Micro-watershed unit partitioning method based on susceptibility purity and hydrological topology dimensionality reduction

By using a method based on susceptibility purity and water system topology dimensionality reduction, the problem of insufficient geological intrinsic constraints in the division of micro-watershed units was solved, generating micro-watershed units with regular morphology and continuous hydrology, thereby improving the accuracy of geological hazard assessment and the robustness of the model.

CN122574387BActive Publication Date: 2026-11-10CHONGQING INST OF GEOLOGY & MINERAL RESOURCES
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202610726570.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-05-25
Publication Date
2026-11-10
Estimated Expiration
2046-05-25

AI Technical Summary

Technical Problem

Existing micro-watershed unit delineation techniques lack geological constraints, leading to distorted evaluation conditions, abnormally fragmented unit morphology, heavy computational burden, and poor reproducibility of results.

Method used

A micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction is adopted. The susceptibility contribution weight of geological factors is calculated by the information content method. One-dimensional attribute sequence optimization is performed by combining the tree-like topology of the water system. A two-order fusion optimization is performed using greedy algorithm and dynamic programming algorithm.

Benefits of technology

Micro-watershed units with regular morphology, continuous hydrology, and homogeneous geological properties were generated, reducing the number of units, improving evaluation accuracy and model robustness, and realizing standardized and high-precision disaster prevention deployment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122574387B_ABST
    Figure CN122574387B_ABST
Patent Text Reader

Abstract

This invention relates to the field of geological disaster prevention and control technology, and particularly to a method for dividing micro-watershed units based on susceptibility purity and drainage system topology dimensionality reduction. First, multi-source data preprocessing and raster-level susceptibility assessment are completed. Then, the minimum catchment area is extracted based on the DEM and a drainage system tree topology is constructed, reducing the dimensionality of the two-dimensional catchment area along the main river channel to a one-dimensional sequence. Using susceptibility purity as the core criterion, a greedy algorithm and dynamic programming are combined to achieve globally optimal segmentation, ultimately mapping and generating geologically homogeneous and hydrologically continuous micro-watershed evaluation units. This invention solves the problem that existing micro-watershed unit division techniques lack geological intrinsic constraints during evaluation, leading to distorted evaluation conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of geological disaster prevention and control technology, and in particular to a method for dividing micro-watershed units based on susceptibility purity and water system topology dimensionality reduction. Background Technology

[0002] The assessment of geological hazard (landslides, debris flows, collapses, etc.) susceptibility and meteorological risk early warning are core components of regional disaster prevention and mitigation systems. The accuracy of the assessment and the granularity of the early warning directly depend on the method of dividing the assessment units. Currently, slope units or micro-watershed units extracted based on hydrological and topographic features have become the mainstream assessment unit type in the field of geological hazard assessment in China because they can fully reflect the hydrodynamic processes and geomorphological evolution patterns of geological hazard formation.

[0003] However, existing micro-watershed unit partitioning techniques still suffer from the following core shortcomings that are difficult to overcome in practical engineering applications: 1. Lack of scale identification standards leads to highly subjective classification results. Traditional microwatershed unit delineation relies on manual experience to set catchment area thresholds, lacking a unified standard for scale adaptation. An excessively large threshold masks local geological heterogeneity within the study area, leading to overly coarse evaluation results; conversely, an excessively small threshold generates numerous highly fragmented units, increasing the computational burden of subsequent modeling and reducing model stability. Furthermore, due to differences in experience among researchers, delineation results for the same study area often vary significantly, exhibiting poor reproducibility and hindering standardized cross-regional applications.

[0004] 2. Lack of spatial dimensionality reduction mechanisms, resulting in abnormally fragmented unit morphology. Existing methods primarily extract unit boundaries directly from high-dimensional topographic features such as slope and catchment area derived from DEM, failing to effectively integrate the spatial distribution patterns of multi-source geographic elements. In mountainous areas with complex terrain and redundant features, the direct superposition of high-dimensional features easily leads to a disconnect between unit boundaries and the actual geomorphic development process, resulting in a large number of fragmented units with irregular shapes and excessively small areas, which seriously affects the convergence and robustness of subsequent evaluation models.

[0005] 3. Lack of internal geological constraints leads to distorted evaluation of basic conditions. This is the most fundamental flaw in existing technology. Traditional micro-watershed delineation is entirely based on the principle of topographic and hydrological continuity, completely ignoring the intrinsic geological factors that play a decisive role in the occurrence of geological disasters, such as stratigraphic lithology, tectonic faults, and material composition. As a result, the same micro-watershed unit often spans vastly different geological units or fault zones, and the unit contains blocks with significantly different susceptibility levels. High-susceptibility and low-susceptibility characteristics cancel each other out, forcibly "homogenizing" the unit's attributes. This "geologically fragmented" delineation results in a spatial distribution of subsequent evaluations that is seriously inconsistent with geological reality, losing spatial specificity and failing to support high-precision targeted disaster prevention deployment. Summary of the Invention

[0006] The technical problem solved by this invention is to provide a micro-watershed unit division method based on susceptibility purity and water system topology dimensionality reduction, so as to solve the problem that the existing micro-watershed unit division technology lacks geological intrinsic constraints during evaluation, resulting in the distortion of the evaluation basis conditions.

[0007] The basic solution provided by this invention is a micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction, comprising: S1: Collect DEM data, geological data, historical geological disaster data and topographic factors of the target study area, and generate a multi-source raster dataset after preprocessing; S2: Based on historical geological disaster point data and geological data from multi-source raster datasets, the information content method is used to calculate the susceptibility contribution weight of each geological factor, and the susceptibility contribution weight is used to weight and superimpose each geological factor according to the weight, and generate a raster-level susceptibility level map covering the entire region according to the preset susceptibility level. S3: Based on DEM data from a multi-source raster dataset, extract the minimum catchment area and construct a tree-like topology of the water system; S4: Based on the tree-like topology of the water system, all minimum catchment areas are topologically sorted and linearly numbered in the order from upstream to downstream along the main river channel. The grid-level susceptibility level map is spatially overlaid with each minimum catchment area. The number of grids with susceptibility level in each catchment area is counted to form the susceptibility level combination vector of the catchment area. Finally, a one-dimensional attribute sequence to be optimized is formed. S5: A two-stage fusion algorithm based on susceptibility purity is used to perform two-stage optimization partitioning on one-dimensional attribute sequences, using susceptibility purity as the core fusion criterion to obtain the optimal segmentation scheme for one-dimensional attribute sequences of the entire study area's catchment area. S6: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to two-dimensional geographic space, merge the corresponding smallest catchment area into the final micro-watershed evaluation unit vector map layer, and attach the attribute information of each unit.

[0008] Furthermore, S2 includes: S2-1: Geological factors are obtained by spatial quantization and hierarchical discretization of geological data in multi-source raster data, combined with topographic-derived factors. S2-2: Based on historical geological disaster data and geological data, the susceptibility contribution value of each geological factor is calculated using the information content method. Specifically, for any geological factor in its k-th state, the information content value is... The expression is:

[0009] in, denoted as the number of historical geological disaster sites within the k-th state, and N represents the total number of geological disaster sites in the study area. Let S be the area of ​​the k-th state, and S be the total area of ​​the study area; S2-3: Weighted overlay of the information values ​​of all geological factors to generate a raster-level susceptibility comprehensive scoring layer covering the entire region, with the comprehensive score for each raster. for:

[0010] Where M represents the total number of geological factors involved in the evaluation. This represents the information content value of the i-th factor state corresponding to this grid. S2-4: Using the natural breakpoint method, the overall susceptibility score is divided into three levels: high susceptibility, medium susceptibility, and low susceptibility, and a raster susceptibility level map is generated.

[0011] Furthermore, S3 includes: S3-1: Extract DEM data from multi-source raster datasets and use a priority queue depression filling algorithm to fill depressions in the DEM data and eliminate false depressions; S3-2: For the DEM data after filling depressions, the D8 single flow direction algorithm is used to determine the water flow direction of each grid cell as the direction with the largest elevation drop among its 8 neighboring grid cells, and store it with direction encoding to output the flow direction grid map of the whole area. S3-3: Based on the flow direction grid map of the whole area, the cumulative weighted flow algorithm is used to calculate the cumulative flow of each grid. S3-4: Threshold segmentation of the flow accumulation of the grid is performed using a preset catchment area threshold. Grids with flow accumulation exceeding the preset catchment area threshold are marked as channel starting points. The smallest catchment area with complete funnel landform morphology in the entire study area is extracted. S3-5: Based on the starting point of the channel in S3-4, the water system of each level of the channel is traced along the raster map of the entire region. The complete water system network of the main channel and the tributary channel is extracted to generate a tree-like topology of the water system.

[0012] Furthermore, S4 includes: S4-1: Based on the tree-like topology of the water system constructed in step S3, all minimum catchment areas are topologically sorted and linearly numbered in order from upstream to downstream along the main river channel. This distributes the catchment areas, originally located in a two-dimensional geographic space, onto a one-dimensional linear sequence axis along the main river channel. The expression for this one-dimensional linear sequence is: , where n represents the total number of minimum catchment areas; S4-2: Spatially overlay the grid-level susceptibility map with each minimum catchment area, and statistically analyze each catchment area. The number of grid cells representing high, medium, and low susceptibility levels within the catchment area forms a susceptibility level composition vector for that catchment area. This constitutes a one-dimensional attribute sequence to be optimized. .

[0013] Furthermore, S5 includes: S5-1: Constructing a susceptibility purity calculation model is the core indicator for measuring the homogeneity of geological properties within a unit. The expression is:

[0014] in, Indicates susceptibility purity. This indicates the number of grid cells within a unit that represent the dominant susceptibility level. This indicates the number of grid cells within a unit that represent the secondary susceptibility level. The weighting of the secondary susceptibility level relative to the dominant level. This represents the total number of grid cells within a single cell. S5-2: Set purity threshold The purity after merging adjacent catchment areas If so, merging is allowed; when the purity is after merging... If the condition is not met, the merger will be rejected, and a unit boundary will be formed at that point. S5-3: The weight W is determined using an objective weighting scheme based on the information content ratio, expressed as follows:

[0015] in, The average information content value corresponding to the dominant susceptibility level. This represents the average information content value corresponding to the secondary susceptibility level. S5-4: A greedy algorithm is used to scan sequentially along the one-dimensional attribute sequence to strip away large-scale high- or low-risk geological entities, thereby obtaining a one-dimensional subsequence of solidified typical micro-watershed units and transitional zones to be optimized. S5-5: The dynamic programming algorithm is used to recursively optimize the one-dimensional subsequence of the transition zone to be optimized, and the optimal cutting point corresponding to each water catchment area is recorded. After the recursive optimization is completed, the optimal division scheme of the transition zone is obtained by backtracking the optimal cutting point. S5-6: Combine the results of the optimal segmentation scheme of the solidified typical micro-watershed unit and the transition zone to generate the optimal segmentation scheme of the one-dimensional attribute sequence of the watershed of the entire study area.

[0016] Furthermore, in S5-4, a greedy algorithm is used to sequentially scan along the one-dimensional attribute sequence, stripping away large-scale high- or low-risk geological entities to obtain a one-dimensional subsequence of solidified typical micro-watershed units and transitional zones to be optimized. Specifically: Perform a greedy scan from left to right along the one-dimensional attribute sequence to initialize the starting position of the current merging unit. ; from The subsequent catchment areas are merged sequentially, and the susceptibility purity of the merged units is calculated in real time. With area ; when and , The purity threshold, If a minimum unit area threshold is preset, then the next catchment area will be merged. If the above conditions are not met, The previous catchment area at the current location is solidified into an independent micro-watershed unit, and updated. For the current position, repeat the above steps until the one-dimensional attribute sequence ends; The solidified typical micro-watershed units and the remaining one-dimensional subsequence of the transition zone to be optimized were obtained.

[0017] Furthermore, in S5-5, a dynamic programming algorithm is used to recursively optimize the one-dimensional subsequence of the transition zone to be optimized, recording the optimal cutting point corresponding to each catchment area. After the recursive optimization is completed, the optimal cutting scheme of the transition zone is obtained by backtracking the optimal cutting point: The optimization objective is defined as minimizing the total cost J, and the cost function is constructed as follows:

[0018] in, This represents the total number of remaining micro-watershed units in the transition zone. Let m be the impurity rate of the m-th unit; This is the penalty coefficient; Build To minimize the total cost after optimally dividing the catchment areas of the first i transition zones, the recursive equation is:

[0019] in, The minimum total cost calculated for the first j catchment areas; The impurity rate after merging the (j+1)th to the ith water catchment areas into a single micro-watershed unit; The fixed cost incurred for this segmentation; The constraints are:

[0020] Initial conditions are ; This represents the total area of ​​the catchment area. The minimum unit area threshold is preset; Calculate sequentially according to the recursive equation. , , ..., , The total number of water catchment areas in the transition zone is recorded, and each is recorded. The corresponding optimal split point j is obtained by backtracking the split point after the recursion is completed, and the global optimal partitioning scheme of the transition region is obtained.

[0021] Furthermore, S6 includes: S6-1: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to the two-dimensional geographic space, and merge the corresponding smallest catchment areas to generate the final micro-watershed evaluation unit vector map layer. S6-2: Each micro-watershed evaluation unit in the micro-watershed evaluation unit vector layer includes basic attribute information, geological attribute information, and topographic attribute information.

[0022] The technical principle of this invention is as follows: This application uses the "minimum catchment area" as the indivisible basic physical unit, and introduces the grid-level geological hazard susceptibility assessment results as a prerequisite geological constraint into the catchment area fusion stage. The quantitative "susceptibility purity" index is used to dynamically determine the merging qualification of adjacent catchment areas. At the same time, by utilizing the natural tree-like topology of the water system, the traditional two-dimensional spatial aggregation problem is reduced to a one-dimensional sequential optimal segmentation problem along the main river channel. Through a two-stage fusion mechanism of "preliminary purification by greedy algorithm + global optimization by dynamic programming", the spatial homogeneity of the geological attributes of the micro-watershed unit is maximized under the premise of ensuring hydrological continuity and topographic integrity. Finally, a standardized micro-watershed assessment unit suitable for high-precision geological hazard assessment and early warning is generated.

[0023] Its core technological logic can be broken down into the following three mutually supporting key components: 1. Principle of Pre-embedding Geological Intrinsic Factors This approach overturns the traditional post-hoc model of "first dividing into units, then assigning geological attributes." It first completes a geological hazard susceptibility assessment at the grid scale, integrating multiple sources such as stratigraphy, lithology, tectonic faults, topography, and hydrology. Then, it uses the susceptibility level as the core constraint for merging catchment areas. This ensures that the unit division process is always guided by the inherent laws governing geological hazard occurrence, completely resolving the fundamental flaw of traditional methods that "fragment the geological environment," and ensuring that the geological attributes within the final micro-watershed units are highly homogeneous.

[0024] 2. Principle of One-Dimensional Dimensionality Reduction Topological Constraints Based on the tree-like topology of the water system, the minimum catchment area is linearly sequenced, and the forced merging operation can only be performed between adjacent numbered units along the main channel. This principle, on the one hand, completely eliminates the "jump merging across catchment areas" problem common in two-dimensional spatial aggregation, ensuring the hydrological continuity and geomorphological integrity of micro-watershed units; on the other hand, it transforms the NP-hard two-dimensional spatial segmentation problem into an efficient one-dimensional sequence partitioning problem, significantly reducing computational complexity and providing the possibility of finding the global optimum.

[0025] 3. Principle of Two-Order Algorithm Collaborative Optimization A two-stage architecture of "preliminary purification using a greedy algorithm + global optimization using dynamic programming" is adopted to balance partitioning efficiency and global optimality. The greedy algorithm quickly scans along a one-dimensional sequence to identify and solidify extreme units with pure properties (extremely high / extremely low susceptibility regions), which greatly reduces the subsequent solution space and improves the algorithm's running efficiency; The dynamic programming algorithm performs globally optimal segmentation of the attribute transition zone. By introducing an L0 penalty term to balance the purity and granularity of the unit, it achieves a balance between local geological specificity and global partitioning rationality. At the same time, by adjusting the penalty coefficient, multi-scale adaptive partitioning can be achieved.

[0026] The beneficial effects of this invention are as follows: This invention addresses the three core defects of existing micro-watershed unit division techniques one by one, resulting in significant technical effects: 1. It solves the problem of "lack of scale identification standards and strong subjectivity in the division results", and realizes standardized and repeatable micro-watershed division: This invention abandons the traditional method of relying on manual experience to set a single catchment area threshold, and instead adopts a two-parameter adaptive scaling mechanism of "susceptibility purity threshold + dynamic programming penalty coefficient". The susceptibility purity threshold ensures the homogeneity of geological attributes within the unit, while the penalty coefficient can flexibly adjust the unit granularity according to the scale of the study area and evaluation needs. The same study area can obtain completely consistent division results under the same parameters, demonstrating strong repeatability and enabling standardized application across regions, thus eliminating the result bias caused by differences in the experience of different researchers.

[0027] 2. It solved the problem of "lack of spatial dimensionality reduction mechanisms and abnormally fragmented unit morphology", and generated micro-watershed units with regular morphology and continuous hydrology: By employing one-dimensional dimensionality reduction of the water system topology and forcibly merging operations according to hydrological confluence patterns, this invention avoids the chaotic unit boundaries and fragmented morphology caused by feature redundancy in two-dimensional spatial aggregation. The micro-watershed units generated by this invention all possess complete funnel-shaped catchment landforms, without isolated fragments or cross-water system merging phenomena. Compared to traditional methods, the number of units is reduced by approximately 40%, and the average unit area is increased by approximately 60%, significantly reducing the computational burden of subsequent evaluation modeling and improving the model's convergence and robustness.

[0028] 3. It solves the problem of "lack of geological internal constraints leading to distorted evaluation conditions," and significantly improves the homogeneity of unit geological attributes: This invention embeds geological intrinsic factors into the entire process of segmentation beforehand, and uses susceptibility purity as the core fusion criterion to ensure a high degree of consistency in the susceptibility characteristics of geological disasters within the same micro-watershed unit. It effectively avoids the "attribute homogenization" problem in traditional methods where high-susceptibility and low-susceptibility characteristics cancel each other out within the same unit. This makes the spatial distribution of subsequent geological disaster susceptibility assessment highly consistent with the actual geological situation, improves the assessment accuracy, and provides a reliable foundation for high-precision targeted disaster prevention deployment. Attached Figure Description

[0029] Figure 1 This is a flowchart of an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the operation of an embodiment of the present invention. Detailed Implementation

[0030] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 and Figure 2 As shown: A micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction includes: S1: Collect DEM data, geological data, historical geological disaster data, and topographically derived factors of the target study area, and generate a multi-source raster dataset after preprocessing; S1 includes: S1-1: Collect the following four types of basic data according to the study area scale and evaluation accuracy requirements: 1. Digital Elevation Model (DEM) Data: Collect digital elevation model data covering the entire study area; the data resolution is determined based on the evaluation accuracy. Regional scale (>1000) The evaluation uses 30m resolution SRTM DEM or ALOS World 3D DEM data; County-level scale (100-1000) The evaluation used 12.5m resolution ALOS DEM data; Key area scale (<100) The evaluation uses DEM data generated by vectorizing a 5m resolution domestic Gaofen-7 DEM or a 1:10,000 topographic map.

[0031] The DEM data must contain complete elevation information and have no large areas of missing data or outliers.

[0032] 2. Geological data: Collect the latest regional geological survey results for the study area, including: A vector-format stratigraphic and lithological map with a scale of no less than 1:50,000, containing attribute fields such as stratigraphic code, geological age, lithological name, lithological description, and stratigraphic thickness; A vector-format geological structure map with a scale of no less than 1:50,000, containing attribute fields such as the location, strike, dip, angle, fracture zone width, activity age, and core and limb boundaries of folds at all levels. Optional supplementary hydrogeological borehole data, including borehole location, groundwater depth, and aquifer type, can be used to improve the accuracy of the evaluation.

[0033] 3. Historical geological disaster data: Vector data of geological hazard points verified on-site over the past 20 years were collected for the study area. Hazard types included landslides, debris flows, and collapses. Each hazard point must include attribute information such as the time of occurrence, location, scale, and cause type of the hazard. The number of historical geological hazard points should meet the statistical significance requirement, and in principle, should be no less than 30, with a relatively even distribution within the study area.

[0034] 4. Topographical factors: Based on the above DEM data, the following terrain-derived factors were derived using geographic information system software: slope, aspect, plane curvature, profile curvature, and topographic humidity index (TWI).

[0035] S1-2: Perform uniform, standardized preprocessing operations on all the collected multi-source data to ensure data spatial consistency and format compatibility: 1. Unified Space Reference System: All vector and raster data should be uniformly registered to the same spatial reference system. The WGS84 / UTM projection coordinate system is recommended, and the corresponding projection zone number should be selected according to the longitude of the study area. For raw data using other coordinate systems, accurate registration should be performed using the seven-parameter coordinate transformation method, with the registration error controlled within one raster.

[0036] 2. Uniform raster resolution: Resample all data to the same spatial resolution, consistent with the DEM data resolution: The vector data to raster conversion method uses the center attribute method, which takes the attribute value of the raster where the center point of the vector feature is located as the value of that raster. For continuous raster data (such as DEM and TWI), bilinear interpolation is used for resampling to ensure data continuity. For discrete raster data (such as lithology and tectonic influence levels), the nearest neighbor method is used for resampling to avoid distortion of attribute values.

[0037] 3. Study area cropping and invalid region handling: Based on the boundary vector layer of the study area, all raster data were cropped to generate a rectangular raster dataset covering the entire study area. For invalid areas in the cropped dataset, such as water bodies, permanent snow cover, and glaciers, which have no risk of geological disasters, a masking process was used to mark them as NoData, and they were not included in subsequent calculations.

[0038] 4. Normalization of continuous factors: For continuous factors such as slope and topographic humidity index, the minimum-maximum normalization method is used to process their values ​​and map them to... The interval eliminates the dimensional differences between different factors, and the normalization formula is:

[0039] In the formula, X is the original value of the factor. This is the minimum value of the factor within the study area. This represents the maximum value of the factor within the study area. These are the normalized factor values.

[0040] 5. Data consistency verification: All preprocessed raster data undergoes a consistency check to ensure that the number of rows and columns, spatial extent, and cell size of all layers are completely consistent. After the check passes, a multi-source raster dataset containing a DEM layer, a geological factor layer, and a terrain-derived factor layer is generated and stored in a common raster format (such as GeoTIFF).

[0041] S2: Based on historical geological disaster point data and geological data concentrated in a multi-source raster dataset, the susceptibility contribution weight of each geological factor is calculated using the information content method. This weight is then used to weight and superimpose the various geological factors to generate a raster-level susceptibility map covering the entire region according to a preset susceptibility level. S2 includes: S2-1: Based on the spatial quantization and hierarchical discretization of geological data in multi-source raster data, geological factors are obtained by combining topographic-derived factors; among which, geological factors include: 1. Stratigraphic Core Factor Generation: First, based on the shear strength, hydrological properties, weathering degree, and disaster risk level of the rock and soil, the subdivided stratigraphic units in the original stratigraphic lithology map are grouped into 5-7 major categories with similar disaster response characteristics, including hard rock groups, relatively hard rock groups, soft rock groups, loose sedimentary rock groups, and special disaster-causing rock groups. Then, using the merged lithological category codes as field values, the vector lithology map is converted into a raster layer with the same resolution as the DEM using the central attribute method. Finally, a unique non-zero integer state code is assigned to each lithological category (e.g., 1=hard rock group, 2=relatively hard rock group) as the input state for subsequent information calculation.

[0042] 2. Generation of Fault Structure Influence Factors: First, based on the scale, activity age, and degree of damage to the rock mass integrity of the faults, the faults in the study area were divided into three levels: Level I: regional active faults (length > 10 km, active since the Quaternary); Level II: secondary active faults (length 3-10 km, active since the Late Pleistocene); and Level III: general faults (length < 3 km, no obvious Quaternary activity). Then, buffer zones with corresponding radii were established for each level of fault. The buffer zone for Level I faults was 0-200 m, and for Level III, 2... The buffer zones for secondary faults are 0-500m and 500-1000m, respectively; the buffer zones for tertiary faults are 0-100m, 100-300m, and 300-500m; and the buffer zones for tertiary faults are 0-50m, 50-150m, and 150-300m. Following the principle that "the closer to the fault, the stronger the impact," each buffer zone is assigned an impact level code of 1-5, with larger values ​​indicating stronger fault impact. Finally, the maximum values ​​of the buffer zone raster layers for each level of fault are superimposed to obtain a raster map of the fault structure impact factors for the entire region.

[0043] 3. Generation of structural unit factors: First, based on the relationship between the characteristics of the tectonic stress field and the stability of the rock mass, the study area is divided into seven typical structural units: anticline core, anticline limb, syncline core, syncline limb, rift basin, fault-block mountain, and fault fracture zone. Then, the structural unit type code is used as the field value and converted into a raster layer with a uniform resolution. Finally, a unique non-zero integer state code is assigned to each structural unit type.

[0044] 4. Preprocessing of topographic factors: The slope, aspect, plane curvature, profile curvature, and topographic humidity index (TWI) derived in step S1 are included in the evaluation factor set. The aspect is discretized into 8 states according to 8 directions, and the other continuous topographic factors are discretized into 5 states according to the natural breakpoint method, and corresponding state codes are assigned.

[0045] S2-2: Based on historical geological disaster data and geological data, the susceptibility contribution value of each geological factor is calculated using the information content method. Specifically, for any geological factor in its k-th state, the information content value is... The expression is:

[0046] in, denoted as the number of historical geological disaster sites within the k-th state, and N represents the total number of geological disaster sites in the study area. Let S be the area of ​​the k-th state, and S be the total area of ​​the study area; The information value corresponding to each state of each factor is stored as an attribute table, and a one-to-one mapping relationship between state code and information value is established.

[0047] S2-3: Weighted superposition of the information values ​​of all geological factors to generate a raster-level comprehensive susceptibility scoring layer covering the entire region: First, traverse each valid grid in the study area and extract the state code of all evaluation factors corresponding to that grid. Then, based on the state code, the corresponding information value is retrieved, and all information values ​​are summed to obtain the overall susceptibility score of the raster. The expression is:

[0048] Where M represents the total number of geological factors involved in the evaluation. This represents the information content value of the i-th factor state corresponding to this grid. The overall score value of all rasters is stored as a raster layer, which yields the raster-level susceptibility overall score layer.

[0049] S2-4: Using the natural breakpoint method, the overall susceptibility score is divided into three levels: high susceptibility, medium susceptibility, and low susceptibility, generating a raster susceptibility level map. Specifically: Natural Breakpoint Grading: Based on the statistical distribution characteristics of the susceptibility comprehensive score, the Jenks natural breakpoint method is used to divide the comprehensive score into three intervals, maximizing the difference between intervals and minimizing the difference within intervals; Level assignment: The three scoring intervals are respectively assigned to three levels: high incidence, medium incidence, and low incidence, and a unique integer code is assigned to each level (e.g., 1=low incidence, 2=medium incidence, 3=high incidence). Final output: A raster-level susceptibility rating map (SGR) is generated, which serves as the core geological constraint layer for the subsequent susceptibility attribute statistics of the catchment area in step S4 and the fusion of the catchment area in step S5.

[0050] S3: Based on DEM data from a multi-source raster dataset, extract the minimum catchment area and construct a tree-like topology for the river system; S3 includes: S3-1: Extract DEM data from multi-source raster datasets, and use a priority queue depression-filling algorithm to fill depressions in the DEM data. This eliminates false depressions caused by data acquisition errors, interpolation biases, or terrain occlusion, ensuring continuous downstream water flow. Specifically: The algorithm works as follows: using the boundary grid of the study area as the initial seed point, all grids are processed in ascending order of elevation; for each grid to be processed, if its elevation is lower than the minimum elevation of the already processed grids in its 8-neighborhood, its elevation is raised to the minimum elevation, otherwise the original elevation is kept unchanged. Depression filling threshold setting: Set the minimum real depression depth threshold to 0.5m, only fill pseudo depressions with a depth less than this threshold, retain real catchment depressions with a depth ≥ 0.5m, and avoid overfilling depressions to avoid damaging the hydrological characteristics of the original topography; Output: Generates a smooth DEM raster layer without false depressions, which serves as the basis for subsequent flow direction calculations.

[0051] S3-2: For the DEM data after filling depressions, the D8 single-flow direction algorithm is used to determine the water flow direction of each grid cell as the direction with the largest elevation drop among its 8 neighboring grid cells, and store it with direction encoding to output a full-area flow direction grid map; In this embodiment, the specific process includes: Flow direction determination rule: For each grid cell, calculate the ratio of its elevation difference to the distance between it and its 8 neighboring grid cells (East, Southeast, South, Southwest, West, Northwest, North, Northeast), i.e., the slope; determine the water flow direction as the direction with the largest slope, i.e., the direction with the largest elevation drop; Directional encoding rules: 8 flow directions are represented by 8-bit binary encoding, with the following correspondence: East=1, Southeast=2, South=4, Southwest=8, West=16, Northwest=32, North=64, Northeast=128; Special case handling: For flat areas (where the elevation difference between 8 neighboring areas is 0), the steepest descent method combined with random perturbation is used to determine the flow direction to avoid the occurrence of grids without flow direction; Output: Generates a flow direction raster map with the same resolution as the original DEM, with each raster value being the corresponding flow direction code.

[0052] S3-3: Based on the flow direction grid map of the whole area, the cumulative weighted flow accumulation algorithm is used to calculate the cumulative flow accumulation of each grid. The calculation principle is as follows: the cumulative flow accumulation represents the total area of ​​all grids upstream of the grid along the flow direction; the initial cumulative flow accumulation of each grid is its own grid area, and then the cumulative flow accumulation of each grid is transferred to the downstream grid along the flow direction. The algorithm execution process is as follows: starting from the upstream source grid, the cumulative flow of each grid is calculated sequentially along the flow direction using a recursive traversal method until the outlet of the study area is reached; The output is: a raster map of the cumulative runoff, where the value of each raster is the cumulative runoff at that location. The larger the cumulative runoff, the stronger the water collection capacity at that location, and the more likely it is to develop a channel system.

[0053] S3-4: Threshold segmentation of the grid's runoff accumulation volume is performed using a preset catchment area threshold. Grids with runoff accumulation exceeding the preset catchment area threshold are marked as channel starting points. The smallest catchment area with a complete funnel landform morphology in the entire study area is extracted. The specific process includes: Catchment area threshold setting: Preset catchment area threshold The range of values ​​is The default value is 0.1. The value can be dynamically adjusted according to the topographic complexity of the study area: a smaller value is used for mountainous areas with steep terrain and dense valleys, and a larger value is used for hilly areas with gentle terrain and sparse valleys. Channel start point marking: Marking the confluence accumulation point with a flow rate greater than or equal to... The corresponding grid mark is the starting point of the channel, which is the location of the outlet of the smallest water collection area. Water catchment area extraction: Starting from the beginning of each channel, trace all upstream grids that flow into that point in the reverse direction of the flow to form a closed water catchment area with a unique outlet, which is a minimum water catchment area. Reasonableness check: Remove area smaller than The pseudo-catchment areas ensure that each minimum catchment area has a complete funnel-shaped water catchment landform and covers the entire study area without overlap or omission. Output: Generates a vector layer of the smallest water catchment area. Each smallest water catchment area contains a unique cell number, area, and outlet location attributes.

[0054] S3-5: Based on the starting point of the channel in S3-4, the water system of each level of the channel is traced along the raster map of the entire area, the complete water system network of the main channel and tributary channels is extracted, and a tree-like topology of the water system is generated; the specific process is as follows: Channel water system tracing: Starting from the starting point of each channel, trace downstream along the flow direction until it flows into other channels or the outlet of the study area, forming a complete channel line segment; Channel classification: The Strahler classification method is used to classify channels: the source channel is classified as level 1, the channel formed by the merger of two channels of the same level is classified by level 1, and the channel formed by the merger of channels of different levels is classified by the higher level. Topology Construction: A tree-like topology is constructed with the final outlet of the study area as the root node, and the following information is stored in the form of an adjacency list: A unique identifier for each minimum catchment area; The set of numbers for the immediate upstream catchment areas of each minimum catchment area; The immediate downstream catchment area of ​​each minimum catchment area is numbered; The Strahler sequence of channels corresponding to each minimum catchment area; Topology verification: Check the integrity and consistency of the topology to ensure that there are no errors such as loop confluence or isolated catchment areas; Output: Generates a tree-like topology data file of the water system, providing topological constraints for the one-dimensional reduction and serialization in the subsequent step S4.

[0055] S4: Based on the tree-like topology of the river system, all minimum catchment areas are topologically sorted and linearly numbered in order from upstream to downstream along the main river channel. The raster-level susceptibility level map is then spatially overlaid with each minimum catchment area. The number of raster cells representing the susceptibility level within each catchment area is counted, forming a susceptibility level combination vector for that catchment area. This ultimately constitutes a one-dimensional attribute sequence to be optimized. S4 includes: S4-1: Based on the tree-like topology of the water system constructed in step S3, all minimum catchment areas are topologically sorted and linearly numbered in order from upstream to downstream along the main river channel. The topological sorting rule is as follows: with the final outlet of the study area as the root node, a post-order traversal method is used to traverse all minimum catchment areas from the source to the outlet along the direction of water system confluence, ensuring that each catchment area is numbered only after all its upstream catchment areas have been traversed, strictly following the order of hydrological confluence.

[0056] During linear numbering, each traversed minimum catchment area is assigned a unique, consecutive integer number in chronological order, forming a one-dimensional linear sequence. , where n represents the total number of minimum catchment areas; The sequence constraints are as follows: the numbering must strictly increase along the direction of the main river channel, and skipping numbers across water systems or channels is prohibited. This ensures that the sequence is completely consistent with the hydrological confluence path and provides topological constraints for subsequent one-dimensional optimal segmentation.

[0057] S4-2: Spatially overlay the grid-level susceptibility map with each minimum catchment area, and statistically analyze each catchment area. The number of grid cells representing high, medium, and low susceptibility levels within the catchment area forms a susceptibility level composition vector for that catchment area. This constitutes a one-dimensional attribute sequence to be optimized. .

[0058] S5: A two-stage fusion algorithm based on susceptibility purity is used to perform two-stage optimization partitioning on the one-dimensional attribute sequence, employing susceptibility purity as the core fusion criterion to obtain the optimal segmentation scheme for the one-dimensional attribute sequence of the entire study area's catchment area; wherein, S5 includes: S5-1: Constructing a susceptibility purity calculation model is the core indicator for measuring the homogeneity of geological properties within a unit. The expression is:

[0059] in, Indicates susceptibility purity. This indicates the number of grid cells within a unit that represent the dominant susceptibility level. This indicates the number of grid cells within a unit that represent the secondary susceptibility level. The weighting of the secondary susceptibility level relative to the dominant level. This represents the total number of grid cells within a single cell. S5-2: Set purity threshold The purity of this application is 0.7, when adjacent catchment areas are merged. If so, merging is allowed; when the purity is after merging... If the condition is not met, the merger will be rejected, and a unit boundary will be formed at that point. S5-3: The weight W is determined using an objective weighting scheme based on the information content ratio, expressed as follows:

[0060] in, The average information content value corresponding to the dominant susceptibility level. This represents the average information content value corresponding to the secondary susceptibility level. S5-4: A greedy algorithm is used to sequentially scan along the one-dimensional attribute sequence, stripping away large-scale high- or low-risk geological entities to obtain a one-dimensional subsequence of solidified typical micro-watershed units and transitional zones to be optimized; specifically: Initialization: Perform a greedy scan from left to right along the one-dimensional attribute sequence to initialize the starting position of the current merged cell. ; Merge segments one by one: from The subsequent catchment areas are merged sequentially, and the susceptibility purity of the merged units is calculated in real time. With area ; Merge determination: when and , The purity threshold, To preset the minimum unit area threshold, this application sets it to 0.5. If so, then continue merging the next catchment area; Unit-based firmware update: If the above conditions are not met, then... The previous catchment area at the current location is solidified into an independent micro-watershed unit, and updated. For the current position, repeat the above steps until the one-dimensional attribute sequence ends; Output results: solidified typical microfluidic units and the remaining one-dimensional subsequence of the transition zone to be optimized are obtained.

[0061] S5-5: A dynamic programming algorithm is used to recursively optimize the one-dimensional subsequence of the transition zone to be optimized, recording the optimal cutting point corresponding to each catchment area. After the recursive optimization is completed, the optimal partitioning scheme of the transition zone is obtained by backtracking the optimal cutting point; specifically: The optimization objective is defined as minimizing the total cost J, and the cost function is constructed as follows:

[0062] in, This represents the total number of remaining micro-watershed units in the transition zone. Let m be the impurity rate of the m-th unit; This is a penalty coefficient used to balance unit purity and partitioning granularity; Build To minimize the total cost after optimally dividing the catchment areas of the first i transition zones, the recursive equation is:

[0063] in, The minimum total cost calculated for the first j catchment areas; The impurity rate after merging the (j+1)th to the ith water catchment areas into a single micro-watershed unit; The fixed cost incurred for this segmentation; The constraints are:

[0064] Initial conditions are ; This represents the total area of ​​the catchment area. The minimum unit area threshold is preset; Calculate sequentially according to the recursive equation. , , ..., , The total number of water catchment areas in the transition zone is recorded, and each is recorded. The corresponding optimal split point j is obtained by backtracking the split point after the recursion is completed, and the global optimal partitioning scheme of the transition region is obtained.

[0065] S5-6: The solidified typical micro-watershed units obtained in S5-4 are spliced ​​and integrated with the global optimal partitioning scheme of the transition zone obtained in S5-5 to generate the optimal one-dimensional attribute sequence partitioning scheme of the entire study area's catchment area that covers the entire study area and meets the requirements of hydrological continuity and geological homogeneity. This serves as the direct basis for the two-dimensional spatial mapping and unit merging in step S6.

[0066] S6: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to two-dimensional geographic space, merge the corresponding smallest catchment areas into a final micro-watershed evaluation unit vector map layer, and attach attribute information for each unit; S6 includes: S6-1: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to two-dimensional geographic space, and perform vector merging of the corresponding smallest catchment areas to generate the final micro-watershed evaluation unit vector map layer; specifically including: 1. Numbering Mapping and Matching: Establish a one-to-one correspondence between one-dimensional linear numbers and unique identifiers of two-dimensional minimum catchment areas. Based on the start and end numbers of the units recorded in the optimal segmentation scheme, group and mark all minimum catchment areas belonging to the same micro-watershed unit.

[0067] 2. Topological fusion and merging: Using the geographic information system vector fusion tool, the smallest catchment areas within the same group are merged without overlap, gaps, or gaps. Internal boundaries are automatically eliminated while external closed boundaries are preserved, forming micro-watershed evaluation units with complete geometric shape and continuous hydrology.

[0068] 3. Generation of global layer: Traverse all merged groups to complete the batch merging of micro-watershed units in the entire study area, and generate the final micro-watershed evaluation unit vector layer. The layer adopts the polygon feature format, covering the entire study area without redundant fragments.

[0069] S6-2: Each micro-watershed evaluation unit in the micro-watershed evaluation unit vector layer includes basic attribute information, geological attribute information, and topographic attribute information; the basic attribute information includes the unit number, unit area, coordinates of the unit's geometric center point, the number of the smallest catchment areas contained therein, and a list of their numbers; Geological attribute information includes susceptibility purity value, dominant susceptibility level, secondary susceptibility level, average geological lithology type, and average fault structure influence level; Topographic attribute information includes unit average slope, unit maximum slope, average topographic humidity index, dominant slope aspect, gully sequence and main gully orientation.

[0070] Based on the above technical solution, the highlights of this application are: 1. The first-ever "susceptibility purity" pre-fusion mechanism fundamentally solves the problem of homogenization distortion within the evaluation unit: This method uses the grid-level susceptibility evaluation results as a pre-geological constraint for merging water catchment areas, and uses a quantitative "purity" index to dynamically drive boundary generation, rather than the post-application superposition mode of first dividing and then adding attributes in the traditional method. This fundamentally ensures the spatial homogeneity of geological attributes within each evaluation unit and eliminates the evaluation distortion problem caused by "neutralization of high and low susceptibility". 2. One-dimensional dimensionality reduction based on water system topology completely avoids the "topological jump" and computational explosion problems of two-dimensional aggregation: This method reduces the traditional two-dimensional spatial aggregation problem to a one-dimensional linear sequence along the main channel. At the physical topology level, it forces the merging operation to be performed only between adjacent numbered units, eliminating illegal merging across water catchment areas. At the same time, it transforms the NP-hard two-dimensional spatial partitioning problem into a one-dimensional sequence partitioning problem that can be solved efficiently by dynamic programming, reducing the computational complexity from exponential to polynomial. 3. A combination of greedy and dynamic programming algorithms to balance local specificity and global optimality: In the first stage, the greedy algorithm quickly identifies and solidifies the extreme units with pure properties (extremely high or low susceptibility regions), reducing the solution space of subsequent dynamic programming; in the second stage, dynamic programming performs a globally optimal segmentation of the transition region sequence, constraining the number of units with an L0 penalty term, ensuring that the partitioning result achieves the optimal balance between maximizing purity and rationalizing unit granularity. The two-stage algorithms are interconnected and together constitute the computational core of this method.

[0071] Advantages and effects: 1. Significantly improves the accuracy of susceptibility assessment and the targeting of early warning: Since the division units have internal homogeneity in susceptibility, the dominant level label highly represents the geological attributes of the unit, and the performance indicators of evaluation models (such as random forest, logistic regression, etc.) such as AUC and F1 are significantly improved compared with the use of traditional micro-watershed units; at the same time, the spatial specificity of high-susceptibility units and low-susceptibility units is fully preserved, and the early warning information can accurately point to truly high-risk plots, avoiding missed warnings caused by "homogenization" dilution; 2. Effectively avoids unit fragmentation and improves modeling stability: The introduction of one-dimensional dimensionality reduction mechanism and minimum area threshold constraint suppresses excessive segmentation caused by local terrain fluctuations at the algorithm level, so that the division results maintain a reasonable unit granularity in geologically complex areas, providing high-quality and stable sample units for subsequent statistical modeling. 3. High versatility and can be embedded into existing GIS workflows: This method only requires standard DEM and raster-level susceptibility evaluation results as input, with no mandatory requirements on data types. It can adapt to DEMs of different precision and multiple types of geological data sources. The algorithm process can be packaged into a standardized tool and embedded into mainstream GIS platforms such as ArcGIS and QGIS. It is suitable for large-scale application and promotion in mountainous, hilly and other areas prone to geological disasters.

[0072] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for dividing micro-watershed units based on susceptibility purity and water system topology dimensionality reduction, characterized in that: include: S1: Collect DEM data, geological data, historical geological disaster data and topographic factors of the target study area, and generate a multi-source raster dataset after preprocessing; S2: Based on historical geological disaster point data and geological data from multi-source raster datasets, the information content method is used to calculate the susceptibility contribution weight of each geological factor, and the susceptibility contribution weight is used to weight and superimpose each geological factor according to the weight. A raster-level susceptibility level map covering the entire region is generated according to the preset susceptibility level classification standard. S3: Based on DEM data from a multi-source raster dataset, extract the minimum catchment area and construct a tree-like topology of the water system; S4: Based on the tree-like topology of the water system, all minimum catchment areas are topologically sorted and linearly numbered in the order from upstream to downstream along the main river channel. The grid-level susceptibility level map is spatially overlaid with each minimum catchment area. The number of grids with susceptibility level in each catchment area is counted to form the susceptibility level combination vector of the catchment area. Finally, a one-dimensional attribute sequence to be optimized is formed. S5: A two-stage fusion algorithm based on susceptibility purity is used to perform two-stage optimization partitioning on the one-dimensional attribute sequence, employing susceptibility purity as the core fusion criterion to obtain the optimal segmentation scheme for the one-dimensional attribute sequence of the entire study area's catchment area; S5 includes: S5-1: Constructing a susceptibility purity calculation model is the core indicator for measuring the homogeneity of geological properties within a unit. The expression is: in, Indicates susceptibility purity. This indicates the number of grid cells within a unit that represent the dominant susceptibility level. This indicates the number of grid cells within a unit that represent the secondary susceptibility level. The weighting of the secondary susceptibility level relative to the dominant level. This represents the total number of grid cells within a single cell. S5-2: Set purity threshold The purity after merging adjacent catchment areas If so, merging is allowed; when the purity is after merging... If the condition is not met, the merger will be rejected, and a unit boundary will be formed at that point. S5-3: The weight W is determined using an objective weighting scheme based on the information content ratio, expressed as follows: in, The average information content value corresponding to the dominant susceptibility level. This represents the average information content value corresponding to the secondary susceptibility level. S5-4: A greedy algorithm is used to scan sequentially along the one-dimensional attribute sequence to strip away large-scale high- or low-risk geological entities, thereby obtaining a one-dimensional subsequence of solidified typical micro-watershed units and transitional zones to be optimized. S5-5: The dynamic programming algorithm is used to recursively optimize the one-dimensional subsequence of the transition zone to be optimized, and the optimal cutting point corresponding to each water catchment area is recorded. After the recursive optimization is completed, the optimal division scheme of the transition zone is obtained by backtracking the optimal cutting point. S5-6: Combine the results of the optimal segmentation scheme of the solidified typical micro-watershed unit and the transition zone to generate the optimal segmentation scheme of the one-dimensional attribute sequence of the watershed of the entire study area. S6: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to two-dimensional geographic space, merge the corresponding smallest catchment area into the final micro-watershed evaluation unit vector map layer, and attach the attribute information of each unit.

2. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 1, characterized in that: S2 includes: S2-1: Based on the centralized geological data of multi-source raster dataset, spatial quantization and hierarchical discretization are performed, and geological factors are obtained by combining terrain-derived factors; S2-2: Based on historical geological disaster data and geological data, the susceptibility contribution weight of each geological factor is calculated using the information content method. Specifically, for any geological factor in its k-th state, its information content value... The expression is: in, denoted as the number of historical geological disaster sites within the k-th state, and N represents the total number of geological disaster sites in the study area. Let S be the area of ​​the k-th state, and S be the total area of ​​the study area; S2-3: Weighted overlay of the information values ​​of all geological factors to generate a raster-level susceptibility comprehensive scoring layer covering the entire region, with the comprehensive score for each raster. for: Where M represents the total number of geological factors involved in the evaluation. This represents the information content value of the i-th factor state corresponding to this grid. S2-4: Using the natural breakpoint method, the overall susceptibility score is divided into three levels: high susceptibility, medium susceptibility, and low susceptibility, generating a raster-level susceptibility level map.

3. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 2, characterized in that: S3 includes: S3-1: Extract DEM data from multi-source raster datasets and use a priority queue depression filling algorithm to fill depressions in the DEM data and eliminate false depressions; S3-2: For the DEM data after filling depressions, the D8 single flow direction algorithm is used to determine the water flow direction of each grid cell as the direction with the largest elevation drop among its 8 neighboring grid cells, and store it with direction encoding to output the flow direction grid map of the whole area. S3-3: Based on the flow direction grid map of the whole area, the cumulative weighted flow algorithm is used to calculate the cumulative flow of each grid. S3-4: Threshold segmentation of the flow accumulation of the grid is performed using a preset catchment area threshold. Grids with flow accumulation exceeding the preset catchment area threshold are marked as channel starting points. The smallest catchment area with complete funnel landform morphology in the entire study area is extracted. S3-5: Based on the starting point of the channel in S3-4, the water system of each level of the channel is traced along the raster map of the entire region. The complete water system network of the main channel and the tributary channel is extracted to generate a tree-like topology of the water system.

4. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 3, characterized in that: S4 includes: S4-1: Based on the tree-like topology of the water system constructed in step S3, all minimum catchment areas are topologically sorted and linearly numbered in order from upstream to downstream along the main river channel. This distributes the catchment areas, originally located in a two-dimensional geographic space, onto a one-dimensional linear sequence axis along the main river channel. The expression for this one-dimensional linear sequence is: , where n represents the total number of minimum catchment areas; S4-2: Spatially overlay the grid-level susceptibility map with each minimum catchment area, and statistically analyze each catchment area. The number of grid cells representing high, medium, and low susceptibility levels within the catchment area forms a susceptibility level combination vector for that catchment area. This constitutes a one-dimensional attribute sequence to be optimized. .

5. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 4, characterized in that: In S5-4, a greedy algorithm is used to scan sequentially along the one-dimensional attribute sequence, stripping away large-scale high- or low-risk geological entities to obtain a one-dimensional subsequence of solidified typical micro-watershed units and transitional zones to be optimized. Specifically: Perform a greedy scan from left to right along the one-dimensional attribute sequence to initialize the starting position of the current merging unit. ; from The subsequent catchment areas are merged sequentially, and the susceptibility purity of the merged units is calculated in real time. With area ; when and , The purity threshold, If a minimum unit area threshold is preset, then the next catchment area will be merged. If the above conditions are not met, The previous catchment area at the current location is solidified into an independent micro-watershed unit, and updated. For the current position, repeat the above steps until the one-dimensional attribute sequence ends; The solidified typical micro-watershed units and the remaining one-dimensional subsequence of the transition zone to be optimized were obtained.

6. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 5, characterized in that: In step S5-5, a dynamic programming algorithm is used to recursively optimize the one-dimensional subsequence of the transition zone to be optimized, recording the optimal cutting point corresponding to each catchment area. After the recursive optimization is completed, the optimal cutting scheme of the transition zone is obtained by backtracking the optimal cutting point: The optimization objective is defined as minimizing the total cost J, and the cost function is constructed as follows: in, This represents the total number of remaining micro-watershed units in the transition zone. Let m be the impurity rate of the m-th unit; This is the penalty coefficient; Build To minimize the total cost after optimally dividing the catchment areas of the first i transition zones, the recursive equation is: in, The minimum total cost calculated for the first j catchment areas; The impurity rate after merging the (j+1)th to the ith water catchment areas into a single micro-watershed unit; The fixed cost incurred for this segmentation; The constraints are: Initial conditions are ; This represents the total area of ​​the catchment area. The minimum unit area threshold is preset; Calculate sequentially according to the recursive equation. , , ..., , The total number of water catchment areas in the transition zone is recorded, and each is recorded. The corresponding optimal split point j is obtained by backtracking the split point after the recursion is completed, and the global optimal partitioning scheme of the transition region is obtained.

7. The micro-watershed unit partitioning method based on susceptibility purity and water system topology dimensionality reduction according to claim 6, characterized in that: S6 includes: S6-1: Map the optimal segmentation scheme of the one-dimensional attribute sequence of the entire study area's catchment area back to the two-dimensional geographic space, and merge the corresponding smallest catchment areas to generate the final micro-watershed evaluation unit vector map layer. S6-2: Each micro-watershed evaluation unit in the micro-watershed evaluation unit vector layer includes basic attribute information, geological attribute information, and topographic attribute information.

Citation Information

Patent Citations

  • Distributed hydrological model sub-basin division method considering multiple elements

    CN117725065A

  • Rapid geological susceptibility evaluation method based on evaluation units with different scales

    CN120104705A