Vegetation index and water quality parameter combined cyanobacterial bloom grading early warning system

Abstract

The invention relates to the technical field of cyanobacterial bloom grading, in particular to a cyanobacterial bloom grading early warning system combining vegetation indexes and water quality parameters. The method comprises the following steps: firstly, deducing a drifting trajectory and accumulated turbulence energy of drifting water mass based on a wind condition so as to calculate a surface algae retention index; constructing a flow direction constraint drifting model of the drifting water mass, solving the model to evaluate a pairing relation between the drifting water mass and neighborhood pixels, further combining a surface algae retention index of the drifting water mass and monitoring water quality, and decoupling an algae spectral response coefficient and a water body turbid background parameter; and determining a grading early warning threshold value to carry out water bloom grading early warning. According to the method, the wind-driven flow field is reconstructed and the turbulence mixing constraint is introduced, so that the problem that fixed-point monitoring and satellite observation are inconsistent in spatial-temporal scale and physical space is solved, and meanwhile, the algae spectral response coefficient and the water turbid background parameter are decoupled to adaptively adjust the grading early warning threshold value, so that the accuracy of grading early warning of cyanobacterial bloom is improved.

Description

Technical Field

[0001] This invention relates to the field of cyanobacterial bloom classification technology, specifically to a cyanobacterial bloom classification and early warning system that combines vegetation index and water quality parameters. Background Technology

[0002] In monitoring technologies for cyanobacterial blooms in large, shallow lakes, fixed-point water quality monitoring and satellite remote sensing are the two main data acquisition methods. Ground stations can provide water quality parameters with high temporal resolution, while satellite remote sensing provides instantaneous areal spectral images covering the entire lake area. Existing graded early warning systems for cyanobacterial blooms typically establish regression models directly based on pixel values ​​in areal spectral images taken during satellite transit and water quality parameters sampled in real time at ground stations, and classify them according to fixed statistical thresholds.

[0003] However, during satellite imaging, water masses collected by ground stations drift with the current, causing a spatial mismatch between point water quality parameters and areal spectral images. Strong winds and waves in shallow lakes can also cause vertical turbulence, drawing cyanobacteria underwater, resulting in surface optical signals observed by satellite being lower than the total biomass of the water column. This difference in horizontal displacement and vertical distribution caused by fluid movement leads to a lack of physical consistency between point water quality data and areal spectral images. Simultaneously, the content of suspended sediment and other background substances in the water body fluctuates in real time with hydrodynamic forces, altering the optical response baseline between chlorophyll concentration and vegetation index. Fixed thresholds cannot adapt to this dynamic environmental background, resulting in low accuracy in grading and early warning of cyanobacterial blooms. Summary of the Invention

[0004] To address the technical problem of low accuracy in graded early warning of cyanobacterial blooms, the present invention aims to provide a graded early warning system for cyanobacterial blooms that combines vegetation index and water quality parameters. The specific technical solution adopted is as follows:

[0005] A grading and early warning system for cyanobacterial blooms that combines vegetation indices and water quality parameters, the system comprising:

[0006] Data acquisition module: used to acquire the vegetation index of each pixel in the satellite spectral image of the lake surface and determine the time window based on the satellite imaging time; acquire water quality data and wind condition data of ground stations at each acquisition time within the time window;

[0007] Drift trajectory extrapolation module: used to determine the drift trajectory coordinates of drifting water masses at ground stations at each acquisition time based on wind data, and to determine the cumulative turbulent energy of each drifting water mass at the satellite imaging time;

[0008] The physical constraint inversion module is used to obtain the surface algae retention index of each drifting water mass based on the accumulated turbulent energy, construct a flow direction-constrained drift model for the drifting water mass, solve the model to obtain the pairing coefficient between each drifting water mass and each neighboring pixel at the drift trajectory coordinates in the satellite spectral image; based on the surface algae retention index of each drifting water mass and the corresponding water quality data, combined with the vegetation index of each neighboring pixel and its pairing weight with each drifting water mass, the algae spectral response coefficient and water turbidity background parameters are obtained.

[0009] The graded early warning module is used to determine the graded early warning threshold based on the algal spectral response coefficient and the water turbidity background parameters, and to conduct graded early warning of algal blooms based on the vegetation index of each pixel.

[0010] Furthermore, the method for obtaining the time window includes:

[0011] Based on the lake flow velocity and the coverage radius of the ground station, the retrospective duration is obtained; taking the satellite imaging time as the center, the time periods of the retrospective duration are obtained on both sides of the time series, and the time range corresponding to the two time periods is used as the time window.

[0012] Furthermore, the method for obtaining the drift trajectory coordinates includes:

[0013] At each acquisition moment within the time window, the lake surface velocity vector is determined based on the wind condition data, the preset wind flow conversion coefficient, and the preset flow direction deflection matrix; the duration between adjacent acquisition moments is taken as the drift duration, and the drift displacement is determined based on the lake surface velocity vector and the drift duration.

[0014] For each drifting water mass, within the time window, the drift trajectory coordinates for the next adjacent acquisition time are determined based on the superposition of the location coordinates of the ground station and the drift displacement at the first acquisition time. This process is repeated to obtain the drift trajectory coordinates for each acquisition time.

[0015] Furthermore, the method for obtaining the accumulated turbulent energy includes:

[0016] At each acquisition moment within the time window, the wind and wave stirring power coefficient is determined based on the wind condition data. The duration between adjacent acquisition moments is taken as the stirring duration. The wind and wave stirring power coefficient and the stirring duration are combined to obtain turbulent energy.

[0017] For each drifting water mass, within the time window, the cumulative turbulent energy of the drifting water mass at the time of satellite imaging is determined based on the sum of the initial turbulent energy and the turbulent energy at all acquisition times prior to the satellite imaging time.

[0018] Furthermore, the method for obtaining the surface algae retention index includes:

[0019] By comparing the cumulative turbulent energy of different drifting water masses, the relative turbulent mixing intensity coefficient of each drifting water mass is obtained; the negative correlation normalization result of the relative turbulent mixing intensity coefficient is used as the surface algae retention index of each drifting water mass.

[0020] Furthermore, the method for obtaining the flow direction constraint drift model includes:

[0021] Using the pairing coefficient between each drifting water mass and each neighboring pixel as the optimization variable, a spatial matching cost term is determined by combining the distribution relationship between each drifting water mass and each neighboring pixel at the corresponding drift trajectory coordinates. A ground distribution constraint term is determined by combining the water quality data of each drifting water mass and the surface algae retention index. A satellite observation constraint term is determined by combining the positional distribution of the neighboring pixels. The spatial matching cost term, the ground distribution constraint term, and the satellite observation constraint term are fused to construct a flow direction constrained drift model.

[0022] Furthermore, the method for obtaining the spatial matching cost term includes:

[0023] For each drifting water mass, the displacement vector pointing to each neighboring cell is determined based on the drift trajectory coordinates of the drifting water mass, and the displacement vector is decomposed into a drift component along the flow and a drift component vertically.

[0024] The drift cost parameter between the drifting water mass and each neighboring pixel is obtained by weighted summing of the modulus of the drifting component along the flow and the modulus of the vertical drift component; the weighting weight of the modulus of the vertical drift component is greater than the weighting weight of the modulus of the drifting component along the flow.

[0025] By weighting the optimization variables of the flow-direction constraint drift model using the drift cost parameter, the spatial matching cost term is obtained.

[0026] Furthermore, the method for obtaining the algal spectral response coefficient and the water turbidity background parameters includes:

[0027] The water quality data includes at least chlorophyll concentration; for each drifting water mass, the chlorophyll concentration is weighted using the surface algae retention index, and the weighted result is taken as the surface chlorophyll concentration.

[0028] A linear regression model of vegetation index for pixels in satellite spectral imagery is constructed; wherein, the variable in the linear regression model is the surface chlorophyll concentration, the regression coefficient is the algal spectral response coefficient, and the bias is the water turbidity background parameter.

[0029] Using the pairing coefficients between each drifting water mass and each neighboring pixel as weights, the regression coefficients and biases in the linear regression model are solved based on the weighted least squares method to obtain the algal spectral response coefficients and water turbidity background parameters.

[0030] Furthermore, the method for obtaining the graded early warning threshold includes:

[0031] A preset number of risk levels are determined, and the spectral response coefficient of the algae is weighted using the preset warning concentration of each risk level. The weighted result is then fused with the water turbidity background parameter as the graded warning threshold for the corresponding risk level.

[0032] Furthermore, the classification and early warning of algal blooms based on the vegetation index of each pixel includes:

[0033] In the satellite spectral image, for each pixel, the warning attribute of the pixel is marked based on the spatial distance between the pixel and the nearest drift trajectory coordinates, wherein the warning attribute with a spatial distance greater than a preset threshold is a confirmed warning, and the warning attribute with a spatial distance less than or equal to the preset threshold is a suspected warning.

[0034] The graded warning interval is determined based on the graded warning threshold of adjacent risk levels. For each pixel, when the vegetation index is within the corresponding graded warning interval, the pixel is marked as the corresponding risk level.

[0035] Algal bloom grading and early warning are based on pixel-level warning attributes and risk levels.

[0036] The present invention has the following beneficial effects:

[0037] This invention first acquires the vegetation index of each pixel in satellite spectral images of the lake surface and determines a time window based on the satellite imaging time, providing a time limit for subsequent extrapolation and retrospection. Then, it acquires water quality and wind data from ground stations at each acquisition time within the time window, providing a data foundation for subsequent graded early warning analysis. Furthermore, based on the wind data, it extrapolates and retrospectively determines the drift trajectory coordinates of each drifting water mass at the satellite imaging time and determines the cumulative turbulent energy of each drifting water mass at the satellite imaging time, thereby obtaining the surface algae retention index of each drifting water mass. Then, it constructs a flow-direction-constrained drift model for the drifting water mass, solves the model to obtain the pairing coefficient between each drifting water mass and each neighboring pixel, and further combines the surface algae retention index of each drifting water mass with the corresponding water quality data to decouple the algae spectral response coefficient and water turbidity background parameters in the satellite spectral signal, thereby determining the graded early warning threshold and conducting graded early warning for algal blooms based on the vegetation index of each pixel. This invention addresses the inconsistency between fixed-point monitoring and satellite observation in terms of spatiotemporal scale and physical space by reconstructing the wind-driven flow field and introducing turbulent mixing constraints. It also decouples the algal spectral response coefficient from the water turbidity background parameters and adaptively adjusts the graded early warning threshold, thereby improving the accuracy of graded early warning for cyanobacterial blooms. Attached Figure Description

[0038] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a block diagram of a cyanobacterial bloom classification and early warning system that combines vegetation index and water quality parameters, provided as an embodiment of the present invention. Detailed Implementation

[0040] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0042] The following description, in conjunction with the accompanying drawings, details a specific scheme for a graded early warning system for cyanobacterial blooms that combines vegetation index and water quality parameters, provided by this invention.

[0043] Please see Figure 1 The diagram shows a block diagram of a cyanobacterial bloom classification and early warning system that combines vegetation index and water quality parameters according to an embodiment of the present invention. The system includes a data acquisition module 101, a drift trajectory inference module 102, a physical constraint inversion module 103, and a classification and early warning module 104.

[0044] This invention addresses the horizontal drift of monitored objects at ground stations caused by wind-driven currents and the vertical mixing of algae caused by strong winds and waves in shallow lakes. It establishes an inversion model based on fluid dynamics constraints, solves the model, and decouples the algal spectral response coefficient, which characterizes the color development ability of algae, from the water turbidity background value, which characterizes the interference of non-biological suspended matter. It then establishes a grading standard for graded early warning of cyanobacterial blooms.

[0045] Data acquisition module 101: used to acquire the vegetation index of each pixel in the satellite spectral image of the lake surface and determine the time window based on the satellite imaging time; acquire water quality data and wind condition data of the ground station at each acquisition time within the time window.

[0046] In one embodiment of the present invention, satellite spectral images of the lake surface are first acquired, and the vegetation index (NDVI) of each pixel in the satellite spectral images is obtained. The vegetation index is a numerical indicator used to quantify the surface vegetation cover and can provide a data basis for subsequent graded early warning. This is a well-known technology and will not be described in detail here. At the same time, water quality data and wind condition data of the lake surface are collected at fixed points through ground-deployed ground stations. The water quality data can help to assess algal bloom information, and the wind condition data can help to assess the impact of wind conditions on the drift of water masses and the underwater entrainment of cyanobacteria.

[0047] The water quality data includes at least the chlorophyll concentration in the collected water samples, and the wind data includes wind speed and wind direction, or can be directly represented by the wind speed vector. The water quality data and wind data are collected every 10-30 minutes. In this embodiment, they are collected every 10 minutes. The implementer can also adjust them, but it is necessary to ensure that the water quality data and wind data are synchronized in time.

[0048] It should be noted that the acquisition of the above-mentioned satellite spectral images, vegetation index, water quality data and wind condition data are all existing technical means, and will not be described in detail here; in the process of graded early warning, there may be multiple ground stations. This embodiment takes one ground station as an example for analysis and description.

[0049] Then, the satellite imaging time of the satellite spectral image is used as the system anchor point or the time of graded early warning. Since the cyanobacteria in the lake water may drift with the current, a certain pixel in the satellite spectral image may correspond to a water mass that drifted from a certain position upstream at a certain time in history. Therefore, the time window is further determined based on the satellite imaging time in order to retrospectively or extrapolate the drift of the water mass.

[0050] It should be noted that the graded early warning method is consistent for each satellite imaging moment, and the analysis and description here will only take any satellite imaging moment as an example.

[0051] Preferably, in one embodiment of the present invention, considering that setting the time window too long not only increases the extrapolation burden but may also introduce errors—for example, theoretically, the long-term drift trajectory of the water mass can be extrapolated, but the water mass may have already undergone biochemical reactions or deteriorated during the long-term drift, resulting in a large error in the extrapolated trajectory; and considering that the lake flow velocity will affect the drift speed of the water mass to a certain extent, and that the monitoring coverage radius of the ground station can be regarded as a representative or reliable extrapolation boundary, thereby determining the extrapolation or retrospective duration; therefore, the method for obtaining the time window includes:

[0052] Based on the lake flow velocity and the coverage radius of the ground station, the retrospective duration is obtained; taking the satellite imaging time as the center, the retrospective duration periods are obtained on both sides of the time series, and the time range corresponding to the two periods is used as the time window.

[0053] As an example, the coverage radius of the ground station is first divided by the lake flow velocity to obtain the extrapolated backtracking time. The lake flow velocity is an average velocity calculated based on historical statistical information, for example, 0.1-0.3 m / s; this embodiment uses an intermediate value. The coverage radius of the ground station needs to be determined based on its design parameters, for example, 5000 m. Then, the time window is determined with the satellite imaging time as the center, i.e. ;in, For the satellite imaging time, To extrapolate the duration of the backtracking.

[0054] Furthermore, it is possible to determine the water quality and wind conditions at each collection point within the time window at the ground station; among them, the monitoring objects (or collected water samples) at different collection points correspond to different drifting water masses.

[0055] To facilitate subsequent drift simulations, the latitude and longitude of the ground measurement points are converted to Cartesian coordinates using either the Gauss-Kruger projection or the universal transverse Mercator projection, denoted as [reference needed]. , where i is the code for the drifting water mass, thus corresponding to the geographic projection coordinates of the satellite spectral image. The projection methods mentioned above are all well-known technologies and will not be elaborated further.

[0056] Drift trajectory extrapolation module 102: used to determine the drift trajectory coordinates of drifting water masses at ground stations at each acquisition time based on wind condition data at the satellite imaging time, and to determine the cumulative turbulent energy of each drifting water mass at the satellite imaging time.

[0057] Since water bodies are driven by wind stress and move continuously, and the wind conditions on the lake surface directly affect the drift direction and speed of the water mass, this embodiment of the invention will determine the drift trajectory coordinates of the drifting water mass at the ground station at each acquisition time based on wind condition data at the time of satellite imaging.

[0058] Preferably, in one embodiment of the present invention, considering that wind energy is not directly converted into water flow power, and that the moving water flow is affected by the Coriolis force (geostrophic force) due to the Earth's rotation, wind condition data can first be analyzed based on a preset wind flow conversion coefficient and a preset flow direction deflection matrix to estimate and determine the flow velocity vector of the lake surface, and then estimate the drift displacement of the drifting water mass at each collection moment to deduce the drift trajectory; therefore, the method for obtaining the drift trajectory coordinates includes:

[0059] At each acquisition moment within the time window, the lake surface velocity vector is determined based on wind condition data, preset wind flow conversion coefficient, and preset flow direction deflection matrix; the duration between adjacent acquisition moments is taken as the drift duration, and the drift displacement is determined based on the lake surface velocity vector and the drift duration.

[0060] For each drifting water mass, within the time window, the drift trajectory coordinates for the next adjacent acquisition time are determined by superimposing the location coordinates of the ground station with the drift displacement at the first acquisition time. This process is repeated to obtain the drift trajectory coordinates for each acquisition time.

[0061] As an example, let's analyze and describe the process using any drifting water mass and any sampling time within a time window:

[0062] First, determine the lake surface current vector. Where k is the data acquisition time number, Let be the lake surface velocity vector at the k-th acquisition time. The airflow conversion coefficient is typically set between 0.02 and 0.04; in this embodiment, an intermediate value is used. For taking the length of the modulus; The wind speed vector is represented as follows: , The horizontal component of wind speed, This represents the vertical component of the wind speed. The preset flow deflection matrix is ​​represented as follows: , The deflection angle (recommended value 20°-45°, used to describe the flow deflection caused by the Coriolis force); Used to characterize the wind speed after the transformation of airflow. Used to characterize the wind speed and direction after the airflow is deflected.

[0063] Due to the satellite imaging time , No. The actual location of the drifting water mass is not at the ground measuring point. If the collection time <Satellite Imaging Time> Explanation of the first The drifting water mass had already drifted downstream of the ground measurement point at the time of satellite imaging; if the acquisition time... >Satellite Imaging Time Explanation of the first If a drifting water mass is still upstream of the ground measurement point at the time of satellite imaging, then the drift inference within the time window also includes forward inference and backward backtracking.

[0064] like (Forward deduction), water mass moves with the current: ;

[0065] like (Reverse flow), water mass flowing backwards: ;

[0066] Where k is the data acquisition time number; Used to characterize the drift trajectory coordinates at the (K+1)th acquisition time; Used to characterize the drift trajectory coordinates at the Kth acquisition time; This represents the lake surface velocity vector at the k-th acquisition time. This refers to the drift duration, which is the duration between adjacent data collection times. .

[0067] Then, the drift trajectory coordinates of each drifting water mass at the time of satellite imaging are determined from the drift simulation trajectory. .

[0068] When the lake surface is calm, cyanobacteria usually float on the surface and can be clearly observed by satellites. However, wind and waves can cause turbulence, which can sweep the cyanobacteria to the bottom, making it possible for satellites to only observe a portion of the cyanobacteria, thus affecting the effectiveness of subsequent graded early warnings. Based on this, this embodiment of the invention further determines the cumulative turbulent energy of each drifting water mass at the time of satellite imaging based on wind condition data, in order to prepare for subsequent assessment of the surface algae retention index of the drifting water mass for accurate graded early warnings.

[0069] It should be noted that in one embodiment of the present invention, it is assumed that the wind field on the lake surface is uniform. Therefore, at each sampling time, even if the drifting water mass drifts to another location, its wind conditions are consistent with the wind conditions monitored by the ground station. In other embodiments, the implementer may also deploy multiple ground stations, assuming that the wind conditions within the coverage radius of each station are consistent. Then, the wind conditions monitored at the ground stations near the drifting water mass during its drift are determined and applied to the analysis and calculation of accumulated turbulent energy.

[0070] Preferably, in one embodiment of the present invention, based on classical empirical relationships in fluid dynamics and ocean physics, it can be deduced that the stirring power of wind-driven water is proportional to the cube of the wind speed. Therefore, the turbulent energy caused by wind and waves at each acquisition moment can be obtained by combining the stirring duration, and then the cumulative turbulent energy of the drifting water mass at the satellite imaging moment can be determined. The method for obtaining the cumulative turbulent energy includes:

[0071] At each acquisition moment within the time window, the wind and wave stirring power coefficient is determined based on the wind condition data. The duration between adjacent acquisition moments is taken as the stirring duration. The wind and wave stirring power coefficient and the stirring duration are combined to obtain the turbulent energy.

[0072] For each drifting water mass, within the time window, the cumulative turbulent energy of the drifting water mass at the time of satellite imaging is determined based on the sum of the initial turbulent energy and the turbulent energy at all acquisition times prior to the satellite imaging time.

[0073] As an example, let's analyze and describe the situation using any drifting water mass as an example:

[0074] Within the time window, the wind speed vector magnitude is collected at each acquisition moment. The cube of the turbulent energy is then multiplied by the stirring duration, and the product is taken as the turbulent energy at the acquisition time. Furthermore, the cumulative turbulent energy is the sum of the turbulent energies at all acquisition times up to the satellite imaging time. The initial turbulent energy of the drifting water mass is 0.

[0075] Physically constrained inversion module 103: This module is used to obtain the surface algae retention index of each drifting water mass based on the accumulated turbulent energy, and to construct a flow direction constrained drift model for the drifting water mass. The model is solved to obtain the pairing coefficient between each drifting water mass and each neighboring pixel at the drift trajectory coordinates in the satellite spectral image. Based on the surface algae retention index of each drifting water mass and the corresponding water quality data, combined with the vegetation index of each neighboring pixel and its pairing weight with each drifting water mass, the algae spectral response coefficient and water turbidity background parameters are obtained.

[0076] Because some algae are swept underwater by wind, waves, and turbulence, satellite observation signals show nonlinear signal loss compared to ground-based measurements. Therefore, this embodiment of the invention further obtains the surface algae retention index of each drifting water mass based on the accumulated turbulence energy. The surface algae retention index is used to characterize the proportion of cyanobacteria in the drifting water mass that can still be monitored by satellites under the influence of wind, waves, and turbulence. This prepares for the subsequent construction of a flow-direction-constrained drift model of the drifting water mass, in order to calculate the true algae spectral response coefficient from the complex flow field environment, and thus prepare for graded early warning.

[0077] Preferably, in one embodiment of the present invention, based on the Rouse Profile theory in fluid mechanics and the competition mechanism between Archimedes' buoyancy and turbulent diffusion, it can be deduced that cyanobacteria will actively rise to the surface under the influence of Archimedes' buoyancy, forming algal blooms, while turbulent eddies will mix the water, thus dispersing the cyanobacteria in the vertical direction; and the greater the turbulent energy, the more uniform the vertical distribution of cyanobacteria in the turbulence, and the lower the proportion remaining on the lake surface; therefore, the method for obtaining the surface algae retention index includes:

[0078] By comparing the cumulative turbulent energy of different drifting water masses, the relative turbulent mixing intensity coefficient of each drifting water mass is obtained; the negative correlation normalization result of the relative turbulent mixing intensity coefficient is used as the surface algae retention index of each drifting water mass.

[0079] As an example, the cumulative turbulent energy of each drifting water mass is first normalized to its maximum and minimum values, and the result of the normalization is used as the relative turbulent mixing intensity coefficient (dimensionless parameter) of the drifting water mass. Alternatively, the implementer can divide the cumulative turbulent energy of each drifting water mass by the maximum cumulative turbulent energy to obtain the relative turbulent mixing intensity coefficient. Furthermore, a variant of the Sigmoid function is used for negative correlation normalization to determine the surface algae retention index of the drifting water mass.

[0080] Specifically, the surface algae retention index ; , where is the surface algae retention index of the i-th drifting water mass; It is an exponential function with the natural constant e as its base; A preset sensitivity coefficient (within the range of 5-15, and 10 in this embodiment) is used to control the steepness of the curve; Let be the relative turbulent mixing intensity coefficient of the i-th drifting water mass; The preset phase transition midpoint (with a value range of 0.4-0.6, and 0.5 in this embodiment) is used to characterize the critical point at which turbulent mixing begins to significantly suppress buoyancy.

[0081] Since the drift simulation of the drifting water mass is only a theoretical estimate, it may also be affected by other local turbulence during the drift process, which may lead to a certain deviation in the drift trajectory coordinates. That is, the pixels corresponding to the drift trajectory coordinates in the satellite spectral image may not be completely consistent with the drifting water mass.

[0082] Based on this, the embodiments of the present invention further construct a flow direction-constrained drift model for drifting water masses, and obtain the pairing coefficient between each drifting water mass and each neighboring pixel at the drift trajectory coordinate in the satellite spectral image by solving the model; the pairing coefficient represents the probability that the pixel corresponds to a drifting water mass, and prepares for the subsequent decoupling of algal spectral response coefficient and water turbidity background parameters.

[0083] It should be noted that for each drifting water mass, the neighboring cell refers to the cell within a preset neighborhood centered on the drift trajectory coordinates; in this embodiment, the size of the preset neighborhood is 5×5, but the implementer can adjust it as needed.

[0084] Considering that drifting water masses usually drift along the direction of water flow, but if the drifting water masses deviate significantly from or jump out of the flow direction, it may violate the laws of nature. Therefore, the spatial matching cost term can be determined first based on the distribution relationship between each drifting water mass and each neighboring cell at the coordinate of the corresponding drift trajectory. The spatial matching cost term allows for small errors in the direction of water flow for drifting water masses, that is, it is desirable to match cells along the direction of water flow. When matching cells perpendicular to the direction of water flow, there will be a large matching cost.

[0085] Considering that wind and waves may sweep cyanobacteria underwater, the actual amount of cyanobacteria or drifting water masses observed by the satellite may be less than the sampling amount at ground measuring points. Therefore, the ground distribution constraint can be determined based on the water quality data of each drifting water mass and the surface algae retention index. The ground distribution constraint allows for data gaps, meaning that some pixels do not need to be rigidly matched with drifting water masses, i.e., there is no need for a one-to-one perfect match, thereby avoiding matching distortion.

[0086] Furthermore, considering that during the matching process, in order to pursue the closest distance between a pixel and a drifting water source, multiple drifting water masses may correspond to the same nearest pixel, and each pixel should ideally only be matched with one drifting water mass. If a pixel is forcibly matched with a neighboring drifting water mass, it may directly distort the matching between drifting water masses and pixels in the entire lake surface flow field; therefore, satellite observation constraints can be determined based on the positional distribution of neighboring pixels. These satellite observation constraints are used to constrain the degree of pixel matching clustering.

[0087] Based on this, in a preferred embodiment of the present invention, the method for obtaining the flow direction constraint drift model includes:

[0088] Using the pairing coefficient between each drifting water mass and each neighboring pixel as the optimization variable, the spatial matching cost term is determined by combining the distribution relationship between each drifting water mass and each neighboring pixel at the corresponding drift trajectory coordinates. The ground distribution constraint term is determined by combining the water quality data and surface algae retention index of each drifting water mass. The satellite observation constraint term is determined by combining the location distribution of neighboring pixels. The flow direction constraint drift model is constructed by fusing the spatial matching cost term, the ground distribution constraint term, and the satellite observation constraint term.

[0089] As an example, the flow direction-constrained drift model can be expressed as follows:

[0090] ;in, A flow direction-constrained drift model; The function is for finding the minimum value; j is the index of the neighboring pixel corresponding to the i-th drifting water mass; is the total number of neighboring pixels corresponding to the i-th drifting water mass; i is the index of the drifting water mass; This represents the total number of drifting water masses; This refers to the spatial matching cost term; A preset regularization parameter (usually set to 0.01-0.1, and 0.1 in this embodiment) is used to control the relaxation of the matching; is the KL divergence function, used for asymmetry measurement, that is, allowing the total mass of two distributions to be unequal; The coefficient matrix, which is the pairing coefficient between each drifting water mass and each neighboring cell, is the optimization variable. The size of the coefficient matrix is ​​I×J, and the matrix elements are (variable) pairing coefficients, which are used to prepare for subsequent optimization solutions. It is the transpose of the coefficient matrix; It is a column vector consisting entirely of 1s, with a length of I; is the demand distribution vector of the pixels, which is also a column vector of all 1s with a length of I, used to constrain the clustering and matching of pixels; To supply the ground distribution vector for the drifting water mass, the pointing element is... A column vector of length I is used for hard matching of pixels.

[0091] For example, matching drifting water clusters with neighboring pixels can be viewed as filling pixels with water clusters;

[0092] Ground distribution constraints middle, Each vector element in the vector represents the total amount of water drift allocated to each neighboring cell. Each vector element in the vector is used to characterize the surface algae residue of each drifting water mass after turbulent mixing. The KL divergence function is then used to evaluate the proximity of the distributions. The closer the distributions are, the smaller the KL divergence function is.

[0093] In satellite observation constraints middle, Each vector element in the vector represents the total amount of all drifting water masses allocated to a given neighborhood cell. Each vector element in the vector represents the total amount that each neighboring cell can receive. The KL divergence function is then used to evaluate the proximity of the assignments. The closer the assignments are, the smaller the KL divergence function is.

[0094] In a preferred embodiment of the present invention, the method for obtaining the spatial matching cost term includes:

[0095] For each drifting water mass, the displacement vector pointing to each neighboring cell is determined based on the drift trajectory coordinates of the drifting water mass, and the displacement vector is decomposed into a drift component along the flow and a drift component vertically.

[0096] The drift cost parameter between the drifting water mass and each neighboring cell is obtained by weighted summing of the modulus of the drifting component along the flow and the modulus of the vertical drift component; the weighting weight of the modulus of the vertical drift component is greater than the weighting weight of the modulus of the drifting component along the flow.

[0097] By weighting the optimization variables of the flow-direction constrained drift model using the drift cost parameter, the spatial matching cost term is obtained.

[0098] Specifically, for each drifting water mass, first determine the displacement vector of the drift trajectory coordinates of the drifting water mass pointing to each neighboring cell, and then decompose the displacement vector into a drift component along the flow and a drift component perpendicular to the flow; wherein, the drift component along the flow is in the same direction as the water flow, and the drift component perpendicular to the water flow.

[0099] Then, the modulus of the along-flow drift component and the modulus of the vertical drift component are weighted and summed to obtain the drift cost parameter between the drifting water mass and each neighboring cell; wherein, the weighting weight of the modulus of the along-flow drift component is set to 1; the modulus of the vertical drift component is set to a value greater than 10, which is 10 in this embodiment, to severely punish matching that violates the laws of fluid dynamics.

[0100] Furthermore, the spatial matching cost term ; Let be the drift cost parameter between the i-th drifting water mass and the j-th neighboring pixel; The variable pairing coefficients between the i-th drifting water mass and the j-th neighboring pixel are matrix elements in the coefficient matrix.

[0101] Then, the flow-direction constrained drift model is solved to minimize S, thereby obtaining the pairing coefficient between each drifting water mass and each neighboring pixel at the drift trajectory coordinates in the satellite spectral image.

[0102] Since the vegetation index of each pixel in satellite spectral image is partly contributed by algae and partly by background water, and the pairing coefficient between the pixel and the drifting water mass can help assess the confidence level of the correspondence between the two, the vegetation index can be further decomposed to separate the algae contribution that is proportional to the algae concentration on the surface of the water mass and the interference from the background water from the mixed satellite signal (pixel vegetation index), thereby obtaining the algae spectral response coefficient and the water turbidity background parameter.

[0103] The algal spectral response coefficient is used to characterize the vegetation index increment generated by a unit effective chlorophyll concentration in satellite spectral images, while the water turbidity background parameter is used to characterize the vegetation index base generated by background substances such as silt and colored dissolved organic matter, in addition to algae.

[0104] In a preferred embodiment of the present invention, the method for obtaining the algal spectral response coefficient and water turbidity background parameters includes:

[0105] The water quality data should include at least chlorophyll concentration; for each drifting water mass, the chlorophyll concentration should be weighted using the surface algae retention index, and the weighted result should be used as the surface chlorophyll concentration.

[0106] A linear regression model of vegetation index for pixels in satellite spectral imagery was constructed. The variable in the linear regression model is the surface chlorophyll concentration, the regression coefficient is the algal spectral response coefficient, and the bias is the water turbidity background parameter.

[0107] Using the pairing coefficients between each drifting water mass and each neighboring pixel as weights, the regression coefficients and biases in the linear regression model are solved based on the weighted least squares method to obtain the algal spectral response coefficients and water turbidity background parameters.

[0108] As an example, firstly, the surface algae retention index is used to weight chlorophyll concentration to calculate the surface chlorophyll concentration of each drifting water mass; then, assuming that the j-th pixel corresponds to the i-th drifting water mass, a structure of the form... The linear regression model; Let be the vegetation index of the j-th pixel; The spectral response coefficient of algae is also known as the regression coefficient. Let be the surface chlorophyll concentration of the i-th drifting water mass, i.e., the variable. The background parameter for water turbidity is the bias.

[0109] Furthermore, the pairing coefficients between drifting water masses and neighboring pixels are used as weights. The regression coefficients and biases in the linear regression model are then solved using the weighted least squares method to minimize the weighted sum of squared residuals. Solve to obtain and .

[0110] The graded early warning module 104 is used to determine the graded early warning threshold based on the algal spectral response coefficient and the water turbidity background parameters, and to carry out graded early warning of algal blooms based on the vegetation index of each pixel.

[0111] After obtaining the algal spectral response coefficient and water turbidity background parameters, the static concentration values ​​specified in the environmental monitoring standards can be further calculated and converted into vegetation indices under satellite spectra. This allows for the determination of graded early warning thresholds, and the implementation of graded early warnings for algal blooms based on the vegetation index of each pixel.

[0112] Preferably, in one embodiment of the present invention, the method for obtaining the graded early warning threshold includes:

[0113] A predetermined number of risk levels are determined. The spectral response coefficient of algae is weighted using the predetermined warning concentration of each risk level. The weighted result is then fused with the background parameters of water turbidity as the graded warning threshold for the corresponding risk level.

[0114] As an example, the preset quantity is set to 3, divided into mild risk level, moderate risk level and severe risk level; then, based on the local environmental emergency standards, the preset warning concentration for each risk level is determined, such as the upper limit of chloroplast concentration corresponding to the risk level, which will not be elaborated further; then, the preset warning concentration for each risk level is multiplied by the algal spectral response coefficient, the product is added to the water turbidity background parameter, and the sum is used as the graded warning threshold for the corresponding risk level.

[0115] Preferably, in one embodiment of the present invention, considering the potential spatial heterogeneity of hydrodynamic and optical characteristics in different areas of a lake, the greater the uncertainty in the inversion of areas farther from the actual drift trajectory, the lower the accuracy of subsequent graded early warnings may be. Therefore, it is also necessary to determine the early warning attributes; then, the risk level is determined by combining the vegetation index; thus, the graded early warning for algal blooms based on the vegetation index of each pixel includes:

[0116] In satellite spectral imagery, for each pixel, the warning attribute of the pixel is marked based on the spatial distance between the pixel and the coordinates of the nearest drift trajectory. The warning attribute with a spatial distance greater than a preset threshold is a confirmed warning, and the warning attribute with a spatial distance less than or equal to the preset threshold is a suspected warning. The graded warning interval is determined based on the graded warning thresholds of adjacent risk levels. For each pixel, when the vegetation index is within the corresponding graded warning interval, the pixel is marked with the corresponding risk level. The graded warning of algal blooms is carried out based on the warning attribute and risk level of the pixel.

[0117] As an example, in satellite spectral imagery, for each pixel, the spatial distance is first measured using Euclidean distance, with a preset threshold ranging from 5 to 10 km. In this embodiment, 5 km is used. When the distance to the drift trajectory coordinates of the nearest drifting water mass is less than 5 km, it can be considered that the matching and inference accuracy of the pixel is relatively higher, and its graded warning accuracy is also relatively higher. Therefore, its warning attribute is set to a definite warning, and the rest are set to a suspected warning.

[0118] Then, the graded warning intervals are determined, for example, [0, A1], (A1, B1], (B1, C1], (C1, ∞); where A1 is the graded warning threshold corresponding to the mild risk level, B1 is the graded warning threshold corresponding to the moderate risk level, and C1 is the graded warning threshold corresponding to the severe risk level; then, the risk level is marked based on the vegetation index, and waterlogging graded warnings are carried out in combination with its warning attributes.

[0119] For example, when a pixel's vegetation index is in the range [0, A1] and its warning attribute is a confirmed warning, the algal bloom warning result for that pixel is confirmed to be without risk; when a pixel's vegetation index is in the range (A1, B1) and its warning attribute is a suspected warning, the algal bloom warning result for that pixel is suspected to be slightly risky; and so on. When there is no risk, no labeling is required; implementers can use different colors to label different risk levels, or they can use solid-color patches and semi-transparent patches to distinguish warning attributes.

[0120] The final output is a graded early warning map of cyanobacterial blooms, which includes clear risk levels and warning attributes (confidence indicators).

[0121] In summary, this invention first acquires the vegetation index of each pixel in the satellite spectral image of the lake surface and determines the time window based on the satellite imaging time; it then acquires water quality and wind data from ground stations at each acquisition time within the time window; based on the wind conditions, it extrapolates the drift trajectory coordinates of each drifting water mass at the satellite imaging time and determines the cumulative turbulence energy, thereby acquiring the surface algae retention index of each drifting water mass; further, it constructs a flow-direction-constrained drift model for the drifting water mass and solves the model to evaluate the pairing relationship between the drifting water mass and neighboring pixels; further, it combines the surface algae retention index of each drifting water mass with the corresponding water quality data to acquire the algae spectral response coefficient and water turbidity background parameters, thereby determining the graded early warning threshold, and conducting graded early warning of algal blooms based on the vegetation index of each pixel. This invention addresses the inconsistency between fixed-point monitoring and satellite observation in terms of spatiotemporal scale and physical space by reconstructing the wind-driven flow field and introducing turbulent mixing constraints. It also decouples the algal spectral response coefficient from the water turbidity background parameters and adaptively adjusts the graded early warning threshold, thereby improving the accuracy of graded early warning for cyanobacterial blooms.

[0122] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A graded early warning system for cyanobacterial blooms combining vegetation index and water quality parameters, characterized in that, The system includes: Data acquisition module: used to acquire the vegetation index of each pixel in the satellite spectral image of the lake surface and determine the time window based on the satellite imaging time; acquire water quality data and wind condition data of ground stations at each acquisition time within the time window; Drift trajectory extrapolation module: used to determine the drift trajectory coordinates of drifting water masses at ground stations at each acquisition time based on wind data, and to determine the cumulative turbulent energy of each drifting water mass at the satellite imaging time; The physical constraint inversion module is used to obtain the surface algae retention index of each drifting water mass based on the accumulated turbulent energy, construct a flow direction-constrained drift model for the drifting water mass, solve the model to obtain the pairing coefficient between each drifting water mass and each neighboring pixel at the drift trajectory coordinates in the satellite spectral image; based on the surface algae retention index of each drifting water mass and the corresponding water quality data, combined with the vegetation index of each neighboring pixel and its pairing weight with each drifting water mass, the algae spectral response coefficient and water turbidity background parameters are obtained. The graded early warning module is used to determine the graded early warning threshold based on the algal spectral response coefficient and the water turbidity background parameters, and to conduct graded early warning of algal blooms based on the vegetation index of each pixel.

2. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the time window includes: Based on the lake flow velocity and the coverage radius of the ground station, the retrospective duration is obtained; taking the satellite imaging time as the center, the time periods of the retrospective duration are obtained on both sides of the time series, and the time range corresponding to the two time periods is used as the time window.

3. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the drift trajectory coordinates includes: At each acquisition moment within the time window, the lake surface velocity vector is determined based on the wind condition data, the preset wind flow conversion coefficient, and the preset flow direction deflection matrix; the duration between adjacent acquisition moments is taken as the drift duration, and the drift displacement is determined based on the lake surface velocity vector and the drift duration. For each drifting water mass, within the time window, the drift trajectory coordinates for the next adjacent acquisition time are determined based on the superposition of the location coordinates of the ground station and the drift displacement at the first acquisition time. This process is repeated to obtain the drift trajectory coordinates for each acquisition time.

4. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the accumulated turbulent energy includes: At each acquisition moment within the time window, the wind and wave stirring power coefficient is determined based on the wind condition data. The duration between adjacent acquisition moments is taken as the stirring duration. The wind and wave stirring power coefficient and the stirring duration are combined to obtain turbulent energy. For each drifting water mass, within the time window, the cumulative turbulent energy of the drifting water mass at the time of satellite imaging is determined based on the sum of the initial turbulent energy and the turbulent energy at all acquisition times prior to the satellite imaging time.

5. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the surface algae retention index includes: By comparing the cumulative turbulent energy of different drifting water masses, the relative turbulent mixing intensity coefficient of each drifting water mass is obtained; the negative correlation normalization result of the relative turbulent mixing intensity coefficient is used as the surface algae retention index of each drifting water mass.

6. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the flow direction constraint drift model includes: Using the pairing coefficient between each drifting water mass and each neighboring pixel as the optimization variable, a spatial matching cost term is determined by combining the distribution relationship between each drifting water mass and each neighboring pixel at the corresponding drift trajectory coordinates. A ground distribution constraint term is determined by combining the water quality data of each drifting water mass and the surface algae retention index. A satellite observation constraint term is determined by combining the positional distribution of the neighboring pixels. The spatial matching cost term, the ground distribution constraint term, and the satellite observation constraint term are fused to construct a flow direction constrained drift model.

7. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 6, characterized in that, The method for obtaining the spatial matching cost term includes: For each drifting water mass, the displacement vector pointing to each neighboring cell is determined based on the drift trajectory coordinates of the drifting water mass, and the displacement vector is decomposed into a drift component along the flow and a drift component vertically. The drift cost parameter between the drifting water mass and each neighboring pixel is obtained by weighted summing of the modulus of the drifting component along the flow and the modulus of the vertical drift component; the weighting weight of the modulus of the vertical drift component is greater than the weighting weight of the modulus of the drifting component along the flow. By weighting the optimization variables of the flow-direction constraint drift model using the drift cost parameter, the spatial matching cost term is obtained.

8. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The methods for obtaining the algal spectral response coefficient and the water turbidity background parameters include: The water quality data includes at least chlorophyll concentration; for each drifting water mass, the chlorophyll concentration is weighted using the surface algae retention index, and the weighted result is taken as the surface chlorophyll concentration. A linear regression model of vegetation index for pixels in satellite spectral imagery is constructed; wherein, the variable in the linear regression model is the surface chlorophyll concentration, the regression coefficient is the algal spectral response coefficient, and the bias is the water turbidity background parameter. Using the pairing coefficients between each drifting water mass and each neighboring pixel as weights, the regression coefficients and biases in the linear regression model are solved based on the weighted least squares method to obtain the algal spectral response coefficients and water turbidity background parameters.

9. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 1, characterized in that, The method for obtaining the graded early warning threshold includes: A preset number of risk levels are determined, and the spectral response coefficient of the algae is weighted using the preset warning concentration of each risk level. The weighted result is then fused with the water turbidity background parameter as the graded warning threshold for the corresponding risk level.

10. The cyanobacterial bloom classification and early warning system combining vegetation index and water quality parameters according to claim 2, characterized in that, Algal bloom classification and early warning based on vegetation index of each pixel includes: In the satellite spectral image, for each pixel, the warning attribute of the pixel is marked based on the spatial distance between the pixel and the nearest drift trajectory coordinates, wherein the warning attribute with a spatial distance greater than a preset threshold is a confirmed warning, and the warning attribute with a spatial distance less than or equal to the preset threshold is a suspected warning. The graded warning interval is determined based on the graded warning threshold of adjacent risk levels. For each pixel, when the vegetation index is within the corresponding graded warning interval, the pixel is marked as the corresponding risk level. Algal bloom grading and early warning are based on pixel-level warning attributes and risk levels.

Images