A landslide dynamic early warning method and system based on space-time evolution

Abstract

This invention relates to the field of geological disaster monitoring technology, specifically to a landslide dynamic early warning method and system based on spatiotemporal evolution. The method includes: collecting target landslide area information using multi-source remote sensing data and integrating it to obtain a spatiotemporal monitoring dataset covering surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences; constructing a landslide state evolution field and coupling multiple monitoring sequences using an improved spatiotemporal fusion algorithm to generate a unified landslide evolution state tensor; building a landslide evolution dynamic graph model based on this tensor; extracting key evolution paths and nodes by combining data evolution characteristics; and conducting critical state assessments on key paths. When the assessment indicators exceed the early warning threshold, an early warning is triggered, outputting dynamic early warning information including high-risk landslide areas and estimated times. This scheme achieves integrated fusion analysis of multi-source monitoring data, completes dynamic analysis of the landslide spatiotemporal evolution process, improves the landslide critical state discrimination system, and realizes refined dynamic early warning of landslides.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring technology, and in particular to a landslide dynamic early warning method and system based on spatiotemporal evolution. Background Technology

[0002] Current landslide dynamic monitoring and early warning systems generally rely on single monitoring indicators for data analysis. Monitoring data such as surface deformation, soil pore pressure, and rainfall intensity are collected and processed independently. Multi-source monitoring data differ in temporal scale and spatial dimension, and there is a lack of standardized coupling processing mechanisms for various types of data. Conventional technical systems have not constructed state analysis fields adapted to the evolution patterns of landslides, and different types of geological and meteorological monitoring sequences cannot achieve synergistic correlation analysis. The decentralized data processing mode cannot fully reflect the continuous temporal variation patterns of landslides as a whole, and the intrinsic correlation characteristics of various landslide inducing factors are difficult to effectively explore.

[0003] Traditional landslide early warning models employ a fixed analytical framework, using a static assessment model to determine landslide risk, which cannot adapt to the spatiotemporal evolution characteristics of landslides over long time scales. Conventional modeling methods cannot rely on continuously changing monitoring data to trace the overall development of landslides, nor can they accurately extract key paths and core nodes in the geological evolution process; the assessment dimensions of landslide critical states are relatively singular. Static assessment models struggle to accurately delineate disaster risk coverage areas, cannot extrapolate the timeline of landslide evolution, and provide limited early warning output. To meet the practical application needs of landslide disaster prevention and control, it is necessary to optimize the fusion and processing of multi-source heterogeneous monitoring data, and build a dynamic analysis model adapted to spatiotemporal changes to compensate for the shortcomings of existing technologies in multi-factor collaborative analysis and dynamic critical state assessment. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a landslide dynamic early warning method and system based on spatiotemporal evolution.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a landslide dynamic early warning method based on spatiotemporal evolution, comprising:

[0006] Spatiotemporal monitoring datasets of the target landslide area are obtained based on multi-source remote sensing data. The spatiotemporal monitoring datasets include surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences.

[0007] A landslide state evolution field is constructed. In the landslide state evolution field, an improved spatiotemporal fusion algorithm is used to couple the surface deformation sequence, soil pore pressure sequence and rainfall intensity sequence to generate a unified landslide evolution state tensor.

[0008] Based on the landslide evolution state tensor, a landslide evolution dynamic graph model is constructed. In the landslide evolution dynamic graph model, key evolution paths and nodes are extracted according to the evolution characteristics of the landslide evolution state tensor.

[0009] Critical state assessments are performed on key evolution paths in the landslide evolution dynamics model. When the assessment indicators of key evolution paths exceed the warning threshold, a warning signal is triggered and dynamic warning information containing high-risk landslide areas and estimated times is generated.

[0010] As a further aspect of the present invention, the working principle of the improved spatiotemporal fusion algorithm is as follows:

[0011] The system receives surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences. The surface deformation sequence is the surface displacement of the target landslide area at multiple consecutive monitoring times. The soil pore pressure sequence is the soil pore water pressure value at multiple consecutive monitoring times. The rainfall intensity sequence is the rainfall data at multiple consecutive monitoring times.

[0012] We constructed corresponding physical quantity fields for the surface deformation sequence, soil pore pressure sequence, and rainfall intensity sequence, respectively, and mapped the data points of different physical quantities at the same time to the same geographical coordinates in the landslide state evolution field.

[0013] Within the landslide state evolution field, a dynamic coupling kernel function is defined. The dynamic coupling kernel function dynamically adjusts the weight coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence.

[0014] Using a dynamic coupling kernel function, the surface deformation value, soil pore pressure value and rainfall intensity value corresponding to the same geographical coordinate are weighted and fused at each monitoring time, and the fused state scalar of the geographical location at the monitoring time is output.

[0015] The fused state scalars of all monitoring times and all geographical coordinates are arranged in chronological order and spatial grid to form the landslide evolution state tensor.

[0016] As a further aspect of the present invention, the construction of a landslide evolution dynamics graphical model based on the landslide evolution state tensor includes:

[0017] From the landslide evolution state tensor, the fused state scalar of each spatial grid point at continuous monitoring time is extracted to form the state evolution trajectory of each spatial grid point;

[0018] Calculate the correlation degree between the state evolution trajectories of any two spatial grid points, identify the two spatial grid points whose correlation degree exceeds the connection threshold as associated nodes, and establish a weighted connection edge between the associated nodes. The weight value of the connection edge is the correlation degree value of the state evolution trajectories between the associated nodes.

[0019] By treating all spatial grid points as nodes and all connecting edges as edges, an initial network structure based on spatial grid points is formed.

[0020] In the initial network structure, the nodes whose fusion state scalar changes most drastically over time are identified and marked as high-activity nodes. The geographical locations corresponding to the spatial grid points of high-activity nodes are then marked as key evolution nodes.

[0021] In the initial network structure, find all connecting edges that connect highly active nodes, and mark the paths formed by these connecting edges as critical evolution paths.

[0022] As a further aspect of the present invention, the calculation of the correlation degree of the state evolution trajectory between any two spatial grid points includes:

[0023] For any two spatial grid points, obtain their respective scalar sequence of fused state over all monitoring periods;

[0024] Alignment and normalization are performed on the two scalar sequences of fused states to eliminate the difference between the dimensions and the initial values;

[0025] The dynamic time warping algorithm is used to calculate the minimum cumulative distance between two normalized fusion state scalar sequences.

[0026] The minimum cumulative distance is converted into a correlation value, where the correlation value is inversely proportional to the minimum cumulative distance.

[0027] As a further aspect of the present invention, the critical state assessment of key evolution paths in the landslide evolution dynamics graphical model includes:

[0028] For each critical evolution path, extract the weight value sequence of all connecting edges that constitute the critical evolution path;

[0029] Perform trend analysis on the weight value sequence and calculate the average rate of change and acceleration of change of the weight value sequence within the most recent monitoring window;

[0030] A comprehensive critical index is constructed based on the weight value sequence, average rate of change, and acceleration of change.

[0031] The comprehensive critical index is compared with a preset warning threshold. When the comprehensive critical index exceeds the warning threshold for a stable duration, the critical evolution path is determined to have entered a critical state.

[0032] The construction of a comprehensive critical index specifically includes:

[0033] Calculate the historical average weight value and the instantaneous weight value at the current moment from the weight value sequence;

[0034] Calculate the historical average rate of change and the instantaneous rate of change at the current moment from the average rate of change;

[0035] Calculate the ratio of the instantaneous weight value to the historical average weight value, and use it as the weight deviation component;

[0036] Calculate the ratio of the instantaneous rate of change to the historical average rate of change, and use it as the acceleration component of change;

[0037] The weighted deviation component, the acceleration component, and the acceleration of change are weighted and summed. The result of the weighted sum is then normalized, and the normalized value is the comprehensive critical index.

[0038] As a further aspect of the present invention, the determination that the critical evolution path has entered a critical state when the duration of the comprehensive critical index exceeding the warning threshold reaches a stable duration condition includes:

[0039] Record the starting moment when the comprehensive critical index first exceeds the warning threshold;

[0040] The comprehensive critical index is continuously monitored from the start time. If the comprehensive critical index remains above the warning threshold during the subsequent preset stable monitoring period, the stable duration condition is met.

[0041] When the stable duration condition is met, a critical state determination record is generated. The critical state determination record is associated with the key evolution path identifier that triggers the critical state and the determination time.

[0042] As a further aspect of the present invention, the triggering of the early warning signal and the generation of dynamic early warning information including high-risk landslide areas and estimated times include:

[0043] An early warning signal is triggered immediately upon generating a critical state determination record;

[0044] Based on the key evolution path identifiers associated in the critical state determination record, the key evolution path is located in the landslide evolution dynamics diagram model;

[0045] Extract all key evolution nodes connected by the key evolution path, and delineate the corresponding high-risk landslide areas in the geographic space of the target landslide area based on the spatial grid coordinates of the key evolution nodes.

[0046] Based on the magnitude and acceleration of the comprehensive critical index exceeding the warning threshold, the estimated time from the current moment to the estimated time of landslide occurrence is calculated through a preset time estimation relationship.

[0047] By integrating the geographical range and estimated time of high-risk landslide areas, dynamic early warning information is generated.

[0048] As a further aspect of the present invention, the dynamic coupling kernel function dynamically adjusts the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence, including:

[0049] Extract the soil pore pressure value at the current monitoring time and its change relative to the previous monitoring time from the soil pore pressure sequence;

[0050] A baseline weight allocation ratio is set, where the surface deformation sequence has baseline weight one and the rainfall intensity sequence has baseline weight two;

[0051] When the soil pore pressure value is higher than the saturation threshold, the weight coefficient of the rainfall intensity sequence is lowered based on the baseline weight two, while the weight coefficient of the surface deformation sequence is raised based on the baseline weight one.

[0052] When the change in soil pore pressure is positive and its absolute value is greater than the change threshold, the weighting coefficient of the rainfall intensity sequence will be increased based on the baseline weighting.

[0053] The weighting coefficients of the adjusted surface deformation sequence and the rainfall intensity sequence are normalized to ensure that the sum of the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence is a constant value. The normalized coefficients are the dynamic weighting coefficients used for the fusion calculation at the current time.

[0054] As a further aspect of the present invention, the method of utilizing a dynamic coupling kernel function to perform a weighted fusion calculation on the surface deformation value, soil pore pressure value, and rainfall intensity value corresponding to the same geographical location coordinates at each monitoring time, and outputting a fusion state scalar of the geographical location at the monitoring time, includes:

[0055] Obtain the surface deformation value, soil pore pressure value, and rainfall intensity value corresponding to the current monitoring time and current geographical location coordinates;

[0056] Using the dynamic coupling kernel function, the dynamic weighting coefficients for surface deformation and rainfall intensity at the current monitoring time are calculated based on the soil pore pressure value and its changing trend at the current monitoring time.

[0057] Soil pore pressure value is used as the basic fusion amount;

[0058] Multiply the surface deformation value and the rainfall intensity value by their corresponding dynamic weighting coefficients to obtain the weighted surface deformation value and the weighted rainfall value;

[0059] The basic fusion quantity, weighted surface deformation, and weighted rainfall are added together, and the sum is the fusion state scalar of the monitoring time and geographical coordinates.

[0060] As a further aspect of the present invention, the present invention also includes a landslide dynamic early warning system based on spatiotemporal evolution. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the landslide dynamic early warning method based on spatiotemporal evolution as described above.

[0061] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0062] A landslide state evolution field was constructed, and an improved spatiotemporal fusion algorithm was used to couple surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences to form a standardized landslide evolution state tensor. The spatiotemporal standards and storage structures of different monitoring data were unified, coordinating the acquisition rhythm and spatial coverage of various monitoring contents, breaking the fragmented processing mode of multi-source monitoring data. Different types of environmental monitoring parameters were integrated and summarized, unifying the data expression format, and completely preserving the continuous change information of each parameter. This approach takes into account the comprehensive changes in landslide deformation, soil environment, and meteorological conditions, mitigating the limitations of independent analysis of single-dimensional data, and objectively presenting the landslide changes under the combined influence of multiple factors.

[0063] A landslide evolution dynamics model is constructed based on the landslide evolution state tensor. Evolutionary paths and key nodes are selected by combining the data's own evolutionary patterns to conduct landslide critical state assessments. The development trajectory of landslides is traced by combining actual data change patterns, accurately locating important changes in the evolutionary process and changing the analytical approach of determining risk based on a single standard. State assessment is completed by combining key evolutionary paths, enriching the analytical perspectives of landslide states, accurately delineating the coverage of disaster risks, and reasonably judging the development pace of geological evolution. This model aligns with the actual attribute of landslides slowly changing in time and space, fully reconstructing the stage-by-stage transformation process of geological states, enriching the composition of early warning information, and meeting the actual analytical needs of dynamic monitoring of geological disasters. Attached Figure Description

[0064] Figure 1 This is a flowchart of a landslide dynamic early warning method based on spatiotemporal evolution as described in this invention;

[0065] Figure 2 A flowchart illustrating the work done to improve the spatiotemporal fusion algorithm;

[0066] Figure 3 A flowchart for constructing a dynamic model of landslide evolution. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0069] See Figure 1 This invention provides a landslide dynamic early warning method based on spatiotemporal evolution, the specific method including:

[0070] A spatiotemporal monitoring dataset containing surface deformation, soil pore pressure, and rainfall intensity sequences for the target landslide area was acquired based on multi-source remote sensing data. A landslide state evolution field was constructed, in which the three sequences were coupled using an improved spatiotemporal fusion algorithm to generate a unified landslide evolution state tensor. Based on this landslide evolution state tensor, a landslide evolution dynamic graph model was further constructed, and key evolution paths and nodes were extracted from the model according to the evolution characteristics of the tensor. Critical state assessments were performed on the key evolution paths in the graph model. When the assessment indicators exceeded the warning threshold, a warning signal was triggered, and dynamic warning information including high-risk landslide areas and estimated times was generated.

[0071] In one embodiment of the present invention, see [reference] Figure 2The improved spatiotemporal fusion algorithm receives surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences. The surface deformation sequence represents the surface displacement of the target landslide area at multiple consecutive monitoring times; the soil pore pressure sequence represents the soil pore water pressure values ​​at corresponding consecutive monitoring times; and the rainfall intensity sequence represents the rainfall data at the corresponding monitoring times. The algorithm constructs corresponding physical quantity fields for each of the surface deformation, soil pore pressure, and rainfall intensity sequences, mapping data points of different physical quantities at the same time to the same geographical coordinates in the landslide state evolution field. Within the landslide state evolution field, a dynamic coupling kernel function is defined. This function dynamically adjusts the weighting coefficients of the surface deformation and rainfall intensity sequences during the fusion process based on the magnitude and trend of the soil pore pressure sequence. Using this dynamic coupling kernel function, at each monitoring time, a weighted fusion calculation is performed on the surface deformation value, soil pore pressure value, and rainfall intensity value corresponding to the same geographical coordinate, outputting the fusion state scalar for that geographical location at that monitoring time. The fused state scalars of all monitoring times and all geographical coordinates are arranged in chronological order and spatial grid to form the landslide evolution state tensor.

[0072] In practical implementation, the improved spatiotemporal fusion algorithm receives surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences. The surface deformation sequence represents the surface displacement of the target landslide area at multiple consecutive monitoring times. The soil pore pressure sequence represents the soil pore water pressure values ​​at multiple consecutive monitoring times, and the rainfall intensity sequence represents the rainfall data at multiple consecutive monitoring times. Taking a specific landslide monitoring area as an example, the surface deformation sequence is acquired using synthetic aperture radar interferometry, the soil pore pressure sequence is collected by a network of pore water pressure sensors deployed within the landslide body, and the rainfall intensity sequence is observed by meteorological radar or rain gauges. The three sequences are temporally aligned and spatially cover the same geographical area. In practical implementation, corresponding physical quantity fields are constructed for the surface deformation sequence, soil pore pressure sequence, and rainfall intensity sequence, respectively. Each physical quantity field maps the values ​​of each monitoring time in the sequence to preset spatial grid points in the landslide state evolution field, so that data points of different physical quantities at the same time are mapped to the same geographical coordinates in the landslide state evolution field, thereby achieving data alignment in time and space. In some embodiments, the landslide state evolution field is defined as a three-dimensional coordinate space covering the target landslide area, where two dimensions represent geographic plane coordinates, one dimension represents time coordinates, and surface deformation value, soil pore pressure value, and rainfall intensity value are used as scalar functions in the field.

[0073] A dynamic coupling kernel function is defined within the landslide state evolution field. This function dynamically adjusts the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence. The input to the dynamic coupling kernel function can be understood as the soil pore pressure value at the current monitoring time and geographical location. and its change relative to the previous monitoring time The output consists of two dynamic weighting coefficients for the surface deformation sequence and the rainfall intensity sequence. and .in, This represents the soil pore pressure value. This represents the change in soil pore pressure. The dynamic weighting coefficients representing the land surface deformation sequence. This represents the dynamic weighting coefficients of the rainfall intensity sequence. The operational logic of the dynamic coupling kernel function includes: reducing the weighting coefficients of the rainfall intensity sequence when the soil pore pressure value exceeds a preset saturation threshold. And increase the weighting coefficient of the surface deformation sequence. When the soil pore pressure changes When the value is positive and its absolute value is greater than the change threshold, increase the weighting coefficient of the rainfall intensity sequence. In practical implementation, the dynamic coupling kernel function calculates the final weights through the following steps: setting a baseline weight allocation ratio, where the surface deformation sequence has a baseline weight... The rainfall intensity sequence has a baseline weight of 2 Based on soil pore pressure values Comparison with saturation threshold, and changes in soil pore pressure The comparison results with the change threshold are used to determine the baseline weight. With benchmark weight two Perform linear adjustment to obtain the adjusted weights. and ;right and Normalization is performed to make Normalized and This refers to the dynamic weighting coefficient.

[0074] Using a dynamic coupling kernel function, a weighted fusion calculation is performed on the surface deformation, soil pore pressure, and rainfall intensity values ​​corresponding to the same geographical location at each monitoring time, outputting a scalar fusion state of the geographical location at the monitoring time. In some embodiments, for the monitoring time... and geographic coordinates Obtain surface deformation values Soil pore pressure value With rainfall intensity value The dynamic weighting coefficients are calculated using a dynamic coupling kernel function. and Then the fusion state scalar It can be calculated using the following formula:

[0075]

[0076] in: Represents the scalar of the fusion state. This represents the soil pore pressure value. Represents the value of surface deformation. Indicates the rainfall intensity value. and These are dynamic weighting coefficients. It can be understood that the weighted fusion calculation uses soil pore pressure as the basic fusion quantity, and then superimposes surface deformation and rainfall intensity values ​​based on their dynamic weighting coefficients, thereby reflecting the coupling effect of multiple factors. In practice, this calculation traverses all geographical coordinates and all monitoring times in the landslide state evolution field, generating a fusion state scalar for each coordinate point at each time point.

[0077] The fused state scalars of all monitoring times and all geographical coordinates are arranged chronologically and spatially to form a landslide evolution state tensor. In practice, the landslide evolution state tensor is a three-dimensional array, where two dimensions correspond to the row and column indices of the spatial grid, and the third dimension corresponds to the monitoring time index. Each array element stores the fused state scalar value at the corresponding location and time. Optionally, the resolution of the spatial grid is consistent with the original spatial resolution of the multi-source remote sensing data, and the monitoring time sequence in the temporal dimension is consistent with the acquisition frequency of the input sequence. Through this arrangement, the landslide evolution state tensor fully expresses the spatiotemporal evolution of the target landslide area's fused state.

[0078] In one embodiment of the present invention, the process of constructing a landslide evolution dynamics graphical model based on the landslide evolution state tensor is described in [reference needed]. Figure 3The fused state scalar of each spatial grid point at continuous monitoring time is extracted from the landslide evolution state tensor to form the state evolution trajectory of each spatial grid point. The correlation degree between the state evolution trajectories of any two spatial grid points is calculated. Two spatial grid points with a correlation degree exceeding a connection threshold are identified as associated nodes, and a weighted connection edge is established between the associated nodes. The weight value of the connection edge is the correlation degree value. All spatial grid points are used as nodes, and all connection edges are used as edges, forming an initial network structure based on spatial grid points. In this initial network structure, the node whose fused state scalar changes most drastically in the time dimension is identified and marked as a high-activity node. The geographical locations corresponding to the spatial grid points of the high-activity nodes are marked as key evolution nodes. In the initial network structure, all connection edges connecting the high-activity nodes are found, and the paths formed by these connection edges are marked as key evolution paths. The specific method for calculating the correlation degree of the state evolution trajectory between any two spatial grid points is as follows: for any two spatial grid points, obtain their respective fused state scalar sequences over all monitoring periods; align and normalize the two fused state scalar sequences to eliminate the difference between the dimensions and the initial values; use the dynamic time warping algorithm to calculate the minimum cumulative distance between the two normalized fused state scalar sequences; and then convert this minimum cumulative distance into a correlation degree value, which is inversely proportional to the minimum cumulative distance.

[0079] In practical implementation, a landslide evolution dynamics graph model is constructed based on the landslide evolution state tensor. The landslide evolution state tensor is a three-dimensional data structure containing fused state scalars of all monitoring times and all geographical coordinates. The fused state scalars of each spatial grid point at consecutive monitoring times are extracted from the landslide evolution state tensor to form the state evolution trajectory of each spatial grid point. For example, for a certain landslide monitoring area, the landslide evolution state tensor contains 100 spatial grid points and 50 consecutive monitoring times, resulting in 100 state evolution trajectories of length 50. The correlation degree between the state evolution trajectories of any two spatial grid points is calculated. For spatial grid point A and spatial grid point B, the fused state scalar sequence of spatial grid point A over all 50 monitoring periods is obtained. Simultaneously, obtain the scalar sequence of the fused state of spatial grid point B. In some embodiments, the fused state scalar sequence of spatial grid point A and the fused state scalar sequence of spatial grid point B are aligned and normalized to eliminate differences caused by different dimensions and initial values. The normalization process maps the values ​​of each sequence to the range of zero to one. A dynamic time warping algorithm is used to calculate the minimum cumulative distance between the two normalized fused state scalar sequences. .

[0080] It can be understood that converting the minimum cumulative distance into a correlation value, the correlation value is inversely proportional to the minimum cumulative distance. It can be calculated using the following formula:

[0081]

[0082] in: This represents the correlation value between spatial grid point A and spatial grid point B. This represents the minimum cumulative distance calculated using the dynamic time warping algorithm. It is a scaling factor greater than zero used to adjust the sensitivity of the correlation degree to distance. In specific implementation, the calculated correlation degree value will be... The correlation value is compared with a preset connection threshold. If the connection threshold is exceeded, spatial grid point A and spatial grid point B are identified as associated nodes, and a weighted connection edge is established between the associated nodes. The weight value of this connection edge is the association degree value. The process involves iterating through all combinations of spatial grid points in the landslide evolution state tensor, repeating the above calculation and comparison process to establish connections between all spatial grid points. Using all spatial grid points as nodes and all connections established under the given conditions as edges, an initial network structure based on these spatial grid points is constructed. This initial network structure forms the underlying topology of the landslide evolution dynamics graph model.

[0083] In the initial network structure, the nodes whose fused state scalar changes most drastically over time are identified. These nodes are determined by calculating the variance or standard deviation of the state evolution trajectory of each node, and the nodes with the largest variance or standard deviation are marked as highly active nodes. In some embodiments, the number of highly active nodes can be proportionally set according to the network size, and the geographical locations corresponding to the spatial grid points of highly active nodes are marked as key evolutionary nodes. In the initial network structure, all connecting edges linking highly active nodes are found, and the paths formed by these connecting edges are marked as key evolutionary paths. Key evolutionary paths describe the main state propagation or influence channels between highly active nodes. Optionally, the extraction of key evolutionary paths can be accomplished using a graph search algorithm, for example, starting from any highly active node, finding all connecting edges with highly active nodes as the starting or ending point, and using the continuous sequence of these edges as key evolutionary paths. The landslide evolution dynamics graph model is ultimately defined by nodes, edges, key evolutionary nodes, and key evolutionary paths.

[0084] In one embodiment of the present invention, a critical state assessment is performed on key evolution paths in a landslide evolution dynamics graph model. For each key evolution path, a sequence of weight values ​​for all connecting edges constituting that path is extracted. Trend analysis is performed on this weight value sequence, calculating the average rate of change and acceleration within the most recent monitoring window. A comprehensive critical index is constructed based on the weight value sequence, average rate of change, and acceleration. This construction process includes calculating the historical average weight value and the current instantaneous weight value from the weight value sequence; calculating the historical average rate of change and the current instantaneous rate of change from the average rate of change; calculating the ratio of the instantaneous weight value to the historical average weight value as the weight deviation component; calculating the ratio of the instantaneous rate of change to the historical average rate of change as the acceleration component; and weighted summing the weight deviation component, acceleration component, and acceleration, then normalizing the result. The normalized value is the comprehensive critical index. The comprehensive critical index is compared with a preset warning threshold. When the comprehensive critical index continuously exceeds the warning threshold for a stable duration, the key evolution path is determined to have entered a critical state. The determination process records the starting moment when the comprehensive critical index first exceeds the warning threshold. Starting from the starting moment, continuous monitoring is performed. If the comprehensive critical index remains above the warning threshold during the subsequent preset stable monitoring period, the stable duration condition is met. When the stable duration condition is met, a critical state determination record is generated. This record is associated with the key evolution path identifier that triggers the critical state and the determination time.

[0085] In practical implementation, critical state assessments are performed on key evolution paths in the landslide evolution dynamics model. For each key evolution path extracted from the model, the weight value sequence of all connecting edges constituting that path is extracted. This sequence records the correlation value of each edge at multiple consecutive monitoring times, forming a time series data set. Trend analysis is performed on the weight value sequence, calculating the average rate of change and acceleration of change within the most recent monitoring window. The average rate of change reflects the average increase or decrease trend of the weight values ​​per unit time, while the acceleration reflects the degree to which the trend of weight value change accelerates or slows down. Based on the weight value sequence, average rate of change, and acceleration, a comprehensive critical index is constructed. The construction of the comprehensive critical index specifically includes the following calculation steps: calculating the historical average weight value from the weight value sequence. and the instantaneous weight value at the current moment. ; Calculate the historical average rate of change from the calculated average rate of change series. and the instantaneous rate of change at the current moment. ; Calculate the ratio of the instantaneous weight value to the historical average weight value This ratio is used as the weight deviation component. ; Calculate the ratio of the instantaneous rate of change to the historical average rate of change. This ratio is used as the acceleration component of change. ; for weight deviation components Acceleration component of change and changing acceleration Perform a weighted summation, and the result of the weighted summation is: ,in These are the preset weighting coefficients; the result of the weighted summation is then normalized, and the normalized value is the comprehensive critical index. Comprehensive critical index The calculation formula can be expressed as:

[0086]

[0087] in: This represents the comprehensive critical index. This represents the instantaneous weight value at the current moment. This represents the historical average weight value. Represents the instantaneous rate of change at the current moment. Indicates the historical average rate of change. Indicates the change in acceleration. These are preset weighting coefficients. This represents a normalization function that maps input values ​​to a fixed range of 0 to 1 or 0 to 100. In some embodiments, the historical average weight value... and historical average rate of change It is the arithmetic mean calculated based on a data window of a fixed time length prior to the current moment.

[0088] Table 1: Weight value sequence fragments of the critical evolutionary path P001 Monitoring time Weight of edge E01 Weight of edge E02 Weight of edge E03 Path weight sequence values ​​(mean) t-4 0.85 0.88 0.82 0.85 t-3 0.87 0.90 0.84 0.87 t-2 0.90 0.92 0.88 0.90 t-1 0.94 0.95 0.92 0.94 t (current) 0.98 0.99 0.96 0.98

[0089] In practical implementation, the weight value sequence extracted for the critical evolution path is the average of the weight values ​​of all connecting edges constituting the path at the same time. Referring to Table 1, the critical evolution path P001 is composed of connecting edges E01, E02, and E03. The weight sequence value of the path at each monitoring time is calculated from the arithmetic mean of the weights of these three edges. Based on this path weight sequence value, trend analysis is performed and a comprehensive critical index is calculated. The process involves comparing the calculated comprehensive critical index with a preset warning threshold. When the comprehensive critical index consistently exceeds the warning threshold for a stable duration, the critical evolution path is determined to have entered a critical state. The determination process records the initial moment when the comprehensive critical index first exceeds the warning threshold. From this initial moment, the comprehensive critical index is continuously monitored. If, during the subsequent preset stable monitoring period, the comprehensive critical index remains above the warning threshold, the stable duration condition is met. Optionally, the stable monitoring period can be set based on the specific geological characteristics of the landslide and the monitoring frequency. When the stable duration condition is met, a critical state determination record is generated. This record associates the critical evolution path identifier that triggered the critical state with the determination time. The critical evolution path identifier is a unique number or code that identifies the path, and the determination time is the moment when the stable duration condition is met and the path is formally determined to have entered a critical state. In some embodiments, the critical state determination record can be stored as a data structure containing information such as the path identifier, determination time, and comprehensive critical index value.

[0090] In one embodiment of the present invention, an early warning signal is triggered and dynamic early warning information containing high-risk landslide areas and estimated times is generated. The early warning signal is triggered immediately upon generating a critical state determination record. Based on the key evolution path identifier associated with the critical state determination record, the key evolution path is located in the landslide evolution dynamics model. All key evolution nodes connected by the key evolution path are extracted, and based on the spatial grid coordinates of these key evolution nodes, the corresponding high-risk landslide area is delineated in the geographic space of the target landslide area. Based on the magnitude and acceleration of the comprehensive critical index exceeding the early warning threshold, the estimated time from the current moment to the estimated landslide occurrence time is calculated using a preset time estimation relationship. The geographic range information of the high-risk landslide area and the estimated time are integrated to form dynamic early warning information.

[0091] In specific implementation, an early warning signal is triggered, generating dynamic early warning information containing high-risk landslide areas and estimated times. The early warning signal is triggered immediately upon generating a critical state determination record, which includes the identifier of the key evolutionary path associated with triggering the critical state and the determination time. Based on the key evolutionary path identifier associated in the critical state determination record, the key evolutionary path is located in the landslide evolution dynamics graph model. The key evolutionary path identifier is a unique code pointing to a specific path in the landslide evolution dynamics graph model. All key evolutionary nodes connected by the key evolutionary paths are extracted. Key evolutionary nodes are spatial grid points corresponding to highly active nodes marked during the construction of the landslide evolution dynamics graph model. Based on the coordinates of the spatial grid points of the key evolutionary nodes, the corresponding high-risk landslide area is delineated in the geographic space of the target landslide area. In some embodiments, the mapping from spatial grid point coordinates to actual geographic coordinates is achieved through pre-stored coordinate transformation parameters. The operation of delineating the high-risk landslide area can be achieved by connecting all the actual geographic coordinate points corresponding to the key evolutionary nodes into polygons to form a closed region.

[0092] Based on the magnitude and acceleration of the comprehensive critical index exceeding the warning threshold, the estimated time from the current moment to the predicted landslide occurrence is calculated using a pre-defined time estimation relationship. This time estimation relationship is a mathematical relationship built upon historical data or a physical model, with the input being the magnitude of the comprehensive critical index exceeding the warning threshold. With changing acceleration The output is the estimated time. Estimated time The calculation formula can be expressed as:

[0093]

[0094] in: This indicates the estimated time from the current moment to the predicted time of the landslide. This indicates the extent to which the comprehensive critical index exceeds the warning threshold (i.e., the difference between the current value of the comprehensive critical index and the warning threshold). This represents the change in acceleration calculated at the time of determination. and It is a positive coefficient calibrated based on historical landslide case data. It integrates the geographical extent information and estimated time of high-risk landslide areas to form dynamic early warning information. Dynamic early warning information is a structured data object containing key evolutionary path identifiers that trigger the warning, the determination time, the set of geographical boundary coordinates of high-risk landslide areas, and the estimated time. Optionally, dynamic early warning information can be further encoded into a specific format for network transmission or visualization.

[0095] Table 2: Information on key evolutionary nodes associated with key evolutionary path P001 Key Evolution Node ID Spatial grid point coordinates (row, column) Actual geographic coordinates (longitude, latitude) KN-101 (15,32) (110.251°E, 30.715°N) KN-102 (15,33) (110.252°E, 30.715°N) KN-107 (16,33) (110.252°E, 30.714°N) KN-108 (16,34) (110.253°E, 30.714°N)

[0096] In practical implementation, taking the "Heifangshan landslide" monitoring scenario as an example, when the critical evolution path marked "P001" is determined to have entered a critical state, a critical state determination record is generated. The critical evolution path marked "P001" is located in the landslide evolution dynamics model, and the four critical evolution nodes connected by the "P001" path are extracted. Node information is shown in Table 2. The latitude and longitude coordinates provided in the "Actual Geographic Coordinates" column are used to sequentially connect these coordinate points on a geographic information system or map; the resulting polygonal area is the high-risk landslide area for this warning. Simultaneously, the extent to which the currently calculated comprehensive critical index exceeds the warning threshold is also considered. and changing acceleration The estimated time is calculated using a preset time estimation relationship. The estimated time is 12 hours. In some embodiments, the final generated dynamic early warning information will include the path identifier "P001", the judgment time "2025-07-2014:00:00", the set of coordinates of high-risk areas (i.e., the sequence of all coordinate points in Table 2), and the estimated time "12 hours". It can be understood that the generation of dynamic early warning information marks the end of the complete process from state analysis to early warning output, providing a clear spatiotemporal orientation for disaster response.

[0097] In one embodiment of the present invention, the dynamic coupling kernel function dynamically adjusts the weighting coefficients based on the magnitude and trend of the soil pore pressure sequence, extracting the soil pore pressure value at the current monitoring time and its change relative to the previous monitoring time from the soil pore pressure sequence. A baseline weighting ratio is set, where the surface deformation sequence has a baseline weight one, and the rainfall intensity sequence has a baseline weight two. When the soil pore pressure value is higher than a saturation threshold, the weighting coefficient of the rainfall intensity sequence is lowered based on baseline weight two, while the weighting coefficient of the surface deformation sequence is raised based on baseline weight one. When the change in soil pore pressure is positive and its absolute value is greater than the change threshold, the weighting coefficient of the rainfall intensity sequence is raised based on baseline weight two. The adjusted weighting coefficients of the surface deformation sequence and the rainfall intensity sequence are normalized to ensure that their sum is a constant value. The normalized coefficients are the dynamic weighting coefficients used for the fusion calculation at the current time. A weighted fusion calculation using a dynamic coupling kernel function outputs a fusion state scalar, obtaining the surface deformation, soil pore pressure, and rainfall intensity values ​​corresponding to the current monitoring time and geographical coordinates. The dynamic coupling kernel function calculates the dynamic weighting coefficients for the surface deformation and rainfall intensity values ​​at the current monitoring time based on the soil pore pressure value and its changing trend. Using the soil pore pressure value as the basic fusion quantity, the surface deformation and rainfall intensity values ​​are multiplied by their corresponding dynamic weighting coefficients to obtain the weighted surface deformation and weighted rainfall, respectively. The basic fusion quantity, weighted surface deformation, and weighted rainfall are then summed, and the result is the fusion state scalar for that monitoring time and geographical coordinates.

[0098] In practical implementation, the dynamic coupling kernel function dynamically adjusts the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence. It extracts the soil pore pressure value at the current monitoring moment and its change relative to the previous monitoring moment from the soil pore pressure sequence. The soil pore pressure value is the pore water pressure measurement value collected by the sensor in real time, and the change is the difference between the soil pore pressure value at the current moment and the soil pore pressure value at the previous moment. A baseline weighting ratio is set, where the surface deformation sequence has a baseline weight one, and the rainfall intensity sequence has a baseline weight two. The initial values ​​of baseline weight one and baseline weight two can be preset based on historical data or expert knowledge. The dynamic coupling kernel function adjusts the baseline weights according to the soil pore pressure value and its change. When the soil pore pressure value is higher than the saturation threshold, the weighting coefficient of the rainfall intensity sequence is lowered based on baseline weight two, while the weighting coefficient of the surface deformation sequence is increased based on baseline weight one. This adjustment reflects the physical mechanism that the importance of deformation factors relatively increases when the soil approaches saturation. When the change in soil pore pressure is positive and its absolute value exceeds the change threshold, the weighting coefficient of the rainfall intensity sequence is increased based on the baseline weighting of two. This reflects the potential increase in the contribution of rainfall infiltration when pore water pressure increases rapidly. It can be understood that the adjusted weighting coefficients of the surface deformation sequence and the rainfall intensity sequence are normalized to ensure that the sum of their weighting coefficients is a constant value. The normalized coefficients are the dynamic weighting coefficients used for the fusion calculation at the current moment. In some embodiments, the dynamic weighting coefficients are output by the dynamic coupling kernel function. and Determined by the following relation:

[0099]

[0100]

[0101] in: The dynamic weighting coefficients representing the land surface deformation sequence. Represents the dynamic weighting coefficients of the rainfall intensity sequence. The baseline weights representing the surface deformation sequence are: The second benchmark weight represents the rainfall intensity sequence. Indicates the weight of the benchmark. The adjustment amount, Indicates the weight of the benchmark two Adjustment amount. Adjustment amount and The specific value is determined by two conditions: whether the soil pore pressure value is higher than the saturation threshold, and whether the change in soil pore pressure is greater than the change threshold.

[0102] Using a dynamic coupling kernel function, a weighted fusion calculation is performed on the surface deformation, soil pore pressure, and rainfall intensity values ​​corresponding to the same geographical coordinates at each monitoring time, outputting a scalar of the fused geographical location at that monitoring time. In practice, the surface deformation, soil pore pressure, and rainfall intensity values ​​corresponding to the current monitoring time and geographical coordinates are obtained; these values ​​are derived from multi-source remote sensing data and sensor networks, respectively. Using the dynamic coupling kernel function, based on the soil pore pressure value and its changing trend at the current monitoring time, the dynamic weighting coefficients for the surface deformation and rainfall intensity values ​​at the current time are calculated. and It is understandable that soil pore pressure is used as the base fusion value, which directly participates in the summation calculation without being multiplied by a weighting coefficient. The surface deformation value and rainfall intensity value are then multiplied by their respective dynamic weighting coefficients. and The weighted surface deformation and weighted rainfall are obtained. The basic fusion quantity, weighted surface deformation, and weighted rainfall are added together, and the sum is the fusion state scalar of the monitoring time and geographical coordinates. The calculation process of the fusion state scalar is a linear weighted summation, and its physical meaning lies in comprehensively reflecting a dimensionless state index of the current spatiotemporal location under the coupled effects of pore pressure, surface deformation, and rainfall. Optionally, the calculation of the dynamic coupling kernel function and the weighted fusion calculation are performed independently and in parallel at each spatial grid point within the landslide state evolution field, generating a landslide evolution state tensor covering the entire monitoring area and all monitoring periods. In some embodiments, the saturation threshold and variation threshold of the soil pore pressure value can be pre-calibrated based on the hydraulic characteristic parameters of the soil type in the landslide area to ensure the scientific nature of the dynamic weight adjustment.

[0103] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A landslide dynamic early warning method based on spatiotemporal evolution, characterized in that, include: Spatiotemporal monitoring datasets of the target landslide area are obtained based on multi-source remote sensing data. The spatiotemporal monitoring datasets include surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences. A landslide state evolution field is constructed. In the landslide state evolution field, an improved spatiotemporal fusion algorithm is used to couple the surface deformation sequence, soil pore pressure sequence and rainfall intensity sequence to generate a unified landslide evolution state tensor. Based on the landslide evolution state tensor, a landslide evolution dynamic graph model is constructed. In the landslide evolution dynamic graph model, key evolution paths and nodes are extracted according to the evolution characteristics of the landslide evolution state tensor. Critical state assessments are performed on key evolution paths in the landslide evolution dynamics model. When the assessment indicators of key evolution paths exceed the warning threshold, a warning signal is triggered and dynamic warning information containing high-risk landslide areas and estimated times is generated.

2. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 1, characterized in that, The working principle of the improved spatiotemporal fusion algorithm is as follows: The system receives surface deformation sequences, soil pore pressure sequences, and rainfall intensity sequences. The surface deformation sequence is the surface displacement of the target landslide area at multiple consecutive monitoring times. The soil pore pressure sequence is the soil pore water pressure value at multiple consecutive monitoring times. The rainfall intensity sequence is the rainfall data at multiple consecutive monitoring times. We constructed corresponding physical quantity fields for the surface deformation sequence, soil pore pressure sequence, and rainfall intensity sequence, respectively, and mapped the data points of different physical quantities at the same time to the same geographical coordinates in the landslide state evolution field. Within the landslide state evolution field, a dynamic coupling kernel function is defined. The dynamic coupling kernel function dynamically adjusts the weight coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence. Using a dynamic coupling kernel function, the surface deformation value, soil pore pressure value and rainfall intensity value corresponding to the same geographical coordinate are weighted and fused at each monitoring time, and the fused state scalar of the geographical location at the monitoring time is output. The fused state scalars of all monitoring times and all geographical coordinates are arranged in chronological order and spatial grid to form the landslide evolution state tensor.

3. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 2, characterized in that, The construction of a landslide evolution dynamics graph model based on the landslide evolution state tensor includes: From the landslide evolution state tensor, the fused state scalar of each spatial grid point at continuous monitoring time is extracted to form the state evolution trajectory of each spatial grid point; Calculate the correlation degree between the state evolution trajectories of any two spatial grid points, identify the two spatial grid points whose correlation degree exceeds the connection threshold as associated nodes, and establish a weighted connection edge between the associated nodes. The weight value of the connection edge is the correlation degree value of the state evolution trajectories between the associated nodes. By treating all spatial grid points as nodes and all connecting edges as edges, an initial network structure based on spatial grid points is formed. In the initial network structure, the nodes whose fusion state scalar changes most drastically over time are identified and marked as high-activity nodes. The geographical locations corresponding to the spatial grid points of high-activity nodes are then marked as key evolution nodes. In the initial network structure, find all connecting edges that connect highly active nodes, and mark the paths formed by these connecting edges as critical evolution paths.

4. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 3, characterized in that, The calculation of the correlation degree between the state evolution trajectories of any two spatial grid points includes: For any two spatial grid points, obtain their respective scalar sequence of fused state over all monitoring periods; Alignment and normalization are performed on the two scalar sequences of fused states to eliminate the difference between the dimensions and the initial values; The dynamic time warping algorithm is used to calculate the minimum cumulative distance between two normalized fusion state scalar sequences. The minimum cumulative distance is converted into a correlation value, where the correlation value is inversely proportional to the minimum cumulative distance.

5. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 3, characterized in that, The critical state assessment of key evolution paths in the landslide evolution dynamics model includes: For each critical evolution path, extract the weight value sequence of all connecting edges that constitute the critical evolution path; Perform trend analysis on the weight value sequence and calculate the average rate of change and acceleration of change of the weight value sequence within the most recent monitoring window; A comprehensive critical index is constructed based on the weight value sequence, average rate of change, and acceleration of change. The comprehensive critical index is compared with a preset warning threshold. When the comprehensive critical index exceeds the warning threshold for a stable duration, the critical evolution path is determined to have entered a critical state. The construction of a comprehensive critical index specifically includes: Calculate the historical average weight value and the instantaneous weight value at the current moment from the weight value sequence; Calculate the historical average rate of change and the instantaneous rate of change at the current moment from the average rate of change; Calculate the ratio of the instantaneous weight value to the historical average weight value, and use it as the weight deviation component; Calculate the ratio of the instantaneous rate of change to the historical average rate of change, and use it as the acceleration component of change; The weighted deviation component, the acceleration component, and the acceleration of change are weighted and summed. The result of the weighted sum is then normalized, and the normalized value is the comprehensive critical index.

6. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 5, characterized in that, When the duration for which the comprehensive critical index continuously exceeds the warning threshold reaches a stable duration, the critical evolution path is determined to have entered a critical state, including: Record the starting moment when the comprehensive critical index first exceeds the warning threshold; The comprehensive critical index is continuously monitored from the start time. If the comprehensive critical index remains above the warning threshold during the subsequent preset stable monitoring period, the stable duration condition is met. When the stable duration condition is met, a critical state determination record is generated. The critical state determination record is associated with the key evolution path identifier that triggers the critical state and the determination time.

7. The landslide dynamic early warning method based on spatiotemporal evolution according to claim 6, characterized in that, The triggering of the early warning signal and the generation of dynamic early warning information containing high-risk landslide areas and estimated times include: An early warning signal is triggered immediately upon generating a critical state determination record; Based on the key evolution path identifiers associated in the critical state determination record, the key evolution path is located in the landslide evolution dynamics diagram model; Extract all key evolution nodes connected by the key evolution path, and delineate the corresponding high-risk landslide areas in the geographic space of the target landslide area based on the spatial grid coordinates of the key evolution nodes. Based on the magnitude and acceleration of the comprehensive critical index exceeding the warning threshold, the estimated time from the current moment to the estimated time of landslide occurrence is calculated through a preset time estimation relationship. By integrating the geographical range and estimated time of high-risk landslide areas, dynamic early warning information is generated.

8. A landslide dynamic early warning method based on spatiotemporal evolution according to claim 2, characterized in that, The dynamic coupling kernel function dynamically adjusts the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence during the fusion process based on the magnitude and trend of the soil pore pressure sequence, including: Extract the soil pore pressure value at the current monitoring time and its change relative to the previous monitoring time from the soil pore pressure sequence; A baseline weight allocation ratio is set, where the surface deformation sequence has baseline weight one and the rainfall intensity sequence has baseline weight two; When the soil pore pressure value is higher than the saturation threshold, the weight coefficient of the rainfall intensity sequence is lowered based on the baseline weight two, while the weight coefficient of the surface deformation sequence is raised based on the baseline weight one. When the change in soil pore pressure is positive and its absolute value is greater than the change threshold, the weighting coefficient of the rainfall intensity sequence will be increased based on the baseline weighting. The weighting coefficients of the adjusted surface deformation sequence and the rainfall intensity sequence are normalized to ensure that the sum of the weighting coefficients of the surface deformation sequence and the rainfall intensity sequence is a constant value. The normalized coefficients are the dynamic weighting coefficients used for the fusion calculation at the current time.

9. A landslide dynamic early warning method based on spatiotemporal evolution according to claim 8, characterized in that, The method utilizes a dynamically coupled kernel function to perform a weighted fusion calculation on the surface deformation value, soil pore pressure value, and rainfall intensity value corresponding to the same geographical location coordinates at each monitoring time, and outputs a fusion state scalar of the geographical location at the monitoring time, including: Obtain the surface deformation value, soil pore pressure value, and rainfall intensity value corresponding to the current monitoring time and current geographical location coordinates; Using the dynamic coupling kernel function, the dynamic weighting coefficients for surface deformation and rainfall intensity at the current monitoring time are calculated based on the soil pore pressure value and its changing trend at the current monitoring time. Soil pore pressure value is used as the basic fusion amount; Multiply the surface deformation value and the rainfall intensity value by their corresponding dynamic weighting coefficients to obtain the weighted surface deformation value and the weighted rainfall value; The basic fusion quantity, weighted surface deformation, and weighted rainfall are added together, and the sum is the fusion state scalar of the monitoring time and geographical coordinates.

10. A landslide dynamic early warning system based on spatiotemporal evolution, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the landslide dynamic early warning method based on spatiotemporal evolution as described in any one of claims 1 to 9.

Images