A method, device, equipment and medium for establishing a shallow collapse disaster early warning model

Abstract

The application discloses a shallow collapse disaster early warning model establishment method, device and equipment and a medium, and comprises the following steps: acquiring historical disaster point data in a region to be warned; dividing a slope unit based on a digital elevation model, and extracting a plurality of disaster-inducing factors; constructing a collapse-prone prediction model based on a gradient boosting decision tree and evaluating the accuracy thereof to obtain a prone partition evaluation result; according to the prone partition evaluation result, statistically analyzing the daily rainfall and the previous rainfall of the historical disaster points in different prone areas, determining the previous rainfall days through a correlation coefficient, and establishing a previous rainfall-daily rainfall threshold model; according to whether the slope body is treated, the vegetation condition and the present stage deformation sign, dividing the slope unit into a prone grade, and correcting the prone grade and the rainfall threshold of the slope unit; and according to the prone grade and the rainfall threshold of the slope unit, constructing a multi-grade space-time early warning model. The method can correct the early warning model of the historical disaster points, and improve the accuracy and reliability of early warning.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for rock and soil disasters, and in particular to a method, apparatus, equipment and medium for establishing an early warning model for shallow collapse disasters. Background Technology

[0002] Shallow soil collapses frequently occur during the flood season, characterized by their numerous locations, wide distribution, small scale, and difficulty in early warning. They pose a serious threat to the lives and property of villagers in mountainous areas and create significant challenges for disaster prevention and mitigation efforts. Every year during the flood season, large-scale evacuations of villagers occur, wasting considerable human and material resources and causing severe disruption to both villagers and disaster relief personnel. The root cause lies in the inaccuracy of early warning thresholds for shallow soil collapses; they are either too insensitive or too sensitive.

[0003] Currently, early warning methods for shallow soil collapses, such as landslides, collapses, or debris flows, mainly employ statistical analysis, numerical simulation, and physical models. Numerical simulation and physical modeling methods are primarily used to study disaster initiation mechanisms and determine single-point disaster warning thresholds, focusing more on the theoretical level, but they have limitations in their application to engineering practice. While models based on statistical analysis lack mechanistic support, they have been widely adopted worldwide after years of development. Statistical models typically only require sufficient rainfall data and disaster site information, without considering the influence of other factors such as slope, vegetation, and lithology on slope instability.

[0004] In summary, the existing methods for establishing early warning models do not consider the impact of slope instability, resulting in low accuracy and reliability of their early warning thresholds. Summary of the Invention

[0005] This invention provides a method, apparatus, equipment, and medium for establishing a shallow landslide disaster early warning model, in order to solve the technical problem that existing early warning model establishment methods do not consider the impact of slope instability, resulting in low accuracy and reliability of their early warning thresholds.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a method for establishing a shallow landslide disaster early warning model, comprising:

[0007] Acquire historical disaster data within the area to be warned, including the location of the disaster, the time of the disaster, and data from nearby rain gauges at the time of the disaster.

[0008] Slope units were divided based on digital elevation model, and multiple disaster-prone factors were extracted based on ArcGIS;

[0009] Using the historical disaster data and the disaster-predating factors as the positive sample set, and selecting the same number of non-collapse points and disaster-predating factors as the negative sample set, a total sample set is formed and divided into a training set, a validation set and a test set according to the proportion.

[0010] A collapse susceptibility prediction model was constructed based on gradient-enhanced decision trees, and its accuracy was evaluated. The susceptibility zoning evaluation results were obtained.

[0011] Based on the susceptibility zoning evaluation results, the daily rainfall and previous rainfall of historical disaster sites in different susceptibility zones were statistically analyzed. The number of days of previous rainfall was determined by the correlation coefficient, and a threshold model of previous rainfall-daily rainfall was established.

[0012] Based on whether the slope has been treated, vegetation conditions, and current signs of deformation, the slope unit is classified into susceptibility levels, and the susceptibility level and rainfall threshold of the slope unit are adjusted accordingly.

[0013] Based on the susceptibility level and rainfall threshold of the slope unit, a multi-level spatiotemporal early warning model is constructed.

[0014] Preferably, the step of dividing the slope into units based on the digital elevation model includes:

[0015] Generate a depression-free area and calculate the flow direction and flow rate;

[0016] Generate a river network and connect the rivers, generate watersheds and convert the watersheds to facets, generate a positive catchment area, and invert the original DEM;

[0017] Repeat the steps of generating a river network and connecting rivers, generating a watershed and turning the watershed into a surface, generating a positive catchment area, and generating a negative catchment area.

[0018] The positive and negative water catchment areas are merged to generate an initial slope unit.

[0019] By correcting unreasonable boundaries, slope elements are obtained.

[0020] Preferably, the extraction of multiple disaster-causing factors based on ArcGIS includes:

[0021] Use the Euclidean distance tool in ArcGIS to calculate the distances from roads, faults, and water systems to various slope factors;

[0022] Take the average distance of all grid cells as the final value of the slope cell;

[0023] Specifically, when extracting elevation, slope, aspect, elevation difference, plane curvature, profile curvature, TWI, and NDVI, the average value of all grid cells within a slope cell is used as the final value for each slope cell. When extracting stratigraphic lithology using slope cells, if a cell involves multiple lithologies, the lithology with the largest area is used as the final value for the slope factor.

[0024] Preferably, the plurality of disaster-causing factors include: elevation, slope, aspect, elevation difference, plane curvature, profile curvature, topographic humidity index (TWI), normalized difference vegetation index (NDVI), lithology, distance to road, distance to fault, and distance to water system.

[0025] Preferably, the step of using the historical disaster site data and the disaster-prone factors as a positive sample set, and selecting the same number of non-collapse points and disaster-prone factors as a negative sample set to form a total sample set, which is then proportionally divided into a training set, a validation set, and a test set, includes:

[0026] The historical disaster data and the disaster-prone factors are used as the positive sample set;

[0027] Outside the buffer zone of the historical disaster point, the same number of non-collapse points are randomly selected to construct a negative sample set;

[0028] The total sample set is formed based on the positive sample set and the negative sample set;

[0029] The total sample set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1, which are used for model training, parameter tuning, and testing, respectively.

[0030] Preferably, the step of constructing a collapse susceptibility prediction model based on a gradient-enhanced decision tree and evaluating its accuracy to obtain susceptibility zoning evaluation results includes:

[0031] Initialize the weak classifier:

[0032]

[0033] f0(x) is the initial weak classifier, L(y) i ,γ) is a differentiable loss function;

[0034] Calculate the negative gradient:

[0035]

[0036] In a regression tree, it is actually yc, where c refers to the average value of the sample labels of each leaf node.

[0037] Using the residual in formula (2) as the new value of the sample, (x) m ,r im), m = 1, 2, 3, ..., M are used as training data for the next tree;

[0038] Based on the new values ​​of the samples, a new weak learner G is trained. k (X) is:

[0039]

[0040] Returning to the step of calculating the negative gradient, until the minimum error requirement is met, the final strong classifier is:

[0041]

[0042] In the formula, f m (x) is the collapse susceptibility prediction model, f0(x) is the initial model, and its value is the minimum value of the loss function; m is the maximum number of iterations, J is the number of leaf nodes of the j-th regression tree, and γ is the optimal fit value of the negative gradient of the loss function in the leaf node region, corresponding to the leaf node region R of the j-th regression tree. jm (j = 1, 2, 3, ..., J), / is the indicator function, R jm It is the leaf node region corresponding to the t-th decision tree; the maximum number of iterations (m) is set to 150, the learning rate is 0.5 when considering the data, and the number of leaf nodes (J) is set to 20.

[0043] Preferably, the step of statistically analyzing the daily rainfall and previous rainfall at historical disaster sites in different disaster-prone areas based on the susceptibility zoning evaluation results, determining the number of previous rainfall days through correlation coefficients, and establishing a previous rainfall-daily rainfall threshold model includes:

[0044] The number of disaster-stricken points, the rainfall on the day the disaster occurred, and the rainfall in the previous nine days were counted separately in different susceptibility zones.

[0045] Correlation analysis was used to determine the number of days with previous rainfall and to establish a threshold model for previous rainfall-to-current rainfall for different susceptibility zones.

[0046] The formula for the correlation coefficient is as follows:

[0047]

[0048] In the formula, r(X,Y) represents the degree of correlation between the two variables; cov(X,Y) represents the covariance of X and Y; var(X) represents the variance of x; and var(Y) represents the variance of y.

[0049] Secondly, the present invention provides a device for establishing a shallow landslide disaster early warning model, comprising:

[0050] The data acquisition module is used to acquire historical disaster data within the area to be warned. The historical disaster data includes the location of the disaster, the time of the disaster, and the data of nearby rain gauges at the time of the disaster.

[0051] The slope division module is used to divide slope units based on the digital elevation model and extract multiple disaster-prone factors based on ArcGIS.

[0052] The sample generation module is used to take the historical disaster point data and the disaster-predatory factors as the positive sample set, select the same number of non-collapse points and disaster-predatory factors as the negative sample set, form a total sample set, and divide it into training set, validation set and test set according to the proportion.

[0053] The model evaluation module is used to construct a collapse susceptibility prediction model based on gradient-enhanced decision trees and evaluate its accuracy, thereby obtaining susceptibility zoning evaluation results.

[0054] The rainfall threshold module is used to statistically analyze the daily rainfall and previous rainfall of historical disaster sites in different susceptibility zones based on the susceptibility zoning evaluation results, determine the number of previous rainfall days through correlation coefficients, and establish a previous rainfall-daily rainfall threshold model.

[0055] The grade correction module is used to classify the slope unit into a susceptibility grade based on whether the slope has been treated, vegetation conditions, and current deformation signs, and to correct the susceptibility grade and rainfall threshold of the slope unit.

[0056] The early warning model building module is used to construct a multi-level spatiotemporal early warning model based on the susceptibility level and rainfall threshold of the slope unit.

[0057] Thirdly, the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the shallow collapse disaster early warning model establishment method described in any one of the above.

[0058] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the shallow collapse disaster early warning model establishment method described in any one of the above.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This invention provides a method for establishing a shallow landslide disaster early warning model, comprising: acquiring historical disaster point data within the area to be warned, the historical disaster point data including the location of the disaster, the time of the disaster, and nearby rain gauge data at the time of the disaster; dividing the slope into units based on a digital elevation model, and extracting multiple disaster-inducing factors based on ArcGIS; using the historical disaster point data and the disaster-inducing factors as a positive sample set, selecting an equal number of non-landslide points and disaster-inducing factors as a negative sample set, forming a total sample set and dividing it proportionally into a training set, a validation set, and a test set; and constructing a landslide early warning model based on a gradient-enhanced decision tree. A landslide susceptibility prediction model was developed and its accuracy evaluated to obtain susceptibility zoning evaluation results. Based on the susceptibility zoning evaluation results, the daily rainfall and previous rainfall at historical disaster sites in different susceptibility zones were statistically analyzed. The number of days of previous rainfall was determined through correlation coefficients, and a previous rainfall-to-daily rainfall threshold model was established. Based on whether the slope has been treated, vegetation conditions, and current deformation signs, the susceptibility levels of the slope units were classified, and the susceptibility levels and rainfall thresholds of the slope units were corrected. Based on the susceptibility levels and rainfall thresholds of the slope units, a multi-level spatiotemporal early warning model was constructed.

[0061] This invention starts with historical disaster sites and uses a combination of statistical analysis and machine learning. Using slope units as the smallest predictive unit, it considers the results of the latest round of risk and hazard point surveys to correct their susceptibility and thresholds. This invention assesses the current stability of key disaster prevention and mitigation targets during the flood season (i.e., existing risk and hazard points) to revise the early warning model established from historical disaster sites, thereby improving the accuracy and reliability of early warnings.

[0062] Meanwhile, the early warning model applied in this invention combines statistical analysis and machine learning algorithms, and is a spatiotemporally coupled prediction method. When new sample information is collected, it can be input into the model for parameter tuning and verification, achieving the effect of dynamic updating. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the method for establishing a shallow landslide disaster early warning model provided in the first embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of the ROC curve provided in an embodiment of the present invention;

[0065] Figure 3 This is a schematic diagram of susceptibility partitioning provided in an embodiment of the present invention;

[0066] Figure 4 This is a schematic diagram of the pre-rainfall-to-current-rainfall threshold model provided in an embodiment of the present invention;

[0067] Figure 5 This is a schematic diagram of the shallow collapse disaster early warning model establishment device provided in the second embodiment of the present invention. Detailed Implementation

[0068] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] Reference Figure 1 The first embodiment of the present invention provides a method for establishing a shallow landslide disaster early warning model, including the following steps:

[0070] S11, Obtain historical disaster data within the area to be warned, including the location of the disaster, the time of the disaster, and data from nearby rain gauges at the time of the disaster.

[0071] S12, based on the digital elevation model, divides slope units and extracts multiple disaster-prone factors based on ArcGIS;

[0072] S13, using the historical disaster point data and the disaster-preparing factors as the positive sample set, and selecting the same number of non-collapse points and disaster-preparing factors as the negative sample set, to form a total sample set and divide it into a training set, a validation set and a test set according to the proportion.

[0073] S14. Construct a collapse susceptibility prediction model based on gradient-enhanced decision trees and evaluate its accuracy to obtain susceptibility zoning evaluation results.

[0074] S15. Based on the susceptibility zoning evaluation results, statistically analyze the daily rainfall and previous rainfall at historical disaster sites in different susceptibility zones, determine the number of days of previous rainfall through correlation coefficients, and establish a threshold model of previous rainfall - daily rainfall.

[0075] S16. Based on whether the slope has been treated, vegetation conditions, and current deformation signs, the slope unit is classified into susceptibility levels, and the susceptibility level and rainfall threshold of the slope unit are corrected.

[0076] S17. Based on the susceptibility level and rainfall threshold of the slope unit, a multi-level spatiotemporal early warning model is constructed.

[0077] In one implementation, the step of dividing the slope units based on the digital elevation model includes:

[0078] Generate a depression-free area and calculate the flow direction and flow rate;

[0079] Generate a river network and connect the rivers, generate watersheds and convert the watersheds to facets, generate a positive catchment area, and invert the original DEM;

[0080] Repeat the steps of generating a river network and connecting rivers, generating a watershed and turning the watershed into a surface, generating a positive catchment area, and generating a negative catchment area.

[0081] The positive and negative water catchment areas are merged to generate an initial slope unit.

[0082] By correcting unreasonable boundaries, slope elements are obtained.

[0083] In one implementation, the extraction of multiple disaster-prone factors based on ArcGIS includes:

[0084] Use the Euclidean distance tool in ArcGIS to calculate the distances from roads, faults, and water systems to various slope factors;

[0085] Take the average distance of all grid cells as the final value of the slope cell;

[0086] Specifically, when extracting elevation, slope, aspect, elevation difference, plane curvature, profile curvature, TWI, and NDVI, the average value of all grid cells within a slope cell is used as the final value for each slope cell. When extracting stratigraphic lithology using slope cells, if a cell involves multiple lithologies, the lithology with the largest area is used as the final value for the slope factor.

[0087] It is worth noting that the various disaster-causing factors include: elevation, slope, aspect, elevation difference, plane curvature, profile curvature, topographic humidity index (TWI), normalized difference vegetation index (NDVI), lithology, distance to road, distance to fault, and distance to water system.

[0088] In one implementation, the step of using the historical disaster site data and the disaster-prone factors as a positive sample set, and selecting the same number of non-collapse points and disaster-prone factors as a negative sample set, to form a total sample set, which is then proportionally divided into a training set, a validation set, and a test set, includes:

[0089] The historical disaster data and the disaster-prone factors are used as the positive sample set;

[0090] Outside the buffer zone of the historical disaster point, the same number of non-collapse points are randomly selected to construct a negative sample set;

[0091] The total sample set is formed based on the positive sample set and the negative sample set;

[0092] The total sample set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1, which are used for model training, parameter tuning, and testing, respectively.

[0093] In one implementation, the step of constructing a collapse susceptibility prediction model based on a gradient-enhanced decision tree and evaluating its accuracy to obtain susceptibility zoning evaluation results includes:

[0094] Initialize the weak classifier:

[0095]

[0096] f0(x) is the initial weak classifier, L(y) i ,γ) is a differentiable loss function;

[0097] Calculate the negative gradient:

[0098]

[0099] In a regression tree, it is actually yc, where c refers to the average value of the sample labels of each leaf node.

[0100] Using the residual in formula (2) as the new value of the sample, (x) m ,r im ), m = 1, 2, 3, ..., M are used as training data for the next tree;

[0101] Based on the new values ​​of the samples, a new weak learner G is trained. k (X) is:

[0102]

[0103] Returning to the step of calculating the negative gradient, until the minimum error requirement is met, the final strong classifier is:

[0104]

[0105] In the formula, f m (x) is the collapse susceptibility prediction model, f0(x) is the initial model, and its value is the minimum value of the loss function; m is the maximum number of iterations, J is the number of leaf nodes of the j-th regression tree, and γ is the optimal fit value of the negative gradient of the loss function in the leaf node region, corresponding to the leaf node region R of the j-th regression tree. jm (j = 1, 2, 3, ..., J), where I is the indicator function and R is the expression. jm It is the leaf node region corresponding to the t-th decision tree; the maximum number of iterations (m) is set to 150, the learning rate is 0.5 when considering the data, and the number of leaf nodes (J) is set to 20.

[0106] In one implementation, the step of statistically analyzing the current day's rainfall and previous rainfall at historical disaster sites in different disaster-prone areas based on the susceptibility zoning evaluation results, determining the number of previous rainfall days through correlation coefficients, and establishing a previous rainfall-current day's rainfall threshold model includes:

[0107] The number of disaster-stricken points, the rainfall on the day the disaster occurred, and the rainfall in the previous nine days were counted separately in different susceptibility zones.

[0108] Correlation analysis was used to determine the number of days with previous rainfall and to establish a threshold model for previous rainfall-to-current rainfall for different susceptibility zones.

[0109] The formula for the correlation coefficient is as follows:

[0110]

[0111] In the formula, r(X,Y) represents the degree of correlation between the two variables; cov(X,Y) represents the covariance of X and Y; var(X) represents the variance of x; and var(Y) represents the variance of y.

[0112] To facilitate understanding of the present invention, some preferred embodiments of the present invention will be further described below. In one embodiment, the following steps are included:

[0113] Step 1: Collect information on historical disaster sites in the survey area, including the location and time of the disaster and the data from nearby rain gauges at the time of the disaster; the data from nearby rain gauges at the time of the disaster are statistically analyzed on a daily basis, collecting rainfall data for the day of the disaster and the nine days prior.

[0114] Step 2: Divide the slope into units based on the high-precision digital elevation model (DEM) and extract multiple disaster-prone factors based on ArcGIS.

[0115] The division of slope units follows these 10 steps: 1. Fill depressions; 2. Calculate flow direction and flow rate; 3. Generate the optimal river network by setting the flow rate threshold multiple times using a raster calculator; 4. Connect rivers; 5. Generate watersheds; 6. Convert watersheds to generate positive catchment areas; 7. Invert the original DEM; 8. Repeat steps 2-6 to generate negative catchment areas; 9. Merge positive and negative catchment areas to initially generate slope units; 10. Manually correct unreasonable boundaries.

[0116] Based on DEM calculations of topographic factors such as elevation, slope, aspect, elevation difference, plane curvature, profile curvature, and topographic humidity index (TWI), the Euclidean distance tool generates buffer maps of roads, faults, and water systems.

[0117] The average value of all raster cells within a slope cell is used as the final value for each slope cell. Elevation, slope, aspect, elevation difference, plane curvature, profile curvature, topographic humidity index (TWI), and distance to roads, faults, and water systems are extracted as disaster-inducing factors. The lithology with the largest area is used as the final value for slope factors.

[0118] Step 3: Using the collapse list and disaster-prone factors as positive sample points, select the same number of non-collapse points and disaster-prone factors as negative sample points to form the total sample data and divide it into training set, validation set and test set according to the proportion.

[0119] Using historical disaster sites as the origin, a 1km distance buffer is generated. Outside the buffer, the same number of non-disaster sites (negative samples) as the disaster sites (positive samples) are randomly selected to form a 1:1 total sample dataset. The total sample set is then randomly divided into a training set, a validation set, and a test set in an 8:1:1 ratio, ensuring that the number of positive and negative samples in each dataset is consistent.

[0120] Step 4: Construct a Gradient Boosting Decision Tree (GBDT) collapse susceptibility prediction model and evaluate its accuracy.

[0121] Input the training set and use the scikit-learn package in Python 3.7 to build a GBDT collapse susceptibility prediction model; input the validation set and use Grid Search for parameter tuning; input the test set to test the tuned model. For example, the AUC value is used to evaluate the model's predictive performance, and the results are as follows: Figure 2 As shown.

[0122] Step 5: Based on the susceptibility zoning assessment results, statistically analyze the daily rainfall and previous rainfall at historical disaster sites in different susceptibility zones, and determine the number of days with previous rainfall through correlation analysis.

[0123] Susceptibility zoning evaluation results are as follows Figure 3 As shown, the study area is divided into four levels: high, medium, low, and non-prone.

[0124] Statistical analysis was performed on the daily rainfall and previous rainfall at historical disaster sites in different disaster-prone areas. Correlation analysis was used to determine the number of days of previous rainfall (4 days in this example). A threshold model of previous rainfall versus daily rainfall was established. The results are shown in the table below. Figure 4 As shown.

[0125]

[0126] Step Six: Based on the slope unit area, and according to whether the slope has been treated, vegetation conditions, and current deformation signs, the units are divided into three levels: stable, relatively stable, and unstable. The susceptibility level and rainfall threshold are then adjusted accordingly. Points that have undergone engineering treatment are classified as stable; points that have not been treated, have sparse vegetation, have exposed surfaces, and show obvious signs of erosion are classified as unstable; and points that have not been treated but have good vegetation cover and have not yet shown obvious signs of instability are classified as relatively stable.

[0127] Furthermore, for points determined to be stable, their susceptibility level needs to be adjusted to low susceptibility; for points determined to be unstable, their susceptibility level needs to be adjusted to high susceptibility; and for points determined to be relatively stable, their susceptibility level needs to be adjusted to medium susceptibility. If the results are consistent with the original model's predictions, no adjustment is necessary.

[0128] The rainfall threshold was also recalculated using the previous rainfall-to-the-day rainfall threshold model, based on the revised high-risk areas.

[0129] Step 7: Based on the local geological disaster meteorological warning level classification standards, construct a multi-level spatiotemporal precision early warning method.

[0130] According to the "Guangdong Province Emergency Warning Information Release Management Measures", geological disaster meteorological risk warnings are divided into four levels from high to low: Level 1, Level 2, Level 3 and Level 4.

[0131] Level 1 (Red): The possibility of a disaster is very high (probability of disaster is greater than 80%), and the risk of geological disasters caused by meteorological factors is very high.

[0132] Level 2 (Orange): High probability of disaster (60%–80%), with a high risk of geological disasters caused by meteorological factors.

[0133] Level 3 (Yellow): The possibility of a disaster is relatively high (probability of disaster is greater than 40% to 60%), and the risk of geological disasters caused by meteorological factors is relatively high.

[0134] Level 4 (Blue): The possibility of a disaster is relatively small (probability of disaster is greater than 20% to 40%), and the risk of geological disasters caused by meteorological factors is low.

[0135] Finally, the multi-level spatiotemporal early warning model is shown in the table below:

[0136]

[0137] Furthermore, when the flood season approaches, new disaster site information can be collected to test the established model and can be used as a validation set input to adjust and correct the model parameters.

[0138] Compared with existing shallow landslide or collapse early warning models, the technical advantages of this invention are as follows:

[0139] 1) This invention implements regional early warning, not monitoring and early warning for single-point landslides or collapses.

[0140] 2) This invention starts from historical disaster sites and uses a combination of statistical analysis and machine learning. Using slope units as the smallest predictive unit, it considers the results of the latest round of risk and hazard point surveys to correct the susceptibility and thresholds. When the next flood season arrives, the monitored risk and hazard points can verify the model's accuracy, and the collected new disaster site sample information can be input into the model for correction. This is a method for establishing a precise rainfall threshold early warning model for shallow landslides based on "point-area-point" correction.

[0141] 3) This invention determines the current stability of the key disaster prevention and mitigation targets during the flood season (i.e., existing risk points) and corrects the early warning model established from historical disaster points, thereby improving the accuracy and reliability of the early warning.

[0142] 4) The early warning model method used in this invention combines local meteorological early warning release standards, which can ensure the synchronization of early warning information.

[0143] 5) The early warning model used in this invention combines statistical analysis and machine learning algorithms. It is a spatiotemporally coupled prediction method. When new sample information is collected, it can be input into the model for parameter tuning and verification, achieving the effect of dynamic updating.

[0144] Reference Figure 5 The second embodiment of the present invention provides a device for establishing a shallow landslide disaster early warning model, comprising:

[0145] The data acquisition module is used to acquire historical disaster data within the area to be warned. The historical disaster data includes the location of the disaster, the time of the disaster, and the data of nearby rain gauges at the time of the disaster.

[0146] The slope division module is used to divide slope units based on the digital elevation model and extract multiple disaster-prone factors based on ArcGIS.

[0147] The sample generation module is used to take the historical disaster point data and the disaster-predatory factors as the positive sample set, select the same number of non-collapse points and disaster-predatory factors as the negative sample set, form a total sample set, and divide it into training set, validation set and test set according to the proportion.

[0148] The model evaluation module is used to construct a collapse susceptibility prediction model based on gradient-enhanced decision trees and evaluate its accuracy, thereby obtaining susceptibility zoning evaluation results.

[0149] The rainfall threshold module is used to statistically analyze the daily rainfall and previous rainfall of historical disaster sites in different susceptibility zones based on the susceptibility zoning evaluation results, determine the number of previous rainfall days through correlation coefficients, and establish a previous rainfall-daily rainfall threshold model.

[0150] The grade correction module is used to classify the slope unit into a susceptibility grade based on whether the slope has been treated, vegetation conditions, and current deformation signs, and to correct the susceptibility grade and rainfall threshold of the slope unit.

[0151] The early warning model building module is used to construct a multi-level spatiotemporal early warning model based on the susceptibility level and rainfall threshold of the slope unit.

[0152] It should be noted that the shallow landslide disaster early warning model establishment device provided in this embodiment of the invention is used to execute all the process steps of the shallow landslide disaster early warning model establishment method in the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0153] This invention also provides a terminal device. The terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a shallow landslide disaster early warning model establishment program. When the processor executes the computer program, it implements the steps in the various shallow landslide disaster early warning model establishment method embodiments described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the model evaluation module.

[0154] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0155] The terminal device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of terminal devices and do not constitute a limitation on the terminal device. It may include more or fewer components than described above, or a combination of certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0156] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0157] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0158] Wherein, if the modules / units integrated in the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0159] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0160] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for establishing a shallow landslide disaster early warning model, characterized in that, include: Acquire historical disaster data within the area to be warned, including the location of the disaster, the time of the disaster, and data from nearby rain gauges at the time of the disaster. Slope units were divided based on digital elevation model, and multiple disaster-prone factors were extracted based on ArcGIS; Using the historical disaster data and the disaster-predating factors as the positive sample set, and selecting the same number of non-collapse points and disaster-predating factors as the negative sample set, a total sample set is formed and divided into a training set, a validation set and a test set according to the proportion. A collapse susceptibility prediction model was constructed based on gradient-enhanced decision trees, and its accuracy was evaluated. The susceptibility zoning evaluation results were obtained. Based on the susceptibility zoning evaluation results, the daily rainfall and previous rainfall of historical disaster sites in different susceptibility zones were statistically analyzed. The number of days of previous rainfall was determined by the correlation coefficient, and a threshold model of previous rainfall-daily rainfall was established. Based on whether the slope has been treated, vegetation conditions, and current signs of deformation, the slope units are classified into susceptibility levels, and the susceptibility levels and rainfall thresholds of the slope units are adjusted accordingly. Based on the susceptibility level and rainfall threshold of the slope unit, a multi-level spatiotemporal early warning model is constructed; The process of constructing a collapse susceptibility prediction model based on gradient-enhanced decision trees and evaluating its accuracy yields susceptibility zoning evaluation results, including: Initialize the weak classifier: （1） f 0 (x) As the initial weak classifier, L ( y i , Let be a differentiable loss function; Calculate the negative gradient: （2） In the regression tree, it is actually... y - c , c This refers to the average value of the sample labels for each leaf node; Using the residual in formula (2) as the new value of the sample, ( x m , r im ), m =1, 2, 3, …, M As training data for the next tree; Based on the new values ​​of the samples, a new weak learner is trained. G k ( X )for: （4） Returning to the step of calculating the negative gradient, until the minimum error requirement is met, the final strong classifier is: （5） In the formula, f m ( x () is a landslide susceptibility prediction model. f 0( x ) represents the initial model, and its value is the minimum value of the loss function; m The maximum number of iterations, J It is the first j The number of leaf nodes in the regression tree, where γ is the optimal fit of the loss function to the negative gradient in the leaf node region, corresponding to the th... j The leaf node region of the regression tree is R jm ( j =1, 2, 3, …, J ), I For indicator functions, R jm It is the first t The leaf node regions corresponding to each decision tree; the maximum number of iterations ( m The learning rate is set to 0.5 when considering the data, and the number of leaf nodes is (150 times). J Set it to 20; Based on the susceptibility zoning evaluation results, the method involves statistically analyzing the daily rainfall and previous rainfall at historical disaster sites within different susceptibility zones, determining the number of days of previous rainfall through correlation coefficients, and establishing a previous rainfall-daily rainfall threshold model, including: The number of disaster-stricken points, the rainfall on the day the disaster occurred, and the rainfall in the previous nine days were counted separately in different susceptibility zones. Correlation analysis was used to determine the number of days with previous rainfall and to establish a threshold model for previous rainfall and current rainfall for different susceptibility zones. The formula for the correlation coefficient is as follows: (6) In the formula, The degree of correlation between the two variables; Let X be the covariance of Y; Let x be the variance; Let y be the variance.

2. The method for establishing a shallow landslide disaster early warning model according to claim 1, characterized in that, The method of dividing slope units based on digital elevation model includes: Generate depression-free areas and calculate flow direction and flow rate; Generate a river network and connect the rivers, generate watersheds and convert the watersheds to facets, generate a positive catchment area, and invert the original DEM; Repeat the steps of generating a river network and connecting rivers, generating a watershed and turning the watershed into a surface, generating a positive catchment area, and generating a negative catchment area. The positive and negative water catchment areas are merged to generate an initial slope unit. By correcting unreasonable boundaries, slope elements are obtained.

3. The method for establishing a shallow landslide disaster early warning model according to claim 1, characterized in that, The extraction of multiple disaster risk factors based on ArcGIS includes: Use the Euclidean distance tool in ArcGIS to calculate the distances from roads, faults, and water systems to various slope factors; Take the average distance of all grid cells as the final value of the slope cell; Specifically, when extracting elevation, slope, aspect, elevation difference, plane curvature, profile curvature, TWI, and NDVI, the average value of all grid cells within a slope cell is used as the final value for each slope cell. When extracting stratigraphic lithology using slope cells, if a cell involves multiple lithologies, the lithology with the largest area is used as the final value for the slope factor.

4. The method for establishing a shallow landslide disaster early warning model according to claim 3, characterized in that, The aforementioned disaster-causing factors include: elevation, slope, aspect, elevation difference, planar curvature, profile curvature, topographic humidity index (TWI), normalized difference vegetation index (NDVI), lithology, distance to road, distance to fault, and distance to water system.

5. The method for establishing a shallow landslide disaster early warning model according to claim 1, characterized in that, The historical disaster site data and the disaster-prone factors are used as the positive sample set, and an equal number of non-collapse points and disaster-prone factors are selected as the negative sample set to form a total sample set, which is then divided into a training set, a validation set, and a test set according to a certain ratio. The historical disaster data and the disaster-prone factors are used as the positive sample set; Outside the buffer zone of the historical disaster point, the same number of non-collapse points are randomly selected to construct a negative sample set; The total sample set is formed based on the positive sample set and the negative sample set; The total sample set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1, which are used for model training, parameter tuning, and testing, respectively.

6. A device for establishing a shallow landslide disaster early warning model, used to implement the method for establishing a shallow landslide disaster early warning model as described in any one of claims 1 to 5, characterized in that, include: The data acquisition module is used to acquire historical disaster data within the area to be warned. The historical disaster data includes the location of the disaster, the time of the disaster, and the data of nearby rain gauges at the time of the disaster. The slope division module is used to divide slope units based on the digital elevation model and extract multiple disaster-prone factors based on ArcGIS. The sample generation module is used to take the historical disaster point data and the disaster-predatory factors as the positive sample set, select the same number of non-collapse points and disaster-predatory factors as the negative sample set, form a total sample set, and divide it into training set, validation set and test set according to the proportion. The model evaluation module is used to construct a collapse susceptibility prediction model based on gradient-enhanced decision trees and evaluate its accuracy, thereby obtaining susceptibility zoning evaluation results. The rainfall threshold module is used to statistically analyze the daily rainfall and previous rainfall of historical disaster sites in different susceptibility zones based on the susceptibility zoning evaluation results, determine the number of previous rainfall days through correlation coefficients, and establish a previous rainfall-current rainfall threshold model. The grade correction module is used to classify the slope unit into a susceptibility grade based on whether the slope has been treated, vegetation conditions, and current deformation signs, and to correct the susceptibility grade and rainfall threshold of the slope unit. The early warning model building module is used to construct a multi-level spatiotemporal early warning model based on the susceptibility level and rainfall threshold of the slope unit.

7. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the shallow landslide disaster early warning model establishment method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the shallow collapse disaster early warning model establishment method as described in any one of claims 1 to 5.

Images