A method and system for identifying debris flow disasters in high-cold mountainous areas

Abstract

The application discloses a kind of high-cold mountainous area debris flow disaster identification method and system, belong to geological disaster identification technical field, the method will be converted to the unified space-time calculation grid of multiple-source data acquisition, and identification period is divided according to preset time step;With reconstructed rainfall forcing field as the upper boundary input of hydrological process model, with reconstructed snowmelt field as additional water source term, with early wet state, vegetation data and soil hydraulic parameter as initial state parameter, surface runoff is simulated using hydrological process model, and is passed to two-phase scouring and filling dynamic model as water source term, to update physical state field every time step, obtain physical cascade state field, and then mark state and space-time connectivity in each grid cell in the unified space-time calculation grid, output the starting position, starting time, movement path and influence area of debris flow in identification period.The method can form cascade transmission inside the model, so that the debris flow identification result has traceable physical basis.

Description

Technical Field

[0001] This invention relates to the field of geological disaster identification technology, and more specifically to a method and system for identifying debris flow disasters in high-altitude and cold mountainous areas. Background Technology

[0002] Accurate identification and early warning of debris flow disasters in high-altitude and cold mountainous areas are crucial for ensuring the safety of downstream transportation routes, water conservancy facilities, and the lives and property of residents. However, the terrain in high-altitude and cold mountainous areas is steep, meteorological conditions are complex, and conventional monitoring facilities such as ground rain gauges and hydrological stations are extremely sparse, making it difficult to capture localized heavy rainfall and snowmelt processes that trigger debris flows. Furthermore, the initiation and evolution of debris flows is a complex physical process involving rainfall infiltration, surface runoff generation, gully erosion, and two-phase fluid movement. Disaster identification cannot rely solely on single rainfall indicators or remote sensing image changes; it requires the construction of physical models that reflect the watershed's hydrological response and gully dynamic evolution. Therefore, effectively integrating multi-source, multi-scale, and heterogeneous observational data into the driving conditions of physical models is a key aspect of research on debris flow identification methods in high-altitude and cold mountainous areas.

[0003] Currently, existing technologies primarily combine multi-source observation data with numerical simulations for debris flow identification. The numerical models used are typically distributed hydrological models and debris flow dynamic models based on physical processes. This type of method first uses radar reflectivity data to invert distributed rainfall fields, extracts topographic parameters from digital elevation models, identifies vegetation cover and loose sediment source distribution using remote sensing imagery, and incorporates soil moisture data as a wetting coefficient into the model. Based on this, a distributed hydrological model is used to calculate surface runoff, which in turn drives the debris flow dynamic model to simulate the movement and deposition processes. The underlying principle is that by processing the aforementioned observation data separately and assigning them as independent parameters or external inputs to different stages of the model, the simulation process can reflect the actual geographical characteristics of the study area to a certain extent.

[0004] However, although multi-source observation data describes different aspects of the debris flow-prone environment in time and space, existing technologies only assign various types of data as independent parameters to different stages of the model, failing to establish an effective transformation relationship between data and model state variables. This results in the model being unable to accurately depict the key stages in the formation process of debris flows in high-altitude and cold mountainous areas, and the reliability and spatial accuracy of the identification results are difficult to meet the needs of practical applications. Summary of the Invention

[0005] To address the problems existing in the above-mentioned fields, this invention proposes a method and system for identifying debris flow disasters in high-altitude and cold mountainous areas. By establishing a conversion link from multi-source observation data to physical model state variables, various types of data enter the corresponding links of the model according to their physical meaning, and obtain the physical cascaded state field through cascade transmission. Then, the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period are output, so that the identification results have traceable physical evidence, effectively improving the spatial accuracy and physical interpretability of debris flow identification in high-altitude and cold mountainous areas.

[0006] To address the aforementioned technical problems, this invention discloses a method for identifying debris flow hazards in high-altitude and cold mountainous areas, comprising the following steps: Acquire radar reflectivity data, surface meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock depth data of the target watershed within the preset identification period; The precipitation forcing field is reconstructed based on radar reflectivity data and ground meteorological data. The snowmelt field and the previous wet state are reconstructed based on wet snow observation data, ground meteorological data and soil moisture data. The precipitation forcing field, snowmelt field, previous wet state, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data are converted into a unified spatiotemporal computing grid. The identified time period is divided into multiple computing time steps according to a preset time step. Using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as the additional water source term, and the previous wet state, vegetation data, and soil hydraulic parameters as the initial state parameters, the hydrological process model is used to simulate surface runoff, and the surface runoff is transferred as the water source term to the two-phase scour and deposition dynamic model; the physical state field is updated step by step through the two-phase scour and deposition dynamic model to obtain the physical cascade state field; Based on the physical cascaded state field, state marking and spatiotemporal connectivity are performed on each grid cell in the unified spatiotemporal computing grid, and the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period are output.

[0007] Preferably, the step of performing state marking and spatiotemporal connectivity on each grid cell in the unified spatiotemporal computing grid, and outputting the starting position, starting time, movement path, and affected area of ​​the debris flow within the identified time period, specifically includes: When no bed erosion occurs in the grid cell and the volumetric sediment concentration is lower than the preset debris flow concentration threshold, it is marked as runoff state; When the grid cell experiences bed erosion and the volumetric sediment concentration is below the debris flow concentration threshold, it is marked as a sediment-carrying transition state. When the volumetric sand concentration of a grid cell reaches or exceeds the debris flow concentration threshold, it is marked as a debris flow state. When a grid cell that was previously in a debris flow state meets the velocity decay condition and the bed elevation increases, it is marked as an accumulation state. The velocity decay condition is that the velocity is lower than a preset accumulation velocity threshold. Spatiotemporal connectivity is performed on grid cells marked as debris flow or deposition states along the direction of topographic flow convergence. The location of the debris flow state grid cluster that first forms spatiotemporal connectivity is determined as the starting location, and the corresponding time is the starting time. The continuous expansion range of the debris flow state grid cluster along the terrain confluence direction is determined as the movement path, and the union of the debris flow state grid and the deposition state grid is determined as the influence area.

[0008] Preferably, the reconstructing of the precipitation forcing field based on radar reflectivity data and surface meteorological data specifically includes: Beam blocking compensation, ground clutter filtering, and vertical reflectivity profile correction are performed on radar reflectivity data. The corrected radar reflectivity data is converted into an initial distributed rainfall field based on the preset reflectivity factor and the relationship between rainfall intensity. During each correction period, the dynamic scale factor is determined based on the ratio of the measured rainfall at the ground rain gauge to the rainfall at the corresponding grid of the initial distributed rainfall field. The initial distributed rainfall field was corrected using a dynamic scaling factor to obtain the rainfall forcing field.

[0009] Preferably, the reconstructing of the snowmelt field and its previous wet state based on wet snow observation data, surface meteorological data, and soil moisture data specifically includes: The distribution area of ​​wet snow was determined by utilizing the backscattering variations in synthetic aperture radar time-series images; The target grid temperature is determined based on the temperature of the ground weather station, the elevation of the target grid, and the vertical temperature lapse rate. When the target grid temperature is higher than the snow melting threshold, the snow melting amount is determined based on the degree-day factor and the remaining snow storage. Using the snowmelt volume as an incremental water source and the soil moisture data as an initial wetting baseline, the initial soil moisture content or initial saturation of the hydrological process model is determined together, thus completing the reconstruction of the snowmelt field and its previous wetting state.

[0010] Preferably, the step of transferring surface runoff as a water source term to the two-phase scour and sedimentation dynamic model specifically means that surface runoff is used as a water source term in the mass conservation equation of the two-phase scour and sedimentation dynamic model, which is specifically a depth-averaged two-phase flow model.

[0011] Preferably, the step of updating the physical state field step-by-step through the two-phase scouring and silting dynamic model to obtain the physical cascaded state field specifically includes: The physical state field includes water depth, flow velocity, volumetric sediment concentration, and rheological resistance; The updated rheological resistance in the physical state field is fed back into the flow velocity and bed scouring and sedimentation calculations in the next calculation step.

[0012] Preferably, the simulation of surface runoff using a hydrological process model specifically includes: The maximum canopy water storage capacity is determined based on the vegetation type and leaf area index in the vegetation data. The canopy interception amount is determined based on the rainfall forcing field and the maximum canopy water storage, and the canopy interception amount is subtracted from the rainfall forcing field to obtain the effective rainfall reaching the surface; Determine the soil infiltration capacity based on the initial soil moisture content and soil hydraulic parameters; Surface runoff is determined based on the effective rainfall and soil infiltration capacity.

[0013] Preferably, before transferring surface runoff as a water source term to the two-phase scour and deposition dynamic model, a parameter constraint inversion step is further included, which specifically includes: Multiple candidate parameter groups are generated within the preset range of rheological and geotechnical parameter values. Each candidate parameter group contains a set of rheological parameters and a set of geotechnical parameters. Each set of candidate parameters is input into the two-phase scouring and silting dynamic model for simulation, and the simulation results corresponding to each set of candidate parameters are output. Based on the actual scouring and deposition range, actual erosion volume, channel cross-sectional changes, and disaster status of different tributaries determined by UAV digital orthophotos and digital elevation models, the simulation results of each group are evaluated separately. Candidate parameter sets whose evaluation results meet preset conditions are selected as the parameter sets when the two-phase scour and sedimentation dynamic model receives the surface runoff.

[0014] Preferably, the debris flow concentration threshold is 0.35 to 0.45.

[0015] Preferably, it also includes a debris flow disaster identification system for high-altitude and cold mountainous areas, comprising: The multi-source observation and acquisition module is used to acquire radar reflectivity data, surface meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock depth data of the target watershed within a preset identification period. The boundary condition reconstruction module is used to reconstruct the rainfall forcing field based on radar reflectivity data and ground meteorological data; to reconstruct the snowmelt field and the previous wet state based on wet snow observation data, ground meteorological data and soil moisture data; and to convert the rainfall forcing field, snowmelt field, previous wet state, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data into a unified spatiotemporal computing grid, and to divide the identified time period into multiple computing time steps according to a preset time step. The physical cascade calculation module is used to simulate surface runoff using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as an additional water source term, and the previous wet state, vegetation data, and soil hydraulic parameters as initial state parameters. The surface runoff is then passed to the two-phase scour and deposition dynamic model as a water source term. The physical state field is updated hourly by the two-phase scour and deposition dynamic model to obtain the physical cascade state field. The identification output module is used to perform state marking and spatiotemporal connectivity of each grid cell in the unified spatiotemporal computing grid according to the physical cascaded state field, and output the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The proposed method for identifying debris flow hazards in high-altitude and cold mountainous areas establishes a transformation link from multi-source observation data to physical model state variables. This allows data such as radar reflectivity, wet snow observation, and soil moisture to be reconstructed into model-driving conditions with clear physical meanings, such as rainfall forcing fields, snowmelt fields, and pre-wetting states. These conditions then enter the corresponding stages of the hydrological process model and the two-phase erosion and deposition dynamics model according to their respective physical roles. This changes the previous situation where multi-source data were used as isolated parameters, enabling snowmelt humidification and pre-wetting states to influence runoff generation and erosion processes according to physical mechanisms. At the same time, it completely preserves the spatial distribution information of rainfall, avoiding spatial misjudgments caused by sparse ground observations. Furthermore, based on the runoff transfer and hourly state feedback between the hydrological process model and the two-phase erosion and deposition dynamics model, a cascade evolution is formed, allowing the complete physical process of debris flow from runoff generation, erosion, movement to deposition to be deduced hourly within the model. As a result, the debris flow identification results have traceable physical evidence, significantly improving the spatial accuracy and physical interpretability of debris flow identification in high-altitude and cold mountainous areas. Attached Figure Description

[0017] Figure 1 This is a flowchart of the debris flow disaster identification method for high-altitude cold mountainous areas proposed in this invention; Figure 2 This is a diagram showing the reconstructed boundary conditions of multi-source observations provided in an embodiment of the present invention. Figure 3 The identification result observation constraint inversion closed-loop diagram provided in the embodiments of the present invention; Figure 4 The system overall architecture diagram provided for embodiments of the present invention; Figure 5 The diagram shows the physical cascaded state variable transfer relationship provided in the embodiments of the present invention. Detailed Implementation

[0018] The following will refer to the appendices in the embodiments of the present invention. Figures 1-5 The technical solutions in the embodiments of the present invention will be clearly and completely described. It should be understood that the terminology used in the present invention is only for describing particular implementation methods and is not intended to limit the present invention.

[0019] Example like Figure 1 As shown, this invention proposes a method for identifying debris flow disasters in high-altitude and cold mountainous areas, which includes the following steps: S1: Acquire radar reflectivity data, surface meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock depth data of the target watershed within the preset identification period.

[0020] Specifically, this includes radar reflectivity data from ground-based Doppler weather radar or phased array weather radar; surface meteorological data from one or more ground meteorological stations (including at least rainfall and temperature data); synthetic aperture radar time-series imagery as wet snow observation data; soil moisture data from satellite or ground observations; digital elevation models as topographic data; land cover type and leaf area index as vegetation data; soil saturated hydraulic conductivity, saturated water content, residual water content, air intake pressure head, and shape parameters as soil hydraulic parameters; and bedrock depth data.

[0021] Specifically, in this embodiment, a high-altitude mountainous watershed with an upstream snow-covered area, two tributaries on the left and right, and downstream settlements is taken as an example. First, radar reflectivity, rainfall and temperature at the watershed outlet meteorological station, synthetic aperture radar wet snow extent, satellite soil moisture, digital elevation model, land cover, leaf area index, soil hydraulic parameters, and bedrock depth are acquired. The radar reflectivity is converted into a distributed rainfall field after quality control, and time-by-time total amount correction is performed using the outlet rain gauge. This correction preserves the localized heavy rainfall center in the upper reaches of the right tributary, rather than uniformly assigning the outlet rainfall to the entire watershed.

[0022] For post-disaster calibration, it is also possible to obtain UAV digital orthophotos, UAV digital elevation models, and on-site trench cross sections.

[0023] Observations in high-altitude and cold mountainous areas may be interrupted, obstructed, or have resolution mismatches. To ensure the continuous operation of the identification process, timestamps, spatial coverage, quality indicators, and resolutions can be recorded for each type of data when acquiring multi-source observation data.

[0024] S2: Reconstruct the precipitation forcing field based on radar reflectivity data and ground meteorological data, and reconstruct the snowmelt field and previous wet state based on wet snow observation data, ground meteorological data and soil moisture data. Then, convert the precipitation forcing field, snowmelt field, previous wet state, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data into a unified spatiotemporal computing grid, and divide the identification period into multiple computing time steps according to the preset time step.

[0025] When the ground rain gauge data is normal and the rainfall in the corresponding radar grid is greater than the minimum effective value, the dynamic correction of formulas (2) and (3) in step S21 is performed. When the ground rain gauge data is missing, there is a physically unreasonable sudden change, or the rainfall in the corresponding radar grid is lower than the minimum effective value, the dynamic scale factor is not calculated, the quality-controlled radar rainfall field is used directly, or the most recent effective scale factor within the preset time window is used, and the confidence level of the rainfall forcing field is reduced.

[0026] When radar data is partially obscured by terrain, low-confidence areas can be marked using beam obscuration quality maps and supplemented by interpolation from adjacent radars, satellite precipitation products, or terrain correlation. Supplementary data is only used for the missing areas and does not cover radar spatial structures of acceptable quality. When radar data is completely unavailable, satellite precipitation or numerical weather prediction precipitation can be used as an alternative forcing, but the appropriate data source and confidence level should be marked during the disaster identification phase.

[0027] When the synthetic aperture radar (SAR) wet snow observation time and the identification period are not completely consistent, the most recent effective wet snow range can be used as a constraint on snowmelt space, and the remaining snow storage can be updated according to the temperature process. When the resolution of soil moisture products is significantly lower than that of the computational grid, they are used as the previous wetting index at the watershed or zonal scale, and high-resolution direct assignment without basis is avoided.

[0028] When numerical calculations encounter local instability at wet-dry boundaries, excessively large time steps, or state variables exceeding limits, the calculation time step can be reduced, non-physical negative water depths can be truncated, and the volumetric sediment concentration can be constrained to a range from 0 to the maximum allowable concentration. Abnormal meshes should also be recorded. Regions that still do not converge after anomaly handling should be marked as low-confidence regions in the results, rather than directly outputting the deterministic debris flow state.

[0029] Specifically, the snowmelt process is calculated based on the extent of wet snow and elevation-corrected temperature, and the higher initial wet conditions are determined using soil moisture.

[0030] S21: Reconstruct the precipitation forcing field based on radar reflectivity data and surface meteorological data, specifically including: Beam blocking compensation, ground clutter filtering, and vertical reflectivity profile correction are performed on radar reflectivity data. The corrected radar reflectivity data is converted into an initial distributed rainfall field based on the preset reflectivity factor and the relationship between rainfall intensity. Within each correction period, the dynamic scale factor is determined based on the ratio of the measured rainfall at the ground rain gauge station in the ground meteorological data to the rainfall at the corresponding grid of the initial distributed rainfall field at the rain gauge station. The initial distributed rainfall field was corrected using a dynamic scaling factor to obtain the rainfall forcing field.

[0031] Specifically, such as Figure 2 As shown, this embodiment first performs quality control on radar reflectivity data, including partial beam obstruction compensation, ground clutter filtering, outlier removal, and reflectivity vertical profile correction.

[0032] The radar reflectivity factor after quality control is denoted as . The initial distributed rainfall field is determined based on the pre-calibrated relationship between reflectivity factor and rainfall intensity. In one alternative implementation, the reflectivity factor satisfies the following relationship with rainfall intensity:

[0033] in, To calculate grid coordinates, For time, These are parameters calibrated based on historical rainfall in the target area.

[0034] To correct for radar precipitation total bias while preserving the spatial differences identified by the radar, a dynamic scaling factor is calculated for each correction period: in, This represents the measured rainfall from the ground rain gauge during the correction period. These are the grid coordinates corresponding to the rain gauge station. To prevent positive numbers with a denominator of zero, This represents the corrected distributed rainfall forcing field. Upper and lower limits can be set for the dynamic scale factor to reduce the impact of rain gauge outliers on the overall rainfall field. When multiple rain gauges are present, the scale factor can be calculated separately, and spatial weighting can be applied based on distance, elevation zoning, or radar quality level.

[0035] To address the issue that sparse rain gauges cannot characterize the spatial heterogeneity of localized heavy rainfall in high-altitude mountainous areas, this invention utilizes this scale factor to correct the entire radar rainfall field. This corrected rainfall field matches the total measured rainfall on the ground while preserving the spatial differences observed by the radar. Thus, even if there is only one or a few rain gauges in the study area, spatially differentiated rainfall distributions can be obtained, avoiding misjudgments of tributary-scale rainfall caused by uniform extrapolation from a single rain gauge.

[0036] S22: Reconstruct the snowmelt field and its previous wet conditions based on wet snow observation data, surface meteorological data, and soil moisture data, specifically including: The distribution area of ​​wet snow can be determined by utilizing the backscattering variation of synthetic aperture radar time-series images in wet snow observation data; The target grid temperature is determined based on the surface meteorological station temperature in the surface meteorological data, the target grid elevation in the topographic data, and the preset vertical temperature lapse rate. When the target grid temperature is higher than the preset snow melting threshold, the snow melting amount is determined based on the preset day factor and the remaining snow storage in the previous time step. Using snowmelt volume as an incremental water source and soil moisture data as an initial wetting baseline, the initial soil moisture content or initial saturation of the hydrological process model is determined together, thus completing the reconstruction of the snowmelt field and its previous wetting state.

[0037] Specifically, in this embodiment, the wet snow region is extracted using the backscattering variation of synthetic aperture radar (SAR) time-series images. Since SAR is less affected by cloud cover, the wet snow region can be used to constrain the snowmelt range in high-altitude areas. Let the temperature at the ground weather station be... The elevation of the weather station is The target grid elevation is The vertical temperature lapse rate is Then the target grid temperature can be determined as:

[0038] when Furthermore, when the grid is located within a snow-bound or wet snow-bound area, the snowmelt amount can be determined as follows: in, This represents the remaining snow reserves from the previous period. For daily living factors, This is for calculating the time step.

[0039] Snowmelt volume is used to update the previous soil moisture content or the previous watershed wetness index. When the synthetic aperture radar wet snow observation time and the identification period are not completely consistent, the most recent effective wet snow range can be used as the snowmelt spatial constraint, and the remaining snow storage can be updated according to the temperature process.

[0040] S23: Transform rainfall forcing field, snowmelt field, pre-wetting status, topographic data, vegetation data, soil hydraulic parameters, and bedrock depth data into a unified spatiotemporal computing grid, specifically including: Topography, land cover, leaf area index, soil hydraulic parameters, initial soil moisture content, and bedrock depth are mapped to a unified grid. The grid resolution can be determined based on the digital elevation model, target channel width, and computing resources, for example, from 1 meter to 30 meters; the temporal resolution for rainfall and temperature can be on the order of minutes or hours, and the internal calculation time step can be automatically determined based on stability conditions.

[0041] When the spatial resolution of soil moisture data is lower than that of the spatiotemporal computing grid, the soil moisture data is used as the watershed pre-wetness index, and the watershed pre-wetness index is spatially downscaled according to topographic, soil or vegetation zones.

[0042] Specifically, in this embodiment, when the spatial resolution of satellite soil moisture data is lower than that of the computational grid, it can be used as the average moisture baseline of the watershed, and then the data can be divided into zones based on slope aspect, slope, soil type, vegetation type and wet snow area, instead of directly interpreting low-resolution pixels as high-resolution accurate values.

[0043] S3: Using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as the additional water source term, and the previous wet state, vegetation data and soil hydraulic parameters as the initial state parameters, the hydrological process model is used to simulate surface runoff, and the surface runoff is transferred to the two-phase scour and deposition dynamic model as the water source term; the physical state field is updated step by step through the two-phase scour and deposition dynamic model to obtain the physical cascade state field.

[0044] To address the issue that multi-source observations such as wet snow, snowmelt, soil moisture, topography, vegetation, and soil parameters are merely used in parallel without forming initial and boundary conditions for the model according to their physical meaning, this invention ensures that each type of data enters different stages of the model according to its own physical meaning. In step S3, the rainfall forcing field is used as the upper boundary flux, the snowmelt field as the water source term, soil moisture as the initial water content, vegetation and soil hydraulic parameters as physical process control parameters, topography as the topography driver, and bedrock depth as the erosion constraint. This establishes an explicit transformation link from observed physical quantities to model state variables, so that multi-source observation data is no longer passively input information, but is organically integrated into the entire simulation process as a model driving force with clear physical functions.

[0045] In the initial stages of operation of the hydrological process model and the two-phase erosion-deposition dynamic model, some rainfall was intercepted by the canopy, and subsequently, the soil infiltration capacity gradually decreased. Continuous surface runoff first occurred in the localized heavy rainfall area of ​​the right tributary. As flow velocity and bed shear increased, bed erosion occurred in the right tributary, and the grid state transitioned from runoff to sediment-laden transition. The eroded material increased the volumetric sediment concentration, and when the volumetric sediment concentration reached the debris flow concentration threshold, the relevant grids transitioned to debris flow state. The left tributary, due to lower rainfall intensity and insufficient bed erosion, maintained either a runoff state or a sediment-laden transition state.

[0046] S31: Surface runoff is simulated using a hydrological process model, specifically including: The maximum canopy water storage capacity is determined based on the vegetation type and leaf area index in the vegetation data. The canopy interception is determined based on the rainfall forcing field and the maximum canopy water storage, and the canopy interception is deducted from the rainfall forcing field to obtain the effective rainfall reaching the surface; Determine the soil infiltration capacity based on the initial soil moisture content and soil hydraulic parameters; Surface runoff is determined based on effective rainfall and soil infiltration capacity.

[0047] Specifically, physical cascading is not a one-time sequential call of multiple modules, but rather the passing and updating of state variables in each computation step.

[0048] In this embodiment, firstly, the maximum canopy water storage is determined based on land cover and leaf area index, and the current canopy interception is calculated. Effective rainfall reaching the surface is the corrected distributed rainfall minus the increase in canopy water storage. Different maximum canopy water storage relationships can be configured for forests, grasslands, cultivated land, or bare land. Once the canopy reaches saturation, additional rainfall enters the surface.

[0049] Subsequently, the infiltration process is calculated based on the initial soil moisture content, pressure head, and soil hydraulic parameters. The infiltration model can employ the variable saturated soil water movement equation and determine the hydraulic conductivity using saturated hydraulic conductivity, saturated water content, residual water content, air inlet pressure head, and shape parameters. Increased initial soil moisture content due to early snowmelt reduces available infiltration reserves, leading to earlier surface runoff.

[0050] S32: Surface runoff is transferred as a water source term to the two-phase scour and deposition dynamic model. Specifically, surface runoff is used as a water source term in the mass conservation equation of the two-phase scour and deposition dynamic model, which is a depth-averaged two-phase flow model.

[0051] In this embodiment, the water depth is set to... h The depth-averaged flow velocity along the two horizontal coordinate directions is u and v The volumetric sand concentration is cThe bed surface elevation is z b Solid phase erosion rate E s Based on the conservation of mass and momentum, the water depth and flow velocity are updated; based on the relationship between bed shear and bed shear strength, the erosion rate is determined, and the bed elevation and volumetric sediment concentration are updated.

[0052] In this embodiment, bed erosion is only allowed to occur within the soil layer. When the cumulative erosion depth reaches the bedrock burial depth, the erodible thickness of the corresponding grid is set to zero. This treatment avoids unreasonable and continuous erosion of the bedrock area by the model when there are no constraints.

[0053] S33: The physical state field is obtained by updating the physical state field step-by-step through the two-phase scouring and sedimentation dynamic model, specifically including: The physical state field includes water depth, flow velocity, volumetric sediment concentration, and rheological resistance; The updated rheological resistance in the physical state field is fed back into the flow velocity and bed scouring and sedimentation calculations in the next calculation step.

[0054] Rheological parameters are updated hourly based on volumetric sediment concentration. The mixture density can be determined by weighting the water density and solid density according to the volumetric sediment concentration. Yield stress and dynamic viscosity can be expressed as an exponential function of the volumetric sediment concentration.

[0055] in, For yield stress, For dynamic viscosity, These are rheological parameters. As bed erosion increases, the volumetric sediment concentration rises, and the yield stress and dynamic viscosity change. The changed rheological resistance affects the flow velocity and bed shear in the next time step, thus affecting subsequent erosion. This forms a closed loop of "effective rainfall - infiltration - surface runoff - bed erosion - sediment concentration - rheological resistance - subsequent erosion".

[0056] The computation can be solved using an explicit difference scheme with decreasing total variation, or using a finite volume scheme, a finite difference scheme, or other numerical schemes that satisfy conservation requirements. Computation can be accelerated using shared memory parallelism, GPU parallelism, region partitioning, or surrogate models.

[0057] When numerical calculations encounter local instability at wet-dry boundaries, excessively large time steps, or state variables exceeding limits, the calculation time step can be reduced, non-physical negative water depths can be truncated, and the volumetric sediment concentration can be constrained to a range from 0 to the maximum allowable concentration. Abnormal meshes should also be recorded. Regions that still do not converge after anomaly handling should be marked as low-confidence regions in the results, rather than directly outputting the deterministic debris flow state.

[0058] The replacement of numerical solution methods does not change the core relationship of converting multi-source observations into boundary conditions and identifying them through physical state cascades.

[0059] To address the lack of clear state transition criteria in hydrological and two-phase scour and deposition simulation results, which makes it difficult to directly identify the starting location, starting time, movement path, and affected area of ​​debris flows, this invention introduces physically meaningful state criteria such as whether the volumetric sediment concentration reaches the debris flow concentration threshold, whether bed erosion occurs, and whether the flow velocity decays and the bed elevation increases. The continuous numerical simulation results are discretized into four identifiable states: runoff, sediment-carrying transition, debris flow, and deposition. Then, through spatiotemporal connectivity, the state grid is connected into a spatially coherent disaster object, enabling the identification conclusions to directly serve disaster emergency response and engineering decision-making.

[0060] S4: Based on the physical cascaded state field, state marking and spatiotemporal connectivity are performed on each grid cell in the unified spatiotemporal computing grid, and the starting position, starting time, movement path and affected area of ​​the debris flow are output within the identification period.

[0061] Based on the continuous debris flow state grid within the right tributary, the starting location and time were identified, and the movement path was formed along the main tributary. In the area where the slope decreases at the tributary mouth, the flow velocity decreases and the bed elevation increases, causing the grid to transform into a depositional state and forming an influence area. The right tributary was identified as the main initiating tributary, the main tributary as the main movement path, and the tributary mouth as the main depositional influence area, while maintaining the identification difference that the left tributary did not form a fully developed debris flow.

[0062] Specifically, for each grid cell in the unified spatiotemporal computation grid, state marking is performed based on the physical cascaded state field of each grid cell, where: When a grid cell does not experience bed erosion and its volumetric sediment concentration is below the preset debris flow concentration threshold, it is marked as a runoff state; when a grid cell experiences bed erosion and its volumetric sediment concentration is below the debris flow concentration threshold, it is marked as a sediment-carrying transition state; when the volumetric sediment concentration of a grid cell reaches or exceeds the debris flow concentration threshold, it is marked as a debris flow state; when a grid cell that was previously in a debris flow state meets the velocity decay condition and the bed elevation increases, it is marked as an deposition state, where the velocity decay condition is that the velocity is below the preset deposition velocity threshold. Spatiotemporal connectivity is performed on grid cells marked as debris flow or deposition states along the direction of topographic flow convergence. The location of the debris flow state grid cluster that first forms a spatiotemporally connected network is determined as the starting location, and the corresponding time is determined as the starting time. The continuous expansion range of the debris flow state grid cluster along the terrain confluence direction is determined as the movement path, and the union of the debris flow state grid and the deposition state grid is determined as the influence area.

[0063] In this embodiment, at each computation time step, the physical cascaded state field corresponding to each grid cell is read. In order to transform the continuous numerical simulation into disaster identification results that can be used in engineering, the physical cascaded state field of each grid cell is marked with a state label.

[0064] In one implementation, the grid state Determined according to the following rules: Where 0 represents runoff state, 1 represents sediment-carrying transition state, 2 represents debris flow state, and 3 represents deposition state. The threshold for debris flow concentration. For the velocity modulus, The accumulation velocity threshold, This represents the current increment of the bed surface elevation.

[0065] The above The mutual exclusion priority criteria are given in the order of deposition state, debris flow state, sand-laden transition state and other cases to avoid conflicts when a grid satisfies multiple state conditions at the same time.

[0066] In one specific implementation, the debris flow concentration threshold A value of 0.40 can be used, or it can be determined within the range of 0.35 to 0.45 based on the regional material composition. This threshold is not used alone for regional early warning, but rather applies to the volumetric sediment concentration state after hydrological, hydrodynamic, and scour-deposition cascade calculations have been completed. Therefore, it can establish a traceable relationship with bed erosion, flow path, and previous wetting conditions.

[0067] When a grid cell has no bed erosion and the volumetric sediment concentration is below the debris flow concentration threshold, it is marked as a runoff state; when a grid cell has bed erosion and the volumetric sediment concentration is below the debris flow concentration threshold, it is marked as a sediment-carrying transition state; when the volumetric sediment concentration reaches or exceeds the debris flow concentration threshold, it is marked as a debris flow state; when a grid cell that was previously in a debris flow state meets the velocity decay condition and the bed elevation increases, it is marked as an accumulation state.

[0068] First, within the same computation time step, debris flow state grids are connected according to eight-neighborhood, flow-direction neighborhood, or channel topology to form debris flow state grid clusters; then, grid clusters with inheritance relationships along the topographic confluence direction in adjacent output time steps are connected.

[0069] The grid cluster that first reaches the preset minimum connectivity area or minimum continuous duration is identified as the effective debris flow initiation object. Correspondingly, its first occurrence position is taken as the starting position, and the corresponding time is taken as the starting time; the subsequent set of grids expanding along the confluence direction is the movement path; the combination of debris flow state grids, grids reaching dangerous water depths or dangerous flow velocities, and deposition state grids can constitute the influence area.

[0070] To reduce false identification caused by isolated and abnormal grid cells, minimum connectivity area, minimum continuity duration, and channel constraints can be set. The minimum connectivity area should be greater than the area of ​​a single computational grid cell; the minimum continuity duration can be set based on the model output interval. For grid cells on the outer slope of the channel, whether to retain them can be determined based on the slope, runoff accumulation, or pre-extracted channel network.

[0071] In this embodiment, the starting area, movement path, and influence area can be generated in vector or raster form, while outputting the start time, the time of arrival at the key section, the maximum water depth, the maximum flow velocity, the maximum erosion depth, and the maximum deposition thickness. The identification results can be used for disaster identification, emergency evacuation range delineation, key tributary ditch screening, or engineering measure deployment.

[0072] The execution process of steps S1-S4 above demonstrates that the method proposed in this invention does not directly provide watershed-level conclusions based on a single rainfall threshold. Instead, it drives hydrological process models and two-phase scour-deposition dynamic models through spatially differentiated rainfall forcing and pre-wetting conditions, and then forms disaster identification objects based on the spatiotemporal transition of state variables. Therefore, even if the rainfall, soil moisture, and sediment source conditions of each tributary within the same watershed are different, their disaster states can still be identified separately.

[0073] To address the issue of ambiguity in rheological and geotechnical parameters, which leads to discrepancies between the identified results and the actual erosion and deposition patterns, this invention does not rely on a single observation indicator (such as outlet flow) for parameter calibration. Instead, it simultaneously constrains the simulation results from four dimensions: deposition range, erosion volume, cross-section, and tributary state. The performance of different parameter groups in different dimensions can be distinguished, ensuring that the final selected parameter group matches actual observations at multiple physical characteristic levels, thereby reducing the impact of parameter ambiguity on spatial identification results.

[0074] The method proposed in this invention, before transferring surface runoff as a water source term to the two-phase scour-deposition dynamic model in step S3, also includes a parameter constraint inversion step, such as... Figure 3 As shown, the parameter constraint inversion step specifically includes: Multiple candidate parameter sets are generated within the preset range of rheological and geotechnical parameter values. Each candidate parameter set contains a set of rheological parameters and a set of geotechnical parameters. Each set of candidate parameters is input into the two-phase scour and sedimentation dynamic model for simulation, and the simulation results corresponding to each set of candidate parameters are output. Based on the actual scouring and deposition range, actual erosion volume, channel cross-sectional changes, and disaster status of different tributaries determined by UAV digital orthophotos and digital elevation models, the simulation results of each group are evaluated separately. Candidate parameter sets whose evaluation results meet the preset conditions are selected as the parameter sets when the two-phase scour and sedimentation dynamic model receives surface runoff.

[0075] Specifically, when there is historical debris flow event data or post-disaster investigation data in the target watershed, in this embodiment, the theoretical range of parameters is first set based on literature, experiments, or regional experience. Then, multiple sets of candidate parameters are generated within the theoretical range, and a complete physical cascade model is executed for each set of candidate parameters.

[0076] UAV digital orthophotos can be used to extract the boundaries of actual erosion and deposition zones. The difference between the UAV digital elevation model and the pre-disaster digital elevation model can be used to estimate bed elevation changes and erosion-deposition volumes. On-site cross-sections can be used to constrain local erosion depths. For watersheds with multiple tributaries, the actual state of each tributary—whether it is a debris flow, a sediment-laden flood, or no significant flow—can also be recorded.

[0077] The comprehensive evaluation function of the candidate parameter group can be expressed as: in, This is a critical success index for simulating the accumulation range and comparing it with the actual accumulation range. To account for the normalized error of erosion volume, For different tributary ditch disaster state errors, For the cross-sectional error of the channel, The weights are non-negative and their sum is 1. Choose the comprehensive evaluation function. The candidate parameter set that is the smallest or less than the preset threshold is used as the parameter set when the two-phase scour and sedimentation dynamic model receives surface runoff.

[0078] Compared to calibration based solely on a single outlet flow rate, simultaneously using depositional space overlap, erosion volume, cross-section, and tributary conditions can constrain different types of simulation biases. For example, a set of candidate parameters might correctly simulate the peak outlet flow rate but produce incorrect debris flow conditions in tributaries where no debris flow has occurred; in this case, tributary condition errors will prevent that set of candidate parameters from being selected. This reduces the impact of parameter ambiguity on spatial identification results.

[0079] For watersheds lacking post-disaster drone data, historical remote sensing imagery, field survey points, known stacking fan boundaries, or parameters from adjacent similar watersheds can be used as constraints. For watersheds without any historical events, multiple parameter scenarios can be provided based on theoretical scope and regional experience, with the identification results and uncertainty ranges output for each scenario.

[0080] This invention also proposes a debris flow disaster identification system for high-altitude and cold mountainous areas, such as... Figure 4 As shown, it includes: The multi-source observation and acquisition module 110 is used to acquire radar reflectivity data, ground meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data of the target watershed within a preset identification period. The boundary condition reconstruction module 120 includes a precipitation field correction unit 121, a snowmelt wetting reconstruction unit 122, and a spatiotemporal grid unification unit 123. The precipitation field correction unit 121 is used to reconstruct the precipitation forcing field based on radar reflectivity data and surface meteorological data. The snowmelt wetting reconstruction unit 122 reconstructs the snowmelt field and the previous wetting state based on wet snow observation data, surface meteorological data, and soil moisture data. The spatiotemporal grid unification unit 123 converts the precipitation forcing field, snowmelt field, previous wetting state, topographic data, vegetation data, soil hydraulic parameters, and bedrock burial depth data into a unified spatiotemporal computing grid, and divides the identification period into multiple computing time steps according to a preset time step. The physical cascade calculation module includes a hydrological cascade calculation module 130, a two-phase scour and deposition calculation module 140, and a state transition identification module 150, wherein: The hydrological cascade calculation module 130 is used to simulate surface runoff using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as the additional water source term, the previous wet state, vegetation data and soil hydraulic parameters as the initial state parameters, and the hydrological process model to simulate surface runoff, and the surface runoff is transferred to the two-phase scour and deposition dynamic model as the water source term.

[0081] like Figure 5 As shown, the hydrological cascade calculation module 130 includes a canopy interception unit 131, a soil infiltration unit 132, and a surface runoff generation unit 133. The canopy interception unit 131 calculates the canopy interception amount and the effective rainfall reaching the surface based on vegetation data and rainfall forcing field, and transmits the effective rainfall to the soil infiltration unit 132. The soil infiltration unit 132 calculates the soil infiltration amount based on the previous wetting state, soil hydraulic parameters, and the received effective rainfall, and transmits the remaining portion of the effective rainfall after deducting the soil infiltration amount to the surface runoff generation unit 133. The surface runoff generation unit 133 determines the surface runoff based on the received remaining effective rainfall and transmits the surface runoff as a water source term to the two-phase scour and sedimentation calculation module 140.

[0082] The two-phase scour and sedimentation calculation module 140 uses a two-phase scour and sedimentation dynamic model to update the physical state field step by step.

[0083] The two-phase scour and deposition calculation module 140 includes a hydrodynamic update unit 141, a bed surface scour and deposition unit 142, and a rheological feedback unit 143. The hydrodynamic update unit 141 updates the water depth and velocity based on the received surface runoff source term, topographic slope, and flow resistance, and transmits the updated water depth and velocity to the bed surface scour and deposition unit 142. The bed surface scour and deposition unit 142 updates the bed surface scour and deposition volume and volumetric sediment concentration based on the water depth, velocity, and bedrock burial depth data. The rheological feedback unit 143 updates the rheological resistance based on the volumetric sediment concentration and feeds the updated rheological resistance back to the hydrodynamic update unit 141 for updating the water depth and velocity in the next calculation step.

[0084] The state transition identification module 150 uses the updated physical state field as the initial state field for the next calculation step and outputs the physical cascaded state field.

[0085] The state transition identification module 150 includes a grid state marking unit 151 and a spatiotemporal connectivity unit 152; wherein, the grid state marking unit 151 is used to mark the state of each grid unit according to the physical cascaded state field; the spatiotemporal connectivity unit 152 is used to spatiotemporally connect the grid units marked as debris flow state or deposition state along the terrain confluence direction.

[0086] The identification output module 170 is used to output the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period based on the state marking results of the grid state marking unit 151 and the spatiotemporal connectivity results of the spatiotemporal connectivity unit 152.

[0087] The system also includes an observation constraint calibration module 160 for performing the above-mentioned parameter constraint inversion steps.

[0088] The method proposed in this invention combines distributed radar rainfall with dynamic scale correction of ground rainfall. While correcting for total rainfall deviations, it preserves the spatial differences in localized heavy rainfall, reducing misidentification of tributaries caused by uniform extrapolation from a single rain gauge. Simultaneously, it converts wet snow, snowmelt, and soil moisture into pre-infiltration conditions for the infiltration model, allowing the impact of snowmelt humidification on runoff initiation time and bed shear to be transmitted along the physical link to debris flow identification conclusions. Based on this, by transmitting hydrological, hydrodynamic, scour and deposition, and rheological states at each calculation time step, it can identify the formation process from clear water runoff to sediment-laden transition flow and then to debris flow, outputting spatially defined disaster targets. Furthermore, by constraining and inverting parameters through UAV and on-site scour and deposition observations, it can simultaneously utilize deposition range, erosion volume, cross-section, and tributary status to correct identification results, improving parameter interpretability and cross-event reuse. Therefore, this invention is applicable to high-altitude and cold mountainous areas where ground monitoring is sparse and difficult to access, forming a closed loop from observation acquisition, physical calculation, disaster identification to result correction.

[0089] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0090] Furthermore, unless otherwise stated, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. All references to this specification are incorporated by way of citation to disclose and describe methods relating to those references. In the event of any conflict with any incorporated reference, the content of this specification shall prevail.

Claims

1. A method for identifying debris flow hazards in high-altitude and cold mountainous areas, characterized in that, Includes the following steps: Acquire radar reflectivity data, surface meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock depth data of the target watershed within the preset identification period; The precipitation forcing field is reconstructed based on radar reflectivity data and ground meteorological data. The snowmelt field and the previous wet state are reconstructed based on wet snow observation data, ground meteorological data and soil moisture data. The precipitation forcing field, snowmelt field, previous wet state, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data are converted into a unified spatiotemporal computing grid. The identified time period is divided into multiple computing time steps according to a preset time step. Using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as the additional water source term, and the previous wet state, vegetation data, and soil hydraulic parameters as the initial state parameters, the hydrological process model is used to simulate surface runoff, and the surface runoff is transferred as the water source term to the two-phase scour and deposition dynamic model; the physical state field is updated step by step through the two-phase scour and deposition dynamic model to obtain the physical cascade state field; Based on the physical cascaded state field, state marking and spatiotemporal connectivity are performed on each grid cell in the unified spatiotemporal computing grid, and the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period are output.

2. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The process of performing state labeling and spatiotemporal connectivity on each grid cell in the unified spatiotemporal computing grid, and outputting the starting position, starting time, movement path, and affected area of ​​the debris flow within the identified time period, specifically includes: When no bed erosion occurs in the grid cell and the volumetric sediment concentration is lower than the preset debris flow concentration threshold, it is marked as runoff state; When the grid cell experiences bed erosion and the volumetric sediment concentration is below the debris flow concentration threshold, it is marked as a sediment-carrying transition state. When the volumetric sand concentration of a grid cell reaches or exceeds the debris flow concentration threshold, it is marked as a debris flow state. When a grid cell that was previously in a debris flow state meets the velocity decay condition and the bed elevation increases, it is marked as an accumulation state. The velocity decay condition is that the velocity is lower than a preset accumulation velocity threshold. Spatiotemporal connectivity is performed on grid cells marked as debris flow or deposition states along the direction of topographic flow convergence. The location of the debris flow state grid cluster that first forms spatiotemporal connectivity is determined as the starting location, and the corresponding time is the starting time. The continuous expansion range of the debris flow state grid cluster along the terrain confluence direction is determined as the movement path, and the union of the debris flow state grid and the deposition state grid is determined as the influence area.

3. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The reconstructing of the precipitation forcing field based on radar reflectivity data and surface meteorological data specifically includes: Beam blocking compensation, ground clutter filtering, and vertical reflectivity profile correction are performed on radar reflectivity data. The corrected radar reflectivity data is converted into an initial distributed rainfall field based on the preset reflectivity factor and the relationship between rainfall intensity. Within each correction period, the dynamic scale factor is determined based on the ratio of the measured rainfall at the ground rain gauge station in the ground meteorological data to the rainfall at the corresponding grid of the initial distributed rainfall field at the rain gauge station. The initial distributed rainfall field was corrected using a dynamic scaling factor to obtain the rainfall forcing field.

4. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The reconstruction of the snowmelt field and its previous wet conditions based on wet snow observation data, surface meteorological data, and soil moisture data specifically includes: The distribution area of ​​wet snow is determined by utilizing the backscattering variation of synthetic aperture radar time-series images in the wet snow observation data; The target grid temperature is determined based on the surface meteorological station temperature in the surface meteorological data, the target grid elevation in the terrain data, and the preset vertical temperature lapse rate. When the target grid temperature is higher than the preset snow melting threshold, the snow melting amount is determined based on the preset day factor and the remaining snow storage in the previous time step. Using the snowmelt volume as an incremental water source and the soil moisture data as an initial wetting baseline, the initial soil moisture content or initial saturation of the hydrological process model is determined together, thus completing the reconstruction of the snowmelt field and its previous wetting state.

5. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The step of transferring surface runoff as a water source term to the two-phase scour and sedimentation dynamic model specifically means that surface runoff is used as a water source term in the mass conservation equation of the two-phase scour and sedimentation dynamic model, which is specifically a depth-averaged two-phase flow model.

6. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The step of updating the physical state field step-by-step through the two-phase scouring and silting dynamic model to obtain the physical cascaded state field specifically includes: The physical state field includes water depth, flow velocity, volumetric sediment concentration, and rheological resistance; The updated rheological resistance in the physical state field is fed back into the flow velocity and bed scouring and sedimentation calculations in the next calculation step.

7. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, The simulation of surface runoff using a hydrological process model specifically includes: The maximum canopy water storage capacity is determined based on the vegetation type and leaf area index in the vegetation data. The canopy interception amount is determined based on the rainfall forcing field and the maximum canopy water storage, and the canopy interception amount is subtracted from the rainfall forcing field to obtain the effective rainfall reaching the surface; Determine the soil infiltration capacity based on the initial soil moisture content and soil hydraulic parameters; Surface runoff is determined based on the effective rainfall and soil infiltration capacity.

8. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 1, characterized in that, Before transferring surface runoff as a water source term to the two-phase scour and deposition dynamic model, a parameter constraint inversion step is included, which specifically includes: Multiple candidate parameter groups are generated within the preset range of rheological and geotechnical parameter values. Each candidate parameter group contains a set of rheological parameters and a set of geotechnical parameters. Each set of candidate parameters is input into the two-phase scouring and silting dynamic model for simulation, and the simulation results corresponding to each set of candidate parameters are output. Based on the actual scouring and deposition range, actual erosion volume, channel cross-sectional changes, and disaster status of different tributaries determined by UAV digital orthophotos and digital elevation models, the simulation results of each group are evaluated separately. Candidate parameter sets whose evaluation results meet preset conditions are selected as the parameter sets when the two-phase scour and sedimentation dynamic model receives the surface runoff.

9. The method for identifying debris flow disasters in high-altitude and cold mountainous areas according to claim 2, characterized in that, The debris flow concentration threshold is 0.35 to 0.

45.

10. A debris flow disaster identification system for high-altitude cold mountainous areas, characterized in that, include: The multi-source observation and acquisition module is used to acquire radar reflectivity data, ground meteorological data, wet snow observation data, soil moisture data, topographic data, vegetation data, soil hydraulic parameters and bedrock depth data of the target watershed within a preset identification period. The boundary condition reconstruction module is used to reconstruct the rainfall forcing field based on radar reflectivity data and ground meteorological data; to reconstruct the snowmelt field and the previous wet state based on wet snow observation data, ground meteorological data and soil moisture data; and to convert the rainfall forcing field, snowmelt field, previous wet state, topographic data, vegetation data, soil hydraulic parameters and bedrock burial depth data into a unified spatiotemporal computing grid, and to divide the identified time period into multiple computing time steps according to a preset time step. The physical cascade calculation module is used to simulate surface runoff using the rainfall forcing field as the upper boundary input of the hydrological process model, the snowmelt field as an additional water source term, and the previous wet state, vegetation data, and soil hydraulic parameters as initial state parameters. The surface runoff is then passed to the two-phase scour and deposition dynamic model as a water source term. The physical state field is updated hourly by the two-phase scour and deposition dynamic model to obtain the physical cascade state field. The identification output module is used to perform state marking and spatiotemporal connectivity of each grid cell in the unified spatiotemporal computing grid according to the physical cascaded state field, and output the starting position, starting time, movement path and affected area of ​​the debris flow within the identification period.

Images