CN116543528A - Regional landslide hazard early warning method based on rainfall threshold - Google Patents
Regional landslide hazard early warning method based on rainfall threshold Download PDFInfo
- Publication number
- CN116543528A CN116543528A CN202310569441.2A CN202310569441A CN116543528A CN 116543528 A CN116543528 A CN 116543528A CN 202310569441 A CN202310569441 A CN 202310569441A CN 116543528 A CN116543528 A CN 116543528A
- Authority
- CN
- China
- Prior art keywords
- landslide
- rainfall
- model
- area
- hazard
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000008878 coupling Effects 0.000 claims abstract description 24
- 238000010168 coupling process Methods 0.000 claims abstract description 24
- 238000005859 coupling reaction Methods 0.000 claims abstract description 24
- 238000007637 random forest analysis Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 238000011156 evaluation Methods 0.000 claims description 31
- 230000001186 cumulative effect Effects 0.000 claims description 14
- 238000010801 machine learning Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 12
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 9
- 238000005315 distribution function Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 8
- 238000011835 investigation Methods 0.000 claims description 7
- 238000007619 statistical method Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 5
- 230000035699 permeability Effects 0.000 claims description 5
- 238000012876 topography Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 238000010835 comparative analysis Methods 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims description 2
- 238000005192 partition Methods 0.000 claims description 2
- 238000009825 accumulation Methods 0.000 claims 1
- 238000009826 distribution Methods 0.000 abstract description 15
- 238000011160 research Methods 0.000 abstract description 6
- 238000005457 optimization Methods 0.000 abstract description 3
- 231100000656 threshold model Toxicity 0.000 description 11
- 239000002689 soil Substances 0.000 description 4
- LTINZAODLRIQIX-FBXRGJNPSA-N cefpodoxime proxetil Chemical compound N([C@H]1[C@@H]2N(C1=O)C(=C(CS2)COC)C(=O)OC(C)OC(=O)OC(C)C)C(=O)C(=N/OC)\C1=CSC(N)=N1 LTINZAODLRIQIX-FBXRGJNPSA-N 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000009137 wuling Substances 0.000 description 2
- 239000005995 Aluminium silicate Substances 0.000 description 1
- 201000004569 Blindness Diseases 0.000 description 1
- 240000002834 Paulownia tomentosa Species 0.000 description 1
- 235000010678 Paulownia tomentosa Nutrition 0.000 description 1
- 241000233805 Phoenix Species 0.000 description 1
- 235000012211 aluminium silicate Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001704 evaporation Methods 0.000 description 1
- 230000008020 evaporation Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008595 infiltration Effects 0.000 description 1
- 238000001764 infiltration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- NLYAJNPCOHFWQQ-UHFFFAOYSA-N kaolin Chemical compound O.O.O=[Al]O[Si](=O)O[Si](=O)O[Al]=O NLYAJNPCOHFWQQ-UHFFFAOYSA-N 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/10—Alarms for ensuring the safety of persons responsive to calamitous events, e.g. tornados or earthquakes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Biomedical Technology (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Molecular Biology (AREA)
- Pure & Applied Mathematics (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Emergency Management (AREA)
- Geology (AREA)
- Game Theory and Decision Science (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Computer Vision & Pattern Recognition (AREA)
Abstract
The invention relates to a regional landslide hazard early warning method based on a rainfall threshold, and belongs to the field of disaster early warning. The method comprises the following steps: s1: establishing an information quantity-random forest IV-RF and information quantity-BP neural network IV-BPNN landslide vulnerability prediction coupling model based on information quantity and negative sample selection; s2: establishing a continuous probability rainfall threshold model based on multi-function fitting; s3: regional landslide hazard early warning based on vulnerability and continuous probability rainfall threshold. According to the invention, IV is respectively coupled with the RF model and the BPNN model to obtain two coupling models, and optimization is carried out in the two models to obtain a landslide susceptibility value of a research area with higher precision. Coupling the landslide susceptibility value with the continuous probability value to obtain the continuous probability landslide risk value, reflecting the space distribution characteristic of rainfall and realizing more accurate landslide risk early warning classification.
Description
Technical Field
The invention belongs to the field of disaster early warning, and relates to a regional landslide hazard early warning method based on rainfall threshold.
Background
The mountain geology and topography conditions are complex, and geological disasters frequently occur. Landslide is the most common type of geological disasters, and threatens people's life and property safety, living environment and production life. In order to meet regional disaster prevention and reduction and homeland resource planning, regional landslide hazard early warning is needed. The regional landslide hazard is generally considered as a coupling result of the space probability and the time probability of landslide occurrence, the landslide susceptibility can represent the space probability, and the accuracy of the landslide hazard pre-warning is determined on the premise of the landslide hazard pre-warning; landslide hazard considers not only internal factors but also the influence of external factor dynamics, such as rainfall and earthquake.
In recent years, statistical analysis and machine learning models are applied to regional landslide susceptibility prediction results are remarkable, and statistical analysis models (such as information quantity models) have the characteristics of simplicity and convenience in operation, but are very dependent on subjective experience, and cannot reflect complex nonlinear relations between landslide and evaluation factors; the machine learning model (such as a random forest model and a BP neural network model) has blindness when selecting landslide non-samples, and predicts a region with complicated geological, topography and geomorphic conditions by adopting a single model, so that the prediction precision and reliability cannot be ensured.
The landslide time probability calculated by the traditional rainfall threshold model based on statistical analysis is discontinuous, the time probability level can be generally and roughly divided, the spatial distribution characteristic of rainfall cannot be reflected, and the accuracy and the efficiency of the traditional landslide hazard early warning are required to be improved.
Therefore, in order to fully play the roles of the statistical analysis model and the machine learning model in landslide vulnerability prediction (space probability), the problem that the traditional rainfall threshold model calculates the discontinuous time probability is solved, and the space probability and the time probability are coupled, so that the regional landslide risk early warning with higher precision and high efficiency is realized.
Disclosure of Invention
In view of the above, the invention aims to provide a regional landslide hazard early warning method based on rainfall threshold. Firstly, a statistical analysis model and a machine learning model are coupled to improve the regional landslide susceptibility prediction precision; and secondly, constructing a continuous probability rainfall threshold model based on multi-function fitting to obtain continuous landslide time probability values. And finally, multiplying the space probability and the time probability to obtain the regional landslide hazard value and grading the regional landslide hazard value so as to embody the space distribution characteristics of rainfall and improve the landslide hazard early warning efficiency.
In order to achieve the above purpose, the present invention provides the following technical solutions:
(1) As shown in fig. 1, the present invention proposes an information volume-random forest (IV-RF) and information volume-BP neural network (IV-BPNN) landslide vulnerability prediction coupling model selected based on the information volume and negative samples. A non-landslide point (0) is randomly selected based on a landslide extremely low and low easy occurrence area preliminarily defined by an IV model, and a 1-0 data set is formed by the non-landslide point (0) and the landslide point (1); and secondly, introducing information magnitude values in the data set as input layers of RF and BPNN, and establishing an IV-RF and IV-BPNN coupling model. The prediction accuracy of the single model and the coupling model is compared by a receiver operation characteristic curve (ROC) and a confusion matrix, and the reasonability of the easily-developed partitions of different models is compared based on the FR model.
a. Initial prediction
Dividing and numbering analysis areas by an evaluation unit (a grid unit or a slope unit); the evaluation factors collect data and information of landslide liability-related areas in aspects of topography, geological conditions, weather hydrology, human engineering activities and the like through means of data collection, remote sensing interpretation, field investigation and the like, preference is carried out according to expert experience or a statistical analysis method, evaluation factors with strong independence and non-collinearity are selected for assignment, various evaluation factors in an evaluation unit are used as landslide independent variables, the landslide liability of the evaluation unit is used as dependent variables, and an information quantity method is adopted for assignment of the evaluation factors (formula 1); the historical landslide information includes: a landslide sample database is established according to the spatial position, area, volume and contour range of the landslide;
wherein N is n Representing the area of landslide occurring in the n-th classification of the evaluation factor; n (N) 0 Indicating the total area where landslide occurs; s is S n Represents the nth classification area; s is S 0 Representing the total area of the investigation region; n (N) n /N 0 Representing the area ratio of landslide occurrence in the classification; s is S n /S 0 Representing the fractional area ratio; IV represents the information amount and reflects the landslide contribution degree.
And accumulating information values given by all evaluation factors of the evaluation units in the analysis area to obtain an area landslide susceptibility value, grading by adopting a natural breakpoint method, and finally obtaining a primary area landslide susceptibility prediction result.
b. Negative sample selection
On the processed basic evaluation factor layer, a proper data set is required to be selected as the input of a landslide susceptibility prediction model. Firstly, based on related vector data obtained by existing landslide point cataloging and field investigation, a landslide grid, namely a positive sample, is obtained through a spatial connection tool of arcgis, and the reference number is 1. Furthermore, since the selected vulnerability prediction model is a supervised machine learning model, a non-landslide grid, i.e., a negative sample, needs to be selected. In order to avoid that the selected non-landslide grids are real non-landslide grids, firstly water systems and residential areas are avoided, secondly on the basis of classification of landslide susceptibility values calculated by the information quantity model, random selection of the non-landslide grids is carried out in the extremely low susceptibility areas or the low susceptibility areas, and meanwhile the non-landslide grids are ensured to be out of the range of a landslide point buffer area, and the label is 0, as shown in fig. 2. For example, from 100 landslide boundary surface vector diagrams, obtaining 500 landslide grids with a size of 30m by 30m through a spatial connection tool of arcgis, as shown in fig. 3, wherein the grid size is determined by a scale and layer precision together; and selecting 500 non-landslide grids according to the negative sample selection mode to form a '0-1' data set containing 1000 data. A 70% sample of the data set was randomly selected as the machine learning training sample set, and a 30% sample test sample set.
c. Model coupling
In order to take the advantages of the information quantity model and the machine learning model into consideration, the two models are coupled, and the specific steps are as follows: taking the information magnitude of the '0-1' data set as an input layer of a random forest model and a BP neural network model, and carrying out model training; and finally grading the landslide susceptibility value predicted by the model.
d. Comparative analysis
After training is completed by using the sample data random forest model and the BP neural network model, whether the accuracy can meet the prediction requirement after the model is required to be quantized. Typically, the area under the ROC curve (subject operating characteristic curve) and AUC curve are used to evaluate the training effect of a supervised machine learning model. In addition, in the field of machine learning, especially for supervised machine learning, confusion matrices may also be employed to evaluate model effects.
e. Optimal model selection
To discuss the predictive effect of the model, the ratio of landslide point area to frequency is used to measure the predictive effect. For the ratio of the area of the landslide point to the frequency, in general, the lower the landslide susceptibility level is, the smaller the ratio of the area of the landslide point to the frequency should be, and the ratio of the area of the landslide point to the frequency should be gradually increased along with the increase of the susceptibility level.
(2) As shown in fig. 4, a continuous probability rainfall threshold model based on multi-function fitting is proposed.
The cumulative proportion P of the number of landslide based on different effective rainfall (EE) in the earlier stage and different rainfall duration days (D) EE And P D Will P EE And P D And multiplying to obtain time probability P, performing P-EE-D fitting in a form of multiplying a cumulative normal distribution function by a power function, and calculating the time probability P of landslide occurrence of a certain point when the early effective rainfall and the rainfall duration days of the certain point are obtained. And randomly selecting a certain number of landslide points which do not participate in modeling, and substituting the landslide points into the established model to verify the model accuracy.
a. Rainfall data
1) Effective rainfall in early stage
The early effective rainfall further considers the water loss caused by evaporation and runoff based on the accumulated rainfall, namely the real water content of the rainfall infiltrated rock-soil body is far lower than the accumulated record value [80] . The water content of the rock-soil body is increased due to rainfall infiltration, the pore water pressure of the rock-soil body is increased, the shear strength is reduced, and finally the slope is unstable to cause landslide. Therefore, the effective rainfall capacity in the prior period can reflect the water content change process of the rock-soil body in the rainfall process, and the expression of EE is as follows:
wherein R is 0 For the daily rainfall of landslide, k is the rainfall permeability coefficient, R i The rainfall on the i-th day before landslide is represented, and n is the number of days in which rainfall is considered. For the rainfall permeability coefficient k, generally according to the study areaDepending on the situation, most of the studies range from 0.72 to 0.86, with 0.84 being the most widely used.
2) Days of continuous rainfall
The duration of rainfall is specified as: the duration days from the beginning to the end of a rainfall event; it is generally considered that if the rainfall is less than 5mm for two consecutive days, i.e. the end of one rainfall event.
b.P EE And P D Probability of rainfall
Counting the accumulated ratio (P) of the number of landslide under different early-stage effective rainfall (EE) EE ) Fitting by adopting an accumulated normal distribution function to obtain P EE -EE curve, as shown in the following formula.
Counting the cumulative ratio (P) of the number of landslide under different rainfall duration days (D) D ) Fitting by using a power function to obtain P D -D curve, as shown in the following formula.
P(D)＝a+b×D c (4)
c. Continuous probability rainfall threshold model based on multi-function fitting
Will P EE And P D The time probability P is obtained through multiplication, and P-EE-D fitting is carried out in the form of the product of a cumulative normal distribution function and a power function, wherein the product is shown in the following formula.
d. Typical rainfall type landslide instance accuracy verification
In order to verify the precision of a continuous probability rainfall threshold model based on multi-function fitting, landslide points of which parts do not participate in modeling are randomly selected and substituted into the established rainfall model, and the higher the precision is, the more accurate the model is proved.
(3) Regional landslide hazard early warning method based on vulnerability and continuous probability rainfall threshold
The continuous probability landslide hazard early warning is to grade the continuous probability landslide hazard value by adopting a natural breakpoint method or an equidistant dividing method. The continuous probability landslide hazard value is obtained by multiplying a landslide vulnerability value by a rainfall type landslide continuous probability value fitted by a nonlinear equation, and different combinations of the landslide vulnerability value and the rainfall continuous probability value can be reflected, as shown in the formula. Compared with the traditional landslide hazard early warning, the continuous probability landslide hazard early warning is not singly and qualitatively overlapped in matrix, but is a quantitative interval division result. After the landslide susceptibility is determined, the landslide hazard early warning level is increased according to the increase of the rainfall continuous probability value, and the landslide hazard early warning level accords with the dynamic change characteristic. The continuous probability landslide hazard values are classified into 5 stages by an equidistant dividing method, and as shown in table 1, the stages are respectively an extremely high hazard zone (1.0-0.8), a high hazard zone (0.8-0.6), a medium hazard zone (0.6-0.4), a low hazard zone (0.4-0.2) and an extremely low hazard zone (0.2-0.0).
H＝P×S (6)
Where H is a continuous probability landslide hazard value, P is a continuous probability value (time probability), and S is a landslide susceptibility value (space probability).
TABLE 1 continuous probability landslide hazard value grading table
| Risk classification | Extremely high risk area | High risk area | Dangerous area in middle | Low risk area | Very low risk area |
The invention has the beneficial effects that: according to the invention, IV is respectively coupled with the RF model and the BPNN model to obtain two coupling models, and optimization is carried out in the two models to obtain a landslide susceptibility value of a research area with higher precision. In addition, the continuous probability value is obtained by fitting rainfall parameters by adopting a nonlinear equation to represent the landslide time probability, and the landslide susceptibility value and the continuous probability value are coupled to obtain the continuous probability landslide risk value, so that the space distribution characteristics of rainfall are reflected, and more accurate landslide risk early warning classification is realized.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objects and other advantages of the invention may be realized and obtained by means of the instrumentalities and combinations particularly pointed out in the specification.
Drawings
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in the following preferred detail with reference to the accompanying drawings, in which:
FIG. 1 is a regional landslide vulnerability prediction model study flow chart;
FIG. 2 is a negative sample selection schematic;
FIG. 3 is a grid cell;
FIG. 4 is a schematic diagram of a continuous probability rainfall threshold model based on multi-function fitting;
FIG. 5 is a study area;
FIG. 6 is a regional landslide vulnerability prediction graph based on a traffic model;
FIG. 7 is a regional landslide vulnerability prediction graph based on BPNN and RF;
FIG. 8 is a regional landslide vulnerability prediction graph based on IV-BPNN and IV-RF;
FIG. 9 is a comparison of ROC curves for four models;
FIG. 10 is a confusion matrix comparison for four models;
FIG. 11 is a graph showing the frequency ratio of the models at different stages;
FIG. 12 is an edge distribution and a cumulative distribution;
FIG. 13 is a cumulative frequency-to-duty fit based on the amount of active rainfall in the early period and the number of days of rainfall duration;
FIG. 14 is a continuous probability rainfall threshold based on a multi-function fit;
fig. 15 is a graph of continuous probability rainfall threshold accuracy verification based on a multi-function fit.
Detailed Description
Other advantages and effects of the present invention will become apparent to those skilled in the art from the following disclosure, which describes the embodiments of the present invention with reference to specific examples. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be noted that the illustrations provided in the following embodiments merely illustrate the basic idea of the present invention by way of illustration, and the following embodiments and features in the embodiments may be combined with each other without conflict.
Wherein the drawings are for illustrative purposes only and are shown in schematic, non-physical, and not intended to limit the invention; for the purpose of better illustrating embodiments of the invention, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the size of the actual product; it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numbers in the drawings of embodiments of the invention correspond to the same or similar components; in the description of the present invention, it should be understood that, if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc., that indicate an azimuth or a positional relationship based on the azimuth or the positional relationship shown in the drawings, it is only for convenience of describing the present invention and simplifying the description, but not for indicating or suggesting that the referred device or element must have a specific azimuth, be constructed and operated in a specific azimuth, so that the terms describing the positional relationship in the drawings are merely for exemplary illustration and should not be construed as limiting the present invention, and that the specific meaning of the above terms may be understood by those of ordinary skill in the art according to the specific circumstances.
The forecasting area is Fuling area of Chongqing city, the geographic position is in the middle of Chongqing city, and the area of the placard reaches 2942.4km 2 East-west width 74.4km, north-south length 70.9km, span 29 DEG 20 '-30 DEG 00' north latitude, east longitude 106 DEG 57 '-107 DEG 44'. On administrative division, its south connects south Sichuan district and Wu Longou, east and adjacent Fengdu county, west connects Banan district.
(1) Regional landslide vulnerability prediction
a. Initial prediction
First, the study area was divided into 30m×30m evaluation units, and 326.9 ten thousand evaluation units were obtained in total. The system collects data of the topography, geological conditions, meteorological hydrology, human engineering activities and the like of the analysis area, and obtains 495 landslide samples according to remote sensing interpretation and field verification, as shown in fig. 5. And (3) reorganizing the acquired site and indoor data by adopting GIS software, and establishing a regional geological environment condition and landslide disaster distribution database, wherein the data source information is shown in Table 2. And (3) optimizing the evaluation factors according to the principal component analysis, the correlation analysis and the multiple collinearity analysis method, selecting the evaluation factors with strong independence and non-collinearity, and finally selecting 15 evaluation factors. And grading and assigning the evaluation factors by adopting an information quantity model according to the relation between the landslide points and the evaluation factors, wherein the grading assignment result is shown in Table 3. The landslide susceptibility prediction is carried out for the first time in the research area by adopting an information quantity method, the susceptibility value is graded by adopting a natural breakpoint method, and the primary prediction result is shown in fig. 6.
TABLE 2 data Source information
TABLE 3 evaluation factor classification results
b. Negative sample selection
Based on the existing landslide point catalog and related vector data obtained by field investigation, 8675 landslide grids are obtained through a spatial connection tool of arcgis. Firstly, water systems and residential areas need to be avoided, secondly, on the basis of the landslide susceptibility value calculated by the IV model and classification, non-landslide grids are randomly selected in the extremely low susceptibility area or the low susceptibility area, and meanwhile, the non-landslide grids are ensured to be out of the 1000m range of the landslide point buffer area. 8675 non-landslide grids are finally generated, and a landslide-non-landslide data set is formed by the known 8675 landslide grids. A 70% landslide-non-landslide dataset was randomly selected as the training set, with the remaining 30% as the test set.
c. Model prediction
1) Single model susceptibility prediction
And (3) adopting a BPNN model and an RF model to predict landslide vulnerability, wherein a training set is required to be used for model training, a testing set is required to be used for model inspection, and finally all raster data of a research area is input into the trained model for prediction. The construction of the BPNN and RF models is carried out by adopting a software package in Matlab software, and the prediction result is shown in figure 7.
2) Coupling model vulnerability prediction
And taking the information magnitude of the landslide-non-landslide data set as an input layer of the RF model and the BPNN model and performing model training. And finally grading the landslide susceptibility value predicted by the model. By adopting the coupling mode, on one hand, the pertinence of data processing and the regularity of data distribution can be improved, and on the other hand, the accuracy of RF and BPNN model prediction can be improved. The coupling model prediction results are shown in fig. 8.
3) Comparative analysis
As shown in fig. 9, AUC values in the BPNN, RF, IV-BPNN coupling model and the IV-RF coupling model are respectively 0.87, 0.88, 0.92 and 0.96,4 models, and ROC curves of the models are all close to the upper left corner, which indicates that 4 models trained by using 14 evaluation factors have higher precision and can accurately predict the space probability of landslide occurrence; secondly, the prediction precision of a single machine learning model can be obviously improved by coupling the information quantity with the BPNN model and the RF model. Compared with a network model, the effect of the information quantity model coupling the tree model is more obvious than that of the coupling network model.
As can be seen from fig. 10, the prediction performance indexes of the RF and BPNN models are both about 85%, which indicates that each model has better prediction performance, but the prediction performance of the RF model is slightly higher than that of the BPNN model; the predictive performance indexes of the IV-RF model and the IV-BPNN model are about 95%, which shows that the information quantity model can greatly improve the predictive performance of the tree model and the network model, and the predictive performance indexes of the IV-RF model are higher than those of the IV-BPNN model.
As can be seen from fig. 11, the landslide susceptibility predicted by the RF, BPNN, IV-RF and IV-BPNN models had a ratio of the landslide area to the total landslide area of 86.72%, 87.08%, 92.75% and 89.14% in the extremely high susceptibility region and the high susceptibility region, respectively, and the area ratios of the extremely high susceptibility region and the high susceptibility region were: 15.89%, 31.52%, 14.17% and 21.08% (it is generally believed that the extremely high and high susceptibility areas to landslide should be as small as possible, and the ratio of the landslide area in the classification to the total landslide area should be as large as possible), indicating that the landslide susceptibility classification predicted by the IV-RF model is more consistent with the spatial distribution of landslide disasters in the hilly areas, and has higher prediction accuracy. In addition, the information quantity model is also shown to have obvious precision improvement and optimization on a single network model and a tree model.
d. Optimal model selection
And comprehensively selecting the susceptibility prediction result of the IV-RF model as a subsequent risk early warning basis according to the ROC, the confusion matrix and the reasonable comparison of the subareas.
(2) Continuous probability rainfall threshold model based on multi-model coupling
a. Rainfall data
According to the geographic position of the Fuling area and the distribution situation of peripheral rainfall stations, rainfall data of 11 rainfall stations are selected, and the method comprises the following steps: 11 sites of Kaolin, changshou, fengshun, yutaishan, harmonious, shuangkou, damu, he Ji, wuling mountain, longtan, as shown in Table 4, rainfall data is downloaded from the China weather center website (http:// data. Cma. Cn /). Taking the characteristic of the spatial uneven distribution of rainfall into consideration, interpolating the rainfall corresponding to all rainfall landslide in a research area in the rainfall duration days (D) by adopting a Splines (space spline interpolation method) in Arcgis software, superposing interpolation results and landslide points, and extracting rainfall of the rainfall landslide points by adopting MATLAB; and the rainfall permeability coefficient of the early-stage effective rainfall (EE) is finally valued to be 0.8 through correlation analysis.
Table 4 11 rainfall station information
| Stop sign | 57520 | 57522 | 57523 | A8108 | A7585 | A8718 | A8104 | A7520 | A8101 | A7474 | A7510 |
| Station name | Long life and long service life | A kind of tomb | Fengdu (Chinese character of Fengdu) | Big wood | Fengcai (Phoenix) | River map | Rain bench | Mountain of Wuling mountain | Peaceful and harmony | Double river mouth | Lung Tam |
| County of district | Long life and long service life | A kind of tomb | Fengdu (Chinese character of Fengdu) | A kind of tomb | Wu Long | Nanchuan (Nanchuan) | A kind of tomb | A kind of tomb | A kind of tomb | Banan (Bananensis) | A kind of tomb |
| Longitude (°) | 107.07 | 107.42 | 107.68 | 107.67 | 107.31 | 107.03 | 107.45 | 107.53 | 107.47 | 106.90 | 107.10 |
| Latitude (°) | 29.83 | 29.75 | 29.87 | 29.62 | 29.40 | 29.31 | 29.72 | 29.52 | 29.87 | 29.61 | 29.45 |
| Elevation (m) | 378 | 274 | 218 | 980 | 616 | 660 | 660 | 1210 | 180 | 220.7 | 705 |
Before different rainfall threshold models are built, landslide samples are checked, and the samples with incomplete landslide point time records or landslide caused by other external factors are removed, so 418 rainfall type landslide points in 1980-2015 are selected for continuous probability rainfall threshold model building, 23 rainfall type landslide points in 2016-2020 are used for model accuracy verification, and four typical rainfall type landslide events in 2020-2021 are selected for danger early warning verification.
b.P EE And P D Probability of rainfall
And counting the early-stage effective rainfall (EE) and the rainfall duration days (D) of 418 take-off and land-off rain type landslide in the 1980-2015 year, wherein the early-stage effective rainfall interval in the research area is 10-130 mm, and the rainfall duration days are within 8 days. Fig. 12 shows the distribution of D and EE values (gray points) in linear coordinates, and the edge distribution (histogram) of D and EE and its cumulative frequency duty cycle (red curve) that result in the onset and onset of the rainy landslide of the Fuling area 418. From the graph, the cumulative frequency of the number of days of rainfall (P D ) Conform to the distribution characteristics of the power function, and the cumulative frequency of the early effective rainfall is the duty ratio (P EE ) Conforming to the characteristic of the cumulative normal distribution function, P EE And P D Fitting can be performed using equations 3 and 4, respectively, with the fitting results shown in fig. 13. The fitted equation is shown in the formulas 7 and 8, and the cumulative frequency ratio P is calculated according to the fitting result EE And P D R of fitting 2 Respectively reaching 0.95 and 0.97 indicates that the effective rainfall at the early stage is relative to P EE Pair of rainfall duration days P D The method has strong interpretation capability, and the mean square error is 0.08 and 0.04 respectively, which shows that the fitting result is very ideal.
P(D)＝1.31-1.08×D -0.6 (8)
c. Continuous probability rainfall threshold model based on multi-function fitting
Cumulative frequency of days of rainfall P D Cumulative frequency to early effective rainfall duty cycle P EE And multiplying to obtain time probability P, and carrying out two-dimensional nonlinear fitting by utilizing MATLAB as fitting data accords with power function characteristics and accumulated normal distribution function characteristics to obtain landslide continuous probability values of the Fuling area under the combination of different early-stage effective rainfall and different rainfall duration days, wherein the fitting result is shown in fig. 14 and 9. R of fitting 2 The arrival of 0.980 shows that the early effective rainfall and the rainfall duration days have strong interpretation capability for P, and the mean square error of 0.03 shows that the fitting result is very ideal.
d. Typical rainfall type landslide instance accuracy verification
23 rainfall landslide points in 2016-2020 are substituted into the rainfall model, so that the time probability corresponding to the landslide points is obtained, as shown in fig. 15. There are 5 landslide point time probabilities located in the range of 80% -100%, 7 landslide point time probabilities located in the range of 60% -80%, 6 landslide point time probabilities located in the range of 40% -60% and 5 landslide point time probabilities located in the range of 20% -40%.
(3) Regional landslide hazard early warning based on vulnerability and continuous probability rainfall threshold
Multiplying the landslide continuous probability value based on the multi-function fitting with the landslide susceptibility value to obtain the continuous probability landslide risk value of the Fuling area.
Four typical rainfall landslide events within 2020-2021 are selected for hazard early warning verification, and landslide hazard calculation results are shown in table 5. The method can predict the space-time distribution of the landslide with higher precision and can reflect the continuous probability risk value of the landslide.
Table 5 table of continuous probability landslide risk values for four typical rainfall landslide events
| Landslide designation | Time | EE | D | Value of vulnerability | Continuous probability | Continuous probability risk value | Risk rating |
| a. Cool water paving landslide | 2020/6/29 | 115.2 | 6 | 0.898 | 0.936 | 0.840 | Extremely high |
| b. Landslide on the list | 2018/8/01 | 96.5 | 7 | 0.875 | 0.930 | 0.814 | Extremely high |
| c. Tung garden landslide | 2020/6/27 | 104.1 | 6 | 0.966 | 0.919 | 0.888 | Extremely high |
| d. Slide for master temple | 2016/6/02 | 103.2 | 3 | 0.891 | 0.733 | 0.653 | High height |
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the present invention, which is intended to be covered by the claims of the present invention.
Claims (4)
1. A regional landslide hazard early warning method based on rainfall threshold is characterized in that: the method comprises the following steps:
s1: establishing an information quantity-random forest IV-RF and information quantity-BP neural network IV-BPNN landslide vulnerability prediction coupling model based on information quantity and negative sample selection;
s2: establishing a continuous probability rainfall threshold model based on multi-function fitting;
s3: regional landslide hazard early warning based on vulnerability and continuous probability rainfall threshold.
2. The regional landslide hazard pre-warning method based on rainfall threshold according to claim 1, wherein the regional landslide hazard pre-warning method is characterized by comprising the following steps: the S1 specifically comprises the following steps: the establishment of the information quantity-random forest IV-RF and information quantity-BP neural network IV-BPNN landslide vulnerability prediction coupling model based on the information quantity and negative sample selection specifically comprises the following steps: a non-landslide point 0 is randomly selected based on a landslide extremely low and low easy occurrence area preliminarily defined by an IV model, and a 1-0 data set is formed by the non-landslide point 0 and the landslide point 1; secondly, introducing information magnitude values in the data set as input layers of RF and BPNN, and establishing an IV-RF and IV-BPNN coupling model; the prediction precision of a single model and a coupling model is compared through a receiver operation characteristic curve ROC and an confusion matrix, and the reasonability of the easily-occurring partition of different models is compared based on an FR model:
s11: initial prediction
Dividing and numbering analysis areas by adopting an evaluation unit; the method comprises the steps of collecting data and information of landslide liability-related areas in terms of topography, geological conditions, weather hydrology and human engineering activities by means of data collection, remote sensing interpretation and field investigation, optimizing according to expert experience or a statistical analysis method, selecting evaluation factors with strong independence and non-collinearity, assigning, taking various evaluation factors in an evaluation unit as landslide independent variables, taking the landslide liability of the evaluation unit as dependent variables, and assigning the evaluation factors by adopting an information quantity method, wherein the formula 1 is shown as follows; the historical landslide information includes: a landslide sample database is established according to the spatial position, area, volume and contour range of the landslide;
wherein N is n Representing the area of landslide occurring in the n-th classification of the evaluation factor; n (N) 0 Indicating the total area where landslide occurs; s is S n Represents the nth classification area; s is S 0 Representing the total area of the investigation region; n (N) n /N 0 Representing the area ratio of landslide occurrence in the classification; s is S n /S 0 Representing the fractional area ratio; IV represents information quantity and reflects landslide contribution degree;
accumulating information values given by all evaluation factors of the evaluation units in the analysis area to obtain an area landslide susceptibility value, and grading by adopting a natural breakpoint method to obtain a primary area landslide susceptibility prediction result;
s12: negative sample selection
On the processed basic evaluation factor layer, a proper data set is required to be selected as the input of a landslide susceptibility prediction model; firstly, acquiring landslide grids, namely positive samples, with the mark of 1, through a space connection tool of arcgis based on related vector data acquired by existing landslide point cataloging and field investigation; selecting a non-landslide grid, namely a negative sample; in order to avoid that the selected non-landslide grids are real non-landslide grids, firstly avoiding water systems and residential areas, secondly carrying out random selection on the non-landslide grids in a very low-probability area or a low-probability area on the basis of classification of landslide probability values calculated by an information quantity model, and meanwhile ensuring that the non-landslide grids are out of a landslide point buffer area range, wherein the label is 0; the grid size is determined by the scale and the layer precision; selecting 500 non-landslide grids according to the negative sample selection mode to form a '0-1' data set containing 1000 data; randomly selecting 70% of samples of the data set as a machine learning training sample set and a test sample set of 30% of samples;
s13: model coupling
Coupling the two models, taking the information magnitude of the '0-1' data set as an input layer of a random forest model and a BP neural network model, and carrying out model training; grading landslide susceptibility values predicted by the model;
s14: comparative analysis
After training a random forest model and a BP neural network model by using sample data is completed, quantifying whether the accuracy after model training can meet the prediction requirement; the area under the ROC curve and the AUC curve is adopted to evaluate the training effect of the supervised machine learning model;
s15: optimal model selection
In order to discuss the prediction effect of the model, the ratio of the area occupation ratio of the landslide points to the frequency is adopted to measure the prediction effect; for the ratio of the area of the landslide point to the frequency, the lower the landslide susceptibility level is, the smaller the ratio of the area of the landslide point to the frequency is, and the ratio of the area of the landslide point to the frequency is gradually increased along with the increase of the susceptibility level.
3. The regional landslide hazard warning method based on rainfall threshold according to claim 2, wherein the regional landslide hazard warning method is characterized in that: the step S2 is specifically as follows: landslide number accumulation duty ratio P based on different early-stage effective rainfall EE and different rainfall duration days D EE And P D Will P EE And P D Multiplying to obtain time probability P, performing P-EE-D fitting in a form of accumulating the product of a normal distribution function and a power function, and calculating the time probability P of landslide occurrence of a certain point when the early-stage effective rainfall and the rainfall duration days of the certain point are obtained; randomly selecting a certain number of landslide points which do not participate in modeling, and substituting the landslide points into the established model to verify the model accuracy;
s21: rainfall data
1) Effective rainfall in early stage
The expression of the early effective rainfall EE is as follows:
wherein R is 0 For the daily rainfall of landslide, k is the rainfall permeability coefficient, R i Indicating the rainfall on the i-th day before landslide, n being the number of days in which rainfall is considered; for the rainfall permeability coefficient k, the value range is 0.72-0.86;
2) Days of continuous rainfall
The duration of rainfall is specified as: the duration days from the beginning to the end of a rainfall event; if the rainfall is less than 5mm for two consecutive days, ending the one rainfall event;
S22：P EE and P D Probability of rainfall
Counting the accumulated duty ratio P of the number of landslide under different early-stage effective rainfall EE EE Fitting by adopting an accumulated normal distribution function to obtain P EE -EE curve, as shown in the following formula;
counting the accumulated proportion P of the number of landslide under different rainfall duration days D D Fitting by using a power function to obtain P D -D curve:
P(D)＝a+b×D c (4)
s23: continuous probability rainfall threshold model based on multi-function fitting
Will P EE And P D Multiplying to obtain time probability P, and performing P-EE-D fitting in the form of the product of a cumulative normal distribution function and a power function, wherein the formula is as follows:
s24: typical rainfall type landslide instance accuracy verification
In order to verify the precision of a continuous probability rainfall threshold model based on multi-function fitting, landslide points of which parts do not participate in modeling are randomly selected and substituted into the established rainfall model, and the higher the precision is, the more accurate the model is proved.
4. The regional landslide hazard warning method based on rainfall threshold according to claim 3, wherein the regional landslide hazard warning method is characterized in that: the step S3 is specifically as follows: the continuous probability landslide hazard early warning is to grade the continuous probability landslide hazard value by adopting a natural breakpoint method or an equidistant dividing method; the continuous probability landslide hazard value is obtained by multiplying a landslide susceptibility value by a rainfall continuous probability value fitted by a nonlinear equation, and different combinations of the landslide susceptibility value and the rainfall continuous probability value can be reflected, as shown in the formula (6); 5-level division is carried out on the continuous probability landslide hazard value by adopting an equidistant division method, wherein the continuous probability landslide hazard value comprises an extremely high hazard zone, H is 1.0-0.8, a high hazard zone, H is 0.8-0.6, a medium hazard zone, H is 0.6-0.4, a low hazard zone, H is 0.4-0.2 and an extremely low hazard zone, and H is 0.2-0.0;
H＝P×S(6)
in the formula, H is a continuous probability landslide hazard value, P is a continuous time probability value, and S is a landslide susceptibility value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310569441.2A CN116543528A (en) | 2023-05-19 | 2023-05-19 | Regional landslide hazard early warning method based on rainfall threshold |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310569441.2A CN116543528A (en) | 2023-05-19 | 2023-05-19 | Regional landslide hazard early warning method based on rainfall threshold |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116543528A true CN116543528A (en) | 2023-08-04 |
Family
ID=87457561
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310569441.2A Pending CN116543528A (en) | 2023-05-19 | 2023-05-19 | Regional landslide hazard early warning method based on rainfall threshold |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116543528A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117935491A (en) * | 2023-12-25 | 2024-04-26 | 甘肃农业大学 | Landslide rainfall threshold prediction method based on space-time probability |
-
2023
- 2023-05-19 CN CN202310569441.2A patent/CN116543528A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117935491A (en) * | 2023-12-25 | 2024-04-26 | 甘肃农业大学 | Landslide rainfall threshold prediction method based on space-time probability |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220043182A1 (en) | Spatial autocorrelation machine learning-based downscaling method and system of satellite precipitation data | |
| CN106651211B (en) | Flood disaster risk assessment method for different scale areas | |
| CN110346517B (en) | Smart city industrial atmosphere pollution visual early warning method and system | |
| CN110766257B (en) | Method for evaluating short-term exposure concentration of air pollutants of crowd | |
| CN111401602B (en) | Assimilation method for satellite and ground rainfall measurement values based on neural network | |
| CN110852577A (en) | Urban flood assessment method based on urban toughness and urban drainage basin hydrological model | |
| CN111666656A (en) | Rainfall estimation method and rainfall monitoring system based on microwave rainfall attenuation | |
| CN103090831B (en) | The determination methods of icing region electric power line ice-covering thickness | |
| CN104484993A (en) | Processing method of cell phone signaling information for dividing traffic zones | |
| CN111144656A (en) | Disaster evaluation analysis method based on GIS | |
| CN112785024A (en) | Runoff calculation and prediction method based on watershed hydrological model | |
| CN114254802B (en) | Prediction method for vegetation coverage space-time change under climate change drive | |
| CN114578457B (en) | Air pollutant concentration space-time prediction method based on evolutionary ensemble learning | |
| CN112258029B (en) | Demand prediction method for sharing bicycles around subway station | |
| CN115994685A (en) | Method for evaluating current situation of homeland space planning | |
| CN116543528A (en) | Regional landslide hazard early warning method based on rainfall threshold | |
| CN116822185B (en) | Daily precipitation data space simulation method and system based on HASM | |
| CN114878748B (en) | CO (carbon monoxide) 2 Emission monitoring method and emission monitoring system | |
| CN118070619B (en) | Urban waterlogging model sensitive parameter identification optimization method by utilizing machine learning | |
| CN114662774A (en) | City block vitality prediction method, storage medium and terminal | |
| Tsakiris et al. | Regional drought identification and assessment. Case study in Crete | |
| CN115033809A (en) | Urban particulate matter space-time management and control method | |
| CN116502748A (en) | Landslide space risk early warning decision mechanism research method integrating multiple influencing factors | |
| CN112200399B (en) | Earthquake disaster risk assessment and economic loss prediction method | |
| CN115860487B (en) | Method for evaluating local vegetation change risk based on vegetation stability risk index |
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 |