Possibilities of reflectance spectroscopy for the assessment of contaminant elements in suburban soils

Due to rapid economic development, high levels of potentially harmful elements are continuously being released into the suburban soils of the Nanjing area, China. Conventional methods for investigating soil potentially harmful element contamination based on raster sampling and chemical analysis are time consuming and relatively expensive. Reflectance spectroscopy within the visible-near-infrared region has been widely used to predict soil constituents due to its rapidity, convenience and accuracy. The objective of this study was to examine the possibility of using soil reflectance spectra as a rapid method to simultaneously assess contaminant metals (Ni, Cr, Cu, Hg, Pb, Zn) and As in the suburban soils of the Nanjing area. One hundred and twenty soil samples were collected for chemical analyses and spectral measurements. Prediction of contaminant elements was achieved by a partial least-square regression (PLSR) approach. According to their relationships with Fe, the seven contaminant elements could be categorized into two groups. The results showed that the prediction accuracy for Group II (Ni, Cr, Cu and Hg) was higher than that for Group I (Pb, Zn and As). This finding was consistent with the fact that the correlation coefficients between Group II and Fe were higher than that between Group I and Fe. It was concluded that element-sorption by spectrally active Fe oxides was the major mechanism by which to predict spectrally featureless contaminant elements. This conclusion was strengthened by the fact that the PLSR regression coefficients, which revealed the most important wavelengths for prediction, were attributed to absorption features of Fe oxides. Future study with real remote sensing data and field measurements are strongly recommended.     

基于高光谱的土壤养分含量反演模型研究

为实现土壤养分（有机质SOM、全氮TN、全磷TP、全硫TS）含量的快速测定,以建三江创业农场为例,对土壤原始反射率进行了一阶微分（FD）、倒数对数（RL）、倒数一阶微分（FDR）、多元散射校正（MSC）和连续统去除（CR）变换,分析6种光谱变量与土壤养分的相关性,将在α=0.01水平上显著相关的波段作为特征波段,运用多元逐步回归（SMLR）、偏最小二乘回归（PLSR）和BP神经网络（BPNN）三种分析方法分别建立有机质、全氮、全磷和全硫的高光谱预测模型,并利用决定系数（R²）、均方根误差（RMSE）和相对分析误差（RPD）对预测模型进行评价.结果显示,PLSR和BPNN建立的土壤养分含量预测模型均优于SMLR,能极好地预测有机质和全氮含量,同时具有粗略估算全硫含量的能力.三种方法中仅有CR-BPNN能对全磷含量进行粗略估算.对有机质、全氮、全磷和全硫预测效果最佳的模型及其验证集决定系数分别为：MSC-PLSR （0.86）、MSC-PLSR （0.75）、CR-BPNN （0.56）、FDR-BPNN （0.67）.     

Spectroscopic-based assessment of the content and geochemical behaviour of arsenic in a highly heterogeneous sulphide-rich mine waste dump

Soil pollution by arsenic is a serious environmental problem in many mining areas. Quick identification of the amount and extent of the pollution is an important basis for developing appropriate remediation strategies. In a case study, 55 soil samples were collected from a highly heterogeneous waste dump around the Sarcheshmeh copper mine, south east Iran. Samples’ visible and near-infrared (VNIR) reflectance spectra were measured, transformed to absorbance and then pre-processed using Savitzky–Golay first-derivative (FD) and Savitzky–Golay second-derivative (SD) transformation methods. The obtained spectra were then subjected to three regression models including principal component regression (PCR), partial least squares regression (PLSR) and support vector regression (SVR) for predicting arsenic concentration. The best prediction accuracies were obtained by SVR and PLSR methods applied on first-derivative pre-processed spectra with R ² values of 0.81 and 0.69, respectively. It was found that VNIR spectroscopy is a successful method for predicting As concentration in contaminated soils of the dumpsites. Study of the prediction mechanism showed that the intercorrelation between arsenic and spectral features of soil including iron oxy/hydroxides and clay minerals was the major mechanism enabling the prediction of arsenic concentration. However, higher values of correlation coefficients at ~460, ~560 and ~590 nm suggested the internal association between arsenic and iron minerals as the more important mechanism for prediction. This conclusion supported previous speciation studies conducted in the same waste dump using improved correlation analysis and chemical sequential extraction method.     

基于可见—近红外反射光谱的典型农田重金属污染风险分类研究

对于人为因素或自然因素造成的农田土壤重金属元素污染,需要进行大面积的土壤环境质量调查和分类管控,然而传统的采样测试方法存在工作量大、代价高等问题。可见—近红外（Vis-NIR）反射光谱是一种快速低成本获取土壤理化信息的手段。为研究Vis-NIR反射光谱预测模型划分土壤重金属污染风险类别的能力,文章以典型人为污染地区（浙江温岭）和典型地质高背景地区（广西横县）的390份农田土壤为样本,测定8种重金属元素（As、Cd、Cr、Cu、Hg、Ni、Pb和Zn）的含量和pH值,并测定土壤Vis-NIR光谱。使用偏最小二乘（PLS）和支持向量机（SVM）算法建立回归模型,对土壤重金属含量和pH值进行预测,并基于预测值进行土壤重金属污染风险分类。结果显示,温岭土壤主要污染元素Cd和Cu的光谱模型回归预测偏差（RPD）分别为1.23和1.19,预测机制与有机质有关。横县土壤主要污染元素As和Cd的RPD分别为1.98和1.93,预测机制与铁氧化物和粘土矿物有关。地质高背景土壤重金属与铁氧化物的正相关性普遍较强,使得光谱模型对重金属含量预测准确度较高。温岭和横县土壤pH值的光谱模型RPD分别为1.76和1.68。土壤重金属污染风险光谱分类的总体 准确度分别为75.0%~100%（温岭）和80.0%~100％（横县）。将Vis-NIR光谱与遥感技术相结合,对农田土壤重金属污染风险进行快速分类总体是可行的。     

Modelling of shear stiffness of unsaturated fine grained soils at very small strains

The shear modulus at very small strains (less than 0.001%) is an important parameter in the design of geotechnical structures subjected to static and cyclic loadings. Although numerous soil models are available for predicting shear modulus of saturated and dry soils, only a few ones can predict shear stiffness at very small strains of unsaturated soils correctly. In this study, a few unsaturated soil models are evaluated critically and compared with a newly developed model. This newly proposed model is verified by using measured shear modulus at very small strains for three different low plasticity fine grained soils available in the literature. It is found that this new model can predict shear modulus at very small strain resulting from an increase and a decrease in mean net stress at constant matric suction for low plasticity fine grained soils. Moreover, this model is able to give a reasonably good prediction on shear stiffness at very small strain during wetting of a collapsible unsaturated soil. In addition, the newly proposed model is illustrated to capture a consistent trend with experimental data of shear stiffness at very small strain for non-collapsible soils obtained during drying–wetting cycles. This evaluation revealed that the newly proposed model has better predictive capabilities than some earlier formulations of the same simplicity. In addition, the proposed model with fewer parameters has similar predictive capability as compared with a more complex model.     

基于粒子群优化支持向量机的矿区土壤有机质含量高光谱反演

为了监测复垦矿区土壤的有机质含量, 综合利用光谱分析、统计学习理论与方法以及智能优化理论与方法, 研究了矿区复垦土壤有机质含量与土壤光谱之间的关系, 在此基础上建立了土壤有机质含量高光谱反演模型, 实现土壤有机质含量定量检测.首先对原始土壤光谱数据进行预处理, 然后进行相关性分析, 提取450 nm、500 nm、650 nm、770 nm、1 460 nm和2 140 nm作为特征波段, 最后利用多元线性回归(multiple linear regression, MLR)、偏最小乘回归(partial least squares regression, PLSR)和粒子群优化支持向量机回归(particle swarm optimization support vector machine regression, PSO-SVM)方法建立了土壤有机质含量的高光谱定量反演模型, 并对模型进行验证.3种模型的验证结果如下: MLR、PLSR和PSO-SVM模型的R2分别为0.79、0.83和0.85, RMSE分别为5.26、4.93和4.76.实验结果表明, 无论从模型的稳定性还是预测能力上, PSO-SVM都要优于其他两个模型.      

Schwertmannite stability in acidified coastal environments

A combination of analytical and field measurements has been used to probe the speciation and cycling of iron in coastal lowland acid sulfate soils. Iron K-edge EXAFS spectroscopy demonstrated that schwertmannite dominated (43-77%) secondary iron mineralization throughout the oxidized and acidified soil profile, while pyrite and illite were the major iron-bearing minerals in the reduced potential acid sulfate soil layers. Analyses of contemporary precipitates from shallow acid sulfate soil groundwaters indicated that 2-line ferrihydrite, in addition to schwertmannite, is presently controlling secondary Fe(III) mineralization. Although aqueous pH values and concentrations of Fe(II) were seasonally high, no evidence was obtained for the Fe(II)-catalyzed crystallization of either mineral to goethite. The results of this study indicate that: (a) schwertmannite is likely to persist in coastal lowland acid sulfate soils on a much longer time-scale than predicted by laboratory experiments; (b) this mineral is less reactive in these types of soils due to surface-site coverage by components such as silicate and possibly, to a lesser extent, natural organic matter and phosphate and; (c) active water table management to promote oxic/anoxic cycles around the Fe(II)-Fe(III) redox couple, or reflooding of these soils, will be ineffective in promoting the Fe(II)-catalyzed transformation of either schwertmannite or 2-line ferrihydrite to crystalline iron oxyhydroxides.     

某水利枢纽厂房大型基坑开挖渗流研究

渗流,特别涉及到自由面或浸润线的确定是岩土工程的重点和难点问题。借鉴砂土坝的渗流分析方法,采用经典分段组合法理论得到基坑开挖边坡渗流简化模型Ⅰ。在此基础上将土层分层,应用土层分界面上的渗流折射定律,得到简化模型Ⅱ。分别求解两种模型得到正常水位和最大水位条件下的渗水量,与工程实际的渗水量进行对比发现,计算渗水量大于实际渗水量,这是由于忽略了围堰对渗水的阻隔作用。在正常水位条件下,模型Ⅱ的结果优于模型Ⅰ。     

Advanced Mathematical Models to Predict the Compaction Properties of Coarse-Grained Soils from Various Physical Properties

An essential task in the process of construction is the determination of compaction properties of soils. Many years of laboratory test experience strengthen our belief in the existence of predictive equations that govern the compaction characteristics of soils. An advanced mathematical model developed in this research in order to uncertain the governing equations. An advanced mathematical model developed in this research in order to uncertain the governing equations. Through a comparative study among a Multiple Linear Regression (MLR) model, an Artificial Neural Network (ANN) model, Extreme Learning Machine (ELM) and a Support Vector Machine (SVM) model, the best predicting model was determined. For this purpose, Six hundred and six (606) samples collected and split into a dataset used for training the models and another used for validation of the derived model. 8 neural networks with a varying number of hidden layers and a varying number of nodes in hidden layers were employed. In ELM 1 hidden layer with varying number of units were employed. It was found that the equations derived from the ELM models described the relationship with superiority over multiple regression, ANN and SVM models for Maximum Dry Density and MLR models described the relationship with superiority over ANN, ELM and SVM models for Optimum Moisture Content.     

The development of a multi-surface soil speciation model for Cd (II) and Pb (II): Comparison of two approaches for metal adsorption to clay fractions

The mobility of toxic metals in soils or sediments is of great concern to scientists and environmentalists since it directly affects the bioavailability of metals and their movement to surface and ground waters. In this study, a multi-surface soil speciation model for Cd (II) and Pb (II) was developed to predict the partition of metals on various soil solid components (e.g. soil organic matter (SOM), oxide mineral, and clay mineral). In previous study, the sorption of metal cations on SOM and oxide minerals has been evaluated by thermodynamically based surface complexation model. However, metal binding to soil clay fractions was normally treated in a simplistic manner: only cation exchange reactions were considered and exchange coefficient was assumed unity. In this study, the binding of metals onto clays was described by a two-site surface sorption model (a basal surface site and an edge site). The model was checked by predicting the adsorption behavior of Cd (II) and Pb (II) onto three selected Chinese soils as a function of pH and ionic strengths. Results showed that the proposed model more accurately predicted the metal adsorption on soils under studied condition, especially in low ionic strength condition, suggesting that adsorption of metals to soil clay fractions need to be considered more carefully when modeling the partition of trace elements in soils. The developed soil speciation model will be useful when evaluating the movement and bioavailability of toxic metals in soil environment.     

Abiotic reduction of uranium by Fe(II) in soil

Structural Fe(II) has been shown to reduce several oxidized environmental contaminants, including NO₃, chlorinated solvents, Cr(VI), and U(VI). Studies investigating reduction of U(VI) by soils and sediments, however, suggest that abiotic reduction of U(VI) by Fe(II) is not significant, and that direct enzymatic reduction of U(VI) by metal-reducing bacteria is required for U(VI) immobilization as U(IV). Here evidence is presented for abiotic reduction and immobilization of U(VI) by structural Fe(II) in a redoximorphic soil collected from a hillside spring in Iowa. Oxidation of Fe(II) in the soil after reaction with U(VI) was demonstrated by Mössbauer spectroscopy and reduction of U(VI) by the pasteurized soil using U L_III-edge X-ray absorption spectroscopy (XAS). XAS indicates that both reduced U(IV) and oxidized U(VI) or U(V) are present after U(VI) interaction with the Fe(II) containing soil. The EXAFS data show the presence of a non-uraninite U(IV) phase and evidence of the oxidized U(V) or U(VI) fraction being present as a non-uranyl species. Little U(VI) reduction is observed by soil that has been exposed to air and oxidation of Fe(II) to goethite has occurred. Soil characterization based on chemical extractions, Mössbauer spectroscopy, and Fe K-edge XAS indicate that the majority of Fe(II) in the soil is structural in nature, existing in clay minerals and possibly a green rust-like phase. These data provide compelling evidence for abiotic reduction of U(VI) by structural Fe(II) from soil near Fe-rich oxic–anoxic boundaries in natural environments. The work highlights the potential for abiotic reduction of U(VI) by Fe(II) in reduced, Fe-rich environments.     

Soil depth reconstruction for the assessment of the susceptibility to shallow landslides in fine-grained slopes

This paper proposes a methodology aimed at reconstructing the maximum thickness mobilized by shallow landslides in fine-grained soils with the aid of geological and geotechnical analyses. The methodology, implemented within a geographic information system (GIS) environment, is composed of two stages for map reconstruction and two stages for map validation. The first stage of map reconstruction is aimed at individuating the soil thickness on the basis of only topographical and geological analyses; the second stage improves the previously obtained map with the aid of morphological and geotechnical analyses that provide a thickness map usable for shallow landslide susceptibility assessment. This map is validated with the aid of both in situ investigations (stage I), and geotechnical models able to back-analyse shallow precipitation-induced landslides over a wide area (stage II). An application of the proposed methodology is provided for a test area of the Calabria region (southern Italy) that is representative of the Catanzaro Strait, where widely diffused shallow landslides in fine-grained soils systematically occur. The results highlight the usefulness and reliability of the geotechnical models when implemented with the aid of a database representative of fine-grained soils while a secondary role is played by in situ investigations that in the test site have been performed only in a few representative and accessible areas.     

Limitations of the ferrozine method for quantitative assay of mineral systems for ferrous and total iron

The quantitative assay of clay minerals, soils, and sediments for Fe(II) and total Fe is fundamental to understanding biogeochemical cycles occurring therein. The commonly used ferrozine method was originally designed to assay extracted forms of Fe(II) from non-silicate aqueous systems. It is becoming, however, increasingly the method of choice to report the total reduced state of Fe in soils and sediments. Because Fe in soils and sediments commonly exists in the structural framework of silicates, extraction by HCl, as used in the ferrozine method, fails to dissolve all of the Fe. The phenanthroline (phen) method, on the other hand, was designed to assay silicate minerals for Fe(II) and total Fe and has been proven to be highly reliable. In the present study potential sources of error in the ferrozine method were evaluated by comparing its results to those obtained by the phen method. Both methods were used to analyze clay mineral and soil samples for Fe(II) and total Fe. Results revealed that the conventional ferrozine method under reports total Fe in samples containing Fe in silicates and gives erratic results for Fe(II). The sources of error in the ferrozine method are: (1) HCl fails to dissolve silicates and (2) if the analyte solution contains Fe³⁺, the analysis for Fe²⁺ will be photosensitive, and reported Fe(II) values will likely be greater than the actual amount in solution. Another difficulty with the ferrozine method is that it is tedious and much more labor intensive than the phen method. For these reasons, the phen method is preferred and recommended. Its procedure is simpler, takes less time, and avoids the errors found in the ferrozine method.     

Predicting selenite adsorption by soils using soil chemical parameters in the constant capacitance model

The constant capacitance model, a chemical surface complexation model, was applied to selenite, Se(IV), adsorption on 36 soils selected for variation in soil chemical properties. The constant capacitance model was able to fit Se(IV) adsorption by optimizing one monodentate Se(IV) surface complexation constant and the surface protonation constant. A general regression model was developed for predicting these surface complexation constants for Se(IV) from easily measured soil chemical characteristics. These chemical properties were inorganic carbon content, organic carbon content, iron oxide content, aluminum oxide content, and surface area. The prediction equations were used to obtain values for the surface complexation constants for four additional soils, thereby providing a completely independent evaluation of the ability of the constant capacitance model to describe Se(IV) adsorption. The model’s ability to predict Se(IV) adsorption was quantitative on one soil and semi-quantitative on three soils. Incorporation of these prediction equations into chemical speciation-transport models will allow simulation of soil solution Se(IV) concentrations under diverse non-calcareous agricultural and environmental conditions without the requirement of soil specific adsorption data and subsequent parameter optimization.     

Capability of vis-NIR spectroscopy and Landsat 8 spectral data to predict soil heavy metals in polluted agricultural land (Iran)

Heavy metals are toxic elements that have hazardous effect on the environment. They cause soil pollution as a result of their toxicity, potential reactivity, and mobility in soils. There are so many methods for the measurement of heavy metal concentrations in soils and aquatic systems. The traditional methods used for detecting heavy metal distribution in soil involve laboratory analysis and raster sampling. Both of them are expensive and time-consuming for large areas. Remote sensing techniques are used for obtaining the earth’s surface information, and these techniques have been used in the investigations of heavy metal distributions in preliminary analysis of soils as a rapid method. Today, near-infrared reflectance spectroscopy (NIRS) of soil characteristics has been of interest as a significant object. The present investigation is focused on the detection of heavy metals in contaminated soils by the application of reflectance spectroscopy in the spectral range of 350 to 2500 nm. This study also discusses the circumstances of the applied current methods for the detection and estimation of arsenic (As), cadmium (Cd), nickel (Ni), and lead (Pb) in contaminated agricultural soils. In the first part of laboratory spectroscopy, estimations were done using heavy metal reflectance spectroscopy and partial least square regression (PLSR) approaches, while in the second part, the heavy metal estimations were done using soil organic carbon reflectance spectroscopy through the PLSR approaches. Similar to the tasks above, estimations of As, Cd, Ni, and Pb by using Landsat 8 images were done in the forms of direct and indirect methods and the distribution of heavy metals in the study area was determined. Finally, the results obtained using direct and indirect methods were compared with the wet chemical measurements of heavy metals and organic carbon. It was found that although the direct detection of heavy metals using the images of Landsat 8 produced more accurate results than the indirect detections, the results obtained from laboratory spectroscopy corresponded more with the results from atomic adsorption spectroscopy. On the other hand, based on the fact that the soil has a complex content, the use of nonlinear methods, such as artificial neural networks in predicting soil heavy metal contents, could be regarded as a trusted method.     

Assessing the propensity of landfill soils to undergo reductive iron dissolution

Naturally occurring iron from soil and aquifer sediments at waste disposal sites often becomes liberated into groundwater as a result of reductive dissolution. Research was conducted to evaluate an appropriate procedure for assessing a soil’s propensity to undergo iron reductive dissolution. Soil samples collected from waste disposal sites in Florida were characterized by pH, organic carbon content, total iron content, amorphous iron content, citrate-dithionite-bicarbonate extractable iron, and qualitative X-ray diffraction analysis, followed by a series of extraction tests designed to simulate the reductive dissolution process. Over a 30-day period, biological reducing tests released 13–260 mg/kg Fe(II) from soils, and a chemical reducing test released 2.2–178 mg/kg Fe(II) from soils. Soil amorphous iron content was shown to be the most effective parameter for assessment of iron reductive dissolution potential through standard soil characterization. These results suggest that biological reducing tests may be helpful for assessing long-term soil iron reductive dissolution potential, and that soil amorphous iron content provides a good indication of the potential for a soil to undergo reductive dissolution at a landfill site.     

非饱和上海软土水力和力学特性耦合弹塑性模拟

目前大多数非饱和土的弹塑性本构模型用非饱和击实土的试验结果进行验证,但现场其他类型的土,如沉积土经常有在非饱和状态下外部环境变化的情况。现有的非饱和土弹塑性模型是否适用于沉积土一类的现场土是需要研究的课题。进行非饱和上海第③层土的吸力控制排水排气三轴剪切试验,使用文中提出的能统一考虑非饱和土水力性状和力学性状的弹塑性本构模型,预测上述三轴试验结果,并与试验数据进行比较。比较结果显示,建立的本构模型能够很好地预测非饱和上海软土的水力和力学性质,说明该模型不仅可以适用击实土的预测,还能够很好地适用于其他类型非饱和土的水力和力学性质的模拟。     

Iron(III) reduction and phosphorous solubilization in humid tropical forest soils

Phosphorus is one of the nutrients most commonly limiting net primary production in soils of humid tropical forests, mainly because insoluble Al and Fe phosphates and strong sorption to Fe(III) (hydr)oxides remove P from the bioavailable pool. Recent field studies have suggested, however, that this loss may be balanced by organic P accumulation under a wet moisture regime (>3350 mm annual precipitation). It has been hypothesized that, as the moisture regime changes from dry to mesic to wet, periods of anoxic soil conditions increase in intensity and duration, depleting Fe(III) (hydr)oxides and releasing sorbed P, but also slowing organic matter turnover, thus shifting the repository of soil P from minerals to humus. Almost no quantitative information is available concerning the coupled biogeochemical behavior of Fe and P in highly weathered forest soils that would allow examination of this hypothesis. In this paper, we report a laboratory incubation study of the effects of biotic Fe(III) (hydr)oxide reduction on P solubilization in a humid tropical forest soil (Ultisol) under a wet moisture regime (3000-4000 mm annual rainfall). The objectives of our study were: (1) to quantify Fe(III) reduction and P solubilization processes in a highly weathered forest soil expected to typify the hypothesized mineral dissolution-organic matter accumulation balance; (2) to examine the influence of electron shuttling on these processes using anthraquinone-2,6-disulfonate (AQDS), a well-known surrogate for the semiquinone electron shuttles in humic substances, as an experimental probe; and (3) to characterize the chemical forms of Fe(II) and P produced under anoxic conditions, both with and without AQDS. Two series of short-term incubation experiments were carried out, one without AQDS and another with an initial AQDS concentration of 150 μM. We measured pH, pE, and the production of Fe(II), total Fe [Fe(II) + Fe(III)], inorganic P, total P (inorganic P + organic P), and biogenic gases (CO₂, H₂ and CH₄). The same positive correlation was found between soluble P release and soluble Fe(II) production throughout incubation, implying that reduction of Fe(III) solubilized P. The Fe(II) produced was mainly particulate, evidently due to the formation of Fe(II) solid phases. Thermodynamic calculations indicated that precipitation of siderite and, in the presence of AQDS, vivianite was favored under the anoxic conditions that developed rapidly in the soil suspensions. Inorganic soluble P released during incubation was very small, indicating that the soluble P produced was mainly in organic form, which is consistent with the hypothesis that P accumulates in soil humus. Our net CO₂ production, H₂ consumption, and Fe(II) production data all suggested that reductive dissolution of Fe(III) (hydr)oxides was a terminal electron-accepting process coupled both to H₂ consumption and organic C oxidation by the native population of microorganisms in the soil. Addition of AQDS accelerated the production of Fe(II) and the release of soluble P, while hastening the decline in H₂ gas levels and suppressing CH₄ production. However, throughout incubation, the same quantitative relationships between soluble Fe(II) and P, and between pE and pH, were found, irrespective of AQDS addition. Thus we conclude that, in our soil incubation experiments, added AQDS functioned with the native microbial population solely as an electron shuttle catalyzing Fe(III) reduction. Whether humic substances in the soil also can act as electron shuttles in this way is a matter for future investigation.     

Modeling soil collapse by artificial neural networks

The feasibility of using neural networks to model the complex relationship between soil parameters, loading conditions, and the collapse potential is investigated in this paper. A back propagation neural network process was used in this study. The neural network was trained using experimental data. The experimental program involved the assessment of the collapse potential using the one-dimensional oedometer apparatus. To cover the broadest possible scope of data, a total of eight types of soils were selected covering a wide range of gradation. Various conditions of water content, unit weights and applied pressures were imposed on the soils. For each placement condition, three samples were prepared and tested with the measured collapse potential values averaged to obtain a representative data point. This resulted in 414 collapse tests with 138 average test values, which were divided into two groups. Group I, consisting of 82 data points, was used to train the neural networks for a specific paradigm. Training was carried out until the mean sum squared error (MSSE) was minimized. The model consisting of eight hidden nodes and six variables was the most successful. These variables were: soil coefficient of uniformity, initial water content, compaction unit weight, applied pressure at wetting, percent sand and percent clay. Once the neural networks have been deemed fully trained its accuracy in predicting collapse potential was tested using group II of the experimental data. The model was further validated using information available in the literature. The data used in both the testing and validation phases were not included in the training phase. The results proved that neural networks are very efficient in assessing the complex behavior of collapsible soils using minimal processing of data. This revised version was published online in July 2006 with corrections to the Cover Date.     

Separating anthropogenic from natural anomalies in environmental geochemistry

Environmental geochemistry has attracted increasing interest during the last decade. In Sweden, geochemical mapping is carried out with methods that allow the data to be used in environmental research, including sampling plant roots and mosses from streams, soils and bedrock. These three sample types form an integrated strategy in environmental research, as well as in geochemical exploration. However, one problem that becomes prominent in geochemical mapping is to distinguish the signals derived from natural sources from those derived from anthropogenic sources. So far, this has mostly been done by using different types of samples, for example, different soil horizons. This is both expensive and time-consuming.We are currently developing alternative statistical solutions to this problem. The method used here is PLSR (partial least squares regression analysis). In this paper, we present an initial discussion on the applicability of PLSR in differentiating anthropogenic anomalies from natural contents.PLSR performs a simultaneous, interdependent principal component analysis decomposition in both X- and Y-matrices, in such a way that the information in the Y-matrix is used directly as a guide for optimal decomposition of the X-matrix. PLSR thus performs a generalized multivariate regression of Y on X overcoming the multicollinearity problem of correlated X-variables. The advantage of PLSR is that it gives optimal prediction ability in a strict statistical sense.Bedrock geochemistry from different lithologies in the mapping area in southern Sweden (Y-matrix) is analyzed together with stream or soil data (X-matrix). By modelling the PLS-regression between these two data sets, separate multivariate geochemical models based on the different bedrock types were developed. This step is called the training or modelling stage of the multivariate calibration. These calibrated models are subsequently used for predicting new (X) geochemical samples and estimating the corresponding Y-variable values. Information is obtained on how much of the metal contents in each new geochemical sample correlate with the different modelled bedrock types.By computing the appropriate X-residuals, we obtain information on the anthropogenic impact that is also carried by these new samples. In this way, it is possible from one single geochemical survey to derive both conventional geochemical background data and anthropogenic data, both of which can be readily displayed as maps.The present study concerns development of data analysis methods. Examples of the applications of the methodology are presented using Pb and U. The results show the share of these contents in different sampling media that is derived from bedrock on the one hand, and from anthropogenic sources, on the other.