Lake algae vertical distribution daily change monitoring method based on satellite remote sensing image

Abstract

The invention discloses a lake algae vertical distribution daily change monitoring method based on a satellite remote sensing image. The invention relates to the technical field of environmental engineering and image processing, and solves the problem that vertical distribution daily variation of algae is difficult to obtain through single observation. The method comprises the following steps: collecting data of algae samples at different depths of a lake, establishing satellite-ground synchronous data, establishing a vertically distributed mathematical model, and parameterizing distribution characteristics of algae in a water depth direction; the remote sensing image is preprocessed, and an inversion model is established based on the relation between the surface chlorophyll concentration in the in-situ data and the remote sensing reflectivity. Utilizing machine learning to optimize model parameters, and combining with a vertical distribution parameterization result to construct an algae vertical distribution remote sensing monitoring model; the method integrates remote sensing data to invert lake algae vertical distribution, performs time sequence analysis on data obtained by inversion, estimates algae vertical distribution and spatial change conditions of a target water area, and provides important scientific basis and technical support for algae bloom prevention and control and environmental protection.

Description

Technical Field

[0001] The present invention relates to the field of environmental engineering and image processing technology, and in particular to a method for monitoring diurnal changes in the vertical distribution of lake algae based on satellite remote sensing images. Background Art

[0002] Eutrophication in lakes and reservoirs leads to rapid algal growth, exacerbating algal blooms and posing a serious threat to the health of lake ecosystems and the safety of drinking water supplies. Therefore, understanding the spatiotemporal dynamics of lake algae is crucial for mitigating water safety risks. Lake algal blooms fluctuate rapidly, appearing and disappearing, and can change rapidly within a single day or even hours. Algal blooms occur when large algae accumulate under certain meteorological conditions, leading to the rise and accumulation of algae. Monitoring only surface blooms can lead to misjudgment of algal conditions. However, combining the different vertical distribution patterns of algae in the water column can more accurately capture the spatiotemporal dynamics of algae. Therefore, vertical algal distribution patterns are a key parameter for accurately estimating algal biomass and assessing lake algal bloom risk. Traditional monitoring methods require collecting water samples at different depths to determine vertical algal distribution, which is labor-intensive and financially costly, and lacks sufficient spatiotemporal coverage. Satellite remote sensing, with its long time series and large-scale coverage, is currently a key tool for monitoring lake algal dynamics and other ecological environments. Using remote sensing to characterize the vertical distribution of lake algae has important research significance and practical value.

[0003] Current methods often use remote sensing to estimate the vertical distribution of algae by constructing functional relationships with parameters of the vertical distribution structure of algae, often focusing on single daily observations such as those from polar-orbiting satellites like MODIS and OLCI. However, due to the influence of meteorological and environmental conditions, as well as factors inherent to the algae particles themselves, algae particles dispersed in the water column rise, aggregate, and migrate under favorable external conditions. This process is continuous and highly dynamic, making it difficult to capture diurnal variations in the vertical distribution of algae with a single observation. Furthermore, under cloudy conditions, observational uncertainty is high. Furthermore, three-dimensional vertical simulations of algae lack the effective support of continuous daily variation data, leading to significant uncertainty in algae risk assessments. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for monitoring the diurnal changes in the vertical distribution of algae in lakes based on satellite remote sensing images, so as to solve the problem that it is difficult to obtain the diurnal changes in the vertical distribution of algae in a single observation in the existing technology, and the observation results are more uncertain under cloudy conditions. The three-dimensional vertical simulation of algae lacks effective support from continuous diurnal change data, which leads to greater uncertainty in algae risk assessment.

[0005] The present invention provides a method for monitoring diurnal changes in the vertical distribution of lake algae based on satellite remote sensing images, which is characterized by comprising:

[0006] Step 1: Obtain observation parameters for each sampling point in the algae-type lake area, record the sampling time and location latitude and longitude coordinates as a basis for matching with satellite data; mathematically describe the vertical distribution type of algae based on the observation parameters; and select descriptive parameters to describe the vertical profile distribution of algae based on the classified vertical distribution type;

[0007] Step 2: Study the spatiotemporal distribution and intraday variation of algae vertical distribution types at different sampling points, and construct a mapping relationship between morphological parameters and environmental factors in combination with observation parameters; analyze the response of environmental factors at the sampling points to the changes in algae vertical distribution, and select sensitive factors for the formation and changes of different algae vertical distribution types; based on field patrol and fixed-point continuous observation test data, simulate the absorption and backscattering, diffuse attenuation coefficient, uplink and downlink radiation brightness, uplink and downlink radiation illumination, remote sensing reflectivity, and photosynthetic active radiation under different Chl-a concentrations and different algae vertical distribution conditions, analyze the response of water optical properties and environmental factors to the formation process of different algae vertical distribution types, and determine the optical and environmental factor thresholds of algae vertical distribution types;

[0008] Step 3: Preprocess remote sensing images of algae-bearing lakes and establish an inversion model based on the relationship between surface chlorophyll concentration and remote sensing reflectance in the in-situ data. Use machine learning to optimize model parameters. Combined with the vertical distribution parameterization results, a remote sensing monitoring model for the vertical distribution of algae is constructed, and uncertainty analysis is performed.

[0009] Step 4: Integrate remote sensing data to invert the vertical distribution of lake algae, perform time series analysis on the inverted data, extract the daily variation characteristics of the vertical distribution of algae, and integrate them to reveal the daily variation patterns.

[0010] Furthermore, in step one, the observation parameters include: water body optical characteristic parameters, water environment parameters, meteorological parameters, and lake physical and hydrodynamic parameters; the water body optical characteristic parameters include remote sensing reflectance, underwater light field and backscattering coefficient; the water environment parameters include Chl-a concentration and transparency; the meteorological parameters include wind speed, air temperature and solar radiation; the lake physical and hydrodynamic parameters include water depth, flow direction and flow velocity.

[0011] Furthermore, in step 1, based on the observed parameters, the vertical distribution type of algae is mathematically described and divided into four types: vertical uniform distribution type, Gaussian distribution type, exponential distribution type and power function distribution type;

[0012] The vertical uniform distribution function relationship is expressed as:

[0013] f1(z)=C (Formula 1)

[0014] Where C is a constant;

[0015] The Gaussian distribution function relationship is expressed as:

[0016]

[0017] Where C0 represents the background value of vertical stratification concentration, σ1 represents the peak width, h represents the peak height, and z1 represents the depth;

[0018] The exponential distribution function relationship is expressed as:

[0019] f3(z)=m1×exp(m2×z1) (Formula 3)

[0020] Where m1 and m2 are exponential morphological parameters, z1 represents the depth;

[0021] The power function distribution function relationship is expressed as:

[0022]

[0023] Where n1 and n2 are both power function parameters, and z1 represents the depth.

[0024] Furthermore, in step 1, the coefficient of variation, patch coefficient, average residence depth, and water column concentration difference parameters are selected to describe the vertical profile distribution of algae. The calculation formula is as follows:

[0025]

[0026] Where CV is the coefficient of variation, CP is the patch coefficient, MRD is the mean residence depth, WCCD is the water column concentration difference, σ2 and m are the standard deviation and mean of the vertical concentration values, respectively, and x z and d z Represents the concentration value and depth value of the zth layer, x max and x min are the minimum and maximum vertical concentrations, respectively.

[0027] Furthermore, in step 2, the spatiotemporal distribution of vertical distribution types of algae at different sample points and the degree of intraday variation are studied. Combined with meteorological parameters, lake physical and hydrodynamic parameters, machine learning algorithms such as SVM, RF, XGB or DNN are used to construct the mapping relationship between morphological parameters and environmental factors.

[0028] Furthermore, in step 2, the HydroLight software was used to simulate the absorption and backscattering, diffuse attenuation coefficient, uplink and downlink radiation brightness, uplink and downlink radiation illuminance, remote sensing reflectivity and photosynthetic active radiation under different Chl-a concentrations of 10ug / L, 20ug / L, 30ug / L, 40ug / L, 50ug / L, 80ug / L and 100ug / L, as well as the vertical distribution of different algae including vertical uniform distribution, Gaussian distribution, exponential distribution and power function distribution.

[0029] Furthermore, step three includes:

[0030] Download the GOCI-II L1B file for later atmospheric correction processing; use the SeaDAS integrated MUMM algorithm, DSF algorithm, or 6SV atmospheric correction module to correct the GOCI-II image. Combined with the field spectral data measured synchronously by the satellite and the ground, the accuracy of different atmospheric correction algorithms is compared, and the optimal algorithm is selected to obtain the GOCI-IIRrs dataset;

[0031] Preprocess the GOCI-IIRrs dataset;

[0032] The Chl-a concentrations in the water bodies of the sampling points at the sampling stations were matched with the pixels of the preprocessed remote sensing image raster data, using the criteria that the time difference between the sampling time and the satellite observation time was no more than 5 hours, half of the pixels in the 3×3 pixel window around the central pixel of the sampling point coordinate had valid data, and the coefficient of variation of these valid data was less than 10%. The paired datasets were then arranged in ascending order of Chl-a concentration, and one sample was selected as a validation sample every three samples to obtain the paired datasets.

[0033] Combined with the existing data set, according to the band characteristics of the GOCI-II sensor, the FLH, TBA and BR Chl-a inversion algorithms were compared, and the best algorithm was selected as the remote sensing inversion algorithm for Chl-a concentration in water bodies at the sampling points:

[0034]

[0035] Where x1 = 101.056, x2 = 0.5057, Rrs(660), Rrs(680) and Rrs(745) are the reflectances at 660 nm, 680 nm and 745 nm, respectively;

[0036]

[0037] Chla=x1FLH+x2 (Formula 12)

[0038] Where x1 = 116.04, x2 = -51.18, Rrs(680), Rrs(745) and Rrs(865) are the reflectances at 680 nm, 745 nm and 865 nm, respectively;

[0039]

[0040] Wherein, x1=1.8875, x2=0.8296, x3=-0.2241, x4=10, Rrs(660) and Rrs(745) are the reflectances at 660nm and 745nm, respectively;

[0041] Use training dataset samples and validation dataset samples to evaluate the prediction accuracy of the model;

[0042] The lake surface Chl-a concentration, spectrum, meteorology and different vertical distribution structure parameters measured synchronously by satellite and ground were used to calibrate and verify the parameters based on the test data set through grid search method with the help of machine learning methods to determine the optimal model parameters and select the best universal algorithm to construct new algae vertical distribution structure parameters.

[0043] The Monte Carlo method is used to simulate a set of normally distributed errors. The error distribution is quantified according to the mean absolute percentage error of the input variable in the 6SV atmospheric correction, and the mean absolute value of the simulated errors is required to be equal to the MAPE in the atmospheric correction. Each spectral index is regarded as an independent variable of the whole. The set of simulated errors is superimposed on each sample point to obtain a series of input variables with additional uncertainty. Finally, the sensitivity of the model to the atmospheric correction error is obtained by comparing the estimated results using the input with additional uncertainty with the original input estimate. The number of Monte Carlo simulation runs for each variable is 100,000, and anomalous simulation values ​​other than three times the standard deviation are excluded.

[0044] Furthermore, step four includes:

[0045] After preprocessing, the GOCI-II image is positioned along the longitude and latitude grids, and the surface chlorophyll, spectrum, and environmental parameters are integrated on a pixel-by-pixel basis. The vertical type and structural parameters of each pixel are then substituted into the vertical model to estimate the vertical type and structural parameters of each pixel. A spatial distribution map at a fixed time is generated to obtain the vertical spatial distribution of algae in the lake at a fixed time.

[0046] Based on the 1-hour resolution of GOCI-II, we inverted vertical parameters scene by scene, calculated the vertical distribution of algae in lakes at different times, constructed long time series, analyzed the diurnal variation cycle, and quantified the daily-scale effect of algae vertical variation:

[0047] Generate the vertical average residence depth of lake algae and the difference in water column concentration, calculate the diurnal vertical change rate of algae, and obtain its diurnal spatial distribution. The calculation formula is as follows:

[0048]

[0049] Among them, Va is the vertical change rate at a certain time of the day, M t1 With M t2 are the average depths at time t1 and time t2, W t1 With W t2 are the water column concentration differences between time t1 and time t2, and △t is the time difference between time t1 and time t2.

[0050] The present invention has the following beneficial effects: The method for monitoring the daily changes in the vertical distribution of lake algae based on satellite remote sensing images of the present invention realizes the estimation of the daily changes in the vertical distribution of lake algae through the lake geostationary satellite remote sensing products. The vertical distribution structure of algae in lake water bodies is processed and analyzed eight times a day using GK2BGOCI-II satellite remote sensing images, and the spatial distribution of the vertical distribution and structural changes of lake algae is obtained based on pixel calculations. The daily changes in the vertical distribution of algae in the entire lake and sub-regions are statistically analyzed, and the automatic estimation of the vertical change rate of lake algae is preliminarily realized. The method of the present invention can comprehensively obtain the spatial distribution of the vertical distribution of lake algae with a spatial resolution of 250m, which has important reference and guiding significance for the assessment of lake eutrophication and early warning of algal blooms, and can provide important scientific basis and technical support for regional algal bloom control and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0052] Figure 1 is a schematic diagram of a flow chart of an embodiment of the present invention;

[0053] Figure 2 is a scatter plot of model-predicted and measured chlorophyll a concentrations;

[0054] Figure 3 (a) is the true color image of GOCI-II at 9:15 on September 27, 2024. Figure 3 (b) is the estimated result of vertical algae type distribution;

[0055] Figure 4 This is a schematic diagram showing the changes in the vertical distribution types of the main algae in a lake from September 23, 2024 to October 6, 2024. DETAILED DESCRIPTION

[0056] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings.

[0057] See also Figure 1 This embodiment takes a lake as an example and uses GK-2B-GOCI-II satellite data to further describe the technical solution of the present invention. Figure 1 As shown, the method for monitoring the diurnal changes in the vertical distribution of lake algae based on satellite remote sensing images disclosed in an embodiment of the present invention includes the following steps:

[0058] (1) Collect data on algae samples from different depths of the lake, establish satellite-ground synchronization data, and establish a mathematical model of vertical distribution, such as normal distribution fitting, to parameterize the distribution characteristics of algae in the water depth direction.

[0059] (1-1) Evenly distribute sampling points across the algae-forming lake. Under clear weather conditions, ensure that parameter measurements are taken at each sampling point at the same time as satellite imagery, striving to obtain synchronized satellite-ground data. Observational parameters include: water optical properties such as remote sensing reflectance, underwater light field, and backscatter coefficient; water environmental parameters such as Chl-a concentration and transparency; meteorological parameters such as wind speed, air temperature, and solar radiation; and lake physical and hydrodynamic parameters such as water depth, flow direction, and velocity. Finally, record the latitude and longitude coordinates of the sampling time and location for matching with satellite data.

[0060] (1-2) Based on the measured data, the vertical distribution types of algae are mathematically described and divided into four types: vertical uniform distribution type; Gaussian distribution type; exponential distribution type; power function distribution type. The vertical uniform distribution function relationship is expressed as:

[0061] f1(z)=C (Formula 1)

[0062] Where C is a constant;

[0063] The Gaussian distribution function relationship is expressed as:

[0064]

[0065] Where C0 represents the background value of vertical stratification concentration, σ1 represents the peak width, h represents the peak height, and z1 represents the depth;

[0066] The exponential distribution function relationship is expressed as:

[0067] f3(z)=m1×exp(m2×z1) (Formula 3)

[0068] Where m1 and m2 are exponential morphological parameters, z1 represents the depth;

[0069] The power function distribution function relationship is expressed as:

[0070]

[0071] Where n1 and n2 are both power function parameters, and z1 represents the depth.

[0072] (1-3) According to the classified vertical types, the coefficient of variation (CV), the coefficient of patchiness (CP), the mean residence depth (MRD), and the water column concentration difference (WCCD) are selected to describe the vertical profile distribution of algae. The calculation formula is as follows:

[0073]

[0074] Where CV is the coefficient of variation, CP is the patch coefficient, MRD is the mean residence depth, WCCD is the water column concentration difference, σ2 and m are the standard deviation and mean of the vertical concentration values, respectively, and x z and d z Represents the concentration value and depth value of the zth layer, x max and x min are the minimum and maximum vertical concentrations, respectively.

[0075] (2) Through multi-dimensional data coupling, the synergistic effects of light, temperature, and nutrients on the vertical distribution of algae were revealed, and their response thresholds were quantified.

[0076] (2-1) Based on the experimental results, the spatiotemporal distribution and intraday variation of algae vertical distribution types at different sample points were studied. Combined with relevant meteorological, hydrological, and lake physical characteristics data such as water depth, machine learning algorithms such as SVM, RF, XGB, and DNN were used to construct the mapping relationship between morphological parameters and environmental factors. By fitting nonlinear associations, quantifying feature weights, and simulating complex interactions, the accuracy of analyzing the response of algae vertical distribution to environmental factors was improved. The response of sample point environmental factors to the change of algae vertical distribution was analyzed, and sensitive factors for the formation and change of different algae vertical distribution types were selected.

[0077] (2-2) Based on the data from field surveys and fixed-point continuous observation tests, the HydroLight software was used to simulate the absorption and backscattering, diffuse attenuation coefficient, uplink and downlink radiation brightness, uplink and downlink radiation illuminance, remote sensing reflectivity, photosynthetic active radiation, etc. at different Chl-a concentrations of 10ug / L, 20ug / L, 30ug / L, 40ug / L, 50ug / L, 80ug / L, and 100ug / L, as well as the vertical uniform distribution type, Gaussian distribution type, exponential distribution type, and power function distribution type. The response patterns of the optical properties of water bodies and environmental factors to the formation process of different algae vertical distribution types were analyzed, and the optical and environmental factor thresholds of the algae vertical distribution types were determined.

[0078] (3) Preprocess the remote sensing images and establish an inversion model based on the relationship between surface chlorophyll concentration and remote sensing reflectance in the in-situ data. Use machine learning to optimize model parameters. Combined with the vertical distribution parameterization results, a remote sensing monitoring model for the vertical distribution of algae is constructed, and uncertainty analysis is performed.

[0079] (3-1) Download the GOCI-II L1B file from the Korea National Ocean Satellite Center (https: / / www.nosc.go.kr / eng / main.do) for subsequent atmospheric correction processing. GOCI-II images were corrected using mature atmospheric correction modules such as the MUMM algorithm, the DSF algorithm, and the 6SV algorithm integrated with SeaDAS. The accuracy of different atmospheric correction algorithms was compared using field spectral data from synchronous satellite and ground measurements, and the optimal algorithm was selected to obtain the GOCI-II Rrs dataset.

[0080] (3-2) Preprocessing the RRS reflectance product data obtained by remote sensing. Since meaningful Chla parameters cannot be retrieved in areas with cyanobacteria blooms, the floating algae index (FAI) proposed by Hu et al. (2010) was used to remove bloom pixels from the image data, with an FAI > -0.004. Furthermore, due to the large amount of aquatic vegetation in the eastern region of Taihu Lake, the method described in this paper is not applicable, so masking was performed to remove this portion of the data.

[0081] (3-3) Chl-a concentrations in water bodies at the Taihu Lake station were matched with pixels in preprocessed remote sensing imagery raster data, using the criteria that the time difference between sampling and satellite observation time was no more than 5 hours, that half of the pixels in a 3×3 pixel window around the central pixel of the sampling point coordinate had valid data, and that the coefficient of variation of these valid data was less than 10%. This paired dataset was then arranged in ascending order of Chl-a concentration, and one sample was selected every three samples as a validation sample to obtain the paired dataset.

[0082] (3-4) Combining existing datasets and based on the GOCI-II sensor band characteristics, we compared the Chl-a inversion algorithms such as Fluorescence Line Height (FLH), Total Broadband Absorption (TBA), and Broadband Reflectance (BR), and selected the best algorithm as the remote sensing inversion algorithm for Chl-a concentration in water bodies at the sampling points:

[0083]

[0084] Where x1 = 101.056, x2 = 0.5057, Rrs(660), Rrs(680) and Rrs(745) are the reflectances at 660 nm, 680 nm and 745 nm, respectively;

[0085]

[0086] Chla=x1FLH+x2 (Formula 12)

[0087] Where x1 = 116.04, x2 = -51.18, Rrs(680), Rrs(745) and Rrs(865) are the reflectances at 680 nm, 745 nm and 865 nm, respectively;

[0088]

[0089] Wherein, x1=1.8875, x2=0.8296, x3=-0.2241, x4=10, Rrs(660) and Rrs(745) are the reflectances at 660 nm and 745 nm, respectively.

[0090] The prediction accuracy of the model is evaluated using the training data set samples and the validation data set samples. The results are as follows: Figure 2 As shown, dots represent training set samples, and boxes represent validation set samples. The root mean square error (RMSE) and unbiased percentage difference (UPD) between the model's predictions and field-measured Chla concentrations are both small, and the scatter plots are evenly distributed on both sides of the 1:1 line, indicating that the model has good predictive ability.

[0091] (3-5) The lake surface Chl-a concentration, spectrum, meteorology and different vertical distribution structure parameters measured synchronously by satellite and ground are used to calibrate and verify the parameters based on the test data set through grid search method with the help of machine learning methods such as random forest (RF), support vector machine (SVM), and artificial neural network (CNN). The optimal model parameters are determined and the universal algorithm for constructing new algae vertical distribution structure parameters is selected.

[0092] Table 1 Machine learning model parameter set

[0093]

[0094] (3-6) First, a Monte Carlo simulation method was used to generate a set of normally distributed errors. The error distribution was quantified based on the mean absolute percentage error (MAPE) of the input variable in the 6SV atmospheric correction. This required that the mean absolute value of the simulated errors be equal to the MAPE in the atmospheric correction. Second, considering the correlation between spectral bands, each spectral index was treated as a global independent variable. The simulated errors were superimposed on each sample point to obtain a set of input variables with additional uncertainty. Finally, the model's sensitivity to atmospheric correction errors was determined by comparing the estimated results using the inputs with additional uncertainty with the original input estimates. Monte Carlo simulations were run 100,000 times for each variable, and outliers exceeding three times the standard deviation were excluded.

[0095] (4) Integrate remote sensing data to invert the vertical distribution of lake algae, perform time series analysis on the inverted data, extract the diurnal variation characteristics of the vertical distribution of algae, and integrate them to reveal the diurnal variation patterns, providing a systematic methodology for the refined study of lake algae ecology.

[0096] (4-1) After preprocessing, the GOCI-II image is positioned along the longitude and latitude grids, and the surface chlorophyll, spectrum and environmental parameters are integrated in pixel units. Vertical models such as Gaussian and exponential distribution fitting are substituted to estimate the vertical type of each pixel, such as surface aggregation and uniformity, as well as structural parameters such as peak depth and distribution width. A fixed-time spatial distribution map is generated to obtain the spatial distribution of algae in the lake at a fixed time. Figure 3 , to achieve vertical control of lake algae.

[0097] (4-2) Based on the 1-hour resolution of GOCI-II, vertical parameters are inverted scene by scene, the vertical distribution of lake algae at different times is calculated, and a long time series is constructed. Figure 4 , analyze the diurnal variation cycle and quantify the daily-scale effects of vertical changes in algae such as migration rate:

[0098] Generate the mean vertical residence depth (MRD) and water column concentration difference (WCCD) of lake algae, calculate the diurnal vertical change rate of algae, and obtain its diurnal spatial distribution. The calculation formula is as follows:

[0099]

[0100] Among them, Va is the vertical change rate at a certain time of the day, M t1 With M t2are the average depths at time t1 and time t2, W t1 With W t2 is the difference in water column concentration between time t1 and time t2, and △t is the time difference between time t1 and time t2. This enables refined monitoring of diurnal changes in vertical distribution, comprehensively analyzes vertical mechanisms, and provides technical support for lake ecological protection.

[0101] The above-described embodiments of the present invention do not limit the protection scope of the present invention.

Claims

1. A method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images, characterized in that: include: Step 1: Obtain observation parameters for each sampling point in the algae-type lake area, record the sampling time and location latitude and longitude coordinates as a basis for matching with satellite data; mathematically describe the vertical distribution type of algae based on the observation parameters; and select descriptive parameters to describe the vertical profile distribution of algae based on the classified vertical distribution type; Step 2: Study the spatiotemporal distribution of algae vertical distribution types at different sampling points and the degree of intraday variation. Combined with observation parameters, construct a mapping relationship between morphological parameters and environmental factors. Analyze the response of environmental factors at the sampling points to the changes in algae vertical distribution, and select sensitive factors for the formation and changes of different algae vertical distribution types. Based on field survey and fixed-point continuous observation test data, the absorption and backscattering, diffuse attenuation coefficient, uplink and downlink radiance, uplink and downlink irradiance, remote sensing reflectance, and photosynthetically active radiation under different Chl-a concentrations and different algae vertical distribution conditions were simulated. The response of water optical properties and environmental factors to the formation process of different algae vertical distribution types was analyzed, and the optical and environmental factor thresholds of algae vertical distribution types were determined. Step 3: Preprocess remote sensing images of algae-bearing lakes and establish an inversion model based on the relationship between surface chlorophyll concentration and remote sensing reflectance in the in-situ data. Use machine learning to optimize model parameters. Combined with the vertical distribution parameterization results, a remote sensing monitoring model for the vertical distribution of algae is constructed, and uncertainty analysis is performed. Step 4: Integrate remote sensing data to invert the vertical distribution of lake algae, perform time series analysis on the inverted data, extract the daily variation characteristics of the vertical distribution of algae, and integrate them to reveal the daily variation patterns.

2. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: In step one, the observation parameters include: water body optical characteristic parameters, water environment parameters, meteorological parameters, and lake physical and hydrodynamic parameters; the water body optical characteristic parameters include remote sensing reflectance, underwater light field and backscattering coefficient; the water environment parameters include Chl-a concentration and transparency; the meteorological parameters include wind speed, air temperature and solar radiation; the lake physical and hydrodynamic parameters include water depth, flow direction and flow velocity.

3. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: In step 1, based on the observed parameters, the vertical distribution type of algae is mathematically described and divided into four types: vertical uniform distribution type, Gaussian distribution type, exponential distribution type and power function distribution type; The vertical uniform distribution function relationship is expressed as: f1(z)=C (Formula 1) Where C is a constant; The Gaussian distribution function relationship is expressed as: Where C0 represents the background value of vertical stratification concentration, σ1 represents the peak width, h represents the peak height, and z1 represents the depth; The exponential distribution function relationship is expressed as: f3(z)=m1×exp(m2×z1) (Formula 3) Where m1 and m2 are exponential morphological parameters, z1 represents the depth; The power function distribution function relationship is expressed as: Where n1 and n2 are both power function parameters, and z1 represents the depth.

4. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: In step 1, the coefficient of variation, patch coefficient, average residence depth, and water column concentration difference parameters are selected to describe the vertical profile distribution of algae. The calculation formula is as follows: Where CV is the coefficient of variation, CP is the patch coefficient, MRD is the mean residence depth, WCCD is the water column concentration difference, σ2 and m are the standard deviation and mean of the vertical concentration values, respectively, and x z and d z Represents the concentration value and depth value of the zth layer, x max and x min are the minimum and maximum vertical concentrations, respectively.

5. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: In step 2, the spatiotemporal distribution of vertical distribution types of algae at different sample points and the degree of intraday variation are studied. Combined with meteorological parameters, lake physical and hydrodynamic parameters, the SVM, RF, XGB or DNN algorithm is used to construct the mapping relationship between morphological parameters and environmental factors.

6. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: In step 2, the HydroLight software was used to simulate the absorption and backscattering, diffuse attenuation coefficient, uplink and downlink radiation brightness, uplink and downlink radiation illuminance, remote sensing reflectivity and photosynthetic active radiation under different Chl-a concentrations of 10ug / L, 20ug / L, 30ug / L, 40ug / L, 50ug / L, 80ug / L and 100ug / L, as well as the vertical distribution of algae with vertical uniform distribution, Gaussian distribution, exponential distribution and power function distribution.

7. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: Step three includes: Download the GOCI-II L1B file for later atmospheric correction processing; use the SeaDAS integrated MUMM algorithm, DSF algorithm, or 6SV atmospheric correction module to correct the GOCI-II image. Combined with the field spectral data measured synchronously by the satellite and the ground, the accuracy of different atmospheric correction algorithms is compared, and the optimal algorithm is selected to obtain the GOCI-IIRrs dataset; Preprocess the GOCI-IIRrs dataset; The Chl-a concentrations in the water bodies of the sampling points at the sampling stations were matched with the pixels of the preprocessed remote sensing image raster data, using the criteria that the time difference between the sampling time and the satellite observation time was no more than 5 hours, half of the pixels in the 3×3 pixel window around the central pixel of the sampling point coordinate had valid data, and the coefficient of variation of these valid data was less than 10%. The paired datasets were then arranged in ascending order of Chl-a concentration, and one sample was selected as a validation sample every three samples to obtain the paired datasets. Combined with the existing data set, according to the band characteristics of the GOCI-II sensor, the FLH, TBA and BR Chl-a inversion algorithms were compared, and the best algorithm was selected as the remote sensing inversion algorithm for Chl-a concentration in water bodies at the sampling points: Where x1 = 101.056, x2 = 0.5057, Rrs(660), Rrs(680) and Rrs(745) are the reflectances at 660 nm, 680 nm and 745 nm, respectively; Chla=x1FLH+x2 (Formula 12) Where x1 = 116.04, x2 = -51.18, Rrs(680), Rrs(745) and Rrs(865) are the reflectances at 680 nm, 745 nm and 865 nm, respectively; Wherein, x1=1.8875, x2=0.8296, x3=-0.2241, x4=10, Rrs(660) and Rrs(745) are the reflectances at 660nm and 745nm, respectively; Use training dataset samples and validation dataset samples to evaluate the prediction accuracy of the model; The lake surface Chl-a concentration, spectrum, meteorology and different vertical distribution structure parameters measured synchronously by satellite and ground were used to calibrate and verify the parameters based on the test data set through grid search method with the help of machine learning methods to determine the optimal model parameters and select the best universal algorithm to construct new algae vertical distribution structure parameters. The Monte Carlo method is used to simulate a set of normally distributed errors. The error distribution is quantified according to the mean absolute percentage error of the input variable in the 6SV atmospheric correction, and the mean absolute value of the simulated errors is required to be equal to the MAPE in the atmospheric correction. Each spectral index is regarded as an independent variable of the whole. The set of simulated errors is superimposed on each sample point to obtain a series of input variables with additional uncertainty. Finally, the sensitivity of the model to the atmospheric correction error is obtained by comparing the estimated results using the input with additional uncertainty with the original input estimate. The number of Monte Carlo simulation runs for each variable is 100,000, and anomalous simulation values ​​other than three times the standard deviation are excluded.

8. The method for monitoring diurnal changes in vertical distribution of lake algae based on satellite remote sensing images according to claim 1, characterized in that: Step 4 includes: After preprocessing, the GOCI-II image is positioned along the longitude and latitude grids, and the surface chlorophyll, spectrum, and environmental parameters are integrated on a pixel-by-pixel basis. The vertical type and structural parameters of each pixel are then substituted into the vertical model to estimate the vertical type and structural parameters of each pixel. A spatial distribution map at a fixed time is generated to obtain the vertical spatial distribution of algae in the lake at a fixed time. Based on the 1-hour resolution of GOCI-II, we inverted vertical parameters scene by scene, calculated the vertical distribution of algae in lakes at different times, constructed long time series, analyzed the diurnal variation cycle, and quantified the daily-scale effect of algae vertical variation: Generate the vertical average residence depth of lake algae and the difference in water column concentration, calculate the diurnal vertical change rate of algae, and obtain its diurnal spatial distribution. The calculation formula is as follows: Among them, Va is the vertical change rate at a certain time of the day, M t1 With M t2 are the average depths at time t1 and time t2, W t1 With W t2 are the water column concentration differences between time t1 and time t2, and △t is the time difference between time t1 and time t2.

Images