Method and apparatus for identifying groundwater level thresholds corresponding to surface deformation

Abstract

This invention discloses a method and apparatus for identifying groundwater level thresholds corresponding to surface deformation, comprising: step S1, acquiring groundwater levels and PS point subsidence rates for different aquifer groups; step S2, quantifying the spatiotemporal response characteristics of groundwater level changes and surface deformation over multiple time periods based on groundwater levels and PS point subsidence rates for different aquifer groups, and identifying the response thresholds of groundwater level changes to surface deformation for different aquifer groups. By employing the technical solution of this invention, the spatiotemporal response characteristics between groundwater level changes and surface deformation for different aquifer groups are quantitatively analyzed, and groundwater level thresholds are identified, providing a scientific basis for regional land subsidence control.

Description

Technical Field

[0001] This invention belongs to the field of land subsidence control technology, and particularly relates to a method and device for identifying groundwater level thresholds corresponding to surface deformation. Background Technology

[0002] Land subsidence is a slowly changing geological hazard caused by soil compression under the influence of natural and anthropogenic factors, resulting in a regional decrease in ground elevation. The main cause of land subsidence is the excessive extraction of groundwater, which leads to the compaction of aquifer systems and consequently, land subsidence. Currently, it is impossible to quantitatively analyze the spatiotemporal response characteristics between groundwater level changes and surface deformation in different aquifer groups, identify groundwater level thresholds, and provide a scientific basis for regional land subsidence control. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method and apparatus for identifying groundwater level thresholds corresponding to surface deformation.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for identifying groundwater level thresholds corresponding to surface deformation, comprising:

[0006] Step S1: Obtain the groundwater level and PS point settlement rate for different aquifer groups;

[0007] Step S2: Based on the groundwater level of different aquifer groups and the settlement rate of PS point, quantify the spatiotemporal response characteristics of water level changes in different aquifers and surface deformation over multiple time periods, and identify the response threshold of water level changes in different aquifer groups to surface deformation.

[0008] Preferably, step S2 includes:

[0009] Step S21: Using the groundwater level of different aquifer groups as the independent variable and the settlement rate of PS point as the dependent variable, the random forest machine learning algorithm is used to obtain the predicted impact of groundwater level of different aquifer groups on surface deformation.

[0010] Step S22: Based on the predicted impact of groundwater level on surface deformation of different aquifer groups, the spatiotemporal response characteristics of groundwater level changes and surface deformation in different aquifers over multiple time periods are obtained through the SHAP interpretable machine learning method, and the groundwater level threshold is identified.

[0011] Preferably, step S22 includes:

[0012] Step S221: Calculate the Shapley value of the predicted influence of groundwater level on surface deformation for different aquifer groups;

[0013] Step S222: Rank the influencing factors according to the Shapley values ​​to obtain the spatiotemporal response characteristics of water level changes and surface deformation in different aquifers;

[0014] Step S223: Based on the Shapley value, use the single-factor dependency of SHAP to identify the response threshold of different aquifer water level changes to surface deformation.

[0015] The present invention also provides a groundwater level threshold identification device corresponding to surface deformation, comprising:

[0016] The acquisition module is used to acquire groundwater levels and PS point settlement rates for different aquifer groups;

[0017] The identification module is used to quantify the spatiotemporal response characteristics of water level changes in different aquifers and surface deformation over multiple time periods based on groundwater levels and PS point settlement rates of different aquifer groups, and to identify the response threshold of water level changes in different aquifer groups to surface deformation.

[0018] Preferably, the identification module includes:

[0019] The first processing unit is used to take the groundwater level of different aquifer groups as the independent variable and the settlement rate of PS point as the dependent variable, and use the random forest machine learning algorithm to obtain the predicted impact of groundwater level of different aquifer groups on surface deformation.

[0020] The second processing unit is used to obtain the spatiotemporal response characteristics of the changes in aquifer water level and surface deformation over multiple time periods by using the SHAP interpretable machine learning method, based on the predicted impact of groundwater level on surface deformation of different aquifer groups, and to identify groundwater level thresholds.

[0021] Preferably, the second processing unit includes:

[0022] The first processing component is used to calculate the Shapley values ​​of the predicted impact of groundwater levels on surface deformation for different aquifer groups.

[0023] The second processing component is used to rank the importance of influencing factors according to the Shapley value, and obtain the spatiotemporal response characteristics of water level changes and surface deformation in different aquifers.

[0024] The third processing component is used to identify the response threshold of different aquifer water level changes to surface deformation based on Shapley values ​​and the single-factor dependency of SHAP.

[0025] This invention uses groundwater levels of different aquifer groups as independent variables and PS point settlement rate as dependent variables. It employs random forest machine learning algorithm and SHAP interpretable machine learning method to quantify the spatiotemporal response characteristics of water level changes in different aquifer groups and surface deformation over multiple time periods, and identifies the response threshold of water level changes in different aquifer groups to surface deformation. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0027] Figure 1 This is a flowchart of the groundwater level threshold identification method corresponding to surface deformation in an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Example 1:

[0031] like Figure 1 As shown, this embodiment of the invention provides a method for identifying groundwater level thresholds corresponding to surface deformation, including:

[0032] Step S1: Obtain the groundwater level and PS point settlement rate for different aquifer groups;

[0033] Step S2: Based on the groundwater level of different aquifer groups and the settlement rate of PS point, quantify the spatiotemporal response characteristics of water level changes in different aquifers and surface deformation over multiple time periods, and identify the response threshold of water level changes in different aquifer groups to surface deformation.

[0034] Preferably, step S2 includes:

[0035] Step S21: Using groundwater levels in different aquifer groups as independent variables and the settlement rate at point PS as the dependent variable, a random forest machine learning algorithm is used to obtain the predicted impact of groundwater levels in different aquifer groups on surface deformation, specifically including:

[0036] S211. Use PS-InSAR to process SAR images and obtain surface deformation information.

[0037] PS-InSAR (Permanent Scatterer Synthetic Aperture Radar Interferometry) is a technique for monitoring surface deformation. PS-InSAR technology performs differential interferometry on long-term SAR image data. By identifying points (permanent scatterers, PS) in the monitoring area that maintain stable reflection characteristics in multiple time phases as target points, differential modeling is performed using PS points, and the residual phase of the terrain and the atmospheric delay phase are removed, thereby calculating the deformation information of the PS points. Its basic principle is to use N+1 registered single-view complex images covering the study area, select one as the main image, and the remaining N images as sub-images. The phase values ​​of the corresponding pixels of the PS points in the main image are subtracted from each of the N differential interferograms. Each pixel in each differential interferogram contains five phase components

[30] :

[0038]

[0039] in, This is a flat phase; For terrain phase; For deformation phase; This is the atmospheric delayed phase; The noise phase is caused by system noise and other errors. SRTM 30m spatial resolution DEM data is selected to remove the terrain phase, and temporal filtering is used to remove atmospheric delay phase and noise phase.

[0040] S212. Based on ArcGIS, preprocessing requires inputting surface deformation data and groundwater level data from a random forest (RF) regression model.

[0041] Available groundwater level data are typically point features or contour lines. ArcGIS Kriging interpolation is used to interpolate the contour-based water level data into surface data. Due to variations in geological structure, the depth of groundwater level is not directly correlated with surface deformation; therefore, it is necessary to calculate the variation in groundwater level data for each layer as input data for the model.

[0042] S213. Using the preprocessed data described above, construct an RF regression model.

[0043] Using different aquifer water levels as feature variables (independent variables) and the deformation rate at the PS point as the target variable (dependent variable), 70% of the data was selected as the training set and 30% as the test set. Multiple sample sets were created by randomly sampling with replacement from the training data, each used to train a decision tree. For each decision tree, during tree construction, a subset of features (usually the square root of the number of features) was randomly selected for each node split. The split point selection metric utilized the minimum mean squared error (MSE), calculated as follows:

[0044]

[0045] Among them, y i This is the actual value. Here, N represents the model's predicted value, and N is the number of samples. The decision tree selects the split point at each node that minimizes the MSE (Minimum Separation Equation) to achieve the best splitting effect. This process is repeated to create a large number of decision trees, each trained on different sample sets and feature subsets. Each tree is trained independently and learns different features of the data. For new data, predictions are made using each decision tree in the forest. The average of the predictions from all trees is then used to obtain the predicted target variable.

[0046]

[0047] Where T is the total number of trees. The final prediction result is obtained by averaging, which in this example is the deformation rate of point PS.

[0048] S214. Use the characteristic importance index of the model to evaluate the contribution of water level changes in different aquifers to surface deformation, and obtain the influence of water level changes in different aquifers on surface deformation.

[0049] Feature importance is measured by the reduction in the mean squared error between the predicted and actual output results corresponding to the input features during model training. The importance of feature j is I. j It can be represented as:

[0050]

[0051] Where j is the j-th aquifer, S j ΔMSE is the set of nodes in decision tree t that are split using feature j. s,t This indicates the amount of MSE reduction caused by the splitting of s.

[0052] Step S22: Based on the predicted impact of groundwater levels on surface deformation of different aquifer groups, the SHAP interpretable machine learning method is used to obtain the spatiotemporal response characteristics of groundwater level changes and surface deformation in different aquifers over multiple time periods, and to identify groundwater level thresholds, specifically including:

[0053] Step S221: Calculate the Shapley value of the predicted impact of groundwater level on surface deformation for different aquifer groups, and further quantify the contribution and impact of water level changes in different aquifers on surface deformation.

[0054] Based on the data trained using the RF model in S213, different aquifer water levels are used as the input features of the model, and the PS point settlement rate is used as the target variable, i.e., the predicted value. The model input feature set is divided into several subsets, each representing a different feature combination. For each feature, the marginal contribution in all possible feature combinations is calculated. The calculation process is as follows: for each feature combination, assuming the influence of adding or removing a feature, the incremental contribution of that feature to the model output is calculated. By calculating the average marginal contribution of all feature combinations, the Shapley value of that feature is obtained. The Shapley values ​​of each feature are summed to obtain the interpretation of the model's predicted value. The SHAP method treats the contribution of each feature as a linear summation model, so that the predicted value can be decomposed into the sum of the contributions of each feature.

[0055] Given a feature i, its Shapley value is calculated using the following formula:

[0056]

[0057] Where: N is the complete set of features, S is a subset of features excluding feature i, f(S) represents the model prediction with only feature subset S included, f(S∪{i}) represents the model prediction with feature i included, and |S| and (|N|-|S|-1)! represent the factorial of the subset size to ensure that the marginal contributions of all feature combinations are weighted fairly.

[0058] The SHAP method decomposes the model's predicted value f(x) into the sum of the Shapley values ​​of each feature:

[0059]

[0060] Where f(x) is the predicted value of the model, This is a constant that explains the model, namely the predicted mean of all training samples. The Shapley value of the i-th feature represents the marginal contribution of this predictor variable to the predicted value of the target variable. By calculating the absolute value of each feature's Shapley value, the importance of the feature—that is, the contribution of different aquifer water level changes to surface deformation—is quantified. The larger the absolute value of the feature's Shapley value, the more significant its impact on the model output. Furthermore, the sign of the feature's Shapley value quantifies its influence on the predicted value. A positive Shapley value indicates a positive impact on the model output; an increase in the feature value leads to a higher predicted value, corresponding to a higher aquifer water level (rising water level), a faster deformation rate, and a more pronounced surface uplift. Conversely, a negative Shapley value indicates a negative impact on the model output; a decrease in the feature value leads to a lower predicted value, corresponding to a lower aquifer water level (falling water level), a slower deformation rate, and a more pronounced surface subsidence.

[0061] Step S222: Rank the influencing factors according to the Shapley values ​​to obtain the spatial response characteristics of water level changes and surface deformation in different aquifers.

[0062] First, based on a single sample, the feature corresponding to the largest absolute value of the Shapley value is selected, i.e., the aquifer water level corresponding to the largest absolute value of the Shapley value at that PS point. Then, the spatial location of the sample is extracted, i.e., the latitude and longitude of the PS. By traversing each sample point, the feature with the greatest spatial impact on each sample point can be obtained, i.e., the aquifer water level corresponding to the greatest spatial impact on surface deformation. Using the above steps, the spatial response characteristics of different aquifer water level changes and surface deformation are obtained.

[0063] Step S223: Based on the clustering characteristics of water level changes in different aquifers and their Shapley values, analyze the threshold effect of water level changes in different aquifers on surface deformation.

[0064] By analyzing the changes in shapley values ​​corresponding to different aquifer water level changes, we can identify the characteristic interval where the shapley value increases from small to large and then remains in equilibrium, which can be used as the response threshold range.

[0065] Example 2:

[0066] This invention also provides a groundwater level threshold identification device corresponding to surface deformation, comprising:

[0067] The acquisition module is used to acquire groundwater levels and PS point settlement rates for different aquifer groups;

[0068] The identification module is used to quantify the spatiotemporal response characteristics of water level changes in different aquifers and surface deformation over multiple time periods based on groundwater levels and PS point settlement rates of different aquifer groups, and to identify the response threshold of water level changes in different aquifer groups to surface deformation.

[0069] As one embodiment of the present invention, the identification module includes:

[0070] The first processing unit is used to take the groundwater level of different aquifer groups as the independent variable and the settlement rate of PS point as the dependent variable, and use the random forest machine learning algorithm to obtain the predicted impact of groundwater level of different aquifer groups on surface deformation.

[0071] The second processing unit is used to obtain the spatiotemporal response characteristics of the changes in aquifer water level and surface deformation over multiple time periods by using the SHAP interpretable machine learning method, based on the predicted impact of groundwater level on surface deformation of different aquifer groups, and to identify groundwater level thresholds.

[0072] As one embodiment of the present invention, the second processing unit includes:

[0073] The first processing component is used to calculate the Shapley values ​​of the predicted impact of groundwater levels on surface deformation for different aquifer groups.

[0074] The second processing component is used to rank the importance of influencing factors according to the Shapley value, and obtain the spatiotemporal response characteristics of water level changes and surface deformation in different aquifers.

[0075] The third processing component is used to identify the response threshold of different aquifer water level changes to surface deformation based on Shapley values ​​and the single-factor dependency of SHAP.

[0076] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for identifying a corresponding groundwater level threshold value at which ground surface deformation occurs, characterized by, The method comprises the following steps: Step S1, obtaining different aquifer group groundwater levels and PS point subsidence rates; Step S2, quantifying the spatiotemporal response characteristics of different aquifer water level changes and surface deformation in multiple time periods according to the different aquifer group groundwater levels and PS point subsidence rates, and identifying the response threshold of different aquifer group water level changes on surface deformation; Step S2 comprises: Step S21, taking different aquifer group groundwater levels as independent variables and PS point subsidence rates as dependent variables, and using a random forest machine learning algorithm to obtain the predicted influence of different aquifer group groundwater levels on surface deformation; Step S21 comprises: S211, obtaining surface deformation information by processing SAR images using PS-InSAR: Using the registered N+1 single-view complex images covering the study area, selecting one of them as the main image, and the remaining N images as the secondary images, respectively subtracting the phase values of the PS points corresponding to the pixels in the main image to obtain N difference interferograms, each pixel in each difference interferogram contains five phase components: wherein, is a flat earth phase; is a terrain phase; is a deformation phase; is an atmospheric delay phase; is a noise phase due to system noise and other errors. S212, preprocessing the surface deformation data and groundwater level data input into the random forest (RF) regression model based on ArcGIS; S213, constructing an RF regression model using the above preprocessed data; The RF regression model takes different aquifer water levels as characteristic variables and PS point deformation rates as target variables to obtain the predicted target variable results: where T is the total number of trees, is the final prediction result obtained by averaging, which is the deformation rate of the PS point; S214, using the feature importance index of the model to evaluate the contribution of different aquifer water level changes to surface deformation to obtain the influence of different aquifer water level changes on surface deformation; Step S22, according to the predicted influence of different aquifer group groundwater levels on surface deformation, obtaining the spatiotemporal response characteristics of different aquifer water level changes and surface deformation in multiple time periods by SHAP interpretable machine learning method, and identifying the groundwater level threshold; Step S22 comprises: Step S221, calculating the shapley value of the predicted influence of different aquifer group groundwater levels on surface deformation: Step S222, sorting the importance of influencing factors according to the shapley value to obtain the spatiotemporal response characteristics of different aquifer water level changes and surface deformation: According to the aquifer water level corresponding to the maximum absolute value of the PS point shapley value, the latitude and longitude of the PS point are extracted; each sample point is traversed to obtain the aquifer water level corresponding to the maximum influence of each sample point on surface deformation in space; Step S223, based on the shapley value, using the single-factor dependence relationship of SHAP to identify the response threshold of different aquifer water level changes on surface deformation.

2. A device for identifying a corresponding groundwater level threshold at which ground deformation occurs, characterized by, The method of claim 1 comprises: An acquisition module for acquiring different aquifer group groundwater levels and PS point subsidence rates; An identification module for quantifying the spatiotemporal response characteristics of different aquifer water level changes and surface deformation in multiple time periods according to the different aquifer group groundwater levels and PS point subsidence rates, and identifying the response threshold of different aquifer group water level changes on surface deformation; The identification module comprises: The first processing unit is configured to use a random forest machine learning algorithm to obtain a prediction influence of different aquifer group water levels on ground surface deformation, with the different aquifer group water levels as independent variables and the PS point subsidence rate as a dependent variable. The second processing unit is configured to obtain a spatiotemporal response feature of the different aquifer water level changes and the ground surface deformation in multiple time periods by using a SHAP interpretable machine learning method according to the prediction influence of the different aquifer group water levels on the ground surface deformation, and identify a groundwater level threshold. The second processing unit comprises: A first processing component configured to calculate a shapley value of the prediction influence of the different aquifer group water levels on the ground surface deformation. A second processing component configured to sort the importance of the influencing factors according to the shapley value, and obtain the spatiotemporal response feature of the different aquifer water level changes and the ground surface deformation. A third processing component configured to identify a response threshold of the different aquifer water level changes on the ground surface deformation based on the shapley value and a single-factor dependency relationship of the SHAP.

Images