基于不同学科解决粗差探测问题的方法研究

通过对已有粗差探测成果的研究,本文按照基于的学科不同将其大致分为4类:数理统计粗差探测方法,基于多元统计分析粗差探测方法,线性代数粗差探测方法,人工智能粗差探测方法。然后从这4个方面辐射出已有的关于粗差发现、定位、定值等相关方法的研究,并且针对国内学者对用人工智能方法实现粗差探测未进行成熟研究的现状,提出可行策略并进行实现从而为人工智能和测量数据处理2个学科的结合开拓新的研究方向。     

Paleoenvironmental interpretations of Quaternary reef deposits based on comparisons of 10 selected modern and fossil larger foraminifera from the Ryukyu Islands, Japan

Abstract A new method for reconstructing depositional environments of larger foraminifera‐bearing limestones is proposed. First, depth and spatial distributions of empty tests of 10 foraminiferal taxa in a 1–2 mm size fraction were examined using 32 surface sediment samples collected from depths shallower than 200 m, located to the west of Miyako Island, Ryukyu Islands, northwest Pacific. Distributional ranges of empty tests in the 1–2 mm size fraction appear to be more limited than those including other size fractions in previous reports, partly because larger empty tests of each taxon are less easily transported than smaller ones. Multivariate analyses (Q‐mode cluster analysis and non‐metric multidimensional scaling ordination) based on binary (presence/absence) data of the 10 taxa delineate four sample groups, each of which corresponds to different depositional environments: nearshore zone/bay inlet; back‐reef to fore‐reef; outer shelf to shelf slope; and shelf slope distant from coral reefs. Next, these modern data were applied to reconstruct the depositional environment of a rock section distributed in the Shiratorizaki area (Irabu Island, Ryukyu Islands), which consists of larger foraminiferal limestone of the Pleistocene Ryukyu Group. Multivariate analyses were performed on the fossil plus the modern foraminiferal data to explore the possible relationship of the fossil associations with variations in modern associations, demonstrating that the fossil foraminiferal associations resemble the modern outer shelf associations. The modern analog technique was also applied to estimate paleobathymetry using fossil foraminiferal data. The results indicate that the section studied was deposited in outer shelf environments at depths between 53.5 and 98.6 m. These paleoenvironmental interpretations are consistent with previous studies based on sedimentary facies and the computer‐based expert system. The modern dataset and methods used in our work would be particularly useful for paleoenvironmental reconstructions of Quaternary reef and shelf carbonates along active margins of the northwest Pacific.     

地下水流量预测的多元线性回归分析模型研究

地下水系统是一个复杂的随机系统,本文根据地下水流量与其影响因素之间存在的相关关系,建立了一个基于线性回归分析法的地下水流量预测模型,并将其用于遵义市海龙坝的地下水流量预测。结果表明预测精度较高,建立的模型较符合本研究区的实际情况。     

Multivariate analysis of landscape wildfire dynamics in a Mediterranean ecosystem of Greece

This paper focuses on spatial distribution of long-term fire patterns versus physical and anthropogenic elements of the environment that determine wildfire dynamics in Greece. Logistic regression and correspondence analysis were applied in a spatial database that had been developed and managed within a Geographic Information System. Cartographic fire data were statistically correlated with basic physical and human geography factors (geomorphology, climate, land use and human activities) to estimate the degree of their influence at landscape scale. Land cover types of natural and agricultural vegetation were the most influential factors for explaining landscape wildfire dynamics in conjunction with topography and grazing.     

多元统计分析在矿床指示元素组合特征研究中的应用——以山东黄埠岭金矿为例

应用相关分析、聚类分析和因子分析方法讨论了黄埠岭矿床中指示元素的组合特征。综合对该矿床具有代表性的7号矿脉和10号矿脉的分析结果，确定了黄埠岭金矿床Au的富集与Ag、Cu、Pb、Zn、Bi密切相关，矿体原生晕头晕组合为As、Sb、Hg，主成矿元素组合Au、Ag、Cu、Pb、Zn、Bi，尾晕元素组合为Co、Ni。为矿床的深边部成矿预测提供参数。     

Evaluation of geochemical data acquired from regular grids

Geochemical data obtained during mineral exploration often are biased by systematic as well as random errors; these may result in failures when usual methods of evaluation are used. This is true particularly in soil surveys carried out in regions where a long history of prospecting and mining activity has occurred and/or where aerial chemical pollution is likely to have occurred.A satisfactory evaluation of geochemical data even in such an unfavorable case requires sampling on a relatively dense grid and utilization of all available knowledge of types of mineralization. The evaluation procedure proposed consists of five consecutive phases: (1). Dividing the area of interest into subareas of a relatively homogeneous geological nature. (2). Processing by multivariate methods (factor analysis, in particular) without consideration of geographic relations. (3). A preliminary interpretation and search for a geochemical explanation of factors. (4). Processing of individual factors in two-dimensional geographic space by directional and frequency linear filtering methods. (5). Final interpretation and construction of a geochemical model. The procedure is illustrated by an example from a geochemical exploration survey in the vicinity of Píbram (Middle Bohemia).     

Multivariate normality in morphometric analysis

Data on some living (salamanders and grasshoppers) and fossil (Devonian brachiopods) animals are analyzed by means of recently developed methods for the large-scale treatment of multivariate normality. Multivariate nonnormality was found to exist in all situations, even if the univariate deviations in the skewness and kurtosis statistics proved to be without significance for the most part. The effect of logarithmically transforming the data appears to be a mixed blessing. Apart from the fact that the investigator is removed one step from the biological relationships in his data by carrying out a transformation of them, the betterment in the multivariate interconnections with respect to normality tends to be slight, despite the general improvement in the univariate values. The relationship between sample size and the multivariate normality measures b_1,p and b_2,p are studied empirically.     

Multivariate morphometrics and analysis of shape

Analysis of size and shape has always figured largely in interpretations of multivariate analyses of measurements on organisms, although on premises that are not always unchallengeable. A recent surge of interest has led to important advances in formal analysis of variations in shape (geometric morphometrics) and phylogenetic significance of such changes. In connection herewith, an improved philosophy for the choice of variables suitable for depicting shape variation is evolving. Important results for statistical validity of multivariate analyses now are being made available by several workers (robust estimates of canonical vectors, etc., stability of eigenvectorial elements, effects of deviations from multivariate normality) with special reference to applications in morphometric studies. Such questions have been little considered in the past with the result that many multivariate analyses of biological and geological data are inexact.Contribution Number 2 in a series of review articles by the Mathematical Geologists of United States. Accepted January 8, 1985. Invited lecture delivered to the thematic conference of Geological Shape Analysis, Wood's Hole, Massachusetts, September 8–12, 1984.     

On the Use of Multivariate Lévy-Stable Random Field Models for Geological Heterogeneity

Increasing attention has been paid to the use of non-Gaussian distributions as models of heterogeneity in sedimentary formations in recent years. In particular, the Lévy-stable distribution has been shown to be a useful model of the distribution of the increments of data measured in well logs. Frequently, the width of this distribution follows a power–law type scaling with increment lag, thus suggesting a nonstationary, fractal, multivariate Lévy distribution as a useful random field model. However, in this paper we show that it is very difficult to formulate a multivariate Lévy distribution with any nontrivial spatial correlations that can be sampled from rigorously in large models. Conventional sequential simulation techniques require two properties to hold of a multivariate distribution in order to work: (1) the marginal distributions must be of relatively simple form, and (2) in the uncorrelated limit, the multivariate distribution must factor into a product of independent distributions. At least one of these properties will break down in a multivariate Lévy distribution, depending on how it is formulated. This makes a rigorous derivation of a sequential simulation algorithm impossible. Nonetheless, many of the original observations that spurred the original interest in multivariate Lévy distributions can be reproduced with a conventional normal scoring procedure. Secondly, an approximate formulation of a sequential simulation algorithm can adequately reproduce the Lévy distributions of increments and fractal scaling frequently seen in real data.