A, parameter synthesis method
Parameter synthesis method is also called expert experience index comprehensive evaluation method. It is the simplest quantitative evaluation method. This model is mainly based on experts' rich experience, and obtains experts' experience knowledge through experts' scoring. Experts select the factors that affect geological disasters and make maps. According to experts' experience, each factor is given appropriate weight, and finally weighted superposition or synthesis is carried out to form a geological disaster risk zoning map.
Its main advantages are: ① a large number of parameters can be considered at the same time; ② It can be used to evaluate the stability of any scale area and single slope; ③ The use of implicit rules is greatly reduced, and the degree of quantification is improved; ④ The whole process can be completed quickly with the support of GIS, which makes data management standardized, short in time and low in cost. The main disadvantages are: ① Subjectivity is strong, and the results obtained by different investigators or experts cannot be compared. The determination of weight still contains different degrees of subjectivity; ② Implicit evaluation rules make it difficult to analyze and update the results; ③ Detailed field investigation is needed; (4) When applied to large-scale regional assessment, the operation is complicated and the model is difficult to popularize.
Second, the mathematical multivariate statistical model method
This method is to study the statistical laws between existing geological disasters and similar unstable phenomena and geological environmental conditions and factors, and establish relevant prediction models, so as to predict the danger of regional geological disasters. There are many methods for this kind of model, such as regression analysis, discriminant analysis and cluster analysis.
The premise of statistical analysis is to understand the distribution of geological disasters in the study area (training area). According to the theory of mathematical statistics, a mathematical statistical model of influencing parameters and whether geological disasters occur is established. After being verified in the experimental area, it is applied to areas with the same or similar geological environment to predict the distribution law of disaster risk in the study area. Therefore, the reliability of the evaluation results of statistical analysis method directly depends on the accuracy of the original data in the experimental area, and the model cannot be popularized in any area. Nevertheless, a large number of studies show that statistical analysis is the most suitable method for regional geological disaster risk assessment and zoning at present. It is based on strict mathematical statistics theory, its mathematical model is simple and easy to understand, and it can be well combined with GIS technology, so that huge data can be managed, analyzed and stored reasonably.
Principal component analysis and factor analysis in multivariate statistical analysis have many successful applications in environmental statistics. Principal component factor analysis, which combines these two methods, can be applied to the study of multivariate factor weights (Wu, 199 1). The main idea of principal component-factor analysis is (Ying Nonggen, Liu Youci, 1987): collect relevant information from all the original variables studied, and synthesize multiple variables into several unrelated principal components by discussing the internal dependence structure of the correlation matrix, so as to reproduce the relationship between the original variables, and further explore the internal reasons of these correlations through the orthogonal or oblique rotation of the axis of the factor load matrix.
This method is suitable for the study of spatial prediction of regional geological disasters, and has a macro-guiding role in land use, land development and urban planning in a certain region.
Third, the analytic hierarchy process
Analytic Hierarchy Process (AHP) is to analyze and evaluate a complex system, which contains many factors and is difficult to quantify accurately. According to the correlation between various factors and their evaluation objectives, the combination mode and level are straightened out, and the structural model and mathematical model of system evaluation are established accordingly. For all kinds of fuzzy factors in the model, according to their intensity and control degree to the affected objects, the scale index and action weight are determined; Taking these indicators as basic parameters, they are substituted into the evaluation model, and quantitative analysis is carried out step by step, and finally the evaluation goal is achieved. According to the composition of geological disaster risk system, the evaluation can be completed through four levels of statistical analysis: grass-roots statistical analysis with various elements as the main body; Transition layer analysis for risk, vulnerability and disaster reduction capacity; Criterion layer analysis for expected loss; Target layer analysis with risk degree or risk grade as the ultimate goal.
Fourthly, fuzzy and grey clustering methods.
Fuzzy clustering discriminant model is based on fuzzy mathematics theory. Due to the complexity of geological disaster system, the objective reality of geological disaster system can not be accurately described by absolute "either/or", and there is a fuzzy phenomenon of "this or that", which can not be described by binary logic of 1 or 0, but expressed by multi-value (or continuous value) logic of interval [0, 1]. Fuzzy mathematics theory is just suitable for the uncertainty of geological disaster system, while membership function is used to describe the transition problem with unclear boundary and the uncertainty of complex system affected by many factors. At present, the commonly used methods include fuzzy comprehensive evaluation method, fuzzy reliability analysis method and fuzzy hierarchical comprehensive evaluation method derived from hierarchical principle. The basic steps of fuzzy cluster comprehensive evaluation are: according to the risk composition of geological disasters, establish factor set, comprehensive evaluation set and weight set, determine membership function, get comprehensive evaluation results, and explain and analyze them.
Grey clustering comprehensive evaluation method is based on grey system theory and is often used to study the problem of "small sample and poor information uncertainty". In the prediction of geological disasters, grey relational analysis can be used to evaluate the influence degree of various influencing factors of slope stability, which can overcome the shortcomings brought by systematic analysis by the usual mathematical statistics method, and has no special requirements for sample size and regularity. Similarly, through the grey whitening weight function clustering in grey clustering, the risk state of each research unit can be judged, and then the risk zoning in spatial prediction can be completed. Various forecasting models of grey system with grey model (GM) as the core also provide an effective way for the analysis of various time series data in geological disaster forecasting, and become one of the commonly used methods for real-time tracking and forecasting of geological disasters at present. The basic steps of grey clustering comprehensive evaluation are: determining whitening number and whitening function, calibrating clustering weight, calculating clustering coefficient, constructing class vector and solving clustering grey number.
Evaluation method of verb (abbreviation of verb) information model
The theoretical basis of this model is information theory. The possibility of geological disasters is expressed by the decrease of entropy in the process of geological disasters. The average reduction of uncertainty caused by the combination of factors is equal to the change of entropy value of geological disaster system. It is considered that the occurrence of geological disasters is related to the quantity and quality of information obtained in the prediction process, which is measured by the amount of information. The greater the amount of information, the greater the possibility of geological disasters. This model forecasting method, like statistical forecasting model, is suitable for small and medium-sized regional forecasting.
Information science has become a widely used science, but it has only a short history of half a century. Shannon's famous paper "Mathematical Theory of Communication" published in 1948 marked the birth of information science. Shannon defined information as "reduction of uncertainty of random events", transplanted mathematical statistics method into communication field, and put forward the concept of information quantity and the mathematical formula of information entropy. The object of information science research is information, and its important task is to study information extraction, information transmission, information processing and information storage. Due to the integration trend of modern natural science development and the mutual penetration and connection of various disciplines, after decades of development, the concepts of information quantity and information entropy have gone far beyond the communication field. Information science is not only used in various natural science fields, but also widely used in management, society and other scientific fields.
The application of information theory in the study of ore deposit prediction in geological field was put forward by Viso Oster Roska (1968) and Cha Jin (1969) successively. Zhao et al. studied the application of information method in regional prospecting in the book Statistical Prediction of Mineral Deposits. Since 1985, Yan and Yin Kunlong have repeatedly explored the application of information methods in the spatial prediction and zoning of regional landslide disasters in southern Shaanxi and the Three Gorges reservoir area of the Yangtze River, and made a comparative study with the research results of other methods (such as cluster analysis, regression analysis, quantitative theoretical methods, etc.). ). Ai Nanshan and Miao Tiande (1987) studied the information entropy of the geomorphic system in the erosion basin. Based on the area-elevation curve of Stroller watershed, they constructed the expression of information entropy of geomorphic system of eroded watershed, and used it as the criterion of watershed stability. Read J. and Harr M.( 1988) first combined the concept of information entropy with slice method to calculate the slope safety factor. Because of the diversity of geological disaster prediction content, the prediction theory and method are not single. Yan et al. (1989) classified it into three types of model prediction methods-deterministic model prediction method, statistical model prediction method and information model prediction method; The first two models can be called "white box" and "black box" models respectively, while the information model is in between.
The geological disaster phenomenon (Y) in Xi is influenced by many factors, each of which is different in size and nature. In different geological environments, geological disasters always have an "optimal combination of factors". Therefore, for the prediction of regional geological disasters, we should not stay on a single factor, but comprehensively study the "best combination of factors". The viewpoint of information prediction holds that the occurrence of geological disasters is related to the quantity and quality of information obtained in the prediction process, so it can be measured by the amount of information:
Theory and practice of geological disaster risk assessment
According to the conditional probability operation, the above formula can be further written as:
Theory and practice of geological disaster risk assessment
I(y, x 1x2xn) is the information (bit) provided by the factor combination x 1x2xn to geological disasters; P(y, x 1x2xn) is the probability of geological disasters under the combination of x 1x2xn and other factors; Ix 1(y, x2) is the information provided by factor x 1 to geological disasters; P(y) is the probability of geological disasters.
Equation (2) shows that the information provided by the factor combination x 1x2xn is equal to the information provided by the factor x 1. After determining the factor x 1, the information provided by the factor x2 is determined until the factor x 1x2xn- 1, which shows the additivity of information.
P (y, x 1x2xn) and P(y) can be expressed by statistical probability, and the information provided by various factor combinations can be positive or negative. When P(y, x1x2xn)&; gt； P(y)，I (y，x 1x2xn)&； gt； 0; On the other hand, I (y, x1x2xn) < 0。 If it is greater than 0, it means that the combination of factors x 1x2xn is beneficial to predict the occurrence of geological disasters, on the contrary, it means that these combinations are not conducive to the occurrence of geological disasters.
The prediction of regional geological disasters is based on the division of grid units in the study area. According to the specific geological and topographic conditions in different regions, the corresponding grid shape and grid size are adopted, and the information statistical analysis is further carried out in combination with the regional geological disaster distribution map. Suppose that an area is divided into n units, and the number of units suffering from geological disasters is N0. The combination of the same factor x 1x2nx has ***M units, and the number of geological disasters in these units is M0. According to the principle that statistical probability represents prior probability, the information provided by factor x 1x2nx in this area is as follows:
Theory and practice of geological disaster risk assessment
If the area ratio is used to calculate the amount of information, Equation (3) can be expressed as:
Theory and practice of geological disaster risk assessment
Where: a is the total area of units in this area; A0 is the sum of unit areas where geological disasters have occurred; S is the total unit area of the combination of the same factor x 1x2xn; S0 is the sum of the unit areas where geological disasters occur in x 1x2xn combined units with the same factor.
Under normal circumstances, because there are many factors acting on geological disasters, the corresponding combination states of the factors are also particularly many, and the sample statistics are often limited, so the simplified single-factor information model is used for step-by-step calculation, and then the corresponding information model is rewritten as:
Theory and practice of geological disaster risk assessment
Where: I is the predicted value of information quantity of a unit in the prediction area; Si is the total unit area occupied by factor xi; S0i is the sum of the unit areas where geological disasters occur in the xi factor.
Six, the empirical weight method
Weights of evidence is a geostatistical method based on binary (existence or non-existence) images proposed by Canadian mathematical geologist Agterberg et al. (1989). It is a quantitative prediction method based on Bayesian rule under the assumption of independent conditions. Bonham-Carter et al. (1990) and Harris et al. (200 1) used WOE method to predict the prospective distribution of minerals. Through the statistical analysis between the prediction factors and response factors of grid cells with known mineralization conditions, the weights are calculated, and then the prediction factors of each grid cell to be predicted are weighted and integrated. Finally, by determining the probability of each unit response factor, different levels of metallogenic prospect areas can be obtained.
Van Westen further applied the model to the field of disaster risk assessment. The main principle of data-driven weight simulation method is to establish the statistical relationship between landslide distribution and influencing factors by using historical landslide distribution data, that is, to determine the contribution rate (weight) of each influencing factor to landslide disaster according to the statistical situation of landslide distribution of each influencing factor in different categories. This method of determining weights with data is called data-driven model. Compared with expert knowledge model, the determination of weight is more scientific and reliable, and the uncertainty caused by expert subjectivity is avoided. Finally, the historical data of landslide distribution in another period is used to test the evaluation results and predict the success rate, and the unreasonable boundary is adjusted to make the evaluation results more credible. The statistical method used in the data-driven weight model based on statistical Bayesian method is more rigorous, which fully considers the relationship between landslide influencing factors and the relationship between each influencing factor and landslide disaster. And the independent analysis of influencing factors, find out the most critical influencing factors. On this basis, the weight of each influencing factor is calculated.
Seven, nonlinear model prediction method
Nonlinear model prediction method, also known as BP neural network method, is to establish a prediction model by transforming the input and output problem of a group of samples into a nonlinear optimization problem.
In view of the complexity of geological disaster system, it is difficult to express it by simple linear equations, so a number of nonlinear prediction models have developed rapidly. For example, fractal theory is to study the movement law of geological disasters by studying the self-similarity of geological disaster systems. Yi Shunmin applied fractal theory to study the self-similar structural characteristics of regional landslide disaster activities, and found that there was an obvious dimension reduction phenomenon before the climax of geological disaster activities. Wu Zhongru and Huang Guoming put forward landslide deformation instability criterion and landslide creep phase space model based on fractal theory, which is a brand-new idea of geological disaster time prediction. The self-organization theory explores how the complex system of geological disasters evolves from disorder to orderly self-organization process. Catastrophe theory mainly describes the catastrophe behavior of nonlinear system in critical instability from a quantitative point of view, which provides a new way for geological disaster time prediction. Fractal theory discusses the self-similarity among all levels in the system from the perspective of geometry, and applies it to the process description and process prediction of geological disasters, making the complexity simple and qualitative and quantitative. Chaos dynamics discusses the irreversibility of nonlinear geological disaster system in its evolution process and the sensitivity of evolution behavior to initial values.
Artificial Neural Network (ANN) is a network composed of a large number of artificial neurons which are similar to natural neurons. The information processing of the network is realized by the interaction between neurons, and the storage of knowledge and information is characterized by the distributed physical connection between network elements. The learning and identification of the network depends on the dynamic evolution process of the connection weight system of each neuron. Artificial neural network is a very large-scale nonlinear continuous-time adaptive information processing system. At present, the application of artificial neural network has penetrated into many fields, providing a new modern way for learning recognition and calculation.
Artificial neural network is easy to use, and its information processing process is similar to the black box of human brain, as shown in figure 1-6. In practical application, people only deal with its surface input and output, and the internal information processing process is invisible. For those who don't understand the internal principle of neural network, they can also give their own problems to this network to solve. As long as you let your examples learn for a period of time, you can solve related problems. This is in line with the basic principles and ideas of geological disaster prediction theory.
Figure 1-6 Schematic diagram of neural network information processing
According to the simulation of different organizational levels and abstract levels of biological nervous system by artificial neural network, artificial neural network can be divided into many types. At present, there are more than 40 artificial neural network models. The multilayer feedforward neural network model (BP model for short) used for geological disaster prediction and evaluation is the most widely used and fastest developing neural network model at present, as shown in figure 1-7. It adopts a hierarchical structure, including an input layer, an output layer and one or more hidden layers.
Graph 1-7 BP network model
In fact, the BP model turns the input-output problem of a group of samples into a nonlinear optimization problem. We can think of this model as a mapping from input to output, which is highly nonlinear. If the number of input nodes is n and the number of output nodes is m, the neural network represents the mapping from N-dimensional Euclidean space to M-dimensional Euclidean space.
In the process of prediction and identification, the selection of standard samples is the key to successful prediction. Generally speaking, it is best for the learning samples to cover all the states of the predicted objects and be widely representative. When determining the network structure, generally speaking, the hidden three-layer BP model can simulate any continuous function with any accuracy. The number of hidden layer nodes is too small to effectively map the relationship between input layer and output layer; Too many, the convergence speed is too slow. Therefore, the selection of the number of nodes in the middle layer needs repeated calculus training to get the ideal number of nodes. In the calculation process, in order to improve the efficiency, the number of input nodes and the dimension of training samples can be appropriately reduced to increase the stability of the network. At the same time, the iterative convergence speed can be improved by adding pulse term method or adaptively adjusting learning rate and yoke gradient method.
When BP model is applied to geological disaster risk zoning, the corresponding prediction network can be established by learning the standard samples in the sample area, so it can be extended to the prediction area for prediction. The variables in the network input layer correspond to the main factors affecting the occurrence of geological disasters. Variables can be binary variables or specific observation data. Of course, because there are differences in units or orders of magnitude between variables, variable data must be standardized or standardized. The output layer corresponds to the classification of geological disaster prediction grade (extremely high, high, medium, low and extremely low), or the specific numerical expression of danger degree, such as stability coefficient and failure probability. This requires higher research accuracy and more detailed indicators in the sample area.
Eight, geological disaster risk analysis and GIS technology
Geographic Information System (GIS) is a new discipline which integrates computer science, information science, modern geography, remote sensing mapping, environmental science, urban science, space science, management science and modern communication technology. Specifically, GIS refers to the technical system of inputting, storing, retrieving, modifying, measuring, calculating, analyzing and outputting various geographic information and its carriers (words, data, charts, thematic maps, etc.). The main functions of GIS are collection, storage, management, analysis, output of various data, data maintenance and update, regional spatial analysis, multi-factor comprehensive analysis and dynamic monitoring. GIS can not only manage digital and literal (attribute) information like traditional database management system (DBMS), but also manage spatial (graphic) information. It can use various spatial analysis methods to comprehensively analyze different information, find the relationship between spatial entities, and analyze and deal with phenomena and processes distributed in a certain area. Modern geographic information system is developing towards intelligent GIS which can provide rich and comprehensive spatial analysis functions. Intelligent GIS has powerful spatial modeling function, which can build various professional, comprehensive and integrated geoscience analysis models, complete specific practical work and solve problems that only geoscientists can solve before.
GIS organically integrates various technologies and disciplines related to spatial information, and combines spatial and non-spatial data from different data sources through spatial operation and model analysis to provide useful information products for planning, management and decision-making. GIS provides us with a new way to know and understand geoscience information, and its powerful spatial analysis function and spatial database management ability provide us with a scientific and convenient new way to study regional geological disasters.
As one of the core technologies of digital earth, GIS has become an increasingly mature spatial data processing technology and method after nearly 40 years of development. It provides a new way to know and understand geoscience information, and has been widely used in land and resources investigation, environmental quality evaluation, regional planning and design, public facilities management and other fields. In the field of geological disaster research, the application of GIS technology has developed from the initial data management, digital input and graphic output of multi-source data collection to the use of digital elevation model and digital ground model, the extended analysis of the combination of GIS and disaster assessment model, the integration of GIS and decision support system, and the application of GIS virtual reality technology. , and has been gradually developed and deeply applied.
All kinds of geological disasters occur in a certain space and time range on the earth's surface. Although different types of geological disasters and different individuals of the same type of geological disasters have different forms and different formation mechanisms, they are all the results of the interaction between disaster breeding environment and trigger factors, and are closely related to spatial information. Using GIS technology can not only manage various geological disasters and their related information, but also analyze the statistical relationship between the occurrence of geological disasters and environmental factors from different time and space scales, and evaluate the probability and possible consequences of various geological disasters. Geological hazard zoning map belongs to comprehensive map, which has static characteristics in a certain period and needs to be updated constantly. Especially when new geological disasters occur, they should be revised in time. Because of the spatial analysis, mapping function and visualization of GIS technology, GIS technology has developed rapidly in the study of geological disaster zoning. The systematic research on the risk, vulnerability and risk assessment of geological disasters based on GIS software has gradually become the development direction in this field, and it may be combined with network technology in the near future.
Foreign countries, especially developed countries, have done a lot of work on the application of GIS technology in the field of geological disasters. Since 1980s, from data management, multi-source data collection, data input and graphic output, to the use of digital elevation model and digital terrain model, the extended analysis of the combination of GIS and disaster assessment model, the integration of GIS and decision support system (DSS) and the use of GIS virtual reality technology, the application of GIS technology has been continuously developed and widely used. In the field of landslide disaster research, the application of GIS technology has been relatively mature, mainly reflected in the following aspects:
(1) Establish a landslide disaster information management system based on GIS. For example, Keane James M. (1992), Bahari Wan (1998), Bliss Norman B. (1998) and so on. GIS is applied to the management of historical landslide disaster data and the graphic expression of prediction results.
(2) GIS technology is combined with various evaluation models and applied to landslide risk prediction. For example, mathura (1987), lekkas e (1995), Randall (1998), Dhaka Amod sagar (1999) and so on. Combining the spatial analysis function of GIS with the prediction model, the spatial superposition of landslide prediction factors is completed and the landslide risk assessment is carried out.
(3) Risk analysis, prediction and management of landslide disaster based on GIS. Such as Ellen (1994), Leroy (1996), Bunza (1996), castaneda Oscar E. (1998), Atkinson (1998). Aleotti(2000) starts from the factors that affect the risk of landslide disaster, uses the spatial analysis function of GIS to superimpose the factors, realizes the risk assessment and manages the disaster information in combination with the information management function of GIS, and finally makes management decisions, thus achieving the purpose of disaster prevention and reduction. At present, the close combination of RS and GIS has been basically realized in the field of landslide disaster prediction abroad, and the combination of 3S technology has also been realized in some projects.
Geological hazard assessment based on GIS technology started late in China, and there is no mature and practical GIS system for geological hazard prediction and assessment. Jiang Yun and Wang Lansheng (1994) applied GIS technology to the management and control of ground rock mass stability in mountainous cities, and took Chongqing as a typical research object to predict the deformation and failure of ground rock mass in time and space. At the same time, by analyzing the restrictive relationship between urban geological environment and land engineering utilization, using GIS information storage, query, spatial superposition operation and DEM model, the soil fertility grade is divided and the comprehensive evaluation zoning map of slope stability is compiled. Lei and Jiang Xiaozhen (1994) applied GIS technology to the evaluation of karst collapse, and completed the risk evaluation and zoning of karst collapse in the study area. Chengdu University of Technology (1998) cooperated with China Geological Environment Monitoring Institute and the Three Gorges Geological Disaster Prevention and Control Headquarters of the Ministry of Land and Resources to develop and test the "Geological Disaster Information System and Prevention Decision Support System", and initially established a national geological disaster investigation and comprehensive evaluation system. China Institute of Land and Resources Economics, China Geo University, Institute of Karst Geology of Chinese Academy of Geological Sciences, and Physical Geology Data Center of Ministry of Land and Resources jointly carried out the key project of "National Geological Disaster Risk Zoning", and conducted the risk assessment of small landslides, mudslides and karst collapses in China based on GIS by using the domestic software MAPGIS (including geological disaster risk assessment, vulnerability assessment and risk zoning). Zhu Liangfeng and others developed a set of geological disaster risk assessment system RISKANLY with domestic copyright on MAPGIS software platform. This set of geological disaster risk analysis based on GIS technology is not only feasible, but also advanced in technology, which represents the development direction of geological disaster risk analysis. Of course, both the risk analysis model of geological disasters and the regional socio-economic vulnerability analysis model need to be further studied and developed in practice, which obviously should be continuously developed with the gradual deepening of human understanding of the essential attributes of geological disasters.
With the rapid development of China's social economy and the acceleration of urbanization, the breadth and depth of geological disasters such as collapse, landslide, debris flow and ground collapse are also increasing rapidly, so it is necessary to pay more attention to the regional spatio-temporal prediction of geological disasters. The causes of geological disasters are complex and there are many related factors, which are closely related to spatial information. Therefore, GIS technology can not only manage all kinds of spatial information related to geological disasters, but also analyze the statistical relationship between the occurrence of geological disasters and environmental factors from different spatial and time scales, and evaluate the occurrence risk and possible disaster scope of geological disasters. Therefore, the risk assessment and zoning of geological disasters based on GIS will play an important role in the future social and economic development of China.
Nine. abstract
The risk assessment of geological disasters involves two important aspects: one is the possibility of geological disasters, and the other is the ability of human, society and environment to resist geological disasters. Therefore, the definition of geological disasters adopts the international term geological disasters. This book follows the principles of scientificity and universality, combines with the representative expressions of terms that have been initially formed in the field of geological disaster risk assessment in China in recent years, and defines the basic terms involved in geological disaster risk assessment on the basis of the unified definition proposed by UNESCO as follows:
(1) danger degree h (danger). The probability of potential geological disasters occurring in a certain area for a certain period of time.
(2) Vulnerability 5 (Vulnerability). Vulnerability can be expressed by 0- 1, where 0 means no loss and 1 means total loss.
(3) Disaster-bearing body E (dangerous element). Various objects threatened by geological disasters in a specific area include population, property, economic activities, public facilities, land, resources and environment.
(4) Risk degree R (risk). The disaster-bearing body may be attacked by various geological disasters, causing direct and indirect economic losses, casualties and environmental damage. Risk is equal to the product of risk, vulnerability and the value of the disaster-bearing body.
Risk degree (R)= risk degree (H)× vulnerability degree (V)× disaster-bearing body value (E)