A method and system for constructing a lake eutrophication index coupling model

Abstract

The present invention discloses a method and system for constructing a lake eutrophication index coupling model, and belongs to the technical field of environmental monitoring. The method comprises the following steps: step S1, obtaining effective band data, and calculating the cyanobacteria index and the band ratio; step S2, determining the observation index from the effective band data, and determining the reference wavelength; step S3, calculating the total absorption coefficient and the backscattering coefficient of the reference wavelength; calculating the total absorption coefficient and the backscattering coefficient corresponding to N target wavelengths; step S4, using the effective band data, the cyanobacteria index, the band ratio, and the total absorption coefficient and the backscattering coefficient corresponding to the N target wavelengths as input parameters to construct a lake eutrophication index coupling model. The lake eutrophication index coupling model provided by the present invention has strong adaptability and noise resistance, helps to realize efficient and accurate lake eutrophication monitoring, and provides support for environmental management and water quality early warning.

Description

Technical Field

[0001] The invention belongs to the technical field of environmental monitoring, and in particular relates to a method and system for constructing a lake eutrophication index coupling model. Background Art

[0002] Lake eutrophication refers to the phenomenon that the excess nutrients (such as nitrogen and phosphorus) in the water body leads to abnormal reproduction of algae and aquatic plants, thus causing water quality deterioration and ecological imbalance. With the increase of agricultural activities, industrial emissions and urban expansion, the problem of lake eutrophication is becoming increasingly serious around the world. This phenomenon not only affects the stability of the water ecosystem, but also has adverse effects on drinking water safety, fishery resources and tourism.

[0003] Empirical models based on remote sensing data are an early method for estimating lake eutrophication, which usually rely on the reflectance of specific bands or band combinations to estimate indicators such as chlorophyll, suspended matter and dissolved organic matter in water bodies. This type of model is simple to calculate and is suitable for preliminary analysis of large-scale water bodies. It has been widely used in the study of water quality parameter estimation using medium-resolution remote sensing data (such as MODIS and Landsat).

[0004] Due to differences in lake types, nutrient sources, and climatic conditions, the spectral characteristics of water bodies in different regions vary significantly, which limits the applicability of the model. Many empirical models and semi-analytical models rely on measured data from specific lakes or regions and are difficult to generalize to other regions, affecting the versatility of the model. In addition, the complexity of the water environment makes it difficult to extract spectral features. The presence of suspended matter, dissolved organic matter, and algae in the water body can easily cause confusion in spectral signals, especially in highly eutrophic water bodies. This mixing effect reduces the accuracy of the model, especially when multiple parameters are inverted simultaneously. Summary of the invention

[0005] Purpose of the invention: In order to solve the above problems, the present invention provides a method and system for constructing a lake eutrophication index coupling model.

[0006] Technical solution: A method for constructing a lake eutrophication index coupling model, comprising the following steps:

[0007] Step S1, obtaining original band data of different optical bands, preprocessing the original band data to obtain effective band data; the effective band data at least includes: remote sensing reflectance and central wavelength; based on the remote sensing reflectance and central wavelength corresponding to different optical bands, calculating the cyanobacteria index and the band ratio;

[0008] Step S2, determining an observation index from the effective band data; judging the observation index to obtain a judgment result, and determining a reference wavelength based on the judgment result;

[0009] Step S3, calculating the total absorption coefficient and backscattering coefficient of the reference wavelength for the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculating the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths;

[0010] Step S4, using the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, constructing a hyperparameter grid through grid search, and constructing a lake eutrophication index coupling model.

[0011] In a further embodiment, the cyanobacteria index in step S1 is calculated using the following formula:

[0012] ;

[0013] In the formula, is the cyanobacteria index; , , They are the remote sensing reflectances in the red, blue and near-infrared bands respectively; , , , They are the center wavelengths of the green, blue, red and near-infrared bands respectively.

[0014] In a further embodiment, the step S2 further comprises the following steps:

[0015] Select the remote sensing reflectance of the red band from the valid band data As an observation indicator; With threshold Make a comparison and get the result:

[0016] like , then the reference wavelength The value of ;

[0017] like , then the reference wavelength The value of .

[0018] In a further embodiment, , the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula:

[0019] ;

[0020] ;

[0021] ;

[0022] In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; is the intermediate process variable; r rs (469), r rs (555) are the remote sensing reflectances at wavelengths of 469 nm and 555 nm respectively; The wavelength The remote sensing reflectance of the location, is the wavelength Remote sensing reflectivity of the location.

[0023] In a further embodiment, , then the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula:

[0024] ;

[0025] In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; It is the remote sensing reflectance at a wavelength of 645nm.

[0026] In a further embodiment, the backscattering coefficient of the reference wavelength in step S3 is calculated using the following formula:

[0027] ;

[0028] ;

[0029] In the formula, , The reference wavelengths are Backscattering coefficient of position, backscattering coefficient of pure water; is an intermediate process variable; The reference wavelength The total absorption coefficient of the location; The reference wavelength Remote sensing reflectance of the location; reference wavelength The value is or .

[0030] In a further embodiment, the backscattering coefficients corresponding to the N target wavelengths in step S3 are calculated using the following formula:

[0031] ;

[0032] In the formula, is the target wavelength The backscatter coefficient at position, .

[0033] In a further embodiment, the total absorption coefficient corresponding to the N target wavelengths in step S3 is calculated using the following formula:

[0034] ;

[0035] ;

[0036] In the formula, is the target wavelength The total absorption coefficient at the position, ; is the target wavelength Pure water backscattering coefficient at position; is an intermediate process variable; The target wavelength Remote sensing reflectance of the location.

[0037] In a further embodiment, the band ratio includes: , , , , , ; is the remote sensing reflectance of the green band.

[0038] In another technical solution, a system for constructing a lake eutrophication index coupling model is provided, which is used to implement the above-mentioned method for constructing a lake eutrophication index coupling model, including:

[0039] The first module is configured to obtain original band data of different optical bands, pre-process the original band data, and obtain effective band data; the effective band data at least includes: remote sensing reflectance and central wavelength; based on the remote sensing reflectance and central wavelength corresponding to different optical bands, the cyanobacteria index and the band ratio are calculated;

[0040] The second module is configured to determine an observation index from the effective band data; judge the observation index to obtain a judgment result, and determine a reference wavelength based on the judgment result;

[0041] The third module is configured to calculate the total absorption coefficient and backscattering coefficient of the reference wavelength for the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculate the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths;

[0042] The fourth module is configured to use the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, construct a hyperparameter grid through grid search, and construct a lake eutrophication index coupling model.

[0043] Beneficial effects:

[0044] (1) The data set used in the lake eutrophication index coupling model in the present invention covers oligotrophic, mesotrophic and eutrophic water states, and the data set comes from 3528 sampling points around the world. The significant optical variability and a large number of data samples provide favorable conditions for the development of a large-scale water eutrophication index (TSI) inversion algorithm.

[0045] (2) The present invention couples machine learning and semi-analytical algorithms to effectively combine the nonlinear fitting capabilities of machine learning with the physical constraints of semi-analytical models, thereby improving prediction accuracy and interpretability and reducing dependence on large-scale training data.

[0046] (3) In a complex water environment, the lake eutrophication index coupling model provided by the present invention has strong adaptability and noise resistance, which helps to achieve efficient and accurate lake eutrophication monitoring and provide support for environmental management and water quality early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a construction diagram and a verification diagram of the semi-analytical model of lake eutrophication index in Example 1 of the present invention;

[0048] Figure 2 It is a construction diagram and a verification diagram of the lake eutrophication index gradient tree boosting model in Example 1 of the present invention;

[0049] Figure 3 It is a fitting diagram of different input variables and measured TSI in Example 1 of the present invention;

[0050] Figure 4 It is a construction diagram and a verification diagram of the lake eutrophication index coupling model in Example 1 of the present invention;

[0051] Figure 5 This is a comparative analysis diagram of various models in Example 1 of the present invention. DETAILED DESCRIPTION

[0052] Example 1

[0053] like Figures 1 to 5 As shown, this embodiment provides a method for constructing a lake eutrophication index coupling model, including steps S1, S2, S3 and S4.

[0054] Step S1, obtaining original band data of different optical bands, preprocessing the original band data to obtain effective band data; the effective band data at least includes: remote sensing reflectance, center wavelength; based on the remote sensing reflectance and center wavelength corresponding to different optical bands, calculating the cyanobacteria index and band ratio.

[0055] The above raw band data includes the MODIS remote sensing reflectance obtained by Aqua apparent reflectance MYD09GA-6.1 version. Each reflectance image has undergone strict quality control, including removing errors and cloud pollution pixels through quality control mark (state_1km), and removing non-water pixels through water extraction index (NDWI>0). The remote sensing reflectance can be obtained by dividing the apparent reflectance by π.

[0056] The above-mentioned cyanobacteria index is calculated using the following formula:

[0057] ;

[0058] In the formula, is the cyanobacteria index; , , They are the remote sensing reflectances in the red, blue and near-infrared bands respectively; , , , They are the center wavelengths of the green, blue, red and near-infrared bands respectively.

[0059] The above band ratios include: , , , , , ; is the remote sensing reflectance of the green band.

[0060] Step S2, determining an observation index from the effective wavelength band data; judging the observation index to obtain a judgment result, and determining a reference wavelength based on the judgment result.

[0061] The step S2 further comprises the following steps:

[0062] Select the remote sensing reflectance of the red band from the valid band data As an observation indicator; With threshold Make a comparison and get the result:

[0063] like , then the reference wavelength The value of ;

[0064] like , then the reference wavelength The value of .

[0065] The following example illustrates the above judgment. For example, for clear water, , then the reference wavelength ; For highly scattering water, , then the reference wavelength .

[0066] Step S3, calculating the total absorption coefficient and backscattering coefficient of the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculating the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths.

[0067] The following is based on the reference wavelength The calculation of the total absorption coefficient of the reference wavelength is explained separately for different values ​​of :

[0068] 1. , the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula:

[0069] ;

[0070] ;

[0071] ;

[0072] In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; is the intermediate process variable; r rs (469), r rs (555) are the remote sensing reflectances at wavelengths of 469 nm and 555 nm respectively; The wavelength The remote sensing reflectance of the location, is the wavelength Remote sensing reflectivity of the location.

[0073] (ii) , then the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula:

[0074] ;

[0075] In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; It is the remote sensing reflectance at a wavelength of 645nm.

[0076] Based on the total absorption coefficient of the reference wavelength, the backscattering coefficient of the reference wavelength is calculated:

[0077] ;

[0078] ;

[0079] In the formula, , The reference wavelengths are Backscattering coefficient of position, backscattering coefficient of pure water; is an intermediate process variable; The reference wavelength The total absorption coefficient of the location; The reference wavelength Remote sensing reflectance of the location; reference wavelength The value is or .

[0080] In this embodiment, three target wavelengths are selected for illustration, and the three target wavelengths are: wavelength 469nm, wavelength 555nm, and wavelength 645nm; that is, the total absorption coefficient at these three wavelength positions needs to be calculated. , , , backscatter coefficient , , .

[0081] The backscattering coefficient corresponding to the target wavelength is calculated using the following formula:

[0082] ;

[0083] In the formula, is the target wavelength The backscatter coefficient at position, . The possible values ​​are 469, 555, and 645.

[0084] The total absorption coefficient corresponding to the target wavelength is calculated using the following formula:

[0085] ;

[0086] ;

[0087] In the formula, is the target wavelength The total absorption coefficient at the position, ; is the target wavelength Pure water backscattering coefficient at position; is an intermediate process variable; The target wavelength Remote sensing reflectance of the location. The possible values ​​are 469, 555, and 645.

[0088] Step S4, using the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, constructing a hyperparameter grid through grid search, and constructing a lake eutrophication index coupling model.

[0089] Valid band data includes single band , , , .

[0090] Band ratios include: , , , , , .

[0091] The lake eutrophication index coupling model uses the above data as the input parameters of the gradient tree boosting algorithm, constructs a hyperparameter (such as learning rate, number of trees, maximum depth, etc.) grid through grid search, and evaluates model performance through cross-validation to obtain the optimal coupling model.

[0092] like Figure 3 As shown in the figure, by analyzing the correlation between the single band, band ratio, cyanobacteria index, total absorption coefficient and backscattering coefficient corresponding to the target wavelength and the eutrophication index (TSI), the results show that the above parameters are significantly correlated with the eutrophication index (TSI) (P<0.001) and can be used as an effective input for the lake eutrophication index coupling model. Among them, Figure 3The variables in (a) to (n) are red light, near infrared, blue light, green light, red light to blue light ratio, red light to green light ratio, blue light to green light ratio, cyanobacteria index, 645 total absorption coefficient, 469 total absorption coefficient, 555 total absorption coefficient, 645 backscattering coefficient, 469 backscattering coefficient, 555 backscattering coefficient.

[0093] The optimal parameters and model are obtained by constructing a grid of hyperparameters (such as learning rate, number of trees, maximum depth, etc.) through grid search and evaluating model performance through cross-validation. Figure 4 As shown in the figure, the validation set R of the lake eutrophication index coupling model 2 is 0.78, the average relative error is 13.12%, and the deviation is 0.24; Figure 4 (a) and (b) are the construction and verification diagrams of the lake eutrophication index coupling model.

[0094] In addition, in order to verify the significant advantages of the lake eutrophication index coupling model, a lake eutrophication index semi-analytical model and a lake eutrophication index gradient tree boosting model were also established.

[0095] The construction of the semi-analytical model of lake eutrophication index is as follows: (1) When matching MODIS satellite data, the sampling data on the day of satellite transit is selected as the matching sample point, and the mean value within the 3 × 3 pixel window at the center of each sampling position is used as the matching value. When the coefficient of variation within the window is greater than 10%, the matching sample point is eliminated as unqualified. According to the above rules, 3528 sample point pairs were finally obtained. The eutrophication index (TSI) of the matching data set ranged from 14 to 98, with a mean of 52, covering oligotrophic, mesotrophic and eutrophic water states. The significant optical variability provides favorable conditions for the development of algorithms for estimating the eutrophication index (TSI) of water bodies over a large range. (2) The total absorption coefficient of the three target wavelengths mentioned above , , , backscatter coefficient , , As the independent variable and the matching point eutrophication index (TSI) as the dependent variable, a semi-analytical model of lake eutrophication index was constructed and the optimal model coefficient was determined. (3) Through multiple linear regression, the semi-analytical model of eutrophication index (TSI) was obtained as follows:

[0096] .

[0097] like Figure 1 As shown in Figure 2, the model validation set R of the semi-analytical model of lake eutrophication index 2 is 0.68, the average relative error is 15.74%, and the deviation is -0.25; Figure 1 (a) and (b) are the construction diagram and verification diagram of the semi-analytical model of lake eutrophication index, respectively.

[0098] The construction of the lake eutrophication index gradient tree boosting model is as follows: , , , , band ratio, and cyanobacteria index were used as model inputs to construct a gradient tree boosting model for lake eutrophication index. The gradient tree boosting model is an ensemble learning method that improves prediction accuracy by gradually fitting residuals. During the model training process, a simple model is first initialized, and then the residuals of the current model are calculated in each iteration. Then, a decision tree is used to fit these residuals, and finally, the prediction results of the new tree are combined with the current model. The gradient tree boosting model can effectively capture the complex nonlinear relationships in the data and show strong prediction capabilities in a variety of regression and classification tasks. In the process of hyperparameter adjustment, a grid of hyperparameters (such as learning rate, number of trees, maximum depth, etc.) is constructed through grid search, and then the GridSearchCV class of the machine learning library (scikit-learn) is used for systematic search to find the best parameter combination and score (as shown in Table 1). This process evaluates model performance through cross-validation to obtain the optimal model.

[0099] Table 1. Hyperparameter debugging of gradient tree boosting model

[0100]

[0101] like Figure 2 As shown in the figure, the lake eutrophication index gradient tree boosting model structure shows that the model validation set R 2 is 0.68, the average relative error is 15.59%, and the deviation is 0.31; Figure 2 (a) and (b) are the construction and verification diagrams of the lake eutrophication index gradient tree boosting model.

[0102] like Figure 5 As shown, the robustness of the above three models is analyzed by Monte Carlo method, and the results of the comprehensive model accuracy and robustness analysis are combined to prove the significant advantages of the lake eutrophication index coupling model provided in this embodiment in this field; wherein, Figure 5 (a) and (b) are comparison diagrams of root mean square error and determination coefficient.

[0103] In each round of simulation, random errors (0~1, step size is 0.1) are added to the test set. The error is a normally distributed noise generated by a random function, and its standard deviation is the current error level. For example, error level = 0.1 means that the generated noise will cause each feature data of the test set to fluctuate by up to 10%. The simulation is run 100 times at each error level. This is to obtain stable estimates of RMSE and R² through multiple experiments. The Monte Carlo simulation results show that the accuracy of the integrated model is significantly improved compared with the semi-analytical and gradient tree boosting models at different error levels. As the error level continues to increase, the accuracy of the semi-analytical and gradient tree boosting models decreases significantly. When the error level = 1, R 2 dropped to below 0.3; while the integrated model accuracy (measured in R 2 The above results show that compared with the semi-analytical model and the gradient tree boosting model, the accuracy of the lake eutrophication index coupling model (measured in R 2 For example, the lake eutrophication index coupling model has obvious advantages in both model accuracy and stability.

[0104] Example 2

[0105] This embodiment provides a system for constructing a lake eutrophication index coupling model, which is used to implement a method for constructing a lake eutrophication index coupling model as described in Example 1, including:

[0106] The first module is configured to obtain original band data of different optical bands, pre-process the original band data, and obtain effective band data; the effective band data at least includes: remote sensing reflectance and central wavelength; based on the remote sensing reflectance and central wavelength corresponding to different optical bands, the cyanobacteria index and the band ratio are calculated;

[0107] The second module is configured to determine an observation index from the effective band data; judge the observation index to obtain a judgment result, and determine a reference wavelength based on the judgment result;

[0108] The third module is configured to calculate the total absorption coefficient and backscattering coefficient of the reference wavelength for the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculate the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths;

[0109] The fourth module is configured to use the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, construct a hyperparameter grid through grid search, and construct a lake eutrophication index coupling model.

Claims

1. A method for constructing a lake eutrophication index coupling model, characterized in that: The following steps are involved: Step S1, obtaining original band data of different optical bands, preprocessing the original band data to obtain effective band data; the effective band data at least includes: remote sensing reflectance and central wavelength; based on the remote sensing reflectance and central wavelength corresponding to different optical bands, calculating the cyanobacteria index and the band ratio; Step S2, determining an observation index from the effective band data; judging the observation index to obtain a judgment result, and determining a reference wavelength based on the judgment result; Step S3, calculating the total absorption coefficient and backscattering coefficient of the reference wavelength for the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculating the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths; Step S4, using the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, constructing a hyperparameter grid through grid search, and constructing a lake eutrophication index coupling model; The step S2 further comprises the following steps: Select the remote sensing reflectance of the red band from the valid band data As an observation indicator; With threshold Make a comparison and get the result: like , then the reference wavelength The value of ; like , then the reference wavelength The value of ; exist , the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula: ； ； ； In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; is the intermediate process variable; r rs (469), r rs (555) are the remote sensing reflectances at wavelengths of 469 nm and 555 nm respectively; The wavelength The remote sensing reflectance of the location, is the wavelength Remote sensing reflectivity of the location; exist , then the reference wavelength The value of When the total absorption coefficient of the reference wavelength in step S3 is calculated using the following formula: ； In the formula, , The wavelengths are Total absorption coefficient of the position, absorption coefficient of pure water; It is the remote sensing reflectance at a wavelength of 645nm.

2. The method for constructing a lake eutrophication index coupling model according to claim 1, characterized in that: In step S1, the cyanobacteria index is calculated using the following formula: ； In the formula, is the cyanobacteria index; , , They are the remote sensing reflectances in the red, blue and near-infrared bands respectively; , , , They are the center wavelengths of the green, blue, red and near-infrared bands respectively.

3. The method for constructing a lake eutrophication index coupling model according to claim 1, characterized in that: The backscattering coefficient of the reference wavelength in step S3 is calculated using the following formula: ； ； In the formula, , The reference wavelengths are Backscattering coefficient of position, backscattering coefficient of pure water; is an intermediate process variable; The reference wavelength The total absorption coefficient of the location; The reference wavelength Remote sensing reflectance of the location; reference wavelength The value is or .

4. The method for constructing a lake eutrophication index coupling model according to claim 3, characterized in that: The backscattering coefficients corresponding to the N target wavelengths in step S3 are calculated using the following formula: ； In the formula, is the target wavelength The backscatter coefficient at position, .

5. The method for constructing a lake eutrophication index coupling model according to claim 4, characterized in that: The total absorption coefficient corresponding to the N target wavelengths in step S3 is calculated using the following formula: ； ； In the formula, is the target wavelength The total absorption coefficient at the position, ; is the target wavelength Pure water backscattering coefficient at position; is an intermediate process variable; The target wavelength Remote sensing reflectance of the location.

6. The method for constructing a lake eutrophication index coupling model according to claim 2, characterized in that: The band ratios include: , , , , , ; is the remote sensing reflectance of the green band.

7. A system for constructing a lake eutrophication index coupling model, used to implement a method for constructing a lake eutrophication index coupling model as claimed in any one of claims 1 to 6, characterized in that: include: The first module is configured to obtain original band data of different optical bands, and pre-process the original band data to obtain effective band data; The effective band data at least includes: remote sensing reflectance and central wavelength; based on the remote sensing reflectance and central wavelength corresponding to different optical bands, the cyanobacteria index and band ratio are calculated; The second module is configured to determine an observation index from the effective band data; judge the observation index to obtain a judgment result, and determine a reference wavelength based on the judgment result; The third module is configured to calculate the total absorption coefficient and backscattering coefficient of the reference wavelength for the reference wavelength; based on the total absorption coefficient and backscattering coefficient of the reference wavelength, calculate the total absorption coefficient and backscattering coefficient corresponding to N target wavelengths; The fourth module is configured to use the effective band data, cyanobacteria index, band ratio, and total absorption coefficient and backscattering coefficient corresponding to N target wavelengths as input parameters of the gradient tree boosting algorithm, construct a hyperparameter grid through grid search, and construct a lake eutrophication index coupling model.

Images