CN114048944A - Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster - Google Patents
Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster Download PDFInfo
- Publication number
- CN114048944A CN114048944A CN202111175541.4A CN202111175541A CN114048944A CN 114048944 A CN114048944 A CN 114048944A CN 202111175541 A CN202111175541 A CN 202111175541A CN 114048944 A CN114048944 A CN 114048944A
- Authority
- CN
- China
- Prior art keywords
- area
- type
- early warning
- geological disaster
- factors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Abstract
The invention discloses a method for estimating evacuation of population and damage of houses under geological disasters induced by rainstorm, which specifically comprises the following steps: the method comprises the steps of collecting basic information of a geological disaster risk prevention area and residential buildings in the prevention area in an area to be researched, obtaining real-time disaster situations of the area to be researched in a certain time range in the future by combining a spatial interpolation technology according to future disaster field forecast information obtained by weather forecast, and obtaining early warning levels of different geological disaster risk prevention areas at different moments, population to be evacuated and total number of damaged houses at the last forecast moment under the forecast situation through combined use of hierarchical clustering, a geographic detector and a mixed naive Bayesian model. According to the three indexes, the risk of inducing the geological disaster by the rainstorm can be comprehensively evaluated, and the method can be used for circularly calculating along with the continuous updating of weather forecast products.
Description
Technical Field
The invention is suitable for the technical field of meteorological disaster prevention and reduction, and particularly relates to an estimation method for people to be evacuated and houses to be damaged under a geological disaster induced by rainstorm.
Background
The method is characterized in that most of southeast hilly areas and western parts of China are mountain environments, geological disasters are often induced under the action of rainstorm disasters, and serious challenges are brought to life health and property safety of people, at present, a corresponding monitoring system is a common means for reducing the danger of the geological disasters, and a geological disaster monitoring and early warning system (publication number: CN112382047A) discloses an early warning system, does not explain the selection process of early warning indexes in detail, and the standard for determining whether to send out early warning is the ratio of the area enclosed by historical data and actual measurement data respectively, so that scientific research is needed; the method for monitoring and early warning of geological disasters under the support of big data (publication number: CN108961688A) can only obtain the probability of occurrence of geological disasters and determine whether to early warn or not according to the probability, does not relate the probability of occurrence of the disasters with consequences, does not describe the basis of selection of historical data, and may bring a series of problems such as multiple collinearity among independent variables. The utility model relates to a geological disaster monitoring and early warning method and a device thereof (publication number: CN106960263A), which is determined only according to the occurrence frequency when selecting the pregnancy disaster factors and does not consider the problem of spatial heterogeneity of various disaster factors.
Generally speaking, the current geological disaster monitoring and early warning system mainly has the following defects: (1) at present, no reasonable step or standard exists in the aspect of selection of disaster-causing factors for a related method for geological disaster early warning, and (2) the related method is based on occurrence of phenomena when historical data are applied, has no specific disaster result and cannot provide related basis for decision-making of decision makers.
Disclosure of Invention
The invention aims to provide an estimation method for evacuating population and damaging houses under a rainstorm induced geological disaster, which aims to solve the problems in the prior art and adopts the following technical scheme:
a prediction method for people to be evacuated and houses to be damaged under a rainstorm induced geological disaster is characterized by comprising the following steps:
step (1), establishing a historical geological disaster database, and acquiring data recorded by monitoring points in each precautionary area;
step (2), classifying all risk prevention areas in the research area by adopting a hierarchical clustering method based on the recorded data;
step (3), determining a static driving factor corresponding to each type of precaution area set through combined iterative analysis of a geographic detector and a mixed naive Bayesian model;
step (4), based on the weather forecast data of a certain time length in the future at the current moment, determining the probability of the damage of structural members of residential buildings of different types in different precautionary areas at different times according to a mixed naive Bayes model which is obtained by training and meets the performance requirements;
step 5, determining people to be evacuated according to the prediction result of the mixed naive Bayes model obtained by training on the early warning level;
step (6), calculating expected values of the number of damaged houses according to the prediction results of the trained mixed naive Bayes model on the early warning levels and the house damage probability;
and (7) according to the updated weather forecast data, circulating the steps 4-6 to obtain the updated estimation results of the corresponding evacuated people and the damaged houses.
The historical geological disaster library in the step (1) refers to:
(1) the method comprises the steps that sample data recorded by real-time monitoring points in all geological disaster risk prevention areas in history comprises early warning grades (0 grade without early warning, the early warning can be in grades 1-4 during early warning, and the early warning grades respectively correspond to existing blue, yellow, orange and red early warnings in China, wherein the early warning grades of each geological disaster risk prevention area at different moments are obtained by obtaining early warning information issued by weather stands at different moments), the rainfall intensity (rainfall per hour), the water content and the temperature of soil at a certain depth and data corresponding to all static driving candidate factors of the monitoring points. (2) The data that unmanned aerial vehicle was obtained by taking photo by plane, including the residential buildings quantity and the type that appear structural member destruction in this risk prevention area.
Geological disaster risk prevention area refer to the risk prevention area that sets up in natural disaster risk general survey, all be equipped with a real-time monitoring point in every different geological disaster risk prevention area, be equipped with automatic meteorological station and soil moisture meter on the real-time monitoring point, be used for measuring the soil water content of the intensity of rain, temperature and certain degree of depth at different moments respectively, these three variables are dynamic disaster-causing factors. The number of different types of residential buildings with structural component damage occurring at different moments is obtained by interpreting high-resolution aerial images obtained by hourly aerial shooting by an unmanned aerial vehicle.
In the step (2), the total time of 4-level early warning and the total number of damaged houses in different precautionary areas can be calculated according to the information recorded in the historical geological disaster database, and the two indexes are firstly standardized, namely the two indexes are subjected to standardization treatment
In the formula xisA standard value, x, of an index x corresponding to the ith risk prevention areaiAn index x corresponding to the ith risk prevention area is indicated,
the mean value, σ, of the values on the index x representing all the guard areasxThe standard deviation of the values on the index x for all the precautionary zones is represented. And clustering by using a hierarchical clustering algorithm, regarding each precautionary area as a class, calculating the Euclidean distance between every two precautionary areas, clustering the two precautionary areas with the shortest distance to form a new class, performing iterative circulation according to a specified class and class distance calculation rule to form a hierarchical clustering tree diagram, and dividing the hierarchical clustering tree diagram into g classes of precautionary area sets (class 1 and class 2 … g classes) with different risk degrees according to actual conditions.
In the step (3), firstly, a geographical detector is utilized to carry out preliminary screening on each type of precautionary area set independently, the static driving factor refers to a factor screened from a static driving candidate factor by the geographical detector, the static driving candidate factor comprises elevation, gradient, slope direction, geological structure type, stratum lithology, slope structure, land type, land covering type and whether passing through an earthquake zone, and the vegetation covering condition is represented by a normalized vegetation index NDVI.
Further, in step (3), when the geographic probe is used, the area element of the whole research area is first converted into a point element by using ArcGIS, the point element is expressed in a grid point form, the attribute value of a point having an intersection relationship with the precaution area set of the type is "1", and the attribute values of the other points are "0". Dividing all static driving candidate factors into type quantities, sorting the candidate factors which are numerical type, and then dividing the candidate factors into different types by adopting a natural breakpoint method or an equidistant method, wherein the dividing quantities of the candidate factor types enable the q value to be the maximum principle, and the q value is calculated according to the following formula:
wherein h is the number of divisions of the candidate factor type, NhAnd N is the number of point elements and the number of all point elements within the layer h, respectively.
And σ2The variance of the point element attribute values in the layer h and the variance of all the point element attribute values are respectively. And after the maximum q value of each static driving candidate factor is obtained, the first n candidate factors with the maximum q value are selected as the static driving factors corresponding to the precaution area set.
Further, in the step (3), a mixed naive Bayesian model is adopted, independent variables in samples to be classified are dynamic disaster-causing factors and static driving factors, conditional probability logarithm type dynamic disaster-causing factors are calculated by adopting a Gaussian Bayesian model, discrete type static driving factors are calculated by adopting a polynomial Bayesian model, dependent variables are early warning grades (for example, the early warning grades are divided into 5 grades in total of 0-4), all samples belonging to the precautionary zone set in a historical geological disaster database are divided into a training set and a testing set according to a certain proportion, and the training set is used for obtaining each univariate changeConditional probability of quantity P (x)j＝xjk|Ci) Calculated as follows:
in the formula, P (x)j＝xjk|Ci) When the early warning level is CiTime jth argument xj(j ═ 1,2, 3..) is given by xjkProbability (k denotes the argument x)jType k divided or take value k), μjiAs belonging to C in the training dataiIndependent variable x of leveljMean value of (a)jiAs belonging to C in the training dataiIndependent variable x of leveljStandard deviation of (2). N is a radical ofikIndicates the early warning level is CiTime independent variable xjGet xjkNumber of samples of (1), NiFor early warning grade to be CiThe number of samples of (1) is taken as alpha as a smoothing coefficient, h represents the number of categories of the early warning level, and if the early warning level is divided into 5 levels of 0-4, n is 5.
Will train the resulting P (x)j＝xjk|Ci) The value is applied to the test set, and for each sample in the test set, the early warning grade is Ci(i＝0～4)The probability of (c) is:
the final prediction result for this sample is P (C)i|x1＝x1k,x2＝x2k,…xj＝xjk) Maximum time corresponding to CiThe value is obtained. Traversing all samples in the test set to obtain the prediction early warning grade of each sample, judging whether the performance of the classifier meets the requirements (namely whether the overall prediction accuracy and the area under the ROC curve exceed a certain threshold value) by calculating the overall prediction accuracy and the area under the ROC curve, and if not, changing the number of the static driving factors from n to n +1, namely changing the front n with the maximum q valueAnd (4) taking the +1 static candidate driving factors as static driving factors, and performing loop iteration until the constructed Bayesian classifier meets the performance requirement.
The static driving factor corresponding to each type of precaution area set (i.e. the static driving factor corresponding to each precaution area) can be obtained by repeating the above steps.
Further, in the step (4), a weather forecast product obtained by the WRF mode system is used to obtain a forecast point of the forecast hourly rainfall of the area to be researched within a certain time (such as 72 hours) in the future with spatial distribution difference and generate a corresponding thiessen polygon, the rainfall information of the forecast point is attached to the thiessen polygon with a spatial intersection relationship, and the hourly rainfall of each geological disaster risk prevention area is the same as the hourly rainfall of the thiessen polygon where the monitoring point is located. The method comprises the steps that a HYDROUS-1D model is utilized to simulate the water content of soil at a certain depth at monitoring points in each precautionary area at different moments, the input of the model is the hourly rainfall obtained by weather forecast, initial conditions need to be set for the model before the model is used according to the land cover types of the monitoring points in the different geological disaster risk precautionary areas and the actually measured initial soil water content at the starting forecast moment, and model parameters are calibrated according to the historical soil water content change conditions. And when the rainfall forecast information at the subsequent moment is updated, synchronously setting the actually measured soil water content at the subsequent initial forecast moment as the initial soil water content. Hourly temperature forecast data of monitoring points in each risk zone are also obtained through a WRF mode system. Determining the corresponding independent variable type according to the precaution area set to which the precaution area z belongs, and judging the probability that a certain precaution area z is in different early warning levels at the moment t according to the value or the type value of each variable:
in the formula
The 1 st argument, which represents the monitoring point in the preventive zone z at time t, will be
C of maximum valueiThe value is determined as the warning level of the precaution area z at that moment.
The hourly aviation image data obtained by the unmanned aerial vehicle can be obtained at different early warning levels C after being interpreted and countediTotal number N of b-type residential building damages in lower g-type geological disaster risk prevention area setbigThe total number N of residential buildings of the type corresponding to the type in the precaution area setbgAnd early warning level CiFor a total time tcigThe ratio of the products is the probability of the destruction of different types of residential buildings under different early warning levels in the set of the g-th class geological disaster risk prevention area, namely
Probability of destruction of the b-type residential building structural member in the risk containment zone z belonging to the g-type containment zone set at time t
Can be calculated as follows:
further, in the step (5), when the precautionary area z has a corresponding evacuation level (level 4, red warning) at any time t, all residents in residential buildings in the precautionary area need to evacuate, and the total number of people to be evacuated under the forecast scenario can be obtained by traversing all the precautionary areas.
Further, in step (6), n is performed on each type b residential building in the precautionary zone based on the monte carlo method according to the probability of the destruction of the type b residential building structural member in the precautionary zone z at the time t obtained in step (4)simuSubsampling, each comprising a time sequence of whole states, the final result being a breakdown if a certain residential building breaks down at time t, if nsimuTotal n in subsamplesfailureSecond damage, the damage probability of b-type buildings under the current forecast situation
The expected value of the number of houses damaged by the b-type buildings in the risk prevention zone z is
The total number of the type b residential buildings in the precaution area z; traversing all the building types in the precaution area z to obtain the total quantity of the damaged buildings in the precaution area; and traversing all the precautionary areas to obtain the total number of the houses damaged in the research area under the forecasting scene.
Further, the method for obtaining rainfall forecast information in the step (4) is repeated after the current forecast time is a certain time (for example, after 3 hours), and the updated rainfall information is substituted into the steps (5) and (6), so that the updated estimation results of the corresponding evacuation population and the damaged house can be obtained.
The invention has the following advantages:
(1) based on the combined use of hierarchical clustering, a geographic detector and a mixed naive Bayes method, the difference of different risk prevention areas in the degree of danger and the spatial heterogeneity of each disaster causing factor are comprehensively considered through an iterative algorithm, and a method for differentially selecting disaster causing factor indexes for different prevention areas is provided, so that the effect of reducing the dimension of the disaster causing factors is realized, the calculation cost in the actual process can be obviously reduced in future pre-disaster forecast, the calculation efficiency is improved, and the method is of great importance for the real-time forecast hour by hour before disaster.
(2) By applying the method and the system, the probability of occurrence of each early warning level at each moment can be forecasted before the disaster, the expected values of the number of people to be evacuated and the number of damaged houses are judged on the basis of the forecasted probability, the probability of occurrence of geological disasters is linked with the actual consequences, a more direct disaster result is obtained, and the visual understanding of a decision maker on possible disasters is enhanced.
Drawings
Fig. 1 is a flow chart of an estimation method for people to be evacuated and houses to be damaged under a rainstorm induced geological disaster provided by the invention.
Fig. 2 is a schematic diagram of a geological disaster risk prevention area and monitoring points in the prevention area.
FIG. 3 is a hierarchical clustering result of 30 risk prevention areas of the area to be studied.
FIG. 4 shows the result of the elevation classification of the area to be studied.
Detailed Description
The method is used for evaluating the geological disaster risk of a certain village of the insane town of the Lingan region in Hangzhou city of Zhejiang as a research region. The research area is located in the Anhui island Stone town of Hangzhou city in Zhejiang province, the east longitude is 118.87-119.02 degrees, the north latitude is 30.19-30.35 degrees, the total area of the whole town is 139.1 square kilometers, 16 administrative villages are shared in the town, and the general population is about 2.6 million people. In the period of the typhoon Liqima in 2019, the island stone town generates serious geological secondary disasters under the influence of typhoon rainstorms, and the severe geological secondary disasters greatly threaten the life health and property safety of local residents. At present, the temporary security area has developed refined geological disaster risk management and control work, hundreds of geological disaster risk prevention areas are defined, one area and one code for each prevention area are realized, and early warning information issued by the weather station corresponding to the prevention areas is accepted.
In order to dynamically evaluate the risk of inducing a geological disaster by rainstorm, the technical route provided by the invention is adopted, as shown in fig. 1, and the steps are as follows:
step 1: and establishing a historical geological disaster database. In the first national natural disaster comprehensive risk screening work, the temporary safety area in Hangzhou city of Zhejiang province carries out the definition work of the geological disaster risk prevention area, a real-time monitoring point is arranged in the precaution area, as shown in figure 2, an automatic weather station and a soil moisture meter are arranged on the real-time monitoring point and are respectively used for measuring the rainfall intensity, the temperature and the soil moisture content at a certain depth at different moments, the three variables are dynamic disaster factors, the detection value with the time mark can be transmitted to a historical geological disaster library in real time through a wireless transmission technology in the measurement process, the corresponding early warning grade of each precaution area at different time is obtained according to the early warning information issued by the meteorological station in the temporary security area, if the meteorological station issues a geological disaster yellow early warning at a certain day 12, before the early warning information of the new level is released, the early warning levels corresponding to all the moments are yellow early warnings (the corresponding level is 2 levels).
The static driving candidate factors comprise elevation, gradient, slope direction, geological structure type, stratigraphic lithology, slope structure, land type, land cover type, vegetation cover condition (normalized vegetation index (NDVI)), seismic zone (whether passing or not) and the like. Elevation data can be DEM data with ALOS resolution of 12.5m, the gradient and the slope direction can be directly calculated through tool kits related to the gradient and the slope direction in ArcGIS, the geological structure type, the stratigraphic lithology, the slope structure and the earthquake zone (whether the earthquake zone passes) can be obtained through consulting geological survey reports of the Hangzhou city temporary region to obtain corresponding survey result vector files, the land type can be land type data obtained through third land survey in China, the land cover type can be obtained through referring to annual land cover products (CLCD) derived based on Landsat satellite images, and the vegetation cover condition (normalized vegetation index (NDVI)) can be obtained through calculation after radiation calibration, atmospheric correction and geometric correction are carried out on remote sensing images of the research region through ENVI software. And obtaining the static driving candidate factor corresponding to the position of the monitoring point in each precaution area by a space connection tool in ArcGIS after each layer of the index is obtained.
In the whole rainfall process of each rainstorm disaster, the unmanned aerial vehicle takes an aerial photograph at the end of each hour, and counts the number and types of the residential buildings damaged by the structural members in each risk prevention area after visual interpretation, so as to obtain the number and corresponding types of the residential buildings damaged at the moment.
Step 2, before the new rainstorm disaster comes, the total time of 4-level early warning and the total number of damaged houses in different precautionary areas can be calculated through the information recorded in the historical geological disaster database, and the two indexes are standardized firstly, namely the two indexes are standardized
In the formula xisA standard value, x, of an index x corresponding to the ith risk prevention areaiIs shown asThe indices x corresponding to the i risk prevention areas,
the mean value, σ, of the values on the index x representing all the guard areasxThe standard deviation of the values on the index x for all the precautionary zones is represented.
Based on the two indexes, clustering is performed by using a hierarchical clustering algorithm, and the number of precaution area sets with different degrees of danger, which need to be divided into all precaution areas, is determined according to actual conditions by using a ' classification ' -system clustering ' module in general software SPSS in the field of statistics, wherein 30 geological disaster risk precaution areas are shared in a certain village in the island and stone town, and can be grouped into 3 types, as shown in FIG. 3, the precaution area sets with high, medium and low degrees of danger are respectively from right to left and respectively correspond to 3, 2 and 1 levels.
And 3, firstly, carrying out preliminary screening on static driving candidate factors by utilizing a geographic detector for each type of precaution area set, taking a high-risk degree precaution area set as an example, firstly, converting the whole to-be-researched area into point elements by utilizing a surface-to-point tool in ArcGIS software on the surface elements of the whole to-be-researched area, wherein the attribute value of a point which has an intersection relation with the high-risk degree precaution area set is '1', and the attribute values of the other points are '0'. Dividing all static driving candidate factors into type quantities, sorting the candidate factors which are numerical type, then dividing the candidate factors into different types by adopting a natural discontinuous point method or an equal spacing method, taking elevation data as an example, wherein the dividing result is shown in figure 4, and performing serial number assignment by using a re-classifying tool in ArcGIS software, wherein the dividing quantities of all candidate factor types enable the maximum q value, determining a point element set in different layers based on the spatial intersection relation between each type surface (layer) and the point element after each index dividing type, and calculating the q value according to the following formula:
wherein h is the number of divisions of the candidate factor type, NhAnd N is the number of point elements and all points in the layer h, respectivelyThe amount of the element.
And σ2The variance of the point element attribute values in the layer h and the variance of all the point element attribute values are respectively. And after the maximum q value of each static driving candidate factor is obtained, the first n candidate factors with the maximum q value are selected as the static driving factors corresponding to the precaution area set. For the high-risk degree precaution area set, after calculation, the first 4 static driving candidate factors with the maximum q value are selected as the static driving factors, wherein the static driving candidate factors are respectively an elevation, a land type, a seismic zone (whether the seismic zone passes through), and the like, and a land coverage type, and the types of the static driving factors are shown in the following table
Based on python software, a mixed naive Bayesian model is adopted, independent variables in samples to be classified are dynamic disaster-causing factors and static driving factors, conditional probability logarithm type dynamic disaster-causing factors are calculated by adopting a Gaussian Bayesian model, discrete type static driving factors are calculated by adopting a polynomial Bayesian model, dependent variables are early warning levels (0-4), all samples belonging to the precautionary area set in a historical geological disaster situation database are divided into a training set and a testing set according to a certain proportion, and the conditional probability P (x) of each univariate is obtained through the training setj＝xjk|Ci) Calculated as follows:
in the formula, P (x)j＝xjk|Ci) When the early warning level is CiTime jth argument xj(j ═ 1,2, 3..) is given by xjkProbability (k denotes the argument x)jType k divided or take value k), μjiAs belonging to C in the training dataiIndependent variable x of leveljMean value of (a)jiFor trainingIn the data belong to CiIndependent variable x of leveljStandard deviation of (2). N is a radical ofikIndicates the early warning level is CiTime independent variable xjGet xjkNumber of samples of (1), NiFor early warning grade to be CiWhere α is a smoothing coefficient, may be taken to be 1, and n represents the number of classes of the warning level, here 5.
Will train the resulting P (x)j＝xjk|Ci) The value is applied to the test set, and for each sample in the test set, the early warning grade is Ci(i＝0～4)The probability of (c) is:
the final prediction result for this sample is P (C)i|x1＝x1k,x2＝x2k,…xj＝xjk) Maximum time corresponding to CiThe value is obtained. And traversing all samples in the test set to obtain the prediction early warning grade of each sample, and judging whether the classifier meets the requirement or not by calculating (1) the overall prediction accuracy and (2) an AUC value obtained by calculating the ROC curve, wherein the model can be considered to meet the requirement when the overall prediction accuracy exceeds 85% and the AUC value is more than 0.9.
If the requirement is not met, the number of the static driving factors can be increased by one, for the embodiment, the first 5 static candidate driving factors with the maximum q value are used as the static driving factors, and the overall prediction accuracy value and the AUC value are recalculated until the constructed bayesian classifier meets the performance requirement.
Repeating the above steps for the other two risk prevention area sets with risk degrees to obtain the static driving factor corresponding to each type of prevention area set (i.e. the static driving factor corresponding to each prevention area).
And 4, obtaining predicted hourly rainfall forecast points of the to-be-researched area with spatial distribution difference in the future 72 hours by using a weather forecast product obtained by a WRF mode system, and generating a corresponding Thiessen polygon, wherein the specific method comprises the following steps: and reading the ncl format file output by the WRF by using an nclread function in Matlab, so that rainfall on grid points with different longitudes and latitudes and at different moments can be obtained. The spatial resolution of the obtained rainfall was 5km and the temporal resolution was 1 h.
Attaching the rainfall information of the forecast points to the Thiessen polygons with space intersection relations, wherein the hourly rainfall of each geological disaster risk prevention area is the same as the hourly rainfall of the Thiessen polygons where the monitoring points are located in the areas. The method comprises the steps that a HYDROUS-1D model is utilized to simulate the water content of soil at a certain depth at monitoring points in each precautionary area at different moments, the input of the model is the hourly rainfall obtained by weather forecast, initial conditions need to be set for the model before the model is used according to the land cover types of the monitoring points in the different geological disaster risk precautionary areas and the actually measured initial soil water content at the starting forecast moment, and model parameters are calibrated according to the historical soil water content change conditions. And when the rainfall forecast information at the subsequent moment is updated, synchronously setting the actually measured soil water content at the subsequent initial forecast moment as the initial soil water content. Hourly temperature forecast data of monitoring points in each risk zone are also obtained through a WRF mode system. Determining the corresponding independent variable type according to the precaution area set to which the precaution area z belongs, and judging the probability that a certain precaution area z is in different early warning levels at the moment t according to the value or the type value of each variable:
in the formula
The 1 st argument, which represents the monitoring point in the preventive zone z at time t, will be
C of maximum valueiThe value is determined as the warning level of the precaution area z at that moment.
The hourly aviation image data obtained by the unmanned aerial vehicle can be obtained at different early warning levels C after being interpreted and countediLower category g geological disaster riskTotal number N of b-type residential building damage in precaution area setbigThe total number N of residential buildings of the type corresponding to the type in the precaution area setbgAnd early warning level CiFor a total time tcigThe ratio of the products is the probability of the destruction of different types of residential buildings under different early warning levels in the set of the g-th class geological disaster risk prevention area, namely
Probability of destruction of the b-type residential building structural member in the risk containment zone z belonging to the g-type containment zone set at time t
Can be calculated as follows:
and 5: determining evacuation-worthy population
When the precaution area z has the grade 4 (red early warning) at any time t, all residents in residential buildings in the precaution area need to evacuate, and the total number of the population to be evacuated under the forecasting scene can be obtained by traversing all the precaution areas.
Step 6: calculating an expected value for the number of damaged houses
According to the probability of the destruction of the b-type residential building structural members in the precaution area z at the moment t obtained in the step 4, n is carried out on each b-type residential building in the precaution area based on the Monte Carlo methodsimuSubsampling, each comprising a time sequence of whole states, the final result being a breakdown if a certain residential building breaks down at time t, if nsimuTotal n in subsamplesfailureSecond damage, the damage probability of b-type buildings under the current forecast situation
The expected value of the number of houses damaged by the b-type buildings in the risk prevention zone z is
The total number of the type b residential buildings in the precaution area z; traversing all the building types in the precaution area z to obtain the total quantity of the damaged buildings in the precaution area; and traversing all the precautionary areas to obtain the total number of the houses damaged in the research area under the forecasting scene.
And 7: and (3) repeating the step (4) after a certain time (such as 3 hours) at the current forecasting time to obtain the rainfall forecasting information, and substituting the updated rainfall information into the steps (5) and (6) to obtain the updated estimation results of the corresponding evacuation population and the damaged house.
Claims (7)
1. A prediction method for people to be evacuated and houses to be damaged under a rainstorm induced geological disaster is characterized by comprising the following steps:
step (1), establishing a historical geological disaster database, and acquiring data recorded by monitoring points in each precautionary area;
step (2), classifying all risk prevention areas in the research area by adopting a hierarchical clustering method based on the recorded data;
step (3), determining a static driving factor corresponding to each type of precaution area set through combined iterative analysis of a geographic detector and a mixed naive Bayesian model;
step (4), based on the weather forecast data of a certain time length in the future at the current moment, determining the probability of the damage of structural members of residential buildings of different types in different precautionary areas at different times according to a mixed naive Bayes model which is obtained by training and meets the performance requirements;
step 5, determining people to be evacuated according to the prediction result of the mixed naive Bayes model obtained by training on the early warning level;
step (6), calculating expected values of the number of damaged houses according to the prediction results of the trained mixed naive Bayes model on the early warning levels and the house damage probability;
and (7) according to the updated weather forecast data, circulating the steps 4-6 to obtain the updated estimation results of the corresponding evacuated people and the damaged houses.
2. The method of claim 1, wherein the historical geological disaster library is characterized by comprising the following steps:
(1) sample data recorded at a real-time monitoring point in all geological disaster risk prevention areas in history, wherein each sample data comprises data corresponding to early warning grade, rain intensity, soil water content and temperature at a certain depth and each static driving candidate factor of the monitoring point recorded at the same time;
(2) data obtained by aerial photography of the unmanned aerial vehicle comprise the number and types of residential buildings with structural component damage in different risk prevention areas;
geological disaster risk prevention area refer to the risk prevention area that sets up in natural disaster risk general survey, all be equipped with a real-time monitoring point in every different geological disaster risk prevention area, be equipped with automatic meteorological station and soil moisture meter on the real-time monitoring point, be used for measuring the soil water content of the intensity of rain, temperature and certain degree of depth at different moments respectively, these three variables are dynamic disaster-causing factors. The number of different types of residential buildings with structural component damage occurring at different moments is obtained by interpreting high-resolution aerial images obtained by hourly aerial shooting by an unmanned aerial vehicle.
3. The method for estimating the number of people to be evacuated and houses to be damaged under the condition of the rainstorm induced geological disaster as claimed in claim 1, wherein in the step (2), the total time of occurrence of the early warning levels corresponding to the evacuation in different precautionary areas and the total number of damaged houses can be calculated according to the information recorded in the historical geological disaster database, the clustering is performed by using a hierarchical clustering algorithm after the standardization processing, and the precautionary areas are divided into g types of precautionary area sets with different risk degrees according to the actual conditions.
4. The method for estimating the number of people to be evacuated and houses to be damaged under the condition of the rainstorm induced geological disaster according to claim 1, wherein in the step (3), firstly, each type of precautionary area set is individually subjected to preliminary screening by using a geographic detector, the static driving factor refers to a factor screened from static driving candidate factors by using the geographic detector, the static driving candidate factors comprise elevation, gradient, slope direction, geological structure type, stratum lithology, slope structure, land type, land cover type, vegetation cover condition and whether an earthquake zone passes through, and the vegetation cover condition is represented by normalized vegetation index NDVI; dividing all static driving candidate factors into type quantities, sorting the candidate factors which are numerical type, and then dividing the candidate factors into different types by adopting a natural breakpoint method or an equidistant method, wherein the dividing quantities of the candidate factor types enable the q value to be the maximum principle, and the q value is calculated according to the following formula:
wherein h is the number of divisions of the candidate factor type, NhAnd N is the number of point elements and the number of all point elements in the layer h respectively;
and σ2Respectively representing the variance of the attribute values of the point elements in the layer h and the variance of the attribute values of all the point elements; after the maximum q value of each static driving candidate factor is obtained, the first n candidate factors with the maximum q value are selected as the static driving factors corresponding to the precaution area set;
adopting a mixed naive Bayes model, wherein independent variables in samples to be classified are dynamic disaster factors and static driving factors, conditional probability logarithm type dynamic disaster factors are calculated by adopting a Gaussian Bayes model, discrete type static driving factors are calculated by adopting a polynomial Bayes model, dependent variables are early warning levels, all samples belonging to the precautionary area set in a historical geological disaster database are divided into a training set and a testing set according to a certain proportion, and the conditional probability P (x) of each univariate is obtained through the training setj＝xjk|Ci) Calculated as follows:
in the formula, P (x)j＝xjk|Ci) When the early warning level is CiTime jth argument xjIs taken as xjkWhere j ═ 1,2, 3.., k denotes the argument xjOf the kth type or take the kth value, mujiAs belonging to C in the training dataiIndependent variable x of leveljMean value of (a)jiAs belonging to C in the training dataiIndependent variable x of leveljStandard deviation of (d); n is a radical ofikIndicates the early warning level is CiTime independent variable xjGet xjkNumber of samples of (1), NiFor early warning grade to be CiThe number of samples of (1) is taken as alpha, a smoothing coefficient is taken as 1, and h represents the number of classes of the early warning level;
will train the resulting P (x)j＝xjk|Ci) The value is applied to the test set, and for each sample in the test set, the early warning grade is Ci(i＝0～4)The probability of (c) is:
the final prediction result for this sample is P (C)i|x1＝x1k，x2＝x2k，…xj＝xjk) Maximum time corresponding to CiA value; traversing all samples in the test set to obtain the prediction early warning grade of each sample, judging whether the performance of the classifier meets the requirement, if not, changing the number of the static driving factors from n to n +1, namely, taking the first n +1 static candidate driving factors with the maximum q value as the static driving factors, and performing cycle test until the constructed Bayesian classifier meets the performance requirement; the static driving factors corresponding to each type of precaution area set, namely the static driving factors corresponding to each precaution area set, can be obtained by repeating the steps.
5. The method for estimating the number of people to be evacuated and houses to be damaged under the condition of the rainstorm induced geological disaster as claimed in claim 1, wherein in the step (4), the WRF mode system is used for obtaining hourly rainfall and temperature of the forecast points in the area to be researched and attaching information to each risk prevention area; simulating the water content of soil at a certain depth at monitoring points of each precaution area at different moments by using a HYDROUS-1D model; when the rainfall forecast information at the subsequent moment is updated, synchronously setting the actually measured soil water content at the subsequent initial forecast moment as the initial soil water content; determining a corresponding independent variable type according to a precaution area set to which a precaution area z belongs, and judging the probability that a certain precaution area z is in different early warning levels at the moment t:
in the formula
The 1 st argument, which represents the monitoring point in the preventive zone z at time t, will be
C of maximum valueiSetting the value as the early warning grade of the precaution area z at the moment;
the hourly aviation image data obtained by the unmanned aerial vehicle can be obtained at different early warning levels C after being interpreted and countediTotal number N of b-type residential building damages in lower g-type geological disaster risk prevention area setbigThe total number N of residential buildings of the type corresponding to the type in the precaution area setbgAnd early warning level CiFor a total time tcigThe ratio of the products is the probability of the destruction of different types of residential buildings under different early warning levels in the set of the g-th class geological disaster risk prevention area, namely
Probability of destruction of the b-type residential building structural member in the risk containment zone z belonging to the g-type containment zone set at time t
Can be calculated as follows:
6. the method for estimating the number of people to be evacuated and the number of damaged houses under the rainstorm induced geological disaster as claimed in claim 1, wherein in the step (5), when the precautionary area z has a corresponding evacuation level at any time t, all residents in residential buildings in the precautionary area are evacuated, and the total number of people to be evacuated under the forecast scenario can be obtained by traversing all the precautionary areas.
7. The method for estimating the number of people to be evacuated and houses to be destroyed in the event of a geologic disaster induced by rainstorm according to claim 4, wherein in the step (6), n is performed on each type b residential building in the precautionary area based on the Monte Carlo method according to the probability of the destruction of the type b residential building structural member in the precautionary area z obtained in the step (4)simuSubsampling, each comprising a time sequence of whole states, the final result being a breakdown if a certain residential building breaks down at time t, if nsimuTotal n in subsamplesfailureSecond damage, the damage probability of b-type buildings under the current forecast situation
The expected value of the number of houses damaged by the b-type buildings in the risk prevention zone z is
The total number of the type b residential buildings in the precaution area z; traversing all the building types in the precaution area z to obtain the total quantity of the damaged buildings in the precaution area; and traversing all the precautionary areas to obtain the total number of the houses damaged in the research area under the forecasting scene.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111175541.4A CN114048944A (en) | 2021-10-09 | 2021-10-09 | Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111175541.4A CN114048944A (en) | 2021-10-09 | 2021-10-09 | Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114048944A true CN114048944A (en) | 2022-02-15 |
Family
ID=80205001
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111175541.4A Pending CN114048944A (en) | 2021-10-09 | 2021-10-09 | Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114048944A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115423157A (en) * | 2022-08-15 | 2022-12-02 | 中国水利水电科学研究院 | Dynamic early warning method and device for mountain torrent disaster risks |
| CN116304963A (en) * | 2023-05-25 | 2023-06-23 | 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) | Data processing system suitable for geological disaster early warning |
| CN116882764A (en) * | 2023-09-07 | 2023-10-13 | 北京国信华源科技有限公司 | Disaster risk management method based on region and historical data machine learning |
-
2021
- 2021-10-09 CN CN202111175541.4A patent/CN114048944A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115423157A (en) * | 2022-08-15 | 2022-12-02 | 中国水利水电科学研究院 | Dynamic early warning method and device for mountain torrent disaster risks |
| CN115423157B (en) * | 2022-08-15 | 2023-04-28 | 中国水利水电科学研究院 | Dynamic early warning method and device for mountain torrent disaster risk |
| CN116304963A (en) * | 2023-05-25 | 2023-06-23 | 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) | Data processing system suitable for geological disaster early warning |
| CN116882764A (en) * | 2023-09-07 | 2023-10-13 | 北京国信华源科技有限公司 | Disaster risk management method based on region and historical data machine learning |
| CN116882764B (en) * | 2023-09-07 | 2023-11-28 | 北京国信华源科技有限公司 | Disaster risk management method based on region and historical data machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mantovani et al. | Remote sensing techniques for landslide studies and hazard zonation in Europe | |
| Zhang et al. | Assimilation of precipitation-affected radiances in a cloud-resolving WRF ensemble data assimilation system | |
| CN114048944A (en) | Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster | |
| CN111666656A (en) | Rainfall estimation method and rainfall monitoring system based on microwave rainfall attenuation | |
| KR100982447B1 (en) | Landslide occurrence prediction system and predicting method using the same | |
| van Westen et al. | From landslide inventories to landslide risk assessment; an attempt to support methodological development in India | |
| CN109782373A (en) | A kind of sand-dust storm forecast method based on improved Naive Bayesian-CNN multiple target sorting algorithm | |
| CN111210483B (en) | Simulated satellite cloud picture generation method based on generation of countermeasure network and numerical mode product | |
| Michaelides et al. | Monitoring and forecasting air pollution levels by exploiting satellite, ground-based, and synoptic data, elaborated with regression models | |
| Ogashawara et al. | The use of optical remote sensing for mapping flooded areas | |
| CN116449331B (en) | Dust particle number concentration estimation method based on W-band radar and meteorological satellite | |
| Xin et al. | Seasonal differences in the dominant factors of surface urban heat islands along the urban-rural gradient | |
| Myint et al. | Categorizing natural disaster damage assessment using satellite-based geospatial techniques | |
| Lee et al. | Estimation of maximum daily fresh snow accumulation using an artificial neural network model | |
| Tyndall et al. | Sensitivity of surface air temperature analyses to background and observation errors | |
| Xu et al. | Identification and quantitative analysis of dust trajectories in the Hexi Corridor | |
| Hu et al. | Objective Verification of Clear-Air Turbulence (CAT) Diagnostic Performance in China Using In Situ Aircraft Observation | |
| Foroushani et al. | Spatial and temporal gradients in the rate of dust deposition and aerosol optical thickness in southwestern Iran | |
| Peterson et al. | The illumination of thunderclouds by lightning: Part 3: Retrieving optical source altitude | |
| Davis et al. | Post-hurricane michael damage assessment using ADCIRC storm surge hindcast, image classification, and LiDAR | |
| CN114722909A (en) | Solar flare time sequence classification method based on low-dimensional convolutional neural network | |
| Jacques | Comparison of the United States Precision Lightning Network™(USPLN™) and the Cloud-to-Ground Lightning Surveillance System (CGLSS) | |
| Ballari et al. | Spatial Association to Characterize The Climate Teleconnection Patterns In Ecuador Based On Satellite Precipitation Estimates | |
| Murata et al. | Assessing goodness of fit to a gamma distribution and estimating future projection on daily precipitation frequency using regional climate model simulations over Japan with and without the influence of tropical cyclones | |
| MORRIS et al. | Assessing the Impact of Non-Conventional Radar and Surface Observations on High-Resolution Analyses and Forecasts of a Severe Hailstorm. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |