Dynamic risk assessment model for flood disaster on a projection pursuit cluster and its application

Aiming at reducing the losses from flood disaster, a dynamic risk assessment model for flood disaster is studied in this article. This model is built upon the projection pursuit cluster principle and risk indexes in the system, proceeding from the whole structure to its component parts. In this study, a fuzzy analytic hierarchy approach is employed to screen out the index system and determine the index weight, while the future value of each index is simulated by an improved back-propagation neural network algorithm. The proposed model adopts a dynamic evaluation method to analyze temporal data and assesses risk development by comprehensive analysis. The projection pursuit theory is used for clustering spatial data. The optimal projection vector is applied to calculate the risk cluster type. Therefore, the flood disaster risk level is confirmed and then the local conditions for presenting the control strategy. This study takes the Tunxi area, Huangshan city, as an example. After dynamic risk assessment model establishment, verification and application for flood disasters between the actual and simulated data from 2001 to 2013, the comprehensive risk assessment results show that the development trend for flood disaster risk is still in a decline on the whole, despite the rise in a few years. This is in accordance with the actual conditions. The proposed model is shown to be feasible for theory and application, providing a new way to assess flood disaster risk.     

Seismic comprehensive forecast based on modified project pursuit regression

In the research of projection pursuit for seismic comprehensive forecast, the algorithm of projection pursuit regression (PPR) is one of most applicable methods. But generally, the algorithm structure of the PPR is very complicated. By partial smooth regressions for many times, it has a large amount of calculation and complicated extrapolation, so it is easily trapped in partial solution. On the basis of the algorithm features of the PPR method, some solutions are given as below to aim at some shortcomings in the PPR calculation: to optimize project direction by using particle swarm optimization instead of Gauss-Newton algorithm, to simplify the optimal process with fitting ridge function by using Hermitian polynomial instead of piecewise linear regression. The overall optimal ridge function can be obtained without grouping the parameter optimization. The modeling capability and calculating accuracy of projection pursuit method are tested by means of numerical emulation technique on the basis of particle swarm optimization and Hermitian polynomial, and then applied to the seismic comprehensive forecasting models of poly-dimensional seismic time series and general disorder seismic samples. The calculation and analysis show that the projection pursuit model in this paper is characterized by simplicity, celerity and effectiveness. And this model is approved to have satisfactory effects in the real seismic comprehensive forecasting, which can be regarded as a comprehensive analysis method in seismic comprehensive forecast.     

Infilling of missing rainfall and streamflow data in the Shire River basin,Malawi – A self organizing map approach

A major requirement for the assessment, development and sustainable use of water resources is the availability of good quality hydrological time series data of sufficiently long duration. However, it is not uncommon to find data that are riddled with gaps, characterized by questionable quality and short durations. Sometimes, the data are just not available. Such situations are most prevalent in developing countries and the consequence is a high degree of uncertainty in the assessed characteristics of water management schemes and ultimately its ineffectual performance. Thus dealing with these problems is an important exercise in hydrological analyses. This paper focuses on the multivariate infilling of gaps for rainfall and streamflow data in the Shire River basin in Malawi, using a self organizing map (SOM) approach, which is a form of unsupervised artificial neural networks. The results show that this approach can produce reliable estimates of hydro-meteorological data thus offering promise for reducing the uncertainties associated with the use of insufficient data for water resources assessment.     

Support vector regression for estimating earthquake response spectra

This study uses support vector regression (SVR), a supervised machine learning algorithm, to model the average horizontal response spectrum as a nonparametric function of a set of predictor ground motion variables. Traditional ground motion prediction equations (GMPEs) are derived using parametric regression, where a fixed functional form is selected, and the model coefficients are determined by minimizing the errors on the training set. The SVR model is nonparametric; there is no need to assume a fixed functional form. Using nonlinear basis functions, the data points are mapped into a high dimensional feature space, where nonlinear input-output relationships can be expressed as a linear combination of nonlinear functions, using a subset of the data points. The combination weights are determined by minimizing the generalization error, using a formal mechanism to characterize the trade-off between the model complexity and the quality of fit to the data. Minimization of the generalization error instead of the fitting error leads to better generation to unseen data, and thus reduces the risk of over-fitting for a given number of data points. The SVR model is not based on a specific probability distribution, and is readily applicable to non-Gaussian data. An example application is presented for vertical strike-slip earthquakes, and the predictions from the SVR model are compared to the recently developed GMPEs. The results demonstrate the validity of the proposed model, and suggest that it can be used as an alternative to the conventional ground motion prediction models.     

新疆天山地区 PP回归综合预报模型研究及预报效能评价

运用基于 PP回归理论的数值型地震综合预报软件系统, 对新疆天山地区进行了 PP回归建模研究。 通过回顾性 PP回归动态建模预测检验,对其预报效能进行了评价。 结果认为,该模型作为时序性数值预报模型具有较好的中短期预报效能。 尤其对地震样本量大、震级分布完备的地区,其预报效能更好。     

Deep neural network application for 4D seismic inversion to changes in pressure and saturation: Optimizing the use of synthetic training datasets

In this work, we tackle the challenge of quantitative estimation of reservoir dynamic property variations during a period of production, directly from four-dimensional seismic data in the amplitude domain. We employ a deep neural network to invert four-dimensional seismic amplitude maps to the simultaneous changes in pressure, water and gas saturations. The method is applied to a real field data case, where, as is common in such applications, the data measured at the wells are insufficient for properly training deep neural networks, thus, the network is trained on synthetic data. Training on synthetic data offers much freedom in designing a training dataset, therefore, it is important to understand the impact of the data distribution on the inversion results. To define the best way to construct a synthetic training dataset, we perform a study on four different approaches to populating the training set making remarks on data sizes, network generality and the impact of physics-based constraints. Using the results of a reservoir simulation model to populate our training datasets, we demonstrate the benefits of restricting training samples to fluid flow consistent combinations in the dynamic reservoir property domain. With this the network learns the physical correlations present in the training set, incorporating this information into the inference process, which allows it to make inferences on properties to which the seismic data are most uncertain. Additionally, we demonstrate the importance of applying regularization techniques such as adding noise to the synthetic data for training and show a possibility of estimating uncertainties in the inversion results by training multiple networks.     

Assessment of social vulnerability to natural hazards in the Yangtze River Delta, China

China is exposed to a wide range of natural hazards, and disaster losses have escalated over the past decade. Owing to the pressure from natural disasters, along with changes in climate, social conditions, and regional environment, assessment of social vulnerability (SV) to natural hazards has become increasingly urgent for risk management and sustainable development in China. This paper presents a new method for quantifying SV based on the projection pursuit cluster (PPC) model. A reference social vulnerability index (SVI) at the county level was created for the Yangtze River Delta area in China for 1995, 2000, 2005, and 2009. The result of social vulnerability assessment was validated using data of actual losses from natural disasters. The primary findings are as follows: (i) In the study area, the major factors that impact SVI are regional per capita GDP and per capita income. (ii) The study area was more vulnerable in 1995 than in later years. SV of the whole region had decreased over the study period. (iii) Most part of Shanghai and the southeast part of Jiangsu Province had been the least vulnerable within the region. From this least vulnerable zone to the periphery of the region, the situation deteriorated. The highest SVI values in all evaluated years were found in the northern, western, or southern tips of the Yangtze River Delta.     

基于PPAR模型视二维地震时间序列预测的初步研究

PP投影寻踪是一种长于分析非正态、非线性的高维数据的新统计方法,它通过投影降维,客观地寻找反映高维数据结构特征的投影方向,从而解决“维数祸根”和高维数据间的非正态、非线性问题。将PP理论和时间序列分析中的自回归(AR(K))模型结合起来,建立投影寻踪自回归预测模型(PPAR)。尝试实现地震震级和时间的视二维预测,即在固定研究区里。实现震级和时间二要素的预测,进而建立视二维地震时间序列的投影寻踪自回归模型。研究中首先选取北天山地区作为实验区,模型的回归拟合和外符检验效果较理想,可实现视二维预测目标。考虑到实际预测意义。即中强地震的预测,又以天山地区为研究区。令其震级序列的震级阈值分别为5．0和5．5,分别以未删除余震和删除余震的序列建立模型。对比分析表明,后所建立的模型要优于前的模型。特别是对时间间隔序列的预测。两外符检验的合格率均较高,故认为对于震级和时间二要素的预测是有一定实效的。     

Model structure exploration for index flood regionalization

Although the nonlinear power form model structure is widely accepted by practitioners in the flood regionalization modelling, there is a lack of studies on whether there is a room for further improvement, and if the answer is yes, what should be done to explore alternative model structures. A framework is proposed in this study towards investigating this issue by the following steps: (i) a universal data‐driven model is utilized to see if there is a room for improvement compared with the conventional model, and (ii) if improvement is achieved, this means that there should exist more effective model structures than the current form. However, because the universal data‐driven models are usually opaque, more explicit model structures should be explored, which are convenient for practical usage. In this study, the proposed framework is applied in a case study using the catchment characteristics from the Flood Estimation Handbook in conjunction with the gamma test, support vector machine (SVM) and genetic programming (GP). First, the gamma test is used for the purpose of input variables selection where no model structure needs to be defined as a priori, and therefore, the result can be applied to any model structures for model building. Second, an SVM, which is a powerful data‐driven nonlinear model capable of modelling a variety of nonlinear systems, is applied to the index flood model for the first time. Once the best model is determined using those two data‐driven tools, GP is employed to find an alternative model structure. As the SVM is not formulated for producing explicit model functional form, the GP offers an advantage at this point where it can infer an explicit mathematical model functional form. Copyright © 2012 John Wiley & Sons, Ltd.     

Reproduction of secondary data in projection pursuit transformation

A challenge when working with multivariate data in a geostatistical context is that the data are rarely Gaussian. Multivariate distributions may include nonlinear features, clustering, long tails, functional boundaries, spikes, and heteroskedasticity. Multivariate transformations account for such features so that they are reproduced in geostatistical models. Projection pursuit as developed for high dimensional data exploration can also be used to transform a multivariate distribution into a multivariate Gaussian distribution with an identity covariance matrix. Its application within a geostatistical modeling context is called the projection pursuit multivariate transform (PPMT). An approach to incorporate exhaustive secondary variables in the PPMT is introduced. With this approach the PPMT can incorporate any number of secondary variables with any number of primary variables. A necessary alteration to the approach to make this numerically practical was the implementation of a continuous probability estimator that relies on Bernstein polynomials for the transformation that takes place in the projections. Stopping criteria were updated to incorporate a bootstrap t test that compares data sampled from a multivariate Gaussian distribution with the data undergoing transformation.     

PUBLICATIONS OF THE IAHS

Abstract

Gauging stations where the stage—discharge relationship is affected by hysteresis due to unsteady flow represent a challenge in hydrometry. In such situations, the standard hydrometric practice of fitting a single-valued rating curve to the available stage—discharge measurements is inappropriate. As a solution to this problem, this study provides a method based on the Jones formula and nonlinear regression, which requires no further data beyond the available stage—discharge measurements, given that either the stages before and after each measurement are known along with the duration of each measurement, or a stage hydrograph is available. The regression model based on the Jones formula rating curve is developed by applying the monoclinal rising wave approximation and the generalized friction law for uniform flow, along with simplifying assumptions about the hydraulic and geometric properties of the river channel in conjunction with the gauging station. Methods for obtaining the nonlinear least-squares rating-curve estimates, while factoring in approximated uncertainty, are discussed. The broad practical applicability and appropriateness of the method are demonstrated by applying the model to: (a) an accurate, comprehensive and detailed database from a hydropower-generated highly dynamic flow in the Chattahoochee River, Georgia, USA; and (b) data from gauging stations in two large rivers in the USA affected by hysteresis. It is also shown that the model is especially suitable for post-modelling hydraulic and statistical validation and assessment.     

Simulating earthquake evacuation using human behavior models

In classical earthquake risk assessment, the human behavior is actually not taken into account in risk assessment. Agent‐based modeling is a simulation technique that has been applied recently in several fields, such as emergency evacuation. The paper is proposing a methodology that includes in agent‐based models the human behavior, considering the anxiety effects generated by the crowd and their influence on the evacuation delays. The proposed model is able to take into account the interdependency between the earthquake evacuation process, and the corresponding damage of structural and non‐structural components that is expressed in term of fragility curves. The software REPAST HPC has been used to implement the model, and as a case study, the earthquake evacuation by a mall located in Oakland has been used. The human behavior model has been calibrated through a survey using a miscellaneous sample from different countries. The model can be used to test future scenarios and help local authorities in situations where the human behavior plays a key role. Copyright © 2017 John Wiley & Sons, Ltd.     

A hybrid approach of artificial neural network and multiple regression to forecast typhoon rainfall and groundwater-level change

ABSTRACT

A forecasting model is developed using a hybrid approach of artificial neural network (ANN) and multiple regression analysis (MRA) to predict the total typhoon rainfall and groundwater-level change in the Zhuoshui River basin. We used information from the raingauge stations in eastern Taiwan and open source typhoon data to build the ANN model for forecasting the total rainfall and the groundwater level during a typhoon event; then we revised the predictive values using MRA. As a result, the average accuracy improved up to 80% when the hybrid model of ANN and MRA was applied, even where insufficient data were available for model training. The outcome of this research can be applied to forecasts of total rainfall and groundwater-level change before a typhoon event reaches the Zhuoshui River basin once the typhoon has made landfall on the east coast of Taiwan.     

A supervised committee machine artificial intelligent for improving DRASTIC method to assess groundwater contamination risk: a case study from Tabriz plain aquifer,Iran

Vulnerability maps are designed to show areas of greatest potential for groundwater contamination on the basis of hydrogeological conditions and human impacts. The objective of this research is (1) to assess the groundwater vulnerability using DRASTIC method and (2) to improve the DRASTIC method for evaluation of groundwater contamination risk using AI methods, such as ANN, SFL, MFL, NF and SCMAI approaches. This optimization method is illustrated using a case study. For this purpose, DRASTIC model is developed using seven parameters. For validating the contamination risk assessment, a total of 243 groundwater samples were collected from different aquifer types of the study area to analyze \( {\text{NO}}_{ 3}^{ - } \) concentration. To develop AI and CMAI models, 243 data points are divided in two sets; training and validation based on cross validation approach. The calculated vulnerability indices from the DRASTIC method are corrected by the \( {\text{NO}}_{3}^{ - } \) data used in the training step. The input data of the AI models include seven parameters of DRASTIC method. However, the output is the corrected vulnerability index using \( {\text{NO}}_{3}^{ - } \) concentration data from the study area, which is called groundwater contamination risk. In other words, there is some target value (known output) which is estimated by some formula from DRASTIC vulnerability and \( {\text{NO}}_{3}^{ - } \) concentration values. After model training, the AI models are verified by the second \( {\text{NO}}_{3}^{ - } \) concentration dataset. The results revealed that NF and SFL produced acceptable performance while ANN and MFL had poor prediction. A supervised committee machine artificial intelligent (SCMAI), which combines the results of individual AI models using a supervised artificial neural network, was developed for better prediction of vulnerability. The performance of SCMAI was also compared to those of the simple averaging and weighted averaging committee machine intelligent (CMI) methods. As a result, the SCMAI model produced reliable estimates of groundwater contamination risk.     

The BEAM-project: prediction and assessment of mixture toxicities in the aquatic environment

Freshwater and marine ecosystems are exposed to various multi-component mixtures of pollutants. Nevertheless, most ecotoxicological research and chemicals regulation focus on hazard and exposure assessment of individual substances only, the problem of chemical mixtures in the environment is ignored to a large extent. In contrast, the assessment of combination effects has a long tradition in pharmacology, where mixtures of chemicals are specifically designed to develop new products, e.g. human and veterinary drugs or agricultural and non-agricultural pesticides. In this area, two concepts are frequently used and are thought to describe fundamental relationships between single substance and mixture effects: Independent Action (Response Addition) and Concentration Addition. The question, to what extent these concepts may also be applied in an ecotoxicological and regulatory context may be considered a research topic of major importance, as the concepts would allow to make use of already existing single substance toxicity data for the predictive assessment of mixture toxicities. Two critical knowledge gaps are identified: (a) There is a lack of environmental realism, as a huge part of our current knowledge about the applicability of the concepts is restricted to artificial situations with respect to mixture composition or biological effect assessment. (b) The knowledge on what exactly is needed for using the concepts as tools for the predictive mixture toxicity assessment is insufficient. Both gaps seriously hamper the necessary, scientifically sound consideration of mixture toxicities in a regulatory context.In this paper, the two concepts will be briefly introduced, the necessity of considering the toxicities of chemical mixtures in the environment will be demonstrated and the applicability of Independent Action and Concentration Addition as tools for the prediction and assessment of mixture toxicities will be discussed. An overview of the specific aims and approaches of the BEAM project to fill in the identified knowledge gaps is given and first results are outlined.     

Hazard Assessment of Co-seismic Landslides Based on Information Value Method: A Case in 2018 MW6.6 Hokkaido Earthquake, Japan

An M_W6.6 earthquake occurred in eastern Hokkaido, Japan on September 6^th, 2018. Based on the pre-earthquake image from Google Earth and the post-earthquake image from high resolution (3 m) planet satellite, we manually interpret 9 293 coseismic landslides and select 7 influencing factors of seismic landslide, such as elevation, slope, slope direction, road distance, flow distance, peak ground acceleration (PGA) and lithology. Then, 9 293 landslide points are randomly divided into training samples and validation samples with a proportion of 7:3. In detail, the training sample has 6 505 landslide points and the validation sample has 2 788 landslide points. The hazard risk assessment of seismic landslide is conducted by using the information value method and the study area is further divided into five risk grades, including very low risk area, low risk area, moderate risk area high risk area and very high risk area. The results show that there are 7 576 landslides in high risk area and very high risk area, accounting for 81.52% of the total landslide number, and the landslide area is 22.93 km², accounting for 74.35% of the total area. The hazard zoning is in high accordance with the actual situation. The evaluation results are tested by using the curve of cumulative percentage of hazardous area and cumulative percentage of landslides number. The results show that the success rate of the information value method is 78.50% and the prediction rate is 78.43%. The evaluation results are satisfactory, indicating that the hazard risk assessment results based on information value method may provide scientific reference for landslide hazard risk assessment as well as the disaster prevention and mitigation in the study area.     

Scenario-Based Seismic Risk Analysis: An Engineering Approach to the Development of Source and Site-Specific Ground Motion Time Histories in Areas of Low Seismicity

Modern engineering design methods require ground motion time histories as input for non-linear dynamic structural analysis. Non-linear dynamic methods of analysis are increasingly applied in the context of probabilistic risk assessments and for cost-effective design of critical infrastructures. In current engineering practice artificial time histories matching deterministic design spectra or probabilistic uniform hazard spectra are most frequently used for engineering analysis. The intermediate step of generation of response spectra can lead to a biased estimate of the potential damage from earthquakes because of insufficient consideration of the true energy content and strong motion duration of earthquakes. Thus, assessment of seismic risk may seem unrealistic. An engineering approach to the development of three-component ground motion time histories has been established which enables consideration of the typical characteristics of seismic sources, regional ground motion attenuation, and the main geotechnical characteristics of the target site. Therefore, the approach is suitable for use in scenario-based risk analysis a larger number of time histories are required for representation of the seismic hazard. Near-field effects are implemented in the stochastic source model using engineering approximations. The approach is suggested for use in areas of low seismicity where ground motion records of larger earthquakes are not available. Uncertainty analysis indicates that ground motions generated by individual earthquakes are well constrained and that the usual lognormal model is not the best choice for predicting the upper tail of the distribution of the ground motions.     

A new site classification approach based on neural networks

Site classification is an important procedure for a reliable site-specific seismic hazard assessment. On the other hand, the site conditions at strong-motion stations are essential for accurate interpretation and analysis of the recorded ground motion data obtained from different regions of the world. For some countries with insufficient data on the subsurface geological settings, the required site condition information is not available. This paper presents a new and efficient approach for site classification based on artificial neural networks (ANN) along with a selected set of representative horizontal to vertical spectral ratio (HVSR) curves for four site classes. The nonlinear nature of ANN and their ability to learn in a complex environment make it highly suitable for function approximation and solving complicated engineering problems. Two types of radial basis function (RBF) neural networks, namely, probabilistic neural networks (PNN) and generalized regression neural networks (GRNN) were chosen in this study, as no separate training phase is required, rendering them particularly suitable for site classification. The proposed approach has been tested using data of the Chi-Chi, Taiwan, earthquake (Mw=7.6) recorded from 87 stations at which the site conditions are known. Analyses show that both the PNN and the GRNN perform very well with similar accuracy in estimating site conditions, with successful rates of 78% and 75%, respectively.