Global river water quality remote sensing assessment method, apparatus and procedure based on water body color

Abstract

The present application relates to a global river water quality remote sensing evaluation method, device and program based on water body color, red, blue and green band remote sensing reflectance data of river section samples of global multiple rivers are acquired; the red, blue and green band remote sensing reflectance data of the representative area of each river section is converted to CIE color space, and the horizontal and vertical chroma coordinate values and the chroma angle are used as input variables; the river section samples are divided into three types of water bodies by using unsupervised clustering algorithm; the boundary function of three types of water bodies is fitted by using the chroma diagram of three types of water body samples respectively; the water body color intensity calculation formula is constructed based on the boundary function of three types of water bodies; the preset percentile value of the color intensity of each type of water body is calculated; the remote sensing reflectance data of the river section samples of any river are acquired, the water body type is judged by using the boundary function, the color intensity is calculated, and the water quality is evaluated based on the percentile value. The method of the present application can quickly and scientifically realize large regional scale water quality remote sensing evaluation.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing science and technology, spatiotemporal dynamic remote sensing monitoring and assessment of river water quality, and specifically to methods, equipment, and procedures for global river water quality remote sensing assessment based on water color. Primarily based on the high spatial resolution Landsat virtual satellite constellation remote sensing reflectance, it first performs color space conversion and constructs a rapid classification method for different types of river water bodies globally. Then, it establishes a method for calculating the color intensity of different water body types. Finally, based on the color intensity, it determines the good water quality threshold for different types of river water bodies and applies it on a global scale. Background Technology

[0002] River water is a vital material foundation for human production and daily life. However, under the influence of intense human activities and natural changes, global river water quality exhibits dramatic spatiotemporal variations, particularly in densely populated areas where water quality deteriorates, often manifesting as excessive nutrients, algal blooms, and black and odorous water bodies. To achieve remote sensing monitoring and assessment of the spatiotemporal dynamics of river water quality, numerous studies in recent years have reported remote sensing assessment methods for river water quality. These methods can be broadly categorized into three types: first, empirical remote sensing models based on the band ratios of satellite remote sensing reflectance to construct different water quality indicators such as nutrient, suspended solids, and chlorophyll concentrations; second, sensitive semi-analysis indices constructed based on satellite remote sensing reflectance in different bands, which are then used for remote sensing inversion of optically sensitive water quality indicators such as chlorophyll and suspended solids concentrations; and third, big data-driven remote sensing models based on machine / deep learning algorithms such as random forests, neural networks, and support vector machines to construct different water quality parameters.

[0003] Although many remote sensing methods for assessing river water quality have been reported, these methods are primarily designed for typical or regional rivers and are difficult to apply to global rivers. While the band ratio method is computationally simple, it lacks a physical basis and has poor regional scalability. Semi-analytical algorithms, although possessing some physical basis, require high accuracy in atmospheric correction of remote sensing reflectance; even small atmospheric correction errors can lead to significant errors in water quality assessment, making them unsuitable for rivers with low atmospheric correction accuracy.

[0004] Furthermore, existing technologies utilize chromaticity angles to indicate water quality parameters, but these studies primarily target specific water quality parameters and are difficult to apply to large-area river bodies containing a mixture of various water quality types. For example, existing technologies directly construct empirical relationships between chromaticity angles and water quality parameters such as chlorophyll a (Chl-a) and total suspended solids (TSM) concentrations, thereby enabling the estimation of water quality parameters using chromaticity angles. Such methods are suitable for water bodies dominated by specific water quality parameters, such as relatively clear water bodies dominated by Chl-a, but are difficult to apply to ultra-high turbidity water bodies, which often have large chromaticity angles but very low Chl-a concentrations. Summary of the Invention

[0005] The present invention aims to overcome the problems existing in the prior art and provide a method, equipment and program for remote sensing assessment of global river water quality based on water color.

[0006] To achieve the above objectives, the present invention adopts the following solution:

[0007] A global river water quality remote sensing assessment method based on water color, the method comprising:

[0008] Obtain remote sensing reflectance data in the red, blue, and green bands of river sections from multiple rivers worldwide;

[0009] The remote sensing reflectance data of the red, blue, and green bands of representative areas of each river section were converted to the CIE color space to obtain the horizontal and vertical chromaticity coordinates and chromaticity angles. , and chromaticity angle As input variables, an unsupervised clustering algorithm was used to classify the river section samples into three types of water bodies;

[0010] River section samples from three different water bodies Chromatography diagrams, fitting boundary functions for three types of water bodies;

[0011] The chromaticity angle ranges of the three types of water bodies are obtained based on the boundary functions of the three types of water bodies; a color intensity calculation formula is constructed for each type of water body, and the color intensity is characterized by the relative position of the chromaticity angle of the current water body within the chromaticity angle range;

[0012] A preset percentile value for the color intensity of each type of water body is calculated as a threshold for classifying water quality as good or bad.

[0013] Remote sensing reflectance data of river sections of any river are obtained, and after color space conversion, the water body type is determined by boundary function and its color intensity is calculated. Water quality is assessed based on the segmentation threshold.

[0014] In some embodiments of the present invention, the river segment sample is a representative area selected with the centroid of the river segment as the center;

[0015] The remote sensing reflectance data for the red, blue, and green light bands are the average values ​​of the remote sensing reflectance in the red, blue, and green light bands for the representative region.

[0016] In some embodiments of the present invention, the K-means method is used to perform unsupervised classification of the river segment samples based on the input variables.

[0017] In some embodiments of the present invention, the boundary functions of the three types of water bodies are determined based on the following method:

[0018] Obtain the CIE color space distribution of all samples from three types of water bodies;

[0019] Obtain the x-coordinate of each sample in the CIE color space The range of values ​​is divided into equal intervals. Given intervals, calculate the value in each interval. The ordinate at the lowest % value The calculated coordinate points are fitted to obtain the lower boundary functions for the three types of water bodies. ;

[0020] In the CIE color space, select water bodies that are classified as Class II and whose x-axis is [missing information]. Less than 1 / 3 of the samples, according to the sample The range of values ​​is divided into equal intervals. For each interval, calculate the Class II water body samples. The ordinate at the low value The values ​​are calculated, and the boundary function between classes I and II is obtained by fitting all coordinate points. ;

[0021] In the CIE color space, select water bodies that are classified as Class II and whose x-axis is [missing information]. More than 1 / 3 of the samples, according to sample The range of values ​​is divided into equal intervals. For each interval, calculate the Class II water body samples. The ordinate at the low value The values ​​are calculated, and the boundary function between classes II and III is obtained by fitting all coordinate points. .

[0022] In some embodiments of the present invention, the method further includes the step of repeatedly constructing the boundary function, in each interval The ordinate at the low value The value is the target for searching. The obtained CIE color space coordinates are compared with the previously obtained boundary function prediction results to evaluate the classification accuracy of different types of river water bodies; wherein, the... .

[0023] In some embodiments of the present invention, the color intensity of the water body is calculated based on the following formula for three types of water bodies:

[0024]

[0025]

[0026]

[0027] In the formula, , , , The chromaticity angles are the boundary functions corresponding to the three types of water bodies.

[0028] In some embodiments of the present invention, the remote sensing reflectance data is acquired based on Landsat satellite.

[0029] The present invention further provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0030] The present invention further provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0031] The present invention further provides a computer program product, including a computer program / instruction that, when executed by a processor, implements the steps of the above-described method.

[0032] The present invention has the following beneficial effects:

[0033] (1) Based on CIE color space conversion and unsupervised clustering results, a boundary function was constructed that can be used for rapid remote sensing identification of different types of rivers;

[0034] (2) Based on the classification of river water bodies, a method for calculating the color intensity of different types of river water bodies was constructed, and the color intensity is significantly related to the water quality; and the global river good water quality boundary line was adopted to provide the color intensity threshold for good water quality of different types of river water.

[0035] (3) The method of the present invention can quickly and scientifically realize water quality remote sensing assessment on a large regional scale (global scale). Attached Figure Description

[0036] Figure 1 The data represents the classification results and average remote sensing reflectance of three types of river water bodies; (a), (b), and (c) are cluster scatter plots of Class I, II, and III river water bodies; (d), (e), and (f) are the average remote sensing reflectance of Class I, II, and III river water bodies in different bands of the Landsat satellite; Blue: blue light band; Green: green light band; Red: red light band; NIR: near-infrared band; SWIR1: shortwave infrared band 1; SWIR2: shortwave infrared band 2.

[0037] Figure 2This represents the correlation between brown intensity and CDOM content retrieved from remote sensing of typical rivers worldwide. Here, a2-f2 are images of different typical rivers, and a1-f1 are the corresponding correlations between brown intensity and CDOM content.

[0038] Figure 3 This represents the correlation between green intensity and Chl-a content retrieved from remote sensing of typical rivers worldwide. Here, a2-f2 are images of different typical rivers, and a1-f1 are the corresponding correlations between green intensity and Chl-a content.

[0039] Figure 4 This represents the correlation between yellow intensity and TSM content in remote sensing inversion of typical rivers worldwide, where a2-f2 are images of different typical rivers; and a1-f1 are the corresponding correlations between yellow intensity and TSM content. Detailed Implementation

[0040] Taking different types of rivers worldwide as examples, this invention details a specific implementation method for global river water quality remote sensing assessment based on water color, with the main steps as follows:

[0041] (1) Global River Segment Sample Extraction. Based on the publicly available Global River Width (GRWL) dataset, this invention extracts the average water reflectance of 3×3 pixel windows from Landsat-5 / 7 / 8 / 9 series satellites from 2013 to 2023, centered on the centroid points of 15,743 river segments worldwide. The average water reflectance includes the average remote sensing reflectance of the blue, green, and red light bands of the water body.

[0042] (2) CIE Color Space Conversion. Based on the remote sensing reflectance of blue, green, and red light bands of global river samples, this invention converts the RGB color space to the CIE color space, making the color changes more consistent with human visual observation and corresponding to the Forel-Ule water classification table. The specific calculation formula is shown in equation (1):

[0043] (1)

[0044] In the formula, R blue R green and R red , respectively, are the Landsat satellite remote sensing reflectances in the blue, green, and red light bands; X, Y, and Z are the three color stimuli; x and y are the horizontal and vertical chromaticity coordinates in the CIE color space; a is the chromaticity angle in the CIE color space, with a value range of -180° to 180°.

[0045] (3) Unsupervised clustering of water body types.

[0046] Using Python's scikit-learn library, unsupervised classification is performed based on x, y, and a in the CIE color space, employing the K-means method.

[0047] The color of river water is determined by its dominant type into three categories: brown water dominated by dissolved organic matter, green water dominated by algae, and yellow water dominated by inorganic suspended matter. Therefore, the number of types in the K-means unsupervised clustering method is set to 3.

[0048] Using x, y, and a in the CIE color space as input parameters, all input samples are divided into three classes (Class I, Class II, and Class III), as shown in the following figure. Figure 1 As shown.

[0049] (4) Determination of the boundary lines of different types of water bodies.

[0050] Based on the K-means clustering results, this invention uses the following method to determine different types of boundary functions:

[0051] ① For the distribution of all samples in the CIE color space across the three categories, this invention divides the x-axis into 50 equal parts, calculates the y-axis value at the lowest 1% value in each part, and fits all points to obtain the lower boundary function y1 for the three types. Figure 1 a);

[0052] ② In the CIE color space, select samples that are classified as Class II in K-means clustering and whose x-coordinate is less than 1 / 3. Divide these samples into 50 equal parts based on the x-coordinate, calculate the y-coordinate of the lowest 1% value in each part, and fit all points to obtain the boundary function y2 between Class I and Class II. Figure 1 b);

[0053] ③ In the CIE color space, select samples that are classified as Class II in K-means clustering and whose x-coordinate is greater than 1 / 3. Divide these samples into 50 equal parts based on the x-coordinate, calculate the y-coordinate of the lowest 1% value in each part, and fit all points to obtain the boundary function y3 between Class II and Class III. Figure 1 c). The three boundary functions are as shown in equation (2):

[0054] (2)

[0055] In global remote sensing inversion, boundary functions can be used to determine the type of river water pixels, enabling rapid pixel classification and making it suitable for long-term global remote sensing. Furthermore, boundary functions are used to determine the chromaticity angle parameter in subsequent water color intensity measurements.

[0056] (5) Evaluation of classification accuracy for different types of water bodies. Step (4) was repeated, and the y-coordinate of the lowest value in each sample was used as the search target. The obtained (x, y) points in the CIE color space were then compared with the boundary function prediction results of Equation (2) to evaluate the classification accuracy of different types of river water bodies. The results showed that the boundary accuracy of the three boundary functions was as high as 0.93, 0.99 and 0.99, respectively, indicating the usability of the three boundary functions.

[0057] (6) Calculation of color intensity for different types of water bodies. K-means clustering results show that: the color of type I water bodies is mainly affected by dissolved organic matter, ranging from dark blue to brown; the color of type II water bodies is mainly affected by algal content, ranging from light green to dark green; and the color of type III water bodies is mainly affected by inorganic suspended matter, ranging from light yellow to dark yellow. Based on these distribution characteristics, this invention constructs calculation formulas for the color intensity of three different types of river water bodies, as shown in formula (3):

[0058]

[0059] (3)

[0060]

[0061] In the formula, 'a' is the chromaticity angle in the CIE color space, with a value range of -180° to 180°; in this embodiment... For the curve, take the intersection of the three boundary functions as the fixed point and construct... The tangent, which is in The chromaticity angle corresponding to the tangent in the direction is -141.96°. The chromaticity angle corresponding to the tangent in the direction is 75.49. The corresponding chromaticity angle is -20.71°; The corresponding chromaticity angle is 20.29°. That is, the chromaticity angle range of Type I water bodies is -141.96° to -20.71°, the chromaticity angle range of Type II water bodies is -20.71° to 20.29°, and the chromaticity angle range of Type III water bodies is 20.29° to 75.49°.

[0062] In this embodiment, remote sensing reflectance data of RGB bands were obtained from typical river sections around the world. After color conversion, water body type was determined based on equation (2), and color intensity was calculated using equation (3). The correlations between brown intensity and the content of colored dissolved organic matter (CDOM) retrieved from water remote sensing, green intensity and the concentration of Chl-a in water remote sensing, and yellow intensity and the concentration of TSM in water remote sensing were compared. The results showed that the correlation r between brown intensity and the CDOM content in water was between 0.62 and 0.84. Figure 2The correlation coefficient (r) between green intensity and Chl-a concentration was between 0.83 and 0.90. Figure 3 The correlation coefficient (r) between yellow intensity and TSM concentration was between 0.74 and 0.92. Figure 4 This indicates that the color intensity calculated by the method of the present invention well reflects the correlation with the dominant elements of the water body.

[0063] (7) Determination of good water quality thresholds for different types of water bodies. Based on the boundary function in step (4), this invention can obtain the number of pixels of different types of rivers in a 1° × 1° grid globally from 1984 to 2023 using Landsat series satellites, thereby obtaining the dominant water body type for different years. Furthermore, for the average color intensity of each type of river in a 1° grid globally from 1984 to 2023, the 80% dividing line is used as the good water quality threshold, thus obtaining the color intensity thresholds for good water quality for different types of rivers (Type I: brown intensity < 65.44%; Type II: green intensity < 45.57%; Type III: yellow intensity < 29.24%).

[0064] The 80% threshold is selected based on the current recommended values ​​for water quality assessment. Specifically, when there is a lack of corresponding water quality parameter thresholds for water quality assessment of a large area, it is recommended to use 80% of the corresponding water quality parameter as the threshold for dividing water quality into good and bad, that is, to consider 80% of the water bodies in a large area as having good water quality.

[0065] (8) Assessment of good water quality rate at different spatial scales. Based on the color intensity thresholds of different types of good water quality obtained in step (7), the proportion of rivers with good water quality in the 1° grid worldwide from 1984 to 2023 can be remotely assessed, and then the proportion of rivers with good water quality in different countries can be assessed.

Claims

1. A global river water quality remote sensing assessment method based on water color, characterized in that, The method includes: Obtain remote sensing reflectance data in the red, blue, and green bands of river sections from multiple rivers worldwide; The remote sensing reflectance data of the red, blue, and green bands of representative areas of each river section were converted to the CIE color space to obtain the horizontal and vertical chromaticity coordinates and chromaticity angles. , and chromaticity angle As input variables, an unsupervised clustering algorithm was used to classify the river section samples into three types of water bodies; River section samples from three different water bodies Chromatography diagrams, fitting boundary functions for three types of water bodies; The chromaticity angle ranges of the three types of water bodies are obtained based on the boundary functions of the three types of water bodies; a color intensity calculation formula is constructed for each type of water body, and the color intensity is characterized by the relative position of the chromaticity angle of the current water body within the chromaticity angle range; A preset percentile value for the color intensity of each type of water body is calculated as a threshold for classifying water quality as good or bad. Remote sensing reflectance data of river sections of any river are obtained, and after color space conversion, the water body type is determined by boundary function and its color intensity is calculated. Water quality is assessed based on the segmentation threshold.

2. The method according to claim 1, characterized in that, The river segment sample is a representative area selected with the centroid of the river segment as the center; The remote sensing reflectance data for the red, blue, and green light bands are the average values ​​of the remote sensing reflectance in the red, blue, and green light bands for the representative region.

3. The method according to claim 1, characterized in that, The K-means method is used to perform unsupervised classification of the river segment samples based on the input variables.

4. The method according to claim 1, characterized in that, The boundary functions for the three types of water bodies are determined based on the following method: Obtain the CIE color space distribution of all samples from three types of water bodies; Obtain the x-coordinate of each sample in the CIE color space The range of values ​​is divided into equal intervals. Given intervals, calculate the value in each interval. The ordinate at the lowest % value The calculated coordinate points are fitted to obtain the lower boundary functions for the three types of water bodies. ; In the CIE color space, select water bodies that are classified as Class II and whose x-axis is [missing information]. Less than 1 / 3 of the samples, according to the sample The range of values ​​is divided into equal intervals. For each interval, calculate the Class II water body samples. The ordinate at the low value The values ​​are calculated, and the boundary function between classes I and II is obtained by fitting all coordinate points. ; In the CIE color space, select water bodies that are classified as Class II and whose x-axis is [missing information]. More than 1 / 3 of the samples, according to sample The range of values ​​is divided into equal intervals. For each interval, calculate the Class II water body samples. The ordinate at the low value The values ​​are calculated, and the boundary function between classes II and III is obtained by fitting all coordinate points. .

5. The method according to claim 4, characterized in that, The method further includes repeating the boundary function construction step, in each interval The ordinate at the low value The value is the target for searching. The obtained CIE color space coordinates are compared with the previously obtained boundary function prediction results to evaluate the classification accuracy of different types of river water bodies; wherein, the... .

6. The method according to claim 1, characterized in that, For the three types of water bodies, the color intensity is calculated based on the following formula: ； ； ； In the formula, , , , The chromaticity angles are the boundary functions corresponding to the three types of water bodies.

7. The method according to claim 1, characterized in that, The remote sensing reflectance data was acquired based on the Landsat satellite.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images