Method for identifying dominant factors of dissolved iron in nearshore sea area based on partial least square regression

Abstract

The application discloses a method for identifying dominant factors of dissolved iron in offshore waters based on partial least squares regression. The method comprises the following steps: obtaining sample data of the sea area to be identified, and grouping the sample data according to seasons and water layers based on the spatio-temporal heterogeneity of the offshore sea area; obtaining standardized sample data of different groups; determining the number of latent variables of different groups based on partial least squares regression and standardized sample data of different groups to determine partial least squares regression models of different groups; calculating the variable importance projection value of environmental factors, and combining the sign of the environmental factors in the final regression coefficient vector to identify the environmental factors that play a dominant role in the concentration of dissolved iron and their influence direction. The application effectively overcomes the multicollinearity problem between environmental factors, realizes the fine and situational analysis of the distribution mechanism of dissolved iron in offshore waters, and the identification result is stable, reliable and has strong interpretability, thereby providing a scientific basis for marine environment research and management.

Description

Technical Field

[0001] This invention relates to the field of marine environmental data analysis technology, specifically to a method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression. Background Technology

[0002] The nearshore marine environment is complex, and dissolved iron, as a key micronutrient, is influenced by a variety of environmental factors. These factors often exhibit high correlations, i.e., multicollinearity, which poses a significant challenge to accurately identifying the dominant factors in dissolved iron.

[0003] Currently, commonly used analytical methods include multiple linear regression and principal component analysis (PCA) and its regression methods. In cases of multicollinearity, the estimation of regression coefficients in multiple linear regression becomes extremely unstable, even exhibiting signs that contradict common physicochemical principles, leading to misjudgments of the dominant factor for dissolved iron. While PCA and its regression methods can eliminate multicollinearity through dimensionality reduction, their dimensionality reduction process only considers the variance information of the independent variables, failing to guarantee a high correlation between the extracted principal components and the target variable, dissolved iron. This results in decreased model prediction accuracy and poor identification of the dominant factor.

[0004] Therefore, there is an urgent need in this field for a method to identify the dominant factor of dissolved iron that can be accurately and stably identified under multicollinearity conditions and can adapt to the complex spatiotemporal changes in nearshore waters. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, this invention provides a method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0007] A method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression includes the following steps:

[0008] Sample data of the sea area to be identified were obtained. Each sample included measurements of dissolved iron concentration and environmental factors. The sample data were grouped by season and water layer based on the spatiotemporal heterogeneity of the nearshore sea area to obtain sample data of different groups.

[0009] The dissolved iron concentration and environmental factor measurements in the sample data of different groups were standardized to obtain standardized sample data for different groups.

[0010] Based on partial least squares regression and standardized sample data of different groups, the number of latent variables in different groups is determined, and the partial least squares regression model for different groups is determined based on the number of latent variables in different groups.

[0011] Based on partial least squares regression models for different groups, the variable importance projection values ​​of environmental factors are calculated. Combined with the signs of environmental factors in the final regression coefficient vector, the environmental factors that play a dominant role in dissolved iron concentration and their direction of influence are identified.

[0012] Furthermore, environmental factors include salinity, turbidity, dissolved organic carbon, colored dissolved organic matter, chlorophyll a, and suspended particulate matter.

[0013] Furthermore, the different groups include the winter surface layer, winter middle layer, winter bottom layer, summer surface layer, summer middle layer, and summer bottom layer.

[0014] Furthermore, based on partial least squares regression and standardized sample data from different groups, the number of latent variables for different groups is determined. The specific process is as follows: using the K-fold cross-validation method, the standardized sample data from different groups are divided into multiple mutually exclusive subsets; each subset is used as the validation set, and the remaining subsets are used as the training set. For each value in the range of latent variable numbers, a partial least squares regression model is established and the corresponding prediction error index is calculated; based on the trend of the prediction error index changing with the number of latent variables, the number of latent variables that minimizes the prediction error index or indicates its inflection point is selected to determine the number of latent variables for different groups.

[0015] Furthermore, a partial least squares regression model is established. The specific process is as follows: determine whether the number of environmental factors is greater than the sample size. If so, the first method is used to establish the partial least squares regression model; otherwise, the second method is used to establish the partial least squares regression model.

[0016] Furthermore, the first sampling method is used to establish a partial least squares regression model, including the following steps:

[0017] A1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data from different groups;

[0018] A2. Select any non-zero column or its first principal component vector from the dissolved iron concentration matrix and determine it as the initial value of the score vector of the first latent variable;

[0019] A3. Calculate the initial loading vector of the environmental factor corresponding to the first latent variable based on the initial value of the score vector of the first latent variable, expressed as:

[0020]

[0021] in: Let be the initial loading vector of the environmental factors corresponding to the first latent variable. This is an environmental factor matrix. This is the matrix transpose operator. The initial values ​​for the score vector of the first latent variable;

[0022] A4. Normalize the initial loading vector of the environmental factor corresponding to the first latent variable to obtain the loading vector of the environmental factor corresponding to the first latent variable.

[0023] A5. Based on the loading vector of the environmental factor corresponding to the first latent variable, iterate the initial value of the score vector of the first latent variable until the iteration condition is met to obtain the score vector of the first latent variable, expressed as:

[0024]

[0025] in: Let be the score vector of the first latent variable. This is the loading vector of the environmental factor corresponding to the first latent variable;

[0026] A6. Based on the score vector of the first latent variable, calculate the load vector of the dissolved iron concentration corresponding to the first latent variable, expressed as:

[0027]

[0028] in: Let be the load vector corresponding to the dissolved iron concentration of the first latent variable. This is the dissolved iron concentration matrix;

[0029] A7. Based on the score vector of the first latent variable, calculate the weight vector corresponding to the first latent variable, expressed as:

[0030]

[0031] in: This is the weight vector corresponding to the first latent variable;

[0032] A8. Based on the score vector of the first latent variable, calculate the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, expressed as:

[0033]

[0034] in: This is the regression coefficient vector for the dissolved iron concentration corresponding to the first latent variable;

[0035] A9. Based on the score vector of the first latent variable and the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, calculate the residual matrix of dissolved iron concentration and the residual matrix of environmental factors, as follows:

[0036] ,

[0037]

[0038] in: The residual matrix of environmental factors, This is the residual matrix of the dissolved iron concentration;

[0039] A10. Determine the residual matrix of dissolved iron concentration as the new dissolved iron concentration matrix, and determine the residual matrix of environmental factors as the new environmental factor matrix, in order to calculate the next latent variable;

[0040] A11. Repeat steps A1-A10 to calculate all latent variables in sequence to obtain the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration.

[0041] A12. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as:

[0042]

[0043] in: This is the final regression coefficient vector. This is the weight matrix. The loading matrix of environmental factors, The load matrix represents the concentration of dissolved iron.

[0044] A13. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as:

[0045]

[0046] in: This is the predicted value for dissolved iron concentration.

[0047] Furthermore, a partial least squares regression model is established using the second method, the specific process of which is as follows:

[0048] B1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data of different groups, and calculate the covariance matrix of dissolved iron concentration matrix and environmental factor matrix respectively;

[0049] B2. Perform eigenvalue decomposition on the covariance matrix of the dissolved iron concentration matrix and the environmental factor matrix to obtain the eigenvalues ​​after decomposition. Select the eigenvectors corresponding to the h largest eigenvalues, where h is the number of latent variables, and construct a weight matrix by selecting the eigenvectors in order of size.

[0050] B3. Calculate the score matrix of the latent variables based on the weight matrix, as follows:

[0051]

[0052] in: The score matrix represents the latent variables. This is an environmental factor matrix. This is the weight matrix;

[0053] B4. Based on the score matrix of latent variables, calculate the loading matrix of environmental factors and the loading matrix of dissolved iron concentration, as follows:

[0054] ,

[0055]

[0056] in: The loading matrix of environmental factors, The load matrix represents the concentration of dissolved iron. This is the dissolved iron concentration matrix;

[0057] B5. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as:

[0058]

[0059] in: This is the final regression coefficient vector. This is the matrix transpose operator;

[0060] B6. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as:

[0061]

[0062] in: This is the predicted value for dissolved iron concentration.

[0063] Furthermore, the variable importance projection values ​​of environmental factors are calculated and expressed as follows:

[0064]

[0065] in: For the first The projected importance values ​​of each environmental factor. This represents the serial number of the environmental factor. The number of environmental factors, The number of latent variables. For the sequence number of the latent variable, For the first The variance of dissolved iron concentration explained by each latent variable. The weight matrix is ​​the first... The latent variable corresponds to the first... The weights of each environmental factor.

[0066] The beneficial effects of this invention are as follows:

[0067] (1) This invention determines the number of latent variables in different groups by using partial least squares regression and standardized sample data of different groups, and determines the partial least squares regression model of different groups based on the number of latent variables in different groups. Then, based on the partial least squares regression model of different groups, the variable importance projection value of environmental factors is calculated, which fundamentally overcomes the model instability problem caused by multicollinearity and makes the identification results more reliable.

[0068] (2) This invention groups the sample data by season and water layer according to the spatiotemporal heterogeneity of nearshore sea areas to obtain sample data of different groups. Then, it adopts an independent modeling strategy by season and water layer to fully capture the spatiotemporal heterogeneity of nearshore sea environment, making the analysis conclusions more refined and more consistent with the actual physical and chemical processes.

[0069] (3) This invention identifies the environmental factors that play a dominant role in the concentration of dissolved iron and their direction of influence by calculating the variable importance projection value of environmental factors and combining the sign of environmental factors in the final regression coefficient vector. This process adopts a combination of dual criteria and gives the importance ranking and direction of influence of environmental factors. The conclusion is intuitive, easy to understand and apply. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the process for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression. Detailed Implementation

[0071] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0072] like Figure 1 As shown, the method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression includes steps S1-S4, as detailed below:

[0073] S1. Obtain sample data of the sea area to be identified. Each sample includes the dissolved iron concentration measurement and environmental factor measurement. Based on the spatiotemporal heterogeneity of the nearshore sea area, the sample data are grouped by season and water layer to obtain sample data of different groups.

[0074] In an optional embodiment of the present invention, environmental factors include salinity, turbidity, dissolved organic carbon, colored dissolved organic matter, chlorophyll a, and suspended particulate matter.

[0075] The different groups include the winter surface layer, winter middle layer, winter bottom layer, summer surface layer, summer middle layer, and summer bottom layer.

[0076] S2. Standardize the dissolved iron concentration and environmental factor measurements in the sample data of different groups to obtain standardized sample data for different groups.

[0077] In an optional embodiment of the present invention, the present invention uses the Z-score standardization method to standardize the dissolved iron concentration measurement value and environmental factor measurement value in the sample data of different groups to obtain standardized sample data of different groups.

[0078] S3. Based on partial least squares regression and standardized sample data of different groups, determine the number of latent variables for different groups, and determine the partial least squares regression model for different groups based on the number of latent variables for different groups.

[0079] In an optional embodiment of the present invention, the present invention determines the number of latent variables for different groups based on partial least squares regression and standardized sample data of different groups. The specific process is as follows: using the K-fold cross-validation method, the standardized sample data of different groups are divided into multiple mutually exclusive subsets; each subset is used as the validation set and the remaining subsets are used as the training set; for each value in the range of the number of latent variables, a partial least squares regression model is established and the corresponding prediction error index is calculated; based on the trend of the prediction error index changing with the number of latent variables, the number of latent variables that minimizes the prediction error index or indicates its inflection point is selected to determine the number of latent variables for different groups.

[0080] The present invention establishes a partial least squares regression model. The specific process is as follows: determine whether the number of environmental factors is greater than the number of samples. If so, the first method is used to establish a partial least squares regression model; otherwise, the second method is used to establish a partial least squares regression model.

[0081] The first sampling method of this invention establishes a partial least squares regression model, including the following steps:

[0082] A1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data from different groups.

[0083] A2. Select any non-zero column or its first principal component vector from the dissolved iron concentration matrix, and determine it as the initial value of the score vector of the first latent variable.

[0084] A3. Calculate the initial loading vector of the environmental factor corresponding to the first latent variable based on the initial value of the score vector of the first latent variable, expressed as:

[0085]

[0086] in: Let be the initial loading vector of the environmental factors corresponding to the first latent variable. This is an environmental factor matrix. This is the matrix transpose operator. The initial values ​​are the score vectors of the first latent variable.

[0087] A4. Normalize the initial loading vector of the environmental factor corresponding to the first latent variable to obtain the loading vector of the environmental factor corresponding to the first latent variable.

[0088] A5. Based on the loading vector of the environmental factor corresponding to the first latent variable, iterate the initial value of the score vector of the first latent variable until the iteration condition is met to obtain the score vector of the first latent variable, expressed as:

[0089]

[0090] in: Let be the score vector of the first latent variable. This is the loading vector of the environmental factor corresponding to the first latent variable.

[0091] Specifically, a threshold for the iterative difference of the score vector is set according to the actual application scenario, and the iteration condition is set to the difference between the score vector of the first latent variable after iteration and the score vector of the first latent variable before iteration being less than the threshold for the iterative difference of the score vector.

[0092] A6. Based on the score vector of the first latent variable, calculate the load vector of the dissolved iron concentration corresponding to the first latent variable, expressed as:

[0093]

[0094] in: Let be the load vector corresponding to the dissolved iron concentration of the first latent variable. This is the dissolved iron concentration matrix.

[0095] A7. Based on the score vector of the first latent variable, calculate the weight vector corresponding to the first latent variable, expressed as:

[0096]

[0097] in: This is the weight vector corresponding to the first latent variable.

[0098] A8. Based on the score vector of the first latent variable, calculate the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, expressed as:

[0099]

[0100] in: This is the regression coefficient vector for the dissolved iron concentration corresponding to the first latent variable.

[0101] A9. Based on the score vector of the first latent variable and the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, calculate the residual matrix of dissolved iron concentration and the residual matrix of environmental factors, as follows:

[0102] ,

[0103]

[0104] in: The residual matrix of environmental factors, This is the residual matrix of the dissolved iron concentration.

[0105] A10. Determine the residual matrix of dissolved iron concentration as the new dissolved iron concentration matrix, and determine the residual matrix of environmental factors as the new environmental factor matrix, in order to calculate the next latent variable.

[0106] A11. Repeat steps A1-A10 to calculate all latent variables in sequence to obtain the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration.

[0107] A12. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as:

[0108]

[0109] in: This is the final regression coefficient vector. This is the weight matrix. The loading matrix of environmental factors, This is the load matrix for the dissolved iron concentration.

[0110] A13. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as:

[0111]

[0112] in: This is the predicted value for dissolved iron concentration.

[0113] This invention employs a second method to establish a partial least squares regression model, the specific process of which is as follows:

[0114] B1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data of different groups, and calculate the covariance matrix of dissolved iron concentration matrix and environmental factor matrix.

[0115] B2. Perform eigenvalue decomposition on the covariance matrices of the dissolved iron concentration matrix and the environmental factor matrix to obtain the decomposed eigenvalues. Select the eigenvectors corresponding to the h largest eigenvalues, where h is the number of latent variables, and construct a weight matrix from the selected eigenvectors in order of size.

[0116] B3. Calculate the score matrix of the latent variables based on the weight matrix, as follows:

[0117]

[0118] in: The score matrix represents the latent variables. This is an environmental factor matrix. This is the weight matrix.

[0119] B4. Based on the score matrix of latent variables, calculate the loading matrix of environmental factors and the loading matrix of dissolved iron concentration, as follows:

[0120] ,

[0121]

[0122] in: The loading matrix of environmental factors, The load matrix represents the concentration of dissolved iron. This is the dissolved iron concentration matrix.

[0123] B5. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as:

[0124]

[0125] in: This is the final regression coefficient vector. This is the matrix transpose operator.

[0126] B6. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as:

[0127]

[0128] in: This is the predicted value for dissolved iron concentration.

[0129] S4. Based on the partial least squares regression models of different groups, calculate the variable importance projection values ​​of environmental factors, and combine the signs of environmental factors in the final regression coefficient vector to identify the environmental factors that play a dominant role in dissolved iron concentration and their direction of influence.

[0130] In an optional embodiment of the present invention, the variable importance projection value of environmental factors is calculated and expressed as:

[0131]

[0132] in: For the first The projected importance values ​​of each environmental factor. This represents the serial number of the environmental factor. The number of environmental factors, The number of latent variables. For the sequence number of the latent variable, For the first The variance of dissolved iron concentration explained by each latent variable. , To calculate the sum of squares, For the first The score vector of each latent variable. For the first The loading vector corresponding to the dissolved iron concentration of each latent variable. This characterizes the independent explanatory power of the latent variable for changes in dissolved iron concentration. The weight matrix is ​​the first... The latent variable corresponds to the first... The weights of each environmental factor.

[0133] Specifically, the variable importance projection value of an environmental factor quantitatively characterizes the overall contribution of that environmental factor in explaining the changes in dissolved iron concentration; the larger the value, the more important the environmental factor.

[0134] This invention obtains the final regression coefficient vector from different groups of partial least squares regression models. Each element in the final regression coefficient vector corresponds to an environmental factor. The sign of each element in the final regression coefficient vector directly indicates the direction of the influence between the environmental factor and the dissolved iron concentration. A positive sign indicates a positive correlation, meaning that as the value of the environmental factor increases, the dissolved iron concentration predicted by the partial least squares regression model tends to increase. A negative sign indicates a negative correlation, meaning that as the value of the environmental factor increases, the dissolved iron concentration predicted by the partial least squares regression model tends to decrease. The absolute value of each element in the final regression coefficient vector reflects the strength of the environmental factor's influence on the dissolved iron concentration after excluding the influence of other environmental factors.

[0135] This invention, based on the calculation results of the variable importance projection values ​​of environmental factors, comprehensively judges each environmental factor to identify the dominant factor and interpret its role. The specific process is as follows: A threshold of 1.0 is set for the variable importance projection value; among all environmental factors, those with a variable importance projection value greater than 1.0 are initially identified as important factors; for each selected important factor, the sign and magnitude of its corresponding element in the final regression coefficient vector are examined, and its sign (positive or negative) is used to determine whether it is a promoting or inhibiting factor. The absolute value of the element is used to rank the influence among environmental factors of similar importance, and the final dominant factor is determined based on the influence ranking results of the important factors.

[0136] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression, characterized in that, Includes the following steps: Sample data were acquired from the sea area to be identified. Each sample included measurements of dissolved iron concentration and environmental factors. Based on the spatiotemporal heterogeneity of the nearshore sea area, the sample data were grouped by season and water layer to obtain sample data for different groups. Environmental factors included salinity, turbidity, dissolved organic carbon, colored dissolved organic matter, chlorophyll a, and suspended particulate matter. The different groups included winter surface layer, winter middle layer, winter bottom layer, summer surface layer, summer middle layer, and summer bottom layer. The dissolved iron concentration and environmental factor measurements in the sample data of different groups were standardized to obtain standardized sample data for different groups. Based on partial least squares regression and standardized sample data from different groups, the number of latent variables in each group is determined, and a partial least squares regression model for each group is determined accordingly. The specific process for determining the number of latent variables in each group is as follows: Using K-fold cross-validation, the standardized sample data from different groups is divided into multiple mutually exclusive subsets. Each subset is used as the validation set, and the remaining subsets are used as the training set. For each value within the range of latent variable numbers, a partial least squares regression model is established, and the corresponding prediction error index is calculated. Based on the trend of the prediction error index changing with the number of latent variables, the number of latent variables that minimizes the prediction error index or indicates its inflection point is selected to determine the number of latent variables in each group. Based on partial least squares regression models for different groups, the variable importance projection values ​​of environmental factors are calculated. Combined with the signs of environmental factors in the final regression coefficient vector, the environmental factors that play a dominant role in dissolved iron concentration and their direction of influence are identified.

2. The method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression according to claim 1, characterized in that, The specific process for establishing a partial least squares regression model is as follows: determine whether the number of environmental factors is greater than the sample size. If so, use the first method to establish a partial least squares regression model; otherwise, use the second method to establish a partial least squares regression model.

3. The method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression according to claim 2, characterized in that, The first sampling method establishes a partial least squares regression model, including the following steps: A1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data from different groups; A2. Select any non-zero column or its first principal component vector from the dissolved iron concentration matrix and determine it as the initial value of the score vector of the first latent variable; A3. Calculate the initial loading vector of the environmental factor corresponding to the first latent variable based on the initial value of the score vector of the first latent variable, expressed as: in: Let be the initial loading vector of the environmental factors corresponding to the first latent variable. This is an environmental factor matrix. This is the matrix transpose operator. The initial values ​​for the score vector of the first latent variable; A4. Normalize the initial loading vector of the environmental factor corresponding to the first latent variable to obtain the loading vector of the environmental factor corresponding to the first latent variable. A5. Based on the loading vector of the environmental factor corresponding to the first latent variable, iterate the initial value of the score vector of the first latent variable until the iteration condition is met to obtain the score vector of the first latent variable, expressed as: in: Let be the score vector of the first latent variable. This is the loading vector of the environmental factor corresponding to the first latent variable; A6. Based on the score vector of the first latent variable, calculate the load vector of the dissolved iron concentration corresponding to the first latent variable, expressed as: in: Let be the load vector corresponding to the dissolved iron concentration of the first latent variable. This is a matrix of dissolved iron concentrations; A7. Based on the score vector of the first latent variable, calculate the weight vector corresponding to the first latent variable, expressed as: in: This is the weight vector corresponding to the first latent variable; A8. Based on the score vector of the first latent variable, calculate the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, expressed as: in: This is the regression coefficient vector for the dissolved iron concentration corresponding to the first latent variable; A9. Based on the score vector of the first latent variable and the regression coefficient vector of the dissolved iron concentration corresponding to the first latent variable, calculate the residual matrix of dissolved iron concentration and the residual matrix of environmental factors, as follows: , in: The residual matrix of environmental factors, This is the residual matrix of the dissolved iron concentration; A10. Determine the residual matrix of dissolved iron concentration as the new dissolved iron concentration matrix, and determine the residual matrix of environmental factors as the new environmental factor matrix, in order to calculate the next latent variable; A11. Repeat steps A1-A10 to calculate all latent variables in sequence to obtain the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration. A12. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as: in: This is the final regression coefficient vector. This is the weight matrix. The loading matrix of environmental factors, The load matrix represents the concentration of dissolved iron. A13. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as: in: This is the predicted value for dissolved iron concentration.

4. The method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression according to claim 2, characterized in that, The second method is used to establish a partial least squares regression model. The specific process is as follows: B1. Construct dissolved iron concentration matrix and environmental factor matrix based on standardized sample data of different groups, and calculate the covariance matrix of dissolved iron concentration matrix and environmental factor matrix respectively; B2. Perform eigenvalue decomposition on the covariance matrix of the dissolved iron concentration matrix and the environmental factor matrix to obtain the eigenvalues ​​after decomposition. Select the eigenvectors corresponding to the h largest eigenvalues, where h is the number of latent variables, and construct a weight matrix by selecting the eigenvectors in order of size. B3. Calculate the score matrix of the latent variables based on the weight matrix, as follows: in: The score matrix represents the latent variables. This is an environmental factor matrix. This is the weight matrix; B4. Based on the score matrix of latent variables, calculate the loading matrix of environmental factors and the loading matrix of dissolved iron concentration, as follows: ， in: The loading matrix of environmental factors, The load matrix represents the concentration of dissolved iron. This is a matrix of dissolved iron concentrations; B5. Based on the weight matrix, the loading matrix of environmental factors, and the loading matrix of dissolved iron concentration, calculate the final regression coefficient vector, expressed as: in: This is the final regression coefficient vector. This is the matrix transpose operator; B6. Based on the final regression coefficient vector, a partial least squares regression model is established, expressed as: in: This is the predicted value for dissolved iron concentration.

5. The method for identifying the dominant factor of dissolved iron in nearshore waters based on partial least squares regression according to claim 1, characterized in that, The variable importance projection values ​​of environmental factors are calculated and expressed as follows: in: For the first The projected importance values ​​of each environmental factor. This represents the serial number of the environmental factor. The number of environmental factors, The number of latent variables. For the sequence number of the latent variable, For the first The variance of dissolved iron concentration explained by each latent variable. The weight matrix is ​​the first... The latent variable corresponds to the first... The weights of each environmental factor.

Images