A method for extracting nutrients from eutrophic water bodies based on Landsat 8

Abstract

This invention discloses a method for extracting eutrophic water bodies based on Landsat 8, comprising the following steps: using a water body as the research object, acquiring data sources within the research area; data preprocessing; determining the eutrophication level of Landsat 8 at different times using validation data (MODIS data); water body extraction and accuracy verification; calculating the Landsat 8 data from different times according to the water body indices MANDWI and ANDWI, respectively, and obtaining the water body extraction results using the Otsu method; through result comparison and random point accuracy verification, the results show that MANDWI has a good water body extraction effect for water bodies with different eutrophication levels. The modified enhanced normalized difference water body index MANDWI of this invention has a good clear water extraction effect and can extract eutrophic water bodies; achieving high-precision water body extraction and expanding the application of remote sensing data in shoreline extraction and water body identification.

Description

Technical Field

[0001] This invention relates to water extraction methods, specifically a method for extracting eutrophic water based on Landsat 8. Background Technology

[0002] Water surface area is a basic hydrological quantity describing changes in lake water conditions. Changes in water surface area reflect the dynamic changes of lakes. Extracting water surface area helps in the rational regulation, utilization and protection of water resources, and is helpful for river and lake ecological governance. It plays an important role in ecological environment construction and social development. Water surface area is also an important indicator for measuring the water storage capacity of reservoirs. Quickly and accurately extracting water surface area from satellite remote sensing images has become an important means of water resource surveys, macro-monitoring of water resources and wetland protection.

[0003] Water body extraction typically employs two methods: single-band thresholding and multi-band computation. Single-band thresholding introduces misleading information when extracting water bodies from areas like mountain shadows, reducing accuracy. Among multi-band computation algorithms, the most commonly used is the interspectral relation method, which extracts water body information by constructing water body indices through combined operations of different bands. This method offers advantages over single-band thresholding. Water body extraction using water body indices requires thresholding to classify the results. The choice of threshold directly affects classification error. Manual threshold selection is highly subjective, while adaptive thresholding algorithms, such as the Otsu method, can automatically obtain the optimal threshold.

[0004] Currently used water indices include the Normalized Difference Water Index (NDWI), the Modified Normalized Difference Water Index (MNDWI), the Automated Water Extraction Index (AWEI), and the Augmented Normalized Difference Water Index (ANDWI). Among these, the Augmented Normalized Difference Water Index (ANDWI), proposed in 2021, has achieved a very high level of accuracy in classification and water body identification. However, in some cases, ANDWI may classify areas with floating aquatic vegetation as non-water areas and may not be suitable for eutrophic water bodies.

[0005] Several studies have been conducted on the extraction of eutrophic water bodies, including extraction models for eutrophic and heavily polluted water bodies based on TM imagery and Landsat 8 imagery. These models are constructed based on the band characteristics and differences between shadows and eutrophic water bodies, thus avoiding the influence of shadows. However, judging water bodies based on the characteristics of only a few bands has an uncertain threshold, which may vary for different regions. Furthermore, the limited number of bands applied restricts the accuracy of the judgment. Summary of the Invention

[0006] Purpose of the invention: In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a high-precision method for extracting eutrophic water samples with a wide range of applicability.

[0007] Technical solution: The present invention provides a method for extracting nutrients from eutrophic water bodies based on Landsat 8, comprising the following steps:

[0008] Step 1, Data Preparation: Taking lakes and rivers as the research objects, and Landsat 8 as the data source;

[0009] Step 2, Data Preprocessing: Cropping the images to obtain image data for the study area;

[0010] Step 3, water surface area extraction: The Landsat 8 band is combined and calculated using the Modified Augmented Normalized Difference Water Index (MANDWI) to obtain the calculated result. Then, the water surface area is extracted by setting a threshold or using the Otsu method.

[0011] Furthermore, in step one, the data source is Landsat 8 imagery.

[0012] Furthermore, in step two, the study area is determined using remote sensing data processing software.

[0013] Furthermore, in step three, the mathematical expression for the modified enhanced normalized difference water index (MANDWI) is:

[0014]

[0015] Wherein, B is the reflectance of the blue band in Landsat 8, G is the reflectance of the green band in Landsat 8, R is the reflectance of the red band in Landsat 8, NIR is the reflectance of the near-infrared band in Landsat 8, SWIR1 is the reflectance of the shortwave infrared 1 band in Landsat 8, and SWIR2 is the reflectance of the shortwave infrared 2 band in Landsat 8.

[0016] The above-mentioned method for extracting eutrophic water bodies also includes step four, which involves visually interpreting false-color images to determine the verification point, comparing the extraction results of each water body extraction method with the reference value, establishing a confusion matrix between the two values, and calculating the following accuracy evaluation indicators.

[0017] Furthermore, the accuracy evaluation indicators include misclassification error (CE), omission error (OE), overall accuracy (OA), and Kappa coefficient.

[0018] Furthermore, the formula for calculating the misclassification error CE is:

[0019]

[0020] In this context, FP represents a sample whose true class is negative, but the model predicts it as positive; TP represents a sample whose true class is positive, and the model also predicts it as positive.

[0021] Furthermore, the formula for calculating the omission error (OE) is as follows:

[0022]

[0023] Where FN represents a sample whose true class is positive, but the model predicts it as negative; TP represents a sample whose true class is positive, and the model also predicts it as positive.

[0024] Furthermore, the formula for calculating the overall accuracy OA is:

[0025]

[0026] Wherein, TP means the true class of the sample is positive and the model predicts it to be positive; TN means the true class of the sample is negative and the model predicts it to be negative; FN means the true class of the sample is positive but the model predicts it to be negative; and FP means the true class of the sample is negative but the model predicts it to be positive.

[0027] Furthermore, the formula for calculating the Kappa coefficient is as follows:

[0028]

[0029] Where P0 is the sum of the number of correctly classified samples in each class divided by the total number of samples, i.e., the overall precision OA; P e Multiply the actual quantity of each class by the sum of the predicted quantities of that class, and divide by the square of the total number of all classes.

[0030] Beneficial effects: Compared with existing technologies, this invention has the following significant features: The method for acquiring water body distribution over a large area through remote sensing imagery does not target specific water samples for testing. It constructs a modified enhanced normalized difference water index (MANDWI), which not only has good clean water extraction performance but can also extract eutrophic water bodies. A new water body extraction model is proposed, solving the problem of eutrophic water body extraction and addressing the issue that conventional water indices easily classify eutrophic water bodies as non-water bodies. This achieves high-precision water body extraction and has good applicability in lake surface information acquisition, expanding the application of remote sensing data in shoreline extraction, water body identification, and other fields. Attached Figure Description

[0031] Figure 1 This is a research technology roadmap for the present invention;

[0032] Figure 2 This invention provides a comparison of extraction results from Taihu Lake water at different times and with different levels of cyanobacterial blooms.

[0033] Figure 3 This is the distribution of verification sample points for the present invention. Detailed Implementation

[0034] like Figure 1 A method for extracting nutrients from eutrophic water bodies based on Landsat 8 includes the following steps:

[0035] Step 1: Data Preparation. Taking Taihu Lake as the research object, remote sensing data within the study area was acquired, including Landsat 8 imagery and MODIS data. The Landsat 8 data from August 26, 2021, contained clouds, and the cloud-free data was downloaded using Google Earth Engine, as shown in Table 1 below.

[0036] Table 1 Research Data

[0037]

[0038] Step 2, Data Preprocessing. MODIS data bands are extracted and reprojected using the MODIS data processing tool MRT. To reduce computational load, all images are cropped to obtain images covering the study area.

[0039] Step 3, Data Validation. To verify the applicability of MANDWI for extracting data from Taihu Lake with different eutrophication levels, the experiment used chlorophyll a concentration and eutrophic water area obtained from MODIS data inversion as grading standards. Four Landsat 8 images with different degrees of algal bloom severity were selected, with image dates of April 28, 2018, August 26, 2021, November 3, 2017, and May 11, 2017.

[0040] Step 4: Water body extraction and accuracy verification. The modified enhanced normalized difference water index (MANDWI) is calculated using the formula to obtain the results after calculating the data from four periods. Then, the water body extraction results for the four periods are obtained by setting a threshold or using the Otsu method. The mathematical expression for the modified enhanced normalized difference water index (MANDWI) is:

[0041]

[0042] Wherein, B is the reflectance of the blue band in Landsat 8, G is the reflectance of the green band in Landsat 8, R is the reflectance of the red band in Landsat 8, NIR is the reflectance of the near-infrared band in Landsat 8, SWIR1 is the reflectance of the shortwave infrared 1 band in Landsat 8, and SWIR2 is the reflectance of the shortwave infrared 2 band in Landsat 8.

[0043] For the Landsat 8 imagery of Taihu Lake, the water surface area was extracted using the Otsu method based on MANDWI, and the water body extraction results of the ANDWI Otsu method were used for comparison and verification. Taihu Lake is located on the southern edge of the Yangtze River Delta, in southern Jiangsu Province. It is one of the five largest freshwater lakes, ranking third, and is situated between 30°55'40"~31°32'58"N and 119°52'32"~120°36'10"E, with an area of ​​approximately 2427.8 square kilometers. The water body extraction results of MANDWI were compared with those of ANDWI, and the results are as follows: Figure 2 As shown. From Figure 2 As can be seen, for water bodies at different times and with different degrees of algal blooms, the main difference between the binary images of MANDWI and ANDWI lies in the framed area. Combined with the false-color image, this part is mainly eutrophic water. Therefore, MANDWI has a better extraction effect on eutrophic water, while ANDWI has a poorer extraction effect on this part of the water.

[0044] To verify the accuracy of MANDWI's Taihu Lake water body extraction, this paper conducts random point accuracy verification on the water body extraction results of the above four phases. 600 verification points were randomly generated in the Taihu Lake area, and the point distribution map is shown below. Figure 3As shown, each verification point was judged through visual interpretation of false-color images. The extraction results of each water body extraction method were compared with the reference values, and a confusion matrix between the two values ​​was established. The confusion matrix method compares and calculates the categories of selected sample points with the corresponding categories in the image classification results, and is a commonly used method in remote sensing land cover classification and evaluation. The confusion matrix is ​​shown in Table 2 below.

[0045] Table 2 Confusion Matrix

[0046]

[0047] The following accuracy evaluation metrics are calculated using the confusion matrix: Commission Error (CE), Omission Error (OE), Overall Accuracy (OA), and Kappa coefficient.

[0048] (1) Commission Error (CE)

[0049] This refers to the probability that a certain type in the classification result is inconsistent with the type of the reference image.

[0050]

[0051] (2) Omission Error (OE)

[0052] This refers to the probability that a certain type in a reference image will be classified as another category.

[0053]

[0054] (3) Overall Accuracy (OA)

[0055] Overall accuracy refers to the percentage of all correctly classified samples out of the total number of samples. It is an important indicator of the quality of land cover classification and its value ranges from 0 to 1.

[0056]

[0057] (4) Kappa coefficient

[0058] The Kappa coefficient is an important indicator of the consistency of land cover classification. It is generally believed that a value of 0.8 or higher indicates a good classification effect.

[0059]

[0060] Where P0 is the sum of the number of correctly classified samples in each class divided by the total number of samples, which is the overall accuracy (OA); P e The sum of the actual quantity in each category multiplied by the predicted quantity in that category is divided by the square of the total number of all categories, as follows:

[0061]

[0062] In the above formula, TP (True Positive): The true class of the sample is positive, and the model predicts that it is also positive; FN (False Negative): The true class of the sample is positive, but the model predicts it as negative; FP (False Positive): The true class of the sample is negative, but the model predicts it as positive; TN (True Negative): The true class of the sample is negative, and the model predicts it as negative.

[0063] Table 3. Accuracy Evaluation of Water Extraction Results

[0064]

[0065] Based on the above calculations of accuracy indicators, the accuracy results are shown in Table 3 above. Statistical results show that as the severity of algal blooms increases, the overall accuracy (OA) and Kappa coefficient of MANDWI remain close to 1, while the overall classification accuracy (OA) of ANDWI decreases from 0.97 to 0.76, and the Kappa coefficient decreases from 0.89 to 0.46. The misclassification error (CE) of MANDWI is close to 0, and the misclassification error (CE) of ANDWI is also 0. As the severity of algal blooms increases, the omission error of MANDWI approaches 0, while the omission error of ANDWI gradually increases from 0.04 to 0.29. Therefore, MANDWI exhibits good accuracy in water extraction from Taihu Lake.

Claims

1. A method for extracting nutrients from eutrophic water bodies based on Landsat 8, characterized in that, Includes the following steps: Step 1: Focusing on lakes and rivers, and using Landsat 8 imagery as the data source; Step 2: Crop the Landsat 8 imagery to obtain image data for the study area. Step 3: Combine the Landsat 8 bands according to the modified enhanced normalized differential water index MANDWI to obtain the calculated results, and then obtain the water extraction results by setting a threshold or the Otsu method. In step three, the mathematical expression for the modified enhanced normalized difference water index (MANDWI) is: Wherein, B is the reflectance of the blue band in Landsat 8, G is the reflectance of the green band in Landsat 8, R is the reflectance of the red band in Landsat 8, SWIR1 is the reflectance of the shortwave infrared 1 band in Landsat 8, and SWIR2 is the reflectance of the shortwave infrared 2 band in Landsat 8.

2. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 1, characterized in that: In step two, the cropping is performed by obtaining the study area range using remote sensing data processing software.

3. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 1, characterized in that: The process also includes step four, which involves visually interpreting false-color images to determine the accuracy of each verification point, comparing the extraction results of each water body extraction method with the reference values, establishing a confusion matrix between the two values, and calculating the accuracy evaluation index.

4. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 3, characterized in that: The accuracy evaluation indicators include misclassification error (CE), omission error (OE), overall accuracy (OA), and Kappa coefficient.

5. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 4, characterized in that: The formula for calculating the misclassification error CE is as follows: In this context, FP represents a sample whose true class is negative, but the model predicts it as positive; TP represents a sample whose true class is positive, and the model also predicts it as positive.

6. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 4, characterized in that: The formula for calculating the omission error (OE) is as follows: In this context, FN represents a sample whose true class is positive, but the model predicts it as negative; TP represents a sample whose true class is positive, and the model also predicts it as positive.

7. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 4, characterized in that: The formula for calculating the overall accuracy OA is: Wherein, TP means the true class of the sample is positive and the model predicts it to be positive; TN means the true class of the sample is negative and the model predicts it to be negative; FN means the true class of the sample is positive but the model predicts it to be negative; and FP means the true class of the sample is negative but the model predicts it to be positive.

8. The method for extracting nutrients from eutrophic water bodies based on Landsat 8 according to claim 4, characterized in that: The formula for calculating the Kappa coefficient is as follows: Where P0 is the sum of the number of correctly classified samples in each class divided by the total number of samples, i.e., the overall precision OA; P e Multiply the actual quantity of each class by the sum of the predicted quantities of that class, and divide by the square of the total number of all classes.

Images