the RF(AUC=95.1%)and LDA(AUC=941.7%)have produced the best performances in comparison to other MLTs.The outcome of this study and the landslide susceptibility maps would be useful for environmental protection.     

Development and future prospects of quantitative mineral assessment in China

Mineral potential assessment at the Earth’s surface has been an important research for geoscientists around the world in the past five decades. The fundamental aspects of mineral assessment at different scales can be associated with the following tasks, e.g., mineral potential mapping and estimation of mineral resources. This paper summarized the history and development in terms of theories, methods technologies and software platforms for quantitative assessment of mineral resources in China, e.g. comprehensive information methodology, geological anomaly, three-component quantitative prediction method, 5P ore-finding area, integrated information assessment method, nonlinear process modeling and fractals, three dimensional mineral potential mapping, etc. At last, to discuss the future of quantitative mineral assessment in an era of big data including platform for 3D visualization, analysis and sharing, new methods and protocols for data cleaning, information enhancement, information integration, and uncertainties and multiple explanations of multi-information.© 2019 China Geology Editorial Office.     

The social housing sector in Prague and Warsaw: trends and future prospects

The paper provides an overview of trends and processes of change affecting new social housing provision in Prague and Warsaw. The local responses are reviewed within the context of changes to the national housing system defining the performance of municipal and non-profit housing sectors. The research analyses the mix of policy instruments implemented in three major policy domains—regulatory, fiscal and financial—to promote the production of new social housing in the two cities. The system of new social housing provision is examined as a dynamic process of interaction between public and private institutions defining housing policy outcomes. The outcomes are evaluated through a series of indicators related to housing output, stability of investment, differentiation of rents, affordability and choice. The overview demonstrates how significant shifts in regulatory and fiscal policy, coupled with decentralization of responsibilities for social housing, limit the opportunities for more efficient performance in the sector and its growth. This is particularly evident in Warsaw, where the sector operates as a social safety net. New social housing in both cities has better quality and remains affordable, but access is constrained and waiting times have increased. The research highlights the problem of declining output, dwindling financial resources, and lack of cost recovery due to universal rent control. This is eroding the sustainability of social housing, potentially leading to lower investment and subsequent privatisation. In Warsaw, housing allowances are a municipal responsibility making the liberalization of rents difficult, while Prague has moved in the direction of rent deregulation with a more robust system of means-tested housing support provided by the central government. Such policy choices map a different trajectory for the future of social housing.     

Heterogeneity in the mantle—its creation, evolution and destruction

Small-scale seismic heterogeneity exists at different levels in the lower mantle, and is detected by methods that analyze scattered–not direct–energy from natural and artificial sources. Its vertical distribution, association with subduction, and its ≤ 10-km characteristic scale length strongly suggest that it is chemical/petrological in nature and originally created by melting and differentiation during mid-ocean ridge formation. What is of interest is that the scale lengths of both upper and lower mantle seismic heterogeneity are similar, which supports the view of a common origin explored here. Unlike the lower mantle however, which is broadly homogeneous in structure, the upper mantle contains things that trap and impede the dispersal and re-mixing of heterogeneity: continental crust, lithosphere and cratonic roots. These probably control the depths, the longevity and the age of heterogeneities at shallow mantle levels, and suggest that heterogeneities observed in continental mantle lithosphere are probably old, trapped by the process that grows continental roots. Alternatively, if crustal heterogeneity is controlled by the details of a magmatic process, it must either be somehow continually renewed, for which there is no recognizable surface expression, or it must be depleted over time and the present is a time when, by luck, we may still witness it.     

内蒙古浩布高铅锌矿床闪锌矿微量元素特征及其地质意义

内蒙古浩布高铅锌矿床位于大兴安岭南段黄岗-甘珠尔庙成矿带, 该带是我国北方重要的铅锌多金属成矿带之一。目前, 浩布高矿床闪锌矿微量元素赋存机制尚不清晰, 矿床成因类型亦存在一定的争议。本研究采用激光剥蚀电感耦合等离子体质谱仪(LA-ICP-MS)原位微量元素分析手段, 结合机器学习方法, 对浩布高闪锌矿微量元素组成特征、赋存机制以及矿床成因类型进行了探讨。LA-ICP-MS分析结果表明, 浩布高矿床中闪锌矿以富集Fe、Mn、Co、Cu、Se、Ag、Cd、In、Sn等元素, 贫Ni、Ga、Ge、As、Mo、Sb、Au、Tl、Pb、Bi等元素为特征。其中, Fe、Mn、In等元素主要以类质同象的方式替代Zn赋存于闪锌矿中, Cu、Ag、Sn等元素含量变化范围较大, 可能部分以微粒包裹体的形式存在。In与Cu含量具有一定的正相关, 推测In在闪锌矿中富集机理可能为Cu⁺+In³⁺↔2Zn²⁺。Cd与Fe含量呈一定的正相关性, 而与Zn含量呈负相关性, 这可能暗示在闪锌矿中Cd主要以类质同象的方式置换Zn元素而非Fe元素。通过穷举闪锌矿微量元素图解发现, 即使得分最高的Co/Ag-Mn图解依然存在较大范围的重叠区域, 因此不能简单地用闪锌矿微量元素二元图解来判别矿床类型。通过测试四种机器学习中经典的不同核函数的支持向量机算法, 最终得出准确率为91.5%的高斯内核支持向量机闪锌矿成矿类型分类器, 可以用于多种矿床类型闪锌矿的研究。浩布高矿床中闪锌矿微量元素组成明显不同于密西西比河谷型(MVT)、沉积喷流型(SEDEX)、火山块状硫化物型(VMS)和浅成低温热液型铅锌矿床, 而与矽卡岩型接近。综合已有矿物学、年代学以及本研究获得的闪锌矿地球化学证据表明, 浩布高应是一个典型的矽卡岩型铅锌矿床。

     

Spatial Variability of Rock Depth in Bangalore Using Geostatistical, Neural Network and Support Vector Machine Models

Geospatial technology is increasing in demand for many applications in geosciences. Spatial variability of the bed/hard rock is vital for many applications in geotechnical and earthquake engineering problems such as design of deep foundations, site amplification, ground response studies, liquefaction, microzonation etc. In this paper, reduced level of rock at Bangalore, India is arrived from the 652 boreholes data in the area covering 220 km². In the context of prediction of reduced level of rock in the subsurface of Bangalore and to study the spatial variability of the rock depth, Geostatistical model based on Ordinary Kriging technique, Artificial Neural Network (ANN) and Support Vector Machine (SVM) models have been developed. In Ordinary Kriging, the knowledge of the semi-variogram of the reduced level of rock from 652 points in Bangalore is used to predict the reduced level of rock at any point in the subsurface of the Bangalore, where field measurements are not available. A new type of cross-validation analysis developed proves the robustness of the Ordinary Kriging model. ANN model based on multi layer perceptrons (MLPs) that are trained with Levenberg–Marquardt backpropagation algorithm has been adopted to train the model with 90% of the data available. The SVM is a novel type of learning machine based on statistical learning theory, uses regression technique by introducing loss function has been used to predict the reduced level of rock from a large set of data. In this study, a comparative study of three numerical models to predict reduced level of rock has been presented and discussed.     

岩性识别：方法、现状及智能化发展趋势

岩性识别是地质工作中一项基础而又重要的工作。传统的岩性识别方法过于依赖经验和地质专业知识积累,不仅耗时长、专业性强,还易受主观因素影响,导致准确率不理想。笔者等首先回顾了传统的岩性识别方法,之后总结了最新涌现的智能化识别方法,最后详细介绍了基于岩石图像、镜下图像、图像与元素信息融合等的智能识别方法。基于岩石图像的识别方法对于文中的岩石识别准确率可达90%以上,基于图像与元素融合的岩性识别方法可以降低图像相似度高、风化破坏表观特征等因素对识别准确度的影响。笔者等认为当前岩性智能化识别研究仍处于初级阶段。综合各类数据源的优势,利用机器学习深度挖掘岩石元素、矿物、光谱和表观特征间的内在关联性,有利于突破单源信息的局限性,实现岩性快速准确识别。     

岩性识别：方法、现状及智能化发展趋势

岩性识别是地质工作中一项基础而又重要的工作。传统的岩性识别方法过于依赖经验和地质专业知识积累,不仅耗时长、专业性强,还易受主观因素影响,导致准确率不理想。笔者等首先回顾了传统的岩性识别方法,之后总结了最新涌现的智能化识别方法,最后详细介绍了基于岩石图像、镜下图像、图像与元素信息融合等的智能识别方法。基于岩石图像的识别方法对于文中的岩石识别准确率可达90%以上,基于图像与元素融合的岩性识别方法可以缓解图像相似度高、风化破坏表观特征等因素对识别准确度的影响。笔者等认为当前岩性智能化识别研究仍处于初级阶段,无法满足工程需求。综合各类数据源的优势,利用机器学习深度挖掘岩石元素、矿物、光谱和表观特征间的内在关联性,有利于突破单源信息的局限性,实现岩性快速准确识别。     

中性岩角闪石成分大数据对裂谷与汇聚活动陆缘背景的鉴别

基于GEOROC数据库全球闪长岩中角闪石的矿物化学成分,利用机器学习中支持向量机分类器,本文对汇聚边缘（CM）和板内裂谷（IV）两种环境中的闪长质岩浆岩进行了角闪石成分对比研究。结果显示：角闪石的化学成分能够区分上述两种构造环境,并依此建立区分不同构造背景的模型。该分类模型建立在超过1 700条来自两种构造背景下的闪长岩中的角闪石数据的训练与测试优化基础上,且分类精确度达到87%。总体而言,通过角闪石的矿物化学成分再计算获得的离子数（如Mg_C、Mn_C、Al^Ⅳ、Fe³⁺等）以及相应的温度、压力与氧逸度是区分CM与IV两类构造背景的主要指标。将相应的分类模型应用于扬子西北缘旺苍米仓山地区形成于~760 Ma的闪长岩与辉长闪长岩中的角闪石,分类结果显示其来自板内裂谷环境。结合前人研究及角闪石主、微量元素地球化学特征,本文认为该区的闪长岩与辉长闪长岩可能形成于板内伸展裂谷环境。

     

The past in dust: current trends and future directions in Pleistocene geoarcheology of European loess

Loess is a main archive of Pleistocene landscapes and environments and therefore has an important connection to the preservation and interpretation of Paleolithic sites. In Europe, anthropogenic sites have been found in loess because of past local occupation. At one extreme, sites are well preserved with minimal disturbance often accompanied by embedded proxies to estimate ecological parameters. On the other hand, loess deposits have undergone post-depositional alterations such as weathering, pedogenesis or bioturbation due to changing environmental conditions or other disturbances that obscure anthropogenic sites. We outline the current state of research and connections between Paleolithic archeology and loess research while introducing a series of subsequent regional case studies as part of a special issue. We also make recommendations for future work to incorporate a wider variety of methods to create more robust inferences on hominin and environmental evolution and their connections.     

Spatial variability of soil water erosion: Comparing empirical and intelligent techniques

Soil water erosion (SWE) is an important global hazard that affects food availability through soil degradation, a reduction in crop yield, and agricultural land abandonment. A map of soil erosion susceptibility is a first and vital step in land management and soil conservation. Several machine learning (ML) algorithms optimized using the Grey Wolf Optimizer (GWO) metaheuristic algorithm can be used to accurately map SWE susceptibility. These optimized algorithms include Convolutional Neural Networks (CNN and CNN-GWO), Support Vector Machine (SVM and SVM-GWO), and Group Method of Data Handling (GMDH and GMDH-GWO). Results obtained using these algorithms can be compared with the well-known Revised Universal Soil Loss Equation (RUSLE) empirical model and Extreme Gradient Boosting (XGBoost) ML tree-based models. We apply these methods together with the frequency ratio (FR) model and the Information Gain Ratio (IGR) to determine the relationship between historical SWE data and controlling geo-environmental factors at 116 sites in the Noor-Rood watershed in northern Iran. Fourteen SWE geo-environmental factors are classified in topographical, hydro-climatic, land cover, and geological groups. We next divided the SWE sites into two datasets, one for model training (70% of the samples = 81 locations) and the other for model validation (30% of the samples = 35 locations). Finally the model-generated maps were evaluated using the Area under the Receiver Operating Characteristic (AU-ROC) curve. Our results show that elevation and rainfall erosivity have the greatest influence on SWE, while soil texture and hydrology are less important. The CNN-GWO model (AU-ROC = 0.85) outperformed other models, specifically, and in order, SVR-GWO = GMDH-GWO (AUC = 0.82), CNN = GMDH (AUC = 0.81), SVR = XGBoost (AUC = 0.80), and RULSE. Based on the RUSLE model, soil loss in the Noor-Rood watershed ranges from 0 to 2644 t ha^–1yr^?1.