Probabilistic parameter estimation and predictive uncertainty based on field measurements for unsaturated soil slope

A key issue in assessment of rainfall-induced slope failure is a reliable evaluation of pore water pressure distribution and its variations during rainstorm, which in turn requires accurate estimation of soil hydraulic parameters. In this study, the uncertainties of soil hydraulic parameters and their effects on slope stability prediction are evaluated, within the Bayesian framework, using the field measured temporal pore-water pressure data. The probabilistic back analysis and parameter uncertainty estimation is conducted using the Markov Chain Monte Carlo simulation. A case study of a natural terrain site is presented to illustrate the proposed method. The 95% total uncertainty bounds for the calibration period are relatively narrow, indicating an overall good performance of the infiltration model for the calibration period. The posterior uncertainty bounds of slope safety factors are much narrower than the prior ones, implying that the reduction of uncertainty in soil hydraulic parameters significantly reduces the uncertainty of slope stability.     

三种水文模型不确定性分析方法比较

模型不确定性研究是水文科学的重要课题。以尼泊尔Bagmati流域为案例,采用了马尔科夫链蒙托卡罗(Markov Chain Monte Carlo)、蒙托卡罗(Monte Carlo)和拉丁超立方体(Latin Hypercube)等三种方法,分析了水箱模型输出成果的不确定性,并将三种方法所获得参数不确定性进行了比较。另外,运用Meta-Gaussian模型计算了总体不确定性,在基于所采用的似然函数基础上,对由参数导致模型输出的不确定性和模型输出的总体不确定性进行了比较。结果显示,模型的不确定性比参数不确定性更为重要,同时也表明,尽管蒙托卡罗和拉丁超立方体两种模拟方法产生几乎相同的结果,但两者都与马尔科夫链蒙托卡罗方法有很大的不同。     

Probabilistic risk assessment of major accidents: application to offshore blowouts in the Gulf of Mexico

Major accidents are low-frequency, high-consequence accidents which are not well supported by conventional statistical methods due to the scarcity of directly relevant data. Modeling and decomposition techniques such as event tree have been proved as robust alternatives as they facilitate incorporation of partially relevant near accident data–accident precursor data—in probability estimation and risk analysis of major accidents. In this study, we developed a methodology based on event tree and hierarchical Bayesian analysis to establish informative distributions for offshore blowouts using data of near accidents, such as kicks, leaks, and failure of blowout preventers collected from a variety of offshore drilling rigs. These informative distributions can be used as predictive tools to estimate relevant failure probabilities in the future. Further, having a set of near accident data of a drilling rig of interest, the informative distributions can be updated to render case-specific posterior distributions which are of great importance in quantitative risk analysis. To cope with uncertainties, we implemented the methodology in a Markov Chain Monte Carlo framework and applied it to risk assessment of offshore blowouts in the Gulf of Mexico.     

Characterising geotechnical model uncertainty by hybrid Markov Chain Monte Carlo simulation

Geotechnical models are usually associated with considerable amounts of model uncertainty. In this study, the model uncertainty of a geotechnical model is characterised through a systematic comparison between model predictions and past performance data. During such a comparison, model input parameters (such as soil properties) may also be uncertain, and the observed performance may be subjected to measurement errors. To consider these uncertainties, the model uncertainty parameters, uncertain model input parameters and actual performance variables are modelled as random variables, and their distributions are updated simultaneously using Bayes’ theorem. When the number of variables to update is large, solving the Bayesian updating problem is computationally challenging. A hybrid Markov Chain Monte Carlo simulation is employed in this paper to decompose the high-dimensional Bayesian updating problem into a series of updating problems in lower dimensions. To increase the efficiency of the Markov chain, the model uncertainty is first characterised with a first order second moment method approximately, and the knowledge learned from the approximate solution is then used to design key parameters in the Markov chain. Two examples are used to illustrate the proposed methodology for model uncertainty characterisation, with insights, discussions, and comparison with previous methods.     

Modified Markov Chain Monte Carlo Method for Dynamic Data Integration Using Streamline Approach

In this paper, the Markov Chain Monte Carlo (MCMC) approach is used for sampling of the permeability field conditioned on the dynamic data. The novelty of the approach consists of using an approximation of the dynamic data based on streamline computations. The simulations using the streamline approach allows us to obtain analytical approximations in the small neighborhood of the previously computed dynamic data. Using this approximation, we employ a two-stage MCMC approach. In the first stage, the approximation of the dynamic data is used to modify the instrumental proposal distribution. The obtained chain correctly samples from the posterior distribution; the modified Markov chain converges to a steady state corresponding to the posterior distribution. Moreover, this approximation increases the acceptance rate, and reduces the computational time required for MCMC sampling. Numerical results are presented.     

Bayesian Modelling of Spatial Data Using Markov Random Fields, With Application to Elemental Composition of Forest Soil

Spatial datasets are common in the environmental sciences. In this study we suggest a hierarchical model for a spatial stochastic field. The main focus of this article is to approximate a stochastic field with a Gaussian Markov Random Field (GMRF) to exploit computational advantages of the Markov field, concerning predictions, etc. The variation of the stochastic field is modelled as a linear trend plus microvariation in the form of a GMRF defined on a lattice. To estimate model parameters we adopt a Bayesian perspective, and use Monte Carlo integration with samples from Markov Chain simulations. Our methods does not demand lattice, or near-lattice data, but are developed for a general spatial data-set, leaving the lattice to be specified by the modeller. The model selection problem that comes with the artificial grid is in this article addressed with cross-validation, but we also suggest other alternatives. From the application of the methods to a data set of elemental composition of forest soil, we obtained predictive distributions at arbitrary locations as well as estimates of model parameters.     

Aquifer hydraulic conductivity prediction via coupling model of MCMC-ANN

Grain-size distribution data,as a substitute for measuring hydraulic conductivity(K),has often been used to get K value indirectly.With grain-size distribution data of 150 sets of samples being input data,this study combined the Artificial Neural Network technology(ANN)and Markov Chain Monte Carlo method(MCMC),which replaced the Monte Carlo method(MC)of Generalized Likelihood Uncertainty Estimation(GLUE),to establish the GLUE-ANN model for hydraulic conductivity prediction and uncertainty analysis.By means of applying the GLUE-ANN model to a typical piedmont region and central region of North China Plain,and being compared with actually measured values of hydraulic conductivity,the relative error ranges are between 1.55%and 23.53%and between 14.08%and 27.22%respectively,the accuracy of which can meet the requirements of groundwater resources assessment.The global best parameter gained through posterior distribution test indicates that the GLUEANN model,which has satisfying sampling efficiency and optimization capability,is able to reasonably reflect the uncertainty of hydrogeological parameters.Furthermore,the influence of stochastic observation error(SOE)in grain-size analysis upon prediction of hydraulic conductivity was discussed,and it is believed that the influence can not be neglected.     

Bayesian Gaussian Mixture Linear Inversion for Geophysical Inverse Problems

A Bayesian linear inversion methodology based on Gaussian mixture models and its application to geophysical inverse problems are presented in this paper. The proposed inverse method is based on a Bayesian approach under the assumptions of a Gaussian mixture random field for the prior model and a Gaussian linear likelihood function. The model for the latent discrete variable is defined to be a stationary first-order Markov chain. In this approach, a recursive exact solution to an approximation of the posterior distribution of the inverse problem is proposed. A Markov chain Monte Carlo algorithm can be used to efficiently simulate realizations from the correct posterior model. Two inversion studies based on real well log data are presented, and the main results are the posterior distributions of the reservoir properties of interest, the corresponding predictions and prediction intervals, and a set of conditional realizations. The first application is a seismic inversion study for the prediction of lithological facies, P- and S-impedance, where an improvement of 30% in the root-mean-square error of the predictions compared to the traditional Gaussian inversion is obtained. The second application is a rock physics inversion study for the prediction of lithological facies, porosity, and clay volume, where predictions slightly improve compared to the Gaussian inversion approach.     

Unraveling reservoir compaction parameters through the inversion of surface subsidence observations

In an attempt to derive more information on the parameters driving compaction, this paper explores the feasibility of a method utilizing data on compaction-induced subsidence. We commence by using a Bayesian inversion scheme to infer the reservoir compaction from subsidence observations. The method’s strength is that it incorporates all the spatial and temporal correlations imposed by the geology and reservoir data. Subsequently, the contributions of the driving parameters are unravelled. We apply the approach to a synthetic model of an upscaled gas field in the northern Netherlands. We find that the inversion procedure leads to coupling between the driving parameters, as it does not discriminate between the individual contributions to the compaction. The provisional assessment of the parameter values shows that, in order to identify adequate estimate ranges for the driving parameters, a proper parameter estimation procedure (Markov Chain Monte Carlo, data assimilation) is necessary.     

基于Monte Carlo-BP神经网络TBM掘进速度预测

预测隧道工程中TBM掘进速度,主要有完全经验的、半理论半经验的模型和人工智能等方法,所用参数均为确定性的,未考虑参数存在的随机性,故导致预测结果的不准确性。基于此,提出了Monte Carlo-BP神经网络TBM掘进速度预测模型,着重考虑了一些重要输入参数的随机性, 其中输入参数重要性的大小通过粗糙集进行计算排序。采用Monte Carlo产生随机数时,由于参量的样本数据的有限,分布函数均采用阶梯形经验分布函数。如果采用的数据是来自不同类型的 TBM,则应当考虑机器性能参数,并重新对参数重要性进行排序。实例计算表明,Monte Carlo-BP神经网络模型预测结果和实测值总体趋势和均值比较一致。     

Identifying parameters of advanced soil models using an enhanced transitional Markov chain Monte Carlo method

Acta Geotechnica - Parameter identification using Bayesian approach with Markov Chain Monte Carlo (MCMC) has been verified only for certain conventional simple constitutive models up to now. This...     

Indicator Variogram Models: Do We Have Much Choice?

Many variogram (or covariance) models that are valid—or realizable—models of Gaussian random functions are not realizable indicator variogram (or covariance) models. Unfortunately there is no known necessary and sufficient condition for a function to be the indicator variogram of a random set. Necessary conditions can be easily obtained for the behavior at the origin or at large distance. The power, Gaussian, cubic or cardinal-sine models do not fulfill these conditions and are therefore not realizable. These considerations are illustrated by a Monte Carlo simulation demonstrating nonrealizability over some very simple three-point configurations in two or three dimensions. No definitive result has been obtained about the spherical model. Among the commonly used models for Gaussian variables, only the exponential appears to be a realizable indicator variogram model in all dimensions. It can be associated with a mosaic, a Boolean or a truncated Gaussian random set. In one dimension, the exponential indicator model is closely associated with continuous-time Markov chains, which can also lead to more variogram models such as the damped oscillation model. One-dimensional random sets can also be derived from renewal processes, or mosaic models associated with such processes. This provides an interesting link between the geostatistical formalism, focused mostly on two-point statistics, and the approach of quantitative sedimentologists who compute the probability distribution function of the thickness of different geological facies. The last part of the paper presents three approaches for obtaining new realizable indicator variogram models in three dimensions. One approach consists of combining existing realizable models. Other approaches are based on the formalism of Boolean random sets and truncated Gaussian functions.     

突发性水污染事件溯源方法

为快速准确地求解突发性水污染溯源问题,在微分进化与蒙特卡罗基础上提出了一种新的溯源方法。该方法将溯源问题视为贝叶斯估计问题,推导出污染源强度、位置和排放时刻等未知参数的后验概率密度函数;结合微分进化和蒙特卡罗模拟方法对后验概率分布进行采样,进而估计出这些未知参数,确定污染源项。通过算例与贝叶斯-蒙特卡罗方法进行对比,结果表明:该方法可使迭代次数有效缩减3/4,污染源强度、位置和排放时刻的平均相对误差分别减少1.23%、2.23%和4.15%,均值误差分别降低0.39%、0.83%和1.49%,其稳定性和可靠性明显高于贝叶斯-蒙特卡罗方法,能较好地识别突发性水污染源,为解决突发水污染事件中的追踪溯源难点问题提供了新的思路和方法。     

An intelligent response surface method for analyzing slope reliability based on Gaussian process regression

Problems in geotechnical engineering inevitably involve many uncertainties in the analysis. Reliability methods are important for evaluating slope stability and can take the uncertainties into consideration. In this paper, a novel intelligent response surface method is proposed in which a machine learning algorithm, namely Gaussian process regression, is used to approximate the high-dimensional and highly nonlinear response hypersurface. An iterative algorithm is also proposed for updating the response surface dynamically by adding the new training point nearest to the limit state surface to the initial training database at each step. The proposed Gaussian process response surface method is used to analyze three different case studies to assess its validity and efficiency. Direct Monte Carlo simulation is also carried out in each case to serve as the benchmark. Comparing with other methods confirms the accuracy and efficiency of the novel intelligent response surface method, which requires fewer performance function calls and avoids the need to normalize the correlative non-normal variables.     

Back analysis of slope failure with Markov chain Monte Carlo simulation

Field observed performance of slopes can be used to back calculate input parameters of soil properties and evaluate uncertainty of a slope stability analysis model. In this paper, a new probabilistic method is proposed for back analysis of slope failure. The proposed back analysis method is formulated based on Bayes’ theorem and solved using the Markov chain Monte Carlo simulation method with a Metropolis–Hasting algorithm. The method is very flexible as any type of prior distribution can be used. The method is also computationally efficient when a response surface method is employed to approximate the slope stability model. An illustrative example of back analysis of a hypothetical slope failure is presented. Effects of jumping distribution functions and number of samples on the efficiency of Markov chains are studied. It is found that the covariance matrix of the jumping function can be set to be one half of the covariance of the prior distribution to achieve a reasonable acceptance rate and that 80,000 samples seem to be sufficient to obtain robust posterior statistics for the example. It is also found that the correlation of cohesion and friction angle of soil does not affect the posterior statistics and the remediation design of the slope significantly, while the type of the prior distribution seems to have much influence on the remediation design.     

Probabilistic Fuzzy Logic Modeling: Quantifying Uncertainty of Mineral Prospectivity Models Using Monte Carlo Simulations

Significant uncertainties are associated with the definition of both the exploration targeting criteria and computational algorithms used to generate mineral prospectivity maps. In prospectivity modeling, the input and computational uncertainties are generally made implicit, by making a series of best-guess or best-fit decisions, on the basis of incomplete and imprecise information. The individual uncertainties are then compounded and propagated into the final prospectivity map as an implicit combined uncertainty which is impossible to directly analyze and use for decision making. This paper proposes a new approach to explicitly define uncertainties of individual targeting criteria and propagate them through a computational algorithm to evaluate the combined uncertainty of a prospectivity map. Applied to fuzzy logic prospectivity models, this approach involves replacing point estimates of fuzzy membership values by statistical distributions deemed representative of likely variability of the corresponding fuzzy membership values. Uncertainty is then propagated through a fuzzy logic inference system by applying Monte Carlo simulations. A final prospectivity map is represented by a grid of statistical distributions of fuzzy prospectivity. Such modeling of uncertainty in prospectivity analyses allows better definition of exploration target quality, as understanding of uncertainty is consistently captured, propagated and visualized in a transparent manner. The explicit uncertainty information of prospectivity maps can support further risk analysis and decision making. The proposed probabilistic fuzzy logic approach can be used in any area of geosciences to model uncertainty of complex fuzzy systems.     

Probabilistic back analysis of slope failure – A case study in Taiwan

In this paper, the authors present a probabilistic back-analysis of a recent slope failure at a site on Freeway No. 3 in northern Taiwan. Post-event investigations of this failure found uncertain strength parameters and deteriorating anchor systems as the most likely causes for failure. Field measurement after the event indicated an average slip surface of inclination 15°. To account for the uncertainties in input parameters, the probabilistic back analysis approach was adopted. First, the Markov Chain Monte Carlo (MCMC) simulation was used to back-calculate the geotechnical strength parameters and the anchor force. These inverse analysis results, which agreed closely with the findings of the post-event investigations, were then used to validate the maximum likelihood (ML) method, a computationally more efficient back-analysis approach. The improved knowledge of the geotechnical strength parameters and the anchor force gained through the probabilistic inverse analysis better elucidated the slope failure mechanism, which provides a basis for a more rational selection of remedial measures.     

A Monte Carlo adapted finite element method for dislocation simulation of faults with uncertain geometry

Dislocation modelling of an earthquake fault is of great importance due to the fact that ground surface response may be predicted by the model. However, geological features of a fault cannot be measured exactly, and therefore these features and data involve uncertainties. This paper presents a Monte Carlo based random model of faults with finite element method incorporating split node technique to impose the effects of discontinuities. Length and orientation of the fault are selected as random parameters in the domain model, and hence geometrical uncertainties are encountered. Mean and standard deviation values, as well as probability density function of ground surface responses due to the dislocation are computed. Based on analytical and numerical calculation of dislocation, two approaches of Monte Carlo simulations are proposed. Various comparisons are examined to illustrate the capability of both methods for random simulation of faults.     

Substitution Random Fields with Gaussian and Gamma Distributions: Theory and Application to a Pollution Data Set

This paper presents random field models with Gaussian or gamma univariate distributions and isofactorial bivariate distributions, constructed by composing two independent random fields: a directing function with stationary Gaussian increments and a stationary coding process with bivariate Gaussian or gamma distributions. Two variations are proposed, by considering a multivariate directing function and a coding process with a separable covariance, or by including drift components in the directing function. Iterative algorithms based on the Gibbs sampler allow one to condition the realizations of the substitution random fields to a set of data, while the inference of the model parameters relies on simple tools such as indicator variograms and variograms of different orders. A case study in polluted soil management is presented, for which a gamma model is used to quantify the risk that pollutant concentrations over remediation units exceed a given toxicity level. Unlike the multivariate Gaussian model, the proposed gamma model accounts for an asymmetry in the spatial correlation of the indicator functions around the median and for a spatial clustering of high pollutant concentrations.