Blue-green algae bloom weather forecasting and early warning method and system based on hierarchical fusion algorithm
Through a method based on hierarchical fusion algorithm, combined with satellite image data and refined grid point numerical forecasting, a meteorological risk index prediction model of cyanobacteria water blooms was constructed, which solved the problems of difficult parameters and poor accuracy in the existing technology, and realized cyanobacteria water blooms. high-precision prediction and risk assessment.
Patent Information
- Application Number
- CN202410914881.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2024-11-12
AI Technical Summary
The existing cyanobacteria bloom prediction and early warning technology lacks effective methods, difficult to obtain parameters, insufficient spatial effect and poor accuracy, so it is impossible to effectively predict the occurrence and development of cyanobacteria blooms.
Using a method based on hierarchical fusion algorithm, the meteorological risk of cyanobacteria blooms is constructed through preprocessing of satellite image data, normalized vegetation index calculation, hierarchical analysis method and fuzzy mathematical judgment matrix method, combined with refined grid point numerical forecasting and machine learning. An exponential prediction model to achieve high-precision distribution and area prediction.
It improves the accuracy and process efficiency of the prediction results, enhances the spatial effect, and can efficiently evaluate the meteorological risks of cyanobacteria blooms, and provides accurate distribution and area prediction.
Smart Images

Figure CN118942228A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of meteorological forecasting and warning, and in particular to a cyanobacteria bloom meteorological forecasting and warning method based on a hierarchical fusion algorithm and a cyanobacteria bloom meteorological forecasting and warning system based on a hierarchical fusion algorithm. Background Art
[0002] Cyanobacteria are one of the earliest aquatic organisms on Earth, and they play an important role in the material cycle and energy flow of aquatic ecosystems. Excessive reproduction of cyanobacteria can lead to cyanobacterial blooms in water bodies, deterioration of water quality, and huge losses to regional landscapes, economy, and ecology. Cyanobacterial blooms have become the most common water environment problem in shallow lakes at home and abroad. Studies have found that at this stage, endogenous nutrients in most inland lakes are sufficient to support the growth of cyanobacteria. Nutrient conditions are no longer the limiting factor for the growth of cyanobacteria, and environmental meteorological conditions may become the main limiting factor. Environmental meteorological conditions generally directly or indirectly affect the number, community, distribution, and life cycle of algae by affecting the stratification of lake water, heat transfer and exchange, and the availability of light and nutrients. Studies have confirmed that meteorological factors such as temperature, wind, light, and precipitation will have an important impact on the growth of cyanobacteria and the formation of water blooms. Therefore, the development of meteorological forecasting and early warning technology for cyanobacterial blooms is of great significance to improving the pertinence and scientificity of government prevention and control of cyanobacterial blooms and minimizing the ecological hazards and health risks of cyanobacterial blooms.
[0003] At present, the prediction and early warning technology of cyanobacteria blooms is still in the development stage, and there is a lack of effective and operational cyanobacteria bloom prediction and early warning methods. The only cyanobacteria bloom prediction technologies mostly use chlorophyll a concentration or algae density as the characterization factors of the prediction model, which mainly include two categories: one is the biological growth mechanism driven model based on the growth process of the algae itself, and the other is the mathematical statistics driven model based on the influence of external factors. The mechanism-driven model mainly starts from the process mechanism of cyanobacteria bloom formation, and establishes the relationship between key variables and other predictable variables through biochemical laws. This model has high data requirements, too many independent variables and cannot fully simulate its real growth state. The process is complex and there are too many influencing factors, making it difficult to obtain effective business methods; the mathematical statistics-driven model mainly analyzes the observation data of cyanobacteria blooms and establishes the relationship between characterization factors and influencing factors to achieve cyanobacteria bloom prediction. Although this method does not consider the complex mechanism of blooms, characterization factors such as chlorophyll a concentration or algae density do not necessarily coincide with the occurrence of cyanobacteria blooms. When the chlorophyll a or algae density in the lake is high, if there is no suitable meteorological conditions, cyanobacteria will not float up to form cyanobacteria blooms. In our daily actual monitoring, this situation is found many times. Therefore, neither of these two methods can effectively predict the occurrence and development of cyanobacteria blooms. Summary of the invention
[0004] Based on this, it is necessary to provide a cyanobacteria bloom meteorological forecasting and early warning method and system based on a hierarchical fusion algorithm to address the problems of difficult acquisition of parameters, insufficient spatial effect and poor accuracy of existing cyanobacteria bloom prediction models.
[0005] The present invention is implemented by the following technical scheme: A cyanobacteria bloom meteorological forecasting and early warning method based on a hierarchical fusion algorithm comprises the following steps:
[0006] S1: Collect satellite image data of cyanobacteria blooms.
[0007] S2: Preprocess satellite image data.
[0008] S3: Calculate the normalized difference vegetation index NDVI value of each satellite image after preprocessing, and identify the distribution range of cyanobacteria blooms through a preset threshold.
[0009] S4: Based on satellite image data, a model prediction method is used to predict the meteorological risk of cyanobacteria blooms. The model prediction method is as follows:
[0010] S41: The analytic hierarchy process is used to analyze the meteorological driving indicators of cyanobacterial blooms, and the meteorological risks of cyanobacterial blooms are divided into multiple levels according to the factor layer range of the meteorological driving indicators of cyanobacterial blooms.
[0011] S42: The judgment matrix method in fuzzy mathematics is used to determine the weights of indicators at each criterion level in the cyanobacteria bloom risk index prediction model.
[0012] S43: Combined with the refined grid numerical forecasting method, the meteorological index impact score and the comprehensive score of the cyanobacteria bloom risk index are calculated.
[0013] S44: Based on the historical long-term satellite monitoring data of cyanobacteria blooms, the frequency distribution of cyanobacteria blooms in lake areas is calculated. According to the degree of spatial distribution of cyanobacteria blooms, a correction coefficient for the frequency of cyanobacteria blooms is formulated to correct the comprehensive score of the cyanobacteria bloom risk index.
[0014] S5: Based on the meteorological risk index prediction model of cyanobacteria bloom, combined with the historical long-term satellite monitoring data of cyanobacteria bloom, a cyanobacteria bloom information prediction model was constructed through machine learning methods, and the cyanobacteria bloom distribution prediction product was obtained using an integrated forecasting approach.
[0015] The above-mentioned cyanobacteria bloom meteorological forecast and early warning method takes the cyanobacteria bloom directly observed by satellite as the core analysis object, combines the refined grid numerical weather forecast method, and uses the hierarchical analysis method to calculate the meteorological risk index of cyanobacteria bloom and construct a cyanobacteria bloom meteorological risk index prediction model. By integrating the model with the machine learning algorithm, high-precision prediction of the distribution and area of cyanobacteria bloom is achieved. The present invention outputs the prediction results through the model, and uses the reinforcement learning method to train and optimize the model, which not only improves the accuracy of the prediction results, but also improves the efficiency, stability and spatial effect of the prediction process.
[0016] Furthermore, in step S2, preprocessing includes projection, cropping, geometric precision correction and atmospheric correction. Among them, cropping is used to remove the part outside the specific study area in the remote sensing image. Geometric precision correction is used to correct image deviation. Atmospheric correction is used to eliminate the influence of atmospheric molecular scattering and absorption and aerosol scattering on remote sensing images. By preprocessing remote sensing images, the amount of data processing is reduced and the amount of deviation is eliminated, thereby improving the accuracy of remote sensing images in reflecting the characteristics of the objects in the study area.
[0017] Furthermore, in step S3, the normalized difference vegetation index NDVI value calculation formula is as follows:
[0018] NDVI=(ρ nir -ρ red ) / (ρ nir +ρ red )
[0019] where ρ nir represents the reflectivity of the satellite near-infrared band, ρ red It represents the reflectivity of the red light band, and NDVI = -0.012 is generally taken as the threshold for identifying cyanobacteria blooms.
[0020] Further, in step S42, a judgment matrix A is constructed, the corresponding eigenvalues are calculated, and the eigenvalues are normalized to obtain the weights W of the relative importance of each indicator of this layer relative to each indicator of the upper layer. i '. The calculation formula is as follows:
[0021]
[0022] In the formula, w i Represents the eigenvalue of the i-th layer of the judgment matrix A.
[0023] Further, in step S43, the risk model of the meteorological influencing factor is expressed as:
[0024]
[0025] Where U is the function expression of the meteorological risk index of cyanobacteria bloom, f(T) is the impact score of temperature forecast factor, f(V) is the impact score of wind speed forecast factor, f(M) is the impact score of precipitation forecast factor, Wi and Ci are the weights of each influencing factor, the expert scoring results based on the relative importance of cyanobacteria bloom meteorological indicators and Saaty, respectively.
[0026] Further, in step S44, the calculation process of the cyanobacteria bloom frequency correction coefficient is as follows:
[0027] S441: The normalized difference vegetation index NDVI value is synthesized pixel by pixel to obtain the long-term change trend and distribution characteristics of cyanobacteria.
[0028] S442: According to the long-term trend and distribution characteristics of cyanobacteria, a two-dimensional entropy method that can reflect the spatial distribution characteristics is introduced on the basis of one-dimensional entropy, and the two-dimensional entropy weight of the frequency distribution map of cyanobacteria blooms is calculated by the entropy weight method. The two-dimensional entropy weight calculation formula is as follows:
[0029] P i,j =f(i,j) / N 2
[0030] Where f(i, j) is the frequency of occurrence of the feature binary (i, j), N is the scale of the image, i represents the pixel value, and j represents the mean of the neighborhood value.
[0031] S443: Normalize the two-dimensional entropy results and formulate the calculation formula for the cyanobacteria frequency correction coefficient as follows:
[0032]
[0033] In the formula, C i,j It represents the result of dividing the two-dimensional entropy by the position of the cyanobacteria bloom frequency (i, j), represents the normalized two-dimensional entropy result of the cyanobacteria bloom frequency at position (i, j), and represent the maximum and minimum values in the image respectively. The two-dimensional entropy result P represents the meteorological risk correction index of cyanobacteria bloom, and C p Represents the rasterized cyanobacteria frequency correction factor.
[0034] Through the analytic hierarchy process, the meteorological risk of cyanobacteria blooms is divided into different levels according to the meteorological driving indicators of cyanobacteria blooms. The principle of membership in fuzzy mathematics and the judgment matrix method are used to quantify the level of meteorological driving mechanism of cyanobacteria blooms. Combined with the refined grid numerical forecasting method, the meteorological risk status of cyanobacteria blooms is evaluated by quantitatively scoring the impact level of meteorological indicators and the cyanobacteria bloom risk index, which has the advantages of high prediction accuracy and strong timeliness.
[0035] Further, in step S5, the process of constructing a cyanobacteria bloom information prediction model using a machine learning method is as follows:
[0036] S51: Five machine learning methods, including regression support vector machine (SVM), random forest (RF), artificial neural networks (ANNs), naive Bayes (NB) and adaptive boosting (Adaboost), were used for model training. The correlation coefficient R and root mean square error RMSE were used to evaluate the model. Then, the ten-fold cross validation method was used to construct five cyanobacteria bloom information prediction models.
[0037] S52: Based on the idea of integrated forecasting, a high-precision integrated prediction model for cyanobacteria blooms is constructed. The formula is as follows:
[0038]
[0039] In the formula, F e is the final integrated prediction value, C Ri is the correlation coefficient, w i is the weight function of the ith ensemble member, F si , are the predicted values of each model and their average values, T oi , are the actual monitoring and its average value, respectively, and i is the number of artificial intelligence prediction models used.
[0040] S53: Calculate the area information of cyanobacteria bloom. The formula is:
[0041] S b =S r N
[0042] In the formula, S b is the area of cyanobacteria. N is the total number of pixels distributed by cyanobacteria blooms. r is the single-pixel cyanobacterial bloom area.
[0043] Five machine learning methods were used to train and optimize the cyanobacteria bloom information prediction model. The results of the five artificial intelligence prediction models were processed by an integrated forecasting method to effectively improve the forecast timeliness and accuracy of the cyanobacteria bloom information prediction model.
[0044] The present invention also provides a cyanobacteria bloom meteorological forecasting and early warning system based on a hierarchical fusion algorithm, comprising a collection module, a data preprocessing module, a cyanobacteria bloom meteorological risk index prediction model and an output module.
[0045] The acquisition module is used to collect satellite image data of cyanobacteria blooms.
[0046] The data preprocessing module is used to preprocess the satellite image data, calculate the normalized difference vegetation index NDVI value of each preprocessed satellite image, and identify the distribution range of cyanobacteria blooms through a preset threshold.
[0047] The cyanobacteria bloom meteorological risk index prediction model includes a driving indicator module, a judgment matrix module, a grid value module, and a coefficient correction module. The driving indicator module classifies meteorological risks by analyzing the cyanobacteria bloom meteorological impact factor layer. The judgment matrix module determines the weight of each indicator in the cyanobacteria bloom meteorological risk index prediction model. The grid value module quantifies the weight of each criterion layer indicator in the prediction model to generate a risk prediction score. The coefficient correction module proposes a cyanobacteria bloom frequency correction coefficient based on the historical data of the spatial distribution of cyanobacteria blooms, and corrects the comprehensive score of the cyanobacteria bloom risk index.
[0048] The output module is used to integrate the cyanobacteria bloom meteorological risk index prediction model and machine learning methods to construct a cyanobacteria bloom information prediction model, and the prediction results are obtained using an integrated forecasting method.
[0049] Compared with the prior art, the present invention has the following beneficial effects:
[0050] The present invention adopts a model prediction method, takes the cyanobacteria bloom directly observed by satellite as the core analysis object, and divides the meteorological risk of cyanobacteria bloom into different levels according to the meteorological driving indicators of cyanobacteria bloom by using the hierarchical analysis method according to the refined grid numerical weather forecast method. The principle of membership in fuzzy mathematics and the judgment matrix method are used to realize the quantification of the level of meteorological driving mechanism of cyanobacteria bloom. By quantitatively scoring the impact level of meteorological indicators and the cyanobacteria bloom risk index, a cyanobacteria bloom meteorological risk index prediction model is constructed to evaluate the meteorological risk status of cyanobacteria bloom. The model is trained and optimized by the reinforcement learning method to realize the high-precision prediction of the distribution and area of cyanobacteria bloom. The present invention not only improves the accuracy of the prediction results, but also improves the efficiency, stability and spatial effect of the prediction process. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 This is a step diagram of a cyanobacteria bloom meteorological forecasting and early warning method based on a hierarchical fusion algorithm according to Embodiment 1 of the present invention;
[0052] Figure 2 for Figure 1 Flow chart of the meteorological forecast and early warning method of cyanobacteria bloom based on hierarchical fusion algorithm;
[0053] Figure 3 The result diagram of the meteorological forecast risk index of blue algae bloom in Example 1;
[0054] Figure 4 This is a diagram showing the distribution prediction results of cyanobacteria blooms in Example 1;
[0055] Figure 5 This is a graph showing the actual monitoring results of the cyanobacteria bloom in Example 1;
[0056] Figure 6 Based on Figure 1 Schematic diagram of the structure of the cyanobacteria bloom meteorological forecast and early warning system based on the hierarchical fusion algorithm. DETAILED DESCRIPTION
[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0058] It should be noted that when a component is referred to as being "mounted on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component at the same time. When a component is considered to be "fixed to" another component, it may be directly fixed on the other component or there may be a central component at the same time.
[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.
[0060] Example 1: Please refer to Figure 1 and Figure 2 , Figure 1 This is a step diagram of a cyanobacteria bloom meteorological forecasting and early warning method based on a hierarchical fusion algorithm according to Embodiment 1 of the present invention; Figure 2 for Figure 1 Flow chart of the meteorological forecast and early warning method for cyanobacteria bloom based on hierarchical fusion algorithm. The meteorological forecast and early warning method for cyanobacteria bloom includes:
[0061] S1: Collect satellite image data of cyanobacteria blooms. This embodiment uses domestic Fengyun series and EOS / MODIS satellite remote sensing image data from 2003 to 2023.
[0062] S2: Preprocess the satellite image data, including projection, cropping, geometric correction, and atmospheric correction. Projection is used to unify the projection methods of different map data and data from different sources; cropping is used to remove the parts outside the specific study area in the remote sensing image; geometric correction is used to correct image deviations, such as the deviation between the geographic information in the image and the actual situation caused by the sensor attitude and the rotation of the earth; atmospheric correction is used to eliminate the effects of atmospheric molecular scattering and absorption and aerosol scattering on remote sensing images. By preprocessing remote sensing images, the amount of data processing is reduced and the amount of deviation is eliminated, so as to improve the accuracy of remote sensing images in reflecting the characteristics of the objects in the study area.
[0063] S3: Calculate the normalized vegetation index NDVI value of each satellite image after preprocessing, and identify the distribution range of cyanobacteria blooms through a preset threshold; the calculation formula of the normalized vegetation index NDVI value is as follows:
[0064] NDVI=(ρ nir -ρ red ) / (ρ nir +ρ red )
[0065] where ρ nir represents the reflectivity of the satellite near-infrared band, ρ red Indicates the reflectivity of the red light band, and NDVI = -0.012 is generally taken as the threshold for identifying cyanobacteria blooms. A negative NDVI value indicates that the ground is covered with clouds, water, snow, etc., while a positive NDVI value indicates that there is vegetation coverage, and it increases with the increase in coverage.
[0066] S4: Based on satellite image data, a model prediction method is used to predict the meteorological risk of cyanobacteria blooms; the model prediction method is as follows:
[0067] S41: The analytic hierarchy process is used to analyze the meteorological driving indicators of cyanobacterial blooms. According to the factor layer range of the meteorological driving indicators of cyanobacterial blooms, the meteorological risks of cyanobacterial blooms are divided into multiple levels. According to the existing research results and experience, the factor layer range classification of the suitability of cyanobacterial bloom meteorological driving is shown in Table 1:
[0068] Table 1: Factor layer range classification table
[0069]
[0070]
[0071]
[0072] S42: Use the judgment matrix method in fuzzy mathematics to determine the weights of indicators at each criterion level in the cyanobacteria bloom risk index prediction model; according to the hierarchical division of the degree of suitability of meteorological driving of cyanobacteria blooms, compare each element by constructing a judgment matrix to determine the weights of different indicators at each criterion level. The judgment matrix A represents the relative importance of all indicators at a certain level to an indicator at the previous level. Assuming that all indicators a1, a2, ..., an at a certain level are compared pairwise with respect to an indicator B at the previous level, the judgment matrix A is expressed as:
[0073] A <![CDATA[a1]]> <![CDATA[a2]]> ... <![CDATA[a n ]]> <![CDATA[a1]]> <![CDATA[r 11 ]]> <![CDATA[r 12 ]]> ... <![CDATA[r 1n ]]> <![CDATA[a2]]> <![CDATA[r 21 ]]> <![CDATA[r 22 ]]> ... <![CDATA[r 2n ]]> ... ... ... <![CDATA[a n ]]> <![CDATA[r n1 ]]> <![CDATA[r n2 ]]> ... <![CDATA[r nn ]]>
[0074] r ij Indicates index a i and indicator a j When compared with indicator B, a i and a j The membership degree has the fuzzy relationship "... is more important than...". The Saaty relative importance scale is used to quantitatively express the relative importance of any two indicators relative to a certain criterion. The Saaty relative importance scale is shown in Table 2:
[0075] Table 2: Saaty relative importance ranking table
[0076]
[0077] Note: If the importance judgment scale of factor i and factor j is aij, then the importance scale of factor j and factor i is aji=1 / aij to establish a judgment matrix.
[0078] After establishing the judgment matrix A, the corresponding eigenvalues are calculated, and the eigenvalues are normalized to obtain the weight Wi of the relative importance of each indicator in this layer relative to an indicator in the upper layer. The specific solution steps are as follows:
[0079] Use linear algebra methods, such as characteristic polynomial method, power method, and Jacobi iteration method, to calculate the eigenvalues and eigenvectors of the judgment matrix A:
[0080]
[0081] Then solve:
[0082]
[0083] Obtain the eigenvalues of the corresponding matrix and normalize the calculated eigenvalues as follows:
[0084]
[0085] In the formula, w i represents the eigenvalue of the i-th layer of the judgment matrix A, Wi 'Indicates the relative importance of each index in this layer relative to the indexes in the upper layer after the eigenvalues are normalized.
[0086] S43: Combined with the refined grid numerical forecast method, the meteorological index impact score and the comprehensive score of the blue algae bloom risk index are calculated; based on the quantitative data of the evaluation factors and their weights, the weighted average method is used to obtain the risk model of the meteorological impact factors. The formula is as follows:
[0087]
[0088] Where U is the function expression of the meteorological risk index of cyanobacteria bloom, f(T) is the impact score of temperature forecast factor, f(V) is the impact score of wind speed forecast factor, f(M) is the impact score of precipitation forecast factor, Wi, C i Respectively represent the weight of each influencing factor, the expert scoring results based on the meteorological indicators of cyanobacteria blooms and Saaty's relative importance.
[0089] S44: Based on the historical long-term satellite monitoring data of cyanobacteria blooms, the frequency distribution results of cyanobacteria blooms in lake areas are calculated; according to the degree of spatial distribution of cyanobacteria blooms, the cyanobacteria bloom frequency correction coefficient is formulated to correct the comprehensive score of cyanobacteria bloom risk index; the calculation process of cyanobacteria bloom frequency correction coefficient is as follows:
[0090] S441: Process the daily cyanobacteria bloom information from 2003 to 2023 according to the Normalized Difference Vegetation Index (NDVI) value method, and synthesize the frequency of cyanobacteria blooms pixel by pixel to obtain the long-term change trend and distribution characteristics of cyanobacteria;
[0091] S442: According to the long-term trend and distribution characteristics of cyanobacteria, a two-dimensional entropy method that can reflect the spatial distribution characteristics is introduced on the basis of one-dimensional entropy, and the two-dimensional entropy weight result of the cyanobacteria bloom frequency distribution map is calculated by the entropy weight method; the neighborhood frequency mean of the cyanobacteria bloom frequency distribution map is selected as the spatial feature quantity of the frequency distribution, and it is combined with the pixel frequency value of the image to form a feature binary group, recorded as (i, j), where i represents the pixel value, j represents the mean value of the neighborhood value, and the two-dimensional entropy weight calculation formula is as follows:
[0092] P i,j =f(i,j) / N 2
[0093] Where f() is the frequency of occurrence of feature binary (i, j), and N is the scale of the image;
[0094] S443: Normalize the two-dimensional entropy results and formulate the calculation formula for the cyanobacteria frequency correction coefficient as follows:
[0095]
[0096] In the formula, C i,j It represents the result of dividing the two-dimensional entropy by the position of the cyanobacteria bloom frequency (i, j), represents the two-dimensional entropy result of the cyanobacteria bloom frequency normalized at position (i, j), C max With C min They represent the maximum and minimum values in the image respectively. The two-dimensional entropy result P represents the meteorological risk correction index of cyanobacteria bloom, and C p Represents the rasterized cyanobacteria frequency correction factor.
[0097] This embodiment uses the analytic hierarchy process to classify the meteorological risk of cyanobacteria blooms into different levels according to the meteorological driving indicators of cyanobacteria blooms, and adopts the principle of membership in fuzzy mathematics and the judgment matrix method to achieve the quantification of the level of meteorological driving mechanism of cyanobacteria blooms. Combined with the refined grid point numerical forecasting method, the meteorological risk status of cyanobacteria blooms is evaluated by quantitatively scoring the impact level of meteorological indicators and the cyanobacteria bloom risk index. This method has high prediction accuracy and strong timeliness.
[0098] S5: Based on the meteorological risk index prediction model of cyanobacteria bloom, combined with the historical long-term satellite monitoring data of cyanobacteria bloom, a cyanobacteria bloom information prediction model was constructed through machine learning methods, and the cyanobacteria bloom distribution prediction product was obtained using an integrated forecasting approach.
[0099] The process of constructing a cyanobacteria bloom information prediction model using machine learning methods is as follows:
[0100] S51: Five machine learning methods, including regression support vector machine (SVM), random forest (RF), artificial neural network (ANNs), naive Bayes (NB) and adaptive boosting (Adaboost), were used for model training. The correlation coefficient R and root mean square error RMSE were used to evaluate the model. Then, the ten-fold cross validation method was used to construct five cyanobacteria bloom information prediction models. The normalized difference vegetation index NDVI value method was used to roll the long-term satellite remote sensing cyanobacteria bloom historical data from 2003 to 2023 according to the time series, and the processed data were matched with the original data pixel by pixel to obtain a sample data set, which was randomly divided into 10 sub-sample sets for machine learning training.
[0101] S52: Based on the idea of integrated forecasting, a high-precision integrated prediction model for cyanobacteria blooms is constructed. The formula is as follows:
[0102]
[0103] In the formula, F e is the final integrated prediction value, C Ri is the correlation coefficient, w i is the weight function of the ith ensemble member, F si , are the predicted values of each model and their average values, T oi , are the actual monitoring and its average value, respectively, and i is the number of artificial intelligence prediction models used;
[0104] S53: Calculate the area information of cyanobacteria bloom; the formula is:
[0105] S b =S r N
[0106] In the formula, S b is the cyanobacteria area; N is the total number of pixels distributed by cyanobacteria blooms; S r is the single-pixel cyanobacteria bloom area.
[0107] This embodiment uses five machine learning methods to train and optimize the cyanobacteria bloom information prediction model, and processes the results of the five artificial intelligence prediction models through an integrated forecasting method, thereby effectively improving the forecast timeliness and accuracy of the cyanobacteria bloom information prediction model.
[0108] In order to solve the problems of the difficulty in obtaining parameters and insufficient spatial effect of the existing cyanobacteria bloom prediction model, this embodiment adopts a model prediction method, takes the cyanobacteria bloom directly observed by satellite as the core analysis object, uses the hierarchical analysis method to quantify and calculate the meteorological risk index of cyanobacteria bloom, and then combines the refined grid numerical weather forecast method to build a cyanobacteria bloom meteorological risk index prediction model. By training the model with machine learning methods, high-precision prediction of the distribution and area of cyanobacteria bloom is achieved; not only the accuracy of the prediction results is improved, but also the efficiency, stability and spatial effect of the prediction process are improved.
[0109] Experimental verification
[0110] First, based on the ECMWF fine grid numerical weather forecast data on October 11, 2023 and October 12, 2023, the cyanobacteria bloom meteorological risk index prediction model integrated with artificial intelligence is used to predict the spatial distribution of cyanobacteria blooms on October 11 and 12, 2023. Then, the distribution of cyanobacteria blooms actually monitored by satellite remote sensing images on October 11, 2023 and October 12, 2023 is compared and verified.
[0111] See also Figure 3 , Figure 4 and Figure 5 , Figure 3 This is a schematic diagram of the meteorological forecast risk index of cyanobacteria bloom in Example 1. In the figure, the larger the value, the higher the risk of cyanobacteria bloom. On the contrary, the smaller the value, the lower the risk of cyanobacteria bloom. Figure 4 This is a diagram showing the distribution prediction results of cyanobacteria blooms in Example 1; Figure 5This is the actual detection result of cyanobacteria bloom. Figure 4 and Figure 5 It can be found that the predicted results of cyanobacteria bloom distribution are basically consistent with the actual satellite observations of cyanobacteria bloom distribution, and the overall prediction effect is good.
[0112] Example 2: Please refer to Figure 6 , which is based on Figure 1 Schematic diagram of the structure of the cyanobacteria bloom meteorological forecast and warning system based on the hierarchical fusion algorithm. Based on the above-mentioned cyanobacteria bloom meteorological forecast and warning method based on the hierarchical fusion algorithm, in order to achieve high-precision prediction of cyanobacteria blooms, this embodiment provides a cyanobacteria bloom meteorological forecast and warning system based on the hierarchical fusion algorithm. The meteorological forecast and warning system includes an acquisition module, a data preprocessing module, a cyanobacteria bloom meteorological risk index prediction model, and an output module.
[0113] The acquisition module is used to collect satellite image data of cyanobacteria blooms. The satellite image data uses the Fengyun series and EOS / MODIS satellite remote sensing image data from 2003 to 2023, and other satellite data (such as Landsat, HJ and Gaofen, etc.) can also be used.
[0114] The data preprocessing module is used to preprocess the satellite image data. Preprocessing includes projection, cropping, geometric correction and atmospheric correction. The normalized vegetation index NDVI value of each satellite image after preprocessing is calculated, and the distribution range of cyanobacteria blooms is identified by a preset threshold. Generally, NDVI = -0.012 is taken as the cyanobacteria bloom identification threshold. A negative NDVI value indicates that the ground is covered with clouds, water, snow, etc., while a positive value indicates that there is vegetation coverage, and it increases with the increase of coverage.
[0115] The cyanobacteria bloom meteorological risk index prediction model is used to output prediction results based on preprocessed satellite image data.
[0116] The cyanobacteria bloom meteorological risk index prediction model includes a driving indicator module, a judgment matrix module, a grid value module, and a coefficient correction module. The driving indicator module classifies meteorological risks by analyzing the cyanobacteria bloom meteorological impact factor layer. The judgment matrix module determines the weight of each indicator in the cyanobacteria bloom meteorological risk index prediction model. The grid value module quantifies the weight of each criterion layer indicator in the prediction model to generate a risk prediction score. The coefficient correction module proposes a cyanobacteria bloom frequency correction coefficient based on the historical data of the spatial distribution of cyanobacteria blooms, and corrects the comprehensive score of the cyanobacteria bloom risk index. Combined with the refined grid numerical prediction method, the weight of each criterion layer indicator in the prediction model is quantified to generate a risk prediction score. Based on the historical long-term satellite monitoring data of cyanobacteria blooms, according to the degree of performance of cyanobacteria blooms in spatial distribution, a cyanobacteria bloom frequency correction coefficient is proposed to correct the comprehensive score of the cyanobacteria bloom risk index.
[0117] The output module is used to integrate the cyanobacteria bloom meteorological risk index prediction model and machine learning methods to build a cyanobacteria bloom information prediction model. The machine learning method uses five machine learning methods, namely regression support vector machine (SVM), random forest (RF), artificial neural network (ANNs), naive Bayes (NB) and adaptive boosting (Adaboost), for model training, and the correlation coefficient R and root mean square error RMSE are used to evaluate the model. Using the ten-fold cross validation method, five cyanobacteria bloom information prediction models are constructed. The ensemble forecasting method is used to obtain the cyanobacteria bloom distribution prediction results.
[0118] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0119] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. A meteorological forecasting and early warning method for cyanobacteria bloom based on a hierarchical fusion algorithm, characterized in that: The steps include: S1: Collect satellite image data of cyanobacteria blooms; S2: preprocessing the satellite image data; S3: Calculate the normalized difference vegetation index NDVI value of each satellite image after preprocessing, and identify the distribution range of cyanobacteria blooms through a preset threshold; S4: Constructing a cyanobacteria bloom meteorological risk index prediction model to predict the cyanobacteria bloom meteorological risk; the meteorological risk prediction method is as follows: S41: Analyze various meteorological driving indicators of cyanobacteria blooms by using the analytic hierarchy process, and divide the meteorological risks of cyanobacteria blooms into multiple levels according to the factor layer range of the meteorological driving indicators of cyanobacteria blooms; S42: The judgment matrix method in fuzzy mathematics is used to determine the weights of the indicators at each criterion level in the prediction model of the meteorological risk index of cyanobacteria bloom; S43: Combined with the refined grid numerical forecast method, the meteorological index impact score and the comprehensive score of the cyanobacteria bloom risk index are calculated; S44: Based on the historical long-term satellite monitoring data of cyanobacteria blooms, the frequency distribution of cyanobacteria blooms in lake areas is calculated; According to the degree of spatial distribution of cyanobacterial blooms, a cyanobacterial bloom frequency correction coefficient is formulated to correct the comprehensive score of the cyanobacterial bloom risk index; S5: Based on the cyanobacteria bloom meteorological risk index prediction model and combined with the historical long-term satellite monitoring data of cyanobacteria bloom, a cyanobacteria bloom information prediction model is constructed through machine learning methods, and the cyanobacteria bloom distribution prediction results are obtained by using an integrated forecasting method.
2. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1 is characterized in that: In step S2, the preprocessing includes projection, cropping, geometric correction and atmospheric correction; wherein the cropping is used to remove the portion outside the specific study area in the satellite image data; the geometric correction is used to correct the image deviation; and the atmospheric correction is used to eliminate the influence of atmospheric molecular scattering and absorption and aerosol scattering on the remote sensing image.
3. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1 is characterized in that: In step S3, the normalized difference vegetation index NDVI value calculation formula is as follows: NDVI=(r nir -r red ) / (ρ nir +r red ) In the formula, ρ nir represents the reflectivity of the satellite near-infrared band, ρ red Indicates the reflectivity of the red light band.
4. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1 is characterized in that: In step S42, a judgment matrix A is constructed, the corresponding eigenvalues are calculated, and the eigenvalues are normalized to obtain the weights W of the relative importance of each indicator of this layer relative to each indicator of the upper layer. i '; The calculation formula is as follows: In the formula, w i Represents the eigenvalue of the i-th layer of the judgment matrix A.
5. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1, characterized in that: In step S43, the risk model of meteorological influencing factors is expressed as: Where U is the function expression of the meteorological risk index of cyanobacteria bloom, f(T) is the impact score of temperature forecast factor, f(V) is the impact score of wind speed forecast factor, f(M) is the impact score of precipitation forecast factor, Wi and Ci are the weights of each influencing factor, the expert scoring results based on the relative importance of cyanobacteria bloom meteorological indicators and Saaty, respectively.
6. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1 is characterized in that: In step S44, the calculation process of the cyanobacteria bloom frequency correction coefficient is as follows: S441: synthesizing the cyanobacteria bloom frequency pixel by pixel on the normalized difference vegetation index NDVI value to obtain the long-term variation trend and distribution characteristics of cyanobacteria; S442: According to the long-term change trend and distribution characteristics of the cyanobacteria, a two-dimensional entropy method that can reflect the spatial distribution characteristics is introduced on the basis of the one-dimensional entropy, and the two-dimensional entropy weight of the cyanobacteria bloom frequency distribution map is calculated by the entropy weight method; the two-dimensional entropy weight calculation formula is as follows: P i,j =f(i,j) / N 2 Where f(i, j) is the frequency of occurrence of the feature binary (i, j), i represents the pixel value, j represents the mean of the neighborhood value, and N is the scale of the image; S443: Normalize the two-dimensional entropy results and formulate the calculation formula for the cyanobacteria frequency correction coefficient as follows: In the formula, C i,j It represents the result of dividing the two-dimensional entropy by the position of the cyanobacteria bloom frequency (i, j), represents the two-dimensional entropy result of the cyanobacteria bloom frequency normalized at position (i, j), C max With C min They represent the maximum and minimum values in the image respectively. The two-dimensional entropy result P represents the meteorological risk correction index of cyanobacteria bloom, and C p Represents the rasterized cyanobacteria frequency correction factor.
7. The method for meteorological forecasting and early warning of cyanobacteria bloom based on hierarchical fusion algorithm according to claim 1, characterized in that: In step S5, the process of constructing a cyanobacteria bloom information prediction model using the machine learning method is as follows: S51: Five machine learning methods, including regression support vector machine (SVM), random forest (RF), artificial neural network (ANNs), naive Bayes (NB) and adaptive boosting (Adaboost), were used for model training, and the correlation coefficient R and root mean square error RMSE were used to evaluate the model; 5 cyanobacteria bloom information prediction models were constructed using the ten-fold cross validation method; S52: Based on the idea of integrated forecasting, a high-precision integrated prediction model for cyanobacteria blooms is constructed. The formula is as follows: In the formula, F e is the final integrated prediction value, C Ri is the correlation coefficient, w i is the weight function of the ith ensemble member, F si , are the predicted values of each model and their average values, T oi , are the actual monitoring and its average value, respectively, and i is the number of artificial intelligence prediction models used; S53: Calculate the area information of cyanobacteria bloom, the formula is: S b =S r N In the formula, S b is the cyanobacteria area; N is the total number of pixels distributed by cyanobacteria blooms; S r is the single-pixel cyanobacteria bloom area.
8. A cyanobacteria bloom meteorological forecast and early warning system based on a hierarchical fusion algorithm, which adopts the cyanobacteria bloom meteorological forecast and early warning method based on a hierarchical fusion algorithm as claimed in any one of claims 1 to 7, characterized in that: The weather forecast and early warning system comprises: A collection module, which is used to collect satellite image data of cyanobacteria blooms; A data preprocessing module, which is used to preprocess the satellite image data; calculate the normalized difference vegetation index NDVI value of each satellite image after preprocessing, and identify the distribution range of cyanobacteria blooms through a preset threshold; The cyanobacteria bloom meteorological risk index prediction model includes a driving index module, a judgment matrix module, a grid value module, and a coefficient correction module; the driving index module classifies meteorological risks by analyzing the cyanobacteria bloom meteorological impact factor layer; the judgment matrix module determines the weight of each indicator in the cyanobacteria bloom meteorological risk index prediction model; the grid value module quantifies the weight of each criterion layer indicator in the prediction model to generate a risk prediction score; the coefficient correction module formulates a cyanobacteria bloom frequency correction coefficient based on historical data on the spatial distribution of cyanobacteria blooms, and corrects the comprehensive score of the cyanobacteria bloom risk index; The output module is used to integrate the cyanobacteria bloom meteorological risk index prediction model and the machine learning method to construct a cyanobacteria bloom information prediction model, and obtain the prediction result by using an integrated forecasting method.
Citation Information
Patent Citations
Vegetation index time-spectrum data-based cyanobacteria bloom monitoring method and device
CN104833637A
Method for screening high-hardness and high-entropy alloy through particle swarm optimization algorithm assisted machine learning
CN116312890A
Cyanobacterial bloom monitoring method and system based on satellite observation data
CN116664947A
Improved cellular automaton cyanobacterial bloom prediction method
CN117408385A
Cited By
Artificial intelligence-based enteromorpha prevention and control early warning method and system
CN119516750A
Meteorological-water quality cooperative driving-based intra-day dynamic prediction method for cyanobacterial bloom
CN121189539A
Water environment cyanobacterial bloom early warning method and system
CN121234091A
River and lake water bloom prediction method and system, medium and product
CN121525961A
A method, system, medium, and product for predicting algal blooms in rivers and lakes.
CN121525961B
