Method suitable for remote sensing inversion of high-turbidity estuary granular organic carbon concentration

Abstract

The invention relates to the technical field of water quality parameter remote sensing inversion, and discloses a simple method suitable for high-turbidity estuary particulate organic carbon concentration remote sensing inversion, which comprises the following steps: collecting and preprocessing suspended particulate matter (SPM) concentration, particulate organic carbon (POC) concentration data and corresponding remote sensing reflectivity data of a water body; according to a wave spectrum response function of each wave band of a common optical satellite sensor, the equivalent remote sensing reflectivity Rrs (lambda i) of each wave band is obtained through simulation, and by comparing the correlation between the remote sensing reflectivity Rrs (lambda i) of the visible light-near infrared wave band and the actually measured suspended particulate matter (SPM) concentration, the POC inversion model constructed by the method adopts a subsection modeling strategy, so that the concentration of the SPM can be calculated. The concentration of the suspended particulate matters with strong correlation is preferably selected as a correlation parameter, so that a relatively large error generated by an existing multi-parameter model is effectively avoided, and the precision of the inversion of the concentration of the particulate organic carbon in the high-turbidity estuary region is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of remote sensing inversion technology for water quality parameters, specifically a method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries. Background Technology

[0002] Remote sensing inversion of water color is a method for detecting the concentration of substances in the surface layer of water bodies at long distances using airborne or spaceborne sensors. It typically establishes a relationship between the water's radiance or reflectance and the substance concentration. However, phosphoric acid (POC) itself is not an optically active component. Inversion is usually performed using the water's inherent / apparent optical properties or correlation parameters. This includes using the ratio or difference of remote sensing reflectance in the blue-green bands to invert POC concentration; estimating POC concentration using light attenuation coefficients, diffuse reflectance coefficients, or backscattering coefficients; and indirectly estimating POC concentration using POC-related suspended particulate matter (SPM), chlorophyll a, and particle size. The first two methods are suitable for clear ocean waters (Class I), while nearshore waters have complex optical properties and significant spatial heterogeneity, and POC sources are abundant, rendering these methods ineffective. Therefore, indirect inversion using correlation parameters has become a feasible solution for remote sensing inversion of POC concentration in nearshore waters.

[0003] However, the strong water-sediment coupling and dramatic geomorphic evolution in the estuary region have shaped complex optical gradients and spatial heterogeneity in the water. The spectral characteristics of the water are dominated by strong scattering by particulate matter (SPM), resulting in a weak spectral signal for chlorophyll and significantly limiting the accuracy of chlorophyll concentration retrieval (>30%). Furthermore, particulate matter size itself does not possess optical properties and often requires stepwise estimation using semi-analytical algorithms, leading to poor retrieval accuracy in existing studies (>35%). Therefore, methods that estimate POC concentration using these two parameters will transmit significant process errors to the final retrieval result.

[0004] Since POCs primarily adhere to suspended particulate matter, indirectly estimating POC concentration using SPM concentration is a simple and effective method. Currently, remote sensing inversion of POC concentration mainly relies on establishing a linear relationship between the two. However, in high-turbidity estuaries, SPM concentrations can span four orders of magnitude, making accurate SPM concentration inversion challenging. The functional relationship between SPM and POC concentrations also becomes more complex, rendering existing algorithms inapplicable. Therefore, accurately estimating SPM concentrations in high-turbidity estuaries from satellite imagery and clarifying the quantitative relationship between estuarine SPM and POC is crucial for remote sensing inversion of POC concentration in high-turbidity estuaries. Summary of the Invention

[0005] Technical problems to be solved

[0006] The purpose of this invention is to address the problem that existing remote sensing inversion algorithms for particulate organic carbon (POC) concentration cannot be applied to turbid estuarine waters. Based on the accurate inversion of SPM concentration, this invention constructs a segmented remote sensing inversion algorithm for POC concentration based on suspended particulate matter concentration, providing a solution for monitoring POC in turbid estuarine waters using optical satellite data.

[0007] Technical solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries, comprising the following steps:

[0009] Step 1: Collect data on suspended particulate matter (SPM) concentration, particulate organic carbon (POC) concentration, and corresponding remote sensing reflectance data of the water body and perform preprocessing.

[0010] Step 2: Based on the spectral response functions of commonly used optical satellite sensors for each band, simulate the equivalent remote sensing reflectance R for each band. rs (λ i );

[0011] Step 3: By comparing the equivalent remote sensing reflectance R in the visible-near-infrared bands mentioned above. rs (λ i The correlation between the measured and actual suspended particulate matter (SPM) concentrations was investigated to construct a remote sensing inversion algorithm for SPM concentrations in water bodies.

[0012] Step 4: Construct a piecewise function between the concentration of suspended particulate matter (SPM) and the concentration of particulate organic carbon (POC), and use the least squares method to calculate the optimal segmentation threshold TH of the piecewise function to establish the quantitative relationship between the concentration of suspended particulate matter (SPM) and particulate organic carbon (POC).

[0013] Step 5: Utilize the on-site measured SPM concentration data and equivalent remote sensing reflectance

[0014] R rs (λ i The data is used to obtain the fitting coefficients α0, α1, α2 and α3 in step 3 based on the least squares method, and the inversion model of suspended particulate matter (SPM) concentration is obtained. Then, the optimal segmentation threshold TH in step 4 is determined using the measured data of suspended particulate matter (SPM) concentration and particulate organic carbon (POC) concentration, and the undetermined coefficients β0, β1, γ0, γ1 and γ2 are fitted to finally determine the inversion model of particulate organic carbon (POC) concentration.

[0015] Step 6: Use on-site synchronous measurement data to evaluate the accuracy of the particulate organic carbon (POC) concentration results retrieved from the satellite, and finally complete the satellite mapping of the particulate organic carbon (POC) concentration in the high-turbidity estuary area.

[0016] As a further description of the above technical solution, the preferred sensitive band in step 3 is selected from the red band, red-edge band, and near-infrared band, and the selection is related to the equivalent remote sensing reflectance R. rs (λ i The three bands with the highest correlation were selected as sensitive bands, and the green band with the weakest correlation was used as the denominator to construct a four-band weighted ratio model.

[0017] As a further description of the above technical solution, the formula for constructing the four-band weighted ratio model is as follows:

[0018]

[0019] Where α0, α1, α2, and α3 are all coefficients to be fitted; R rs (Red), R rs (Red_Edge), R rs (NIR) represent the equivalent remote sensing reflectance in the red band, red edge band, and near-infrared band, respectively; w1, w2, and w3 represent R... rs (Red), R rs (Red_Edge) and R rs The weighting coefficients of (NIR).

[0020] As a further description of the above technical solution, in step 4, a piecewise function between the suspended particulate matter (SPM) concentration and the particulate organic carbon (POC) concentration is constructed using linear and quadratic polynomial functions. Specifically, the logarithm of the SPM concentration (log(SPM)) is used as the segmentation object, and the coefficient of determination R is used as the dividing line. 2 The root mean square error (RMSE) and the mean relative error (RPD) are used as indicators to evaluate the model accuracy, such that particulate organic carbon (POC) = f(log 10 SPM achieves optimal accuracy, at which point the coefficient of determination R0 is at its highest. 2 The root mean square error (RMSE) and the mean relative error (RPD) reach their maximum values, while the root mean square error (RMSE) and the mean relative error (RPD) reach their minimum values.

[0021] As a further description of the above technical solution, in step 4, when calculating the two-segment function using the least squares method, to prevent abrupt changes in the satellite inversion results, the segmentation threshold TH is finely adjusted, i.e., TH±0.05. A smoothing function is used for the transition between the two segments. The specific calculation process is as follows:

[0022] When log(SPM)≥(TH+0.05), the following function is used:

[0023] log(POC high )=β0+β1×log(SPM)

[0024] Among them, β0 and β1 are coefficients to be fitted;

[0025] When log(SPM) ≤ (TH - 0.05), the following function is adopted:

[0026] log(POC low ) = γ0 + γ1 × log(SPM) + γ2 × [log(SPM)] 2

[0027] Among them, γ0, γ1 and γ2 are coefficients to be fitted;

[0028] When (TH - 0.05) < log(SPM) < (TH + 0.05), the following function is adopted:

[0029] log(POC med ) = δ × log(POC low ) + (1 - δ) × log(POC high )

[0030]

[0031] Among them, δ is the weight coefficient for the transition of the piecewise model, and TH is the segmentation threshold of log(SPM).

[0032] As a further description of the above technical solution, in the accuracy evaluation index of the particulate organic carbon POC concentration inversion model;

[0033] The calculation formula of the coefficient of determination R 2 is as follows:

[0034] [[ID=4,2]]

[0035] The calculation formula of the root mean square error RMSE is as follows:

[0036]

[0037] The calculation formula of the average relative error RPD is as follows:

[0038]

[0039] Among them, φ i is the concentration of particulate organic carbon POC measured on site; φ' i is the concentration of particulate organic carbon POC obtained by推算 according to the above functional relationship; is the average value of the concentrations of particulate organic carbon POC for all samples; n is the number of samples of the particulate organic carbon POC concentration measured on site.

[0040] As a further description of the above technical solution, in step 6, the satellite data is preprocessed such as radiometric calibration and atmospheric correction to obtain a remote sensing reflectance image. Then, the particulate organic carbon (POC) concentration inversion model is applied to this satellite image, and the accuracy of the satellite-retrieved POC concentration results is evaluated using data from synchronous field measurements. The evaluation index used in the accuracy evaluation is R. 2 RMSE and RPD.

[0041] As a further description of the above technical solution, in step 1, the dataset matching the concentration of suspended particulate matter (SPM) with remote sensing reflectance consists of 214 sets, and the dataset matching the concentration of suspended particulate matter (SPM) with the concentration of particulate organic carbon (POC) consists of 457 sets.

[0042] As a further description of the above technical solution, in step 2, the spectral response function of each band of the commonly used optical satellite sensor is the spectral response function of each band of the Sentinel-2MSI land satellite, and the equivalent remote sensing reflectance R of each band is calculated. rs (λ i The calculation formula for ) is as follows:

[0043]

[0044] Where i is the band number of each satellite sensor; λ i R(λ) represents the corresponding center wavelength. i S(λ) represents the remote sensing reflectance of the water body measured on-site; i R is the spectral response function; rs (λ i ) represents the equivalent remote sensing reflectance after integration.

[0045] As a further description of the above technical solution, step 5 employs a four-fold cross-validation method to evaluate the performance of the particulate organic carbon (POC) inversion model. The four-fold cross-validation method involves randomly dividing the samples into four groups, using any three groups to train the coefficients of the particulate organic carbon (POC) segmentation model, and then using the remaining group for validation. This process is repeated four times. The final evaluation metric is still R0. 2 RMSE and RPD.

[0046] Beneficial effects

[0047] Compared with existing technologies, this invention provides a method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries, which has the following beneficial effects:

[0048] 1. This invention fully considers the complex optical characteristics of high-turbidity estuarine waters and the non-photosensitive properties of particulate organic carbon itself, and preferentially selects strongly correlated suspended particulate matter concentration as the correlation parameter, effectively avoiding the large errors produced by existing multi-parameter models. At the same time, this algorithm is simpler and more effective, and has strong universality in other high-turbidity estuarine areas.

[0049] 2. To address the issue of large variations in suspended particulate matter concentration in high-turbidity estuaries, this invention fully considers the optical characteristics of water bodies ranging from highly turbid to clear, selects the most sensitive bands, and employs a four-band weighted ratio algorithm to construct an inversion model for suspended particulate matter concentration in estuaries, thereby effectively improving the remote sensing inversion accuracy of particulate organic carbon concentration in estuaries.

[0050] 3. Considering the huge gradient changes in the composition of high-turbidity estuarine water, a segmented modeling strategy was adopted to construct a high-precision quantitative relationship between suspended particulate matter and particulate organic carbon. Compared with the existing simple linear model, this invention significantly improves the accuracy of particulate organic carbon concentration inversion in high-turbidity estuarine areas. Attached Figure Description

[0051] Figure 1 This is a technical roadmap of the present invention; Figure 2 Spectral curves of water bodies were measured on-site at the Yellow River estuary; Figure 3 The relationship between suspended particulate matter (SPM) concentration and particulate organic carbon (POC) concentration; Figure 4 The optimal segmentation threshold for the piecewise function between suspended particulate matter (SPM) concentration and particulate organic carbon (POC) concentration; Figure 5 To evaluate the accuracy of the particulate organic carbon (POC) concentration inversion model; Figure 6 A graph for assessing the accuracy of satellite-retrieved particulate organic carbon (POC) concentration results; Figure 7 Map showing the POC concentration distribution in the Yellow River Estuary during (a) the flood season and (b) the dry season in 2022; Figure 8 This is a comparison chart showing the accuracy of the algorithm proposed in this invention compared to existing algorithms. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates the remote sensing inversion algorithm for particulate organic carbon concentration in high-turbidity estuarine waters proposed in this invention. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this invention.

[0060] Example:

[0061] See attached document Figure 1-7 This embodiment provides a remote sensing inversion algorithm for particulate organic carbon concentration in high-turbidity Yellow River estuary waters applicable to optical satellites (taking Sentinel-2MSI as an example), specifically including the following steps:

[0062] Step 1: Collect data on suspended particulate matter (SPM) concentration, particulate organic carbon (POC) concentration, and corresponding remote sensing reflectance data in the Yellow River estuary. Figure 2 The dataset was preprocessed, with 214 datasets matching SPM concentration with remote sensing reflectance and 457 datasets matching SPM concentration with particulate organic carbon (POC) concentration.

[0063] Step 2: Based on the spectral response functions of each band of a commonly used optical satellite sensor (Sentinel-2MSI land satellite), simulate the equivalent remote sensing reflectance R of each band. rs (λ i ), calculate the equivalent remote sensing reflectance R for each band. rs (λ i The calculation formula for ) is as follows:

[0064]

[0065] Where i is the band number of each satellite sensor (MSI); λ i R(λ) represents the corresponding center wavelength. i S(λ) represents the remote sensing reflectance of the water body measured on-site; i R is the spectral response function; rs (λ i () represents the equivalent remote sensing reflectance after integration;

[0066] Step 3: By comparing the equivalent remote sensing reflectance R in the visible-near-infrared bands mentioned above. rs (λ iThe correlation between the measured and actual suspended particulate matter (SPM) concentrations was used to construct a remote sensing inversion algorithm for SPM concentration in water bodies, as shown in the following formula:

[0067]

[0068] Where α0, α1, α2, and α3 are all coefficients to be fitted; R rs (Red), R rs (Red_Edge), R rs (NIR) represent the equivalent remote sensing reflectance in the red band, red edge band, and near-infrared band, respectively; w1, w2, and w3 represent R... rs (Red), R rs (Red_Edge) and R rs The weighting coefficients of (NIR).

[0069] For example, choosing the equivalent remote sensing reflectance R rs (λ i The red band (665nm), red-edge band (740nm), and near-infrared band (865nm), which have high correlation, are used as the denominator, and a four-band weighted ratio model is constructed as follows:

[0070]

[0071] Using 214 sets of SPM-R rs The dataset was fitted using the least squares method, and the undetermined coefficients were α0 = 0.3652, α1 = 2.066, α2 = 2.4327, and α3 = 3.9384.

[0072] Step 4: Construct a piecewise function between the SPM concentration and the POC concentration. This is done using both linear and quadratic polynomial functions, specifically using the logarithm of the SPM concentration (log(SPM)) as the segmentation object and the coefficient of determination R... 2 The root mean square error (RMSE) and the mean relative error (RPD) are used as indicators to evaluate the model accuracy, such that particulate organic carbon (POC) = f(log 10 SPM achieves optimal accuracy, at which point the coefficient of determination R0 is at its highest. 2 The root mean square error (RMSE) and the mean relative error (RPD) reach their maximum values, while the root mean square error (RMSE) and the mean relative error (RPD) reach their minimum values.

[0073] From the distribution maps of POC concentration and SPM concentration ( Figure 3) It can be seen that when the SPM is low, the relationship between log(POC) and log(SPM) is non-linear; when the SPM is high, the relationship between log(POC) and log(SPM) is a simple linear relationship. Therefore, the present invention presets the relationship between POC and SPM as a quadratic polynomial plus a linear equation. Taking log(SPM) as the segmentation object, with the coefficient of determination R 2 , the root mean square error (RMSE), and the average relative error (RPD) as the accuracy evaluation indicators, the least squares method is used to calculate the optimal segmentation threshold (TH = 3.18) of the two-segment function, so that the accuracy of log(POC) = f(log(SPM)) is optimal. At this time, R 2 reaches the maximum (0.96), and RMSE (12.25) and RPD (32.53%) reach the minimum ( Figure 4 ). In addition, to prevent sudden changes in the satellite inversion results, the segmentation threshold is finely tuned, that is, TH ± 0.05, and a smoothing function is used for transition between the two segments. The final relationship is as follows:

[0074] When log(SPM) ≥ 3.23, the following function is used:

[0075] log(POC high ) = β0 + β1 × log(SPM)

[0076] where β0 and β1 are -1.8338 and 1.0925 respectively.

[0077] When log(SPM) ≤ 3.13, the following function is used:

[0078] log(POC low ) = γ0 + γ1 × log(SPM) + γ2 × [log(SPM)] 2

[0079] where γ0, γ1, and γ2 are - 0.2768, -0.7509, and 0.4264 respectively.

[0080] When 3.13 < log(SPM) < 3.23, the following function is used:

[0081] log(POC med ) = δ × log(POC low ) + (1 - δ) × log(POC high )

[0082] [[ID=4!5]] <!

[0083] where δ is the weight coefficient for the transition of the segmented model, and TH is the segmentation threshold of log(SPM).

[0084] In the accuracy evaluation index of the particulate organic carbon (POC) concentration inversion model;

[0085] Coefficient of determination R 2 The calculation formula is as follows:

[0086]

[0087] The formula for calculating the root mean square error (RMSE) is as follows:

[0088]

[0089] The formula for calculating the mean relative error RPD is as follows:

[0090]

[0091] Where, φ i The concentration of particulate organic carbon (POC) measured on-site; φ′ i The particulate organic carbon (POC) concentration obtained from the above functional relationship; is the average particulate organic carbon (POC) concentration of all samples; n is the number of samples for which particulate organic carbon (POC) concentration was measured on-site.

[0092] Step 5: Utilize the on-site measured SPM concentration data and equivalent remote sensing reflectance

[0093] R rs (λ i Based on the data, the coefficients α0, α1, α2, and α3 to be fitted in step 3 are obtained using the least squares method, thus obtaining the inversion model of suspended particulate matter (SPM) concentration. Then, using the measured SPM and POC concentration data, the optimal segmentation threshold TH in step 4 is determined, and the undetermined coefficients β0, β1, γ0, γ1, and γ2 are fitted to obtain the final POC concentration inversion model. The performance of the POC inversion model is evaluated using a four-fold cross-validation method. The four-fold cross-validation method involves randomly dividing the samples into four groups, using any three groups to train the coefficients of the POC segmentation model, and then using the remaining group for validation. This process is repeated four times. The final evaluation index is still R. 2 RMSE and RPD.

[0094] Step 6: Evaluate the accuracy of the satellite-retrieved particulate organic carbon (POC) concentration results using synchronous on-site measurement data. Preprocess the satellite data, including radiometric calibration and atmospheric correction, to obtain a remotely sensed reflectance image. Then, apply the POC concentration retrieval model to this satellite image and evaluate the accuracy of the satellite-retrieved POC concentration results using synchronous on-site measurement data. The evaluation metric used for accuracy evaluation is R0.2 RMSE and RPD ultimately completed the satellite mapping of particulate organic carbon (POC) concentration in the estuary area.

[0095] Experimental example:

[0096] Reference Figures 5-7 This experimental example uses the particulate organic carbon (POC) concentration inversion model constructed in the above embodiments. Using a dataset of 457 measured SPM-POC concentrations, the POC inversion model is evaluated based on the four-fold cross-validation method. The evaluation metric remains R. 2 The results of RMSE and RPD indicate that the algorithm of the present invention has high accuracy (R... 2 =0.96, RMSE=12.21mg / L and RPD=32.63%);

[0097] Cloudless Sentinel-2MSI Level-1b satellite imagery was selected and preprocessed, including radiometric calibration and atmospheric correction, to obtain remote sensing reflectance data. The above POC concentration inversion model was applied to the satellite imagery, and the accuracy of the satellite inversion results was evaluated using field-measured POC concentration data collected within two hours of the satellite's transit. Figure 6 This indicates that satellite inversion based on the algorithm of this invention also achieves high accuracy (R). 2 =0.86 and RPD=28.12%), finally completing the satellite mapping of POC concentration in the Yellow River Estuary region ( Figure 7 );

[0098] Using 216 sets of field-measured suspended particulate matter (SPM) concentration data and equivalent remote sensing reflectance data, the undetermined coefficients α0 = 1.678, α1 = 0.514, β0 = -0.091, β1 = 2.338, and β2 = 4.475 in the piecewise model were obtained based on least squares fitting. The final suspended particulate matter concentration inversion algorithm was then derived, and the accuracy of the inversion algorithm was evaluated. Figure 5 ).

[0099] Currently, for estuarine areas with high turbidity, existing studies mainly rely on constructing a simple linear relationship between SPM and POC to invert POC concentration. The main model types are as follows:

[0100]

[0101] in, ω and ω are undetermined coefficients. To compare the advantages and disadvantages of the existing algorithms with the algorithm proposed in this invention, the coefficients of the above models were recalibrated using a dataset of 457 measured SPM-POC concentrations, and then R was used. 2 The accuracy was evaluated using RMSE and RPD metrics. The results are as follows:

[0102]

[0103] As can be seen, the accuracy of Model 1 (Liu et al., 2015) and Model 2 (Mou et al., 2017) is significantly lower than that of the algorithm of this invention. Model 1 exhibits a large error at low POC, while Model 2 exhibits a large error at high POC. The algorithm of this invention performs well at both low and high POC.

[0104] It should be noted that the term "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries, comprising the following steps: Step 1: Collect data on the concentration of suspended particulate matter (SPM), the concentration of particulate organic carbon (POC) in the water body, and the corresponding remote sensing reflectance data, and perform preprocessing; Step 2: Based on the spectral response functions of commonly used optical satellite sensors for each band, simulate the equivalent remote sensing reflectance R for each band. rs (λ i ); Step 3: By comparing the equivalent remote sensing reflectance R in the visible-near-infrared bands mentioned above. rs (λ i The correlation between the concentration of suspended particulate matter (SPM) and the measured concentration of SPM is analyzed, and bands with higher correlation are selected as sensitive bands to construct a remote sensing inversion algorithm for the concentration of SPM in water areas. Step 4: Construct a piecewise function between the concentration of suspended particulate matter (SPM) and the concentration of particulate organic carbon (POC), use the least squares method to calculate the optimal segmentation threshold TH of the piecewise function, and establish a quantitative relationship between the concentration of suspended particulate matter (SPM) and particulate organic carbon (POC); Step 5: Utilize the on-site measured SPM concentration data and equivalent remote sensing reflectance R rs (λ i The data is used to fit the coefficients α0, α1, α2 and α3 in step 3 to obtain the inversion model of suspended particulate matter (SPM) concentration. Then, using the measured data of suspended particulate matter (SPM) concentration and particulate organic carbon (POC) concentration, the optimal segmentation threshold TH in step 4 is determined, and the coefficients β0, β1, γ0, γ1 and γ2 are fitted to obtain the coefficients β0, β1, γ0, γ1 and γ2. Finally, the inversion model of particulate organic carbon (POC) concentration is determined. Step 6: Use the on-site synchronous measurement data to evaluate the accuracy of the satellite-inverted particulate organic carbon (POC) concentration results, and finally complete the satellite mapping of the particulate organic carbon (POC) concentration in the high-turbidity estuary area.

2. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 1, characterized in that: In step 3, the preferred sensitive band is selected from the red band, red-edge band, and near-infrared band, and the selection is related to the equivalent remote sensing reflectance R. rs (λ i The three bands with the highest correlation were selected as sensitive bands, and the green band with the weakest correlation was used as the denominator to construct a four-band weighted ratio model.

3. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 2, characterized in that: The formula for constructing the four-band weighted ratio model is as follows: Where α0, α1, α2, and α3 are all coefficients to be fitted; R rs (Red), R rs (Red_Edge), R rs (NIR) represent the equivalent remote sensing reflectance in the red band, red edge band, and near-infrared band, respectively; w1, w2, and w3 represent R... rs (Red), R rs (Red_Edge) and R rs The weighting coefficients of (NIR).

4. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 1, characterized in that: In step 4, a piecewise function is constructed between the suspended particulate matter (SPM) concentration and the particulate organic carbon (POC) concentration using linear and quadratic polynomial functions. Specifically, the logarithm (log(SPM)) of the SPM concentration is used as the segmentation object, and the coefficient of determination R is used as the dividing line. 2 The root mean square error (RMSE) and the mean relative error (RPD) are used as indicators to evaluate the model accuracy, such that particulate organic carbon (POC) = f(log 10 SPM achieves optimal accuracy, at which point the coefficient of determination R0 is at its highest. 2 The root mean square error (RMSE) and the mean relative error (RPD) reach their maximum values, while the root mean square error (RMSE) and the mean relative error (RPD) reach their minimum values.

5. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 4, characterized in that: When calculating the two-segment function by the least squares method in Step 4, to prevent sudden changes in the satellite inversion results, consider fine-tuning the segmentation threshold TH, that is, TH±0.05, and use a smoothing function to transition between the two-segment functions. The specific calculation process is as follows: When log(SPM)≥(TH + 0.05), use the following function: log(POC high )=β0+β1×log(SPM) Where, β0 and β1 are coefficients to be fitted; When log(SPM)≤(TH - 0.05), use the following function: log(POC low )=γ0+γ1×log(SPM)+γ2×[log(SPM)] 2 Where, γ0, γ1 and γ2 are coefficients to be fitted; When (TH - 0.05)<log(SPM)<(TH + 0.05), use the following function: log(POC med )=δ×log(POC low )+(1-δ)×log(POC high ) Where, δ is the weight coefficient for the transition of the piecewise model, and TH is the segmentation threshold of log(SPM).

6. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 4, characterized in that: Among the accuracy evaluation indicators of the particulate organic carbon (POC) concentration inversion model; Coefficient of determination R 2 The calculation formula is as follows: The formula for the root mean square error (RMSE) is as follows: The formula for the average relative error (RPD) is as follows: Where, φ i The concentration of particulate organic carbon (POC) measured on-site; φ′ i The particulate organic carbon (POC) concentration obtained from the above functional relationship; is the average particulate organic carbon (POC) concentration of all samples; n is the number of samples for which particulate organic carbon (POC) concentration was measured on-site.

7. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 1, characterized in that: In step 6, satellite data undergoes preprocessing such as radiometric calibration and atmospheric correction to obtain a remotely sensed reflectance image. Then, the particulate organic carbon (POC) concentration inversion model is applied to this satellite image. The accuracy of the satellite-retrieved POC concentration results is evaluated using data from synchronous field measurements. The evaluation metric used for accuracy assessment is R0. 2 RMSE and RPD.

8. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 1, characterized in that: In step 2, the equivalent remote sensing reflectance R of each band is calculated. rs (λ i The calculation formula for ) is as follows: Where i is the band number of each satellite sensor; λ i R(λ) represents the corresponding center wavelength. i S(λ) represents the water body remote sensing reflectance measured on-site; i R is the spectral response function; rs (λ i ) represents the equivalent remote sensing reflectance after integration.

9. The method for remote sensing inversion of particulate organic carbon concentration in high-turbidity estuaries according to claim 1, characterized in that: In Step 5, the four-fold cross-validation method is used to evaluate the performance of the particulate organic carbon (POC) inversion model.

Images