基于数据科学的矿产资源定量预测的理论与方法探索

矿产资源预测已从定性走向了定量,从数据稀疏型走向了数据密集型,亟须数据科学支撑。本文在前人研究基础上,讨论了基于数据科学的矿产资源定量预测理论与方法,该方法的理论基础为相关性理论与异常理论,前者采用监督的机器学习方法挖掘地质找矿大数据与矿床的相关性为预测未发现矿床提供了理论基础;后者采用非监督的机器学习方法识别地质找矿大数据蕴含的地质异常为预测矿床提供了理论依据。该理论与方法强调地质找矿大数据和机器学习的重要性,其中,数据种类的多样性及数据精度和质量会影响预测结果的好坏,机器学习可提高特征提取与信息集成融合效率。此外,本文讨论了基于数据科学的矿产资源定量预测理论与方法的技术框架、特征提取、数据集成融合方法,以及该理论与方法引入的不确定性。     

磁铁矿元素地球化学大数据构建及其在矿床成因分类中的应用

磁铁矿广泛分布在岩浆、热液及沉积等各类矿床中,其地球化学元素组成往往受温度、氧逸度等物理化学条件的影响,能反映矿床形成环境并指示矿床成因类型,是一种重要的勘查指示矿物。自20世纪60年代以来,磁铁矿的主微量元素数据被用来构建不同的判别图,试图来区分矿床的成因类型。然而,由于矿床成因类型的多样性以及同一类型矿床的磁铁矿的主微量元素地球化学组成的复杂性,以往基于少数磁铁矿的主微量元素地球化学成分构建的矿床成因类型判别图存在一定的局限性。基于此,本文收集了前人发表在国内外期刊上的主要矿床类型的磁铁矿的元素地球化学数据(7 388条),初步构建了基于电子探针(EPMA)和激光剥蚀-电感耦合等离子体质谱(LA-ICP-MS)磁铁矿元素地球化学大数据集,建立了基于随机森林算法的矿床成因分类模型,并对磁铁矿主微量元素在矿床成因分类中的重要性做出排序。研究结果表明,基于磁铁矿大数据和机器学习算法构建的判别模型,能有效区分主要矿床类型,整体分类准确度高达95%。由于LA-ICP-MS磁铁矿数据集的测试元素多,分析精度高,使得基于LA-ICP-MS磁铁矿数据集的矿床成因分类模型精度高于基于EPMA数据集,表明磁铁矿中元素种类多少和数据测试精度影响矿床成因分类精度。同时,研究发现V元素在矿床成因分类过程中起到了较为重要的作用。此外,基于大数据和机器学习建立的判别模型对新的磁铁矿数据进行测试,可给出该数据属于每种矿床类型的概率,能有效判别矿床成因类型。     

Application of fractal models to characterization of vertical distribution of geochemical element concentration

Characterization of the vertical distribution of geochemical element concentration is essential for economic planning in the mining industry. 10 mineralized boreholes and 1 non-mineralized borehole from the Qulong copper deposit, Tibet, western China, were collected to identify the vertical distribution properties of Cu values using fractal models. The vertical distribution of Cu values in mineralized and non-mineralized boreholes shows a positive skewed distribution in the former and multimodal distribution in the latter. The results obtained by the box counting method show that the vertical distributions of Cu values in mineralized and non-mineralized boreholes exhibit self-similarity with box dimensions ranging from 1.28 to 1.37. The box dimensions of mineralized boreholes are greater than that of Cu values in the non-mineralized borehole, indicating that the mineralization makes the distribution of Cu values more irregular. The power-law frequency analysis reveals that Cu values in mineralized boreholes are bifractal. The two portions of the plot define a crossover point at 0.33%, for Cu values less than and greater than 0.33, fractal dimensions range from 0.1 to 0.65, in non-mineralized rocks, and range from 2.71 to 5.79, in the mineralized rocks. Hurst exponents for mineralized boreholes occur at 0.8, which are greater than 0.5, indicating that Qulong copper deposit has a good continuity of mineralization.     

Evaluation of the uncertainty in estimation of metal resources of skarn tin in Southern China

This study is concerned with the problem of how many undiscovered mineral deposits can be expected to occur in the vicinity of any known deposit, especially if the deposit is nearly mined out. Skarn tin deposits in southern China were chosen to demonstrate that fractal modeling can be a useful tool to characterize the spatial–temporal distribution of mineral deposits, and to quantify their grades and tonnages. The results show that the spatial–temporal distribution of skarn tin deposits as well as their grades and tonnages satisfy fractal statistic, and suggest that 14 skarn tin deposits could be found around a known skarn tin deposit within a radius of 80 km. Monte Carlo simulation was used to combine the number of deposits and the frequency distributions of grade and tonnage and to capture the uncertainty in estimation of metal resources. At the 90%, 50% and 10% confidence levels, tin metal resources amount to 6 ton, 200 ton, and 1.0 × 10⁴ ton around a known tin deposit within a radius of 80, respectively.     

中国东部夏季雨带类型与前期北半球500 hPa环流异常的关系

根据1951～1986年的资料,在文献[1]对中国东部夏季（6～8月）雨带分布类型的基础上,分析了各雨型与前期北半球500 hPa环流异常的关系。重点研究了各雨型的长期天气过程。发现不同雨型前期环流有不同的长期演变过程。指出中高纬和低纬环流不同季节变异及其相互作用,可能是导致中国东部各种雨带分布类型的一个重要因素。并提出了一些预报线索,可供夏季我国大范围旱涝趋势的长期预报参考。     

Causes and Predictability of the 2021 Spring Southwestern China Severe Drought

In the spring of 2021, southwestern China (SWC) experienced extreme drought, accompanied by the highest seasonal-mean temperature record since 1961. This drought event occurred in the decaying phase of a La Ni?a event with negative geopotential height anomalies over the Philippine Sea, which is distinct from the historical perspective. Historically, spring drought over SWC is often linked to El Ni?o and strong western North Pacific subtropical high. Here, we show that the extreme drought in the spring of 2021 may be mainly driven by the atmospheric internal variability and amplified by the warming trend. Specifically, the evaporation increase due to the high temperature accounts for about 30% of drought severity, with the contributions of its linear trend portion being nearly 20% and the interannual variability portion being about 10%. Since the sea surface temperature forcing from the tropical central and eastern Pacific played a minor role in the occurrence of drought, it is a challenge for a climate model to capture the 2021 SWC drought beyond one-month lead times.     

太平洋各区海温异常对中国东部夏季雨带类型的共同影响

根据1951～1986年的海温资料,作者在对中国东部夏季（6～8月）雨带分布分型的基础上,分析了各雨型与同期和前期各季太平洋（60°N～50°S, 120°E～70°W）海温异常的关系。分析表明,对应不同雨型,无论同期或前期整个太平洋海温距平分布的基本形势都有明显不同的特点,特别是西北太平洋黑潮暖流和热带太平洋赤道冷水带海温的异常。这两个海区的海温变化明显,而且与其周围（西北太平洋呈“>”形,热带太平洋呈“<”形）的海温存在反相的变化。这种结构既反映了海温的纬向热力差异又反映了海温的经向热力差异。指出西北太平洋和热带太平洋各种海温距平不同配置的共同影响,是导致中国东部各种雨带分布类型的一个重要因素。分析还表明,各雨型海温距平的配置有很强的持续性,而且与前期各主要洋流区海温的变异有关。根据前期太平洋大范围海温距平分布的特征和演变的过程以及ENSO预测,可为夏季我国大范围旱涝趋势的长期预报提供依据。     

Shortwave radiation�CSST relationship over the mid-latitude North Pacific during boreal summer in climate models

Over the mid-latitude North Pacific, there is a close relationship between interannual variations of the sea surface temperature (SST) and surface shortwave radiation during boreal summer. The present study evaluates this relationship in coupled model simulations, forced model simulations, and retrospective forecasts. It is found that the simulation of this relationship in climate models is closely related to the model biases in the meridional gradients of mean SST and surface shortwave radiation. A southward shift in the region of large mean meridional gradients leads to a similar southward shift in the region of large correlation between the SST and shortwave radiation variations. The relationship is enhanced (weakened) when the mean meridional gradients are stronger (weaker) compared to observations. The shortwave radiation?CSST correlation is weak in individual forced simulations because of the interference of internally generated shortwave radiation variations. The shortwave radiation?CSST correlation increases significantly in the ensemble mean due to reduction of internally generated variability. The long-lead Climate Forecast System (CFS) forecasts have a better simulation of the shortwave radiation?CSST correlation compared to the short-lead forecasts. Estimation based on the CFS ensemble forecasts indicates that the high-frequency atmospheric variations contribute importantly to the SST variability over the mid-latitude North Pacific during boreal summer.     

Caribbean Sea rainfall variability during the rainy season and relationship to the equatorial Pacific and tropical Atlantic SST

The present study investigates the Caribbean Sea rainfall variability during the early and late rainy seasons and its association with sea surface temperature (SST) and air?Csea interaction based on observational estimates, the NCEP Climate Forecast System (CFS) and Global Forecast System (GFS) simulations, and the CFS retrospective forecasts. Analysis of the observational estimates indicates that air?Csea interaction is important over the Caribbean Sea, whereas the atmospheric forcing of SST dominates over the Gulf of Mexico. The CFS simulation captures the basic elements of this observed air?Csea relationship. The GFS simulation produces spurious SST forcing of the atmosphere over the Gulf of Mexico largely due to prescribing SST. The CFS forecasts capture the air?Csea relationship in the late rainy season (August?COctober), but cannot reproduce the SST forcing of atmosphere over the Caribbean Sea in the early rainy season (May?CJuly). An empirical orthogonal function (EOF) analysis indicates that the leading modes of percent anomalies of the rainy season precipitation have the largest loading in the southern Caribbean Sea in observations. The model simulations and forecasts skillfully reproduce the spatial pattern, but not the temporal evolution. The Caribbean Sea rainfall variability in the early rainy season is mainly due to the tropical North Atlantic (TNA) SST anomalies in observations, is contributed by both the TNA and eastern equatorial Pacific (EEP) SST anomalies in the CFS simulation, and has an overly large impact from the EEP SST anomalies in the GFS simulation and the CFS forecasts. The observed Caribbean Sea rainfall variability in the late rainy season has a leading impact from the EEP SST anomalies, with a secondary contribution from the TNA SST anomalies. In comparison, the model simulations and forecasts overestimate the impacts of the EEP SST anomalies due to an earlier development and longer duration of the El Ni?o-Southern Oscillation in the CFS compared to observations.