Early warning system of cyanobacteria bloom in Lake Taihu based on multi-factor analysis and time lag strategy

Abstract

The present invention relates to the field of data analysis technology, and in particular, to a Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy. It includes a data collection and monitoring unit for monitoring and collecting data from various lake areas of Taihu Lake; a data analysis unit determines the key driving factors and time lag response mechanism and lag duration of the cyanobacteria bloom outbreak, and constructs a cyanobacteria bloom outbreak prediction model for different lake areas of Taihu Lake based on the key driving factors and time lag response mechanism of the cyanobacteria bloom outbreak; a risk assessment and early warning unit uses the cyanobacteria bloom characterization parameter results predicted by the cyanobacteria bloom outbreak prediction model to obtain the future cyanobacteria bloom risk. The system can identify the key factors that directly affect the growth of cyanobacteria, and reveal the mechanism of action of those indirect influencing factors. At the same time, it can evaluate the immediate impact of potential variables on the occurrence of cyanobacteria blooms, capture their cumulative effects at different time points, and provide more accurate information for the early warning system.

Description

Technical Field

[0001] The invention relates to the technical field of data analysis, and in particular to a Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy. Background Art

[0002] The outbreak of cyanobacteria blooms is usually affected by water quality conditions, meteorology and nutrients. The biogeochemical processes such as growth and death of aquatic communities such as lake algae are inconsistent with the time scale of climate change and material circulation, which leads to the time-delay response of causal associations and response relationships in aquatic ecosystems. For example, material circulation can be measured in hours, but biomass changes require days or longer time scales. The current cyanobacteria bloom outbreak warning and forecasting models often use traditional ecological mechanism models, which require understanding of the driving process of the ecosystem and use mathematical equations to explicitly simulate key physical, chemical and biological processes. In order to represent the system process, a large amount of data and a lot of time are required to calibrate the model parameters. In addition, few models can simulate the strong nonlinearity of ecological data, which makes it difficult to further improve the prediction accuracy of the mechanism model. In addition, the model often only focuses on the immediate effect and ignores the cumulative impact of environmental factors over time. Therefore, a Taihu cyanobacteria bloom warning system based on multi-factor analysis and time lag strategy is designed. Summary of the invention

[0003] The purpose of the present invention is to provide a Taihu cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy, so as to solve the problem that few models mentioned in the above background technology can simulate the strong nonlinearity of ecological data, resulting in difficulty in further improving the prediction accuracy of mechanism models, and the models often only focus on immediate effects and ignore the cumulative effects of environmental factors changing over time.

[0004] To achieve the above object, the present invention aims to provide a Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy, comprising:

[0005] A data collection and monitoring unit, which is used to monitor and collect data from various lake areas of Taihu Lake, establish a basic ecological environment database, and store the collected data from various lake areas of Taihu Lake in the ecological environment database;

[0006] A data analysis unit, the data analysis unit is used to determine the key driving factors of the cyanobacteria bloom outbreak based on the data of various lake areas of Taihu Lake collected in the ecological environment database, and at the same time, determine the time-lag response mechanism and lag duration between the cyanobacteria bloom state variables and the key driving factors of the cyanobacteria bloom outbreak in different lake areas of Taihu Lake, and construct a cyanobacteria bloom outbreak prediction model in different lake areas of Taihu Lake based on the key driving factors and time-lag response mechanism of the cyanobacteria bloom outbreak;

[0007] Among them, the data analysis unit includes a key driving factor identification module, a time lag effect analysis module, and a prediction model construction and verification module;

[0008] The risk assessment and early warning unit is used to use the future climate change prediction results as input parameters to run the model, use the cyanobacteria bloom characterization parameter results predicted by the cyanobacteria bloom outbreak prediction model to obtain the future cyanobacteria bloom risk, and determine the risk level, risk trigger conditions and risk thresholds.

[0009] As a further improvement of the present technical solution, the data of each lake area in the data collection and monitoring unit includes hydrological characteristics, water quality parameters, sediment parameters, cyanobacterial bloom characterization characteristics and meteorological data.

[0010] As a further improvement of the technical solution, in the data analysis unit:

[0011] The key driving factor identification module is used to generate sequence data of different time scales based on the data of various lake areas of Taihu Lake collected in the ecological environment database, analyze the temporal pattern and spatial distribution characteristics of cyanobacteria bloom outbreaks, and then determine the key driving factors of cyanobacteria bloom outbreaks in different lake areas of Taihu Lake, and use the structural equation model to rank the key driving factors according to their importance;

[0012] The time lag effect analysis module is used to introduce lag terms into the structural equation model, and then determine the time lag response mechanism and lag duration between the state variables of cyanobacteria blooms in different lake areas of Taihu Lake and the key driving factors of cyanobacteria blooms.

[0013] The prediction model construction and verification module is used to establish a cyanobacterial bloom outbreak prediction model based on the key driving factors in the key driving factor identification module and the time lag response mechanism in the time lag effect analysis module, and to verify the cyanobacterial bloom outbreak prediction model.

[0014] As a further improvement of the technical solution, the key driving factor identification module has the following specific steps:

[0015] S1. Preprocess the data of each lake area of ​​Taihu Lake obtained from the data collection and monitoring units, including removing outliers, filling missing values, and adjusting data of different dimensions to the same level through normalization;

[0016] S2. Based on the preprocessed data set, the correlation matrix and principal component analysis method were used to preliminarily screen out the variables most relevant to the outbreak of cyanobacteria blooms. The random forest algorithm was then used to streamline the feature set, calculate the importance score of each feature, and retain several key driving factors that contribute to the prediction results. In the process of using the random forest algorithm to streamline the feature set and calculate the importance score of each feature, the feature correlation factor, node splitting criterion and tree complexity were introduced for optimization;

[0017] S3. Based on the results of the key driving factors selection, the structural equation model was used to establish the causal relationship network between several key driving factors and the outbreak of cyanobacteria, and to clarify which key driving factors directly promoted the growth of cyanobacteria or which key driving factors indirectly promoted the growth of cyanobacteria;

[0018] S4. Based on the results of S3, obtain the final importance ranking of several key driving factors.

[0019] As a further improvement of this technical solution, S2 is specifically as follows:

[0020]

[0021] in, is the Pearson correlation coefficient between each feature pair and the cyanobacterial bloom outbreak; is the sample size; For the Samples in the feature The value on For all samples in the feature The average value on For the Samples in the feature The value on For all samples in the feature The average value on ; To characterize the impact of cyanobacteria blooms; It is a characteristic of the outbreak of cyanobacteria bloom;

[0022] ;

[0023] in, is the covariance matrix; for The matrix of is the number of features; for The transposed matrix of

[0024] ;

[0025] in, Features In the Nodes in a tree Importance rating; For Node The Gini index of For Node Gini index of the left child node after splitting; For Node Gini index of the right child node after splitting;

[0026] ;

[0027] in, Features The total importance score of is the number of decision trees in the random forest; Features The set of nodes that appear in all trees.

[0028] As a further improvement of the technical solution, in S2, in the process of using the random forest algorithm to simplify the feature set and calculate the importance score of each feature, the feature correlation factor, the node splitting criterion and the complexity of the tree are introduced for optimization, and the optimization is specifically as follows:

[0029] ;

[0030] in, feature Impact factor; To adjust the parameters; For the characteristics of cyanobacteria blooms; For and Features There are interactive features; Features and Features The mutual information between

[0031] );

[0032] in, The optimized features The total importance score of is the splitting criterion factor; is the complexity factor;

[0033] ;

[0034] in, The optimized features The total importance score of .

[0035] As a further improvement of this technical solution, the specific steps of S3 are as follows:

[0036] According to the known key driving factor selection results, the structural equation model is used to construct a preliminary causal network; latent variables are introduced into the structural equation model, and the measurement model is used to define the relationship between observed variables and latent variables. For each latent variable, its corresponding observation indicator is determined, and the path relationship between latent variables is specified using the structural model. Finally, the parameters of the structural equation model are estimated using the maximum likelihood method.

[0037] As a further improvement of the technical solution, the parameters of the structural equation model include path coefficients, error terms and loads of latent variables.

[0038] As a further improvement of the technical solution, hidden variables are introduced into the structural equation model, the relationship between the observed variables and the latent variables is defined by using the measurement model, the corresponding observation indicators are determined for each latent variable, and the path relationship between the latent variables is specified by using the structural model, as follows:

[0039] Measurement Model:

[0040] ;

[0041] in, For the Observation indicators; is a hidden variable; is a hidden variable For The factor loading of each observation indicator; For the The measurement error of each observation indicator;

[0042] Structural model:

[0043] ;

[0044] in, It is a potential variable for the occurrence of cyanobacterial blooms; is the path coefficient of the latent variable of temperature; is the latent variable of temperature; is the path coefficient of the latent variable of light intensity; is the latent variable of light intensity; is the path coefficient of the latent variable of nutrient concentration; is the potential variable of nutrient concentration; is the path coefficient of the latent variable; is the residual term for the occurrence of cyanobacterial blooms.

[0045] As a further improvement of the technical solution, the time lag effect analysis module is specifically as follows:

[0046] The measurement model remains unchanged, and the structural model becomes:

[0047]

[0048] in, For the period Potential variables for the occurrence of cyanobacterial blooms; The temperature in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The temperature latent variable; is the temperature in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The latent variable of temperature; is the light intensity in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The latent variable of light intensity; is the light intensity in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The latent variable of light intensity; is the nutrient concentration in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The potential variable of nutrient concentration; is the nutrient concentration in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The potential variable of nutrient concentration; For the period Hidden variables of For the period The residual term for the occurrence of cyanobacterial blooms.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. In the Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy, feature correlation factors, node splitting criteria and tree complexity are introduced to optimize the process of calculating the importance score of each feature, which can better reflect the impact of node splitting criteria and tree depth and complexity on feature importance scores, thereby making the final key driving factor evaluation more accurate and reliable;

[0051] 2. In the Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy, the introduction of hidden variables and lag terms in the structural equation model can not only identify the key factors directly affecting the growth of cyanobacteria, but also reveal the mechanism of action of those indirect influencing factors. At the same time, it can evaluate the immediate impact of each potential variable on the occurrence of cyanobacteria blooms and capture their cumulative effects at different time points, which will help to better understand the complex mechanism of cyanobacteria blooms and provide more accurate information for the early warning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is the overall flow chart of the present invention;

[0053] The meaning of each number in the figure is:

[0054] 1. Data collection and monitoring unit; 2. Data analysis unit; 21. Key driving factor identification module; 22. Time lag effect analysis module; 23. Prediction model construction and verification module; 3. Risk assessment and early warning unit. DETAILED DESCRIPTION

[0055] 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.

[0056] See also Figure 1 As shown, a Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy is provided, including a data collection and monitoring unit 1, a data analysis unit 2, and a risk assessment and early warning unit 3;

[0057] The data collection and monitoring unit 1 is used to monitor and collect data from various lake areas of Taihu Lake, establish an ecological environment basic database, and store the collected data from various lake areas of Taihu Lake in the ecological environment database;

[0058] The data of each lake area in the data collection monitoring unit 1 include hydrological characteristics, water quality parameters, sediment parameters, blue algae bloom characteristics and meteorological data; among them, the hydrological characteristics are the water level, flow velocity, flow direction and other hydrological characteristics of each lake area of ​​the lake body, the water quality parameters include water temperature, nitrogen, phosphorus, dissolved oxygen, suspended particles and chlorophyll and other conventional water quality, meteorological data include light, temperature, air pressure, rainfall, wind speed, wind direction, and the establishment of a basic database of water ecological environment can strengthen the accumulation of monitoring data and continuous support capabilities;

[0059] Data analysis unit 2 is used to determine the key driving factors of cyanobacteria bloom outbreaks based on the data of various lake areas of Taihu Lake collected in the ecological environment database. At the same time, the time-lag response mechanism and lag duration between the cyanobacteria bloom state variables and the key driving factors of cyanobacteria bloom outbreaks in different lake areas of Taihu Lake are determined, and according to the key driving factors and time-lag response mechanism of cyanobacteria bloom outbreaks, a cyanobacteria bloom outbreak prediction model for different lake areas of Taihu Lake is constructed;

[0060] The data analysis unit 2 includes a key driving factor identification module 21, a time lag effect analysis module 22 and a prediction model construction verification module 23;

[0061] Data Analysis Unit 2:

[0062] The key driving factor identification module 21 is used to generate sequence data of different time scales based on the data of various lake areas of Taihu Lake collected in the ecological environment database, analyze the time pattern and spatial distribution characteristics of the outbreak of cyanobacteria blooms, and then determine the key driving factors of the outbreak of cyanobacteria blooms in different lake areas of Taihu Lake, and use the structural equation model to rank the key driving factors according to their importance;

[0063] The specific steps of key driver identification module 21 are as follows:

[0064] S1. Preprocess the data of each lake area of ​​Taihu Lake obtained from the data collection and monitoring unit 1, including removing outliers, filling missing values, and adjusting data of different dimensions to the same level through normalization to ensure the quality of subsequent analysis and the consistency of model construction;

[0065] S2. Based on the preprocessed data set, the correlation matrix and principal component analysis method were used to preliminarily screen out the variables most relevant to the outbreak of cyanobacteria blooms. The random forest algorithm was then used to streamline the feature set, calculate the importance score of each feature, and retain several key driving factors that contribute to the prediction results. In the process of using the random forest algorithm to streamline the feature set and calculate the importance score of each feature, the feature correlation factor, node splitting criterion and tree complexity were introduced for optimization;

[0066] S2 is as follows:

[0067] Use the correlation matrix to evaluate the linear relationship between features;

[0068] ;

[0069] in, is the Pearson correlation coefficient between each feature pair and the cyanobacterial bloom outbreak; is the sample size; For the Samples in the feature The value on For all samples in the feature The average value on For the Samples in the feature The value on For all samples in the feature The average value on ; To characterize the impact of cyanobacteria blooms; It is a characteristic of the outbreak of cyanobacteria bloom;

[0070] The value range of Pearson correlation coefficient is from -1 to 1. When it is close to 1, it means and When it is close to -1, it indicates a strong positive correlation; and when it is close to 0, it indicates that there is no obvious linear relationship between the two.

[0071] Next, principal component analysis (PCA) was applied to further screen out the features that best explain the variability in the data. PCA achieves dimensionality reduction by finding the directions with the largest variance in the data (i.e., the principal components);

[0072] ;

[0073] in, is the covariance matrix, which is used to describe the relationship between the features in the data set; for The matrix of is the number of features; for The transposed matrix of

[0074] Through Perform eigenvalue decomposition to obtain the eigenvalue and the corresponding eigenvector The eigenvalue reflects the importance of the principal component represented by the corresponding eigenvector, so the previous principal components;

[0075] After using PCA to identify the main features, the random forest algorithm is used to evaluate the importance of features. Each tree in the random forest will randomly select a part of the features when building, and generate different training sets through the bagging method. At each node, the optimal feature is selected for splitting according to the Gini index or other splitting criteria;

[0076] ;

[0077] in, Features In the Nodes in a tree Importance rating; For Node The Gini index of For Node Gini index of the left child node after splitting; For Node Gini index of the right child node after splitting;

[0078] ;

[0079] in, Features The total importance score of is the average importance score of the feature on all trees; is the number of decision trees in the random forest; Features The set of nodes that appear on all trees; finally, the importance scores of all features are normalized to ensure that the sum of the scores is 1, which is convenient for comparing the importance of different features. This method that combines the correlation matrix, PCA, and random forest can not only effectively reduce the feature dimension, but also retain the features that contribute most to the prediction results, providing a solid foundation for the subsequent identification of key driving factors.

[0080] In S2, when using the random forest algorithm to streamline the feature set and calculate the importance score of each feature, we introduce the feature correlation factor, node splitting criterion, and tree complexity for optimization, because different splitting criteria (such as Gini index, information gain, mean square error, etc.) will affect the importance score of the feature. We can introduce a splitting criterion factor To adjust the impact of different splitting criteria. Assume is a coefficient between 0 and 1 that varies depending on the chosen splitting criterion. For example, if you use the Gini index as the splitting criterion, you can set ; If other criteria are used, they can be adjusted according to their effects The value of

[0081] At the same time, limiting the maximum depth of the tree can prevent overfitting, but a tree that is too shallow may not be able to fully capture the complex patterns in the data. Limiting the minimum number of samples in the leaf nodes can also prevent overfitting, but too high a threshold may cause the tree to be too simple, so a complexity factor is introduced. , which is also a coefficient between 0 and 1. When the tree structure is more complex, Larger; when the tree structure is simpler, Smaller; after optimization:

[0082] ;

[0083] in, feature Impact factor; To adjust the parameters; For the characteristics of cyanobacteria blooms; For and Features There are interactive features; Features and Features The mutual information between

[0084] );

[0085] in, The optimized features The total importance score of is the splitting criterion factor, which is used to adjust the influence of different splitting criteria; is the complexity factor, which is used to adjust the depth and complexity of the tree;

[0086] Among them, the splitting criterion factor This can be determined based on the actual splitting criterion used. For example, if the Gini index is used, you can set ; If information gain is used, it can be adjusted appropriately based on experience or experimental results The value of .

[0087] Complexity Factor This can be determined by the maximum depth of the tree and the minimum number of leaf node samples. For example, if the maximum depth of the tree is large and the minimum number of leaf node samples is small, you can set ; If the maximum depth of the tree is small and the minimum number of leaf node samples is large, you can set ;

[0088] By introducing the splitting criterion factor and complexity factor , which can better reflect the impact of node splitting criteria and tree depth and complexity on feature importance scores, making the final feature importance assessment more accurate and reliable. The specific values ​​of these factors need to be set according to actual conditions and may need to be tuned through methods such as cross-validation.

[0089] ;

[0090] in, The optimized features The total importance score of

[0091] S3. Based on the results of the key driving factors selection, the structural equation model was used to establish the causal relationship network between several key driving factors and the outbreak of cyanobacteria, and to clarify which key driving factors directly promoted the growth of cyanobacteria or which key driving factors indirectly promoted the growth of cyanobacteria;

[0092] The specific steps of S3 are as follows:

[0093] Based on the known key driving factor selection results, a preliminary causal network is constructed using the structural equation model; for example, it is assumed that temperature, light intensity, and nutrient concentration are key driving factors that directly affect the growth of cyanobacteria; and precipitation may indirectly promote the growth of cyanobacteria by affecting the input of nutrients; introduce hidden variables into the structural equation model, use the measurement model to define the relationship between observed variables and latent variables, determine the corresponding observation indicators (actual measured data) for each latent variable (such as temperature and light intensity), and use the structural model to specify the path relationship between latent variables. This step requires clarifying which variables are dependent variables and which are independent variables, as well as the direct or indirect relationship between them. Finally, the maximum likelihood method is used to estimate the parameters of the structural equation model;

[0094] Structural equation models can clearly identify which environmental factors directly promote the growth of cyanobacteria, such as temperature, light intensity and nutrient concentration, and which factors affect the growth of cyanobacteria through indirect pathways, such as precipitation promoting the growth of cyanobacteria by increasing nutrient input. Introducing latent variables such as changes in microbial community structure can capture factors that are difficult to measure directly but have an important impact on the outbreak of cyanobacterial blooms. Using measurement models to define the relationship between observed variables and latent variables, and specifying the path relationship between these latent variables through structural models, helps to build a more comprehensive and accurate causal network. This not only improves the model's understanding of complex ecological processes, but also enhances the accuracy of predictions.

[0095] The parameters of the structural equation model include the path coefficient (indicating the strength of the relationship between variables), the error term (unexplained variation), and the loading of the latent variable (the extent to which the observed variable affects the latent variable);

[0096] Hidden variables are introduced into the structural equation model, and the relationship between observed variables and latent variables is defined by using the measurement model. For each latent variable, its corresponding observation indicator is determined, and the path relationship between latent variables is specified by using the structural model, as follows:

[0097] Measurement Model:

[0098] ;

[0099] in, For the Observation indicators; is a hidden variable; is a hidden variable For The factor loading of each observation indicator; For the The measurement error of each observation indicator can be calculated; this can evaluate the impact of each potential variable (such as temperature, light intensity and nutrient concentration) on the corresponding observation indicator and take into account the existence of measurement errors, which helps to improve the accuracy and reliability of the model.

[0100] Structural model:

[0101] ;

[0102] in, It is a potential variable for the occurrence of cyanobacterial blooms; is the path coefficient of the latent variable of temperature; is the latent variable of temperature; is the path coefficient of the latent variable of light intensity; is the latent variable of light intensity; is the path coefficient of the latent variable of nutrient concentration; is the potential variable of nutrient concentration; is the path coefficient of the latent variable; is the residual term for the occurrence of cyanobacterial blooms; this formula describes how different potential variables (such as temperature, light intensity, nutrient concentration and other factors not directly observed) work together to affect the occurrence of cyanobacterial blooms. This method can not only identify the key factors that directly affect the growth of cyanobacteria, but also reveal the mechanism of action of those indirect factors. In this way, a more complex but more realistic causal network can be established, which can provide more scientific support for predicting the outbreak of cyanobacterial blooms and provide a basis for formulating effective prevention and control measures.

[0103] S4. Based on the results of S3, obtain the final importance ranking of several key driving factors;

[0104] The time lag effect analysis module 22 is used to introduce lag terms into the structural equation model, and then determine the time lag response mechanism and lag duration between the cyanobacteria bloom state variables and the key driving factors of cyanobacteria bloom outbreaks in different lake areas of Taihu Lake;

[0105] The time lag effect analysis module 22 is specifically as follows:

[0106] The measurement model remains unchanged, and the structural model becomes:

[0107]

[0108] in, For the period Potential variables for the occurrence of cyanobacterial blooms; The temperature in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The temperature latent variable; is the temperature in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The temperature latent variable; is the light intensity in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The latent variable of light intensity; is the light intensity in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The latent variable of light intensity; is the nutrient concentration in the current period The immediate path coefficient for the occurrence of cyanobacterial blooms; For the period The potential variable of nutrient concentration; is the nutrient concentration in the previous period lagged one-period path coefficient for the occurrence of cyanobacterial blooms; For the period The potential variable of nutrient concentration; For the period Hidden variables of For the period The residual term for the occurrence of cyanobacterial blooms reflects the part that is not explained by the model;

[0109] By introducing environmental factors such as temperature, light intensity and nutrient concentration, and the period before it By using the latent variables of these factors, we can capture the immediate and lagged effects of these factors on the occurrence of cyanobacterial blooms. and Represents the temperature and light intensity at the current moment The direct force on the occurrence of cyanobacterial blooms is and These two factors reflect the This helps us understand how these environmental factors gradually affect the development of cyanobacterial blooms over time.

[0110] In addition, nutrients are another important driving factor, and its immediate path coefficient and the one-period lagged path coefficient It also provides key information about how changes in nutrient levels affect the growth dynamics of cyanobacteria. By estimating the above parameters, we can quantitatively analyze the relative importance of each driving factor at different time points and determine their specific lag time for the outbreak of cyanobacterial blooms. This not only takes into account the immediate effects of environmental factors, but also fully incorporates their historical impacts, thereby providing a scientific basis for more accurate prediction of cyanobacterial blooms and helping to develop more effective prevention and control strategies.

[0111] In this way, not only can the immediate impact of each potential variable on the occurrence of cyanobacterial blooms be assessed, but their cumulative effects at different time points can also be captured. This helps to better understand the complex mechanisms of cyanobacterial blooms and provide more precise information for early warning systems. It is worth noting that in practical applications, it may be necessary to adjust the lag order according to the specific circumstances of the data to ensure the rationality and accuracy of the model.

[0112] The prediction model construction and verification module 23 is used to establish a cyanobacteria bloom outbreak prediction model based on the key driving factors in the key driving factor identification module 21 and the time lag response mechanism in the time lag effect analysis module 22, and to verify the cyanobacteria bloom outbreak prediction model; 70% of the total number of samples are randomly selected as the training set, and the remaining 30% of the data are input into the model as the verification set samples to train and debug the model.

[0113] The risk assessment and early warning unit 3 is used to use the future climate change prediction results as input parameters to run the model, and use the cyanobacteria bloom characterization parameter results predicted by the cyanobacteria bloom outbreak prediction model to obtain the future cyanobacteria bloom risk and determine the risk level, risk triggering conditions and risk thresholds; it helps decision makers to take timely preventive measures to reduce the impact of cyanobacteria blooms on the environment and socio-economy, and provides a scientific basis for the formulation of long-term plans, thereby enhancing the ability to respond to cyanobacteria bloom outbreaks under the background of climate change.

[0114] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy, characterized by: include: A data collection and monitoring unit (1), the data collection and monitoring unit (1) is used to monitor and collect data from various lake areas of Taihu Lake, establish an ecological environment basic database, and store the collected data from various lake areas of Taihu Lake in the ecological environment database; A data analysis unit (2), wherein the data analysis unit (2) is used to determine the key driving factors of the cyanobacteria bloom outbreak based on the data of each lake area of ​​Taihu Lake collected in the ecological environment database, and at the same time, determine the time-lag response mechanism and lag duration between the cyanobacteria bloom state variables and the key driving factors of the cyanobacteria bloom outbreak in different lake areas of Taihu Lake, and construct a cyanobacteria bloom outbreak prediction model for different lake areas of Taihu Lake based on the key driving factors and time-lag response mechanism of the cyanobacteria bloom outbreak; The data analysis unit (2) includes a key driving factor identification module (21), a time lag effect analysis module (22) and a prediction model construction and verification module (23). The time lag effect analysis module (22) is used to introduce lag terms into the structural equation model, thereby determining the time lag response mechanism and lag duration between the cyanobacteria bloom state variables and the key driving factors of cyanobacteria bloom outbreaks in different lake areas of Taihu Lake. A risk assessment and early warning unit (3) is used to use the future climate change prediction results as input parameters to run the model, use the blue algae bloom characterization parameter results predicted by the blue algae bloom outbreak prediction model to obtain the future blue algae bloom risk, and determine the risk level, risk triggering conditions and risk thresholds.

2. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 1 is characterized by: The data of each lake area in the data collection and monitoring unit (1) includes hydrological characteristics, water quality parameters, sediment parameters, blue algae bloom characteristics and meteorological data.

3. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 2 is characterized in that: In the data analysis unit (2): The key driving factor identification module (21) is used to generate sequence data of different time scales based on the data of various lake areas of Taihu Lake collected in the ecological environment database, analyze the temporal pattern and spatial distribution characteristics of cyanobacterial blooms, and then determine the key driving factors of cyanobacterial blooms in different lake areas of Taihu Lake, and use the structural equation model to rank the key driving factors according to their importance; The prediction model construction and verification module (23) is used to establish a cyanobacteria bloom outbreak prediction model based on the key driving factors in the key driving factor identification module (21) and the time lag response mechanism in the time lag effect analysis module (22), and to verify the cyanobacteria bloom outbreak prediction model.

4. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 3 is characterized in that: The key driving factor identification module (21) has the following specific steps: S1. Preprocessing the data of various lake areas of Taihu Lake obtained from the data collection and monitoring unit (1), including removing outliers, filling missing values, and adjusting data of different dimensions to the same level through normalization; S2. Based on the preprocessed data set, the correlation matrix and principal component analysis method were used to preliminarily screen out the variables most relevant to the outbreak of cyanobacteria blooms. The random forest algorithm was then used to streamline the feature set, calculate the importance score of each feature, and retain several key driving factors that contribute to the prediction results. In the process of using the random forest algorithm to streamline the feature set and calculate the importance score of each feature, the feature correlation factor, node splitting criterion and tree complexity were introduced for optimization; S3. Based on the results of the key driving factors selection, the structural equation model was used to establish the causal relationship network between several key driving factors and the outbreak of cyanobacteria, and to clarify which key driving factors directly promoted the growth of cyanobacteria or which key driving factors indirectly promoted the growth of cyanobacteria; S4. Based on the results of S3, obtain the final importance ranking of several key driving factors.

5. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 4 is characterized in that: The specific calculation process of S2 is as follows: Among them, r X,Y is the Pearson correlation coefficient between each feature pair and the outbreak of cyanobacteria bloom; n is the number of samples; X i0 is the value of the i-th sample on feature X; is the average value of all samples on feature X; Y i0 is the value of the i-th sample on feature Y; is the average value of all samples on feature Y; i0 = 1, 2, …, n; X is the feature that affects the outbreak of cyanobacteria bloom; Y is the feature of the outbreak of cyanobacteria bloom; Where C is the covariance matrix; Z is an n×p matrix, where p is the number of features; Z T is the transposed matrix of Z; in, For feature X j Importance score of node q in the i-th tree; is the Gini index of node q; is the Gini index of the left child node after node q is split; is the Gini index of the right child node after node q is split; in, For feature X j is the total importance score of feature X; I is the number of decision trees in the random forest; Q is the feature X j The set of nodes that appear in all trees.

6. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 5, characterized in that: In S2, in the process of using the random forest algorithm to simplify the feature set and calculate the importance score of each feature, the feature correlation factor, the node splitting criterion and the complexity of the tree are introduced for optimization. After optimization, the specific results are: a jq =1+β·MI(X j ,X k ); Among them, α jq Feature X j The influencing factor of j It is a characteristic of algae blooms; k For feature X j There are interactive features; MI(X j ,X k ) is feature X j and feature X k The mutual information between in, is the optimized feature X j is the total importance score of; γ is the splitting criterion factor; δ is the complexity factor; in, is the optimized feature X j The total importance score of .

7. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 6, characterized in that: The specific steps of S3 are as follows: According to the known key driving factor selection results, the structural equation model is used to construct a preliminary causal network; latent variables are introduced into the structural equation model, and the measurement model is used to define the relationship between observed variables and latent variables. For each latent variable, its corresponding observation indicator is determined, and the path relationship between latent variables is specified using the structural model. Finally, the maximum likelihood method is used to estimate the parameters of the structural equation model.

8. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 7, characterized in that: The parameters of the structural equation model include path coefficients, error terms, and loadings of latent variables.

9. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 8, characterized in that: Hidden variables are introduced into the structural equation model, and the relationship between observed variables and latent variables is defined by using the measurement model. For each latent variable, its corresponding observation indicator is determined, and the path relationship between latent variables is specified by using the structural model, as follows: Measurement Model: x uk =λ uk or u +d uk ; Among them, x uk is the kth observation index; η u is a hidden variable; uk is the hidden variable η u The factor loading of the kth observed indicator; δ uk is the measurement error of the kth observation indicator; Structural model: or algae =β1η temp +β2η light +β3η nutrient +β4η u +g algae ; Among them, η algae is the potential variable of cyanobacteria bloom; β1 is the path coefficient of the latent variable of temperature; η temp is the latent variable of temperature; β2 is the path coefficient of the latent variable of light intensity; η light is the latent variable of light intensity; β3 is the path coefficient of the latent variable of nutrient concentration; η nutrient is the latent variable of nutrient concentration; β4 is the path coefficient of latent variable; ζ algae is the residual term for the occurrence of cyanobacterial blooms.

10. The Taihu Lake cyanobacteria bloom early warning system based on multi-factor analysis and time lag strategy according to claim 9, characterized in that: The time lag effect analysis module (22) is specifically as follows: The measurement model remains unchanged, and the structural model becomes: or a l gae,t =b 1,0 or temp,t +b 1,1 or temp,t-1 +b 2,0 or light,t +b 2,1 or light,t-1 +b 3,0 or nutrient,t +b 3,1 or nutrient,t-1 +β4η u,t +g algae,t ; Among them, η algae,t is the potential variable for the occurrence of cyanobacterial blooms at period t; 1,0 is the instantaneous path coefficient of temperature on the occurrence of cyanobacterial blooms in the current period t; η temp,t is the latent variable of temperature at period t; β 1,1 is the lagged one-period path coefficient of temperature on the occurrence of cyanobacterial blooms in the previous period t-1; η temp,t-1 is the latent variable of temperature at period t-1; β 2,0 is the instantaneous path coefficient of light intensity on the occurrence of cyanobacterial blooms at the current period t; η light,t is the latent variable of light intensity at period t; β 2,1 is the lagged one-period path coefficient of the light intensity on the occurrence of cyanobacterial bloom in the previous period t-1; η light,t-1 is the latent variable of light intensity at period t-1; β 3,0 is the instantaneous path coefficient of nutrient concentration on the occurrence of cyanobacterial blooms at the current period t; η nutrient,t is the latent variable of nutrient concentration at period t; β 3,1 is the lagged one-period path coefficient of the nutrient concentration on the occurrence of cyanobacterial blooms in the previous period t-1; η nutrient,t-1 is the latent variable of nutrient concentration at period t-1; η u,t is the latent variable of period t; algae,t is the residual term for the occurrence of cyanobacterial bloom in period t.

Images