CN-121980895-A - Flood vulnerability identification method and system

CN121980895ACN 121980895 ACN121980895 ACN 121980895ACN-121980895-A

Abstract

The invention relates to the field of hydrology and discloses a method and a system for identifying flood liability, wherein the method comprises the steps of obtaining flood point data, initially constructing a meteorological factor, a topography factor and an environmental factor data set which influence flood, screening factors which influence flood liability through Spearman rank correlation coefficient, tolerance and variance expansion factor, evaluating flood liability of a research area through an information quantity model to obtain a disaster liability partition map, selecting non-flood points in a low-incidence area and a lower-incidence area to construct a non-flood point data set, constructing a coupling information quantity model and a machine-learned flood liability model by utilizing the flood points and the non-flood point data, constructing a sample set of the non-flood data and the flood data, identifying an optimal flood liability model, and constructing a flood liability model from two angles of uncertainty of model simulation and uncertainty of data according to the invention and research on flood protection of different decision areas.

Inventors

YAO RUI
SUN PENG
GU XIHUI
LIU RONGHUA
ZHANG XIAOLEI
YANG HUILIN

Assignees

安徽师范大学

Dates

Publication Date: 20260505
Application Date: 20251126

Claims (9)

1. A flood vulnerability recognition method is characterized in that the method comprises the following steps, Step S1, collecting flood point data, namely acquiring a historical flood area of a research area and constructing flood point data of the research area; step S2, constructing an influence factor data set, and preliminarily constructing meteorological factors, topography factors and environmental factor data sets for influencing flood; Step 3, identifying main factors influencing the flood susceptibility of the research area, screening meteorological, topographic and environmental factor data sets by a Spearman rank correlation coefficient and tolerance and variance expansion factor method, and identifying the main factors influencing the flood susceptibility of the research area; step S4, constructing non-flooding point data according to the information quantity model, evaluating the flooding susceptibility of the research area through the information quantity model to obtain a disaster susceptibility partition map, selecting non-flooding points in the low-susceptibility area and the lower-susceptibility area, and constructing the non-flooding point data; Step5, constructing a flood susceptibility model, namely constructing a coupling information quantity model and a machine-learned flood susceptibility model by utilizing flood point and non-flood point data; Step S6, selecting non-flooding points, analyzing the influence of non-flooding point data on the accuracy of the flood-prone simulation, randomly selecting equal amount of non-flooding point data N times from a non-flooding area to reduce the influence of the non-flooding point data on the accuracy of the model, and constructing a sample set of the non-flooding data and the flooding data; And S7, analyzing the quality and the flood susceptibility grade of the flood susceptibility model, identifying the optimal flood susceptibility model by utilizing the ROC curve, accuracy, precision, recall and the F1 score, and grading the flood susceptibility of the research area.
2. The method according to claim 1, wherein in step S1, the flood inundation points include historical statistical inundation points and remote sensing extracted inundation points.
3. The method for identifying the susceptibility to flooding of claim 1, wherein in step S2, the meteorological factor, the topography factor and the environmental factor data set affecting the flooding are preliminarily constructed, including, The climate factors comprise maximum 1 day precipitation, maximum 3 days precipitation, maximum 5 days precipitation and average flood season precipitation; The topography factors comprise DEM data, gradient, slope direction, curvature, surface roughness, topography humidity, river power index and river transport index; Environmental factors include road index, distance from river, water network density, vegetation health index, land utilization, and soil type.
4. The method for identifying the susceptibility to flooding of claim 1, wherein in step S3, the data sets of meteorological factors, topography and environmental factors are screened by a Spearman rank correlation coefficient and tolerance and variance expansion factor method to identify main factors affecting the susceptibility of a flooding study area, specifically, Judging multiple collinearity among influencing factors by utilizing a Spearman rank correlation coefficient, tolerance and variance expansion factor; In case one, when Spearman is greater than a preset first threshold, indicating a high degree of collinearity; Secondly, when the tolerance value is smaller than a preset second threshold value, the existence of colinear between the independent variable and other independent variables is indicated; thirdly, when the variance expansion factor is smaller than a preset third threshold value, the variance expansion factor is larger than or equal to a preset fourth threshold value, and strong multiple collinearity exists; when any one of cases one to three is satisfied, there is a collinearity between the presentation factors, one of which is arbitrarily deleted, and the other is retained.
5. The method for identifying the susceptibility to flooding according to claim 1, wherein in step S4, the susceptibility to flooding of the research area is evaluated through the information quantity model to obtain a disaster susceptibility partition map, non-flooding points are selected in the low susceptibility area and the lower susceptibility area, non-flooding point data is constructed, specifically the following formula, Wherein: representing the total information amount weighted by various flood evaluation indexes, The number of flood points contained in a single evaluation index special grading interval is represented; representing the total number of flood points; representing the grid number in a specific grading area of a single flood evaluation index; representing the total grid number of the research area; After the R value is calculated, the Natural break method in the ArcGIS is utilized to divide the extremely high, medium, low and extremely low 5 grades, and non-flooding point data are randomly selected in the low and extremely low risk areas.
6. The method for identifying the susceptibility to flooding of claim 1, wherein in step S5, the flooding susceptibility model is constructed by using the data of the flooding points and the non-flooding points, wherein the model is coupled with the information quantity model and the machine learning flooding susceptibility model, specifically, Perfecting a non-flood point data set by utilizing the information quantity model; five machine learning models, namely random forest RF, XGBoos, catBoost, decision tree and logistic regression, are selected to analyze flood susceptibility and simultaneously The Stacking integration method is adopted to improve the simulation precision of the machine learning model.
7. The method for identifying the susceptibility to flooding of claim 1, wherein in step S6, the equal amount of non-flooding point data is randomly selected N times from the non-flooding area, and a sample set of the non-flooding data and the flooding data is constructed, specifically, And randomly selecting equal amount of non-flood point data N times from the non-flood area, wherein the longitude and latitude positions of the non-flood points at the last time are completely different from those of the non-flood points at the previous time.
8. The method of claim 1, wherein in step S7, the ROC curve, accuracy, precision, recall and F1 score are used to identify the optimal flood susceptibility model.
9. A flood susceptibility identification system is characterized in that the system comprises: The flood point data acquisition module is used for acquiring a historical flood area of the research area and constructing flood point data of the research area; the influence factor data set construction module is used for preliminarily constructing meteorological factors, topography factors and environment factor data sets influencing flood; the main factors influencing the flood initiation of the research area are identified by screening meteorological, topography and environmental factor data sets through a Spearman rank correlation coefficient, tolerance and variance expansion factor method; the non-flood point data construction module is used for evaluating flood liability of a research area through an information quantity model to obtain a disaster liability partition map, selecting non-flood points in a low liability area and a lower liability area and constructing non-flood point data; The flood susceptibility model construction module is used for constructing a coupling information quantity model and a machine learning flood susceptibility model by utilizing flood point and non-flood point data; The non-flooding data and sample set construction module of the non-flooding data comprises the steps of selecting non-flooding points, analyzing the influence of the non-flooding point data on the flooding susceptibility simulation precision, randomly selecting equal amount of non-flooding point data from a non-flooding area for N times to reduce the influence of the non-flooding point data on the model precision, and constructing a sample set of the non-flooding data and the flooding data; And the analysis module for the flood susceptibility model quality and the flood susceptibility grade is used for identifying the optimal flood susceptibility model by utilizing the ROC curve, accuracy, precision, recall and the F1 score and grading the flood susceptibility of the research area.

Description

Flood vulnerability identification method and system Technical Field The invention relates to the field of hydrology, in particular to a flood vulnerability identification method and system. Background Most floods are caused by strong precipitation, and global temperature increases lead to increased rainfall intensity and frequency, indirectly increasing the likelihood of flood events. The past research susceptibility evaluation method mainly comprises a historical disaster situation mathematical statistics method, an index system method, a scene simulation method and a method based on a geographic information system and remote sensing. The method has the characteristics of strong subjectivity, complex model construction, long time consumption and the like. The machine learning method can carry out numerical reproduction on the nonlinear process of the flood by analyzing the historical data, does not need to know the physical process of the bottom layer, has strong objectivity and high precision, and can quickly and widely evaluate the flood susceptibility. ML such as random forests, support vector machines, and logistic regression have been increasingly applied to flood susceptibility analysis. Previous studies have found that optimal machine learning models for simulating flood susceptibility in different areas are different, and that a single machine learning model cannot meet the flood susceptibility simulation in complex climates and under-the-floor conditions. The flood susceptibility evaluation method is converted from a single model to a mixed model so as to improve the accuracy of flood prediction. The hybrid model combines two or more models, and in particular, the advantages of different models can be interdependent to optimize the evaluation result and improve the prediction accuracy. Uncertainty is an unavoidable component of the decision process, and besides uncertainty in the model, the choice of sample data is also a major source of uncertainty, such as inconsistent choice of test/training set and non-flooded samples with no explicit acquisition method. At present, the influence of non-flood point data on a flood vulnerability model is still to be further clarified. In the identification and simulation of the flood susceptibility, the uncertainty of the flood susceptibility is effectively reduced by improving the identification of non-flood points, the simulation precision of the flood susceptibility model is greatly improved, and scientific basis is provided for disaster risk assessment, emergency management, urban planning and disaster prevention and reduction. Disclosure of Invention The invention aims to provide a method and a system for identifying flood liability, which solve the technical problem of low accuracy of identifying flood liability in the prior art. Specifically, the invention discloses a flood vulnerability identification method and a flood vulnerability identification system, wherein the method comprises the following steps: Step S1, collecting flood point data, namely acquiring a historical flood area of a research area and constructing flood point data of the research area; step S2, constructing an influence factor data set, and preliminarily constructing meteorological factors, topography factors and environmental factor data sets for influencing flood; Step 3, identifying main factors influencing the flood susceptibility of the research area, screening meteorological, topographic and environmental factor data sets by a Spearman rank correlation coefficient and tolerance and variance expansion factor method, and identifying the main factors influencing the flood susceptibility of the research area; step S4, constructing non-flooding point data according to the information quantity model, evaluating the flooding susceptibility of the research area through the information quantity model to obtain a disaster susceptibility partition map, selecting non-flooding points in the low-susceptibility area and the lower-susceptibility area, and constructing the non-flooding point data; Step5, constructing a flood susceptibility model, namely constructing a coupling information quantity model and a machine-learned flood susceptibility model by utilizing flood point and non-flood point data; Step S6, selecting non-flooding points, analyzing the influence of non-flooding point data on the accuracy of the flood-prone simulation, randomly selecting equal amount of non-flooding point data N times from a non-flooding area to reduce the influence of the non-flooding point data on the accuracy of the model, and constructing a sample set of the non-flooding data and the flooding data; And S7, analyzing the quality and the flood susceptibility grade of the flood susceptibility model, identifying the optimal flood susceptibility model by utilizing the ROC curve, accuracy, precision, recall and the F1 score, and grading the flood susceptibility of the research area. The system comprises: The f