机器学习与成矿预测:以闽西南铁多金属矿预测为例

作为近年来爆炸式发展的方法模型,机器学习为地质找矿提供了新的思维和研究方法。本文探讨矿产预测研究的理论方法体系,总结机器学习在矿产预测领域的特征信息提取和信息综合集成两个方面的应用现状,并讨论机器学习在矿产资源定量预测领域面临的训练样本稀少且不均衡、模型训练中缺乏不确定性评估、缺少反哺研究、方法选择等困难和挑战。进一步以闽西南马坑式铁矿为实例论述基于机器学习方法的矿产预测基本流程:(1)通过成矿系统研究建立成矿模型,确定矿床控矿要素;(2)通过勘查系统研究建立找矿模型,并为评价预测提供相关的勘查数据;(3)通过预测评价系统研究,建立预测模型,并提取预测要素;(4)利用机器学习模型对预测要素进行信息综合集成,获取成矿有利度图;(5)对预测性能和结果进行不确定性评估;(6)找矿靶区/成矿远景区圈定及资源量估算。最后,总结建立以地学大数据和地球系统理论为指导,以“地球系统-成矿系统-勘查系统-预测评价系统”为研究路线的基于地学大数据的矿产资源定量预测理论和方法体系的研究愿景。     

层次分析法在矿床大比例尺定位预测中的应用--以铜陵凤凰山铜矿为例

通过铜陵凤凰山铜矿成矿规律研究,利用找矿分析成果和多元找矿信息,运用层次分析法,开展隐伏矿床大比例尺定位预测研究,建立了矿床定位预测模型,并对预测单元进行了成矿有利度计算和评价。已知矿体和工程验证结果表明,该方法可较好地用于大比例尺定位预测分析和成矿预测单元的分级评价。     

贵州铜仁地区锰矿定量预测与评价

应用地质、化探、物探、遥感等手段和资料进行多元信息成矿预测已成为找矿工作的重要方法。以贵州铜仁地区锰矿为研究对象,基于研究区与成矿相关的多元信息,建立区域找矿模型;运用GIS技术对多元信息进行综合处理并提取有利成矿信息,形成找矿标志专题图层,构成找矿预测指标集;利用证据权重法建立锰矿综合预测模型,结合研究区成矿地质条件、已知锰矿床（点）等因素,确定找矿有利地区,并圈定7个远景区,为发现大规模锰矿产地提供参考。     

云南澜沧坝老铅多金属矿综合信息成矿预测

本文基于矿床地质特征研究的基础上,采用“MORPAS”软件平台,对矿区地质、化探、遥感等地学多元信息进行综合分析,厘定了找矿模型,开展了研究区的综合信息成矿预测,圈定了4个找矿靶区,并根据矿床值随各控矿因素及其组合的变化规律,选取一定的模型单元,建立控矿因素及找矿标志的综合值——信息量与资源量之间的函数关系,并对圈定的4个靶区的资源量进行定量估算,对研究区进行资源潜力评价,对该区下一步的找矿工作具有重要的指导意义.     

初论综合信息成系列预测

综合信息成矿系列预测系指在成矿系列理论指导下,研究成矿系列产出的地质背景特征、成矿作用特征和矿床共生组合规律,建立成矿系列的区域成矿模式;以成矿模式为基础,统计对比分析同一成矿系列已知矿床的地质标志、地球化学标志、地球物理和遥感地质标志,建立以成矿系列综合标志信息为特征的综合信息成矿系列找矿模型;运用找矿模型实现对矿化单元的成矿系列预测.     

综合信息成矿系列预测的基本原理和方法

文章主要论述了综合信息区域成矿预测的基本原理和方法，基本原理包括：成矿学原理、地球化学原理、地球物理和遥感地质原理。文章认为，地球化学信息对任何矿种都是直接找矿信息，其信息特征在很大程度上直接反映了矿产资源体的特征，从区域综合异常评价的角度，综合3住处成矿系列预测，实质上是运用数学模型对地球化学异常及其产出的地质背景（地层、岩体和构造）的综合定量评价，从而提高了异常评价的有效性。基本方法主要包括综合信息成矿系列预测图的编制和综合信息成矿系列找矿模型建立。文中强调在成矿系列理论指导下，通过对异常及其产出地质背景的综合分析，以地质体或矿产资源体为单元，进行信息提取、关联和转换，最终建立综合信息成矿系列找矿模型，通过找矿模型这民矿系列的定量定位预测。     

三维成矿预测数据整合过程不确定性研究——以宁芜盆地钟姑矿田为例

基于多元地学大数据的三维成矿预测方法是开展深部找矿预测的新方法和新手段,也是当前成矿预测领域的研究热点之一。然而,大数据具有高维、混杂、非精确等特点,其分析处理过程面临多重不确定性。多元地学大数据整合是三维成矿预测的最终环节,其存在的不确定性将直接作用于预测结果,影响进一步的找矿应用和风险评估。本文以宁芜盆地钟姑矿田为例,从大数据思维出发,定量分析和度量预测要素和数学模型在数据整合过程中存在的不确定性及对三维成矿预测结果的影响。结果显示,断裂构造、背斜轴部等预测要素的不确定性对三维成矿预测结果的影响最为强烈;数据整合模型中,较之Logistic回归模型和证据权重模型,神经网络模型可能具有更高的不确定性程度。进一步工作可通过增强上述预测要素的可靠性和有效性、采用更多的数据整合模型进行更为全面的不确定性分析和评价,以获得更为可靠的三维成矿预测成果,从而降低成矿预测和找矿勘探风险。     

内蒙古额济纳旗红石山金矿三维成矿条件分析与成矿预测

分析了内蒙古额济纳旗红石山金矿床地质特征和控矿地质条件,总结了研究区成矿规律及找矿标志在深部的变化规律,建立了该区域的找矿地质模型,利用Surpac软件对区内地层、断裂构造、岩体、已知矿体、物化探异常等成矿预测因子进行了三维实体建模;采用立方体预测模型法对三维实体模型进行空间分割,并依据建立的找矿地质模型,给每一个立方体单元块赋相应属性值,建立数字找矿模型;进而对研究区地质、地球物理、地球化学等9个预测变量进行三维成矿有利条件分析与提取,采用三维信息法计算并统计各预测变量所包含的信息量,确定成矿有利组合的空间部位,并结合研究区内实际地质情况、已有见矿工程分布等因素圈定了3处找矿靶区;最后采用体积估计法计算得到3处找矿靶区的总资源量为4.67 t。     

模糊综合评判模型在会泽铅锌矿隐伏矿定位预测中的应用

模糊综合评判(FCA)模型是基于模糊数学理论并结合隐伏矿预测理论而建立的一个找矿模型.在会泽铅锌矿隐伏矿定位预测中,运用FCA模型,进行了成矿因素评价集的确定,评价因素权重的确定,并根据流程作出找矿预测的异常图.预测结果与实际工程施工验证相吻合,证明所构造的模型有效.该模型还可以推广到其他矿山进行隐伏矿定位预测.     

基于证据权法的安徽东溪金矿大比例尺成矿预测与评价

应用GIS技术结合地质、地球物理、地球化学等资料进行综合信息成矿预测是目前找矿的趋势。以霍山县东溪金矿及外围1∶10 000大比例尺地区为研究对象,在野外地质调查的基础上,结合空间位置统计结果和矿区成矿特征,利用GIS平台提取地质、地球物理、地球化学等多元地质找矿信息,建立矿区找矿预测模型。选择符合条件的独立预测因子作为证据图层,以5m×5m为网格单元,分析各证据图层与矿体空间分布关系。运用证据权法对矿区及外围进行成矿预测,通过图形方式直观表达成矿高概率的空间分布,对圈定的找矿靶区进行综合评述。研究区找矿靶区内地质、物探、化探找矿信息良好,各靶区内均有较好的金矿化信息。     

Determination of earthquake-risky areas in Yalova City Center (Marmara region,Turkey) using GIS-based multicriteria decision-making techniques (analytical hierarchy process and technique for order preference by similarity to ideal solution)

Disaster mitigation as a pre-disaster measure within the scope of disaster risk management is significant in the sense of reducing the adverse effects of earthquakes in the context of earthquake-sensitive risk planning. In the urban planning context, the existence of numerous decision makers and alternatives, which are depending on many criteria, makes decision-making process difficult. This difficulty was overcomed through geographical information systems (GIS). In the context of GIS-based multicriteria decision-making (MCDM) analysis, we used analytic hierarchy process (AHP) and technique for order preference by similarity to ideal solution (TOPSIS) to determine earthquake-risky areas in Yalova City Center. First, AHP analysis related to geological and superstructure/infrastructure criteria was conducted and two separate AHP maps were obtained. Then, we conducted TOPSIS analysis to consider both criteria in the sense of earthquake risk-sensitive planning. Then, overall earthquake risk map obtained which could be used as an input for disaster mitigation processes.

     

Monte Carlo simulation-aided analytical hierarchy process (AHP) for flood susceptibility mapping in Gabes Basin (southeastern Tunisia)

Flash floods are among the most severe hazards which have disastrous environmental, human, and economic impacts. This study is interested in the characterization of flood hazard in Gabes Catchment (southeastern Tunisia), considered as an important step for flood management in the region. Analytical hierarchy process (AHP) and geographic information system are applied to delineate and characterize flood areas. A spatial database was developed based on geological map, digital elevation model, land use, and rainfall data in order to evaluate the different factors susceptible to affect flood analysis. However, the uncertainties that are associated with AHP techniques may significantly impact the results. Flood susceptibility is analyzed as a function of weights using Monte Carlo (MC) simulation and Global sensitivity analysis. AHP and MC–AHP models gave similar results. However, compared to AHP approach, MC–AHP confidence intervals (95%) of the overall scores had small overlaps. Results obtained were validated by remote sensing data for the zones that showed very high flood hazard during the extreme rainfall event of June 2014 that hit the study basin.     

Exploring spatial autocorrelation of life expectancy in Poland with global and local statistics

The paper presents an exploratory analysis of the geographical patterns of life expectancy at birth by sex in Poland. For the purpose of spatial statistical analysis the geographical variation of life expectancy is partitioned into two components: regional (global) trend and the local variation (or residuals). The spatial autocorrelations of the residuals from the trend surface analysis are examined using global and local statistics. The results indicate that there are statistically significant geographical clusters of high and low life expectancies for men and women in Poland.     

Delineation of potential groundwater zones based on multicriteria decision making technique

Groundwater is the most prioritized water source in India and plays an indispensable role in India's economy. The groundwater potential mapping is key to the sustainable groundwater development and management. A hybrid methodology is applied to delineate potential groundwater zones based on remote sensing, geographical information systems(GIS) and analytic hierarchy process(AHP) as on multicriteria decision making. For the purpose of demonstrating field application, Chittar watershed, Tamilnadu, India is studied as an example. The important morphological characteristics considered in the study are lithology, geomorphology, lineament density, drainage density, slope, and Soil Conservation Service–Curve Number(SCS-CN). These six thematic layers are generated in a GIS platform. Based on intersecting the layers, AHP method, the values for adopting the pairwise comparison normalized weight and normalized subclasses weightage were given. The normalized subclass weightage is input into each layer subclass. Then, weighted linear combination method is used to add the data layers in GIS platform to generate groundwater potential Index(GWPI) map. The GWPI map is validated based on the net recharge computed from the differences of measured groundwater levels between the pre-monsoon and post-monsoon in the year 2018. The kappa statistics are used to measure level spatial consistency between the GWPI and net recharge map. The overall average spatial matching accuracy between the two data sets is 0.86, while the kappa coefficient for GWPI with net recharge, 0.78. The results show that in Chittar watershed about 870 km~2 area is divided into high potential zone(i.e. sum of very high and high potential zone), 667 km~2 area, as the moderate one and the rest 105 km~2 area, as the poor zone(i.e. sum of very poor and poor potential zone).     

An interval matrix method used to optimize the decision matrix in AHP technique for land subsidence susceptibility mapping

The analytical hierarchy process (AHP) is one of the most effective methods for criteria ranking/weighting to have been successfully incorporated into GIS analyses. We present a new method for optimizing pairwise comparison decision-making matrices in AHP method, which has been developed on the basis of an interval pairwise comparison matrix (IPCM) derived from expert knowledge. The method has been used for criteria ranking in land subsidence susceptibility mapping (LSSM) as a practical test case, for which an interval matrix was generated by pairwise comparison. To compare the capability of the AHP method (a traditional approach) with that of the proposed IPCM method (a novel approach), 11 creations of LSSM were ranked using each approach in turn. The criteria weightings obtained were then used to produce LSSM maps based on each of these approaches. The results were tested against a data set of known land subsidence occurrences, indicating an improvement in accuracy of about 14% in the LSSM map that was developed using the IPCM method. This improvement was achieved by minimizing the uncertainty associated with criteria ranking/weighting in a traditional AHP and could form a basis for future research into minimizing the uncertainty in weightings derived using the AHP method. Our results will be of considerable importance for researchers involved in GIS-based multi-criteria decision analysis (MCDA) and those dealing with GIS-based spatial decision-making methods.     

GIS与遥感支持下基于层次分析法的土壤水资源分布研究——以石家庄市为例

开发了一个在GIS与遥感支持下基于层次分析法获得区域土壤水资源分布的新方法。本方法首先根据待研究区的特征选取土壤水影响因子,用层次分析法确定因子权重,建立土壤水资源评价模型,以2000年石家庄市为例,选取6个土壤水影响因子,分别是降水、蒸发、海拔、坡度、土壤孔隙度、土地利用方式,最后基于GIS的空间分析功能实现了石家庄市土壤水资源分级。该方法可为快速获取区域土壤水资源分布提供新的途径。     

Proximity,subjectivity, and space: Rethinking distance in human geography

This paper critically reviews the current status of the concept of distance in human geography in order to argue that recent experimentally-driven work in construal-level theory offers ample opportunities for recasting distance as a key geographical trope. After analysing the four entangled dimensions of distance revealed by construal-level theory (spatial distance; temporal distance; social distance; and hypothetical distance), the paper articulates this research program from experimental psychology with geographical work on non-representational theory, geographical imaginations/imaginative geographies, learning as a geographical process, TimeSpace theorising, and ontogenetic understandings of space. It is argued that the subjective understanding of distance afforded by construal-level theory can rescue distance from its entrenched association with positivistic geography and spatial analysis.     

基于GIS和AHP的南通市主城区地下水地源热泵适宜性评价

将南通市主城区的地质和水文地质条件、地下水动力场、温度场与热物性、地下水化学场和环境地质等自然因素作为层次分析法(AHP)模型的评价因子,应用AHP获得各影响因子的权重值,并结合GIS的空间分析技术,将研究区地下水地源热泵适宜性分为适宜、较适宜和不适宜3个大区,其中适宜区和较适宜区占总调查区的97.78%。经检验,评价方法适合该研究区。     

基于地理因子的现代降水空间模拟方法探讨

将经度、纬度、海拔、坡度、坡向、分区(干燥度)和植被等七个因子作为影响现代降水的地理因子,结合R语言程序、Arcgis空间分析和SPSS统计分析工具,探讨对现代降水进行空间模拟时,其影响因子与重建方法的选择。基于广义可加模型(GAM)用R语言实现了对现代降水影响因子的非线性分析,得到除坡度外其它各因子均与多年平均降水量之间大体呈线性相关;继而对各地理因子进行共线性检验,通过检验的因子使用最小二乘法建立与多年平均降水之间的多元线性回归模型。计算结果表明:逐次叠加地理因子进行回归分析时,其方程的R2均有提升,显示地理因子的类型对降水空间模拟具有一定的影响;非线性和线性建模中坡向因子的P值都大于0.05,说明在本研究中其对降水的影响不显著;线性回归建模的模型拟合优度R2为0.849,非线性模型的解释达到89.6%,两种建模结果都具有一定的可信度,对分析今后中国大范围区域中现代降水的影响因子及空间模拟方法具有重要的参考意义。     

Municipal solid waste landfill siting by using GIS and analytical hierarchy process (AHP): a case study in Qom city,Iran

Qom is the eighth most populated city in center of Iran, and its population growth rate is among the highest in this country. Th presence of a Great Salt Lake, petroleum potential and tourism attractions in this city sheds light on the importance of how solid waste landfill locations should be disposed, located and managed as an environmental issue. Considering the key parameters in landfill site selection, in this study a series of location analysis have been conducted to locate optimum regions for municipal solid waste disposal, using analytical hierarchy process (AHP) and geographical information system (GIS). The main factors in selecting the suitable location for waste disposal include geomorphology–hydrography, environmental–social factors and design criteria, each of which are subdivided into several categories. Criteria are selected according to the regional condition; therefore, important factors such as distance from sea and forested areas were not considered. In the next step, digital layers are weighted and classified according to the available standards and expert judgment. Then, analytical multi-criteria decision-making algorithms as AHP and weighted linear combination are applied upon existing layers in GIS. The results show that by implementing the AHP method in this region only 7% of the study area has a very good and appropriate condition for landfill location and the field observation confirms them. Finally, considering the environmental effects of landfill, appropriate locations are suggested.