Flash flood ecological restoration simulation method and system coupled with attenuation dynamics

Abstract

The present application relates to mountain torrent data processing technical field, especially to a kind of mountain torrent ecological restoration simulation method and system coupled with attenuation dynamics;Gradient boosting regression model coupled with attenuation dynamics provided by the present application breaks through the limitation that traditional static regression model only establishes mapping, successfully embeds the recovery inertia of ecological system in prediction logic, so that the prediction result no longer appears the discontinuous jump that violates physical common sense, truly reflects the continuity characteristics of ecological succession.Secondly, in view of the problem that the prior art cannot distinguish spatial heterogeneity, according to the disturbance zoning result of step S2, the recovery rate parameter β and the attenuation coefficient of heavy, light and non-disturbance area are differentially set, the fine simulation of zoning is realized, so that the model can accurately depict the different dynamics processes of deep water area recovery extremely slowly, shallow water area recovery faster and non-submerged area mainly driven by climate, and provide a more scientific and accurate technical means for post-disaster ecological restoration evaluation of mountain torrent.

Description

Technical Field

[0001] This invention relates to the field of flash flood data processing technology, and in particular to a method and system for simulating flash flood ecological restoration based on coupled attenuation dynamics. Background Technology

[0002] Flash floods are special types of torrents that occur in mountainous areas, caused by torrential rain, snowmelt, or other water sources. They are short-lived, rapidly rising and falling, carrying large amounts of mud, sand, and rocks, and possess immense destructive power. The analysis of the formation mechanisms, risk assessment, and response strategies of flash flood disasters have been a research hotspot in the international field of disaster science. The formation and evolution of flash flood risk is the result of a complex interplay of climate, geomorphology, and human activities. Current research has shifted from analyzing single natural factors to multi-dimensional comprehensive assessments, with particular attention to the synergistic effects of urbanization and climate change. Global warming is exacerbating flood disasters by altering precipitation patterns and the frequency of extreme events, while rapid urbanization, leading to natural surface hardening, green space fragmentation, and the expansion of impermeable surfaces, significantly alters surface hydrological processes, further amplifying the risk of flash floods. Current research mainly focuses on the impacts of climate change and human activities on floods, assessing urban and social flood risks and vulnerabilities, and improving urban flood control capabilities. Research on the long-term impacts of floods on the ecological environment, risk warnings, and differentiated prevention measures remains limited.

[0003] Existing technologies for constructing ecological prediction models to address the recovery process after flash flood disturbances typically use the same attenuation coefficient for all regions. However, in severely disturbed areas, vegetation root systems die, leading to extremely slow recovery; while in undisturbed areas, the impact may be limited to microclimate, resulting in rapid recovery. Using a single coefficient can lead to significant spatial prediction bias in the model. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for simulating flash flood ecological restoration based on coupled attenuation dynamics, which solves the problems existing in the prior art.

[0005] This invention provides a method for simulating flash flood ecological restoration based on coupled attenuation dynamics, the method comprising the following steps: S1: Obtain historical data for mountain torrent ecological restoration simulation; S2: Extract the flood inundation range and zon the disturbance zone of the study area; S3: Construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; Specifically, S3 refers to the overall architecture of the gradient boosting regression model of the coupled decay dynamics, which is a hybrid architecture, expressed as: ; In the formula, Let be the ecological and environmental indicators for year t. For the ecological and environmental indicators of year t-1, β is the theoretical value predicted by the gradient boosting regression model, and β is the partition adaptive recovery rate parameter. The expression for the gradient boosting regression model is: ; In the formula, Let t be the average precipitation in year t. Let be the average temperature of year t. Let be the average evaporation in year t. Encoding the interference region, Let t be the ecological degradation impact value in year t; The formula for calculating the ecological degradation impact value is as follows: ; In the formula, Let be the ecological degradation impact value for year t. The initial impact intensity of the flash flood event. The attenuation coefficient for the impact of flooding, Based on the base year, The initial impact values ​​are those from non-flood factors. The attenuation coefficient is the factor for non-flood impacts. S4; Train the gradient boosting regression model of the coupled decay dynamics and optimize the model hyperparameters; S5: Simulate flash flood ecological restoration based on the gradient boosting regression model of the coupled attenuation dynamics; Preferably, the calculation process of the partition adaptive recovery rate parameter β is as follows: Calculate the ecological deviation of the study area; The expression is: ; In the formula, This represents the historical average ecological index. These are measured ecological indicators after a flash flood. Ecological deviation; An ecological elastic potential function is constructed based on the ecological deviation of the study area; the expression is: ; In the formula, Let be the ecological elastic potential energy function, and k be the ecological stiffness coefficient; The dynamic recovery rate parameter is calculated based on the ecological elastic potential energy function; the expression is: ; In the formula, η is the dynamic recovery rate parameter; η is the energy conversion efficiency constant; γ is the damping coefficient. For partition weights; Boundary constraints and final values ​​are applied to the dynamic recovery rate parameter; The final assignment is obtained by compressing using the Sigmoid function, and the expression is: .

[0006] Preferably, S2 specifically comprises: S2.1: Acquire key image data for flood inundation range extraction and disturbance zoning; S2.2: Extract water body indexes based on the key image data; S2.3: Based on the water body index, perform refined extraction of the flood inundation range to obtain inundation area data; S2.4: Divide the flooded area data into disturbance zones, namely, a heavily disturbed zone, a lightly disturbed zone, and an undisturbed zone.

[0007] Preferably, in step S2.2, an improved dynamic normalized difference method is used to extract the water index from the key image data; specifically, the water index is calculated using the green band data and mid-infrared band data from the key image data, expressed as: ; In the formula, DMNDWI is the water quality index. The reflectance of the green band data, This represents the reflectance of mid-infrared band data.

[0008] Preferably, in step S2.3, a water index threshold is set to distinguish between water bodies and land. Post-disaster image data exceeding the water index threshold is initially identified as flooded area data, while other image data is identified as land data.

[0009] Preferably, in step S2.4, the water depth of the flooded area is first calculated using the following formula: ; In the formula, Let (x, y) be the water depth in the flooded area. For flood water level elevation, Elevation data in the DEM model; Then, the disturbance intensity index is calculated based on the water depth of the flooded area; The calculation formula is: ; In the formula, Let (x, y) be the perturbation intensity index. This represents the maximum water depth within the study area. Let (x, y) be the Euclidean distance from the main channel to the coordinates (x, y). This is the largest Euclidean distance within the study area. and These are the corresponding weight coefficients; Finally, based on the distribution of the disturbance intensity index, the study area is divided into a heavily disturbed area, a lightly disturbed area, and a non-disturbed area.

[0010] Preferably, The area is a severely disturbed zone. The area in question is designated as a slightly disturbed zone, while the other areas are designated as undisturbed zones.

[0011] Preferably, in step S4, historical data from 2000 to 2024 is selected as the input data source, with the t-th... One year of historical data is used as input, and the actual ecological variable observations in year t are used as the prediction target; The model hyperparameters include tree depth, learning rate, partition adaptive recovery rate parameter, flood impact attenuation coefficient, and non-flood impact attenuation coefficient. The optimization objective is to minimize the root mean square error (RMSE). The optimization process is as follows: several sets of initial parameters are randomly sampled in the parameter space, and the model is trained to establish a Gaussian process surrogate model. In each iteration, the Bayesian optimizer uses the Gaussian process surrogate model to predict the expected loss of the remaining parameter space and selects the parameter combination with the minimum expected loss for evaluation. The model is trained and the Gaussian process surrogate model is updated until the preset maximum number of iterations or loss convergence threshold is reached. Finally, the parameter combination that minimizes the RMSE is output.

[0012] Preferably, in step S5, meteorological forecast data for future years and fixed flood zoning data are input, the gradient boosting regression model of the coupled attenuation dynamics is run, and the spatial distribution map and recovery curve of ecological and environmental indicators for each year are output.

[0013] According to another aspect of the present invention, a flash flood ecological restoration simulation system based on coupled decay dynamics is provided. The system employs the aforementioned method for simulating flash flood ecological restoration based on coupled decay dynamics, and the system comprises: The data acquisition module acquires historical data for simulation of flash flood ecological restoration. The partitioning module is used to extract the flood inundation range and partition the disturbance area of ​​the study area. The model building module is used to construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; The model training module is used to train the gradient boosting regression model of the coupled decay dynamics and to optimize the model hyperparameters. The model prediction module is used to simulate flash flood ecological restoration based on the gradient boosting regression model of the coupled decay dynamics.

[0014] The present invention has the following technical effects: The gradient boosting regression model based on coupled decay dynamics provided by this invention breaks through the limitations of traditional static regression models that only establish mappings. It successfully embeds the recovery inertia of the ecosystem into the prediction logic, ensuring that the prediction results no longer exhibit discontinuous jumps that violate physical common sense, and truly reflects the continuous characteristics of ecological succession. Secondly, addressing the problem that existing technologies cannot distinguish spatial heterogeneity, this invention, based on the perturbation zoning results, differentiates the recovery rate parameter β and decay coefficient for severely, lightly, and undisturbed areas, achieving refined simulation of zoning. This allows the model to accurately characterize the differentiated dynamic processes of extremely slow recovery in deep water areas, faster recovery in shallow water areas, and climate-driven processes in non-submerged areas, providing a more scientific and accurate technical means for assessing ecological recovery after flash floods.

[0015] This invention decouples flooding and non-flooding factors when calculating the ecological degradation impact value. The flooding impact attenuation coefficient and the non-flooding impact attenuation coefficient in the formula correspond to the recovery characteristics of the disturbed and undisturbed areas, respectively. The flooding impact attenuation coefficient for severely disturbed areas is smaller, resulting in a slow decay of the first term, thus preserving a long-term ecological deficit in the prediction. Conversely, the non-flooding impact attenuation coefficient for undisturbed areas is larger, causing the second term to rapidly approach zero, reflecting short-term impacts. The final ecological degradation impact value is a linear superposition of the two impacts, considering both the long-term structural damage caused by floods and the short-term dynamic changes in the uninundated areas.

[0016] This embodiment addresses the problem that traditional linear models cannot accurately describe the nonlinear response of ecosystems under extreme disturbances by constructing a quadratic function relationship based on physical mechanisms. The scheme defines the ecological restoration process as the bottom of the potential well in the historical baseline state, and the ecological deviation caused by flooding as the displacement. By calculating the ecological elastic potential energy, the recovery energy stored in the system when it deviates from equilibrium is quantified. Simultaneously, it reveals the non-monotonicity of the β value: in the mildly disturbed zone, the system is in the elastic deformation stage, and the potential energy accumulates rapidly with increasing deviation, resulting in a corresponding increase in the recovery force β, indicating a strong self-repairing dynamic of the ecosystem. However, in the severely disturbed zone, the system enters the plastic deformation stage, and structural damage leads to a decrease in the stiffness coefficient k. Although the potential energy continues to increase, the effective energy available for recovery decreases, causing the recovery force β to decrease instead. This quadratic curve relationship accurately depicts the physical reality that mild damage is easily recoverable while severe damage is difficult to recover, avoiding the systematic error of traditional models that still predict high recovery rates in severely damaged zones. Furthermore, by introducing a damping correction factor γ to adjust the partition weights, the model can adaptively adjust the energy loss rate of different regions according to the perturbation partition results, so that the recovery process of the severely perturbated region exhibits significant stagnation characteristics, further improving the consistency between the simulation results and the actual ecological succession process. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a flash flood ecological restoration simulation method based on coupled attenuation dynamics provided in an embodiment of the present invention; Figure 2 This is a flowchart of the process for extracting the flood inundation range and zoning the disturbance zone of the study area, provided by an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0020] Example 1, Figure 1 A flowchart of a flash flood ecological restoration simulation method based on coupled attenuation dynamics is shown, as follows: Figure 1 As shown, a method for simulating flash flood ecological restoration based on coupled attenuation dynamics includes the following steps: S1: Obtain historical data for mountain torrent ecological restoration simulation; In this step, processed data from January 1, 2000 to December 31, 2025 are used as historical data. This period is chosen because it covers a long-term climate change cycle of nearly 25 years, and the Landsat series satellites have maintained good data continuity during this period, which can fully reflect the long-term steady-state relationship between climate and vegetation.

[0021] The meteorological factors of the historical data include precipitation, average temperature, evapotranspiration, and DEM data.

[0022] S2: Extract the flood inundation range and zon the disturbance zone of the study area; Traditional methods typically divide areas solely based on administrative or watershed boundaries, or simply use flood inundation extent as the sole partitioning method. This approach ignores the differences in the degree of damage to ecosystems caused by inundation duration and depth. For example, within the same inundated area, vegetation mortality rates are significantly higher in deeper water areas than in shallower areas, but traditional binary partitioning fails to reflect this gradient difference, leading to the inability of subsequent models to distinguish between severely damaged and lightly damaged areas. Therefore, this embodiment extracts flood inundation extent and performs disturbance partitioning based on recent historical data; specifically, as... Figure 2 As shown, S2 specifically refers to: S2.1: Acquire key image data for flood inundation range extraction and disturbance zoning; Obtain cloudless or partly cloudy Sentinel-2L2A multispectral images from the month preceding the flash flood of the previous year as baseline background data. Obtain Sentinel-2L2A images from the month following the flash flood of the previous year as post-disaster image data.

[0023] Although the L2A-level data has undergone atmospheric correction, relative radiometric normalization processing is required for the baseline background data and the post-disaster image data to eliminate differences in sensor response and ensure that the reflectance of the two images is comparable.

[0024] S2.2: Extract water body indexes based on the key image data; In this step, to overcome the problem of traditional NDWI easily confusing mountain shadows and built-up areas, this embodiment uses an improved dynamic normalized difference method to extract the water index from the key image data; specifically: The water index is calculated using the green band and mid-infrared band data from the key image data, and the expression is: ; In the formula, DMNDWI is the water quality index. The reflectance of the green band data, This represents the reflectance of mid-infrared band data.

[0025] It should be emphasized that in this step, water body index is extracted from the baseline background data and the post-disaster image data respectively. The water body index DMNDWI is two-dimensional data corresponding to the key image data. The water body index DMNDWI reflects the difference between high vegetation / water reflectance in the green band and strong water absorption and vegetation / soil reflectance in the mid-infrared band, thereby enhancing the water body signal and suppressing the land background.

[0026] S2.3: Based on the water body index, perform refined extraction of the flood inundation range to obtain inundation area data; A water index threshold is set to distinguish between water bodies and land. Post-disaster image data that exceeds the water index threshold is initially identified as flooded area data, while other image data is identified as land data. Then, the difference between the flooded area data from the post-disaster image data and the flooded area data from the baseline background data is calculated to obtain the flooded area data; only the newly added water bodies are retained as the flooded area of ​​the flash flood. This step excludes the area of ​​perennial river channels and reservoirs, and accurately locates the disaster-affected area.

[0027] S2.4: Divide the flooded area data into disturbance zones, namely, a heavily disturbed zone, a lightly disturbed zone, and an undisturbed zone; In this step, the water depth of the flooded area is first calculated using the following formula: ; In the formula, Let (x, y) be the water depth in the flooded area. For flood water level elevation, Elevation data in the DEM model.

[0028] Then, the disturbance intensity index is calculated based on the water depth of the flooded area; The calculation formula is: ; In the formula, Let (x, y) be the perturbation intensity index. This represents the maximum water depth within the study area. Let (x, y) be the Euclidean distance from the main channel to the coordinates (x, y). This is the largest Euclidean distance within the study area. and These are the corresponding weighting coefficients.

[0029] Finally, based on the distribution of the disturbance intensity index, the study area is divided into a heavily disturbed area, a lightly disturbed area, and a non-disturbed area.

[0030] in, The area is a severely disturbed zone. The area in question is designated as a slightly disturbed zone, while the other areas are designated as undisturbed zones.

[0031] S3: Construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; Existing technologies typically use the same attenuation coefficient k for all regions when constructing ecological prediction models to handle the recovery process after flash flood disturbances. However, vegetation roots die in severely disturbed areas, resulting in extremely slow recovery; while undisturbed areas may only be affected by microclimate and recover very quickly. Using a single coefficient can lead to significant spatial prediction bias in the model. Therefore, this step uses the partitioning results of severely disturbed, slightly disturbed, and undisturbed areas generated in step S2 as core constraints to construct a machine learning prediction model that integrates physical dynamics.

[0032] Specifically, S3 is as follows: The overall architecture of the gradient boosting regression model based on the coupled decay dynamics is a hybrid architecture, expressed as follows: ; In the formula, Let be the ecological and environmental indicators for year t. For the ecological and environmental indicators of year t-1, β is the theoretical value predicted by the gradient boosting regression model, and β is the partition adaptive recovery rate parameter.

[0033] The value of the partition adaptive recovery rate parameter is determined based on the perturbation partitioning result. If it is a heavily perturbated area, then β=0.15; if it is a lightly perturbated area, then β=0.35; and if it is an unperturbated area, then β=0.65.

[0034] As a preferred embodiment, setting the zonal adaptive recovery rate parameter β as a fixed constant, such as 0.15 for severely damaged areas and 0.35 for lightly damaged areas, ignores the nonlinear response characteristics of the ecosystem. That is, when the system is severely damaged, its self-repair ability is almost lost, and recovery should be extremely slow; while when the system is lightly damaged and close to equilibrium, the recovery force should be strongest. Based on this, this embodiment proposes an elastic potential energy-resilience model. The healthy state of the ecosystem is defined as the bottom of the potential well, and flood disturbances cause it to deviate from the bottom. The further it deviates, the higher the potential energy, but the resilience is not monotonically changing, but conforms to a nonlinear relationship of a quadratic function. Specifically, the calculation process of the zonal adaptive recovery rate parameter β is as follows: Calculate the ecological deviation of the study area; The expression is: ; In the formula, This represents the historical average ecological index. These are measured ecological indicators after a flash flood. This represents the degree of ecological deviation.

[0035] An ecological elastic potential function is constructed based on the ecological deviation of the study area; In this step, Hooke's Law is used to construct the potential energy function for an ecosystem deviating from equilibrium. A higher potential energy indicates a stronger desire for the study area to return to equilibrium; however, due to physical structural damage, the actual recovery efficiency is somewhat compromised. The expression is: ; In the formula, Let be the ecological elastic potential energy function, and k be the ecological stiffness coefficient. When the study area is a heavily disturbed area, k = 0.5; when the study area is a lightly disturbed area, k = 1.0; and when the study area is an undisturbed area, k = 2.0.

[0036] The dynamic recovery rate parameter is calculated based on the ecological elastic potential energy function. In this step, potential energy is converted into recovery rate, and a damping correction factor γ is introduced to simulate the drag effect of physical obstruction caused by flooding on the recovery rate. The expression is: ; In the formula, η is the dynamic recovery rate parameter; η is the energy conversion efficiency constant, an empirical value used to convert physical quantities into the 0-1 range, and in this embodiment, it is set to 0.3; γ is the damping coefficient, set to 2.0 in the heavily disturbed region, 1.0 in the lightly disturbed region, and 0.1 in the undisturbed region. The partition weights are set to 2.0 for the heavily disturbed region, 1.0 for the lightly disturbed region, and 0 for the undisturbed region. It should be emphasized that the above formula is set for the heavily disturbed region, the lightly disturbed region, and the undisturbed region. When the above formula is used to solve for the dynamic recovery rate parameter of the undisturbed region, the partition weight is 0, and the damping coefficient has no effect.

[0037] Boundary constraints and final values ​​are applied to the dynamic recovery rate parameter; To prevent the dynamic recovery rate parameter from exceeding the physically allowed range (0, 1), it is compressed using the Sigmoid function to obtain the final value, expressed as: .

[0038] This embodiment addresses the problem that traditional linear models cannot accurately describe the nonlinear response of ecosystems under extreme disturbances by constructing a quadratic function relationship based on physical mechanisms. The scheme defines the ecological restoration process as the bottom of the potential well in the historical baseline state, and the ecological deviation caused by flooding as the displacement. By calculating the ecological elastic potential energy, the recovery energy stored in the system when it deviates from equilibrium is quantified. Simultaneously, it reveals the non-monotonicity of the β value: in the mildly disturbed zone, the system is in the elastic deformation stage, and the potential energy accumulates rapidly with increasing deviation, resulting in a corresponding increase in the recovery force β, indicating a strong self-repairing dynamic of the ecosystem. However, in the severely disturbed zone, the system enters the plastic deformation stage, and structural damage leads to a decrease in the stiffness coefficient k. Although the potential energy continues to increase, the effective energy available for recovery decreases, causing the recovery force β to decrease instead. This quadratic curve relationship accurately depicts the physical reality that mild damage is easily recoverable while severe damage is difficult to recover, avoiding the systematic error of traditional models that still predict high recovery rates in severely damaged zones. Furthermore, by introducing a damping correction factor γ to adjust the partition weights, the model can adaptively adjust the energy loss rate of different regions according to the perturbation partition results, so that the recovery process of the severely perturbated region exhibits significant stagnation characteristics, further improving the consistency between the simulation results and the actual ecological succession process.

[0039] Furthermore, The value is predicted by the gradient boosting regression model, which is expressed as follows: ; In the formula, Let t be the average precipitation in year t. Let be the average temperature of year t. Let be the average evaporation in year t. Encoding the interference region, where, if it is a heavily disturbed region, then... If the value is 0, then it is a region of slight disturbance. If the value is 1, then if it is a non-disturbance region, then It is 2. Let t be the ecological degradation impact value for year t.

[0040] To address the problem that existing technologies cannot distinguish the different attenuation patterns between flooded and non-flooded areas, this embodiment proposes a specific calculation formula for the ecological attenuation impact value, as follows: ; In the formula, Let be the ecological degradation impact value for year t. The initial impact intensity of the flash flood event is dimensionless, representing the total amount of direct physical damage suffered by the severely / slightly disturbed area in the year of the flood, and serves as the basis for attenuation calculations. The flood impact attenuation coefficient controls the rate of ecological impact attenuation over time in areas of severe / slight disturbance. Based on the base year, The initial impact value of non-flood factors is dimensionless and represents the indirect impact or background ecological fluctuations experienced by the undisturbed area. The attenuation coefficient is the factor for non-flood impact.

[0041] Furthermore, the process for assigning the initial impact intensity of a flash flood event is as follows: For the severely disturbed and lightly disturbed areas delineated in step S2, each area is divided into several pixels. The measured ecological index value of each pixel in the severely disturbed or lightly disturbed area during the year of the flood is calculated and compared with the average value of the same period in the historical baseline period. Then, the ecological deviation of each pixel in the severely disturbed or lightly disturbed area is calculated. Finally, the ecological deviation of all pixels is calculated and normalized to obtain the initial impact intensity of the severely disturbed or lightly disturbed area. The process for assigning the initial impact value of non-flood factors is as follows: For the non-disturbed area, the deviation of similar ecological indicators caused by non-flood factors such as temperature anomaly, precipitation anomaly, and background drought index in the same year is normalized to obtain the initial impact intensity of the non-disturbed area.

[0042] The gradient boosting regression model with coupled decay dynamics provided in this embodiment breaks through the limitations of traditional static regression models that only establish mappings. It successfully embeds the recovery inertia of the ecosystem into the prediction logic, so that the prediction results no longer show discontinuous jumps that violate physical common sense, and truly reflect the continuous characteristics of ecological succession. Secondly, in view of the problem that existing technologies cannot distinguish spatial heterogeneity, this invention sets the recovery rate parameter β and decay coefficient differently for severely disturbed, lightly disturbed and undisturbed areas based on the disturbance partitioning results in step S2. This achieves refined simulation of partitioning, enabling the model to accurately characterize the differentiated dynamic processes of extremely slow recovery in deep water areas, faster recovery in shallow water areas, and climate-driven non-submerged areas. This provides a more scientific and accurate technical means for ecological recovery assessment after flash floods.

[0043] In this embodiment, the flood and non-flood factors are decoupled when calculating the ecological degradation impact value. The flood and non-flood impact degradation coefficients in the formula correspond to the recovery characteristics of the disturbed and undisturbed areas, respectively. The flood impact degradation coefficient value corresponding to the severely disturbed area is smaller, resulting in a slow decay of the first term, thus retaining a long-term ecological deficit in the prediction. The non-flood impact degradation coefficient value corresponding to the undisturbed area is larger, resulting in a rapid approach to zero of the second term, reflecting the short-term impact. The final ecological degradation impact value is a linear superposition of the two impacts, taking into account both the long-term structural damage caused by floods and the short-term dynamic changes in the non-inundated area.

[0044] S4; Train the gradient boosting regression model of the coupled decay dynamics and optimize the model hyperparameters; In this step, historical data from 2000 to 2024 is selected as the input data source. This period covers the complete climate cycle and possible disturbance events, ensuring that the gradient boosting regression model of coupled decay dynamics can learn the ecological evolution patterns under normal conditions. (The last sentence appears to be incomplete and possibly refers to a specific data point or time period.) One year of historical data is used as input, and the actual ecological variable observation value in year t is used as the prediction target.

[0045] The model hyperparameters include tree depth, learning rate, partition adaptive recovery rate parameter, flood impact attenuation coefficient, and non-flood impact attenuation coefficient. The optimization objective is to minimize the root mean square error (RMSE). The optimization process is as follows: several sets of initial parameters are randomly sampled in the parameter space, and the model is trained to establish a Gaussian process surrogate model. In each iteration, the Bayesian optimizer uses the Gaussian process surrogate model to predict the expected loss of the remaining parameter space and selects the parameter combination with the minimum expected loss for evaluation. The model is trained and the Gaussian process surrogate model is updated until the preset maximum number of iterations or loss convergence threshold is reached. Finally, the parameter combination that minimizes the RMSE is output.

[0046] S5: Simulate flash flood ecological restoration based on the gradient boosting regression model of the coupled attenuation dynamics; Input meteorological forecast data for future years and fixed flood zoning data, run the gradient boosting regression model with coupled attenuation dynamics, and output the spatial distribution map and recovery curve of ecological and environmental indicators for each year.

[0047] Example 2: This invention also provides a flash flood ecological restoration simulation system based on coupled decay dynamics. The system employs a flash flood ecological restoration simulation method based on coupled decay dynamics from Example 1. The system includes: The data acquisition module acquires historical data for simulation of flash flood ecological restoration. The partitioning module is used to extract the flood inundation range and partition the disturbance area of ​​the study area. The model building module is used to construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; The model training module is used to train the gradient boosting regression model of the coupled decay dynamics and to optimize the model hyperparameters. The model prediction module is used to simulate flash flood ecological restoration based on the gradient boosting regression model of the coupled decay dynamics.

[0048] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.

[0049] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0050] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and a processor may execute the program instructions to implement the coupled decay dynamics-based flash flood ecological restoration simulation method described above in any embodiment of this application, and / or other desired functions. Various contents such as initial extrinsic parameters and thresholds may also be stored in the computer-readable storage medium.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for simulating flash flood ecological restoration based on coupled attenuation dynamics, characterized in that, The method includes the following steps: S1: Obtain historical data for mountain torrent ecological restoration simulation; S2: Extract the flood inundation range and categorize the study area into three types: heavily disturbed area, lightly disturbed area, and no disturbed area. S3: Construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; Specifically, S3 refers to the overall architecture of the gradient boosting regression model of the coupled decay dynamics, which is a hybrid architecture, expressed as: ； In the formula, Let be the ecological and environmental indicators for year t. For the ecological and environmental indicators of year t-1, β is the theoretical value predicted by the gradient boosting regression model, and β is the partition adaptive recovery rate parameter. The calculation process for the partition adaptive recovery rate parameter β is as follows: Calculate the ecological deviation of the study area; The expression is: ; In the formula, This represents the historical average ecological index. These are measured ecological indicators after a flash flood. Ecological deviation; An ecological elastic potential function is constructed based on the ecological deviation of the study area; the expression is: ； In the formula, Let be the ecological elastic potential energy function, and k be the ecological stiffness coefficient; The dynamic recovery rate parameter is calculated based on the ecological elastic potential energy function; the expression is: ； In the formula, η is the dynamic recovery rate parameter; η is the energy conversion efficiency constant; γ is the damping coefficient. For partition weights; Boundary constraints and final values ​​are applied to the dynamic recovery rate parameter; The final assignment is obtained by compressing using the Sigmoid function, and the expression is: ； The expression for the gradient boosting regression model is: ； In the formula, Let t be the average precipitation in year t. Let be the average temperature of year t. Let be the average evaporation in year t. Encoding the interference region, Let t be the ecological degradation impact value in year t; The formula for calculating the ecological degradation impact value is as follows: ； In the formula, Let be the ecological degradation impact value for year t. The initial impact intensity of the flash flood event. The attenuation coefficient for the impact of flooding, Based on the base year, The initial impact values ​​are those from non-flood factors. The attenuation coefficient is the factor for non-flood impacts. S4: Train the gradient boosting regression model of the coupled decay dynamics and optimize the model hyperparameters; in S4, historical data from 2000 to 2024 are selected as the input data source, with the t-th... One year of historical data is used as input, and the actual ecological variable observations in year t are used as the prediction target; The model hyperparameters include tree depth, learning rate, partition adaptive recovery rate parameter, flood impact attenuation coefficient, and non-flood impact attenuation coefficient. The optimization objective is to minimize the root mean square error (RMSE). The optimization process is as follows: Several sets of initial parameters are randomly sampled in the parameter space, and the model is trained to establish a Gaussian process surrogate model. In each iteration, the Bayesian optimizer uses the Gaussian process surrogate model to predict the expected loss of the remaining parameter space and selects the parameter combination with the minimum expected loss for evaluation. The model is trained and the Gaussian process surrogate model is updated until the preset maximum number of iterations or loss convergence threshold is reached. Finally, the parameter combination that minimizes the RMSE is output. S5: Simulate the ecological restoration of flash floods based on the gradient boosting regression model of the coupled decay dynamics.

2. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 1, characterized in that, Specifically, S2 is: S2.1: Acquire key image data for flood inundation range extraction and disturbance zoning; S2.2: Extract water body indexes based on the key image data; S2.3: Based on the water body index, perform refined extraction of the flood inundation range to obtain inundation area data; S2.4: Divide the flooded area data into disturbance zones, namely, a heavily disturbed zone, a lightly disturbed zone, and an undisturbed zone.

3. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 2, characterized in that, In step S2.2, an improved dynamic normalized difference method is used to extract the water body index from the key image data; specifically, the water body index is calculated using the green band data and mid-infrared band data from the key image data, expressed as: ； In the formula, DMNDWI is the water quality index. The reflectance of the green band data, This represents the reflectance of mid-infrared band data.

4. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 2, characterized in that, In S2.3, a water index threshold is set to distinguish between water bodies and land. Post-disaster image data that exceeds the water index threshold is initially identified as flooded area data, while other image data is identified as land data.

5. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 2, characterized in that, In step S2.4, the water depth of the flooded area is first calculated using the following formula: ； In the formula, Let (x, y) be the water depth in the flooded area. For flood water level elevation, Elevation data in the DEM model; Then, the disturbance intensity index is calculated based on the water depth of the flooded area; The calculation formula is: ； In the formula, Let (x, y) be the perturbation intensity index. This represents the maximum water depth within the study area. Let (x, y) be the Euclidean distance from the main channel to the coordinates (x, y). This is the largest Euclidean distance within the study area. and These are the corresponding weight coefficients; Finally, based on the distribution of the disturbance intensity index, the study area is divided into a heavily disturbed area, a lightly disturbed area, and a non-disturbed area.

6. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 5, characterized in that, The area is a severely disturbed zone. The area in question is designated as a slightly disturbed zone, while the other areas are designated as undisturbed zones.

7. The method for simulating flash flood ecological restoration based on coupled attenuation dynamics according to claim 1, characterized in that, In step S5, meteorological forecast data for future years and fixed flood zoning data are input, the gradient boosting regression model of the coupled attenuation dynamics is run, and the spatial distribution map and recovery curve of ecological and environmental indicators for each year are output.

8. A simulation system for flash flood ecological restoration based on coupled attenuation dynamics, characterized in that, The system employs a flash flood ecological restoration simulation method based on coupled attenuation dynamics as described in any one of claims 1-7, and the system comprises: The data acquisition module acquires historical data for simulation of flash flood ecological restoration. The partitioning module is used to extract the flood inundation range and partition the disturbance area of ​​the study area. The model building module is used to construct a gradient boosting regression model with coupled decay dynamics based on the perturbation partitioning results; The model training module is used to train the gradient boosting regression model of the coupled decay dynamics and to optimize the model hyperparameters. The model prediction module is used to simulate flash flood ecological restoration based on the gradient boosting regression model of the coupled decay dynamics.

Images