A fitting method for soil heavy metal background values in basalt area based on major element geochemical constraints
By constructing a multivariate log-linear regression model based on major element geochemical constraints, the problems of accuracy and repeatability in fitting heavy metal background values in basalt areas were solved, and the accurate determination of soil heavy metal background values was achieved. This model is applicable to pollution assessment and ecological risk assessment in basalt areas.
Patent Information
- Application Number
- CN202610375067.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-25
- Publication Date
- 2026-07-03
AI Technical Summary
Existing technologies struggle to accurately distinguish between natural accumulation and human input of heavy metals in basalt areas, and the lack of standardized modeling processes results in insufficient accuracy and repeatability in fitting heavy metal background values.
A method based on geochemical constraints of major elements was adopted. Target major elements were screened through data preprocessing, correlation and collinearity analysis, a multivariate log-linear regression model was constructed, outliers were removed and the model was iteratively optimized to accurately determine the background values of heavy metals in soil.
It improves the fitting accuracy and model repeatability of heavy metal background values in basalt soils, adapts to unique geochemical characteristics, and provides reliable pollution assessment and ecological risk assessment data.
Smart Images
Figure CN122337376A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of soil environmental geochemistry, specifically to a method for fitting background values of heavy metals in basaltic soils based on major element geochemical constraints. Background Technology
[0002] Background values of heavy metals in soil are crucial parameters for soil pollution assessment, environmental quality classification, and ecological risk assessment. The scientific validity of their determination directly impacts the rationality of environmental regulation and governance decisions. Basaltic soil parent material is characterized by its simple composition, high iron and magnesium content, and natural enrichment of heavy metals. Furthermore, significant differences in weathering and leaching levels across different regions, coupled with systematic variations in the distribution of major elements, lead to large natural variations in heavy metal content. Moreover, heavy metals introduced by human activities easily overlap with high natural background levels, making the distinction between natural background and human disturbance in this region far more difficult than in other lithological areas.
[0003] Currently, the determination of soil heavy metal background values mainly employs regional statistical methods, iterative elimination methods, and multivariate statistical auxiliary analysis methods. These methods are mostly based on the statistical characteristics of heavy metal concentrations themselves, without fully considering the geochemical constraints of major elements on heavy metal migration and enrichment. They struggle to effectively distinguish between natural heavy metal enrichment and anthropogenic input in basalt areas, easily leading to overestimation or underestimation of background values. Furthermore, existing modeling studies largely remain at the empirical statistical level, lacking standardized procedures. The models suffer from insufficient repeatability and fitting accuracy, failing to adapt to the soil geochemical characteristics of basalt areas and thus failing to meet the practical needs for accurate background value determination in this region. Summary of the Invention
[0004] To address the aforementioned problems, the present invention aims to provide a method for fitting heavy metal background values in basalt soils based on major element geochemical constraints. This method overcomes the technical deficiencies of traditional methods, accurately determines heavy metal background values in basalt soils, effectively distinguishes between natural enrichment and human input, and improves model fitting accuracy and process repeatability.
[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: a method for fitting background values of heavy metals in basaltic soils based on major element geochemical constraints, comprising the following steps:
[0006] S1. Data Collection: Obtain soil geochemical survey datasets for the basalt area to be tested;
[0007] S2. Data preprocessing: After removing invalid sample data from the soil geochemical survey dataset, a natural logarithmic transformation was performed.
[0008] S3. Correlation analysis of indicators: Correlation analysis and collinearity analysis were performed on the soil geochemical survey dataset after data preprocessing to obtain the correlation analysis results and collinearity analysis results;
[0009] S4. Screening of major elements: Based on the results of correlation analysis and collinearity analysis, and combined with the geochemical significance of major elements, a set of target major element indicators is screened.
[0010] S5. Regression Model Construction: Based on the target principal element index set, construct a multiple log-linear regression model between each heavy metal element and the target principal element index set to obtain the multiple log-linear regression model of each heavy metal element.
[0011] S6. Outlier identification and removal: The fitting results of the multiple log-linear regression models for each heavy metal element are evaluated to determine whether to rebuild the model and convert the background value of heavy metals in the soil.
[0012] S7. Conversion of Soil Heavy Metal Background Values: Substitute the concentration data of the target major element index set corresponding to each data sampling point in the basalt area to be tested into the multivariate log-linear regression model to calculate the corresponding heavy metal predicted values. The mean of all heavy metal predicted values of the same heavy metal element category is used as the natural background value of soil heavy metals in the basalt area to be tested.
[0013] Furthermore, the soil geochemical survey dataset consists of several soil geochemical survey data points, and each soil geochemical survey data point uniquely corresponds to a data sampling point; the soil geochemical survey data includes major element concentration data and heavy metal element concentration data, wherein the major element categories in the major element concentration data include at least... and The heavy metal element categories in the heavy metal element concentration data should include at least the following: and .
[0014] Furthermore, the invalid sample data removal refers to removing soil geochemical survey data containing null values for element concentrations or incomplete element categories. Null values for element concentrations refer to cases where the concentration data of major elements or heavy metals in the soil geochemical survey data are empty. Incomplete element categories refer to cases where the major element categories in the soil geochemical survey data do not meet the requirement of including at least [missing information]. and The situation or the heavy metal element category does not meet the requirement of at least including and In this case, the natural logarithmic transformation is performed on the major element concentration data and heavy metal element concentration data of all soil geochemical survey data retained after removing invalid sample data. This transformation is used to weaken the right-skewed distribution characteristics that are common in geochemical data, so as to improve the normality and modelability of the soil geochemical survey dataset after data preprocessing.
[0015] Furthermore, the correlation analysis employs Pearson linear correlation analysis to conduct quantitative analysis on the preprocessed soil geochemical survey dataset, and simultaneously analyzes the calculated correlation coefficients. Significance tests were performed, with <0.05 as the significant correlation threshold and <0.01 as the highly significant correlation threshold. Major element indicators that reached the significant correlation threshold with heavy metal elements were included in the subsequent modeling candidate independent variable range, resulting in candidate major element indicators, i.e., correlation analysis results. This is used to ensure that there is a real geochemical correlation between the modeling independent variables and heavy metal elements, and to avoid model fitting bias caused by the introduction of irrelevant indicators. The analysis objects of the correlation analysis are major element indicators and heavy metal elements. The major element indicators include major elements and derived indicators, wherein the derived indicators include at least one of the Parker weathering index (WIP), chemical alteration index (CIA), and chemical weathering index (CIX).
[0016] The formula for calculating the correlation coefficient using the Pearson linear correlation analysis method is as follows:
[0017] ;
[0018] in, This represents the total number of soil geochemical survey data points in the preprocessed soil geochemical survey dataset. Major element indices in the i-th soil geochemical survey data in the preprocessed soil geochemical survey dataset Concentration value after natural logarithmic transformation Major element indices in all soil geochemical survey data within the preprocessed soil geochemical survey dataset. The arithmetic mean of the concentration values after natural logarithmic transformation. For the heavy metal elements in the i-th soil geochemical survey data in the preprocessed soil geochemical survey dataset Concentration value after natural logarithmic transformation Heavy metal elements in all soil geochemical survey data in the preprocessed soil geochemical survey dataset. The arithmetic mean of the concentration values after natural logarithmic transformation. Main element indicators With heavy metal elements The correlation coefficient;
[0019] The derived indices are constructed using major element concentration data, specifically referring to the literature "A Review of Commonly Used Chemical Weathering Indicators: Application and Prospect" (Journal of Geology of Chinese Universities, 2022, 28(1):51-63). The construction formulas for each derived index are as follows:
[0020] Parker Weathering Index (WIP): ;
[0021] Chemical Alteration Index (CIA): ;
[0022] Chemical Weathering Index (CIX): ;
[0023] Where n is the molar content of the corresponding oxide, It is CaO in soil silicate minerals.
[0024] Furthermore, the collinearity analysis employs either the correlation coefficient matrix analysis method or the variance inflation factor (VIF) analysis method, and the analysis object of the collinearity analysis is the principal element index.
[0025] The correlation coefficient matrix analysis method uses Pearson linear correlation analysis to calculate the correlation coefficient between any two candidate principal element indicators. At the same time, a strong multicollinearity judgment threshold of |r| is set. If the absolute value of the correlation coefficient between two candidate principal element indicators satisfies |r|, it is determined that there is strong multicollinearity between the two. The candidate principal element indicators with strong multicollinearity are recorded as the collinearity analysis result.
[0026] The variance inflation factor (VIF) analysis method constructs a univariate linear regression model with any candidate principal element index as the dependent variable and all other candidate principal element indices as independent variables. The VIF value of the candidate principal element index is calculated to determine its degree of collinearity with other candidate principal element indices. Tolerance level (TOL) is introduced as an auxiliary indicator; TOL is inversely related to VIF, and its mathematical expression is: In the formula, To determine the tolerance of the index of the j-th candidate principal element, the collinearity criterion is set as follows: <10 and When the index is greater than 0.1, it is determined that there is no significant multicollinearity between the j-th candidate principal element index and other candidate principal element indices. ≥10 and When the index is ≤0.1, it is determined that there is strong multicollinearity between the j-th candidate principal element index and other candidate principal element indices, and the candidate principal element indices with strong multicollinearity are recorded as the collinearity analysis results.
[0027] The formula for calculating the variance inflation factor (VIF) is as follows:
[0028] ;
[0029] in: Let be the variance inflation factor of the j-th candidate principal element index. Let be the coefficient of determination of a univariate linear regression model constructed with the j-th candidate principal variable as the dependent variable and the remaining candidate principal variables as independent variables. The value range is [0,1].
[0030] Furthermore, the multivariate log-linear regression model is obtained by importing the concentration index data corresponding to the target principal element index set and the preprocessed heavy metal element concentration data into statistical analysis software. The statistical analysis software also simultaneously outputs the coefficient of determination corresponding to the multivariate log-linear regression model. And the significance level p, used to determine the fitting results;
[0031] The concentration index data corresponding to the target major element index set are the major element concentration data of each effective data sampling point in the basalt area to be tested corresponding to the target major element index set after S2 data preprocessing, and the weathering-derived index calculated from the major element concentration data; the heavy metal element concentration data after data preprocessing are the heavy metal element concentration data corresponding to each effective data sampling point in the soil geochemical survey dataset after data preprocessing.
[0032] Furthermore, the fitting result is determined as follows:
[0033] When the coefficient of determination If the value is greater than 0.8 and the significance level is less than 0.001, then the corresponding multiple log-linear regression model will be applied to the conversion of soil heavy metal background values.
[0034] When the coefficient of determination If the significance level is ≤0.8 or p≥0.001, the corresponding multivariate log-linear regression model will be reconstructed.
[0035] Furthermore, the model reconstruction employs the residual relative deviation threshold method to conduct residual analysis on the multivariate log-linear regression model. The residual analysis involves calculating the residual relative deviation of each data sampling point from the heavy metal element concentration data in the preprocessed soil geochemical survey dataset. The calculation formula is: In the formula, Let represent the measured concentration of heavy metal elements at the i-th data sampling point. The predicted concentrations of heavy metal elements are obtained from a multiple log-linear regression model. A threshold for the residual relative deviation threshold method is set according to the number of model reconstructions. The calculated residual relative deviations of each data sampling point are then used to determine the concentrations. By comparing the threshold of the residual relative deviation threshold method, the soil geochemical survey data corresponding to the abnormal data sampling points in the preprocessed soil geochemical survey dataset are accurately identified and removed, resulting in a new soil geochemical survey dataset after removing the abnormal data sampling points. Based on this new soil geochemical survey dataset and the correlation and collinearity analysis results already completed in the original S3, there is no need to re-conduct the correlation and collinearity analysis of the indicators. Instead, the multiple log-linear regression model between each heavy metal element and the target major element indicator set is directly reconstructed. The newly constructed multiple log-linear regression model should have a determination coefficient that is different from the previously constructed multiple log-linear regression model. If the increase is less than 0.03, then the coefficient of determination is not met. If the increase is less than 0.03, the model should be rebuilt. In this case, the threshold of the residual relative deviation threshold method should be reset according to the number of model rebuilds.
[0036] The threshold value range of the residual relative deviation threshold method is 5%-50%. The threshold value of the residual relative deviation threshold method set according to the number of model reconstructions is as follows: the threshold value of the residual relative deviation threshold method is 50% for the first model reconstruction, 20% for the second model reconstruction, 10% for the third model reconstruction, and 5% for the fourth model reconstruction.
[0037] Furthermore, the coefficient of determination of the fourth model reconstruction of the multiple log-linear regression model compared to the previously constructed multiple log-linear regression model... If the increase is not less than 0.03, then based on the new soil geochemical survey dataset reconstructed by the fourth model, the correlation analysis of the new S3 index is carried out again to obtain the new correlation analysis results and the new collinearity analysis results. Then, based on the new soil geochemical survey dataset and the new correlation analysis results and the new collinearity analysis results, S4-S7 are carried out.
[0038] This invention represents a significant technological advancement over traditional methods for determining soil heavy metal background values. Addressing the shortcomings of traditional methods—such as failing to consider the geochemical constraints of major elements on heavy metals, difficulty in distinguishing between naturally high heavy metal backgrounds and anthropogenic inputs in basalt areas, and lack of standardized modeling processes, resulting in insufficient fitting accuracy and repeatability—this invention deeply integrates major element geochemical constraints with multivariate log-linear statistical modeling. It incorporates basic major elements and derived indicators reflecting basalt weathering characteristics, fundamentally resolving the challenge of distinguishing between natural and anthropogenic interference. Furthermore, a standardized and complete process from data processing to model iterative reconstruction is designed. Through correlation and collinearity analysis, quantitative fitting judgment, and iterative optimization using the residual relative deviation threshold method, the model's fitting accuracy and repeatability are ensured, adapting to the unique soil geochemical characteristics of basalt areas. The resulting soil heavy metal background values are significantly improved in scientific validity and stability, providing more reliable basic data support for soil pollution assessment and ecological risk assessment in basalt areas. Attached Figure Description
[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0040] Figure 1 This is a schematic diagram of the process of the present invention.
[0041] Figure 2 This is a schematic diagram showing the geographical location of the basalt area to be tested and the distribution of soil geochemical survey data sampling points in an embodiment of the present invention.
[0042] Figure 3 This is a statistical diagram illustrating the frequency distribution of major elements and heavy metal elements in basalt soil after data preprocessing in an embodiment of the present invention. Detailed Implementation
[0043] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0044] Figure 1 A flowchart illustrating a method for fitting background heavy metal values in basaltic soils based on major element geochemical constraints, according to an embodiment of the present invention, is shown. Figure 1 As shown, the method includes:
[0045] S1. Data Collection: Obtain soil geochemical survey datasets for the basalt area to be tested;
[0046] S2. Data preprocessing: After removing invalid sample data from the soil geochemical survey dataset, a natural logarithmic transformation was performed.
[0047] S3. Correlation analysis of indicators: Correlation analysis and collinearity analysis were performed on the soil geochemical survey dataset after data preprocessing to obtain the correlation analysis results and collinearity analysis results;
[0048] S4. Screening of major elements: Based on the results of correlation analysis and collinearity analysis, and combined with the geochemical significance of major elements, a set of target major element indicators is screened.
[0049] S5. Regression Model Construction: Based on the target principal element index set, construct a multiple log-linear regression model between each heavy metal element and the target principal element index set to obtain the multiple log-linear regression model of each heavy metal element.
[0050] S6. Outlier identification and removal: The fitting results of the multiple log-linear regression models for each heavy metal element are evaluated to determine whether to rebuild the model and convert the background value of heavy metals in the soil.
[0051] S7. Conversion of Soil Heavy Metal Background Values: Substitute the concentration data of the target major element index set corresponding to each data sampling point in the basalt area to be tested into the multivariate log-linear regression model to calculate the corresponding heavy metal predicted values. The mean of all heavy metal predicted values of the same heavy metal element category is used as the natural background value of soil heavy metals in the basalt area to be tested.
[0052] Furthermore, the working principle of the present invention will be illustrated below through embodiments:
[0053] This embodiment uses the Chengmai basalt area in Hainan as the basalt area to be tested. This area is a typical Cenozoic basalt distribution area in my country. The parent material of the soil is mainly weathering products of basalt. The soil is rich in iron and magnesium and has significant natural enrichment of heavy metals. In addition, the study area is disturbed by agricultural planting and human activities along transportation routes. The superposition effect of natural high background of heavy metals and human input is obvious, which is suitable for verifying the effectiveness and practicality of the method of the present invention.
[0054] S1. Data Collection: A soil geochemical survey dataset was acquired from the basalt area of Chengmai, Hainan. This dataset consists of 4488 soil geochemical survey data points, each corresponding to one sampling point, and each sampling point corresponding to one soil sample. In this embodiment, the sampling point density is 10 points / km². The geographical location and distribution of the sampling points are detailed below. Figure 2 , Figure 2 To clearly demonstrate that the basalt area under test is located in the core distribution area of basalts in Chengmai County, Hainan Province, data sampling points uniformly cover the entire study area without significant spatial bias, ensuring the spatial representativeness and statistical validity of the soil geochemical survey data. All soil samples were tested, and the tested indicators included the concentrations of eight basic major elements and eight typical heavy metals, among which the major elements were... Heavy metal elements are The test results of all soil samples were integrated to form a soil geochemical survey dataset.
[0055] S2. Data Preprocessing: First, invalid sample data removal was performed, strictly eliminating soil geochemical survey data with missing element concentrations, missing detection data, and incomplete element categories. This ensured that the retained soil geochemical survey data covered the concentrations of the aforementioned 8 major elements and 8 heavy metals. No invalid sample data was removed from this soil geochemical survey dataset. Next, the major element concentration data and heavy metal concentration data of all soil geochemical survey data in the dataset were uniformly transformed using the natural logarithm ln transformation to obtain the preprocessed soil geochemical survey dataset. This operation was used to weaken the right-skewed distribution characteristic commonly found in geochemical data from basalt areas, improving the normality and modelability of the soil geochemical survey dataset. The frequency distribution statistics of each element category after transformation are shown in [the table below]. Figure 3 , Figure 3 The horizontal axis represents the concentration of each element after ln transformation, and the vertical axis represents the frequency of occurrence of the corresponding value. μ represents the mean of the element's ln-transformed value, and SD represents the standard deviation. Figure 3 It is evident that after the natural logarithm transformation, the numerical distribution of each element, which was originally right-skewed, is significantly optimized and becomes closer to a normal distribution. This effectively reduces the interference of extreme high values on subsequent modeling and lays a reliable data foundation for the correlation analysis of indicators and the construction of regression models.
[0056] S3. Correlation Analysis: First, a correlation analysis was conducted. Pearson linear correlation analysis was used to quantitatively analyze the preprocessed soil geochemical survey dataset. The analysis focused on major element indicators (including 8 major elements + Chemical Alteration Index (CIA)) and 8 heavy metals. The CIA was calculated from the concentration data of each major element according to the formula for constructing the CIA. The results of the correlation analysis of the preprocessed soil geochemical survey dataset are shown in Table 1 below. The correlation matrix between heavy metals and major element indicators after taking the natural logarithm shows that there is a general and significant correlation between heavy metals and many major elements. Among them, Cu, Ni, Cr, and Zn show significant correlations with... , MgO, MnO The correlation between As and major element indicators such as CIA is strong, and several representative major elements can be selected as candidate major element indicators; As correlates with MnO and Cd showed a relatively significant correlation with MgO, CaO, and The correlation between major elements is good; Hg and and There is a significant correlation between them, while Pb is mainly associated with The correlation is related to major elements such as MnO and MgO; the significance of the correlation coefficient was tested simultaneously, and the results showed that Hg is related to... The correlation between the two elements was not significant (p=0.433), while the correlations between the other elements were significant (p<0.01). Based on this, the major element indicators that reached the significant correlation threshold with heavy metal elements were included in the candidate independent variable range, and the candidate major element indicators were obtained, which are the correlation analysis results.
[0057] Table 1: Correlation coefficients of heavy metal elements and major element indices after taking the natural logarithm.
[0058] Subsequently, collinearity analysis was conducted, and the correlation coefficient matrix analysis method was used to detect collinearity of the candidate principal element indicators. The strong collinearity judgment threshold was set as |r|, and the correlation coefficient between any two candidate principal element indicators was calculated. The results are shown in Table 2 below, which shows the correlation coefficient of candidate principal element indicators.
[0059] Table 2: Correlation Coefficients of Candidate Principal Element Indicators
[0060] Note: ** indicates p < 0.01, * indicates p < 0.05;
[0061] From Table 2, we can see that , and The correlation coefficients among them are all greater than |r|=0.7, indicating strong multicollinearity. This suggests that the three indicators have significant collinearity characteristics. Therefore, during model construction, only one of these indicators needs to be retained as the independent variable to avoid the impact of multicollinearity on model stability. Furthermore... Strong multicollinearity was also observed between the macroelement and MnO. When selecting variables, it is necessary to make reasonable choices based on geochemical significance and model stability. The candidate macroelement indices with strong multicollinearity were recorded as the results of the collinearity analysis.
[0062] S4. Major Element Screening: Based on the results of correlation and collinearity analysis, and combined with the geochemical significance of the major element indices, the target major element indices were screened. Existing research indicates that in basaltic soils, heavy metals such as Cu, Cr, Ni, and Zn are mainly inherited from the parent material, and their spatial distribution characteristics are significantly controlled by the chemical composition of the parent rock. Furthermore, these elements typically exhibit strong correlations with each other. Basalt exhibits relatively stable chemical properties during weathering, effectively characterizing the composition of its parent material. Based on this, and combined with the aforementioned correlation and collinearity analysis results, elements with high correlation to the aforementioned heavy metals and clear geochemical significance were initially selected. MgO, MnO, and CIA were used as independent variables in a multiple log-linear regression model for fitting multiple log-linear regression models of Cu, Cr, Ni, and Zn. For As, considering both its correlation coefficients with the principal elements and the collinearity among the independent variables, it was selected as... MnO and MgO were used as independent variables in the multiple log-linear regression model; for Cd, variables with strong correlation and controllable collinearity were selected. MgO, CaO and As the independent variable in the multiple log-linear regression model, the same applies to Hg; MnO and MgO were selected as independent variables in the multiple log-linear regression model; for Pb, K2O, MnO, MgO and... The variables were used as independent variables in a multivariate log-linear regression model to fit the data; based on this, a set of target principal element indicators consisting of the independent variables of the multivariate log-linear regression model corresponding to each heavy metal element was obtained.
[0063] S5. Regression Model Construction: Based on the independent variables of the multiple log-linear regression model corresponding to each heavy metal element, the concentration index data corresponding to the target principal element index set and the preprocessed heavy metal element concentration data are imported into the statistical analysis software Excel to construct multiple log-linear regression models for each heavy metal element. In this embodiment, the multiple log-linear regression models for each heavy metal element are as follows:
[0064] ln(Cu) = −8.894 + 0.877ln( )+0.261ln(MgO)+0.063ln(MnO)+2.201ln(CIA), =0.840, p<0.001;
[0065] ln(Cr) = −12.287 + 0.560ln( )+0.291ln(MgO)+0.156ln(MnO)+3.364ln(CIA), =0.817, p<0.001;
[0066] ln(Ni) = −10.436 + 0.641ln( )+0.487ln(MgO)+0.118ln(MnO)+2.756ln(CIA), =0.833, p<0.001;
[0067] ln(Zn) = −7.918 + 0.802ln )+0.214ln(MgO)+0.172ln(MnO)+1.984ln(CIA), =0.809, p<0.001;
[0068] ln(As) = −1.781 + 0.750ln( )+0.479ln(MnO)−0.756ln(MgO), =0.432, p<0.001;
[0069] ln(Cd) = 4.504 + 0.304ln( )−0.040ln(MgO)+0.390ln(CaO)+0.128ln( ), =0.60, p<0.001;
[0070] ln(Hg) = 3.179 + 0.233ln( )+0.024ln(MnO)+0.006In(MgO), =0.240, p<0.001;
[0071] ln(Pb) = 1.143 + 0.483ln( )+0.277ln(MnO)−0.087ln(MgO)−0.172ln( ), =0.610, p<0.001.
[0072] S6. Outlier Identification and Removal: Based on the coefficient of determination and significance level of the multiple log-linear regression model, without removing outlier data sampling points, the coefficient of determination of the multiple log-linear regression model for the four heavy metal elements Cu, Cr, Ni, and Zn is determined. All values were greater than 0.8, and the significance level was p < 0.001, indicating that the natural background values of these four heavy metal elements could be well fitted by the major element index, and their spatial distribution was mainly controlled by the basaltic geological background and the geochemical constraints of major elements. In contrast, the determination coefficients of the multiple log-linear regression models for heavy metal elements such as As, Cd, Hg, and Pb were significantly lower. The values are all less than 0.8, indicating that the model has a large dispersion and relatively low model fitting accuracy. Model reconstruction is required to improve the stability and explanatory power of the multivariate log-linear regression model.
[0073] For As, the coefficients are determined based on one model reconstruction. The increase was less than 0.03. Based on this, soil geochemical survey data with significant deviations from the As concentration in the preprocessed soil geochemical survey dataset were gradually removed. After removing soil geochemical survey data with relative residual deviations greater than 50% (accounting for 18.0% of the total soil geochemical survey data), a new soil geochemical survey dataset was obtained. Based on this, the regression model was reconstructed, resulting in the new multiple log-linear regression model: ln(As) = −1.079 + 0.873ln( )+0.391ln(MnO)−0.811ln(MgO), = 0.461, p < 0.001;
[0074] For Cd, the coefficient of determination is based on three model reconstructions. The increase was less than 0.03. Based on this, soil geochemical survey data that deviated significantly from the multiple log-linear regression model in the new soil geochemical survey dataset corresponding to the third model reconstruction were gradually removed. After removing soil geochemical survey data with a relative residual deviation greater than 10% (accounting for 13.3% of the total soil geochemical survey data), a new soil geochemical survey dataset was obtained. Based on this, the regression model was reconstructed, resulting in the new multiple log-linear regression model: ln(Cd) = 4.8012 + 0.2507ln( )-0.0148ln(MgO)+0.4323ln(CaO)+0.1158ln( ), =0.625, p<0.001
[0075] For Hg, the coefficient of determination is not satisfied in any of the four model reconstructions. The increase was less than 0.03. Therefore, S3 and S4 were recalculated based on the new soil geochemical survey dataset. CIA was further introduced into the original multivariate log-linear regression model based on the independent variables. The coefficient of determination was then determined based on the two model reconstructions. The increase was less than 0.03. Based on this, soil geochemical survey data that deviated significantly from the multiple log-linear regression model in the new soil geochemical survey dataset corresponding to the first model reconstruction were gradually removed. After removing soil geochemical survey data with a relative residual deviation greater than 20% (accounting for 13.9% of the total soil geochemical survey data), a new soil geochemical survey dataset was obtained. Based on this, the regression model was reconstructed, resulting in the new multiple log-linear regression model: ln(Hg) = 2.384 + 0.300ln( )+0.075ln(MnO)-0.006ln(MgO)+0.053ln(CIA), =0.470, p<0.001;
[0076] For Pb, the coefficient of determination is not satisfied in any of the four model reconstructions. The increase was less than 0.03. Therefore, S3 and S4 were recalculated based on the new soil geochemical survey dataset. TFe2O3 was further introduced into the original multivariate log-linear regression model as an independent variable, and the coefficient was determined based on the two model reconstructions. The increase was less than 0.03. Based on this, soil geochemical survey data that deviated significantly from the multiple log-linear regression model in the new soil geochemical survey dataset corresponding to the first model reconstruction were gradually removed. After removing soil geochemical survey data with a relative residual deviation greater than 20% (accounting for 5.9% of the total soil geochemical survey data), a new soil geochemical survey dataset was obtained. Based on this, the regression model was reconstructed, and the new multiple log-linear regression model was obtained: ln(Pb) = 1.172 + 0.546ln(( )+0.243ln(MnO)-0.126ln(MgO)-0.196ln( )+0.065ln( ), =0.751, p<0.001; As can be seen above, with the removal of outlier data sampling points, the determination coefficients of the multiple log-linear regression models are significantly improved.
[0077] S7. Conversion of Soil Heavy Metal Background Values: The concentration data of the target principal element index set corresponding to each data sampling point in the basalt area of Chengmai, Hainan, are substituted into the corresponding multiple log-linear regression model to calculate the predicted heavy metal value of each heavy metal element. The mean of all predicted heavy metal values of the same heavy metal element category is used as the natural background value of soil heavy metals in the basalt area to be tested under that heavy metal element category. Finally, the natural background values of soil heavy metals of all heavy metals in the basalt area of Chengmai, Hainan are obtained.
[0078] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A basaltic area soil heavy metal background value fitting method based on major element geochemical constraints, characterized by, The method includes the following steps: S1. Data Collection: Obtain soil geochemical survey datasets for the basalt area to be tested; S2. Data preprocessing: After removing invalid sample data from the soil geochemical survey dataset, a natural logarithmic transformation was performed. S3. Correlation analysis of indicators: Correlation analysis and collinearity analysis were performed on the soil geochemical survey dataset after data preprocessing to obtain the correlation analysis results and collinearity analysis results; S4. Screening of major elements: Based on the results of correlation analysis and collinearity analysis, and combined with the geochemical significance of major elements, a set of target major element indicators is screened. S5. Regression Model Construction: Based on the target principal element index set, construct a multiple log-linear regression model between each heavy metal element and the target principal element index set to obtain the multiple log-linear regression model of each heavy metal element. S6. Outlier identification and removal: The fitting results of the multiple log-linear regression models for each heavy metal element are evaluated to determine whether to rebuild the model and convert the background value of heavy metals in the soil. S7. Conversion of Soil Heavy Metal Background Values: Substitute the concentration data of the target major element index set corresponding to each data sampling point in the basalt area to be tested into the multivariate log-linear regression model to calculate the corresponding heavy metal predicted values. The mean of all heavy metal predicted values of the same heavy metal element category is used as the natural background value of soil heavy metals in the basalt area to be tested.
2. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 1, characterized in that, The soil geochemical survey data set is composed of a plurality of soil geochemical survey data, and each soil geochemical survey data uniquely corresponds to a data sampling point; the soil geochemical survey data includes major element concentration data and heavy metal element concentration data, wherein the major element categories in the major element concentration data at least include and , and the heavy metal element categories in the heavy metal element concentration data at least include and .
3. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 2, characterized in that, The invalid sample data removal refers to removing soil geochemical survey data containing missing element concentrations or incomplete element categories from the soil geochemical survey dataset; the natural logarithmic transformation involves applying a natural logarithmic ln transformation to the major element concentration data and heavy metal element concentration data of all soil geochemical survey data retained after invalid sample data removal.
4. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 1, characterized in that, The correlation analysis adopts the Pearson linear correlation analysis method; the analysis objects of the correlation analysis are major element indicators and heavy metal elements, the major element indicators include major elements and derived indicators, wherein the derived indicators include at least one of the Parker weathering index, chemical alteration index and chemical weathering index.
5. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 4, characterized in that, The collinearity analysis employs either the correlation coefficient matrix analysis method or the variance inflation factor (VIF) analysis method, and the analysis object of the collinearity analysis is the candidate principal element index.
6. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 1, characterized in that, The multivariate logarithmic linear regression model is obtained by importing the concentration index data corresponding to the target major element index set and the data pre-processed heavy metal element concentration data into a statistical analysis software, and the statistical analysis software also synchronously outputs the determination coefficient corresponding to the multivariate logarithmic linear regression model and a significance level p. Preprocessed target principal element index set data and heavy metal element concentration data.
7. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 6, characterized in that, The fitting result is determined as follows: When the coefficient of determination If the value is greater than 0.8 and the significance level is less than 0.001, then the corresponding multiple log-linear regression model will be applied to the conversion of soil heavy metal background values. When the coefficient of determination If the significance level is ≤0.8 or p≥0.001, the corresponding multivariate log-linear regression model will be reconstructed.
8. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 7, characterized in that, The model reconstruction involves performing residual analysis on the multiple log-linear regression model using the residual relative deviation threshold method. This identifies and removes soil geochemical survey data corresponding to outlier sampling points in the preprocessed soil geochemical survey dataset, resulting in a new soil geochemical survey dataset. The regression model is then reconstructed based on this new dataset, along with the correlation and collinearity analysis results from the original S3 dataset, until the newly constructed multiple log-linear regression model corresponds to the coefficient of determination. The increase was less than 0.
03.
9. The basaltic region soil heavy metal background value fitting method based on major element geochemistry constraints according to claim 8, characterized in that, The threshold value range of the residual relative deviation threshold method is 5%-50%.