Remote sensing monitoring method for different life forms of aquatic plants

Abstract

The present application relates to remote sensing monitoring method of different life forms of aquatic plants, comprising: based on field measurement data, using high spatial resolution satellite data for remote sensing monitoring of aquatic plants, using the remote sensing monitoring result verified by precision as modeling true value data, and putting into modeling true value data set; selecting Landsat series data full-band remote sensing reflectivity data and MODIS Aqua b1-b7 band remote sensing reflectivity data as modeling input parameter, modeling true value data set as output, and training convolutional neural network model; acquiring Landsat and MODIS data of the water body to be measured, inputting the convolutional neural network model, and obtaining the distribution of different life forms of aquatic plants in the time to be measured. Based on the method, the interannual and intermonthly variation law and spatial distribution of different life forms of aquatic plants since 2000 can be accurately obtained.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, and in particular to remote sensing monitoring methods for aquatic plants of different life forms. Background Technology

[0002] Macrophytes are a group within the ecological framework, encompassing all aquatic plant groups except for small algae. They are higher plant groups that are partially or entirely submerged in water or floating on the water surface for several months of the year. This includes aquatic groups of seed plants, ferns, and bryophytes, as well as macrophytes with rhizoids. Macrophytes represent a phenotype of convergent adaptation formed by different taxa of plants over a long period of adaptation to the aquatic environment (Hu Jinliang, 2012). The concept of aquatic plants in this study refers only to macrophytes and does not include small algae.

[0003] Aquatic plants and phytoplankton are important primary producers in lake ecosystems, providing water, habitats, and food for fish, waterbirds, and other organisms (Hugheset et al., 2009; Peng et al., 2004). Aquatic plants play a vital role in aquatic ecosystems. For example, in terms of water quality, aquatic plants can effectively enrich pollutants, reducing the pollutant load in rivers (Shinkareva et al., 2019), reduce turbidity, and improve water transparency (Hestiret et al., 2015), while also promoting their own growth (Scheffer, 1999). In terms of ecology, aquatic plants have the function of storing carbon, enabling them to regulate climate change and serving as important ecosystem indicators (McLeodet et al., 2011). In terms of species, aquatic plants can inhibit phytoplankton growth, providing important habitats for zooplankton and fish, and playing a role in protecting biodiversity (Ozimeket et al., 1990; Takamura et al., 2003). However, excessive growth of aquatic plants can have many negative effects. For example, overly dense aquatic vegetation can lead to losses in biodiversity, ecosystem services, and economic activity (Jia et al., 2016), and when plant density exceeds a certain threshold, planktonic fish may avoid the plants (Jeppesene et al., 1998). Control measures are typically implemented to maintain a dynamic balance of aquatic plants and ensure the long-term stability of aquatic ecosystems (Gao et al., 2019). Therefore, the area, spatial distribution, structure, and composition of aquatic plants all have varying degrees of impact on aquatic ecosystems. Accurately understanding the spatial distribution of aquatic plants in aquatic ecosystems is crucial for ensuring their stability and for scientifically understanding the characteristics and potential for carbon sequestration in inland water bodies.

[0004] Based on their habitat, aquatic plants can be divided into emergent macrophytes, free-floating macrophytes, and submerged macrophytes (Kalff, 2011). Emergent macrophytes are those that take root in the mud at the bottom of the water and whose leaves grow above the water surface; free-floating macrophytes are those whose bodies are suspended on the water or whose leaves float on the water surface; and submerged macrophytes are those whose roots are in the mud at the bottom of the water or whose bodies are submerged in the water (Zheng Chong and Wang Hongyan, 2009). There are many differences among aquatic plants of different life forms. In terms of distribution, emergent plants are found from the shore to water depths of 1-2 meters, while submerged plants can reach deeper waters. Submerged plants can grow as long as the light intensity reaches 2%-10% of the incident light (PAR) (Kalff, 2011). In terms of environmental function, emergent plants have the best reoxygenation performance (Duan et al., 2002). Floating-leaved plants have a stronger ability to remove nitrogen from water and inhibit cyanobacteria than submerged plants, while submerged plants have a greater water purification capacity and release more gases into the water than floating-leaved plants (Deng et al., 2018; Tong et al., 2004; Wang et al., 2009; Zhang et al., 2015). In terms of carbon sequestration, submerged plants have low photosynthetic efficiency, and the C and N content in their leaves is lower than that of emergent and floating-leaved plants. However, due to the δ¹⁸O⁻ of inorganic carbon in the water… 13 C is higher than that in air, and its leaves contain δ 13 C is actually higher than that of emergent plants and floating-leaved plants (Huang Liang et al., 2003).

[0005] Traditional methods for monitoring aquatic plants involve field surveys, laboratory analysis, and literature reviews to assess their spatiotemporal distribution. These methods are time-consuming and highly susceptible to environmental conditions, making them unsuitable for large-scale, long-term monitoring (Lu, 1984; Meng, 1979). Remote sensing technology, compared to traditional methods, offers advantages such as high timeliness, wide monitoring range, low cost, and comprehensive data comparability. It can obtain ground information on a large area of ​​aquatic plants in a short period, trace long-term growth and distribution, and analyze the spatiotemporal changes of different life forms. Therefore, with the rapid development of remote sensing technology, combining it with field quadrat monitoring has become a crucial method for aquatic plant monitoring.

[0006] Therefore, monitoring different life forms of aquatic plants based on remote sensing technology helps to understand the spatiotemporal evolution process and patterns of aquatic plants, explore their main environmental driving forces, and at the same time, aquatic plants are also important carbon sources and sinks. Accurate monitoring of the distribution of different life forms of aquatic plants will also provide important support for accurately calculating the carbon sources and sinks of aquatic ecosystems. Summary of the Invention

[0007] The purpose of this invention is to provide a remote sensing monitoring method for aquatic plants with different life forms by utilizing multi-source satellite data and machine learning algorithms.

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

[0009] Remote sensing monitoring methods for different life forms of aquatic plants include:

[0010] Based on field measurement data, aquatic plants are remotely monitored using high spatial resolution satellite data. The remote sensing monitoring results that have passed accuracy verification are used as the modeling ground truth data and put into the modeling ground truth dataset.

[0011] Landsat series full-band remote sensing reflectance data and MODIS Aqua b1-b7 band remote sensing reflectance data were selected as modeling input parameters, and the modeling ground value dataset was used as output to train the convolutional neural network model.

[0012] Landsat and MODIS data of the water body to be tested are obtained and input into the convolutional neural network model to obtain the distribution of different life forms of aquatic plants at the time of testing.

[0013] As a preferred embodiment, the high spatial resolution satellite data refers to remote sensing data with a resolution of less than 10m.

[0014] As a preferred implementation, the high spatial resolution satellite data used is Sentinel-2 MSI data. Sentinel-2 MSI data has a large data volume and is completely free and open, which can reduce application costs.

[0015] In a preferred embodiment, the high spatial resolution satellite data, Landsat series full-band remote sensing reflectance data, and MODIS Aqua b1-b7 band remote sensing reflectance data are radiometrically calibrated, geometrically corrected, and precisely atmospherically corrected data.

[0016] As a preferred implementation, the remote sensing reflectance data of the MODIS Aqua b1-b7 bands are resampled to 30m and used together with all bands of the Landsat series data as input parameters.

[0017] As a preferred implementation, the method for obtaining the modeling ground truth data is to conduct remote sensing monitoring of aquatic plants using the maximum likelihood estimation method based on high spatial resolution satellite data, and to verify the results using field measurement data. The verified data is then used as the ground truth for training the model.

[0018] As a preferred implementation method, satellite data within one week before and after the survey time corresponding to the field measurement data are selected for remote sensing monitoring of aquatic plants, ignoring the changes in aquatic plants within the time difference range.

[0019] In a preferred embodiment, the loss function of the convolutional neural network model is the cross-entropy loss function, and the optimizer is the Adam optimization algorithm.

[0020] In a preferred embodiment, the parameters of the convolutional neural network model are set to iteration number = 100 and learning rate = 0.01.

[0021] As a preferred implementation, the trained convolutional neural network model is embedded into the GEE cloud computing platform. Based on the remote sensing reflectance data of Landsat and MODIS on the cloud platform, after cloud removal, maximum water boundary clipping, and water body boundary extraction, the model is run to obtain the spatial distribution results of aquatic plants in different life forms within the study area during the monitoring time.

[0022] The present invention has the following beneficial effects:

[0023] (1) Selecting high spatial resolution satellite data and combining it with measured data to jointly construct a modeling ground truth dataset. Existing technologies usually use satellite-ground synchronous data for model training, that is, measured data as output and remote sensing data as input, so as to train the remote sensing inversion model. However, the amount of measured data needs to be collected manually, and collecting a large amount of data is extremely labor-intensive. Machine learning usually requires a large number of samples for training to obtain a more accurate model. This application uses high spatial resolution satellite data and combined with measured data to jointly construct a modeling ground truth dataset, which makes up for the lack of data volume of satellite-ground synchronous data.

[0024] (2) Using multi-source satellite remote sensing data, the same model is used to realize the remote sensing monitoring method of different life forms of aquatic plants. The multi-source satellite remote sensing data selected are Landsat series data and MODIS Aqua B1~B7 band data. Landsat series satellite data is a wide band data, which is limited in its ability to capture the spectral differences of different life forms of aquatic plants. MODIS Aqua data has been available every day since 2001. The remote sensing reflectance of all bands of Landsat series data and MODIS Aqua b1-b7 bands are used as the modeling input parameters. Combined with the modeling ground value dataset, the model is trained using the convolutional neural network (CNN) machine learning algorithm. This can accurately obtain the spatiotemporal distribution of aquatic vegetation in lakes and reservoirs, accurately assess the spatiotemporal change trend of aquatic plants, correctly grasp the situation of lake and reservoir ecosystems, improve prediction accuracy, and provide important scientific and technological support for water environment management and decision-making of eutrophic lakes.

[0025] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below may be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other. Furthermore, all combinations of the claimed subject matter are considered part of the inventive subject matter of this disclosure.

[0026] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0027] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein:

[0028] Figure 1 It is a major data source for remote sensing identification of aquatic plants, based on literature statistics.

[0029] Figure 2 It is a modeling strategy for remote sensing monitoring of different aquatic plant life forms based on multi-source satellite data and machine learning algorithms.

[0030] Figure 3 This is the result of remote sensing identification of aquatic plants of different life forms based on Sentinel-2 MSI data.

[0031] Figure 4 These are remote sensing monitoring results of different aquatic plant life forms based on the method of this invention.

[0032] In the aforementioned Figures 1-4, the coordinates, symbols, or other representations expressed in English are all well-known in the field and will not be elaborated upon in this example. Detailed Implementation

[0033] To better understand the technical content of the present invention, specific embodiments are described below in conjunction with the accompanying drawings.

[0034] Various aspects of the invention are described in this disclosure with reference to the accompanying drawings, in which numerous illustrative embodiments are shown. The embodiments of this disclosure are not necessarily intended to encompass all aspects of the invention. It should be understood that the various concepts and embodiments described above, as well as those described in more detail below, can be implemented in any of many ways, as the concepts and embodiments disclosed herein are not limited to any particular implementation. Furthermore, some aspects of the invention disclosed may be used alone or in any suitable combination with other aspects of the invention disclosed.

[0035] This embodiment uses Taihu Lake as an example to further describe the method of the present invention.

[0036] Step 1: Selecting satellite remote sensing data;

[0037] The selection of satellite remote sensing data takes into account the temporal, spatial, and spectral resolution of existing satellite data, as well as the available time limit of the data. Sentinel-2 MSI data (10m; 2016-), Landsat series data (30m; 1970s-), and MODIS Aqua data (250 / 500 / 1000m; 2001-) are selected as the basic remote sensing data for the construction method.

[0038] Figure 1 This is a statistical analysis of data sources for remote sensing monitoring of aquatic plants, with Sentienl-2 MSI data and Landsat series data being the most commonly used. Based on the field survey time, remote sensing data within one week before and after the survey were selected as synchronous data, ignoring changes in aquatic plants within the time frame.

[0039] Step 2, Modeling Strategy;

[0040] The modeling strategy specifically refers to using the identification results of high spatial resolution satellite data as the source of modeling data, and integrating the remote sensing reflectance of all bands of Landsat series data and MODIS Aqua data b1-b7 bands as modeling input parameters to jointly construct a remote sensing monitoring method for different life forms of aquatic plants. Figure 2 This is a technical flowchart of the modeling strategy. The key challenges are the requirements of machine learning methods regarding the number of modeling samples, and the fact that Landsat series satellite data, being broad-band data, has limited ability to capture spectral differences among aquatic plants of different life forms. Since MODIS Aqua data has been available daily since 2001, the remote sensing reflectance of all bands from the Landsat series data and MODIS Aqua's b1-b7 bands is used together as the modeling input parameters.

[0041] Step 3: Constructing the modeling truth dataset;

[0042] The modeling ground truth dataset is based on field measurement data and uses Sentinel-2 MSI data for remote sensing monitoring of aquatic plants. After verification, the results are used as the modeling ground truth dataset. The amount of data in this dataset can ensure the stability and recognition accuracy of the machine learning algorithm.

[0043] In this embodiment, based on Sentinel-2 MSI data, the maximum likelihood method is used to conduct remote sensing identification of aquatic plants with different life forms. After verifying the accuracy of the field survey results, the ground truth dataset for the next step of modeling is constructed. Figure 3 This is an example of the method.

[0044] Step 4: Construct remote sensing monitoring methods for different types and life forms of aquatic plants based on machine learning algorithms;

[0045] By spatially resampling MODIS Aqua data bands b1-b7 to 30 m and using them together with all bands of Landsat series data as input parameters, a remote sensing monitoring method for different life forms of aquatic plants was constructed. Based on this method, the interannual and monthly variation patterns and spatial distribution of different life forms of aquatic plants since 2000 can be accurately obtained.

[0046] according to Figure 2 The modeling and decision-making process can realize remote sensing monitoring of different aquatic plants with different life forms in the whole image.

[0047] The specific process is as follows:

[0048] ① Obtain surface reflectance data from Sentinel-2 MSI, Landsat series, and MODIS Aqua data;

[0049] ②Based on Sentinel-2 MSI data, a high spatial resolution remote sensing identification method for aquatic plants with different life forms is constructed using the maximum likelihood method;

[0050] ③ Based on the results of field surveys, the accuracy of remote sensing identification methods for aquatic plants with different life forms with high spatial resolution was verified, and a ground truth database was constructed;

[0051] ④ Resample the MODIS Aqua data to 30m, select bands b1-b7 and the entire Landsat series data as input parameters, and train a convolutional neural network model based on the ground truth database. The basic parameters of the convolutional neural network (CNN) are: number of iterations = 100, loss function: cross-entropy loss, optimizer: Adam algorithm, learning rate = 0.01. After accuracy verification, the trained convolutional neural network model is obtained.

[0052] ⑤ The trained convolutional neural network model is embedded into the GEE cloud computing platform. Based on the Landsat and MODIS remote sensing reflectance data on the cloud platform, and after preprocessing such as cloud removal, maximum water boundary clipping, and water body boundary extraction, the model is run to obtain the spatial distribution results of aquatic plants in different life forms within the study area during the monitoring period. Figure 4 By repeating this step, the spatiotemporal variation patterns of different life forms of aquatic plants in the study area over a long period of time can be obtained, and the influence of environmental driving forces can be analyzed.

[0053] The above methods can achieve precise synchronization of the spatiotemporal distribution of aquatic vegetation of different life forms, accurately assess the extent of aquatic plant zones, correctly grasp the situation of lake and reservoir ecosystems, improve prediction accuracy, and provide important scientific and technological support for water environment management and decision-making in eutrophic lakes.

[0054] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art to which this invention pertains can make various modifications and refinements without departing from the spirit and scope of the invention.

Claims

1. A remote sensing monitoring method for aquatic plants of different life forms, characterized in that, include: Based on field measurement data, aquatic plants are remotely monitored using high spatial resolution satellite data. The remote sensing monitoring results that have passed accuracy verification are used as the modeling ground truth data and put into the modeling ground truth dataset. The method for obtaining the true data for modeling is to use the maximum likelihood estimation method based on high spatial resolution satellite data to conduct remote sensing monitoring of aquatic plants, and to verify the results using field measurement data. The verified data is then used as the true value for training the model. Landsat series full-band remote sensing reflectance data and MODIS Aqua b1-b7 band remote sensing reflectance data were selected as modeling input parameters, and the modeling ground value dataset was used as output to train the convolutional neural network model. Landsat and MODIS data of the water body to be tested are obtained and input into the convolutional neural network model to obtain the distribution of different life forms of aquatic plants at the time of the test.

2. The method according to claim 1, characterized in that, The high spatial resolution satellite data refers to remote sensing data with a resolution of less than 10m.

3. The method according to claim 1 or 2, characterized in that, The high spatial resolution satellite data used is Sentinel-2 MSI data.

4. The method according to claim 1, characterized in that, The high spatial resolution satellite data, Landsat series full-band remote sensing reflectance data, and MODIS Aqua b1-b7 band remote sensing reflectance data are radiometrically calibrated, geometrically corrected, and precisely atmospherically corrected.

5. The method according to claim 1, characterized in that, The remote sensing reflectance data of the MODIS Aqua b1-b7 bands were resampled to 30m and used together with all bands of the Landsat series data as input parameters.

6. The method according to claim 1, characterized in that, Based on the survey time corresponding to the field measurement data, satellite data within one week before and after the survey were selected for remote sensing monitoring of aquatic plants, ignoring the changes in aquatic plants within the time difference range.

7. The method according to claim 1, characterized in that, The loss function of the convolutional neural network model is the cross-entropy loss function, and the optimizer is the Adam optimization algorithm.

8. The method according to claim 1 or 7, characterized in that, The parameters of the convolutional neural network model are set to iteration number = 100 and learning rate = 0.

01.

9. The method according to claim 1, characterized in that, The trained convolutional neural network model was embedded into the GEE cloud computing platform. Based on the remote sensing reflectance data of Landsat and MODIS on the cloud platform, after cloud removal, maximum water boundary clipping, and water body boundary extraction, the model was run to obtain the spatial distribution results of aquatic plants in different life forms within the study area during the monitoring time.

Images