Search

CN-122020268-A - Intelligent evaluation and early warning method for secondary disaster risk of glacier collapse

CN122020268ACN 122020268 ACN122020268 ACN 122020268ACN-122020268-A

Abstract

The invention relates to the technical field of intersection of artificial intelligence and geological disaster prevention and control, and discloses an intelligent glacier collapse secondary disaster risk assessment and early warning method. The method comprises the steps of fusing multi-source remote sensing and in-situ sensing data to construct glacier three-dimensional deformation fields and internal structure evolution features, dynamically modeling the instability probability of a key area based on a space-time attention-driven graph neural network, establishing a risk evolution equation of multi-physical field coupling by combining with a hydrological meteorological factor, generating a space-time continuous risk level map and triggering a grading early warning instruction. The system comprises modules of multi-source data acquisition, three-dimensional deformation field construction, internal structure inversion, water melting layer identification, space-time diagram construction, instability probability prediction, risk evolution modeling, risk map generation, early warning instruction triggering and the like. According to the intelligent system, accurate, real-time and preposed early warning of secondary disasters such as glacier collapse and landslide and debris flow induced by the intelligent model based on multi-source heterogeneous data depth fusion and physical mechanism constraint is realized.

Inventors

  • WANG SHIJIN
  • MA XINGGANG
  • CHE YANJUN

Assignees

  • 中国科学院西北生态环境资源研究院

Dates

Publication Date
20260512
Application Date
20260414

Claims (10)

  1. 1. The intelligent glacier collapse secondary disaster risk assessment and early warning method is characterized by comprising the following steps of: Acquiring multi-source observation data of a glacier area; Constructing a space-time diagram structure which takes glacier key areas as nodes and takes physical adjacent relations and mechanical conduction paths as edges; Deploying a space-time attention-driven graph neural network model on the space-time graph structure, wherein the model dynamically updates the instability probability of each node through an autoregressive gating mechanism on a time dimension and an anisotropic message transmission mechanism on a space dimension; The multi-source observation data comprise weather hydronic station observation data, accumulated rainfall, daily average air temperature, snow ablation rate and upstream inflow water flow in the weather hydronic station observation data are introduced as external driving variables, and a risk evolution equation of multi-physical field coupling is constructed, wherein the risk evolution equation is defined as that the time derivative of the instability probability is equal to the weighted sum of a deformation energy accumulation item, a crack expansion acceleration item, a melting water lubrication effect item and an external hydronic disturbance item; Based on the numerical solution of the risk evolution equation, generating a space-time continuous risk level map covering the whole glacier basin, wherein the risk levels are divided into four levels of low risk, medium risk, high risk and extremely high risk; When the risk level of any grid cell reaches a high risk or extremely high risk threshold, triggering an early warning instruction of a corresponding level, wherein the early warning instruction comprises a risk position coordinate, an expected collapse time window, a potential influence range and suggested emergency response measures.
  2. 2. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 1, wherein the multi-source observation data further comprises synthetic aperture radar interferometry data, laser radar point cloud data, ground-based synthetic aperture radar monitoring data, ice-quake instrument record data and ground temperature gradient sensor data; Constructing a glacier surface high-precision three-dimensional deformation field based on the synthetic aperture radar interferometry data and the laser radar point cloud data, wherein the three-dimensional deformation field comprises a displacement component along the sight line direction, an elevation change rate along the vertical direction and a displacement vector along the horizontal direction; Inverting the development depth of the internal fissures of glaciers, the expansion rate of the fissures and the spatial distribution density of ice micro-fracture events based on the ground-based synthetic aperture radar monitoring data and ice-quake instrument recording data; calculating the internal temperature profile of the glacier based on the data of the geothermal gradient sensor, and determining the thickness of the water-melt layer of the rock interface and the release strength of the latent heat of phase change; And constructing a space-time diagram structure which takes the glacier key area as a node and takes a physical adjacent relation and a mechanical conduction path as edges by taking the three-dimensional deformation field, the crack development parameter, the ice micro-cracking event density and the water melting layer thickness as node characteristics.
  3. 3. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 2, wherein constructing a glacier surface high-precision three-dimensional deformation field based on the synthetic aperture radar interferometry data and the laser radar point cloud data comprises: performing a least squares phase unwrapping algorithm on the synthetic aperture radar interferometry data to obtain a sequence of phase changes along a line-of-sight direction; correcting the terrain phase by utilizing a digital surface model generated by the laser radar point cloud to eliminate the influence of atmospheric delay and orbit error; Decomposing the sight line displacement component into a vertical direction elevation change rate and a horizontal direction displacement vector; the deformation information of 3 directions is fused to form a three-dimensional deformation field with the spatial resolution of 10 meters and the time resolution of 6 days.
  4. 4. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 3, wherein inversion of the depth of development of cracks in glaciers, the expansion rate of cracks and the spatial distribution density of ice micro-cracking events is performed based on the ground-based synthetic aperture radar monitoring data and ice-quake meter recording data, and comprises the following steps: identifying a surface fracture opening width based on a texture entropy change rate of the synthetic aperture radar amplitude image; Extracting microseismic events with main frequencies between 10 and 50 Hz from ice-quake instrument record data through short-time Fourier transform; Determining three-dimensional space coordinates of a microseismic event by using a double-difference positioning algorithm by combining the P wave travel time difference with the S wave travel time difference with the ice longitudinal wave speed and the ice transverse wave speed; the deepest boundary of the concentrated distribution of the microseismic events is defined as the development depth of the cracks, the crack expansion rate is obtained by dividing the position change quantity of the front end of the cracks in a continuous time period by the time interval, and the number of the microseismic events in a unit volume is defined as the spatial distribution density of the ice body microseismic events.
  5. 5. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 4, wherein calculating a glacier internal temperature profile based on the ground temperature gradient sensor data and determining a glacier interface water-melt layer thickness and a latent heat of phase change release strength comprises: fitting a one-dimensional temperature profile according to the depth of the temperature values measured by the ground temperature gradient sensors distributed every 5 meters along the depth of the drilling hole; defining the vertical distance from the isothermal surface at 0 ℃ to the rock-rock interface in the temperature profile as the thickness of the water-melting layer; the heat flux through the isothermal surface at 0 degrees celsius was calculated according to fourier's law and defined as the latent heat of phase change release intensity.
  6. 6. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 5, wherein constructing a space-time diagram structure with glacier key areas as nodes and physical adjacency and mechanical conduction paths as sides by taking the three-dimensional deformation field, crack development parameters, ice micro-cracking event density and water-melting layer thickness as node characteristics comprises: Performing self-adaptive mesh dissection on the glacier region according to the spatial variation coefficient of the deformation rate gradient and the fracture density, setting the mesh side length to be not more than 20 meters in the region with the deformation rate of more than 5 cm per day or the fracture density of more than 0.5 events per cubic meter, and setting the mesh side length to be not more than 100 meters in the rest regions; defining each grid center point as a node, wherein the node characteristic vector consists of an average three-dimensional deformation field component, crack development depth, crack expansion rate, micro-cracking event density and water melting layer thickness in the grid; If the two nodes are adjacent in space and the elevation difference is smaller than 50 meters, or the historical collaborative deformation correlation coefficient is larger than 0.7, establishing edge connection; and setting the edge weight as a mechanical coupling coefficient, wherein the mechanical coupling coefficient is equal to the product of the inverse of the elastic modulus of the ice body and the square of Euclidean distance between the nodes.
  7. 7. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 6, wherein a space-time attention-driven graph neural network model is deployed on the space-time graph structure, and the instability probability of each node is dynamically updated, comprising: Adopting 3 layers of graph roll lamination, wherein each layer integrates a time gating unit based on a long-short-term memory network and a space attention weight matrix based on Euclidean distance, elevation difference and history cooperative deformation correlation coefficient between nodes; Aggregating the neighbor node characteristic weighted sum through an anisotropic message passing mechanism and fusing with the self history state to generate a new node embedding; The nodes are embedded and mapped into a instability probability value with the value range of 0 to 1 through the full connection layer.
  8. 8. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 7, wherein the integrated rainfall, the average daily air temperature, the snow ablation rate and the upstream inflow water flow in the observation data of the meteorological hydrologic station are introduced as external driving variables to construct a risk evolution equation of multi-physical field coupling, and the method comprises the following steps: The time derivative of the instability probability is expressed as a weighted sum of a deformation energy accumulation term, a crack expansion acceleration term, a water melting lubrication effect term and an external hydrologic disturbance term; The deformation energy accumulation term is defined as the square of the ratio of the current deformation rate to the historical maximum deformation rate; the fracture expansion acceleration term is defined as the absolute value of the fracture density time derivative; The melt lubrication effect term is defined as the product of the thickness of the melt layer and the deviation of the freezing point temperature; the external hydrologic disturbance term is defined as a weighted sum of the cumulative rainfall over the past 72 hours and the number of hours the average daily air temperature exceeds 0 ℃.
  9. 9. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 8, wherein generating a spatiotemporal continuous risk level map covering the entire glacier basin based on a numerical solution of the risk evolution equation comprises: discretizing and solving a risk evolution equation by a finite difference method, wherein the time step is 1 hour, and the initial condition is the initial instability probability output by the graph neural network; Mapping the risk probability value of the discrete node to a regular geographic grid with the spatial resolution of 10 meters through common kriging interpolation, wherein the variation is set to be 50 meters, and the block gold effect is set to be 0.05; Generating a risk evolution prediction for 12 hours in the future by adopting linear extrapolation in the time dimension; dividing risk grades according to the risk probability values: <0.3 is low risk, 0.3≤ <0.7 Is stroke risk, 0.7≤ <0.9 Is a high risk and, And the risk is extremely high at equal to or more than 0.9.
  10. 10. The intelligent glacier collapse secondary disaster risk assessment and early warning method according to claim 9, wherein when the risk level of any grid cell reaches a high risk or extremely high risk threshold, triggering an early warning instruction of a corresponding level comprises: When the risk probability of the same grid unit in the continuous 3 time steps is greater than 0.7, judging that the risk is high, and starting yellow early warning; When the risk probability is greater than 0.9 and the deformation rate exceeds 10 cm per day, judging that the risk is extremely high, and starting red early warning; The early warning instruction comprises longitude and latitude coordinates of a risk position, an expected collapse time window extrapolated based on the time of a risk probability peak value, a potential influence range obtained through the simulation of the volume estimation of a collapse body and a terrain diffusion model, and suggested emergency response measures.

Description

Intelligent evaluation and early warning method for secondary disaster risk of glacier collapse Technical Field The invention belongs to the technical field of intersection of artificial intelligence and geological disaster prevention and control, and particularly relates to an intelligent evaluation and early warning method for secondary disaster risk of glacier collapse. Background With the global warming and aggravation, the rapid ablation of the high-mountain glaciers, the glacier collapse and secondary disasters such as landslide, mud-rock flow and the like are frequently caused, and the safety and infrastructure of downstream residents are seriously threatened. Glacier disaster risk assessment has become a key link in a geological disaster prevention and control and emergency response system. The traditional risk assessment method mainly depends on historical disaster statistics, topography and relief analysis and ground monitoring data of limited points, and is difficult to comprehensively describe the dynamic evolution process of the glacier system. In recent years, the development of remote sensing technology promotes the monitoring of the change of glaciers in a large range, but single data source has obvious limitations that an optical satellite image is easily shielded by cloud and fog and cannot penetrate through an ice body, internal structure information is difficult to obtain, and a ground sensor is high in layout cost and limited in coverage range, so that global real-time sensing is difficult to realize. Therefore, it is needed to construct an intelligent assessment framework capable of fusing multi-source heterogeneous observation data, reconstructing glacier three-dimensional dynamic evolution process and embedding physical mechanisms. The intelligent evaluation of the secondary disaster risk of glacier collapse is characterized in that the accurate identification and prediction of glacier deformation fields, crack development, ice displacement rate and potential instability areas are realized. This requires the system to not only have external morphology perception capability with high spatial-temporal resolution, but also deep analysis of glacier internal stress distribution and rheological properties. Currently, satellite remote sensing can provide earth surface coverage change information with long time sequence, an unmanned aerial vehicle-mounted LiDAR system can acquire three-dimensional point cloud data with centimeter-level precision, and emerging computer vision technologies such as a nerve radiation field (NeRF) and the like show strong potential for reconstructing continuous geometry and appearance from a sparse view angle. However, the existing method mostly uses the technology in isolation, lacks an effective alignment and semantic fusion mechanism of cross-modal data, and does not embed physical priors such as glacier dynamics equations into a reconstruction and prediction model, so that an evaluation result lacks physical consistency and long-term interpretability. The prior art has 3 defects in glacier disaster risk assessment: Firstly, the data perception dimension is single, and the glacier surface deformation and internal structure evolution information cannot be synchronously acquired, so that the critical instability state is judged and delayed; Secondly, the three-dimensional reconstruction model is mostly driven based on pure data, the glacier rheology constraint is ignored, non-object understanding is easy to generate during extrapolation prediction, and the early warning reliability is reduced; Thirdly, multi-source data fusion stays at a characteristic splicing or simple weighting layer, a uniform characterization space with consistent space-time is not established, and high-precision dynamic risk quantification is difficult to support. Especially in the context of frequent extreme weather and acceleration of glacier retraction, the problems described above make it difficult for existing systems to meet the urgent need for accurate early warning of sudden collapse events from hours to days in advance. Therefore, development of an intelligent glacier disaster risk assessment and early warning method which integrates multi-modal sensing and embedding physical mechanisms and supports dynamic three-dimensional reconstruction is urgent. Disclosure of Invention The invention provides an intelligent evaluation and early warning method for secondary disaster risk of glacier collapse, which comprises the steps of constructing a joint characterization system of a three-dimensional dynamic deformation field and internal structure evolution characteristics of glaciers by fusing multi-source heterogeneous remote sensing data, in-situ sensing observation data and a high-resolution digital elevation model, carrying out multi-scale state sensing and evolution trend modeling on key unstability areas of the glaciers by adopting a space-time attention-driven graph neural network o