Remote sensing inversion method for suspended particulate matter in high-turbidity water bodies using high-resolution satellites

Abstract

The present invention discloses a remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites, which relates to the technical field of suspended particulate matter monitoring. The method comprises the following steps: collecting water suspended particulate matter concentration data and corresponding water remote sensing reflectivity data, pre-processing the data to obtain equivalent remote sensing reflectivity data; constructing a segmented water suspended particulate matter concentration inversion model; using the remote sensing reflectivity of the red band as the segmentation object, and determining the coefficient R 2 The optimal segmentation threshold was calculated using the root mean square error (RMSE) as model accuracy evaluation metrics. The final segmentation model coefficients for the segmented suspended particulate matter concentration inversion model were obtained using the collected data through least squares fitting. The segmented suspended particulate matter concentration inversion model was then evaluated. This paper employs a segmented modeling strategy, combining a semi-analytical model with an empirical model and using weighted band ratios as model input, significantly improving the accuracy and universality of the inversion algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of suspended particulate matter monitoring, and in particular relates to a remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites. Background Art

[0002] Suspended particulate matter is an important water quality parameter. It is also the main carrier of organic carbon, nutrients, pollutants, etc. in water, and the material basis for the evolution of erosion and deposition in nearshore waters. Therefore, accurate monitoring of the spatiotemporal dynamics of suspended particulate matter is of great significance.

[0003] Traditional methods of onboard sampling and analysis, as well as fixed-point instrument monitoring, require significant human, financial, and material resources. They can only capture changes in suspended particulate matter within a limited area and specific time periods. Sparsely distributed stations also make it difficult to fully reflect the spatial distribution of suspended particulate matter across an entire region. In contrast, satellite remote sensing technology offers advantages such as wide coverage, high spatial resolution, fixed revisit times, a rich data source, low cost, and easy access. It has become an increasingly important tool for large-scale, quantitative, refined, and long-term continuous monitoring of suspended particulate matter.

[0004] Commonly used satellite data can be divided into two main categories: ocean color satellite data (such as SeaWiFS, MODIS, MERIS, GOCI, OLCI, etc.) and land satellite data (Landsat-TM / ETM+ / OLI, Sentinel-2MSI, PlanetScope, etc.). Ocean color satellites have bands for monitoring the optical properties of aquatic features, but their spatial resolution is relatively low (>250m), limiting their application in small and medium-sized water bodies, such as small lakes and reservoirs, river channels and estuaries, and near-shore ports. In recent years, the performance of high-resolution satellite (Landsat) sensors has continued to improve, not only providing continuous data on an interdecadal scale but also capable of capturing detailed changes in suspended particulate matter concentrations. However, high-resolution satellites often have fewer bands, lacking ultraviolet or red-edge bands. Selecting appropriate satellite bands and their combinations as input to suspended particulate matter concentration inversion algorithms is currently a challenge in modeling.

[0005] Remote sensing inversion of suspended particulate matter is a method that uses satellite-borne sensors to remotely detect the concentration of suspended particulate matter on the surface of water bodies. Its core is to establish a relationship between the water's off-water radiance or remote sensing reflectance and the suspended particulate matter concentration. In highly turbid waters, due to the strong scattering effect of suspended particulate matter in the water, the water's off-water radiance is less affected by chlorophyll and yellow matter. Its spectral characteristics are mainly dominated by suspended particulate matter. As the concentration of suspended particulate matter increases, the sensitive band gradually shifts from the green band to the red band and near-infrared band, which is called the "redshift phenomenon" (see Figure 2). Based on this, the suspended particulate matter concentration inversion algorithms currently used in commercial applications are mostly based on the green band or the red band, but they are not applicable to extremely turbid water bodies (suspended particulate matter concentrations span four orders of magnitude) because the remote sensing reflectivity of the green band and the red band tends to saturation as the suspended particulate matter concentration increases. Therefore, the above models often underestimate in highly turbid waters. In addition, the red band or the near-infrared band are not sensitive in low turbidity water bodies, so the inversion algorithms using the red band or the near-infrared band may fail in low turbidity water bodies. Therefore, it is necessary to develop a remote sensing inversion algorithm for suspended particulate matter in high-turbidity water bodies suitable for high-resolution satellites based on the spectral characteristics of different types of water bodies and taking into account the band settings of high-resolution satellites.

[0006] This invention aims to address the inability of existing remote sensing algorithms for suspended particulate matter concentration inversion to be applied to highly turbid waters, as well as the difficulty in modeling high-resolution satellite data due to its limited bandwidth. A segmented suspended particulate matter concentration inversion algorithm is proposed, providing a solution for the application of high-resolution satellite data in suspended particulate matter monitoring in highly turbid waters. Summary of the Invention

[0007] The purpose of the present invention is to provide a remote sensing inversion method for suspended particulate matter in high turbidity water bodies suitable for high-resolution satellites, so as to solve the problem that the remote sensing inversion algorithm of suspended particulate matter concentration in the existing technology proposed in the above background technology is not applicable to high turbidity waters and the problem that high-resolution satellite data is difficult to model due to the small number of bands.

[0008] To achieve the above objectives, the present invention adopts the following technical solutions:

[0009] The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies using high-resolution satellites includes the following steps:

[0010] S1. Collect the on-site measured water suspended particulate matter concentration data and the corresponding water remote sensing reflectance data, and pre-process the water remote sensing reflectance data to obtain equivalent remote sensing reflectance data;

[0011] S2. Constructing a segmented water suspended particle concentration inversion model; the segmented water suspended particle concentration inversion model includes a low turbidity water suspended particle concentration inversion model, a high turbidity water suspended particle concentration inversion model, and a transition smoothing function; calculating the backscattering coefficient of the green band based on the equivalent remote sensing reflectance data, and establishing a low turbidity water suspended particle concentration inversion model based on the backscattering coefficient of the green band;

[0012] S3. Selection of segmented water suspended particle concentration inversion model; using the remote sensing reflectance R rs (red) is the segmentation object, and the coefficient R is determined 2The root mean square error (RMSE) was used as the model accuracy evaluation index to calculate the optimal segmentation threshold of the inversion model of suspended particle concentration in low turbidity water body and the inversion model of suspended particle concentration in high turbidity water body.

[0013] S4. Determine the final segmented water suspended particle concentration inversion model; using the water suspended particle concentration data and equivalent remote sensing reflectance data in S1, obtain the undetermined coefficients in the segmented water suspended particle concentration inversion model based on the least squares fitting method, obtain the final segmented water suspended particle concentration inversion model, and evaluate the accuracy of the segmented water suspended particle concentration inversion model;

[0014] S5. Obtain the suspended particle concentration in water based on the segmented water suspended particle concentration inversion model; apply the segmented water suspended particle concentration inversion model to high-resolution satellite imagery, and use on-site synchronous measurement data to evaluate the accuracy of the inversion results of the segmented water suspended particle concentration inversion model, and finally complete the detailed mapping of the suspended particle concentration in the area of ​​interest.

[0015] Preferably, in S1, the water body remote sensing reflectance data is preprocessed to obtain equivalent remote sensing reflectance data, specifically as follows:

[0016] According to the spectral response function of each band of the high-resolution satellite sensor, the equivalent remote sensing reflectivity R of each band is simulated. rs (λ i );

[0017] The calculation formula is as follows:

[0018]

[0019] Where i is the band number of each satellite sensor; λ i is the corresponding central wavelength; R(λ i ) is the remote sensing reflectivity of water measured on site; S(λ i ) is the spectral response function; R rs (λ i ) is the integrated equivalent remote sensing reflectivity.

[0020] Preferably, the segmented water suspended particulate matter concentration inversion model is constructed in S2, specifically as follows:

[0021] S201, calculating the backscattering coefficient of the green band;

[0022] The calculation process is as follows:

[0023]

[0024]

[0025]

[0026]

[0027] a(green)=a w (green)+a nw (green)

[0028]

[0029] Among them, g1 and g2 are empirical coefficients, a w (green) is the absorption coefficient of pure water in the green band; b bw (green) is the backscattering coefficient of pure water in the green band; p0-p3 are also empirical coefficients, b bp (green) is the backscattering coefficient of the green band;

[0030] S202. Establish an inversion model for suspended particulate matter concentration in low turbidity water bodies;

[0031] The inversion model for suspended particulate matter concentration in low turbidity water is as follows:

[0032] log(TSM low )=α0+α1×log(b bp (green))

[0033] Among them, α0 and α1 are the coefficients to be fitted;

[0034] S203, establishing an inversion model for suspended particulate matter concentration in high turbidity water bodies;

[0035] The inversion model for suspended particulate matter concentration in high turbidity water is as follows:

[0036]

[0037]

[0038]

[0039] Among them, β0, β1, and β2 are the coefficients to be fitted; R rs (red) and R rs (NIR) are the equivalent remote sensing reflectances of the red band and the near infrared band respectively; w1 and w2 are weight coefficients;

[0040] S204, establishing a transition smoothing function for low-to-high turbidity water body transition;

[0041] The calculation formula of the transition smoothing function is as follows:

[0042] TMSmed =μ×TSM high +(1-μ)×TSM low

[0043]

[0044] Among them, μ is the weight factor, R rs (red) segmentation threshold.

[0045] Preferably, the empirical coefficient of the backscatter coefficient in the green band has the following value:

[0046] g1 and g2 take values ​​of 0.0895 and 0.1247;

[0047] The values ​​of p0-p3 are different for different high-resolution satellites, as follows:

[0048] For Landsat-5TM, the values ​​are: -1.08523, -1.10967, -0.23418, -10249;

[0049] For Landsat-7ETM+, the values ​​are: -1.14697, -1.15697, -0.26497, -0.08258;

[0050] For Landsat-8 / 9 OLI, the values ​​are: -1.15467, -1.17869, -0.24566, and -0.06989 respectively.

[0051] Preferably, the determination coefficient R in S3 2 And the root mean square error RMSE, as follows:

[0052] Coefficient of determination R 2 The calculation formula is as follows:

[0053]

[0054] The calculation formula of root mean square error RMSE is as follows:

[0055]

[0056] Among them, y i is the concentration of suspended particulate matter in water measured on site; i ′ is the concentration of suspended particulate matter in water obtained by model inversion; is the average value of the suspended particulate matter concentration in water of all samples; m is the number of samples of suspended particulate matter concentration in water measured on site.

[0057] Preferably, the selection of the inversion model for the suspended particulate matter concentration in the segmented water body in S3 is as follows:

[0058] like Adopting the inversion model of suspended particulate matter concentration in low turbidity water;

[0059] like The inversion model of suspended particulate matter concentration in highly turbid water bodies is used;

[0060] like Adopt transition smoothing function.

[0061] Preferably, the evaluation in S4 is performed using the determination coefficient R 2 The accuracy of the segmented water suspended particulate matter concentration inversion model was evaluated using the root mean square error (RMSE).

[0062] Preferably, the undetermined coefficients in S4 include α0 and α1 in the inversion model of suspended particle concentration in low turbidity water bodies and β0, β1 and β2 in the inversion model of suspended particle concentration in high turbidity water bodies.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] (1) The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies proposed in this invention, which is suitable for high-resolution satellites, fully considers the differences in optical properties between low-turbidity water bodies and high-turbidity water bodies, adopts a targeted segmented modeling strategy, combines the semi-analytical model with the empirical model, and significantly improves the accuracy and universality of the suspended particulate matter concentration inversion algorithm.

[0065] (2) The method of the present invention takes into account the limited number of bands of high-resolution satellites and only uses the red, green and blue bands to calculate the backscattering coefficient of the green band. The red band and near-infrared band that are common to high-resolution satellites are used as the input of the model in the form of weighted band ratios, which effectively improves the applicability of this method to the vast majority of high-resolution satellites, thereby realizing the fine observation of suspended particulate matter concentration. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of the remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites in the present invention;

[0067] Figure 2 It is a remote sensing reflectivity curve corresponding to different suspended particulate matter concentrations in the prior art;

[0068] Figure 3 This is a spectrum curve of the water body measured on-site at the Yellow River Estuary in Example 1 of the present invention.

[0069] Figure 4Schematic diagram of calculating the optimal segmentation threshold based on the least squares method in Example 1 of the present invention;

[0070] Figure 5 Schematic diagram of the accuracy of the inversion method in Example 1 of the present invention;

[0071] Figure 6 This is the distribution diagram of suspended particulate matter concentration at the Yellow River Estuary in Example 1 of the present invention (October 3, 2018). DETAILED DESCRIPTION

[0072] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0073] Example 1:

[0074] like Figure 1 As shown, this embodiment provides a remote sensing inversion method for suspended particulate matter in high turbidity water bodies applicable to high-resolution satellites (taking Landsat-8OLI as an example), which specifically includes the following steps:

[0075] Step 1: Collect the suspended particulate matter concentration data and the corresponding remote sensing reflectance data of the water area to be monitored; this example takes the Yellow River estuary with high sediment content as the research example, and a total of 216 sets of measured data sets ( Figure 3 ).

[0076] Step 2: Calculate the equivalent remote sensing reflectance R of each band based on the spectral response function of each band of Landsat-8OLI rs (λ i ).

[0077] The calculation formula is as follows:

[0078]

[0079] Where i is the band number of the OLI sensor; λ i is the corresponding central wavelength; R(λ i ) is the remote sensing reflectivity of water measured on site; S(λ i ) is the spectral response function; R rs (λ i ) is the integrated equivalent remote sensing reflectivity.

[0080] Step 3: Calculate the backscattering coefficient b of the green band bp (561nm);

[0081] The calculation process is as follows:

[0082]

[0083]

[0084]

[0085]

[0086] a(561nm)=a w (561nm)+a nw (561nm)

[0087]

[0088] Among them, g1 and g2 are empirical coefficients, with values ​​of 0.0895 and 0.1247 respectively; a w (561nm) is the absorption coefficient of pure water in the green band, which is 0.06236; b bw (561nm) is the backscattering coefficient of pure water in the green band, which is 0.001352; the values ​​of p0-p3 are: -1.15467, -1.17869, -0.24566, and -0.06989 respectively.

[0089] Step 4: Determine the inversion model of suspended particulate matter concentration in segmented water bodies;

[0090] The inversion model for suspended particulate matter concentration in low turbidity water is as follows:

[0091] log(TSM low )=α0+α1×log(b bp (561nm)

[0092] Among them, α0 and α1 are the coefficients to be fitted.

[0093] The inversion model for suspended particulate matter concentration in high turbidity water is as follows:

[0094]

[0095]

[0096]

[0097] Among them, β0, β1 and β2 are the coefficients to be fitted; R rs (655nm) and R rs (865nm) are the equivalent remote sensing reflectances of the red band and the near-infrared band respectively; w1 and w2 are weight coefficients;

[0098] Transition smoothing function, the calculation formula is as follows:

[0099] TMS med =μ×TSM high +(1-μ)×TSM low

[0100]

[0101] Among them, μ is the weight factor, R rs (655nm) segmentation threshold.

[0102] Step 5: Use the remote sensing reflectance R of the red band rs (655nm) is the segmentation object, and the coefficient R is determined 2 The root mean square error (RMSE) was used as the model accuracy evaluation index, and the least square method was used to calculate the optimal segmentation threshold of the low turbidity water body inversion algorithm and the high turbidity water body inversion algorithm. Make the accuracy of the entire inversion algorithm optimal (corresponding to the maximum R 2 , minimum RMSE).

[0103] like Adopting the inversion model of suspended particulate matter concentration in low turbidity water;

[0104] like Adopting the inversion model of suspended particulate matter concentration in low turbidity water;

[0105] like Adopt transition smoothing function;

[0106] In this embodiment, R is calculated rs (655nm)=0.023sr -1 is the optimal segmentation threshold ( Figure 4 ).

[0107] Step 6: Determine the final inversion model for suspended particulate matter concentration in water and evaluate the inversion model for suspended particulate matter concentration in water;

[0108] Using 216 sets of field-measured suspended particulate matter 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 above segmented model were obtained based on the least squares fitting method. The final suspended particulate matter concentration inversion algorithm was obtained, and the determination coefficient R was used. 2 The accuracy of the inversion algorithm is evaluated by the root mean square error (RMSE). Figure 5 ).

[0109] Step 7: Apply the above suspended particulate matter inversion algorithm to high-resolution satellite images, and use on-site synchronous measurement data to evaluate the accuracy of satellite inversion results, and finally complete the fine mapping of suspended particulate matter concentration in the Yellow River Estuary area ( Figure 6 ).

[0110] The above description is only used to help understand the method and core essence of the present invention, but the scope of protection of the present invention is not limited thereto. For those skilled in the art, equivalent replacements or modifications based on the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention should be included in the scope of protection of the present invention. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A remote sensing inversion method for suspended particulate matter in high-turbidity water bodies using high-resolution satellites, characterized by: The steps include: S1. Collect the on-site measured water suspended particulate matter concentration data and the corresponding water remote sensing reflectance data, and pre-process the water remote sensing reflectance data to obtain equivalent remote sensing reflectance data; S2. Constructing a segmented water suspended particle concentration inversion model; the segmented water suspended particle concentration inversion model includes a low turbidity water suspended particle concentration inversion model, a high turbidity water suspended particle concentration inversion model, and a transition smoothing function; calculating the backscattering coefficient of the green band based on the equivalent remote sensing reflectance data, and establishing a low turbidity water suspended particle concentration inversion model based on the backscattering coefficient of the green band; S3. Selection of segmented water suspended particle concentration inversion model; using the remote sensing reflectance R rs (red) is the segmentation object, and the coefficient R is determined 2 The root mean square error (RMSE) was used as the model accuracy evaluation index to calculate the optimal segmentation threshold of the inversion model of suspended particle concentration in low turbidity water body and the inversion model of suspended particle concentration in high turbidity water body. S4. Determine the final segmented water suspended particle concentration inversion model; using the water suspended particle concentration data and equivalent remote sensing reflectance data in S1, obtain the undetermined coefficients in the segmented water suspended particle concentration inversion model based on the least squares fitting method, obtain the final segmented water suspended particle concentration inversion model, and evaluate the accuracy of the segmented water suspended particle concentration inversion model; S5. Obtain the suspended particle concentration in water based on the segmented water suspended particle concentration inversion model; apply the segmented water suspended particle concentration inversion model to high-resolution satellite imagery, and use on-site synchronous measurement data to evaluate the accuracy of the inversion results of the segmented water suspended particle concentration inversion model, and finally complete the detailed mapping of the suspended particle concentration in the area of ​​interest.

2. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies suitable for high-resolution satellites according to claim 1 is characterized in that: In S1, the water body remote sensing reflectance data is preprocessed to obtain equivalent remote sensing reflectance data, specifically as follows: According to the spectral response function of each band of the high-resolution satellite sensor, the equivalent remote sensing reflectivity R of each band is simulated. rs (λ i ); The calculation formula is as follows: Where i is the band number of each satellite sensor; λ i is the corresponding central wavelength; R(λ i ) is the remote sensing reflectivity of water measured on site; S(λ i ) is the spectral response function; R rs (λ i ) is the integrated equivalent remote sensing reflectivity.

3. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies suitable for high-resolution satellites according to claim 2 is characterized in that: The segmented water suspended particulate matter concentration inversion model is constructed in S2, as follows: S201, calculating the backscattering coefficient of the green band; The calculation process is as follows: a(green)=a w (green)+a nw (green) Among them, g1 and g2 are empirical coefficients, a w (green) is the absorption coefficient of pure water in the green band; b bw (green) is the backscattering coefficient of pure water in the green band; p0-p3 are also empirical coefficients, b bp (green) is the backscattering coefficient of the green band; S202. Establish an inversion model for suspended particulate matter concentration in low turbidity water bodies; The inversion model for suspended particulate matter concentration in low turbidity water is as follows: Among them, α0 and α1 are the coefficients to be fitted; S203, establishing an inversion model for suspended particulate matter concentration in high turbidity water bodies; The inversion model for suspended particulate matter concentration in high turbidity water is as follows: Among them, β0, β1, and β2 are the coefficients to be fitted; R rs (red) and R rs (NIR) are the equivalent remote sensing reflectances of the red band and the near infrared band respectively; w1 and w2 are weight coefficients; S204, establishing a transition smoothing function for low-to-high turbidity water body transition; The calculation formula of the transition smoothing function is as follows: TMS med =μ×TSM high +(1-μ)×TSM low Among them, μ is the weight factor, R rs (red) segmentation threshold.

4. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies suitable for high-resolution satellites according to claim 3 is characterized in that: The empirical coefficients of the backscatter coefficients in the green band are as follows: g1 and g2 take values ​​of 0.0895 and 0.1247; The values ​​of p0-p3 are different for different high-resolution satellites, as follows: For Landsat-5TM, the values ​​are: -1.08523, -1.10967, -0.23418, -10249; For Landsat-7ETM+, the values ​​are: -1.14697, -1.15697, -0.26497, -0.08258; For Landsat-8 / 9 OLI, the values ​​are: -1.15467, -1.17869, -0.24566, and -0.06989 respectively.

5. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies suitable for high-resolution satellites according to claim 4 is characterized in that: The determination coefficient R in S3 2 And the root mean square error RMSE, as follows: Coefficient of determination R 2 The calculation formula is as follows: The calculation formula of root mean square error RMSE is as follows: Among them, y i is the concentration of suspended particulate matter in water measured on site; i ′ is the concentration of suspended particulate matter in water obtained by model inversion; is the average value of the suspended particulate matter concentration in water of all samples; m is the number of samples of suspended particulate matter concentration in water measured on site.

6. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites according to claim 5 is characterized in that: The selection of the inversion model for the suspended particulate matter concentration in the segmented water body in S3 is as follows: like Adopting the inversion model of suspended particulate matter concentration in low turbidity water; like The inversion model of suspended particulate matter concentration in highly turbid water bodies is used; like Adopt transition smoothing function.

7. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites according to claim 5 is characterized in that: The S4 was evaluated using the coefficient of determination R 2 The accuracy of the segmented water suspended particulate matter concentration inversion model was evaluated using the root mean square error (RMSE).

8. The remote sensing inversion method for suspended particulate matter in high-turbidity water bodies applicable to high-resolution satellites according to claim 3 is characterized in that: The undetermined coefficients in S4 include α0 and α1 in the inversion model of suspended particle concentration in low turbidity water bodies and β0, β1 and β2 in the inversion model of suspended particle concentration in high turbidity water bodies.

Images