An intelligent soil erosion deduction method and system based on AIGC technology

Abstract

The application discloses a kind of water and soil loss intelligent deduction method and system based on AIGC technology, including step one: the water and soil loss data of target area is collected;Step two: water and soil loss data is preprocessed;Step three: water and soil loss generation model is generated to generate water and soil loss deduction data;Step four: water and soil loss scene generation module is updated by dynamic feedback mechanism;Step five: through the improved spatiotemporal graph convolution network, the soil erosion modulus in future preset period is deduced;Step six: multi-scenario water and soil loss intelligent deduction is executed, and water and soil loss deduction data under different scenarios is generated;Step seven: water and soil loss risk grade is divided, and uncertainty quantification analysis is carried out using Dropout and Bootstrap sampling based on Monte Carlo.The application improves the credibility of water and soil loss deduction result, multi-scenario adaptability and spatial expression precision.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and prediction technology, and in particular to an intelligent simulation method and system for soil and water loss based on AIGC technology. Background Technology

[0002] With the widespread deployment of remote sensing technology, UAV mapping platforms, and ground monitoring networks in the field of soil and water conservation, soil erosion monitoring and prediction based on multi-source observation data has received increasing attention. Existing research typically relies on a combination of satellite remote sensing inversion, ground sensor observations, and numerical simulation models to assess and extrapolate soil erosion processes. However, in practical applications involving complex terrain, extreme climates, and data-sparse areas, the following problems still commonly exist:

[0003] First, while traditional physical mechanism-based erosion models possess clear interpretability, they heavily rely on continuous boundary conditions, refined parameter inputs, and extensive experimental data. In regions with limited observational conditions, model robustness is difficult to guarantee, and the computational demands and slow response times of model derivation make it difficult to meet the requirements of high-frequency dynamic prediction.

[0004] Second, while purely data-driven deep learning models can fit certain statistical regularities when there are abundant samples, their generalization ability is weak, and they are prone to significant biases when dealing with unseen landforms, rainfall patterns, or human disturbance scenarios. Furthermore, these models typically lack the ability to express the physical constraints of erosion processes, making it difficult for the extrapolated results to remain consistent with actual hydrodynamic processes.

[0005] Third, existing studies often treat "data generation" and "state extrapolation" separately, lacking a unified modeling framework between data augmentation, missing data completion, and erosion state prediction. The synthetic data generated by generative models often cannot establish a closed-loop correction mechanism with real observations, resulting in a deviation between the generated scenario and the actual physical process, thus affecting the credibility of the extrapolation results.

[0006] Fourth, multi-source observation data exhibit significant heterogeneity in terms of spatiotemporal resolution, sampling density, and semantic structure. Remote sensing imagery offers wide coverage but has limited temporal resolution, UAVs offer high spatial resolution but have long flight cycles, and ground sensors have high sampling frequencies but limited spatial coverage. The difficulty in achieving a unified representation across different data sources makes it challenging to balance the depiction of macroscopic trends with the detailed representation of fine-scale erosion in the extrapolation results.

[0007] The aforementioned shortcomings make it difficult for existing technologies to achieve rapid, robust, and highly reliable prediction of soil erosion processes under scenarios such as sudden heavy rainfall, complex terrain disturbances, or engineering disturbances, thus failing to meet the urgent needs for forward-looking early warning and intelligent simulation in soil and water conservation management. Therefore, how to provide a method and system for intelligent soil erosion simulation based on AIGC technology is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] One objective of this invention is to propose an intelligent soil erosion simulation method and system based on AIGC technology. This invention comprehensively utilizes generative artificial intelligence models, spatiotemporal graph convolutional networks, and multi-source observation data fusion technology to construct an intelligent soil erosion simulation system capable of self-learning, self-correction, and cross-scenario simulation. By introducing multi-scale feature fusion, conditional generation models, dynamic feedback update mechanisms, and graph convolutional simulation structures with embedded physical equation constraints, this invention can continuously output highly reliable soil erosion simulation results under conditions of sparse observation data, complex terrain, or abnormal climate, effectively improving simulation accuracy and stability, enhancing model adaptability and physical consistency, and achieving highly reliable soil erosion prediction under multiple scenarios.

[0009] A method for intelligent simulation of soil erosion based on AIGC technology according to an embodiment of the present invention includes the following steps:

[0010] Step 1: Collect soil and water loss data for the target area;

[0011] Step 2: Perform spatiotemporal alignment and standardization on the soil erosion data to obtain a multi-scale spatiotemporal feature vector sequence, and use an attention-based feature fusion network to perform weighted fusion of the multi-scale spatiotemporal feature vector sequence to generate a spatiotemporal fusion feature tensor.

[0012] Step 3: Construct a soil erosion generation model based on AIGC technology. The soil erosion generation model includes a soil erosion scene generation module and a discrimination module. Input the spatiotemporal fusion feature tensor into the soil erosion scene generation module to generate soil erosion simulation data, and conduct adversarial training between the real observation data and the soil erosion simulation data through the discrimination module.

[0013] Step 4: Establish a dynamic feedback mechanism to update the soil erosion scenario generation module online and obtain an optimized soil erosion generation model;

[0014] Step 5: Input the soil erosion projection data and the physically corrected observation features into the improved spatiotemporal graph convolutional network to generate the soil erosion modulus for the future preset time period, and output the soil erosion intensity layer and spatial distribution layer. The improved spatiotemporal graph convolutional network introduces physical constraint terms.

[0015] Step 6: Introduce scenario-driven factors and fuse them with spatiotemporal fusion feature tensors. Input the fusion into the optimized soil and water loss generation model, perform multi-scenario intelligent soil and water loss simulation, generate soil and water loss simulation data under different scenarios, and output soil and water loss intensity layer sequence and soil and water loss evolution trend map.

[0016] Step 7: Map the soil erosion intensity layer sequence to a geospatial layer sequence and classify the soil erosion risk level; use Monte Carlo-based Dropout and Bootstrap sampling to perform uncertainty quantification analysis.

[0017] Optionally, step one specifically includes:

[0018] Soil and water loss data of the target area are collected collaboratively by satellite remote sensing platform, UAV aerial survey system and ground sensor network. The soil and water loss data includes high-resolution multispectral imagery, lidar point cloud, soil moisture, rainfall intensity, surface roughness and vegetation coverage.

[0019] Optionally, step two specifically includes:

[0020] Atmospheric and geometric corrections were performed on the high-resolution multispectral images to extract the normalized vegetation index, slope, aspect, and soil brightness index.

[0021] The lidar point cloud is reconstructed in three dimensions to generate a digital surface model, and the surface roughness features and fine-scale vegetation cover features are calculated based on the moving window method.

[0022] Outlier removal, unit standardization, and timestamp synchronization were performed on soil moisture, rainfall intensity, surface roughness, and vegetation cover to obtain a standardized hydrophysical quantity sequence.

[0023] Using spatial grid cells and time steps as indices, normalized vegetation index, slope, aspect and soil brightness index, surface roughness characteristics, fine-scale vegetation cover characteristics and standardized hydrophysical quantity sequences are resampled and interpolated according to spatial location and time label to obtain multi-scale feature vector sequences.

[0024] A multi-scale feature vector sequence is input into an attention-based feature fusion network. Attention weights are assigned to feature vectors from different data sources and weighted fusion is performed to generate a spatiotemporal fusion feature tensor.

[0025] Optionally, step three specifically includes:

[0026] The soil erosion scenario generation module adopts a conditional variational autoencoder structure, which learns the potential distribution of the soil erosion process and generates soil erosion inference data.

[0027] The encoder of the soil erosion scene generation module includes four downsampling stages. Each downsampling stage extracts local spatial features from the spatiotemporal fusion feature tensor through a 3×3 convolutional layer to obtain a downsampled feature map. The spatial dimension of the downsampled feature map is reduced through a 2×2 max pooling layer. The number of channels of the encoder increases exponentially with each downsampling stage, from the initial 64 to 128, 256 and 512, to enhance the semantic expressive power of deep features.

[0028] The decoder of the soil erosion scene generation module is set with four upsampling stages. In each upsampling stage, the spatial resolution is gradually restored through transposed convolutional layers. Skip connections are used to concatenate the downsampled feature maps of each downsampling stage of the encoder with the feature maps of the corresponding upsampling stages to generate upsampled fusion feature maps, which are then used as the input feature maps for the next upsampling stage.

[0029] The discrimination module adopts a multi-scale convolutional neural network structure, including a first-scale discrimination branch, a second-scale discrimination branch, and a third-scale discrimination branch; each scale discrimination branch is composed of several 3×3 convolutional layers, batch normalization layers, and LeakyReLU activation layers stacked sequentially, and different downsampling times are set in each scale discrimination branch;

[0030] Original resolution images, 2x downsampled images, and 4x downsampled images were constructed for real observation data and soil erosion projection data, respectively.

[0031] The first scale discrimination branch receives the original resolution image of the real observation data and the original resolution image of the soil erosion inference data, respectively, outputs the first scale real discrimination feature map and the first scale generated discrimination feature map, and uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain the first scale real response map and the first scale generated response map.

[0032] The second-scale discrimination branch receives a 2x downsampled image of the real observation data and a 2x downsampled image of the soil erosion inference data, respectively, and outputs a second-scale real discrimination feature map and a second-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a second-scale real response map and a second-scale generated response map.

[0033] The third-scale discrimination branch receives a 4x downsampled image of the real observation data and a 4x downsampled image of the soil erosion inference data, respectively, and outputs a third-scale real discrimination feature map and a third-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a third-scale real response map and a third-scale generated response map.

[0034] The loss function for adversarial training includes least squares loss and feature matching loss;

[0035] The least squares loss is calculated by: calculating the expected value of the squared difference between the true response map at each of the three scales and 1, and the expected value of the squared difference of the generated response map at each of the three scales, and then averaging and summing them.

[0036] The feature matching loss is the expectation of the element-wise absolute difference between the real discriminant feature maps and the generated discriminant feature maps at three scales, and then summed.

[0037] Optionally, the dynamic feedback mechanism specifically includes: setting a sliding time window, continuously receiving real-time observation data, and calculating the root mean square error between the soil erosion simulation data and the real-time observation data as a performance loss function; based on the performance loss function, using an adaptive moment estimation optimization algorithm to update the parameters of the soil erosion scenario generation module, with an initial learning rate of 0.001, which decays exponentially with the training rounds or error convergence.

[0038] Optionally, step five specifically includes:

[0039] Based on soil erosion data, a physical correction observation feature was constructed using a general soil loss equation. The physical correction observation feature includes rainfall erosivity factor, soil erodibility factor, and elevation.

[0040] The soil erosion simulation data is mapped into a graph node feature matrix, and a graph structure relationship matrix is ​​constructed based on terrain connectivity.

[0041] Based on physically corrected observation characteristics, physical constraint terms are constructed, specifically as follows:

[0042] ;

[0043] in, Represents physical constraint terms. As a soil erodibility factor, For elevation, Indicates elevation gradient, As the erosivity factor of rainfall, The constraint strength coefficient, Represents the square of the L2 norm;

[0044] The improved spatiotemporal graph convolutional network introduces physical constraint terms, specifically:

[0045] The physical constraint terms are concatenated with the graph node feature matrix in the feature dimension to form an extended node feature matrix; the graph structure relation matrix is ​​normalized to obtain the terrain connectivity normalized adjacency matrix; and graph convolution is performed based on the terrain connectivity normalized adjacency matrix and the extended node feature matrix to obtain the updated graph node feature matrix.

[0046] Based on the updated graph node feature matrix, temporal convolution is used for time modeling to obtain the predicted hidden feature matrix for a future preset time period; the future preset time period includes several prediction time steps, and each prediction time step corresponds to a predicted hidden feature vector.

[0047] For each prediction time step, the predicted hidden feature vector is linearly mapped to generate the soil erosion modulus, and the soil erosion modulus is rearranged into a raster form according to the correspondence between graph nodes and spatial grids to obtain the soil erosion intensity layer and spatial distribution layer for the current prediction time step.

[0048] Optionally, step six specifically includes:

[0049] The scenario-driving factors include climate scenario parameters, land use change parameters, and engineering intervention parameters.

[0050] The climate scenario parameters include temperature and rainfall changes; the land use change parameters represent the spatial proportion changes of different land types; the engineering intervention parameters include spatial configuration quantitative indicators for vegetation restoration, terrace construction, and silt-trapping dam deployment; and the scenario driving factors are reduced to 10 principal components through principal component analysis.

[0051] The scenario-driven factors and the spatiotemporal fusion feature tensor after dimensionality reduction are concatenated in the feature dimension to form the scenario condition feature tensor. The scenario condition feature tensor is then input into the optimized soil and water loss scenario generation module to generate soil and water loss inference data for the corresponding scenario.

[0052] The soil erosion projection data is rearranged according to spatial grid units and time series to construct a soil erosion intensity layer sequence. The soil erosion intensity layer sequence contains multiple prediction time steps, and each prediction time step corresponds to a soil erosion intensity raster layer.

[0053] Based on the soil erosion intensity layer sequence, the soil erosion evolution trend index is calculated. The soil erosion evolution trend index includes the spatial cumulative erosion, the erosion center offset trajectory and the expansion rate of high-risk areas, and a soil erosion evolution trend map corresponding to the scenario is generated.

[0054] The sequence of soil erosion intensity layers and the trend map of soil erosion evolution are visualized.

[0055] Optionally, step seven specifically includes:

[0056] The soil erosion intensity raster layer in the soil erosion intensity layer sequence is mapped to the corresponding geographic coordinate system according to the spatial grid unit to generate a geospatial layer.

[0057] Set a threshold for the risk level of soil and water loss, classify the soil and water loss risk level according to the threshold for the soil and water loss intensity value corresponding to each spatial grid unit, and label the level within each threshold range based on the soil erosion modulus value to generate a soil and water loss risk level map.

[0058] The soil and water loss risk levels include extremely low risk, low risk, medium risk, high risk and extremely high risk;

[0059] During the inference process, the Dropout layer of the soil erosion generation model is kept active. N forward inferences are performed on the same model input to obtain N sets of soil erosion simulation data. The mean and variance of the N sets of soil erosion simulation data are calculated on each spatial grid cell, and confidence intervals are constructed based on the mean and variance to generate a Monte Carlo Dropout confidence interval map.

[0060] Bootstrap resampling is performed on the soil erosion data to obtain several sampled datasets; inference is performed on the several sampled datasets to obtain several sets of soil erosion inference data, and the distribution of the several sets of soil erosion inference data on each spatial grid cell is statistically analyzed, quantiles are calculated, and Bootstrap confidence interval plots are generated.

[0061] Spatial visualization of soil erosion risk level map, Monte Carlo Dropout confidence interval map, and Bootstrap confidence interval map.

[0062] According to an embodiment of the present invention, a soil and water loss intelligent simulation system based on AIGC technology includes:

[0063] Soil and water loss data acquisition module: used to collect soil and water loss data in the target area;

[0064] Spatiotemporal feature construction module: used to preprocess soil erosion data and generate spatiotemporal fusion feature tensors through a feature fusion network based on an attention mechanism;

[0065] AIGC Soil and Water Loss Modeling Module: Used to build soil and water loss generation models and generate soil and water loss projection data;

[0066] Dynamic feedback update module: used to update the soil and water loss generation model online;

[0067] Physical constraint graph convolutional extrapolation module: used to extrapolate future time periods from soil erosion extrapolation data and physically corrected observation features, and generate soil erosion modulus;

[0068] Multi-scenario simulation module: used to construct scenario condition feature tensors and input them into the optimized soil and water loss scenario generation module to generate soil and water loss simulation data for the corresponding scenarios;

[0069] Risk and Uncertainty Analysis Module: Used to classify soil erosion risk levels and employ Monte Carlo Dropout and Bootstrap sampling to perform uncertainty quantification analysis on soil erosion projection data.

[0070] The beneficial effects of this invention are:

[0071] First, this invention achieves a unified expression of multi-scale features of soil erosion by collaboratively acquiring and aligning multi-source heterogeneous observation data and constructing a spatiotemporal fusion feature tensor using a feature fusion network based on an attention mechanism. This solves problems such as inconsistent spatiotemporal resolution and heterogeneous semantic information between remote sensing images, UAV images, and sensor data.

[0072] Secondly, this invention utilizes a soil erosion generation model based on AIGC technology, applies spatiotemporal fusion feature tensors to conditional generative inference, and constructs an adversarial training system between real observation data and soil erosion inference data through a discrimination module. This effectively improves the authenticity, spatial detail representation ability, and adaptability to complex landforms of soil erosion inference data in areas with scarce data or observation gaps.

[0073] Furthermore, this invention establishes a dynamic feedback mechanism to achieve online updates to the soil erosion scenario generation module. This allows the soil erosion generation model to continuously correct its generation distribution based on real-time observation data, improving the dynamic adaptability of the model under scenarios such as sudden rainfall and rapid runoff changes. Simultaneously, by incorporating an improved spatiotemporal graph convolutional network with introduced physical constraints, the generated soil erosion modulus maintains the expressive power of the deep model while conforming to the physical laws of the general soil loss equation, thus ensuring the stability, physical consistency, and interpretability of the simulation results.

[0074] In summary, this invention enables intelligent simulation of soil erosion under multiple scenarios through scenario-driven factors, and employs Monte Carlo Dropout and Bootstrap sampling for uncertainty quantification analysis. It can output highly reliable soil erosion intensity layer sequences and soil erosion evolution trend maps under different climate, land use, and engineering intervention scenarios, achieving highly timely, robust, and reliable intelligent simulation of soil erosion processes under complex scenarios. It has significant engineering application value and promotional significance. Attached Figure Description

[0075] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0076] Figure 1 This is a schematic diagram of a method and system for intelligent simulation of soil and water loss based on AIGC technology proposed in this invention;

[0077] Figure 2 This invention presents a method for intelligent simulation of soil erosion based on AIGC technology and a flowchart of the soil erosion generation model structure in the system.

[0078] Figure 3 This invention presents a method for intelligent inference of soil erosion based on AIGC technology and a flowchart of the improved spatiotemporal graph convolutional network structure in the system.

[0079] Figure 4 This invention presents an intelligent simulation method for soil erosion based on AIGC technology, and a flowchart of intelligent simulation of soil erosion under multiple scenarios in the system. Detailed Implementation

[0080] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0081] refer to Figures 1-4 A method for intelligent simulation of soil erosion based on AIGC technology includes the following steps:

[0082] Step 1: Collect soil and water loss data for the target area;

[0083] Step 2: Perform spatiotemporal alignment and standardization on the soil erosion data to obtain a multi-scale spatiotemporal feature vector sequence, and use an attention-based feature fusion network to perform weighted fusion of the multi-scale spatiotemporal feature vector sequence to generate a spatiotemporal fusion feature tensor.

[0084] Step 3: Construct a soil erosion generation model based on AIGC technology. The soil erosion generation model includes a soil erosion scene generation module and a discrimination module. Input the spatiotemporal fusion feature tensor into the soil erosion scene generation module to generate soil erosion simulation data, and conduct adversarial training between the real observation data and the soil erosion simulation data through the discrimination module.

[0085] Step 4: Establish a dynamic feedback mechanism to update the soil erosion scenario generation module online and obtain an optimized soil erosion generation model;

[0086] Step 5: Input the soil erosion projection data and the physically corrected observation features into the improved spatiotemporal graph convolutional network to generate the soil erosion modulus for the future preset time period, and output the soil erosion intensity layer and spatial distribution layer. The improved spatiotemporal graph convolutional network introduces physical constraint terms.

[0087] Step 6: Introduce scenario-driven factors and fuse them with spatiotemporal fusion feature tensors. Input the fusion into the optimized soil and water loss generation model, perform multi-scenario intelligent soil and water loss simulation, generate soil and water loss simulation data under different scenarios, and output soil and water loss intensity layer sequence and soil and water loss evolution trend map.

[0088] Step 7: Map the soil erosion intensity layer sequence to a geospatial layer sequence and classify the soil erosion risk level; use Monte Carlo-based Dropout and Bootstrap sampling to perform uncertainty quantification analysis.

[0089] In this invention, step five quantitatively extrapolates the future soil erosion trend under a single scenario. Its core involves inputting the soil erosion extrapolation data and physically corrected observation features into an improved spatiotemporal graph convolutional network that incorporates physical constraints. By combining spatial topology and physical priors, it generates soil erosion moduli for a predetermined future time period, outputting refined soil erosion intensity layers and spatial distribution layers to achieve high-precision simulation of short-term evolution trends. Step six focuses on multi-scenario simulation, introducing scenario-driving factors and fusing them with spatiotemporal fusion feature tensors before inputting them into an optimized soil erosion generation model. Multi-scenario simulation is then performed, generating soil erosion extrapolation data under multiple conditions. The output includes a sequence of soil erosion intensity layers and an evolution trend map covering multiple hypothetical scenarios, aiming to reveal the impact of different external conditions on the soil erosion process.

[0090] In this embodiment, step one specifically includes:

[0091] Soil and water loss data of the target area are collected collaboratively by satellite remote sensing platform, UAV aerial survey system and ground sensor network. The soil and water loss data includes high-resolution multispectral imagery, lidar point cloud, soil moisture, rainfall intensity, surface roughness and vegetation coverage.

[0092] In this embodiment, step two specifically includes:

[0093] Atmospheric and geometric corrections were performed on the high-resolution multispectral images to extract the normalized vegetation index, slope, aspect, and soil brightness index.

[0094] The lidar point cloud is reconstructed in three dimensions to generate a digital surface model, and the surface roughness features and fine-scale vegetation cover features are calculated based on the moving window method.

[0095] Outlier removal, unit standardization, and timestamp synchronization were performed on soil moisture, rainfall intensity, surface roughness, and vegetation cover to obtain a standardized hydrophysical quantity sequence.

[0096] Using spatial grid cells and time steps as indices, normalized vegetation index, slope, aspect and soil brightness index, surface roughness characteristics, fine-scale vegetation cover characteristics and standardized hydrophysical quantity sequences are resampled and interpolated according to spatial location and time label to obtain multi-scale feature vector sequences.

[0097] A multi-scale feature vector sequence is input into an attention-based feature fusion network. Attention weights are assigned to feature vectors from different data sources and weighted fusion is performed to generate a spatiotemporal fusion feature tensor.

[0098] In this embodiment, step three specifically includes:

[0099] The soil erosion scenario generation module adopts a conditional variational autoencoder structure, which learns the potential distribution of the soil erosion process and generates soil erosion inference data.

[0100] The encoder of the soil erosion scene generation module includes four downsampling stages. Each downsampling stage extracts local spatial features from the spatiotemporal fusion feature tensor through a 3×3 convolutional layer to obtain a downsampled feature map. The spatial dimension of the downsampled feature map is reduced through a 2×2 max pooling layer. The number of channels of the encoder increases exponentially with each downsampling stage, from the initial 64 to 128, 256 and 512, to enhance the semantic expressive power of deep features.

[0101] The decoder of the soil erosion scene generation module is set with four upsampling stages. In each upsampling stage, the spatial resolution is gradually restored through transposed convolutional layers. Skip connections are used to concatenate the downsampled feature maps of each downsampling stage of the encoder with the feature maps of the corresponding upsampling stages to generate upsampled fusion feature maps, which are then used as the input feature maps for the next upsampling stage.

[0102] The discrimination module adopts a multi-scale convolutional neural network structure, including a first-scale discrimination branch, a second-scale discrimination branch, and a third-scale discrimination branch; each scale discrimination branch is composed of several 3×3 convolutional layers, batch normalization layers, and LeakyReLU activation layers stacked sequentially, and different downsampling times are set in each scale discrimination branch;

[0103] Original resolution images, 2x downsampled images, and 4x downsampled images were constructed for real observation data and soil erosion projection data, respectively.

[0104] The first scale discrimination branch receives the original resolution image of the real observation data and the original resolution image of the soil erosion inference data, respectively, outputs the first scale real discrimination feature map and the first scale generated discrimination feature map, and uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain the first scale real response map and the first scale generated response map.

[0105] The second-scale discrimination branch receives a 2x downsampled image of the real observation data and a 2x downsampled image of the soil erosion inference data, respectively, and outputs a second-scale real discrimination feature map and a second-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a second-scale real response map and a second-scale generated response map.

[0106] The third-scale discrimination branch receives a 4x downsampled image of the real observation data and a 4x downsampled image of the soil erosion inference data, respectively, and outputs a third-scale real discrimination feature map and a third-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a third-scale real response map and a third-scale generated response map.

[0107] The loss function for adversarial training includes least squares loss and feature matching loss;

[0108] The least squares loss is calculated by: calculating the expected value of the squared difference between the true response map at each of the three scales and 1, and the expected value of the squared difference of the generated response map at each of the three scales, and then averaging and summing them.

[0109] The feature matching loss is the expectation of the element-wise absolute difference between the real discriminant feature maps and the generated discriminant feature maps at three scales, and then summed.

[0110] In this embodiment, the dynamic feedback mechanism specifically includes: setting a sliding time window, continuously receiving real-time observation data, and calculating the root mean square error between the soil erosion simulation data and the real-time observation data as a performance loss function; based on the performance loss function, using an adaptive moment estimation optimization algorithm to update the parameters of the soil erosion scenario generation module, with an initial learning rate of 0.001, which decays exponentially with the training rounds or error convergence.

[0111] In this embodiment, step five specifically includes:

[0112] Based on soil erosion data, a physical correction observation feature was constructed using a general soil loss equation. The physical correction observation feature includes rainfall erosivity factor, soil erodibility factor, and elevation.

[0113] The soil erosion simulation data is mapped into a graph node feature matrix, and a graph structure relationship matrix is ​​constructed based on terrain connectivity.

[0114] Based on physically corrected observation characteristics, physical constraint terms are constructed, specifically as follows:

[0115] ;

[0116] in, Represents physical constraint terms. As a soil erodibility factor, For elevation, Indicates elevation gradient, As the erosivity factor of rainfall, This is the constraint strength coefficient;

[0117] The improved spatiotemporal graph convolutional network introduces a physical constraint term to guide the network in updating graph node features, making the prediction results more consistent with the actual soil erosion mechanism. This physical constraint term is used to construct physical residual signals during graph convolution operations and serves as an additional input for updating graph node features. This ensures that the graph node features are constrained by the coupling relationship between topographic gradient, soil erodibility, and rainfall erosion force during spatial propagation. Specifically:

[0118] The physical constraint terms are concatenated with the graph node feature matrix in the feature dimension to form an extended node feature matrix; the graph structure relation matrix is ​​normalized to obtain the terrain connectivity normalized adjacency matrix; and graph convolution is performed based on the terrain connectivity normalized adjacency matrix and the extended node feature matrix to obtain the updated graph node feature matrix.

[0119] Based on the updated graph node feature matrix, temporal convolution is used for time modeling to obtain the predicted hidden feature matrix for a future preset time period; the future preset time period includes several prediction time steps, and each prediction time step corresponds to a predicted hidden feature vector.

[0120] For each prediction time step, the predicted hidden feature vector is linearly mapped to generate the soil erosion modulus, and the soil erosion modulus is rearranged into a raster form according to the correspondence between graph nodes and spatial grids to obtain the soil erosion intensity layer and spatial distribution layer for the current prediction time step.

[0121] In this embodiment, step six specifically includes:

[0122] The scenario-driving factors include climate scenario parameters, land use change parameters, and engineering intervention parameters.

[0123] The climate scenario parameters include temperature and rainfall changes; the land use change parameters represent the spatial proportion changes of different land types; the engineering intervention parameters include spatial configuration quantitative indicators for vegetation restoration, terrace construction, and silt-trapping dam deployment; and the scenario driving factors are reduced to 10 principal components through principal component analysis.

[0124] The scenario-driven factors and the spatiotemporal fusion feature tensor after dimensionality reduction are concatenated in the feature dimension to form the scenario condition feature tensor. The scenario condition feature tensor is then input into the optimized soil and water loss scenario generation module to generate soil and water loss inference data for the corresponding scenario.

[0125] The soil erosion projection data is rearranged according to spatial grid units and time series to construct a soil erosion intensity layer sequence. The soil erosion intensity layer sequence contains multiple prediction time steps, and each prediction time step corresponds to a soil erosion intensity raster layer.

[0126] Based on the soil erosion intensity layer sequence, the soil erosion evolution trend index is calculated. The soil erosion evolution trend index includes the spatial cumulative erosion, the erosion center offset trajectory and the expansion rate of high-risk areas, and a soil erosion evolution trend map corresponding to the scenario is generated.

[0127] The sequence of soil erosion intensity layers and the trend map of soil erosion evolution are visualized.

[0128] In this embodiment, step seven specifically includes:

[0129] The soil erosion intensity raster layer in the soil erosion intensity layer sequence is mapped to the corresponding geographic coordinate system according to the spatial grid unit to generate a geospatial layer.

[0130] A threshold for soil erosion risk level is set. The soil erosion intensity value corresponding to each spatial grid cell is then classified into soil erosion risk levels according to this threshold. Based on the soil erosion modulus value, the levels are labeled within each threshold range to generate a soil erosion risk level map. The soil erosion risk levels include extremely low risk, low risk, medium risk, high risk, and extremely high risk.

[0131] During the inference process, the Dropout layer of the soil erosion generation model is kept active. N forward inferences are performed on the same model input to obtain N sets of soil erosion simulation data. The mean and variance of the N sets of soil erosion simulation data are calculated on each spatial grid cell, and confidence intervals are constructed based on the mean and variance to generate a Monte Carlo Dropout confidence interval map.

[0132] Bootstrap resampling is performed on the soil erosion data to obtain several sampled datasets; inference is performed on the several sampled datasets to obtain several sets of soil erosion inference data, and the distribution of the several sets of soil erosion inference data on each spatial grid cell is statistically analyzed, quantiles are calculated, and Bootstrap confidence interval plots are generated.

[0133] Spatial visualization of soil erosion risk level map, Monte Carlo Dropout confidence interval map, and Bootstrap confidence interval map.

[0134] A smart soil erosion simulation system based on AIGC technology includes:

[0135] Soil and water loss data acquisition module: used to collect soil and water loss data in the target area;

[0136] Spatiotemporal feature construction module: used to preprocess soil erosion data and generate spatiotemporal fusion feature tensors through a feature fusion network based on an attention mechanism;

[0137] AIGC Soil and Water Loss Modeling Module: Used to build soil and water loss generation models and generate soil and water loss projection data;

[0138] Dynamic feedback update module: used to update the soil and water loss generation model online;

[0139] Physical constraint graph convolutional extrapolation module: used to extrapolate future time periods from soil erosion extrapolation data and physically corrected observation features, and generate soil erosion modulus;

[0140] Multi-scenario simulation module: used to construct scenario condition feature tensors and input them into the optimized soil and water loss scenario generation module to generate soil and water loss simulation data for the corresponding scenarios;

[0141] Risk and Uncertainty Analysis Module: Used to classify soil erosion risk levels and employ Monte Carlo Dropout and Bootstrap sampling to perform uncertainty quantification analysis on soil erosion projection data.

[0142] Example 1

[0143] To verify the feasibility of this invention in practice, the method was applied to a typical watershed in a hilly area. This watershed has significant topographic relief, with slopes mostly ranging from 15° to 35°, a long rainy season, and vegetation cover highly dependent on artificial restoration. This region consistently faces challenges such as sparse observational data, long UAV aerial survey cycles, and significant cloud interference from satellite remote sensing. Traditional soil erosion monitoring methods in such scenarios generally suffer from large extrapolation biases, slow response times, and an inability to characterize fine-scale erosion details. This invention primarily evaluates its performance under complex geomorphic conditions, insufficient observational data, and significant climate change scenarios.

[0144] In practical deployment, 10-meter resolution multispectral imagery is acquired via a satellite remote sensing platform, and over 20 high-resolution images from UAV aerial surveys conducted over the past three years are collected. Simultaneously, ground sensors provide real-time observation data on rainfall intensity, soil moisture, and runoff. Due to the low temporal resolution of satellite data, the discontinuous coverage of UAV data, and significant spatiotemporal scale differences among multiple data sources, this invention first utilizes an attention mechanism fusion network to unify these data into a single spatiotemporal fusion feature tensor, enabling it to simultaneously express spatial texture features and temporal dynamic trends. In contrast, traditional methods typically use only a single data source or simple alignment methods, making it difficult to accurately reflect erosion conditions during heavy rain events or human disturbances.

[0145] In this scenario, the soil erosion generation model constructed based on AIGC technology is used to supplement missing erosion scene data. For example, in canyon areas that cannot be covered by UAV aerial surveys, the AIGC generation model can generate soil erosion projection images consistent with the actual terrain. The authenticity is verified by a discrimination module, ensuring that the slope aspect texture and erosion gully morphology of the generated scene are highly consistent with actual observations. The generation model introduces a multi-scale discrimination structure in adversarial training, enabling it to capture multi-level information from micro-scale gully erosion texture to macro-scale slope erosion distribution, greatly improving data consistency and detail fidelity.

[0146] To adapt to extreme rainfall events, this invention maintains a dynamic feedback mechanism during actual operation, enabling the model to automatically update parameters upon receiving real-time sensor data. For example, in a short-duration rainstorm of 112 mm / h, real-time monitoring showed that the soil moisture at the top of the slope reached saturation within 25 minutes. The model immediately incorporated this change into its parameter updates, ensuring that the generated scenario and subsequent simulation results match the actual erosion dynamics during the rainstorm. Traditional static models often take several hours to complete a resimulation, making it difficult to meet the needs of rapid early warning. To ensure the physical reliability of the simulation results, this invention inputs both physically corrected observation features and soil erosion simulation data into a spatiotemporal graph convolutional network that incorporates physical constraints. This allows the model to automatically follow soil loss mechanisms when predicting future erosion trends. For example, during a sudden surge in rainfall intensity, the model automatically enhances the influence of the physical constraints, causing the predicted curve to exhibit a steep increase consistent with the measured data, avoiding the smoothing prediction problem that is common in pure deep learning models.

[0147] To further evaluate the adaptability of this invention under future changing scenarios, a multi-scenario simulation mechanism was constructed. Different climate change parameters, land use change parameters, and engineering measure parameters were set to simulate the soil erosion modulus change trends under three typical scenarios: continuous rainfall increase, arable land expansion, and artificial vegetation restoration, over the next five years. The results show that this invention can stably output erosion evolution processes with consistent physical logic under various scenarios, enabling users to identify high-risk areas in advance and optimize the layout of engineering measures.

[0148] To verify the improvement effect of this invention in practical applications, 120 typical slopes in the same watershed were selected as test samples. This invention was compared and analyzed with the USLE-R physical model and a CNN-based deep learning method. The comparison indicators included the average error of the erosion modulus, spatial matching degree, accuracy of fine-scale gully erosion identification, extrapolation response time, stability under data missing conditions, and uncertainty interval width. The experimental results are shown in Table 1.

[0149] Table 1. Performance Comparison Results of the Invention and the Comparative Scheme in Soil and Water Erosion Simulation Task

[0150]

[0151] As shown in Table 1, the present invention significantly outperforms the comparative schemes in multiple comparative indicators. Regarding the accuracy of erosion modulus prediction, the average error of the present invention is only 12.4 t / ha·yr, a reduction of 55.4% compared to the USLE-R physical model and 36.7% compared to the CNN-based deep learning scheme. In terms of spatial matching, the IoU of the present invention reaches 0.87, enabling more accurate reconstruction of the spatial pattern of slope erosion. In terms of fine-scale gully erosion identification performance, the present invention improves upon the USLE-R physical model by more than 20 percentage points. Furthermore, in terms of simulation response speed, the present invention, relying on the soil erosion generation model and the improved spatiotemporal graph convolutional network, can complete a full simulation in just 3.6 seconds, significantly shortening the computation time. In scenarios with missing data, the present invention automatically fills in the gaps through the soil erosion generation model, improving stability to 94.5%; the uncertainty interval width is significantly reduced, indicating that the output of the soil erosion generation model is more robust and reliable.

[0152] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent simulation of soil erosion based on AIGC technology, characterized in that, Includes the following steps: Step 1: Collect soil and water loss data for the target area; Step 2: Perform spatiotemporal alignment and standardization on the soil erosion data to obtain a multi-scale spatiotemporal feature vector sequence, and use an attention-based feature fusion network to perform weighted fusion of the multi-scale spatiotemporal feature vector sequence to generate a spatiotemporal fusion feature tensor. Step 3: Construct a soil erosion generation model based on AIGC technology. The soil erosion generation model includes a soil erosion scene generation module and a discrimination module. Input the spatiotemporal fusion feature tensor into the soil erosion scene generation module to generate soil erosion simulation data, and conduct adversarial training between the real observation data and the soil erosion simulation data through the discrimination module. Step 4: Establish a dynamic feedback mechanism to update the soil erosion scenario generation module online and obtain an optimized soil erosion generation model; Step 5: Input the soil erosion projection data and the physically corrected observation features into the improved spatiotemporal graph convolutional network to generate the soil erosion modulus for the future preset time period, and output the soil erosion intensity layer and spatial distribution layer. The improved spatiotemporal graph convolutional network introduces physical constraint terms. Step 6: Introduce scenario-driven factors and fuse them with spatiotemporal fusion feature tensors. Input the fusion into the optimized soil and water loss generation model, perform multi-scenario intelligent soil and water loss simulation, generate soil and water loss simulation data under different scenarios, and output soil and water loss intensity layer sequence and soil and water loss evolution trend map. Step 7: Map the soil erosion intensity layer sequence to a geospatial layer sequence and classify the soil erosion risk level; use Monte Carlo-based Dropout and Bootstrap sampling to perform uncertainty quantification analysis.

2. The intelligent soil erosion simulation method based on AIGC technology according to claim 1, characterized in that, Step one specifically includes: Soil and water loss data of the target area are collected collaboratively by satellite remote sensing platform, UAV aerial survey system and ground sensor network. The soil and water loss data includes high-resolution multispectral imagery, lidar point cloud, soil moisture, rainfall intensity, surface roughness and vegetation coverage.

3. The intelligent soil erosion simulation method based on AIGC technology according to claim 1, characterized in that, Step two specifically includes: Atmospheric and geometric corrections were performed on the high-resolution multispectral images to extract the normalized vegetation index, slope, aspect, and soil brightness index. The lidar point cloud is reconstructed in three dimensions to generate a digital surface model, and the surface roughness features and fine-scale vegetation cover features are calculated based on the moving window method. Outlier removal, unit standardization, and timestamp synchronization were performed on soil moisture, rainfall intensity, surface roughness, and vegetation cover to obtain a standardized hydrophysical quantity sequence. Using spatial grid cells and time steps as indices, normalized vegetation index, slope, aspect and soil brightness index, surface roughness characteristics, fine-scale vegetation cover characteristics and standardized hydrophysical quantity sequences are resampled and interpolated according to spatial location and time label to obtain multi-scale feature vector sequences. A multi-scale feature vector sequence is input into an attention-based feature fusion network. Attention weights are assigned to feature vectors from different data sources and weighted fusion is performed to generate a spatiotemporal fusion feature tensor.

4. The intelligent simulation method for soil erosion based on AIGC technology according to claim 1, characterized in that, Step three specifically includes: The soil erosion scenario generation module adopts a conditional variational autoencoder structure, which learns the potential distribution of the soil erosion process and generates soil erosion inference data. The encoder of the soil erosion scene generation module includes four downsampling stages. Each downsampling stage extracts local spatial features from the spatiotemporal fusion feature tensor through a 3×3 convolutional layer to obtain a downsampled feature map. The spatial dimension of the downsampled feature map is reduced through a 2×2 max pooling layer. The number of channels of the encoder increases exponentially with each downsampling stage, from the initial 64 to 128, 256 and 512, to enhance the semantic expressive power of deep features. The decoder of the soil erosion scene generation module is set with four upsampling stages. In each upsampling stage, the spatial resolution is gradually restored through transposed convolutional layers. Skip connections are used to concatenate the downsampled feature maps of each downsampling stage of the encoder with the feature maps of the corresponding upsampling stages to generate upsampled fusion feature maps, which are then used as input feature maps for the next upsampling stage. The discrimination module adopts a multi-scale convolutional neural network structure, including a first-scale discrimination branch, a second-scale discrimination branch, and a third-scale discrimination branch; each scale discrimination branch is composed of several 3×3 convolutional layers, batch normalization layers, and LeakyReLU activation layers stacked sequentially, and different downsampling times are set in each scale discrimination branch; Original resolution images, 2x downsampled images, and 4x downsampled images were constructed for real observation data and soil erosion projection data, respectively. The first scale discrimination branch receives the original resolution image of the real observation data and the original resolution image of the soil erosion inference data respectively, outputs the first scale real discrimination feature map and the first scale generated discrimination feature map, and uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain the first scale real response map and the first scale generated response map. The second-scale discrimination branch receives a 2x downsampled image of the real observation data and a 2x downsampled image of the soil erosion inference data, respectively, and outputs a second-scale real discrimination feature map and a second-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a second-scale real response map and a second-scale generated response map. The third-scale discrimination branch receives a 4x downsampled image of the real observation data and a 4x downsampled image of the soil erosion inference data, respectively, and outputs a third-scale real discrimination feature map and a third-scale generated discrimination feature map. It then uses a 1×1 convolutional layer to perform pixel-by-pixel discrimination to obtain a third-scale real response map and a third-scale generated response map. The loss function for adversarial training includes least squares loss and feature matching loss; The least squares loss is calculated by: calculating the expected value of the squared difference between the true response map at each of the three scales and 1, and the expected value of the squared difference of the generated response map at each of the three scales, and then averaging and summing them. The feature matching loss is the expectation of the element-wise absolute difference between the real discriminant feature maps and the generated discriminant feature maps at three scales, and then summed.

5. The intelligent simulation method for soil erosion based on AIGC technology according to claim 1, characterized in that, The dynamic feedback mechanism specifically includes: setting a sliding time window, continuously receiving real-time observation data, and calculating the root mean square error between the soil erosion simulation data and the real-time observation data as a performance loss function; based on the performance loss function, using an adaptive moment estimation optimization algorithm to update the parameters of the soil erosion scenario generation module, with an initial learning rate of 0.001, which decays exponentially with the training rounds or error convergence.

6. The intelligent simulation method for soil erosion based on AIGC technology according to claim 1, characterized in that, Step five specifically includes: Based on soil erosion data, a physical correction observation feature was constructed using a general soil loss equation. The physical correction observation feature includes rainfall erosivity factor, soil erodibility factor, and elevation. The soil erosion simulation data is mapped into a graph node feature matrix, and a graph structure relationship matrix is ​​constructed based on terrain connectivity. Based on physically corrected observation characteristics, physical constraint terms are constructed, specifically as follows: ； in, Represents physical constraint terms. As a soil erodibility factor, For elevation, Indicates elevation gradient, As the erosivity factor of rainfall, The constraint strength coefficient, Represents the square of the L2 norm; The improved spatiotemporal graph convolutional network introduces physical constraint terms, specifically: The physical constraint terms are concatenated with the graph node feature matrix in the feature dimension to form an extended node feature matrix; the graph structure relation matrix is ​​normalized to obtain the terrain connectivity normalized adjacency matrix; and graph convolution is performed based on the terrain connectivity normalized adjacency matrix and the extended node feature matrix to obtain the updated graph node feature matrix. Based on the updated graph node feature matrix, temporal convolution is used for time modeling to obtain the predicted hidden feature matrix for a future preset time period; the future preset time period includes several prediction time steps, and each prediction time step corresponds to a predicted hidden feature vector. For each prediction time step, the predicted hidden feature vector is linearly mapped to generate the soil erosion modulus, and the soil erosion modulus is rearranged into a raster form according to the correspondence between graph nodes and spatial grids to obtain the soil erosion intensity layer and spatial distribution layer for the current prediction time step.

7. The intelligent simulation method for soil erosion based on AIGC technology according to claim 1, characterized in that, Step six specifically includes: The scenario-driving factors include climate scenario parameters, land use change parameters, and engineering intervention parameters. The climate scenario parameters include temperature and rainfall changes; the land use change parameters represent the spatial proportion changes of different land types; the engineering intervention parameters include spatial configuration quantitative indicators for vegetation restoration, terrace construction, and silt-trapping dam deployment; and the scenario driving factors are reduced to 10 principal components through principal component analysis. The scenario-driven factors and the spatiotemporal fusion feature tensor after dimensionality reduction are concatenated in the feature dimension to form the scenario condition feature tensor. The scenario condition feature tensor is then input into the optimized soil and water loss scenario generation module to generate soil and water loss inference data for the corresponding scenario. The soil erosion projection data is rearranged according to spatial grid units and time series to construct a soil erosion intensity layer sequence. The soil erosion intensity layer sequence contains multiple prediction time steps, and each prediction time step corresponds to a soil erosion intensity raster layer. Based on the soil erosion intensity layer sequence, the soil erosion evolution trend index is calculated. The soil erosion evolution trend index includes the spatial cumulative erosion, the erosion center offset trajectory and the expansion rate of high-risk areas, and a soil erosion evolution trend map corresponding to the scenario is generated. The sequence of soil erosion intensity layers and the trend map of soil erosion evolution are visualized.

8. The intelligent simulation method for soil erosion based on AIGC technology according to claim 1, characterized in that, Step seven specifically includes: The soil erosion intensity raster layer in the soil erosion intensity layer sequence is mapped to the corresponding geographic coordinate system according to the spatial grid unit to generate a geospatial layer. Set a threshold for the risk level of soil and water loss, classify the soil and water loss risk level according to the threshold for the soil and water loss intensity value corresponding to each spatial grid unit, and label the level within each threshold range based on the soil erosion modulus value to generate a soil and water loss risk level map. The soil and water loss risk levels include extremely low risk, low risk, medium risk, high risk and extremely high risk; During the inference process, the Dropout layer of the soil erosion generation model is kept active. N forward inferences are performed on the same model input to obtain N sets of soil erosion simulation data. The mean and variance of the N sets of soil erosion simulation data are calculated on each spatial grid cell, and confidence intervals are constructed based on the mean and variance to generate a Monte Carlo Dropout confidence interval map. Bootstrap resampling is performed on the soil erosion data to obtain several sampled datasets; inference is performed on the several sampled datasets to obtain several sets of soil erosion inference data, and the distribution of the several sets of soil erosion inference data on each spatial grid cell is statistically analyzed, quantiles are calculated, and Bootstrap confidence interval plots are generated. Spatial visualization of soil erosion risk level map, Monte Carlo Dropout confidence interval map, and Bootstrap confidence interval map.

9. A soil erosion intelligent simulation system based on AIGC technology, executing the soil erosion intelligent simulation method based on AIGC technology as described in any one of claims 1 to 8, characterized in that, include: Soil and water loss data acquisition module: used to collect soil and water loss data in the target area; Spatiotemporal feature construction module: used to preprocess soil erosion data and generate spatiotemporal fusion feature tensors through a feature fusion network based on an attention mechanism; AIGC Soil and Water Loss Modeling Module: Used to build soil and water loss generation models and generate soil and water loss projection data; Dynamic feedback update module: used to update the soil and water loss generation model online; Physical constraint graph convolutional extrapolation module: used to extrapolate future time periods from soil erosion extrapolation data and physically corrected observation features, and generate soil erosion modulus; Multi-scenario simulation module: used to construct scenario condition feature tensors and input them into the optimized soil and water loss scenario generation module to generate soil and water loss simulation data for the corresponding scenarios; Risk and Uncertainty Analysis Module: Used to classify soil erosion risk levels and employ Monte Carlo Dropout and Bootstrap sampling to perform uncertainty quantification analysis on soil erosion projection data.

Images