Flood disaster loss real-time estimation method and system

Abstract

The invention discloses a flood disaster loss real-time estimation method and system. Relates to the technical field of flood disaster calculation methods. According to the method, remote sensing satellite data is combined, a loss evaluation system of flood disaster-inducing factors and interaction of the flood disaster-inducing factors is constructed, cross-platform efficient processing can be carried out, precision is improved through multi-algorithm fusion, multi-factor dynamic modeling is established, real-time monitoring is carried out, and an open framework is extensible and has high universality; the method comprises the following steps: carrying out statistical analysis before a disaster by utilizing historical disaster situation data, carrying out a damage situation estimation experiment on the basis of a built flood disaster hydrodynamic model under the condition that disaster-affected units are coupled with disaster-affected body types, constructing a loss state vulnerability curve and a loss rate curve through interaction of a plurality of disaster-inducing factors, and calculating the damage situation of the flood disaster according to the loss state vulnerability curve and the loss rate curve. Regional loss rate calculation and loss amount estimation are carried out, the loss amount is rapidly estimated in real time when an actual disaster occurs, a scientific basis is provided for emergency response and risk management, and then decision makers are helped to formulate scientific disaster response measures.

Description

Technical Field

[0001] The present invention relates to the technical field of flood disaster calculation methods, and in particular to a real-time flood disaster loss estimation method and system. Background Art

[0002] Floods are one of the most common natural disasters worldwide. Their suddenness and destructiveness bring huge losses to the social economy and people's lives. Traditional flood loss assessment methods mainly rely on post-disaster surveys and statistics, which have problems such as time lag and inaccurate data. In existing technologies, the construction of vulnerability curves is usually based on historical data and statistical models, which makes it difficult to reflect the actual situation when the disaster occurs in real time. Step S13

[0003] Existing vulnerability curves are primarily constructed by combining observational data, statistical models, and physical models to quantify the impact of floods on buildings, infrastructure, populations, or ecosystems. The calculation of vulnerability curves relies heavily on highly accurate loss surveys, but in practice, the uncertainty of loss rate data obtained through field surveys and interviews is inherent. Traditional vulnerability curves employ methods such as simple linear models, polynomial models, linear-logarithmic models, log-linear models, double-logarithmic models, logistic curve models, and cumulative probability curve models for fitting regressions. These methods often suffer from simple model forms, difficulties ensuring goodness of fit, and difficulty interpreting parameters. Regional limitations are a significant drawback in vulnerability curve calculations. This is because directly calculating vulnerability curves using models directly utilizes statistical results, which can downplay the impact of regional characteristics, resulting in results calculated in Region A being inapplicable in Region B. In short, existing methods suffer from regional limitations in data processing and model construction, making it difficult to effectively integrate multi-source data for real-time assessments. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a real-time estimation method and system for flood disaster losses. The technical problems solved are: the existing flood disaster vulnerability curve research over-relies on high-precision disaster loss survey data, but the loss rate obtained on the spot has subjective estimation bias and data uncertainty; the traditional model is insufficient in fitting goodness due to its simplified form, and the physical meaning of the parameters is unclear; the model directly relies on regional statistical characteristics, resulting in insufficient universality and difficulty in cross-regional migration and application. The present invention provides a real-time estimation method and system for flood disaster losses, which collects data through multiple data sources before the disaster, analyzes the target area for risk assessment, establishes multiple disaster factor relationships to find hidden indexes, finds the relationship between hidden indexes, disaster loss intensity and loss degree to establish a loss rate curve, establishes a hydrodynamic model to invert a large amount of experimental data, and uses the loss rate curve to estimate the disaster based on real-time remote sensing satellite data. The real-time estimation method and system for flood disaster losses with high universality solves the above problems.

[0005] The purpose of the present invention is mainly achieved through the following technical solutions: Step S1: Acquire data of the target area and extract water body information, disaster-causing factors, and damage factors. The target area includes vulnerable units; preprocess the data and perform spatial coupling to construct a coupled historical data set; Step S2: within the vulnerable unit, construct a hydrodynamic model based on the coupled historical data set, perform inversion simulation, and extract test water body information; determine the relationship between two or more disaster-causing factors based on the test water body information, and construct a relationship formula between the disaster-causing factors; Step S3: construct the relationship between the loss state and two or more disaster-causing factors as the loss state vulnerability curve; calculate the exceedance probability of each loss state, and combine the loss state vulnerability curve to obtain the relationship between the loss rate and the disaster intensity as the loss rate curve; Among them, disaster intensity is the value of the damage factor, and loss rate is the specific quantitative value of disaster intensity. The loss status of the disaster is divided according to the loss rate; Step S4: construct a hydrodynamic model based on flood disasters under different simulation conditions, and invert it to obtain a coupled experimental data set; use the loss rate curve based on the coupled experimental data set to estimate the loss rate under different simulation conditions, combine the disaster factors and select the disaster-bearing body type to construct a total loss formula; calculate the estimated loss amount of flood disasters under different simulation conditions according to the total loss formula as the quasi-real-time total loss amount of the disaster area.

[0006] In order to further optimize the above technical solution, the calculation formula of the hidden index I is expressed as: , Where X represents any hazard factor, d and v represent the hazard intensity of the maximum hazard factor A and the hazard intensity of the maximum hazard factor B, respectively. is a random error term that obeys the standard normal distribution; As a parameter, we can get the following through exponential transformation: .

[0007] In order to further optimize the above technical solution, the relationship between the loss state and two or more disaster factors is constructed as the loss state vulnerability curve, which specifically includes: Step S33: Assuming a normal distribution relationship exists between the hidden index and the disaster-causing factor, the relationship between the hidden index and the loss state is constructed based on the Lognormal distribution function as a loss state vulnerability curve; The relationship between the hidden index I and the disaster loss intensity is constructed using the lognormal distribution function, a multi-parameter regression analysis model of vulnerable units is constructed, and the hidden index is used to I The logarithmic relationship between the hazard factor and the vulnerability curve of different loss states is calculated, and the formula is expressed as: , in, In the disaster factor d The loss rate under the conditions of d is any hazard factor, is the fragility curve, is a hidden index, which is determined by the disaster factor d The disaster factor is calculated by d The logarithmic linear function of the loss state represents the probability of exceeding a certain loss state, that is, the probability that the loss state reaches or exceeds a certain threshold. Disaster factors d The continuous change of is mapped to the probability value in the interval [0, 1] to characterize different disaster factors d The probability of the next loss state occurring; and It represents the undetermined parameter used to calculate the probability of exceeding a certain loss state. The probability of exceeding a loss state is used to analyze the loss amount of each vulnerable unit during the construction of the vulnerability curve.

[0008] In order to further optimize the above technical solution, in step S3, the values ​​of the damage factors are collected based on historical disaster data as the disaster intensity, and the exceedance probability of each loss state is calculated. The loss state vulnerability curve is used to perform weighted calculations to obtain the relationship between the loss rate and the disaster intensity, which is used as the loss rate curve. Specifically, it includes: Step S34: Constructing a relationship between disaster intensity and loss status based on historical disaster data; formulating a classification standard for disaster loss status based on historical disaster data, and calculating the probability of occurrence of a specific loss status using the exceedance probability difference between adjacent loss statuses; To calculate the probability of occurrence of each loss state, it is necessary to calculate the value between two adjacent loss states, that is, to calculate the probability of occurrence of the loss state. The formula is expressed as: , Step S35: Weight each loss state to obtain a loss rate curve; calculate the detailed loss rate under different disaster factors and different disaster intensities according to the differentiation intervals of different loss states. , and introduce weight values Calculate the possible loss rate range, the formula is expressed as: , in Will be expressed in is the disaster loss rate under a certain value of a certain hazard factor; A, B, C, D, E, and F represent six loss states: no loss, slight loss, moderate loss, heavy loss, severe loss, and complete loss, respectively; P represents the corresponding probability of occurrence. The values ​​of are 0, 0.5 or 1, which represent the upper bound, lower bound and mean of the loss rate of the loss state respectively.

[0009] In order to further optimize the above technical solution, in step S3, the maximum likelihood method is used to solve the undetermined parameters, which specifically includes: Step S36: Use the maximum likelihood method to solve the undetermined parameters; suppose there are N samples of historical disaster data, and the hidden index of each sample is , the probability of belonging to a certain loss state is given by Indicates that, taking the state without loss as an example, the likelihood function L is expressed as follows: .

[0010] The present invention also includes a real-time flood disaster loss estimation system, which is applied to the method and includes a data acquisition unit, a loss estimation unit, a flood simulation unit, and a loss rate curve calculation unit connected in series between the data acquisition unit and the loss estimation unit; wherein: The data acquisition unit includes a data analysis module, a remote sensing satellite data processing module, a water body information processing module, and a historical data processing module, which are electrically connected in sequence. The data acquisition unit obtains population, economic, and land data from multiple data sources. The data analysis module uses the collected population data, economic data, and land data, including geographic information, meteorological data, remote sensing satellite data, damage factors, and disaster-causing factors. The remote sensing satellite data processing module extracts water body information from the collected data. The water body information processing module divides the target area into vulnerable units and non-vulnerable units based on the water body information. The historical data processing module preprocesses the collected population data, economic data, and land data, and performs spatial coupling to construct a coupled historical data set. The flood simulation unit includes a hydrodynamic model establishment module, an inversion module, an experimental data processing module, and a disaster factor relationship establishment module that are electrically connected in sequence; the hydrodynamic model establishment module establishes a regional comprehensive hydrodynamic model of flood disasters under precipitation conditions based on remote sensing satellite data and geographic information; the inversion module predicts flood conditions under different precipitation conditions by setting different precipitation conditions in the hydrodynamic model, and extracts test water body information based on the inversion under the hydrodynamic model simulation; the experimental data processing module couples the test water body information with the coupled historical data set to obtain a coupled experimental data set; the disaster factor relationship establishment module determines the relationship between disaster factors in the vulnerable unit based on the test water body information, constructs a relationship formula between disaster factors, and introduces a hidden index I; The loss rate curve calculation unit includes a vulnerability curve construction module and a loss rate curve construction module that are electrically connected in sequence; the vulnerability curve construction module uses regression analysis to construct a relationship between a loss state and two or more disaster-causing factors in a vulnerable unit as a loss state vulnerability curve; based on historical disaster data, each value of the damage factor is collected as the disaster intensity, and the probability of exceeding each loss state is calculated; the loss rate curve construction module uses the loss state vulnerability curve to perform weighted calculations to obtain a relationship between the loss rate and the disaster intensity as the loss rate curve; The loss estimation unit constructs flood disasters caused by different precipitation conditions based on the hydrodynamic model of the HEC-RAS simulation software, and based on this inversion, obtains a coupled experimental data set and uses the loss rate curve to estimate the loss rate of the hydrodynamic model under different simulation conditions. The total loss formula is constructed according to the coupling of disaster factors and disaster-bearing body types, and the estimated loss amount of flood disasters under different simulation conditions is calculated as the quasi-real-time total loss amount of the disaster area.

[0011] In summary, the present invention has the following beneficial effects compared with the prior art: 1. The present invention features efficient cross-platform processing, and the technical process can be performed simultaneously on multiple satellite data processing platforms. Based on the Google Earth Engine (GEE) platform, it supports rapid screening, calling, and automated processing of remote sensing data, significantly shortening data acquisition and analysis time. Compared with existing technologies, it has the advantages of multi-platform compatibility and efficient processing.

[0012] 2. The present invention integrates multiple algorithms to improve accuracy, perform high-precision water body extraction and water depth inversion, integrate the maximum threshold method Otsu for threshold segmentation and the FwDET tool terrain correction algorithm, optimize the accuracy of water body boundary recognition, and dynamically combine terrain data to invert water depth distribution, solving the problem that traditional reflection signal methods are greatly affected by environmental interference and have low accuracy.

[0013] 3. The present invention is a multi-factor dynamic loss assessment model that breaks through the limitations of traditional single-factor statistical models. It supports users to customize multiple influencing factors (such as water depth, flow velocity, vulnerability of disaster-bearing body type, and other disaster-causing factors) and their nonlinear relationships (such as logarithmic correlation), thereby realizing a refined assessment of flood disaster loss rates.

[0014] 4. The present invention can monitor flooding and predict disasters in real time. It can dynamically track the evolution of flooded areas based on remote sensing data and quickly estimate the losses of each disaster-prone body (such as GDP and population) in combination with the loss rate curve. This will reduce the workload of post-disaster statistics, accurately locate high-risk areas, and assist in the targeted deployment of emergency resources.

[0015] 5. In the model of the present invention, model parameters (such as loss status thresholds and weight factors) support manual correction and expert experience embedding, adapt to the disaster characteristics of different regions, provide an open architecture for subsequent algorithm optimization (such as machine learning integration), and are flexible and scalable. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The figures described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the figures: Figure 1 A step diagram of a real-time flood disaster loss estimation method; Figure 2 This is a flow chart of a real-time flood disaster loss estimation system; Figure 3 is the exceedance probability of the disaster under different disaster intensities at different loss levels under flood disasters Figure 4 is the probability of occurrence of each loss intensity under different disaster intensities under flood disasters; Figure 5 is the loss rate under flood disaster, i.e., the vulnerability curve; Figure 6 The estimated values ​​of flood depth, land classification, GDP distribution and total GDP loss for a city under the coupled model; Figure 7 Comparison of the extent of losses in nine counties of a city and the intensity of losses within the county under flood disasters; Figure 8 The estimated losses within one square kilometer of a city due to flood disasters; Figure 9 This is an hourly estimate of disaster losses in a certain county from 01:00 to 03:00 due to flood disasters. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the examples and figures. The illustrative embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention. The embodiments of the present application will be described below in conjunction with the figures.

[0018] This embodiment provides a method and system for real-time estimation of flood disaster losses, with the aim of specifically addressing the problems existing in existing flood vulnerability curve research, namely, over-reliance on high-precision disaster loss survey data, but the loss rates obtained on the spot are subject to subjective estimation bias and data uncertainty; traditional models (such as linear / logistic regression) have insufficient goodness of fit due to their simplified form, and the physical meaning of the parameters is unclear; the model directly relies on regional statistical characteristics, resulting in insufficient universality and difficulty in cross-regional migration and application. The present invention provides a method and system for real-time estimation of flood disaster losses that can be processed efficiently across platforms, improves accuracy through multi-algorithm fusion, establishes multi-factor dynamic modeling, performs real-time monitoring and reduces the burden, and has an open architecture that can be expanded with high universality, thereby solving the above problems.

[0019] In Example 1, referring to Figure 1-Figure 5 , which is the first embodiment of the present invention. This embodiment provides a method for real-time estimation of flood disaster losses, such as Figure 1 As shown, it mainly includes the following steps: Step S1: Obtain population data, economic data, and land data of the target area, including geographic information, meteorological data, remote sensing satellite data, damage factors, and disaster-causing factors; extract water body information through remote sensing satellite data, and divide the target area into vulnerable units and non-vulnerable units; preprocess the collected population data, economic data, and land data, and perform spatial coupling to construct a coupled historical data set.

[0020] Step S2: In the vulnerable unit, a comprehensive hydrodynamic model of regional flood disasters under precipitation conditions is constructed based on remote sensing satellite data and geographic information; by setting different precipitation conditions in the hydrodynamic model, the flood conditions under different precipitation conditions are predicted, and the test water body information including water depth and flow velocity is extracted by inversion under the hydrodynamic model simulation; the test water body information is coupled with the coupled historical data set to obtain a coupled experimental data set; the relationship between the disaster-causing factors in the vulnerable unit is determined based on the test water body information, the relationship between the disaster-causing factors is constructed, and the hidden index I is introduced according to the disaster-causing factor d and the disaster-causing factor v.

[0021] Step S3: Use regression analysis to construct the relationship between the loss state and two or more disaster-causing factors in the vulnerable unit as the loss state vulnerability curve; based on historical disaster data, collect the values ​​of the damage factors as the disaster intensity, calculate the exceedance probability of each loss state, and use the loss state vulnerability curve to weight the relationship between the loss rate and the disaster intensity as the loss rate curve.

[0022] Step S4: Based on the HEC-RAS simulation software hydrodynamic model, construct flood disasters caused by different precipitation conditions. Based on this inversion, by obtaining a coupled experimental data set, use the loss rate curve to estimate the loss rate of the hydrodynamic model under different simulation conditions, and construct a total loss formula based on the coupling of disaster factors and disaster-bearing body types. The estimated loss amount of flood disasters under different simulation conditions is calculated as the quasi-real-time total loss amount of the disaster area.

[0023] like Figure 2 As shown, this embodiment provides a real-time estimation system for flood disaster losses, which is applied to the above method and includes a data acquisition unit, a loss estimation unit, a flood simulation unit and a loss rate curve calculation unit respectively connected in series between the data acquisition unit and the loss estimation unit.

[0024] The data acquisition unit includes a data analysis module, a remote sensing satellite data processing module, a water body information processing module and a historical data processing module which are electrically connected in sequence; wherein the data acquisition unit obtains population, economic and land data from multiple data sources; the data analysis module uses the population data, economic data and land data collected by the data, including geographic information, meteorological data, remote sensing satellite data, damage factors and disaster-causing factors; the remote sensing satellite data processing module extracts water body information by collecting data; the water body information processing module divides the target area into vulnerable units and non-vulnerable units based on the water body information; the historical data processing module pre-processes the collected population data, economic data and land data, and performs spatial coupling to construct a coupled historical data set.

[0025] The flood simulation unit includes a hydrodynamic model establishment module, an inversion module, an experimental data processing module, and a disaster factor relationship establishment module that are electrically connected in sequence; the hydrodynamic model establishment module constructs a regional comprehensive hydrodynamic model of flood disasters under precipitation conditions based on remote sensing satellite data and geographic information; the inversion module sets different precipitation conditions in the hydrodynamic model, uses hydrological simulation software such as HEC-RAS to simulate flood inundation, generates data such as inundation depth and flow rate, and supports adjustment of simulation parameters to cope with different geographical environments and precipitation conditions, to predict flood conditions under different precipitation conditions, and extracts test water body information including water depth and flow rate based on inversion under hydrodynamic model simulation; the experimental data processing module couples the test water body information with the coupled historical data set to obtain a coupled experimental data set; the disaster factor relationship establishment module determines the relationship between disaster factors in the vulnerable unit based on the test water body information, constructs a relationship formula between disaster factors, and introduces a hidden index I based on the disaster factor d and the disaster factor v.

[0026] The loss rate curve calculation unit includes a vulnerability curve construction module and a loss rate curve construction module that are electrically connected in sequence; the vulnerability curve construction module uses regression analysis to construct the relationship between the loss state and two or more disaster-causing factors in the vulnerable unit as a loss state vulnerability curve; based on historical disaster data, the values ​​of the damage factors are collected as the disaster intensity, and the probability of exceeding the occurrence of each loss state is calculated; the loss rate curve construction module uses the loss state vulnerability curve to perform weighted calculation to obtain the relationship between the loss rate and the disaster intensity as the loss rate curve.

[0027] The loss estimation unit constructs flood disasters caused by different precipitation conditions based on the hydrodynamic model of the HEC-RAS simulation software, and based on this inversion, obtains a coupled experimental data set, uses the loss rate curve to estimate the loss rate of the hydrodynamic model under different simulation conditions, constructs a total loss formula based on the coupling of disaster factors and disaster-bearing body types, calculates the estimated loss amount of flood disasters under different simulation conditions, and uses the preset loss rate curve as the quasi-real-time total loss amount of the disaster area based on the flooded water depth, land classification and GDP distribution data. The disaster loss is estimated, and the total loss amount of the selected disaster-bearing body is evaluated through the loss rate calculation method.

[0028] like Figure 2 As shown, the workflow of the method applied by the system of this embodiment is as follows: Step S1: Conduct risk assessment based on historical data. Collect and preprocess data; obtain remote sensing satellite data, geographic information, and meteorological data. In step S1, obtain population, economic, and land data for the target area, as well as meteorological data, remote sensing satellite data, damage factors, and hazard factors. Extract water body information from remote sensing satellite data, and divide the target area into vulnerable and non-vulnerable units. Preprocess the collected population, economic, and land data, and perform spatial coupling to construct a coupled historical dataset.

[0029] Step S11: collect data to obtain population data, economic data and land data of the target area. The land data includes historical disaster information, geographic information, meteorological data and remote sensing satellite data.

[0030] Population data includes census data and statistical yearbooks; economic data includes regional GDP, industrial structure, and the economic contribution of various industries. GDP refers to the value of all final goods and services produced by a country or region's economy over a quarter or year. This value is often used to directly reflect economic losses before and after a disaster. In this invention, this value is expressed in 10,000 yuan.

[0031] Land data includes historical disaster information, geographic information, meteorological data, and remote sensing satellite data.

[0032] Collect historical disaster data and combine it with local meteorological, geographical and other natural environmental factors to provide support for subsequent analysis.

[0033] Acquire appropriate remote sensing satellite data, mainly including satellite image data using Sentinel-1, Sentinel-2, and Landsat data call methods, and ensure that the time span is as short as possible to reflect the situation before and after the disaster.

[0034] Among them, historical disaster data are obtained and summarized based on sources such as those provided by local governments. The historical disaster data include casualty data, building loss data, and economic production reduction data, and the specific damaged measurement units of the research object, i.e. the target area, are established.

[0035] In this example, damage statistics from a single flood disaster event are often insufficient as sufficient experimental data. This is because investigators often focus on counting vulnerable units within a certain number of locations after field visits. Therefore, statistical data covering a sufficient number of times, days, and ranges is required for cross-reference and reuse. Furthermore, individual loss feedback often contains ambiguity and exceeds previous loss estimates, so such data should be excluded. In actual use, data other than the illustrated figures may be included or added.

[0036] The geographic information includes the geographic coordinates, geographic parameters, soil classification, and land classification of the target area; the geographic parameters include slope, slope direction, and distance from the river; the land classification includes building land, woodland, farmland, water area, and unused land, which is used to identify the type of disaster-prone objects and their vulnerability; the disaster-prone object types include casualty data, direct or indirect economic losses, direct or indirect losses of housing and construction, direct or indirect losses of business suspension, and storage losses; the meteorological data includes precipitation data and evaporation data, wherein precipitation data is processed as a time series, and evaporation data is used for long-term simulation.

[0037] The land in the target area is divided into units according to the measurement unit. The units in the target area include vulnerable units and non-vulnerable units. The measurement units include regular grids, administrative divisions and land classifications. The regular grids are divided according to the set size, and the administrative divisions are divided according to the boundaries of counties, towns and villages. For example, a city is divided into 30m grids and 5 types of land classification are extracted, including towns, farmland, forest land, water area and unused land. For example, a city is divided into 30m grids and 5 types of land classification (including towns, farmland, forest land, water area and unused land) are extracted, and 1,000 vulnerable units are marked.

[0038] Historical disaster data also includes damage factors and disaster-causing factors. Data on casualties, structural losses, and economic losses in the target area serve as damage factors, while water depth, flow velocity, duration of inundation, and discharge serve as disaster-causing factors. Detailed geographic and meteorological data within the target area are analyzed, and these factors are combined based on their respective characteristics. The use of precipitation data and land classification is also discussed. Precipitation data from different sources can suffer from errors, mismatches in accuracy, and missing data points. Therefore, high-precision and high-accuracy precipitation data should be selected, and the integration and reuse of multiple precipitation data sets should be considered. The same applies to land classification. Land classification should not be overly detailed. Land classification often has strong timeliness, so appropriate classification levels and timeframes should be selected to achieve a relatively balanced effect over a longer timeframe. In practical applications, consideration may be given to including or excluding data other than the diagrams.

[0039] The remote sensing satellite data includes disaster images, water body information and DEM data of digital elevation models. The remote sensing satellite data is obtained by integrating multiple remote sensing satellite data sources and using the GEE platform to call data with a time span set within a threshold time range before and after the disaster in real time and extract water body information; wherein, the remote sensing satellite data source includes satellite image data of Sentinel-1, Sentinel-2, and Landsat satellites; SAR satellites or light-sensing satellites are used to extract water bodies through the real scene of the disaster to obtain water body information including flooded areas; and DEM data is obtained by a set of ordered numerical arrays that realize digital simulation of the ground terrain through terrain elevation data. The technical process in this embodiment can be carried out simultaneously on multiple satellite data processing platforms, and the GEE platform, namely Google Earth Engine, is used here. On the GEE platform, real-time calling of remote sensing satellite data, rapid screening, calling, response and arrangement processing of target remote sensing satellite data, etc. can be achieved. At the same time, this technology can call data other than Sentinel-1 satellite data. In this embodiment, the calling code of Sentinel-1 is ee.ImageCollection("COPERNICUS / S1_GRD"); the calling code of Sentinel-2 is ee.ImageCollection("COPERNICUS / S2,_SR_HARMONIZED"); and the calling code of Landsat satellite is ee.ImageCollection("LANDSAT / LC08 / C02 / T1_TOA").

[0040] Remote sensing data is used to obtain the actual scene of flood disasters within the corresponding time period. SAR satellites can be used, or light-sensing satellites can be used to extract water body information. When selecting remote sensing satellite data, it is necessary to follow the principles of time consistency, same events, location fit, and fewer clouds. For some data obscured by clouds, cloud recognition and cloud removal can be used. Cloud detection methods include: threshold method, spatial feature method, deep learning method, etc.; cloud removal methods include: imaging model method, image restoration method, deep learning method, and sensor-assisted method. DEM data is a digital elevation model. DEM data is a digital simulation of the ground terrain through limited terrain elevation data, that is, a digital expression of the terrain surface morphology. It is a physical ground model that represents the ground elevation in the form of a set of ordered numerical arrays. It often carries various digital information and can be used for data fusion and connectivity from multiple data sources.

[0041] Step S12: Process the remote sensing satellite data to extract water body information. Based on the remote sensing satellite data, the time span is locked to a set period before and after the disaster. Water body information is extracted and combined with the DEM data to perform terrain analysis, laying the foundation for the construction of the hydrodynamic model.

[0042] Use Sentinel-1 GRD data to calculate the SDWI water extraction index or the maximum threshold method Otsu to process remote sensing satellite data to improve the extraction of water body distribution information. Use the FwDET tool in combination with DEM data to perform water depth inversion and extract water body information; use the FwDET tool to adjust the slope filter and number of iterations, and select the appropriate slope filter strength and number of iterations in the FwDET tool to improve the accuracy of water depth inversion.

[0043] Extract water information based on Sentinel-1 GRD data (SAR synthetic aperture radar), that is, calculate the SDWI water extraction index based on Sentinel-1 dual-polarization data: , where SDWI represents the water extraction index, VV and VH represent the dual polarization modes in the IW polarization mode . SDWI stands for Sentinel-1 Dual-PolarzedWaterIndex, and its Chinese translation is Sentinel-1 Dual-PolarzedWaterIndex. This method conducts in-depth research on the relationship between water body information extraction between Sentinel-1 dual-polarization data (VV and VH), so as to enhance the characteristics of water bodies, make water bodies more clearly distinguished, and eliminate the interference caused by soil and vegetation in water body extraction. The guiding principle of the formula is to multiply the VV and VH polarized images and multiply them by 10 to magnify the difference between water bodies and other land objects, and then use the natural logarithm as the function. VV and VH represent the polarization mode of the signal on the satellite when it is transmitted and received. For example, VH indicates that the signal is vertically polarized when it is transmitted and horizontally polarized when it is received.

[0044] Binarization processing is performed based on the maximum threshold method Otsu to extract water body information: , where S represents the sum of the minimized internal variances, A and B represent the partition results of the binary classification of remote sensing data based on the threshold T, so Represents the pixel points in area A, the corresponding Corresponding to area B; when the VV value of a pixel point is greater than the threshold T, it will be divided into area A, otherwise it will be divided into area B; and Representing the total number of pixels in their respective regions. The Otsu method, proposed by Ōtsu Nobuyuki in 1979, is a maximum inter-class variance method. After image binarization and segmentation using the threshold value obtained by this method, the inter-class variance between the foreground and background images is maximized. Water body extraction is performed by combining remote sensing imagery with the maximum threshold Otsu method (maximum inter-class variance method). By setting the optimal threshold for image binarization and segmentation, the water area in flooded areas can be accurately extracted. This method can automatically adjust the threshold for distinguishing water from background, thereby providing accurate water distribution information for real-time flood disaster assessments.

[0045] The FwDET tool is a flooded area depth estimator. This tool quickly calculates water depth data for flooded areas. The water body information extracted by the FwDET tool focuses on spatial distribution and physical parameters, covering key indicators such as flooded area, water depth, volume, and dynamic changes. It is suitable for disaster assessment, water resource management, and ecological research. Combining the multi-source remote sensing data disaster imagery and DEM data in this example further improves inversion accuracy and its scope of application.

[0046] The slope filter and the number of iterations are the Slope filtering procedure and boundary cells moistening iteration functions introduced in the FwDET tool. Combining these two tools for adjustment can effectively reduce irregular water depth results and sharply changing water body images during water depth inversion.

[0047] Remote sensing satellite technology is used to observe flood disaster water bodies in real time, and the FwDET tool is used to extract the water body of a single flood disaster to obtain historical data on disaster-causing factors, thereby realizing the measurement of physical parameters other than water depth.

[0048] Step S13: The FwDET tool performs elevation analysis in combination with DEM data to extract water body information including flooded water body inundation areas and boundaries, formulas, inundated volumes, and dynamic changes. A distribution map of water depth, flow velocity, and inundation time is drawn based on the extracted water body information.

[0049] Flood water body inundation areas and boundaries: Generate raster or vector boundary data of water body coverage; Formula: Calculate the water depth and flow velocity of each inundated area through spatial analysis of DEM and inundated area data to generate a water depth raster map; Inundated volume and dynamic changes: Based on the water depth and inundated area, the FwDET tool can further estimate the total inundated volume and support multi-temporal analysis to show the dynamic changes of the inundated volume over time; Spatiotemporal dynamic characteristics: The FwDET tool can output inundation time distribution maps, such as hourly inundation depth changes and the spatiotemporal distribution of maximum inundation depth. Combined with the HEC-RAS simulation software, it can display a three-dimensional view of the inundated area and water depth at different time points.

[0050] HEC-RAS simulation software is a river hydraulic calculation program developed by the Engineering Hydrology Center in the United States. HEC-RAS currently supports one- and two-dimensional hydrodynamic models, one-dimensional moving-bed sediment transport models, and one-dimensional water quality models. It also features the ability to couple hydraulic structures (dams, dikes, weirs, culverts, bridges, etc.).

[0051] Step S14: Calculate the hydrological characteristics of the water collection capacity and drainage capacity of each unit in the target area based on geographic information, meteorological data, and water body information. Extract the flooded area through remote sensing images and combine them with hydrological feature identification. Weighted calculation can also be used to classify the units in the target area into vulnerable units and non-vulnerable units. The hydrological feature calculation specifically includes: , where the catchment area is the area of ​​the flooded area, the precipitation intensity is based on spatial interpolation of meteorological data, and the permeability coefficient is set to a specific value according to the land classification; ,in, It is a weight coefficient of slope and distance to the river set according to historical data, and the drainage efficiency coefficient is set to a specific value according to the land classification. When the water collection capacity is greater than the drainage capacity, it is judged as a non-vulnerable unit; when the water collection capacity is less than or equal to the drainage capacity, it is judged as a vulnerable unit.

[0052] Step S15: pre-process the collected population data, economic data and land data.

[0053] First, historical disaster matching is carried out, and vulnerable units in the target area are selected as research sites and target events as cases. Then, according to the historical disaster data, they are matched to each unit and compared with the cases in the historical disaster data, that is, compared with cases at the same location but different times, and based on this, high-credibility samples are screened, such as the data of multiple flood events in the historical disaster data of a certain county, and abnormal samples are eliminated to enhance data reliability; units in the target area are selected as research sites and target events as cases, and face-to-face questionnaires and other methods are used to visit damaged enterprises to collect information such as the enterprises' production capacity (operating) loss rate, flooding depth, and flooding duration. Through data sorting and analysis, invalid samples such as missing values ​​and outliers are eliminated; then the units with missing data are supplemented; the vulnerable units with missing data are used as missing points, and the missing points are supplemented through interpolation or the mean of adjacent units, and compared with cases at the same location but different times, and high-credibility samples are selected to supplement the units with missing data; finally, standardization is performed, and the accuracy of the data is unified through interpolation and resampling methods. In the embodiment, according to actual needs, for example, 1km GDP data can be resampled to 30m to ensure the spatial consistency of different data.

[0054] Step S16: Couple the preprocessed data with geographic information to obtain spatially coupled data, and perform spatial coupling to obtain a coupled historical dataset. The coupled historical dataset is constructed by obtaining population, economic, and land data. Each unit is numbered and georeferenced using geographic coordinates using GIS technology. The population, economic, and land data are matched with vulnerable units to ensure that each vulnerable unit corresponds to a set of survey data. Attribute association is performed, adding fields to each vulnerable unit and labeling the population, economic, and land data. Land classification, historical data on hazard factors, and damage factors are also labeled. These different types of data are spatially aligned and spatially coupled to form a unified coupled dataset, which serves as the coupled historical dataset.

[0055] This component utilizes remote sensing satellite data, geographic information, and meteorological data, along with demographic, economic, and land data, to obtain information on land classification, population distribution, and economic activity distribution across urban areas. This information is then spatially aligned to construct a coupled historical dataset. This data is processed and accessed through the GEE platform, helping to analyze land classification in different regions, including agriculture and urban areas, and their corresponding economic activities, including GDP distribution. This data is combined with disaster influencing factors such as precipitation conditions and flooding depth, and spatial data analysis and visualization using Geographic Information System (GIS) technology to construct a coupled dataset for each unit.

[0056] Step S2: In the vulnerable unit, a comprehensive hydrodynamic model of regional flood disasters under precipitation conditions is constructed based on remote sensing satellite data and geographic information; test water body information including water depth and flow velocity is extracted based on the hydrodynamic model inversion; the test water body information is coupled with the coupled historical data set to obtain a coupled experimental data set; the relationship between disaster-causing factors in the vulnerable unit is determined based on the test water body information, and a relationship between the disaster-causing factors is constructed. A hidden index I is introduced based on the disaster-causing factor d and the disaster-causing factor v to quantitatively express the relationship, thereby realizing a normalized causal expression between the degree of loss and a certain disaster factor.

[0057] This embodiment takes water depth and flow velocity as examples of disaster-causing factors, where the water depth is denoted as d and the flow velocity is denoted as v.

[0058] Based on water body information extracted from remote sensing images and terrain data from DEM data, a regional flood disaster simulation model was constructed using HEC-RAS simulation software. By setting precipitation conditions and geographic information variables in the simulation model, the water depth and flow rate of the water body were carefully divided, and hydrodynamic simulation was performed to predict flood conditions under different precipitation intensities.

[0059] Step S21: Based on meteorological data and remote sensing satellite data, including water body information extracted from remote sensing images, terrain data, land classification, soil type, and distance from rivers from DEM data, a regional flood disaster simulation model is constructed using HEC-RAS simulation software, which is recorded as a hydrodynamic model.

[0060] Among them, DEM data is used to define the terrain definition boundary conditions of the calculation domain, and land classification is used to assign the Manning coefficient; in a specific embodiment, when using the HEC-RAS simulation software, the soil type and the distance from the river are also used to calculate the infiltration, and the distance from the river is used to define the river channel.

[0061] Step S22: Each unit obtains the test water body information under simulation conditions through inversion of the hydrodynamic model, including water depth, flow velocity, submergence time, and flow rate.

[0062] In this embodiment, in order to simulate flood disasters based on precipitation conditions, it is necessary to classify precipitation conditions.

[0063] Controllable parameters include the delineation of upper and lower boundaries and the fineness of the simulated affected cell division. If precipitation data that varies with the location of the affected cell is unavailable, the area-averaged rainfall can be used as the rainfall input source. If a dataset with affected cells or grids is available, relevant precipitation data can be imported using the MeteorologicalData component. Spatially variable wind data in addition to precipitation can be added to the HEC-RAS simulation software using the Unsteady Flow Boundary Condition Editor.

[0064] Step S23: Compare the data with the water body information annotated in the coupled historical dataset. This comparison is with the measured inundation area. Because the coupled historical dataset is constructed from remote sensing imagery and data collected from field surveys, when using the HEC-RAS simulation software to simulate flooding under precipitation conditions, adjustable parameters are adjusted during operation and validation within the hydrodynamic model to improve accuracy. These adjustable parameters include the Manning coefficient, boundary conditions, and model input / output time intervals. These adjustable parameters affect the coefficient (slope) and intercept of the hazard factor under the logarithmic conditions of each term in the hidden index.

[0065] Step S24: Similar to the construction of the coupled historical data set, first use coordinates for positioning, then replace the water body information extracted from the historical disaster data and remote sensing satellite data with the experimental water body information obtained by hydrodynamic model simulation, obtain the corresponding experimental data of the disaster-causing factors based on the experimental water body information, mark the experimental data in the corresponding units, and construct the coupled experimental data set.

[0066] Step S25: The relationship between the extracted disaster-causing factors is judged and transformed through a large amount of historical water body information and the experimental water body information obtained by hydrodynamic model simulation; the disaster-causing factors are judged by whether there is a linear or nonlinear numerical relationship between each other. If there is a relationship, a numerical transformation is performed to convert the independent variable factors into a single variable as much as possible; if there is no relationship, new parameters are matched as coefficients to introduce the hidden index I.

[0067] Determining and transforming the relationships among the extracted hazard factors is the core step in this implementation. Identifying multiple hazard factors and analyzing their interrelationships is crucial for improving disaster assessment accuracy, revealing disaster mechanisms, and optimizing disaster prevention strategies. Natural disasters are often triggered by the combined effects of multiple hazard factors. In flood disasters, precipitation, topography, river siltation, and tides are all causative factors. By using simulation software to identify multiple factors and analyze their relationships to determine whether they are linear or nonlinear, the triggering and evolution mechanisms of disaster chains can be revealed. For example, heavy rainfall in mountainous areas is often exacerbated by terrain uplift. By combining factors such as slope and vegetation cover, the connections between precipitation, runoff, landslides, and debris flows can be quantified. Furthermore, multi-factor analysis can screen for core factors that play a dominant role in disasters.

[0068] This embodiment overcomes the limitations of single-factor assessment. Traditional single-factor assessments (such as predicting flood losses solely based on precipitation intensity) ignore the synergistic effects between factors, often leading to biased assessment results. This embodiment enhances the model's generalization and predictive robustness by simultaneously considering disaster-causing factors such as water depth and flow velocity in flood loss assessments, allowing for a more realistic loss rate curve. For example, at the same water depth, a 1m / s increase in flow velocity can increase the building damage loss rate by 15%-20%. Furthermore, based on multi-factor relationships, targeted monitoring and early warning equipment can be deployed. By simulating multi-factor combination scenarios, a probability distribution range for disaster losses can be generated, providing a quantitative basis for emergency resource stockpiling, enabling targeted intervention and cost-effective optimization. For example, if analysis reveals that the primary causes of urban flooding are aging drainage networks and high rates of surface hardening, priority can be given to network renovation and the addition of permeable pavement, rather than comprehensively increasing flood levee heights, to reduce remediation costs.

[0069] If the hazard factors selected for study have no relationship or a nonlinear numerical relationship, new parameters are matched as coefficients to introduce a hidden index I. The introduction of a hidden index is the most critical technique in disaster risk assessment in this embodiment. Its core value lies in transforming complex multi-factor problems into a concise single-variable analysis framework through mathematical transformation, while preserving the combined impact of multiple factors. The introduction of a hidden index integrates dimensionality reduction with multi-factor information, simplifying and normalizing multiple variables. In flood disasters, hazard factors (such as maximum inundation depth, flow velocity, and inundation duration) often have different dimensions and complex interactions.

[0070] This example uses two disaster factors as an example to illustrate, and uses water depth as an independent variable to build a model with vulnerability. The water depth of the disaster factor is extracted as d and the flow velocity is recorded as v. The relationship between d and v is determined; if d and v are related, then It is composed of d and v=f(d); if d and v are not related, then I is composed of d and v; at this time, the hidden index is introduced, and the calculated I is used as the hidden index.

[0071] The hidden index I is calculated as follows: , where d and v are the maximum flooding depth hazard intensity and the maximum velocity hazard intensity, respectively. is a random error term that obeys the standard normal distribution; is a parameter, n=disaster factor+1, and through exponential transformation we can get: This structure allows for simultaneous analysis of the combined impact of multiple independent variables on losses. This patent uses water depth d as the main independent variable to construct a model related to vulnerability.

[0072] The hidden index transforms multiple factors into dimensionless comprehensive indicators through logarithmic linear combination, eliminates the dimensional differences of the physical quantities of different disaster-causing factors, and enables different factors to be directly linearly superimposed to achieve dimensional unification; Automatically assign weights to each factor, for example, , which shows that the influence of water depth on loss is greater than that of flow velocity, and information compression can be achieved without artificial subjective weighting.

[0073] Random error term introduced in the hidden exponential model , which follows a standard normal distribution, is essentially a statistical representation of unobserved factors, such as local topographical fluctuations and differences in building structure. For example, at the same water depth d and flow velocity v, the actual losses of different buildings may fluctuate depending on the unquantified factor of whether they are equipped with flood control facilities. This can capture such uncertainties and make the model more realistic.

[0074] If a new disaster factor is added, just add an item to the hidden index And re-estimate the parameters without reconstructing the entire model framework, and factors can be added or subtracted conveniently.

[0075] According to this step, the loss estimation problem affected by multiple independent variables is transformed into a problem of hidden index I and loss rate, and the loss estimation problem affected by multiple independent variables is transformed into a univariate regression problem of hidden index I and loss rate. The logarithmic transformation and exponential structure are used to achieve the fusion of multi-parameter information.

[0076] In addition, the multiplication structure allows for simultaneous analysis of more independent variables that may affect the evaluation results. This formula shows the structure for estimating two independent variables.

[0077] Step S3: Conduct vulnerability analysis and construct a loss curve.

[0078] In step S3, a multi-parameter flood disaster vulnerability loss rate curve is constructed under regression analysis; specifically, the following steps are performed: constructing a relationship between the loss state of the vulnerable unit and the disaster-causing factors under the hydrodynamic model simulation, as a loss state vulnerability curve; calculating the exceedance probability of each loss state; and finally, using the loss state vulnerability curve to weight the relationship between the loss rate and the disaster intensity by setting the proportion of each parameter, as a loss rate curve.

[0079] In step 3, the parameter preprocessing is performed to transform multiple flooding parameters into a single hidden index I through logarithmic transformation and linear combination, thus simplifying the complexity of the regression model.

[0080] Then the distribution assumption and model construction map the probability, using the cumulative function of the Lognormal distribution , construct the mapping relationship between disaster-causing factors and loss state exceedance probability, form the probability framework of the vulnerability curve, convert the continuous disaster intensity into the probability of occurrence of discrete loss state, and characterize the statistical law that the higher the intensity, the more serious the loss state.

[0081] The lognormal model is a probability distribution that states that the logarithmic value of a random variable follows a normal distribution. This distribution characteristic is generally suitable for situations where the random variable is always greater than 0. Its results are often skewed to the right and exhibit a long tail. The lognormal regression model can be used to predict future values ​​or assess risk based on fitted parameters. Furthermore, the model can be used to multiply multiple variables to better reflect actual conditions.

[0082] In the calculation process, the loss rate is integrated, and the probability of each loss state is integrated through weight distribution to generate a loss rate curve containing upper and lower bounds and a mean, providing a quantitative basis for flood disaster loss assessment.

[0083] Finally, parameter estimation and verification are carried out, driven by historical data, and the judgment results based on disaster samples are obtained. , the model parameters are estimated by the maximum likelihood method to ensure that the model parameters are optimally matched with historical data, and the curve fitting is consistent with the actual loss distribution, and finally a disaster intensity-loss state probability relationship curve that can be used for prediction is formed.

[0084] This method integrates multi-parameter information through hidden index, realizes multi-parameter dimensionality reduction through hidden index, handles uncertainty with the help of the statistical characteristics of log-normal distribution, and combines the statistical characteristics of historical data. The final constructed loss rate curve can not only reflect the probabilistic relationship between disaster intensity and loss status, realize the probabilistic mapping from disaster physical intensity to loss status, but also flexibly adjust the conservatism and accuracy of the assessment results through weights, which is suitable for quantitative analysis of flood disaster vulnerability.

[0085] Step S31: Select disaster-prone object types Y1 and Y2 based on historical disaster data and meteorological data. Disaster-prone object types include casualty data, direct or indirect economic losses, direct or indirect losses to buildings and structures, direct or indirect losses from business closures, and storage losses. Any two of the selected disaster-prone object types are denoted as Y1 and Y2. Generally speaking, in flood disaster process and event simulations, economic losses are of primary concern to affected residents and are one of the primary outputs of many assessment models. However, the magnitude of economic losses should only be part of a disaster assessment; the impact of factors such as casualty data, building damage, traffic congestion, and social stagnation must be considered based on local characteristics and actual damage losses.

[0086] Step S32: After selecting the disaster-prone body types Y1 and Y2, similar attributes are classified. This means that the fusion of the collected data is considered and vulnerable units with the same attributes or similar geographical locations are aggregated, thereby reducing the specificity of the affected units and improving versatility. Identical attributes that can be classified include: land classification, building groups, etc. Specifically, the disaster data within the same county is divided according to land classification, and different attribute bands are assigned to the disaster data for different land classifications. This operation facilitates the subsequent determination of loss rates for different land classifications. Specifically, it facilitates the generalization of loss rate differences caused by regional and climatic differences in counties or cities to direct or indirect links with land classification, thereby reducing the specificity of the affected units and improving versatility.

[0087] Step S33: Assuming a normal distribution relationship exists between the hidden index and the disaster-causing factor, the relationship between the hidden index and the loss state is constructed based on the Lognormal distribution function as the loss state vulnerability curve, that is, the loss rate.

[0088] The lognormal distribution function was used to establish the relationship between the hidden index (I) and disaster loss intensity, such as flood depth, and to construct a multi-parameter regression analysis model for vulnerable units. By analyzing the relationship between various influencing factors, such as water area, flow velocity, and economic development level, and flood disaster losses, regression analysis was used to construct flood vulnerability curves for different regions, converting each influencing factor into quantifiable loss assessment parameters.

[0089] This example uses the Lognormal distribution function to construct a probabilistic vulnerability model. This model associates a hidden index (representing the potential impact of a disaster) with the loss intensity of flood disasters (e.g., flood depth), describing the relationship between the intensity of disaster losses and the hidden index I. Using a regression analysis framework, the relationship between loss intensity (e.g., water depth) and the hidden index I is modeled. The probabilities of different loss states are then estimated using the Lognormal distribution. Assuming a logarithmic relationship between the hidden index and disaster intensity, the probability of each loss state occurring is calculated. A mapping relationship is then established between water depth d and the probability of a loss state being exceeded, allowing for the calculation of vulnerability curves for different loss states.

[0090] In their study, Shinozuka et al. (2000) proposed a probabilistic vulnerability curve model based on the Lognormal distribution for earthquake-related bridge structure loss data. This two-parameter Lognormal distribution can independently or simultaneously estimate the probabilities of multiple loss states. This example incorporates the probabilistic vulnerability model based on the Lognormal distribution into a regression analysis framework and uses the logarithmic relationship between the hidden index I and the hazard factor to calculate the vulnerability curves for different loss states: .in, The loss rate under the condition of water depth d, The fragility curve is also the cumulative distribution function of the standard normal distribution. It is a hidden index calculated from water depth and parameters. It is a logarithmic linear function of water depth and represents the probability of exceeding a specific loss state, that is, the probability that the loss state reaches or exceeds a certain threshold. The continuous change of water depth is mapped to a probability value in the interval [0, 1] to characterize the probability of occurrence of loss state under different water depths.

[0091] In this embodiment, the hidden index I and the exceedance probability of the loss state Through the standard normal distribution function , so the extreme value distribution of disaster losses often approaches the logarithmic normal form, for example, if a few high-loss events dominate the total losses, , Directly express the probability that the loss level exceeds a certain threshold, making it easier for decision makers to understand the risk level.

[0092] and Represents the undetermined parameters used to calculate the probability of exceeding a certain loss state, which control the position (such as the median water depth) and shape (such as the slope, reflecting the sensitivity of losses to changes in water depth) of the vulnerability curve. The loss state exceedance probability is used in the process of constructing the vulnerability curve to analyze the loss amount of each vulnerable unit in detail.

[0093] Step S34: Construct a relationship between disaster intensity and loss status based on historical disaster data. Based on the historical disaster data, the damage factor values ​​in the data are used as specific quantitative values ​​of disaster intensity. A standard for classifying disaster loss status is established, and the probability of occurrence of a specific loss status is calculated using the exceedance probability difference between adjacent loss statuses.

[0094] Using historical disaster data, the loss rate is divided into six continuous and mutually exclusive intervals from 0 to 1. Each interval corresponds to a different loss state. From 0 to 1, the loss level is divided into six categories: no loss (0-0.1), slight loss (0.1-0.3), moderate loss (0.3-0.5), heavy loss (0.5-0.7), severe loss (0.7-0.9), and complete loss (0.9-1.0). Loss states appear sequentially, meaning that complete loss must have occurred before other loss states. Therefore, a loss state curve reflects the probability of at least one loss state occurring under certain natural conditions when the hazard factor is at a certain value. The value of the loss state curve at this time is the loss state exceedance probability.

[0095] When the loss status is no loss, the loss rate range , the disaster has no obvious impact, and the property or function loss can be ignored; when the loss status is slight loss, the loss rate range is , local minor damage, low repair cost, and short-term function recovery; when the loss state is moderate loss, the loss rate range is , moderate damage, requires a certain amount of time and resources to repair, and the function is partially limited; when the loss state is a heavy loss, the loss rate range is , severe damage, high repair cost, and long time to restore function; when the loss state is severe loss, the loss rate range is , close to complete damage, basic functions lost, need large-scale reconstruction; when the loss state is complete loss, the loss rate range , completely damaged, beyond repair, and completely lost its function.

[0096] Since the calculation principle of this regression model is to use the integral result based on probability density as the probability of at least one loss state occurring, the value between two adjacent loss states must be calculated to calculate the probability of each loss state: ,in, Indicates based on and The calculated hidden index, which, along with the corresponding two unknown parameters, will be used to calculate the probability of exceeding at least this loss state; Indicates based on and The calculated hidden index, along with the corresponding two unknown parameters, is used to calculate the probability of a exceedance of at least the previous loss state.

[0097] Each set of loss states (e.g. no loss, slight loss) corresponds to an independent set of hidden index parameters 、 、 and , can be estimated in parallel, flexibly adapting to the differentiated vulnerability of different loss levels. If more factors are added, they are all calculated independently without the need to restructure the entire model framework.

[0098] Take the loss status as no loss as an example, The result difference of is the probability of the occurrence of the no-loss state, which will represent the probability of the occurrence of the no-loss state, that is, For example, when the water depth When, through calculate , and then substitute The exceedance probability of each loss state is obtained, and the probability of the specific state is obtained by difference.

[0099] Correspondingly, the specific probability of occurrence of six loss levels, namely no loss, slight loss, moderate loss, heavy loss, severe loss, and complete loss, at different water depths can be calculated. For each disaster intensity (such as different water depths d), the probability of occurrence of all loss states is calculated to form a water depth-loss state probability curve.

[0100] Step S35: Weight each loss state to obtain a loss rate curve.

[0101] Loss rate curve constructed by hidden index , is essentially a proxy variable for the combined effects of multiple factors. For example, when the water depth increases, I increases accordingly. It approaches 1 quickly, indicating an increasing probability of high losses. This process implies that the deeper the water depth → the faster the flow velocity → the longer the flooding duration, without the need to explicitly model all factors.

[0102] The loss status is only a rough estimate of the specific loss amount and cannot be directly used to calculate the possible detailed loss amount. Even if it is used, it can only obtain a rough range of the approximate loss amount. This embodiment proposes to calculate the detailed loss rate under different disaster factors and different disaster intensities based on the differentiation intervals of different loss statuses. , and introduce weight values Calculate the possible loss rate range, the formula is expressed as:

[0103] in Will be expressed in is the disaster loss rate under a certain value.

[0104] in The values ​​of are 0, 0.5 or 1, which represent the upper bound, lower bound and mean of the loss rate of the loss state respectively.

[0105] Step S36: Use the maximum likelihood method to solve the undetermined parameters.

[0106] Assume that there are N samples of historical disaster data, and the hidden index of each sample is , the probability of belonging to a certain loss state is given by Indicates that, taking the state without loss as an example, the likelihood function L is expressed as follows: , where N represents the number of all reported samples and i represents the i-th sample; Indicates that each sample is based on and The calculated value of; the loss state of each sample is determined by the discriminant variable Indicates the result. When the reported loss rate of a sample is 0~0.1, =1, that is, no loss, otherwise =0.

[0107] In actual calculations, since there may be cases where the undetermined parameters are negative, it is better to start the basic machine reference value from [0, 0] to avoid negative values ​​affecting the calculation stability. By maximizing the likelihood function Solving for undetermined parameters and Each set of loss states corresponds to an independent set of parameters, resulting in six sets totaling 12 undetermined parameters. Each set of undetermined parameters is independent of each other and has no direct connection, and is estimated separately using the aforementioned method. This approach integrates the loss states and hazard factors of all historical samples to avoid bias in subjective parameter settings and utilizes full sample information. Initial parameters are set to [0, 0], and a logarithmic transformation is combined to calculate the log-likelihood function, effectively handling negative parameter values ​​and ensuring convergence of the iterative optimization process.

[0108] Step S4: Flood disasters caused by precipitation are constructed based on the hydrodynamic model of the HEC-RAS simulation software. Based on this inversion, a coupled experimental data set is obtained. The loss rate of the hydrodynamic model under different simulation conditions is calculated using the loss rate curve. According to the coupling results of the disaster-causing factors and the disaster-bearing body type, a total loss formula is constructed to calculate the estimated loss amount of flood disasters under different simulation conditions as the quasi-real-time total loss amount of the disaster area.

[0109] Step S4: Flood disasters caused by precipitation are constructed based on the hydrodynamic model of the HEC-RAS simulation software. Based on this inversion, a coupled experimental data set is obtained. The loss rate of the hydrodynamic model under different simulation conditions is calculated using the loss rate curve. According to the coupling results of the disaster-causing factors and the disaster-bearing body type, a total loss formula is constructed to calculate the estimated loss amount of flood disasters under different simulation conditions as the quasi-real-time total loss amount of the disaster area.

[0110] The steps for calculating the loss amount specifically include: obtaining a loss rate curve, separating it according to different land classifications or geographical locations and constructing loss rate curves in parallel, so as to obtain feasible loss rate curves under different separation conditions, and modifying the curve according to manual experience to obtain a loss rate information table under different disaster intensities under different flood physical disaster factors, and using the information table in combination with the coupled experimental data set to obtain detailed disaster loss data.

[0111] The key to calculating losses lies in converting precipitation into flooding depth using the hydrodynamic model of the HEC-RAS simulation software. Combined with a compartmentalized, corrected loss rate curve, total losses are then accumulated by affected unit to arrive at the total losses. This process, comprised of four steps: model simulation, curve construction, data coupling, and quantitative calculation, achieves a precise inversion from meteorological disaster factors to economic losses, achieving both scientific validity and engineering practicality.

[0112] Using the HEC-RAS simulation software hydrodynamic model as a foundational tool, we constructed flood disaster scenarios under different precipitation-induced conditions. Based on this inversion, the HEC-RAS simulation software hydrodynamic model construction and flood disaster simulation can simulate precipitation-induced flood disasters and inundation processes, outputting high-precision two-dimensional inundation parameters, including maximum inundation depth d, flow velocity v, and inundated area. Model inputs include DEM data of the affected units, land classification, and precipitation data, including rainfall amount, intensity, and duration.

[0113] Using precipitation data as a driving condition, the model calculates the surface runoff of precipitation and simulates flood evolution under different precipitation scenarios. Output the flood depth of vulnerable units (corresponding to the water depth value of each disaster-stricken unit i), forming a flooded water depth matrix covering the disaster-stricken units, and providing disaster factor intensity data for subsequent loss calculations.

[0114] Simulation conditions were set to vary, including precipitation intensity, duration, and inundation area, to generate multiple inundation parameter datasets covering varying disaster severity levels. The accuracy of the two-dimensional vulnerable cells simulated in the HEC-RAS simulation software model was ensured to be consistent with the accuracy of the affected cells in data such as DEMs and GDP distributions. For data with inconsistent accuracy, interpolation was used to unify the spatial resolution to avoid calculation errors caused by data misalignment.

[0115] By obtaining the loss rate curves reflecting the relationship between different disaster-causing factors and loss rates, the loss rate curves are separated and constructed in parallel according to different land classifications or geographical locations.

[0116] Classify the affected units according to their hazard-bearing body types and separate them spatially. Divide the affected units into several sub-areas based on land classification or geographical location (such as administrative divisions, watershed divisions, and altitude zones). For example, a city can be divided into central business districts, peripheral residential areas, and suburban farmland, and a loss rate curve can be constructed for each sub-area. , reflecting the different vulnerabilities of different disaster-bearing body types to flooding depth.

[0117] For each sub-region, the hidden index model constructed in the previous step is obtained , fitting results in a sub-region-specific loss rate curve. For example, the loss rate in a commercial area might reach 30% at a water depth of 1 meter, while the loss rate in farmland at the same water depth is only 10%, reflecting the impact of the characteristics of the hazard-bearing body on losses.

[0118] Actual loss data from each affected unit was used to validate and weight the loss rate information table, ensuring that the model's calculations aligned with the actual disaster situation. The curve was then corrected based on manual experience, taking into account regional specificity and historical case calibration factors. Domain experts were organized to manually revise the initially constructed loss rate curve. Regional specificity refers to factors such as local building flood control standards, economic density differences, and emergency response capabilities. Historical case calibration refers to loss data from actual disasters in recent years to adjust the curve's offset or slope.

[0119] The information table is then combined with the coupled experimental data set for calculation, and the corrected loss rate curve is discretized into a table form to record the loss rate of each sub-region corresponding to different disaster intensities, that is, different values ​​of disaster factors. .

[0120] The calculation of the loss amount also depends on the accuracy and scope of the sub-area. To ensure the credibility of the experimental results, in the first step, we should try to select GDP distribution or DEM data with strong timeliness and high accuracy, and ensure that the two-dimensional simulation area in the HEC-RAS simulation software is consistent with the accuracy of DEM and other data. For inconsistent situations, we can use interpolation and other methods to optimize. Divide the disaster-stricken unit into N disaster-stricken units that are consistent with the simulation grid of the HEC-RAS simulation software, and extract the GDP data of each disaster-stricken unit before the disaster. If other types of disaster-prone objects are involved (such as population and number of buildings), corresponding data (such as population of the affected unit and building asset value) must be prepared separately and classified according to loss type (personnel casualties, building damage).

[0121] This embodiment takes the water depth d as an example and GDP as the disaster-bearing body type for explanation. The total amount of direct or indirect economic losses is calculated using the formula: ,in, It is expressed as the total quasi-real-time GDP loss in the affected unit due to the flood depth. This formula is expressed as the GDP data of the i-th disaster-stricken unit. This formula realizes the spatial explicit calculation of the economic losses of the disaster-stricken units by accumulating the disaster-stricken units.

[0122] Among them, when calculating the losses of different types of disaster-prone objects, such as population and building damage, they can be differentiated according to the earlier statistics of the loss extent, so as to achieve the estimation of different disaster-prone losses.

[0123] This system integrates the entire process from precipitation-driven flood simulation to loss quantification. Through the HEC-RAS simulation software's high-precision hydrodynamic simulation and detailed modeling of compartmentalized loss rate curves, it can invert flood losses under different precipitation conditions, providing a scientific basis for disaster warning, emergency resource allocation, and post-disaster loss assessment. It is particularly suitable for high-resolution urban flood risk assessment and supports refined disaster prevention and mitigation planning.

[0124] In the specific embodiment, steps S5 and S6 are also included, which are specifically expressed as follows: Step S5: Perform multi-scale simulation and high-precision assessment. By simulating by region and time period, the accuracy and real-time performance of loss calculations are improved. Multi-scale applications can be implemented, supporting refined loss assessments at the city level, county level (as in Example 4), small regional level (as in Example 5), or hourly level (as in Example 6). Simulation accuracy is improved by refining the simulation scope (e.g., county, town, and village levels). Large areas can be further divided, and independent calculations performed for each county, town, and village to enhance assessment accuracy. Regarding simulation accuracy, the refinement of simulation results is further improved by resampling the 100m resolution simulation results to a higher precision of 30m.

[0125] Step S6: Conduct real-time monitoring and feedback, and visualize the disaster assessment results to facilitate decision support.

[0126] It provides a real-time monitoring interface that can display real-time precipitation, flood water levels, inundated areas and other information, and promptly feedback the scope of disaster impact and loss trends to assist in disaster emergency response. It also uses the GEE platform, HEC-RAS simulation software and visualization interface to display loss assessment results in the form of maps, charts, etc., to help decision makers understand the specific situation of flood disaster losses in various regions and support post-disaster rescue, reconstruction and other decisions.

[0127] In a specific embodiment, a visualization and reporting module is also included.

[0128] The Visualization and Reporting module generates disaster loss estimation reports and provides real-time map visualization, helping users intuitively understand the disaster losses at different time points. It also provides post-disaster loss trend charts, showing how the disaster situation has evolved over time, and supports filtering and viewing disaster information by region, time, and other dimensions. Based on the loss estimation results, the system can provide decision-making support to governments and post-disaster response teams, prioritizing response to high-loss areas and formulating appropriate emergency and rescue strategies.

[0129] In Example 2: The second embodiment of the present invention is different from the previous embodiment in that: When the functions are implemented as software functional units and can be sold or used as independent products, they can be stored in a computer-readable storage medium. Therefore, the technical solution of the present invention, in essence, either its contribution to the prior art or part of the solution, can be presented in the form of a computer software product.

[0130] The software product is stored in a storage medium and contains a series of instructions that enable a computer device, such as a personal computer, server, or network device, to execute all or part of the steps of the method described in each embodiment of the present invention. Such storage medium includes any type of medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0131] The logic and steps presented in the flowchart or described in other forms can be considered as an ordered set of executable instructions that perform logical functions. They can be embodied in any computer-readable medium for use by an instruction execution system, device, or apparatus (such as a computer-based system, a system containing a processor, or other system that can obtain and execute instructions from an instruction execution system, device, or apparatus), or for use in conjunction with such instruction execution systems, devices, or apparatuses. Within the scope of this specification, "computer-readable medium" refers to any device that can contain, store, communicate, propagate, or transmit a program for use by an instruction execution system, device, or apparatus, or for use in conjunction with such instruction execution systems, devices, or apparatuses.

[0132] The following are more specific examples of computer-readable media, but this list is not exhaustive: an electrical connection with one or more wires (part of an electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable media on which the program is printed. This is because the program can be obtained in electronic form by optically scanning the paper or other media and then editing, deciphering, or otherwise processing it in an appropriate manner, as necessary, and then stored in a computer memory.

[0133] It should be understood that various aspects of the present invention can be implemented via hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented via software or firmware stored in a memory and executed by an adapted instruction execution system. For example, if hardware is used for implementation, as in another embodiment, any one or a combination of the following technologies well known in the art can be utilized: discrete logic circuits with logic gates that implement logical functions of data signals, application-specific integrated circuits with appropriate combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0134] In Example 3, this example performs a city-level assessment based on the real-time flood disaster loss estimation method of Example 1, and simulates and evaluates the rainfall and flooding in a city from 06:00 to 12:00. Figure 6 As shown in Figure 1, a typical example of the implementation effect of the method of the present invention is shown. It is the city-wide loss estimation distribution result obtained based on the coupled model of flood coverage, GDP, and land use type. (a)-(d) are respectively the flood coverage, GDP distribution, land use type distribution, and city-wide loss estimation distribution under certain precipitation conditions. From this figure, we can see that: (a) Based on a 6-hour continuous rainfall scenario with an average rainfall intensity of 10 mm / h, a 100-meter-resolution inundation coverage model was constructed using the HEC-RAS watershed hydrodynamic algorithm. Through parameter optimization and dynamic boundary condition correction techniques, the model accurately depicts the spatial distribution of inundation at depths ranging from 0.001 mm to 1 m. Simulation results show that the river floodplain and confluence scour zone exhibit significant high inundation risk. The inundation area around the river channel is reduced by 12.5% ​​compared to traditional algorithms, validating the model's ability to accurately control the flood boundary. To meet the needs of multi-source data fusion, an innovative "HEC-RAS simulation-bilinear interpolation" joint upscaling method was proposed, improving the original simulation results from 100 m to 30 m accuracy. (b) (c) Using spatial resampling techniques, GDP distribution data and land use classification data, originally at 1 km resolution, were uniformly upscaled to 30 m resolution. Economic spatial data reveals a typical "core-periphery" structure, with the central urban area contributing 72.6% of the county's total GDP. 83% of high-economic-density areas are located along major rivers calibrated by the HEC-RAS model, forming a hydrological-economic spatial coupling. A detailed analysis of the coupling model reveals that most highly developed economic areas, such as towns, are located adjacent to perennial rivers. Combined with the aforementioned precipitation and flood model results, which show significant flooding in river areas, this suggests that these cities face a certain risk of economic losses under the precipitation and flood model.

[0135] In fact, in the process of simulating rainfall and flooding and estimating losses, in order to avoid the overlap of the catchment area and the perennial water storage area caused by the simulation, which would lead to the calculation of additional loss estimates and thus reduce the accuracy of the assessment, the traditional method is used to misjudge the water storage area. The calculation of loss values ​​for two sensitive areas, farmland (NDVI>0.6) and impervious surfaces (ISA>50%), is specifically limited. Under this calculation premise, this paper uses the loss rate function independently developed, and finally generates a 30m precision spatial distribution map of economic losses. Figure 6 (d) shows that the primary losses, due to the uneven distribution of GDP, are concentrated in economically developed regions. This means that, without considering emergency response measures, highly developed areas such as towns and cities will be hardest hit by economic losses. Simulation results show that due to the combined effects of high-precision simulated flood boundaries and economic density, the central urban area accounts for 89.7% of the losses, with a total estimated loss of 340 million yuan (95% confidence level).

[0136] In Example 4, this example uses the real-time flood disaster loss estimation method of Example 1 to perform county-level scale optimization.

[0137] In actual application, too low simulation accuracy often seriously affects the value and reliability of the evaluation results. Taking into account the model operation time, in this example, the simulation area in Example 3 will be respectively simulated according to the administrative area division standard for each county for 6 hours of precipitation with the same precipitation intensity as in Example 3, resulting in flood disaster simulation. In order to improve the evaluation accuracy, the municipal assessment unit is subdivided into 9 county-level units (Counties A-Ren), and 30m high-precision simulation is used to carry out zoning assessment. The results show that the economic losses in County Wu reached a peak because the main river channel runs through the urban economic zone; although Counties A and C are mainly agricultural, the complex terrain exacerbates the depth of waterlogging, and the losses reached 44 million and 48 million yuan respectively. The cumulative losses at the county level are 314 million yuan, which is 7.65% higher than the overall evaluation accuracy at the municipal level, proving that the refinement of spatial units can effectively reduce model errors. Appendix Figure 7 The county-level loss distribution shows that economic development level and topographic characteristics are key factors influencing the spatial differentiation of losses. In practical applications, users can further increase the number of target area divisions based on their needs, such as further dividing counties into towns, towns into villages, and so on.

[0138] In Example 5, this embodiment uses the real-time flood disaster loss estimation method of Example 1 to perform customized regional assessment.

[0139] The method demonstrates its flexibility in micro-scale application, supports user-defined minimum units (minimum 1 km²), and can be customized to conduct a selective investigation of the damage status of the required research area according to user needs. The minimum unit area covers the area to conduct damage assessment at any point of interest (ROI) within the city. Figure 8 As shown, Figure 8 As shown in Figure 2, an assessment of a 1 km² agricultural-dominated area (>90% farmland) showed that crop damage resulted in an economic loss of RMB 240,000.

[0140] This model provides technical support for risk assessment of critical infrastructure or special economic units, and as can be seen from Examples 3-5, the real-time estimation method and system for flood disaster losses of this embodiment can realize the construction of a macro-meso-micro multi-level assessment system.

[0141] In Example 6, this embodiment uses a real-time estimation method for flood disaster losses in Example 1 to perform a time-series dynamic assessment of the flood inundation situation hour by hour under precipitation conditions. When higher-resolution precipitation data is available, the simulation accuracy can be further improved, such as every 30 minutes or even every minute.

[0142] In this example, hourly economic losses can be estimated based on the hourly flooding and inundation conditions. The relationship between the loss rate curve and inundation depth can be used to estimate the change in losses for a particular county under hourly simulation conditions. This example uses County E from Example 4 as an example, selecting the period from 1:00 AM to 12:00 PM as an implementation example.

[0143] like Figure 9 As shown in the figure, a time series assessment model was constructed based on hourly precipitation data. The dynamic evolution of losses in Wu County from 01:00 to 12:00 was analyzed. In this figure, the green / blue / red colors represent the extent of losses in Wu County under the simulated flood disaster conditions at 01:00, 06:00, and 12:00, respectively. A time series model of loss rate and inundation depth revealed a nonlinear increase in losses (700,000 → 1.4 million → 2 million), with an initial growth rate of 35% / hour over the first three hours and a decrease of 8.3% / hour in the later stages. Spatiotemporal overlay analysis showed that the cumulative losses at 12:00 were 18% less than the instantaneous peak, confirming that dynamic assessments can more accurately reflect the actual course of the disaster. Analysis of the estimated loss curve reveals that the estimated losses increase rapidly in the early stages of the simulation, but the growth trend slows in the later stages until the estimated loss curve flattens.

[0144] It can be seen from Example 6 that the estimated loss value increases faster over time in the early stage of simulation. Therefore, the real-time estimation method for flood disaster losses in Example 1 performs time series analysis and is suitable for rapid risk assessment when a disaster occurs.

[0145] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for real-time estimation of flood disaster losses, characterized in that: The following steps are involved: Step S1: Acquire data of the target area and extract water body information, disaster-causing factors and damage factors. The target area includes vulnerable units. Preprocess the data and perform spatial coupling to construct a coupled historical data set; Step S2: within the vulnerable unit, construct a hydrodynamic model based on the coupled historical data set, perform inversion simulation, and extract test water body information; determine the relationship between two or more disaster-causing factors based on the test water body information, and construct a relationship formula between the disaster-causing factors; Step S3: construct the relationship between the loss state and two or more disaster-causing factors as the loss state vulnerability curve; calculate the exceedance probability of each loss state, and combine the loss state vulnerability curve to obtain the relationship between the loss rate and the disaster intensity as the loss rate curve; Among them, disaster intensity is the value of the damage factor, and loss rate is the specific quantitative value of disaster intensity. The loss status of the disaster is divided according to the loss rate; Step S4: constructing a hydrodynamic model for flood disasters under different simulation conditions, and inverting it to obtain a coupled experimental data set; using the loss rate curve based on the coupled experimental data set, estimating the loss rate under different simulation conditions, and constructing a total loss formula by combining the disaster-causing factors and selecting the disaster-bearing body type; The estimated losses of flood disasters under different simulation conditions are calculated according to the total loss formula and used as the quasi-real-time total losses of the disaster area.

2. A method for real-time estimation of flood disaster losses according to claim 1, characterized in that: In step S1, the data of the target area obtained includes population data, economic data and land data; Among them, land data includes historical disaster data, geographic information, meteorological data and remote sensing satellite data. Historical disaster data includes damage factors and disaster-causing factors; the historical disaster data includes casualty data, building loss data and economic production reduction data; the geographic information includes the geographic coordinates, geographic parameters, soil classification and land classification of the target area; the geographic parameters include slope, aspect and distance to the river; the meteorological data includes precipitation data and evaporation data, and the precipitation data is processed into a time series; the casualty data, building loss data and economic production reduction data of the target area are used as damage factors, and water depth, flow velocity, inundation time and flow rate are used as disaster-causing factors; the remote sensing satellite data also includes disaster images, water body information and DEM data of digital elevation models; In step S1, water body information is extracted through remote sensing satellite data, and the target area is divided into vulnerable units and non-vulnerable units; Based on remote sensing satellite data, the time span is locked within the set time before and after the disaster to extract water body information; the SDWI water extraction index is calculated using Sentinel-1 GRD data or the maximum threshold method Otsu is used to process remote sensing satellite data, and the FwDET tool is used to combine DEM data to perform water depth inversion and extract water body information; The target area is divided into multiple units. Based on geographic information, meteorological data, and water body information, the hydrological characteristics of the water collection capacity and drainage capacity of each unit in the target area are calculated. By identifying the flooded area and combining hydrological characteristics, the units in the target area are divided into vulnerable units and non-vulnerable units. The hydrological characteristic calculation formula is expressed as: , , where the catchment area is the area of ​​the flooded area, the precipitation intensity is based on spatial interpolation of meteorological data, and the permeability coefficient is set according to the land classification; is the weight coefficient set for slope and distance to the river, and the drainage efficiency coefficient is set according to land classification; when the water collection capacity is greater than the drainage capacity, it is determined to be a non-vulnerable unit; when the water collection capacity is less than or equal to the drainage capacity, it is determined to be a vulnerable unit; In step S1, the collected population data, economic data, and land data are preprocessed and spatially coupled with geographic information to construct a coupled historical data set; Each unit is numbered and geo-referenced using geographic coordinates through the Geographic Information System (GIS) technology. The population, economic, and land data are matched with the vulnerable units so that each vulnerable unit corresponds to a set of data and attributes are associated. The population, economic, and land data corresponding to the vulnerable units, land classification, historical data on disaster-causing factors, and damage factors are marked. Different types of data are spatially aligned, and spatial coupling is completed to integrate them into a unified coupled data set as a coupled historical data set.

3. A method for real-time estimation of flood disaster losses according to any one of claims 1 and 2, characterized in that: In step S2, within the vulnerable unit, a regional hydrodynamic model under precipitation conditions is constructed based on remote sensing satellite data and geographic information coupled with the historical data set; and test water body information is extracted by inversion based on the hydrodynamic model simulation. The test water body information is coupled with the coupled historical data set to obtain a coupled experimental data set; the relationship between the disaster-causing factors in the vulnerable unit is determined based on the test water body information, and the relationship between the disaster-causing factors is constructed; step S2 specifically includes: Step S21: Based on water body information, DEM data, and land classification, a hydrodynamic model is constructed using HEC-RAS simulation software; wherein the DEM data is used to define boundary conditions, and the land classification is used to assign the Manning coefficient; Step S22: test water body information of each affected unit obtained through hydrodynamic model inversion under different simulation conditions; Step S23: Compare the water body information with the water body information marked in the coupled historical data set. Adjusting controllable parameters in the hydrodynamic model to improve accuracy, including the Manning coefficient, boundary conditions, and model input / output time intervals; The adjustable parameters affect the coefficient and equation intercept of the disaster factor under each logarithmic condition of the hidden index; Step S24: Similar to the construction of the coupled historical data set, the water body information extracted from the historical disaster data and remote sensing satellite data is replaced with the experimental water body information obtained by the hydrodynamic model simulation, and the corresponding experimental data of the disaster-causing factors are obtained based on the experimental water body information to construct the coupled experimental data set; Step S25: perform relationship judgment and transformation on the extracted disaster-causing factors; judge by whether there is a linear or nonlinear numerical relationship between the disaster-causing factors. If there is a relationship, perform numerical transformation processing to transform the independent variable factors into a single variable as much as possible; if there is no relationship, match new parameters as coefficients respectively to introduce hidden indexes.

4. The method for real-time flood disaster loss estimation according to claim 3, wherein the step S25 comprises the following steps for any two disaster factors, disaster factor A and disaster factor B: Let disaster factor A be the independent variable and be denoted as d, and disaster factor B be denoted as v; Determine whether d is related to v; if d is not related to v, then the hidden index I is composed of d and v; if d is related to v, then the hidden index I is composed of d and v = f(d).

5. A method for real-time estimation of flood disaster losses according to claim 4, characterized in that: The calculation formula of the hidden index I is expressed as: ， Where X represents any hazard factor, d and v represent the hazard intensity of the maximum hazard factor A and the hazard intensity of the maximum hazard factor B, respectively. is a random error term that obeys the standard normal distribution; As a parameter, we can get the following through exponential transformation: .

6. A method for real-time estimation of flood disaster losses according to any one of claims 1 and 2, characterized in that: In step S3, the relationship between the loss state and two or more disaster-causing factors is constructed as a loss state vulnerability curve, which specifically includes: Step S33: Assuming a normal distribution relationship exists between the hidden index and the disaster-causing factor, the relationship between the hidden index and the loss state is constructed based on the Lognormal distribution function as a loss state vulnerability curve; The relationship between the hidden index I and the disaster loss intensity is constructed using the lognormal distribution function, a multi-parameter regression analysis model of vulnerable units is constructed, and the hidden index is used to I The logarithmic relationship between the hazard factor and the vulnerability curve of different loss states is calculated, and the formula is expressed as: ， in, In the disaster factor d The loss rate under the conditions of d is any hazard factor, is the fragility curve, is a hidden index, which is determined by the disaster factor d The disaster factor is calculated by d The logarithmic linear function of the loss state represents the probability of exceeding a certain loss state, that is, the probability that the loss state reaches or exceeds a certain threshold. Disaster factors d The continuous change of is mapped to the probability value in the interval [0, 1] to characterize different disaster factors d The probability of the next loss state occurring; and It represents the undetermined parameter used to calculate the probability of exceeding a certain loss state. The probability of exceeding a loss state is used to analyze the loss amount of each vulnerable unit during the construction of the vulnerability curve.

7. A method for real-time estimation of flood disaster losses according to claim 6, characterized in that: In step S3, the values ​​of the damage factors are collected as disaster intensity based on historical disaster data, and the exceedance probability of each loss state is calculated. The loss state vulnerability curve is used to perform weighted calculations to obtain the relationship between the loss rate and the disaster intensity, which is used as the loss rate curve. Specifically, the following steps are performed: Step S34: Constructing a relationship between disaster intensity and loss status based on historical disaster data; formulating a classification standard for disaster loss status based on historical disaster data, and calculating the probability of occurrence of a specific loss status using the exceedance probability difference between adjacent loss statuses; Using historical disaster data, the loss rate is divided into six continuous and mutually exclusive intervals from 0 to 1. Each interval corresponds to a different loss state. From 0 to 1, the loss state is divided into six categories: no loss, slight loss, moderate loss, heavy loss, severe loss, and complete loss. The occurrence of loss states is sequential, that is, the occurrence of complete loss must have occurred before other loss states. When the loss status is no loss, the loss rate range ; When the loss status is a slight loss, the loss rate range is ; When the loss status is moderate loss, the loss rate range ; When the loss status is a heavy loss, the loss rate range ; When the loss status is severe loss, the loss rate range ; When the loss status is complete loss, the loss rate range ; To calculate the probability of occurrence of each loss state, it is necessary to calculate the value between two adjacent loss states, that is, to calculate the probability of occurrence of the loss state. The formula is expressed as: ， in Represents the undetermined parameter based on the probability of exceeding the loss state X1 and The calculated hidden index, which, along with the corresponding two undetermined parameters, will be used to calculate the probability of exceeding at least this loss state; Represents the undetermined parameter based on the probability of exceeding the loss state X2 and The calculated hidden index, which, along with the corresponding two undetermined parameters, will be used to calculate the probability of a exceedance of at least the previous loss state; Step S35: weighting each loss state to obtain a loss rate curve; According to the differentiation range of different loss states, the detailed loss rate is calculated under different disaster factors and different disaster intensities. , and introduce weight values Calculate the possible loss rate range, the formula is expressed as: ， in Will be expressed in is the disaster loss rate under a certain value of a certain hazard factor; A, B, C, D, E, and F represent six loss states: no loss, slight loss, moderate loss, heavy loss, severe loss, and complete loss, respectively; P represents the corresponding probability of occurrence. The values ​​of are 0, 0.5 or 1, which represent the upper bound, lower bound and mean of the loss rate of the loss state respectively.

8. The method for real-time estimation of flood disaster losses according to claim 6, characterized in that: In step S3, the maximum likelihood method is used to solve the undetermined parameters, which specifically includes: Step S36: Use the maximum likelihood method to solve the undetermined parameters; suppose there are N samples of historical disaster data, and the hidden index of each sample is , the probability of belonging to a certain loss state is given by Indicates that, taking the state without loss as an example, the likelihood function L is expressed as follows: ， Where N represents the number of all reported samples, and i represents the i-th sample; Indicates that each sample is based on and The calculated value of; the loss state of each sample is determined by the discriminant variable Indicates the result. When the reported loss rate of a sample is 0~0.1, =1, that is, no loss, otherwise =0.

9. The method for real-time estimation of flood disaster losses according to claim 6, characterized in that: In step S3, a total loss formula is constructed based on the coupling of disaster factors and disaster-bearing body types, specifically including: Classify the target area, divide the disaster-affected unit into N disaster-affected units, select the disaster-bearing body type as direct or indirect economic loss, and extract the GDP data of each disaster-affected unit using the coupled experimental data set; each sub-area uses the loss rate curve to estimate different disaster intensity That is, the loss rate corresponding to different values ​​of disaster factor i , and discretized into a table form, the information table is combined with the coupled experimental data set to estimate the total loss; For each hazard factor under simulation conditions , estimate the total amount of direct or indirect economic losses, the formula is expressed as: ， in, It is expressed as the total quasi-real-time GDP loss in the disaster-affected unit caused by the disaster factor. It represents the GDP data of the i-th disaster-stricken unit.

10. A real-time flood disaster loss estimation system, characterized in that: The system is applied to a real-time flood disaster loss estimation method according to any one of claims 1 to 9, comprising a data acquisition unit, a loss estimation unit, a flood simulation unit and a loss rate curve calculation unit connected in series between the data acquisition unit and the loss estimation unit, respectively; wherein: The data acquisition unit includes a data analysis module, a remote sensing satellite data processing module, a water body information processing module, and a historical data processing module, which are electrically connected in sequence. The data acquisition unit obtains population, economic, and land data from multiple data sources. The data analysis module uses the collected population data, economic data, and land data, including geographic information, meteorological data, remote sensing satellite data, damage factors, and disaster-causing factors. The remote sensing satellite data processing module extracts water body information from the collected data. The water body information processing module divides the target area into vulnerable units and non-vulnerable units based on the water body information. The historical data processing module preprocesses the collected population data, economic data, and land data, and performs spatial coupling to construct a coupled historical data set. The flood simulation unit includes a hydrodynamic model establishment module, an inversion module, an experimental data processing module, and a disaster factor relationship establishment module that are electrically connected in sequence; the hydrodynamic model establishment module establishes a regional comprehensive hydrodynamic model of flood disasters under precipitation conditions based on remote sensing satellite data and geographic information; the inversion module predicts flood conditions under different precipitation conditions by setting different precipitation conditions in the hydrodynamic model, and extracts test water body information based on the inversion under the hydrodynamic model simulation; the experimental data processing module couples the test water body information with the coupled historical data set to obtain a coupled experimental data set; the disaster factor relationship establishment module determines the relationship between disaster factors in the vulnerable unit based on the test water body information, constructs a relationship formula between disaster factors, and introduces a hidden index I; The loss rate curve calculation unit includes a vulnerability curve construction module and a loss rate curve construction module that are electrically connected in sequence; the vulnerability curve construction module uses regression analysis to construct a relationship between a loss state and two or more disaster-causing factors in a vulnerable unit as a loss state vulnerability curve; based on historical disaster data, each value of the damage factor is collected as the disaster intensity, and the probability of exceeding each loss state is calculated; the loss rate curve construction module uses the loss state vulnerability curve to perform weighted calculations to obtain a relationship between the loss rate and the disaster intensity as the loss rate curve; The loss estimation unit constructs flood disasters caused by different precipitation conditions based on the hydrodynamic model of the HEC-RAS simulation software, and based on this inversion, obtains a coupled experimental data set and uses the loss rate curve to estimate the loss rate of the hydrodynamic model under different simulation conditions. The total loss formula is constructed according to the coupling of disaster factors and disaster-bearing body types, and the estimated loss amount of flood disasters under different simulation conditions is calculated as the quasi-real-time total loss amount of the disaster area.

Images