Quantifying volcanic ash fall hazard to electricity infrastructure

Quantifying the potential ash fall hazards from re-awakening volcanoes is a topic of great interest. While methods for calculating the probability of eruptions, and for numerical simulation of tephra dispersal and fallout exist, event records at most volcanoes that re-awaken sporadically on decadal to millennial cycles are inadequate to develop rigorous forecasts of occurrence, much less eruptive volume. Here we demonstrate a method by which eruption records from radiocarbon-dated sediment cores can be used to derive forecasting models for ash fall impacts on electrical infrastructure. Our method is illustrated by an example from the Taranaki region of New Zealand. Radiocarbon dates, expressed as years before present (B.P.), are used to define an age-depth model, classifying eruption ages (with associated errors) for a circa 1500–10 500 year B.P. record at Mt. Taranaki (New Zealand). In addition, data describing the youngest 1500 years of eruption activity is obtained from directly dated proximal deposits. Absence of trend and apparent independence in eruption intervals is consistent with a renewal model using a mix of Weibulls distributions, which was used to generate probabilistic forecasts of eruption recurrence. After establishing that interval length and tephra thickness were independent in the record, a thickness–volume relationship (from [Rhoades, D.A., Dowrick, D.J., Wilson, C.J.N., 2002. Volcanic hazard in New Zealand: Scaling and attenuation relations for tephra fall deposits from Taupo volcano. Nat. Hazards, 26:147–174]) was inverted to provide a frequency–volume relationship for eruptions. Monte Carlo simulation of the thickness–volume relationship was then used to produce probable ash fall thicknesses at any chosen site. Several critical electrical infrastructure sites in the Taranaki Region were analysed. This region, being the only gas and condensate-producing area in New Zealand, is of national economic importance, with activities in and around the area depending on uninterrupted power supplies. Forecasts of critical ash thicknesses (1 mm wet and 2 mm dry) that may cause short-circuiting, surges or power shutdowns in substations show that the annual probabilities of serious impact are between ~ 0.5% and 27% over a 50 year period. It was also found that while large eruptions with high ash plumes tend to affect “expected” areas in relation to prevailing winds, the direction impacts of small ash falls are far less predictable. In the Taranaki case study, areas out of normal downwind directions, but close to the volcano, have probabilities of impact for critical thicknesses of 1–2 mm of around half to 60% of those in downwind directions and therefore should not be overlooked in hazard analysis. Through this method we are able to definitively show that the potential ash fall hazard to electrical infrastructure in this area is low in comparison to other natural threats, and provide a quantitative measure for use in risk analysis and budget prioritisation for hazard mitigation measures.     

Impact of explosive eruption scenarios at Vesuvius

In the paper the first attempt at the definition of a model to assess the impact of a range of different volcanic hazards on the building structures is presented. This theoretical approach has been achieved within the activities of the EXPLORIS Project supported by the EU. A time history for Sub-Plinian I eruptive scenario of the Vesuvius is assumed by taking advantage of interpretation of historical reports of volcanic crises of the past [Carafa, G. 1632. In opusculum de novissima Vesuvij conflagratione, epistola isagogica, 2^a ed. Napoli, Naples; Mascolo, G.B., 1634. De incendio Vesuvii excitato xvij. Kal. Ianuar. anno trigesimo primo sæculi Decimiseptimi libri X. Cum Chronologia superiorum incendiorum; & Ephemeride ultimi. Napoli; Varrone, S., 1634. Vesuviani incendii historiae libri tres. Napoli], numerical simulations [Neri, A., Esposti Ongaro, T., Macedonio, G., Gidaspow, D., 2003. Multiparticle simulation of collapsing volcanic columns and pyroclastic flows. J. Geophys. Res. Lett. 108, 2202. doi:10.1029/2001 JB000508; Macedonio, G., Costa, A., Longo, A., 2005. HAZMAP: a computer model for volcanic ash fallout and assessment of subsequent hazard. Comput. Geosci. 31,837–845; Costa, A., Macedonio, G., Folch, A., 2006. A three-dimensional Eulerian model for transport and deposition of volcanic ashes. Earth Planet. Sci. Lett. 241,634–647] and experts' elicitations [Aspinall, W.P., 2006. Structured elicitation of expert judgment for probabilistic hazard and risk assessment in volcanic eruptions. In: Mader, H.M. Coles, S.G. Connor, C.B. Connor, L.J. (Eds), Statistics in Volcanology. Geological Society of London on behalf of IAVCEI, pp.15–30; Woo, G., 1999. The Mathematics of Natural Catastrophes. Imperial College Press, London] from which the impact on the building structures is derived. This is achieved by an original definition of vulnerability functions for multi-hazard input and a dynamic cumulative damage model. Factors affecting the variability of the final scenario are highlighted. The results show the high sensitivity of hazard combinations in time and space distribution and address how to mitigate building vulnerability to subsequent eruptive phenomena [Baxter, P., Spence, R., Zuccaro, G., 2008-this issue. Risk mitigation and emergency measures at Vesuvius].     

Probabilistic impedance of foundation: Impact of the seismic design on uncertain soils

In the linear seismic design of buildings, the (deterministic) substructure method is a customary and efficient approach. However, the existence of spatial variability in the parameters of the mechanical model of the soil, as well as parametric errors, calls for the use of probabilistic approaches in order to provide a reliable design of the structure. The construction of probabilistic models of the soil impedance matrix provides a natural path to such approaches within the context of the substructure method. Two main techniques are described in this paper: a parametric one, typically using the stochastic finite element method, and a nonparametric one, which was introduced more recently. The latter is explored more specifically, and the possibilities it offers in terms of seismic design are presented. In particular, it is shown that it allows for the estimation of quantiles of the quantities of interest, rather than confidence intervals, which lead to highly conservative design. Copyright © 2008 John Wiley & Sons, Ltd.     

Probabilistic aftershock hazard assessment I: numerical testing of methodological features

Probabilistic aftershock hazard assessment (PAHA, Wiemer, Geoph Res Lett 27:3405–3408, 2000), provided for California within the frame of the STEP project, is based on a methodology, two features of which are addressed in detail: (1) the parameters of Omori’s law and (2) application of attenuation relations in evaluating peak ground velocity exceedence probability. Concerning the first point, we perform a simple parametric study. We assume the generalized Omori’s law by Shcherbakov et al. (Geoph Res Lett 31:L11613, 2004), in which characteristic time c scales with aftershock magnitude. The study shows that, among all the parameters, the hazard is most sensitive to the choice of m* (controlling the overall aftershock productivity) and least sensitive to the scaling of c. We also conclude that the hazard is mainly due to very early (less than 1 day) aftershocks. As regards the second point, we employ various attenuation relations from different tectonic areas to study their effect on the hazard analysis. We conclude that the resulting variations are relatively large, comparable to those obtained for varying m*.     

Probabilistic aftershock hazard assessment II: application of strong ground motion simulations

Probabilistic aftershock hazard assessment (PAHA) has been introduced by Wiemer (Geophys Res Lett 27:3405–3408, 2000). The method, in its original form, utilizes attenuation relations in evaluating peak ground velocity (PGV) exceedence probability. We substitute the attenuation relations together with their uncertainties by strong ground motion simulations for a set of scenarios. The main advantage of such an approach is that the simulations account for specific details of the aftershock source effects (faulting style, slip distribution, position of the nucleation point, etc.). Mean PGVs and their standard deviations are retrieved from the simulation results obtained by the new hybrid k-squared source model, and they are used for the PAHA analysis at a station under study. The model chosen for the testing purposes is inspired by the Izmit A25 aftershock (M _w  = 5.8) that occurred 26 days after the mainshock. The PAHA maps are compared with (1) those obtained by the use of attenuation relations and (2) the peak values of ten selected strong-motion recordings written by the aftershock at epicentral distances <50 km. We conclude that, although the overall hazard decay with increasing fault distance is similar, the PAHA maps obtained by the use of simulations exhibit remanent radiation pattern effect and prolongation in the strike direction due to the directivity effect pronounced for some of the scenarios. As regard the comparison with real data, we conclude that the PAHA maps agree with observed peak values due to appropriate attenuation model adopted in the analysis.     

利用多目标地球化学数据识别第四纪沉积物类型——基于概率神经网络方法

针对利用多目标地球化学数据研究第四纪沉积物类型问题,提出了基于概率神经网络的分类识别模型,并给出地球化学特征指标选取、指标归一化、神经网络设置和训练的具体方法、步骤。在吉林省中西部松嫩平原应用表明,该方法识别出8类不同成因的第四纪沉积物,较好地解决了该区第四纪沉积物成因归属问题。概率神经网络模型对第四纪沉积物类型的识别能力远高于常规多元统计方法,且结构简单、训练快捷。     

基于概率分析的渗流计算方法及应用

以随机理论为基础,把岩土体渗透系数视为随机变量。根据渗透系数的概率分布,把岩土体视为按概率分布的不同渗透系数介质的混合体,不同渗透系数介质对渗流场产生贡献的大小由其概率的大小来确定。在确定岩土体渗透系数后,通过渗流方程就可求出不同渗透系数作用下区域内的水头分布和断面流量,再根据渗透系数的概率大小进行叠加,求出整个区域内的水头分布和断面流量。工程实例表明,该方法较确定性有限元法误差小,可作为渗流计算的又一方法用于渗流分析。     

基于邻域最优概率的GRAPES快速更新循环同化预报系统逐时降水订正方法

针对高分辨率数值天气预报的时空不确定性, 利用邻域最优概率方法对华南区域GRAPES快速更新循环同化预报系统的24 h预报进行逐时降水订正和检验评估。结果表明: (1)邻域法能改善模式降水预报的空间不确定性, 最优邻域半径随降水等级增加而减小, 强降水的最优邻域半径约为60 km; (2)通过引入时间滞后因子, 可进一步改善模式不同时间起报的不确定性, 结合Brier评分确定了时间滞后窗为4 h; (3)提出基于邻域最优概率阈值的降雨进行分级订正方法, 有效提升了降水客观预报能力, 晴雨预报较模式全部为正技巧, TS评分达到0.89以上, 总体提升幅度约5.3%;强降水预报同样均为正技巧, TS评分呈先降后升趋势, 在12 h时效前后预报效果最优, 进一步提升了GRAPES快速更新循环同化预报系统的业务预报水平。      

关联向量机及其在入侵检测中的应用探讨

介绍一种稀疏的贝叶斯学习算法——关联向量机(RVM)，它在再生核希尔伯特空间中学习，利用贝叶斯方法推理，推广能力好，与支持向量机相比不仅解更为稀疏而且不需要调整超参数。应用RVM的对小样本的良好分类能力，提出一种基于RVM的入侵检测原型系统。