A regional soil erosion vulnerability assessment method based on the combined weighting method

Abstract

The regional soil erosion vulnerability evaluation method based on the combined weighting method disclosed by the present invention is specifically as follows: A combined weighting method integrating the FAHP and BPNN models is proposed to fully consider the objectivity of data and the subjective opinions of experts on evaluation indicators, thereby effectively overcoming the deficiencies of single subjective evaluation or objective weighting, significantly reducing errors, and having high evaluation accuracy and persuasiveness. This weighting method first calculates the subjective weights of each evaluation indicator based on FAHP, then calculates the objective weights of each indicator according to BPNN, and then obtains the final weights of each indicator according to the combined weight calculation formula. At the same time, in the present invention, the selection of evaluation indicators fully considers the typicality, representativeness, and accessibility of data, and has high persuasiveness, credibility, universality, and applicability for promotion to large-scale ranges. The proposed method fills the gap in the quantitative evaluation of soil erosion vulnerability in the existing natural disaster evaluation system and can be used to solve the problems existing in the existing methods.

Description

Technical Field

[0001] The present invention relates to the technical field of quantitative evaluation of regional soil erosion, and particularly relates to a method for evaluating the vulnerability of regional soil erosion based on a combined weighting method. Background Art

[0002] Soil erosion is a global ecological environment problem. Unreasonable production, lifestyle and other human activities are the main reasons for soil loss, and significantly affect natural resources, agricultural production and the ecological environment. Therefore, establishing a reasonable soil erosion evaluation method is crucial for efficiently and accurately identifying the spatial distribution pattern and its evolution law of regional soil erosion vulnerable areas, and regulating soil erosion vulnerable areas. The method for evaluating the vulnerability of regional soil erosion established based on the human-earth coupling system has wide application value in soil and water conservation planning, land use management, environmental protection and other aspects, and has been successfully applied to the ecological environment restoration and evaluation programs in ecologically vulnerable areas such as the Northeast Black Soil Region, the Northwest Loess Region and the Southwest Red Soil Region. However, since the theoretical basis of the method for evaluating the vulnerability of regional soil erosion is based on the general laws and regional characteristics of soil erosion at the basin scale, it is necessary to consider the comprehensive influence of various natural and social factors such as climate, terrain, soil, vegetation, land use, human activities, social economy, and government management. For the universal research of the method for evaluating the vulnerability of regional soil erosion at the regional scale, there has been a lack of systematic and scientific research results for many years.

[0003] At present, in the research on the method for evaluating the vulnerability of regional or larger-scale soil erosion, the method for evaluating the vulnerability of regional soil erosion based on the combined weighting method is relatively few. Although some scholars have established a method for evaluating the vulnerability of soil erosion for some ecologically vulnerable areas, there are the following deficiencies: At present, the methods for weighting the evaluation indicators of the vulnerability of regional soil erosion by domestic and foreign scholars mainly use single subjective weighting methods and objective weighting methods, such as the expert scoring method, the analytic hierarchy process, the entropy weight method, the grey relational method, etc. However, the single subjective weighting method cannot reflect the objectivity of the data, and the single objective weighting method cannot take into account expert experience. Therefore, these methods are not applicable to the non-linear analysis of complex systems and cannot meet the needs of evaluating the vulnerability of regional soil erosion; at present, most of the vulnerability research focuses on the research of sudden natural disasters such as landslides and debris flows, while less research has been done on conventional natural disasters such as desertification and soil and water loss. In addition, the existing methods for evaluating the vulnerability of regional soil erosion focus on small watersheds or small-scale ecologically vulnerable areas. The evaluation indicators are oriented to the unique characteristics of ecologically vulnerable areas, and there are certain limitations, and they do not have the universality and practicability for promotion to regions and large-scale administrative regions. The present invention proposes a technical method to solve the problems existing in the prior art. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to propose a regional soil erosion vulnerability evaluation method based on the combined weighting method to solve the problems existing in the prior art.

[0005] The present invention adopts the following technical solutions to solve its technical problems:

[0006] The regional soil erosion vulnerability evaluation method based on the combined weighting method provided by the present invention is as follows: first, calculate the subjective weights of each index based on FAHP, then calculate the objective weights of each index according to BPNN, and then obtain the final weights of each index according to the combined weight calculation formula.

[0007] In the above method, the modified universal soil loss equation RUSLE model can be used to calculate the soil erosion amount of the study area, and then multiply the normalized data of each evaluation index of the regional soil erosion vulnerability by its corresponding index weight, and calculate the soil erosion vulnerability value of the study area according to "vulnerability = exposure + sensitivity - adaptability". Based on the natural break method in ArcGIS, the soil erosion vulnerability is divided into 5 categories: low, relatively low, moderate, relatively high, and high, and a soil erosion vulnerability classification map is generated. The classification map can be used to intuitively understand the spatial distribution pattern and temporal variation of the soil erosion vulnerability in the regional and large-scale ranges, and provide theoretical and technical support for the reasonable allocation of soil and water conservation measures and soil and water conservation planning in the regional and large-scale ranges.

[0008] The regional soil erosion vulnerability evaluation method based on the combined weighting method provided by the present invention includes: first, obtaining social and economic data, data preprocessing, training using the BPNN machine learning network model, weighting each index layer using FAHP, calculating the combined weight, and then determining the soil erosion modulus according to the digital elevation model data, normalized difference vegetation index data, land use data, soil data, and rainfall data.

[0009] In the above method, the following method can be adopted for data preprocessing: perform radiometric calibration, atmospheric correction, and image fusion tasks on the downloaded remote sensing images through ENVI software; import the social and economic data into ArcGIS and save it as a raster file.

[0010] In the above method, the following method can be adopted to train the BPNN machine learning network model and obtain the objective weight:

[0011] (1) Standardization and classification of input data:

[0012] The range normalization method is used to normalize the data of each index in the index layer, and the normalized data is used as the input data; then the input data is divided into a training data set, a test data set and a validation data set; the training data set is presented to the network during training and accounts for 70% of the sample data; the test data set does not affect the training and provides an independent measurement of the network performance during and after training, which explains 15% of the sample data; the validation data set is used to measure the network generalization and stop training when the generalization stops improving, accounting for 15% of the sample data;

[0013] (2) Construct the network and start training:

[0014] Set the indicator layer to the number of input neurons, set the target layer to the number of output neurons, and set the number of hidden layers to 10; by constructing a 13-10-1 network structure, based on the BPNN algorithm program written in Matlab software, where the initial weight and threshold matrices are randomly generated, the activation function uses the sigmoid function, the learning rate is 0.1, the maximum number of iterations is 1000, and the expected error is 0.5×10 3 , and the performance function is the mean square error; after the training data set propagates forward, if the error does not meet the preset accuracy, the error will be propagated backward from the output layer to the input layer, and the error will be gradually reduced by adjusting the weights and thresholds, thereby obtaining the objective weight V of each index layer;

[0015] (3) Test the accuracy of the training:

[0016] Use the Pearson product-moment correlation coefficient r value to measure the performance of the model network. The closer the r value is to 1, the better the training performance of the BPNN network model.

[0017] In the above method, the FAHP can be used to assign weights to each index layer by the following method:

[0018] (1) Establish a fuzzy complementary judgment matrix:

[0019] When using the fuzzy analytic hierarchy process, first use the relative importance of a certain factor compared with other factors to make a quantitative judgment of the importance of 2 factors, and compare each factor pairwise to obtain the fuzzy judgment matrix A=(a ij ) n×n ; then compare and take values for the factors a1,…a n to obtain the following fuzzy complementary matrix:

[0020] In the formula, n represents the order of the matrix;

[0021] (2) Determine the weight of the judgment matrix:

[0022] After summing the rows of the fuzzy judgment matrix A, perform a mathematical substitution: Let the obtained consistency matrix R = (r ij ) m×n , and use the row-sum normalization method to find the sorting vector from the consistency matrix R, w = (w1, w2,... w n ) T , where the w vector satisfies:

[0023]

[0024] In the formula, a ij is an element in the fuzzy judgment matrix, and i and j represent its row number and column number respectively.

[0025] Thus, the subjective weight U obtained by FAHP is obtained.

[0026] In the above method, the following method can be used to test the consistency of the fuzzy complementary judgment matrix:

[0027] Define the geometry of all n-order fuzzy complementary judgment matrices as G n ; Let A = (a ij ) n×n and B = (b ij ) n×n ∈G n , and use the norm to represent the distance between A and B, denoted as ρ(A, B) = ||A - B||;

[0028] Let A = (a ij ) n×n and B = (b ij ) n×n both be fuzzy judgment matrices, and call the compatibility degree between A and B; The compatibility index between A and B is

[0029] Define w = (w1, w2,... w n ) T as the weight vector of matrix A, as the characteristic matrix of judgment matrix A;

[0030] For the decision maker's attitude α, when the compatibility index I(A, W * ) ≤ α, it is considered to meet the consistency requirement; The smaller the value of α, the higher the requirement of the project decision maker for consistency. Usually, α = 0.1 is taken without special circumstances.

[0031] In the above method, the combination weight calculation can be carried out by the following method:

[0032] (1) First, calculate the subjective weights of each index based on FAHP. The specific process is:

[0033] 1) Make and distribute questionnaires based on the evaluation index system to obtain the opinions of experts in the field of soil erosion and water and soil conservation.

[0034] 2) Summarize the scoring results of each expert on the relative importance of each index, and generate a fuzzy complementary matrix and the corresponding subjective weights.

[0035] 3) Conduct a consistency test on the index weights and modify the weights to obtain a reasonable subjective weight vector.

[0036] (2) Calculate the objective weights of each index based on BPNN. The specific process is as follows:

[0037] 1) Divide the evaluation index data normalized by the range standardization method into training data sets, test data sets, and validation data sets.

[0038] 2) Set the number of output layers and hidden layers, and construct a reasonable network structure.

[0039] 3) Run the program to repeatedly loop until the output conditions are met, and finally output reasonable objective weights.

[0040] (3) Couple the subjective weights and objective weights according to the formula to calculate the combined weights, that is, the final weights of the evaluation indexes.

[0041] In the above method, the calculation formula of the combined weight C can be used to obtain the final weights of the indexes.

[0042]

[0043] In the formula, n is the number of index factors; U i is the subjective weight of the i-th index factor; V i is the objective weight of the i-th index factor.

[0044] The above method provided by the present invention is used for the evaluation of soil erosion vulnerability, and its steps include:

[0045] The first step is to determine the values of each parameter factor of the CSLE model according to rainfall data, soil data, digital elevation model data, normalized difference vegetation index data, and land use data: rainfall erosivity factor, soil erodibility factor, slope length factor, slope gradient factor, vegetation cover factor, and water and soil conservation measure factor.

[0046] The second step is to calculate the soil erosion modulus of each grid cell by using the Revised Universal Soil Loss Equation (RUSLE) model based on the calculated values of soil erosion factors on the ArcGIS platform, and calculate and superimpose the rainfall erosivity factor R, soil erodibility factor K, slope length factor L, slope gradient factor S, vegetation cover factor C, and water and soil conservation measure factor P.

[0047] In the third step, based on the ArcGIS platform, multiply the data of each evaluation index of the normalized regional soil erosion vulnerability by the corresponding index weight obtained in step S5. According to "soil erosion vulnerability = exposure + sensitivity - adaptability", finally obtain the soil erosion vulnerability value of the typical mountainous area (such as the mountainous area in southwestern Hubei).

[0048] The present invention has the following main advantages:

[0049] 1. A proposed method for evaluating regional soil erosion vulnerability makes up for the deficiencies and gaps in the current domestic evaluation in the field of natural disasters, which mainly focuses on the evaluation of sudden disasters.

[0050] 2. Aiming at the single weight assignment method in the previous evaluation methods, the combined weight assignment method can reduce the error of the evaluation index weight and fully take into account the directivity of the subjective weight assignment method and the objectivity of the objective weight assignment method.

[0051] 3. The proposed soil erosion vulnerability evaluation index system has universality and applicability for promotion to the regional and large-scale ranges, and can provide a scientific basis and technical support for the reasonable planning of regional land use, the effective management of water and soil resources, the restoration of the ecological environment, and the construction of a modern model of harmonious coexistence between humans and nature. Brief Description of the Drawings

[0052] Figure 1 It is a training flow chart of the BPNN (Back Propagation Neural Network) network model of the present invention.

[0053] Figure 2 It is a soil erosion vulnerability grading map of typical mountainous areas in Hubei Province from 2010 to 2020 (taking the mountainous area in southwestern Hubei as an example). Detailed Embodiment

[0054] The regional soil erosion vulnerability evaluation method based on the combined weighting method disclosed by the present invention. At present, in the research on regional and large-scale soil erosion vulnerability evaluation schemes, the regional soil erosion vulnerability evaluation method based on the combined weighting method is relatively rare. Although some scholars have established soil erosion vulnerability evaluation methods for some ecologically fragile areas, there are the following deficiencies: At present, the methods used by domestic and foreign scholars to weight evaluation indicators mainly use single subjective weighting methods and objective weighting methods, such as expert scoring method, analytic hierarchy process, entropy weight method, grey relational method, etc. These methods are not applicable to the non-linear analysis of complex systems and cannot meet the needs of soil erosion vulnerability evaluation; At present, most of the vulnerability research focuses on the research of sudden natural disasters such as landslides and debris flows, while less research is done on conventional natural disasters such as desertification and soil and water loss. In addition, the existing regional soil erosion vulnerability evaluation methods focus on small watersheds or small-scale ecologically fragile areas. The evaluation indicators are oriented to the unique characteristics of ecologically fragile areas, which have certain limitations and do not have the universality and practicability for promotion to regions and large-scale administrative regions. The present invention proposes a soil erosion vulnerability evaluation method coupling the human-land relationship based on the combined weighting method. The selection of evaluation indicators and the establishment of the evaluation system have the universality and applicability for promotion to regions and large-scale ranges. The proposed technical method is used to solve the problems existing in the prior art.

[0055] The present invention will be further described below in conjunction with embodiments and drawings, but is not limited to the following content.

[0056] The regional soil erosion vulnerability evaluation method based on the combined weighting method provided by the present invention includes the following steps:

[0057] Step S1, obtain digital elevation model (DEM, Digital Elevation Model) data, topographic map data, normalized difference vegetation index (NDVI, Normalized Digital Vegetation Index) data, land use data, soil data, rainfall data, GDP data, population data, road network data and other required socio-economic data, and select the typical mountainous area in Hubei Province (taking the mountainous area in southwestern Hubei as an example) as the research area.

[0058] The digital elevation model data is: The digital elevation model (DEM, Digital Elevation Model) data processed based on ASTGTM data with a resolution of 30m is used to calculate the L and S factors.

[0059] The normalized difference vegetation index (NDVI) data is the NDVI data processed from Landsat 7 with a resolution of 30m, which is used to calculate the C factor.

[0060] The land use data is the land use data with a resolution of 30m, which is used to calculate the P factor.

[0061] The soil data is the soil physical and chemical property data of the Harmonized World Soil Database (HWSD) with a resolution of 1km, which is used to calculate the K factor.

[0062] The rainfall data is the daily precipitation monitoring data of each monitoring station in the study area, which is used to calculate the R factor.

[0063] The GDP data is obtained from the statistical yearbooks of each year in the study area and is used to calculate the economic density.

[0064] The population data is obtained from the statistical yearbooks of each year in the study area and the population census data, which is used to calculate the population density.

[0065] The road network data is obtained from OpenStreetMap (OSM) with a resolution of 1:250,000, which is used to calculate the road network density.

[0066] The social and economic data are the various social and economic factors in the method for evaluating the vulnerability of regional soil erosion.

[0067] In the present invention, a resampling tool is used to set the spatial resolution of all the above data to 30m and unify the coordinate systems of the used data to WGS1984, which is beneficial to calculating the erosion modulus and performing spatial analysis in combination with the RUSLE model.

[0068] Step S2, data preprocessing.

[0069] It is completed by using ENVI professional software. Through ENVI software, radiometric calibration, atmospheric correction, and image fusion tasks are performed on the downloaded remote sensing images; the social and economic data is imported into ArcGIS and saved as a raster file.

[0070] Step S3, training of the BPNN machine learning network model, including the following steps:

[0071] Step S31, normalization and classification of input data.

[0072] The range normalization method is used to normalize the data of each index in the index layer, and the normalized data is used as the input data. Then the input data is divided into a training data set, a test data set, and a validation data set. The training data set is presented to the network during the training process and accounts for 70% of the sample data. The test data set does not affect the training and provides an independent measurement of the network performance during and after training, which accounts for 15% of the sample data. The validation data set is used to measure the network generalization and stop training when the generalization stops improving, accounting for 15% of the sample data.

[0073] Step S32, construct the network and start training.

[0074] Set the indicator layer to the number of input neurons, set the target layer to the number of output layers, and set the number of hidden layers to 10. By constructing a 13-10-1 network structure, a BPNN algorithm program written based on Matlab software, where the initial weight and threshold matrices are randomly generated, the activation function uses the sigmoid function (S-shaped function), the learning rate is 0.1, the maximum number of iterations is 1000, and the expected error is 0.5×10 3 , and the performance function is the mean square error (MSE). After the training data set propagates forward, if the error does not meet the preset accuracy, the error will propagate backward from the output layer to the input layer, and the error will be gradually reduced by adjusting the weights and thresholds, thereby obtaining the objective weight V of each index layer.

[0075] Step S33, test the accuracy of the training.

[0076] The r value is used to measure the performance of the model network. r is the Pearson product-moment correlation coefficient. The closer the r value is to 1, the better the training performance of the BPNN network model.

[0077] Step S4, assign weights to each index layer using FAHP (Fuzzy Analytic Hierarchy Process).

[0078] Step S41, establish a fuzzy complementary judgment matrix.

[0079] When using the fuzzy analytic hierarchy process, generally, the relative importance of a certain factor compared to other factors is used to quantitatively judge the importance of two factors. The factors are compared pairwise to obtain the matrix A=(a ij ) n×n , and determine whether it has the following two properties:

[0080] 1) 0 < a ij < 1 ( i = 1, 2,..., n)

[0081] 2) a ij + a ji = 1 (i = 1, 2, ..., n; j = 1, 2, ..., n)

[0082] Where: aij is the element in the fuzzy judgment matrix, and i and j represent its row number and column number respectively.

[0083] If the matrix meets the above property requirements, then this matrix is called a fuzzy complementary judgment matrix. In order to quantitatively describe the importance degree of each index, the 0.1 - 0.9 scale method is used to calibrate the relative importance degree between the indexes, and the scale meaning is shown in Table 1.

[0084] Table 1 The meaning of the quantitative scale of the fuzzy complementary judgment matrix

[0085]

[0086] Based on the above numbers as the scale basis, the values of a1, … a n and other factors are compared and obtained, and the fuzzy complementary matrix is as shown in Equation (1):

[0087]

[0088] Where: n represents the order of the matrix.

[0089] Step S42, determine the matrix weight.

[0090] Sum the fuzzy judgment matrix A by rows:

[0091]

[0092] Perform mathematical substitution:

[0093]

[0094] Where: r ij represents the element after summing by rows in the fuzzy judgment matrix, and i and j represent its row number and column number respectively.

[0095] The consistency matrix r = (r ij ) m×n , and the sorting vector w = (w1, w2, … w n ) T is obtained from the consistency matrix r by the row - sum normalization method, where the w vector satisfies:

[0096]

[0097] Where, a ij is the element in the fuzzy judgment matrix, and i and j represent its row number and column number respectively.

[0098] Thus, the subjective weight U obtained based on FAHP is obtained.

[0099] Step S43, Consistency test of fuzzy complementary judgment matrix

[0100] Define the geometry of all n-order fuzzy complementary judgment matrices as G n . Let A = (a ij ) n×n and B = (b ij ) n×n ∈G n , and use the norm to represent the distance between A and B, denoted as ρ(A, B) = ||A - B||.

[0101] Let A = (a ij ) n×n and B = (b ij ) n×n both be fuzzy judgment matrices, and call the compatibility degree between A and B. The compatibility index between A and B is

[0102] Define w = (w1, w2,... w n ) T as the weight vector of matrix A, is the eigenmatrix of judgment matrix A.

[0103] For the attitude α of the decision maker, when the compatibility index I(A, W * ) ≤ α, it is considered to meet the consistency requirement. The smaller the value of α, the higher the requirement of the project decision maker for consistency. Usually, α = 0.1 is taken without special circumstances.

[0104] Step S5, Calculate the combined weight. The calculation formula of the combined weight C i is as follows:

[0105]

[0106] In the formula, n is the number of index factors; U i is the subjective weight of the i-th index factor; V i is the objective weight of the i-th index factor.

[0107] Step S6, Determine the soil erosion modulus according to the digital elevation model data, normalized difference vegetation index data, land use data, soil data, and rainfall data.

[0108] The specific steps are as follows:

[0109] In the first step, determine the values of the parameter factors of the CSLE model based on rainfall data, soil data, digital elevation model data, normalized difference vegetation index data, and land use data: rainfall erosivity factor, soil erodibility factor, slope length factor, slope gradient factor, vegetation cover factor, and soil and water conservation measure factor.

[0110] The RUSLE model is as follows:

[0111] A = R·K·L·S·CP

[0112] In the formula, A is the soil erosion modulus, with the unit t / (hm 2 ·a); R is the rainfall erosivity factor, with the unit MJ·mm / (hm 2 ·h·a), K is the soil erodibility factor, with the unit t·hm 2 ·h / (hm 2 ·MJ·mm); L is the slope length factor, dimensionless; S is the slope gradient factor, dimensionless; C is the vegetation cover factor, dimensionless; P is the soil and water conservation measure factor, dimensionless.

[0113] (1) Rainfall erosivity factor R:

[0114] First, obtain the half-monthly precipitation data using the spatial interpolation method based on the daily precipitation monitoring data. To reduce the error value of the precipitation data, at the same time, obtain the half-monthly precipitation data using the IDW and Kriging interpolation methods, and compare the half-monthly precipitation data obtained by the two different methods with the measured data of the hydrological monitoring station, and select the one with smaller error to calculate the R value.

[0115]

[0116] In the formula, Z*(S0) is the interpolation result of the interpolation point S0; Z(S i ) is the measured value of the measured point S i ; N is the number of measured points participating in the calculation; λ i is the weight coefficient; d i0 is the distance between the measured point S i and the interpolation point S0; p is the power of the distance, generally taking p = 1 or p = 2.

[0117]

[0118] In the formula, Z*(S0) is the interpolation result of the interpolation point S0; Z(S i ) is the measured value of the measured point S i ; N is the number of measured points participating in the calculation; λ i is the Kriging weight coefficient, obtained from the variogram, rather than obtained through the distance between the measured point and the interpolation point; to ensure unbiased estimation, the sum of the weights is 1, that is, Σλi = 1.

[0119] The mean absolute error, mean relative error, root mean square error, and Pearson product-moment correlation coefficient were used as evaluation indicators to compare the measured and interpolated results of the half-month precipitation at each station.

[0120]

[0121]

[0122]

[0123]

[0124] Where p i , o i are the interpolated result and the measured value of the i-th station, respectively; n is the number of verification stations; are the means of the n interpolated results and measured values, respectively. By comparing the measured and interpolated results of the half-month precipitation at each station, a more accurate interpolated result was determined to calculate the R value.

[0125]

[0126]

[0127]

[0128] Where is the annual average rainfall erosivity over the years, MJ·mm / (hm 2 ·h·a); k takes values from 1, 2,..., 24, indicating that a year is divided into 24 half-months: is the rainfall erosivity of the k-th half-month, MJ·mm / (hm 2 ·h); i takes values from 1, 2,..., N; N refers to the time series from 2000 to 2020; j takes values from 0, 1,..., m; m is the number of erosive rainfall days in the k-th half-month of the i-th year (erosive rainfall days refer to days with daily rainfall greater than or equal to 10 mm); P i is the erosive rainfall amount of the j-th erosive rainfall in the k-th half-month of the i-th year, mm; if there is no erosive rainfall amount in a certain half-month of a certain year, that is, j = 0, then let P i,o,k = 0; α is a parameter, taking 0.3937 in the warm season (May - September) and 0.3101 in the cold season (October - December, January - April); is the average rainfall erosivity of the k-th half-month accounting for the proportion of the annual average rainfall erosivity over the years .

[0129] (2) Soil erodibility factor (K):

[0130]

[0131] Wherein, SIL is the soil silt content (%), CLA is the soil clay content (%), SAN is the soil sand content (%), C is the soil organic carbon content (%), SNI is a constant, and SNI = 1 - SAN / 100.

[0132] (3) Slope factor (S):

[0133]

[0134] Wherein, θ represents the slope (°).

[0135] (4) Slope length factor (L):

[0136]

[0137]

[0138] Wherein, L is the slope length factor; λ is the slope length; m is the slope length index, and θ is the slope.

[0139] (5) Vegetation cover factor (C):

[0140]

[0141] The calculation formula for the vegetation coverage fc is as follows:

[0142]

[0143] Wherein, fc is calculated based on the pixel dichotomy method according to NDVI data, NDVI soil represents the NDVI value of a pure bare soil pixel, NDVI max is the NDVI value of a pure vegetation pixel.

[0144] (6) Soil and water conservation measure factor (P)

[0145] Refer to the soil and water conservation measure assignment table for different land use types (Table 2) to obtain the soil and water conservation measure factor value. As shown in Table 2, assign values to the soil and water conservation measure factors for different land types.

[0146] Table 2 Assignment table of soil and water conservation measure factors under different land use types

[0147]

[0148]

[0149] In the second step, based on the ArcGIS platform, using the calculated values of soil erosion factors, the Revised Universal Soil Loss Equation (RUSLE model) is applied to calculate and superimpose the rainfall erosivity factor R, soil erodibility factor K, slope length factor L, slope gradient factor S, vegetation cover factor C, and soil and water conservation measure factor P to obtain the soil erosion modulus of each grid cell.

[0150] In the third step, based on the ArcGIS platform, multiply the normalized data of each evaluation index of regional soil erosion vulnerability by the corresponding index weights obtained in step S5. According to "soil erosion vulnerability = exposure + sensitivity - adaptability", finally obtain the soil erosion vulnerability value of the typical mountainous areas in Hubei Province (taking the mountainous areas in southwestern Hubei as an example). Use the natural breakpoint method to divide the soil erosion vulnerability into 5 categories: low, relatively low, moderate, relatively high, and high, and generate a soil erosion vulnerability classification map. Using the classification map, the spatial distribution pattern and temporal changes of soil erosion vulnerability in the typical mountainous areas in Hubei Province (taking the mountainous areas in southwestern Hubei as an example) can be intuitively understood, and it provides theoretical and technical support for the rational allocation of soil and water conservation measures and soil and water conservation planning in Hubei Province.

[0151] Appendix:

[0152] Table 3 Regional Soil Erosion Vulnerability Evaluation Index System and Its Weights

[0153]

[0154] Application Example 1: The present invention calculates the weights of each index involved in the regional soil erosion vulnerability evaluation index system based on the combined weighting method.

[0155] ① First, calculate the subjective weights of each index based on FAHP. The specific process is as follows:

[0156] 1. Make and distribute questionnaires based on the evaluation index system to obtain the opinions of experts in the field of soil erosion and soil and water conservation.

[0157] 2. Summarize the scoring results of each expert on the relative importance of each index, generate a fuzzy complementary matrix and the corresponding subjective weights.

[0158] 3. Conduct a consistency test on the index weights and modify the weights to obtain a reasonable subjective weight vector.

[0159] ② Then, calculate the objective weights of each index based on BPNN. The specific process is as follows:

[0160] 1. Divide the prepared data into training data sets, test data sets, and validation data sets.

[0161] 2. Set the number of output layers and hidden layers to construct a reasonable network structure.

[0162] 3. The running program loops repeatedly until the output condition is met, and finally outputs reasonable objective weights.

[0163] ③Finally, based on the combined weight calculation formula, the subjective weight and the objective weight are coupled to calculate the final weights of each index.

[0164] Application Example 2: The present invention selects a typical hilly area in Hubei Province (taking the hilly area in southwestern Hubei as an example) as the study area, and the area is planned to be about 27,900 km 2 , and quantitatively evaluates the soil erosion vulnerability of the typical hilly area in Hubei Province (taking the hilly area in southwestern Hubei as an example) and conducts soil erosion vulnerability grading.

[0165] ①The modified universal soil loss equation (RUSLE) is used to calculate the soil erosion amount in the study area. The specific process is as follows:

[0166] 1. Through the ENVI software, perform radiometric calibration, atmospheric correction, and image fusion tasks on the downloaded remote sensing images; import the social and economic data into ArcGIS and save it as a raster file.

[0167] 2. According to the rainfall data, soil data, digital elevation model data, normalized difference vegetation index data, and land use data, combine the formula to determine the values of each parameter factor of the RUSLE model: rainfall erosivity factor, soil erodibility factor, slope length factor, slope gradient factor, vegetation cover factor, and soil and water conservation measure factor.

[0168] 3. Based on the ArcGIS platform, using the calculated values of the soil erosion factors, apply the modified universal soil loss equation (RUSLE) to calculate and superimpose the rainfall erosivity factor R, soil erodibility factor K, slope length factor L, slope gradient factor S, vegetation cover factor C, and soil and water conservation measure factor P to obtain the soil erosion modulus of each grid cell.

[0169] ②Then calculate the regional soil erosion vulnerability value. The specific process is as follows:

[0170] 1. Based on the ArcGIS platform, multiply the normalized data of each evaluation index of the regional soil erosion vulnerability by its corresponding index weight to calculate the values of the index layer and factor layer in the study area.

[0171] 2. According to the formula "vulnerability = exposure + sensitivity - adaptability", obtain the soil erosion vulnerability value of the typical hilly area in Hubei Province (taking the hilly area in southwestern Hubei as an example).

[0172] 3. In ArcGIS, based on the natural breaks method, divide the soil erosion vulnerability into 5 categories: low, relatively low, moderate, relatively high, and high, and generate a soil erosion vulnerability grading map, as Figure 2 shown.

[0173] ③ Analyze the spatio-temporal changes in the soil erosion vulnerability of typical mountainous areas in Hubei Province (taking the mountainous areas in southwestern Hubei as an example).

[0174] By using analysis tools such as classification maps combined with the geographical detector model, the spatial distribution pattern and temporal changes of the soil erosion vulnerability in typical mountainous areas of Hubei Province (taking the mountainous areas in southwestern Hubei as an example) can be intuitively understood, the dominant factors affecting soil vulnerability can be explored, and theoretical and technical support can be provided for the rational allocation of soil and water conservation measures and soil and water conservation planning in Hubei Province.

[0175] It is worth reminding that:

[0176] 1. The present invention is not limited to the embodiments described above. The description of the specific embodiments above is intended to describe and illustrate the technical solutions of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the spirit of the present invention and the scope protected by the claims, those of ordinary skill in the art can make many specific transformations in various forms under the inspiration of the present invention, and these all fall within the protection scope of the present invention.

[0177] 2. The replacement of the calculation methods of each factor of the RUSLE model should belong to the protection scope of the present invention.

[0178] 3. The differences in the weight due to different evaluation scoring criteria of the subjective weighting method in the combined weighting method and the differences in the calculation methods due to the differences in the specific program design of the objective weighting method should belong to the scope of the present invention.

[0179] 4. The replacement and addition or subtraction of some evaluation indicators in the evaluation index system provided by the present invention should belong to the protection scope.

Claims

1. A method for evaluating the vulnerability of regional soil erosion based on the combined weighting method, characterized in that: First, calculate the subjective weights of each index based on the Fuzzy Analytic Hierarchy Process (FAHP). The process includes assigning weights to each index layer using FAHP. Then, calculate the objective weights of each index based on the Back Propagation Neural Network (BPNN). The process includes steps such as normalizing and classifying the input data, constructing the network, and starting to train and test the training accuracy. Then, obtain the final weights of each index according to the combined weight calculation formula. (1) Establish a fuzzy complementary judgment matrix. (2) Determine the weights of the judgment matrix. Among them, the following method is used to test the consistency of the fuzzy complementary judgment matrix: Define the geometry of all n-order fuzzy complementary judgment matrices as G n ; Let A = (a ij ) n×n and B = (b ij ) n×n ∈ G n , use the norm to represent the distance between A and B, denoted as ρ(A, B) = ||A - B||, where a ij is an element in the fuzzy judgment matrix, and i and j represent its row number and column number respectively. Let \(A=(a ij ) n×n and \(B=(b ij ) n×n both be fuzzy judgment matrices. It is called the compatibility degree between \(A\) and \(B\); the compatibility index between \(A\) and \(B\) is Define \(w=(w_1, w_2, \ldots, w\) n ) T as the weight vector of matrix \(A\), as the eigenmatrix of the judgment matrix \(A\). For the attitude α of the decision maker, when the compatibility index I(A, W * ) ≤ α, it is considered to meet the consistency requirement. Among them, A is the fuzzy judgment matrix, A = (a ij ) n×n , and W* is the eigenmatrix of the judgment matrix A. The smaller the value of α, the higher the requirement of the project decision maker for consistency, and α = 0.

1.

2. The regional soil erosion vulnerability evaluation method according to claim 1, characterized in that: Use the Revised Universal Soil Loss Equation (RUSLE) model to calculate the soil erosion amount in the study area. Then, combine the specific evaluation index system and its corresponding weights to calculate the soil erosion vulnerability value of the study area. Based on the natural break method in ArcGIS, the soil erosion vulnerability is divided into 5 categories: low, relatively low, moderate, relatively high, and high, and a soil erosion vulnerability classification map is generated. Using the classification map, the spatial distribution pattern and temporal variation characteristics of the soil erosion vulnerability in the region and large-scale range can be intuitively understood, and it provides theoretical and technical support for the reasonable allocation of soil and water conservation measures and soil and water conservation planning in the region and large-scale range.

3. The regional soil erosion vulnerability evaluation method according to claim 1, characterized in that: First, obtain social and economic data, perform data preprocessing, train using the BPNN machine learning network model, assign weights to each index layer using FAHP, calculate the combined weights, and then determine the soil erosion modulus based on digital elevation model data, normalized difference vegetation index data, land use data, soil data, and rainfall data.

4. The regional soil erosion vulnerability evaluation method according to claim 3, characterized in that The following method is used for data preprocessing: For the downloaded remote sensing images, perform radiometric calibration, atmospheric correction, and image fusion tasks through ENVI software; import the social and economic data into ArcGIS and save it as a raster file.

5. The regional soil erosion vulnerability evaluation method according to claim 3, characterized in that The following method is used to train using the BPNN machine learning network model and obtain the objective weights: (1) Normalize and classify the input data: Use the range normalization method to normalize the data of each index in the index layer and use the normalized data as the input data; then divide the input data into a training data set, a test data set, and a validation data set. The training data set is presented to the network during training and accounts for 70% of the sample data; the test data set does not affect training and provides an independent measurement of the network performance during and after training, which explains 15% of the sample data; the validation data set is used to measure the network generalization and stop training when the generalization stops improving, accounting for 15% of the sample data. (2) Construct the network and start training: Set the indicator layer to the number of input neurons, set the target layer to the number of output neurons, and set the number of hidden layers to 10; by constructing a 13-10-1 network structure, a BPNN algorithm program written based on Matlab software, where the initial weight and threshold matrices are randomly generated, the activation function uses the sigmoid function, the learning rate is 0.1, the maximum number of iterations is 1000, and the expected error is 0.5×10 3 , and the performance function is the mean square error; After the training data set is propagated forward, if the error does not meet the preset accuracy, the error will be propagated backward from the output layer to the input layer, and the error will be gradually reduced by adjusting the weights and thresholds, thereby obtaining the objective weights V of each index layer. (3) Test the training accuracy: Use the Pearson product-moment correlation coefficient r value to measure the performance of the model network. The closer the r value is to 1, the better the training performance of the BPNN network model.

6. The regional soil erosion vulnerability evaluation method according to claim 3, wherein The following method is used to assign weights to each index layer using FAHP: (1) Establish a fuzzy complementary judgment matrix: When using the fuzzy analytic hierarchy process, first, the relative importance of two factors is quantitatively judged by comparing one factor with other factors. The fuzzy judgment matrix A=(a ij ) n×n is obtained by pairwise comparison of all factors; then, the values of factors a1,…a n are compared and taken to obtain the following fuzzy complementary matrix: where n represents the order of the matrix; (2) Determine the weights of the judgment matrix: After summing the rows of the fuzzy judgment matrix A, perform a mathematical substitution: Let the obtained consistency matrix R = (r ij ) m×n , and use the row sum normalization method for the consistency matrix R to find the sorting vector, w = (w1, w2, … w n ) T , where the w vector satisfies: where a ij is an element in the fuzzy judgment matrix, and i and j represent its row number and column number respectively; Thus, obtain the subjective weight U obtained by FAHP.

7. The regional soil erosion vulnerability evaluation method according to claim 3, characterized in that The combined weights of evaluation indicators are calculated using the following method: (1) First, calculate the subjective weights of each indicator based on FAHP. The specific process is as follows: 1) Make and distribute questionnaires based on the evaluation indicator system to obtain the opinions of experts in the field of soil erosion and water and soil conservation; 2) Summarize the scoring results of each expert on the relative importance of each indicator to generate a fuzzy complementary matrix and the corresponding subjective weights; 3) Conduct a consistency test on the indicator weights and modify the weights to obtain a reasonable subjective weight vector; (2) Calculate the objective weights of each indicator according to BPNN. The specific process is as follows: 1) Divide the evaluation indicator data normalized using the range standardization method into training data sets, test data sets, and validation data sets; 2) Set the number of output layers and hidden layers to construct a reasonable network structure; 3) Run the program to loop repeatedly until the output conditions are met, and finally output reasonable objective weights; (3) Couple the subjective weights and objective weights according to the formula to calculate the combined weights.

8. The regional soil erosion vulnerability assessment method according to claim 7, characterized in that Use the calculation formula of the combined weight C’: where n is the number of index factors; U i’ is the subjective weight of the i'-th index factor; V i’ is the objective weight of the i'-th index factor.

9. The method according to any one of claims 1 to 8, characterized in that For the evaluation of regional soil erosion vulnerability, the steps include: The first step is to determine the values of each parameter factor of the CSLE model based on rainfall data, soil data, digital elevation model data, normalized difference vegetation index data, and land use data: rainfall erosivity factor, soil erodibility factor, slope length factor, slope gradient factor, vegetation cover factor, and water and soil conservation measure factor; The second step is to use the calculated values of soil erosion factors based on the ArcGIS platform and apply the Revised Universal Soil Loss Equation (RUSLE) to calculate and superimpose the rainfall erosivity factor R, soil erodibility factor K, slope length factor L, slope gradient factor S, vegetation cover factor C, and water and soil conservation measure factor P to obtain the soil erosion modulus of each grid cell; The third step is to multiply the normalized data of each evaluation indicator of regional soil erosion vulnerability obtained based on the ArcGIS platform by the corresponding indicator weights obtained in step S5. According to "soil erosion vulnerability = exposure + sensitivity - adaptability", finally obtain the soil erosion vulnerability value of the typical mountainous area of the sample.

Images