The pillars of cosmology: a short history and assessment

Is the big bang model of the universe ascendant and unshakable, or declining and outdated? This paper cursorily explores the state of cosmology today.     

Asymmetric vegetation responses to mid-Holocene aridity at the prairie-forest ecotone in south-central Minnesota

The mid-Holocene (ca. 8000-4000 cal yr BP) was a time of marked aridity throughout much of Minnesota, and the changes due to mid-Holocene aridity are seen as an analog for future responses to global warming. In this study, we compare the transition into (ca. 9000-7000 yr ago) and out of (ca. 5000-2500 yr ago) the mid-Holocene (MH) period at Kimble Pond and Sharkey Lake, located along the prairie forest ecotone in south-central Minnesota, using high resolution (∼ 5-36 yr) sampling of pollen, charcoal, sediment magnetic and loss-on-ignition properties. Changes in vegetation were asymmetrical with increasing aridity being marked by a pronounced shift from woodland/forest-dominated landscape to a more open mix of grassland and woodland/savanna. In contrast, at the end of the MH, grassland remained an important component of the landscape despite increasing effective moisture, and high charcoal influxes (median 2.7-4.0 vs. 0.6-1.7 mm² cm^− 2 yr^− 1 at start of MH) suggest the role of fire in limiting woodland expansion. Asymmetric vegetation responses, variation among and within proxies, and the near-absence of fire today suggest caution in using changes associated with mid-Holocene aridity at the prairie forest boundary as an analog for future responses to global warming.     

新疆哈密五堡乡缺水地区找水经验

通过在新疆哈密五堡乡南大草滩严重缺水地区的水地质勘察，找到了一个日供水量为3000m^3/d的小型水源地，为在同类地区找水提供了一定的找水经验。     

Estimation of regional material yield from coastal landslides based on historical digital terrain modelling

High‐resolution historical (1942) and recent (1994) digital terrain models were derived from aerial photographs along the Big Sur coastline in central California to measure the long‐term volume of material that enters the nearshore environment. During the 52‐year measurement time period, an average of 21 000 ± 3100 m³ km^?1 a^?1 of material was eroded from nine study sections distributed along the coast, with a low yield of 1000 ± 240 m³ km^?1 a^?1 and a high of 46 700 ± 7300 m³ km^?1 a^?1. The results compare well with known volumes from several deep‐seated landslides in the area and suggest that the processes by which material is delivered to the coast are episodic in nature. In addition, a number of parameters are investigated to determine what influences the substantial variation in yield along the coast. It is found that the magnitude of regional coastal landslide sediment yield is primarily related to the physical strength of the slope‐forming material. Coastal Highway 1 runs along the lower portion of the slope along this stretch of coastline, and winter storms frequently damage the highway. The California Department of Transportation is responsible for maintaining this scenic highway while minimizing the impacts to the coastal ecosystems that are part of the Monterey Bay National Marine Sanctuary. This study provides environmental managers with critical background data on the volumes of material that historically enter the nearshore from landslides, as well as demonstrating the application of deriving historical digital terrain data to model landscape evolution. Published in 2005 by John Wiley & Sons, Ltd.     

大数据环境下内蒙古浩布高地区铅锌多金属矿智能矿产预测研究

传统机器学习算法已广泛应用于矿产预测,但面对地质大数据的高维稀疏、不平衡小样本等特性仍缺乏有效处理和分析的方法,设计适合地质大数据特点的机器学习算法是智能矿产预测亟需解决的新问题。本文以内蒙古浩布高地区的铅锌多金属矿产预测为例,提出了一种面向地质大数据的半监督协同训练矿产预测模型。首先对研究区地质找矿信息和地球化学异常信息进行定量分析,提取断裂构造、二叠系地层、燕山期侵入岩、地层与岩体接触带、围岩蚀变及Pb、Zn、Sn、Cu地球化学异常共9种找矿因子。然后利用递归特征消除法优选找矿因子组合,不包括Sn异常在内的8个找矿因子组合被选为最优组合。最后,利用支持向量机和随机森林算法作为基分类器进行半监督协同训练矿产预测,绘制成矿概率分布图。ROC曲线和预测度曲线分析结果表明,半监督协同训练模型的AUC值和预测效率都高于随机森林和支持向量机模型。研究结果也为大数据环境下的智能矿产预测提供了一种新的思路。     

卷积神经网络及其在矿床找矿预测中的应用——以安徽省兆吉口铅锌矿床为例

大数据人工智能地质学刚刚起步,基于大数据智能算法的地质研究是非常有意义的探索性实验。利用大数据和机器学习解决矿产预测问题,有助于人们克服不能全面考虑地质变量的困难及评估当前模型在已有数据中的可靠性。元素地表分布特征量主要受原岩成分、成矿作用影响和地表过程的影响,它们携带某些指示矿体就位的信息,即矿体在地下空间就位时在地表的响应,且未在地表过程中消失。以往的地球化学勘查工作仅仅识别异常,但未能发现矿体在地表响应的成矿特征量。本文以安徽省兆吉口铅锌矿床为例,通过机器学习,利用卷积神经网络算法,不断挖掘元素Pb分布特征与矿体地下就位空间的耦合相关性。经过1000次训练后,可以得到准确率0. 93,损失率0. 28的卷积神经网络模型。这种神经网络模型就是矿体在地下就位时元素在地表分布的响应,可以用来进行矿产资源预测。应用该模型对未知区进行预测,结果显示第53号区域具有很大概率存在尚未发现的矿体。     

大数据助地质腾飞:岩石学报2018第11期大数据专题“序”

文中提出,大数据有广义与狭义之分:狭义的大数据以4V特点为标志,而符合大数据三个技术取向的、采用全数据模式的、从数据出发的研究是广义的大数据研究。大数据为什么应运而生?是因为科学发展遇到了瓶颈,遇到了难以解决的问题,大数据不仅开辟了科学研究的新方法,新思路,还引发了对科学哲学的反思。文中强调从理论驱动模式到数据驱动模式的转变,是研究方法和研究思路一个巨大的转变,这种转变开辟了新的科学创新之路。文中指出,在大数据时代,凡是能够用数据化表述的学科才称之为科学,而不能用数据化表述的学科就不是科学,能否被数据化是科学与非科学的分水岭。文中讨论了矿床学研究的目的,认为矿床学研究应当专注于查明矿床形成的规律,指导矿床的找矿,提高经济价值。提出在矿床学研究中应当加强对相关关系的研究。一个矿床的成因大家究竟是如何关注的,与成矿有关的因素很多,成矿究竟与哪些因素有关,在许多情况下可能就是一个相关关系的命题,而大数据研究的就是相关关系。因此,大数据与矿床学研究的思路是天然相通的。     

全球N-MORB和E-MORB分类方案对比

N-MORB与E-MORB是大洋中脊玄武岩常用的分类,二者地球动力学意义不同,备受学术界关注。对于N-MORB与E-MORB的分类识别标志,不同作者有不同的见解。MORB中可以根据Rb/Nd≤0.15、K/Ti≤0.11、(La/Sm)_N≤0.8、K_2O/TiO_20.09、ΔNb=1.74+lg(Nb/Y)-1.92lg(Zr/Y)0、(La/Sm)_N1、100K_2O/TiO_2≤13等7种指标来识别N-MORB,否则为E-MORB。究竟何种标志区分效果较好、比较适合大多数MORB的情况?学术界对此还较少有人讨论。为此,本文尝试利用大数据方法,采用全球全体扩张中心数据,对上述7种标志进行对比,发现(La/Sm)_N1的标志比较适合大多数MORB的情况。为此,我们将(La/Sm)_N1和(La/Sm)_N≥1的所有数据,选取La、Ce、Pr、Nd、Sm、Eu、Gd、Tb、Dy、Ho、Er、Tm、Yb、Lu、Ba、Cs、Hf、K、Nb、Pb、Rb、Sc、Sr、Ta、Th、Ti、Tl、U、V、Y、Zr等31个元素,利用两两元素对数比值进行投图,并计算85%置信度的置信椭圆交叠率,共得出36856个元素对组合,根据最小交叠率的原则,得出使用稀土元素La、Ce、Pr、Sm和其他高场强元素Nb、Zr、Hf、Y之间的比值关系判别效果较好。我们又利用以上得出的8种元素进行投图判定,发现以La为分子或以La/Hf、La/Zr元素比值做为区分标志可以得出更好的结果。因此建议考虑应用以上元素之间的相关关系共同判定N-MORB与E-MORB。     

Revisiting the death of geography in the era of Big Data: the friction of distance in cyberspace and real space

Many scholars have argued that the importance of geographic proximity in human interactions has been diminished by the use of the Internet, while others disagree with this argument. Studies have noted the distance decay effect in both cyberspace and real space, showing that interactions occur with an inverse relationship between the number of interactions and the distance between the locations of the interactors. However, these studies rarely provide strong evidence to show the influence of distance on interactions in cyberspace, nor do they quantify the differences in the amount of friction of distance between cyberspace and real space. To fill this gap, this study used massive amounts of social media data (Twitter) to compare the influence of distance decay on human interactions between cyberspace and real space in a quantitative manner. To estimate the distance decay effect in both cyberspace and real space, the distance decay function of interactions in each space was modeled. Estimating the distance decay in cyberspace in this study can help predict the degree of information flow across space through social media. Measuring how far ideas can be diffused through social media is useful for users of location-based services, policy advocates, public health officials, and political campaigners.     

A hierarchical indexing strategy for optimizing Apache Spark with HDFS to efficiently query big geospatial raster data

ABSTRACT

Earth observations and model simulations are generating big multidimensional array-based raster data. However, it is difficult to efficiently query these big raster data due to the inconsistency among the geospatial raster data model, distributed physical data storage model, and the data pipeline in distributed computing frameworks. To efficiently process big geospatial data, this paper proposes a three-layer hierarchical indexing strategy to optimize Apache Spark with Hadoop Distributed File System (HDFS) from the following aspects: (1) improve I/O efficiency by adopting the chunking data structure; (2) keep the workload balance and high data locality by building the global index (k-d tree); (3) enable Spark and HDFS to natively support geospatial raster data formats (e.g., HDF4, NetCDF4, GeoTiff) by building the local index (hash table); (4) index the in-memory data to further improve geospatial data queries; (5) develop a data repartition strategy to tune the query parallelism while keeping high data locality. The above strategies are implemented by developing the customized RDDs, and evaluated by comparing the performance with that of Spark SQL and SciSpark. The proposed indexing strategy can be applied to other distributed frameworks or cloud-based computing systems to natively support big geospatial data query with high efficiency.