A method for early warning of ecological environment risk of a river basin

Abstract

The present application relates to the technical field of risk early warning, and discloses a kind of basin ecological environment risk early warning method, comprising the following steps: obtaining meteorological hydrology, water quality soil and biological behavior high-frequency time series data in target basin;Based on the ecological characteristics of target basin, determine the key risk indicators and weights;According to the time delay embedding theorem, the key risk indicators are embedded in the ecological phase space, and the distance between the distribution of the ecological state of the target basin and the historical equilibrium state attractor is calculated to determine the critical distance index;Based on the critical distance index, combined with the phase transition classification threshold, the early warning level of the ecological environment of the target basin is determined, and the real-time early warning atlas and response decision of the target basin can be visualized.The present application solves the defects of the existing basin ecological environment risk early warning method in data utilization, model adaptability, index scientificity, spatial precision and emergency response, etc., and realizes the whole-chain optimization of ecological risk from identification to disposal.

Description

Technical Field

[0001] This invention relates to the field of risk warning technology, and more specifically, to a method for early warning of watershed ecological and environmental risks. Background Technology

[0002] The watershed ecological environment is a complex and dynamic system composed of multiple elements such as hydrology, water quality, soil, and biology. Its health is influenced by the interaction of multiple factors, including climate fluctuations, human activities, and natural succession. For example, the Yangtze River Basin, as an important ecological barrier in my country, spans 11 provinces and municipalities and includes various ecological types such as forests, wetlands, and rivers. Its ecosystem not only undertakes functions such as water conservation and biodiversity maintenance, but also needs to cope with the pressure of pollution emissions and over-exploitation of resources brought about by industrialization and urbanization. Such systems exhibit significant spatiotemporal heterogeneity. For instance, instantaneous changes in the structure of plankton communities may reflect sudden changes in water quality, while seasonal fluctuations in river flow velocity are closely related to the watershed's hydrological cycle. However, the multi-scale correlation and nonlinear evolutionary characteristics of the watershed ecological environment necessitate a comprehensive consideration of information across the entire chain, from microscopic biological behavior to macroscopic ecological structure, for risk warning.

[0003] However, existing watershed ecological and environmental risk early warning methods are mostly based on static thresholds or linear regression methods, which are difficult to characterize the nonlinear phase transition process of ecosystems from a steady state to a critical state. For example, some early warning systems only set fixed thresholds for a single water quality parameter (such as ammonia nitrogen concentration), ignoring the spatiotemporal differences between pollutant migration paths and ecosystem buffering capacity. Such methods often suffer from delayed early warnings or misjudgments when responding to climate change or human disturbances.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] In response to the problems in related technologies, this invention proposes a watershed ecological environment risk early warning method to overcome the aforementioned technical problems existing in the existing related technologies.

[0006] Therefore, the specific technical solution adopted by the present invention is as follows:

[0007] A method for early warning of watershed ecological and environmental risks includes the following steps:

[0008] S1. Acquire high-frequency time-series data on meteorology, hydrology, water quality, soil, and biological behavior within the target watershed, and perform standardized cleaning, spatiotemporal interpolation, and wavelet noise reduction on the acquired data.

[0009] S2. Based on the ecological characteristics of the target watershed, key risk indicators are identified, and the weights of the key risk indicators are determined using principal component analysis combined with permutation entropy algorithm.

[0010] S3. Based on the time delay embedding theorem, embed key risk indicators into the ecological phase space, calculate the distribution distance between the target watershed ecological state and the historical equilibrium attractor, and determine the critical distance index.

[0011] S4. Based on the critical distance index and combined with the phase transition classification threshold, determine the early warning level of the ecological environment of the target watershed, and visualize the real-time early warning map and response decision of the target watershed.

[0012] Furthermore, the acquisition of high-frequency time-series data on meteorological and hydrological conditions, water quality, soil conditions, and biological behavior within the target watershed, and the standardization, spatiotemporal interpolation, and wavelet denoising processing of the acquired data, includes the following steps:

[0013] S11. Acquire basic data on rainfall intensity, river flow velocity, water pressure changes, and water quality parameters within the target watershed, as well as high-frequency time-series data on biological behavior, including photosynthetic response signals of plankton and three-dimensional movement trajectories of fish.

[0014] S12. Data standardization and spatiotemporal interpolation techniques are used to process the acquired basic data and high-frequency time series data of biological behavior, and wavelet denoising algorithm is used to denoise the high-frequency time series data of biological behavior.

[0015] Furthermore, the denoising process for high-frequency time-series data on biological behavior using wavelet denoising algorithm includes the following steps:

[0016] By combining the characteristics of planktonic photosynthetic response signals and fish three-dimensional motion trajectory data, the characteristic frequency range in high-frequency time series data of biological behavior was determined;

[0017] Based on the characteristics of high-frequency time series data of biological behavior, the wavelet basis function and the number of decomposition levels are determined. Then, using the determined wavelet basis function and the number of decomposition levels, the high-frequency time series data of biological behavior is decomposed into high-frequency coefficients and low-frequency coefficients at different scales.

[0018] Based on the distribution characteristics of high-frequency coefficients, and combined with preset threshold rules and sizes, the high-frequency coefficients are truncated or attenuated. Then, the high-frequency coefficients after thresholding are combined with the low-frequency coefficients by wavelet inverse transform to reconstruct the processed high-frequency time series data of biological behavior.

[0019] Furthermore, the process of determining key risk indicators based on the ecological characteristics of the target watershed and using principal component analysis combined with permutation entropy algorithm to determine the weights of these key risk indicators includes the following steps:

[0020] S21. Based on the core ecological attributes of the target watershed, screen initial risk indicators related to the ecological vulnerabilities of the watershed from basic data and high-frequency time-series data of biological behavior.

[0021] S22. Calculate the correlation coefficients among the initial risk indicators, eliminate redundant risk indicators based on the calculation results, and verify the abnormal response characteristics of the initial risk indicators before the occurrence of risks by combining the historical ecological risk event records of the target watershed. Based on the verification results, determine the candidate indicator set.

[0022] S23. Use principal component analysis to extract the comprehensive correlation weights of candidate indicators, calculate the dynamic sensitivity weights of candidate indicators based on permutation entropy algorithm, and determine the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights.

[0023] Furthermore, the process of extracting the comprehensive correlation weights of candidate indicators using principal component analysis, calculating the dynamic sensitivity weights of candidate indicators based on the permutation entropy algorithm, and determining the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights includes the following steps:

[0024] S231. Based on the standardized candidate indicators, construct the covariance matrix between the indicators, solve the matrix eigenvalues ​​and eigenvectors, obtain each principal component and contribution rate, and determine the principal components based on the cumulative contribution rate.

[0025] S232. Calculate the comprehensive correlation weight of the candidate indicators based on the percentage of the absolute values ​​of the loadings of the candidate indicators in the principal components; calculate the permutation entropy of the candidate indicators based on the preset embedding dimension and delay time, and perform normalization processing to obtain the dynamic sensitivity weight of the candidate indicators.

[0026] S233. Using the weighted average method, combined with the comprehensive correlation weight and dynamic sensitivity weight of the candidate indicators, calculate the comprehensive weight of the candidate indicators. Based on the ranking results of the comprehensive weight of the candidate indicators, screen key risk indicators and determine the weight of the key risk indicators.

[0027] Furthermore, the step of embedding key risk indicators into the ecological phase space according to the time delay embedding theorem, and calculating the distance between the distribution of the target watershed ecological state and the historical equilibrium attractor to determine the critical distance index includes the following steps:

[0028] S31. Based on the time delay embedding theorem, the time delay of high-frequency time series data of key risk indicators is determined by combining the autocorrelation function method, and the embedding dimension is determined by using the spurious nearest neighbor method.

[0029] S32. Based on the determined time delay and embedding dimension, perform phase space mapping on the high-frequency time series data of key risk indicators, transforming one-dimensional time series data into points in a high-dimensional phase space.

[0030] S33. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, and use the kernel density estimation method to analyze the phase space point set and fit the attractor distribution model of the ecosystem in the phase space under the historical equilibrium state.

[0031] S34. Obtain high-frequency time-series data of key risk indicators in the target watershed, embed them into the ecological phase space to obtain the current state point, and calculate the distance between the current state point and the attractor distribution based on the historical equilibrium state attractor distribution model and Mahalanobis distance.

[0032] S35. Obtain ecological state data of the target watershed before the occurrence of ecological risk events in history, calculate the distance between the ecological state point and the historical equilibrium attractor, and determine the critical distance value before the risk occurs by combining the threshold method, so as to obtain the critical distance index.

[0033] Furthermore, the process of determining the time delay of high-frequency time-series data for key risk indicators based on the time delay embedding theorem and combined with the autocorrelation function method, and determining the embedding dimension using the spurious nearest neighbor method, includes the following steps:

[0034] S311. Calculate the autocorrelation coefficient of the high-frequency time series data of key risk indicators, and select the time interval when the autocorrelation coefficient first drops to the preset initial value as the time delay to ensure that the correlation between adjacent points in the phase space is reasonable.

[0035] S312. Starting from the initial dimension, gradually increase the dimension. When the proportion of false nearest neighbors drops to a preset percentage, use that dimension as the optimal embedding dimension to ensure the dynamic characteristics of the ecological environment are fully reconstructed in the phase space.

[0036] Furthermore, the process of obtaining key risk indicator data from the historical stable period of the target watershed's ecological state, constructing a phase space point set for the corresponding period, and analyzing this phase space point set using the kernel density estimation method to fit an attractor distribution model of the ecosystem in the phase space under historical equilibrium states includes:

[0037] S331. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, calculate the average distance of each point to all other points, and remove outliers in the phase space point set based on the average distance.

[0038] S332. Based on the distribution characteristics of the phase space point set, the Gaussian kernel function is selected as the basic function for kernel density estimation, and the cross-validation method is used to calculate the estimation error corresponding to different bandwidths. The optimal bandwidth for kernel density estimation is determined based on the calculation results of the estimation error.

[0039] S333. Based on the selected Gaussian kernel function and optimal bandwidth, the kernel density of the phase space point set of the historical equilibrium state after outlier removal is estimated. Based on the calculation results of the kernel density estimation, the region in the phase space with a density value higher than a set threshold is defined as the attractor core region. The parameters of the kernel density function are saved to form a complete attractor distribution model of the historical equilibrium state.

[0040] Furthermore, the process of determining the early warning level of the target watershed's ecological environment based on the critical distance index and combined with the phase transition classification threshold, and visualizing the real-time early warning map and response decision of the target watershed, includes the following steps:

[0041] S41. Obtain the critical distance index records corresponding to different ecological risk levels in the target watershed in history, combine them with the phase transition characteristics of the ecosystem from stability to imbalance, and determine the phase transition classification threshold through statistical analysis.

[0042] S42. Based on the comparison between the current critical distance index and the phase transition classification threshold of the target watershed, determine the early warning level of the ecological environment of the target watershed;

[0043] S43. Identify areas with clustered critical distance index values ​​through spatial analysis, and locate ecologically fragile and sensitive zones by combining phase change thermal layers; construct a real-time early warning map of the target watershed, and generate corresponding measures by calling the preset decision library according to the early warning level.

[0044] Furthermore, the phase transition grading threshold includes a first-level threshold, a second-level threshold, and a third-level threshold, with the first-level threshold, the second-level threshold, and the third-level threshold increasing sequentially. The warning levels include safe, attention, warning, and emergency.

[0045] When the current critical distance index of the target watershed is less than or equal to the first-level threshold, the early warning level of the ecological environment of the current target watershed is determined to be safe.

[0046] When the current critical distance index of the target watershed is greater than the first-level threshold and less than or equal to the second-level threshold, the warning level of the ecological environment of the current target watershed is determined to be of concern.

[0047] When the current critical distance index of the target watershed is greater than the secondary threshold and less than or equal to the tertiary threshold, the warning level of the ecological environment of the current target watershed is determined to be a warning.

[0048] When the current critical distance index of the target watershed is greater than the level 3 threshold, the early warning level of the ecological environment of the current target watershed is determined to be emergency.

[0049] Compared with existing technologies, this invention provides a method for early warning of watershed ecological and environmental risks, which has the following beneficial effects:

[0050] (1) By integrating technologies such as high-frequency data processing, dynamic modeling, intelligent analysis and visual decision-making, this invention systematically solves the shortcomings of existing watershed ecological environment risk early warning methods in terms of data utilization, model adaptability, indicator scientificity, spatial accuracy and emergency response, and realizes the whole-chain optimization of ecological risk from identification to disposal.

[0051] (2) This invention, based on the construction of ecological phase space using the time-delay embedding theorem and the fitting of the historical equilibrium attractor distribution model, breaks through the limitations of traditional static thresholds and linear models. By calculating the distance between the current ecological state and the historical equilibrium attractor, it can accurately capture the nonlinear phase transition process of the ecosystem from stability to imbalance, solving the problems of existing models being unable to characterize the dynamic evolution of the ecosystem, and the problems of delayed or misjudged warnings. Thus, it can identify the critical state of the ecosystem in advance, buying time for risk intervention.

[0052] (3) This invention combines principal component analysis and permutation entropy algorithm for indicator selection and weight determination, overcoming the shortcomings of traditional subjective weighting and fragmented indicator analysis. Principal component analysis extracts the comprehensive correlation characteristics between indicators, while permutation entropy algorithm reflects the dynamic sensitivity of indicators to risk. The combination of the two ensures the objectivity of the indicator system and highlights the early warning value of sensitive indicators such as biological behavior. It solves the problems of arbitrary indicator selection and neglect of dynamic correlation in weight allocation in traditional methods, making the risk assessment results more consistent with the actual response patterns of the ecosystem. Attached Figure Description

[0053] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0054] Figure 1 This is a flowchart of a watershed ecological environment risk early warning method according to an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] According to an embodiment of the present invention, a method for early warning of watershed ecological and environmental risks is provided.

[0057] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1As shown, according to an embodiment of the present invention, a method for early warning of watershed ecological and environmental risks is provided, comprising the following steps:

[0058] S1. Acquire high-frequency time-series data on meteorology, hydrology, water quality, soil, and biological behavior within the target watershed, and perform standardized cleaning, spatiotemporal interpolation, and wavelet noise reduction on the acquired data.

[0059] The process of acquiring high-frequency time-series data on meteorological and hydrological conditions, water quality, soil conditions, and biological behavior within the target watershed, and then performing standardized cleaning, spatiotemporal interpolation, and wavelet denoising on the acquired data, includes the following steps:

[0060] S11. Acquire basic data on rainfall intensity, river flow velocity, water pressure changes, and water quality parameters within the target watershed, as well as high-frequency time-series data on biological behavior, including photosynthetic response signals of plankton and three-dimensional movement trajectories of fish.

[0061] S12. Data standardization and spatiotemporal interpolation techniques are used to process the acquired basic data and high-frequency time series data of biological behavior, and wavelet denoising algorithm is used to denoise the high-frequency time series data of biological behavior.

[0062] Specifically, the denoising process for high-frequency time-series data on biological behavior using wavelet denoising algorithm includes the following steps:

[0063] First, the characteristic frequency range of high-frequency time-series data on biological behavior is determined. Combining the characteristics of planktonic photosynthetic response signals and fish three-dimensional motion trajectory data, and through preliminary data observation and analysis, the approximate frequency range of effective signals in the two types of biological behavior data is identified. This serves as a reference for subsequent wavelet denoising, avoiding the accidental deletion of effective information during the denoising process.

[0064] Secondly, select suitable wavelet basis functions and decomposition levels. Based on the characteristics of continuity and abrupt changes in high-frequency time-series data of biological behavior, select wavelet basis functions with good time-frequency localization properties, such as the db series wavelets; then, based on the sampling frequency of the data and the degree of noise interference, determine the appropriate number of wavelet decomposition levels through experiments to ensure that the signal and noise can be fully separated.

[0065] Then, wavelet decomposition is performed on the high-frequency time-series data of biological behavior. Using the selected wavelet basis function and decomposition level, the original data is decomposed into high-frequency and low-frequency coefficients at different scales. The high-frequency coefficients mainly correspond to noise components, while the low-frequency coefficients mainly correspond to effective signal components, preparing for subsequent thresholding processing.

[0066] Next, a threshold is set for the high-frequency coefficients and threshold processing is performed. Based on the distribution characteristics of the high-frequency coefficients obtained from the decomposition, an appropriate threshold rule (such as hard threshold or soft threshold) and threshold size are selected to truncate or attenuate the high-frequency coefficients, suppressing the high-frequency coefficients corresponding to noise, retaining the high-frequency components related to the effective signal, and reducing the interference of noise on the data.

[0067] Finally, the denoised biological behavior data was obtained through wavelet reconstruction. The high-frequency and low-frequency coefficients, after thresholding, were subjected to inverse wavelet transform to reconstruct the processed high-frequency time-series data of biological behavior. The resulting data retains effective information from the photosynthetic response signals of planktonic organisms and the three-dimensional movement trajectories of fish while reducing the impact of noise, providing a reliable data foundation for subsequent data analysis and processing.

[0068] S2. Based on the ecological characteristics of the target watershed, key risk indicators are identified, and the weights of the key risk indicators are determined using principal component analysis combined with permutation entropy algorithm.

[0069] The process of determining key risk indicators based on the ecological characteristics of the target watershed and using principal component analysis combined with permutation entropy algorithm to determine the weights of these key risk indicators includes the following steps:

[0070] S21. Based on the core ecological attributes of the target watershed, screen initial risk indicators related to the ecological vulnerabilities of the watershed from basic data and high-frequency time-series data of biological behavior.

[0071] Specifically, the core ecological attributes of the target watershed should be clearly defined. For example, in mountainous watersheds, the focus should be on soil erosion rate and slope runoff coefficient related to soil erosion; while in plain river network watersheds, the focus should be on river network connectivity and eutrophication indicators. From three types of data—meteorological and hydrological (such as peak flow and drought duration), water and soil quality (such as COD concentration and soil heavy metal content), and biological behavior (such as changes in benthic community structure)—indicators directly related to the ecological vulnerabilities of the watershed should be preliminarily screened to ensure that candidate indicators cover key links in the ecosystem.

[0072] S22. Calculate the correlation coefficients among the initial risk indicators, eliminate redundant risk indicators based on the calculation results, and verify the abnormal response characteristics of the initial risk indicators before the occurrence of risks by combining the historical ecological risk event records of the target watershed. Based on the verification results, determine the candidate indicator set.

[0073] Specifically, by calculating the correlation coefficients between indicators, redundant indicators that are highly collinear are eliminated (such as rainfall, which is easier to monitor, if there is a strong correlation between rainfall and runoff). At the same time, by combining the historical ecological risk event records of the watershed, the abnormal response characteristics of candidate indicators before the occurrence of risks are verified, such as the significant fluctuation of the nitrogen-phosphorus ratio in the water body before a certain cyanobacterial bloom. In this way, indicators with actual risk indication significance are retained, forming a simplified candidate indicator set.

[0074] S23. Use principal component analysis to extract the comprehensive correlation weights of candidate indicators, calculate the dynamic sensitivity weights of candidate indicators based on permutation entropy algorithm, and determine the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights.

[0075] Specifically, the steps of extracting the comprehensive correlation weights of candidate indicators using principal component analysis, calculating the dynamic sensitivity weights of candidate indicators based on the permutation entropy algorithm, and determining the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights include the following:

[0076] S231. Based on the standardized candidate indicators, construct the covariance matrix between the indicators, solve the matrix eigenvalues ​​and eigenvectors, obtain each principal component and contribution rate, and determine the principal components according to the cumulative contribution rate (select the principal components with a cumulative contribution rate of more than 85%).

[0077] S232. Calculate the comprehensive correlation weight of the candidate indicators based on the percentage of the absolute values ​​of the loadings of the candidate indicators in the principal components; calculate the permutation entropy value of the candidate indicators based on the preset embedding dimension and delay time (the higher the entropy value, the higher the complexity of the indicator sequence and the more sensitive it is to ecological disturbances) and normalize it to obtain the dynamic sensitivity weight of the candidate indicators.

[0078] S233. Using the weighted average method (in this embodiment, the comprehensive correlation weight accounts for 60% and the dynamic sensitivity weight accounts for 40%), combined with the comprehensive correlation weight and dynamic sensitivity weight of the candidate indicators, calculate the comprehensive weight of the candidate indicators. Based on the ranking result of the comprehensive weight of the candidate indicators, screen the key risk indicators (select the top 20%-30% of the indicators as key risk indicators) and determine the weight of the key risk indicators.

[0079] S3. Based on the time delay embedding theorem, embed key risk indicators into the ecological phase space, calculate the distribution distance between the target watershed ecological state and the historical equilibrium attractor, and determine the critical distance index.

[0080] The step of embedding key risk indicators into the ecological phase space according to the time delay embedding theorem, and calculating the distance between the distribution of the target watershed ecological state and the historical equilibrium attractor to determine the critical distance index includes the following steps:

[0081] S31. Based on the time delay embedding theorem, the time delay of high-frequency time series data of key risk indicators is determined by combining the autocorrelation function method, and the embedding dimension is determined by using the spurious nearest neighbor method.

[0082] Specifically, the process of determining the time delay of high-frequency time-series data for key risk indicators based on the time delay embedding theorem and combined with the autocorrelation function method, and determining the embedding dimension using the spurious nearest neighbor method, includes the following steps:

[0083] S311. Calculate the autocorrelation coefficient of the high-frequency time series data of key risk indicators, and select the time interval when the autocorrelation coefficient first drops to the preset initial value 1 / e (about 0.368) as the time delay to ensure that the correlation between adjacent points in the phase space is reasonable.

[0084] S312. Starting from the initial dimension, gradually increase the dimension (i.e., start from the low dimension and gradually increase the dimension). When the proportion of false nearest neighbors drops to a preset percentage (5% in this embodiment), the dimension is taken as the optimal embedding dimension to ensure the dynamic characteristics of the ecological environment are fully reconstructed in the phase space.

[0085] S32. Based on the determined time delay and embedding dimension, perform phase space mapping on the high-frequency time series data of key risk indicators, transforming one-dimensional time series data into points in a high-dimensional phase space.

[0086] Specifically, using a defined time delay and embedding dimension as parameters, phase space mapping is performed on the high-frequency time-series data of each key risk indicator: For time-series data of length N {x(t)} (t=1,2,...,N), a point set {Y(t)} is constructed in the phase space, where Y(t)=[x(t),x(t+τ),x(t+2τ),...,x(t+(m-1)τ)], τ is the time delay, m is the embedding dimension, and t+(m-1)τ≤N. Through this operation, one-dimensional time-series data is transformed into points in a high-dimensional phase space, enabling an intuitive representation of the dynamic state of the watershed ecosystem;

[0087] S33. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, and use the kernel density estimation method to analyze the phase space point set and fit the attractor distribution model of the ecosystem in the phase space under the historical equilibrium state.

[0088] Specifically, the process of obtaining key risk indicator data for the historical stable ecological state of the target watershed, constructing a phase space point set for the corresponding period, and analyzing the phase space point set using the kernel density estimation method to fit an attractor distribution model of the ecosystem in the phase space under historical equilibrium includes:

[0089] S331. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, calculate the average distance of each point to all other points, and remove outliers in the phase space point set according to the average distance, that is, judge the points whose distance mean exceeds 3 times the standard deviation as outliers and remove them, to ensure that the point set input to the kernel density estimation can truly reflect the distribution characteristics of the stable state of the ecosystem.

[0090] S332. Based on the distribution characteristics of the phase space point set (the attractor point set under historical equilibrium state usually exhibits continuous and locally clustered characteristics), the Gaussian kernel function is selected as the basic function for kernel density estimation (the Gaussian kernel function can effectively fit the continuously distributed point set, and its bell curve characteristics can accurately capture the local density differences of the point set, avoiding density estimation distortion caused by improper selection of kernel function). The cross-validation method is used to calculate the estimation error corresponding to different bandwidths, and the optimal bandwidth for kernel density estimation is determined based on the calculation results of the estimation error.

[0091] Specifically, cross-validation is used to calculate the estimation error corresponding to different bandwidths: the historical equilibrium phase space point set is randomly divided into a training set and a validation set. Kernel density estimation is performed based on the training set under different bandwidths. The deviation between the estimated value and the actual point set distribution is calculated using the validation set, and the bandwidth with the smallest deviation is selected as the optimal bandwidth. Too large a bandwidth will lead to an overly smooth density distribution, losing the local features of attractors; too small a bandwidth will lead to overfitting, amplifying the influence of random fluctuations. Therefore, the optimal bandwidth needs to balance smoothness and detail preservation.

[0092] S333. Based on the selected Gaussian kernel function and optimal bandwidth, the kernel density of the phase space point set of the historical equilibrium state after outlier removal is estimated. Based on the calculation results of the kernel density estimation, the region in the phase space with a density value higher than the set threshold is defined as the attractor core region. The parameters of the kernel density function are saved to form a complete attractor distribution model of the historical equilibrium state.

[0093] Specifically, using the selected Gaussian kernel function and optimal bandwidth as parameters, the density of the preprocessed historical equilibrium phase space point set is calculated: for any position z in the phase space, its density estimate is obtained by weighted summation of points zi in all historical point sets using the kernel function. This process generates the density distribution surface of the entire phase space by superimposing the density contributions of each historical point to the surrounding region;

[0094] Based on the calculated kernel density estimates, regions in phase space with density values ​​higher than a set threshold (such as the 90th quantile of the density distribution) are designated as attractor core regions. These regions correspond to the set of ecosystem states most frequently occurring under historical equilibrium conditions. Simultaneously, the parameters of the kernel density function (kernel function type, bandwidth, core region boundary) are preserved to form a complete historical equilibrium attractor distribution model, providing a benchmark model for subsequent calculations of the distance between the current ecological state and the attractors.

[0095] S34. Obtain high-frequency time series data of key risk indicators in the target watershed, embed them into the ecological phase space to obtain the current state point, and calculate the distance between the current state point and the attractor distribution based on the historical equilibrium state attractor distribution model and the Mahalanobis distance (considering the correlation between indicators). The larger the distance value, the higher the degree of deviation of the current ecological state from the historical stable state and the greater the potential risk.

[0096] S35. Obtain ecological state data before ecological risk events (such as water quality deterioration and biological community decline) occurred in the target watershed in history, calculate the distance between the ecological state point and the historical equilibrium attractor, and determine the critical distance value before the risk occurs by combining the threshold method, so as to obtain the critical distance index. When the distance between the current ecological state and the attractor exceeds the critical value, it indicates that the watershed ecosystem is close to the risk triggering state, and provides a quantitative basis for subsequent early warning level determination.

[0097] S4. Based on the critical distance index and combined with the phase transition classification threshold, determine the early warning level of the ecological environment of the target watershed, and visualize the real-time early warning map and response decision of the target watershed.

[0098] The process of determining the early warning level of the target watershed's ecological environment based on the critical distance index and combined with the phase transition classification threshold, and visualizing the real-time early warning map and response decision of the target watershed, includes the following steps:

[0099] S41. Obtain the critical distance index records corresponding to different ecological risk levels in the target watershed in history, combine them with the phase transition characteristics of the ecosystem from stability to imbalance, and determine the phase transition classification threshold through statistical analysis.

[0100] S42. Based on the comparison between the current critical distance index and the phase transition classification threshold of the target watershed, determine the early warning level of the ecological environment of the target watershed;

[0101] Specifically, the phase transition grading threshold includes a first-level threshold, a second-level threshold, and a third-level threshold, with the first-level threshold, the second-level threshold, and the third-level threshold increasing sequentially. The warning levels include safety, attention, warning, and emergency.

[0102] When the current critical distance index of the target watershed is less than or equal to the first-level threshold, the warning level of the ecological environment of the current target watershed is determined to be safe, that is, the ecosystem is in historical equilibrium, there is no risk, and the visualization color is green.

[0103] When the current critical distance index of the target watershed is greater than the first-level threshold and less than or equal to the second-level threshold, the warning level of the ecological environment of the current target watershed is determined to be of concern, that is, the ecosystem is slightly deviating from the equilibrium state and potential risks need to be paid attention to, and the visualization color is yellow;

[0104] When the current critical distance index of the target watershed is greater than the secondary threshold and less than or equal to the tertiary threshold, the warning level of the ecological environment of the current target watershed is determined to be warning, that is, the ecosystem is close to the critical state, the risk probability has increased significantly, and the visualization color is orange.

[0105] When the current critical distance index of the target watershed is greater than the level 3 threshold, the warning level of the ecological environment of the current target watershed is determined to be emergency, that is, the ecosystem is close to ecological imbalance and requires emergency intervention, and the visualization color is red.

[0106] S43. Identify areas with clustered critical distance index values ​​through spatial analysis, and locate ecologically fragile and sensitive zones by combining phase change thermal layers; construct a real-time early warning map of the target watershed, and generate corresponding measures by calling the preset decision library according to the early warning level.

[0107] In summary, by utilizing the above-mentioned technical solutions of this invention, this invention systematically solves the deficiencies of existing watershed ecological environment risk early warning methods in terms of data utilization, model adaptability, indicator scientificity, spatial accuracy, and emergency response by integrating technologies such as high-frequency data processing, dynamic modeling, intelligent analysis, and visual decision-making. It achieves full-chain optimization of ecological risks from identification to disposal.

[0108] Furthermore, this invention, based on the construction of ecological phase space using the time-delay embedding theorem and the fitting of the historical equilibrium attractor distribution model, overcomes the limitations of traditional static thresholds and linear models. By calculating the distance between the current ecological state and the historical equilibrium attractor, it can accurately capture the nonlinear phase transition process of the ecosystem from stability to imbalance, solving the problems of existing models being unable to characterize the dynamic evolution of ecosystems, and the issues of delayed or misjudged warnings. This allows for the early identification of critical states of ecosystems, buying time for risk intervention.

[0109] Furthermore, this invention combines principal component analysis (PCA) with permutation entropy algorithm for indicator selection and weight determination, overcoming the shortcomings of traditional subjective weighting and fragmented indicator analysis. PCA extracts the comprehensive correlation characteristics between indicators, while permutation entropy algorithm reflects the dynamic sensitivity of indicators to risk. The combination of the two ensures the objectivity of the indicator system while highlighting the early warning value of sensitive indicators such as biological behavior. This solves the problems of arbitrary indicator selection and neglect of dynamic correlation in weight allocation in traditional methods, making the risk assessment results more consistent with the actual response patterns of the ecosystem.

[0110] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it includes the steps described in the above methods. The storage medium may be, for example, ROM / RAM, magnetic disk, optical disk, etc.

[0111] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for early warning of watershed ecological and environmental risks, characterized in that, Includes the following steps: S1. Acquire high-frequency time-series data on meteorology, hydrology, water quality, soil, and biological behavior within the target watershed, and perform standardized cleaning, spatiotemporal interpolation, and wavelet noise reduction on the acquired data. The process of acquiring high-frequency time-series data on meteorological and hydrological conditions, water quality, soil conditions, and biological behavior within the target watershed, and then performing standardized cleaning, spatiotemporal interpolation, and wavelet denoising on the acquired data, includes the following steps: S11. Acquire basic data on rainfall intensity, river flow velocity, water pressure changes, and water quality parameters within the target watershed, as well as high-frequency time-series data on biological behavior, including photosynthetic response signals of plankton and three-dimensional movement trajectories of fish. S12. Data standardization and spatiotemporal interpolation techniques are used to process the acquired basic data and high-frequency time series data of biological behavior, and wavelet denoising algorithm is used to denoise the high-frequency time series data of biological behavior. S2. Based on the ecological characteristics of the target watershed, key risk indicators are identified, and the weights of the key risk indicators are determined using principal component analysis combined with permutation entropy algorithm. S3. Based on the time delay embedding theorem, embed key risk indicators into the ecological phase space, calculate the distribution distance between the target watershed ecological state and the historical equilibrium attractor, and determine the critical distance index. The step of embedding key risk indicators into the ecological phase space according to the time delay embedding theorem, and calculating the distance between the distribution of the target watershed ecological state and the historical equilibrium attractor to determine the critical distance index includes the following steps: S31. Based on the time delay embedding theorem, the time delay of high-frequency time series data of key risk indicators is determined by combining the autocorrelation function method, and the embedding dimension is determined by using the spurious nearest neighbor method. S32. Based on the determined time delay and embedding dimension, perform phase space mapping on the high-frequency time series data of key risk indicators, transforming one-dimensional time series data into points in a high-dimensional phase space. S33. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, and use the kernel density estimation method to analyze the phase space point set and fit the attractor distribution model of the ecosystem in the phase space under the historical equilibrium state. S34. Obtain high-frequency time-series data of key risk indicators in the target watershed, embed them into the ecological phase space to obtain the current state point, and calculate the distance between the current state point and the attractor distribution based on the historical equilibrium state attractor distribution model and Mahalanobis distance. S35. Obtain ecological state data of the target watershed before the occurrence of ecological risk events in history, calculate the distance between the ecological state point and the historical equilibrium attractor, and combine the threshold method to determine the critical distance value before the risk occurs, and obtain the critical distance index. S4. Based on the critical distance index and combined with the phase transition classification threshold, determine the early warning level of the ecological environment of the target watershed, and visualize the real-time early warning map and response decision of the target watershed.

2. The watershed ecological environment risk early warning method according to claim 1, characterized in that, The method of using wavelet denoising algorithm to denoise high-frequency time series data of biological behavior includes the following steps: By combining the characteristics of planktonic photosynthetic response signals and fish three-dimensional motion trajectory data, the characteristic frequency range in high-frequency time series data of biological behavior was determined; Based on the characteristics of high-frequency time series data of biological behavior, the wavelet basis function and the number of decomposition levels are determined. Then, using the determined wavelet basis function and the number of decomposition levels, the high-frequency time series data of biological behavior is decomposed into high-frequency coefficients and low-frequency coefficients at different scales. Based on the distribution characteristics of high-frequency coefficients, and combined with preset threshold rules and sizes, the high-frequency coefficients are truncated or attenuated. Then, the high-frequency coefficients after thresholding are combined with the low-frequency coefficients by wavelet inverse transform to reconstruct the processed high-frequency time series data of biological behavior.

3. The watershed ecological environment risk early warning method according to claim 1, characterized in that, The process of identifying key risk indicators based on the ecological characteristics of the target watershed and determining their weights using principal component analysis combined with permutation entropy algorithm includes the following steps: S21. Based on the core ecological attributes of the target watershed, screen initial risk indicators related to the ecological vulnerabilities of the watershed from basic data and high-frequency time-series data of biological behavior. S22. Calculate the correlation coefficients among the initial risk indicators, eliminate redundant risk indicators based on the calculation results, and verify the abnormal response characteristics of the initial risk indicators before the occurrence of risks by combining the historical ecological risk event records of the target watershed. Based on the verification results, determine the candidate indicator set. S23. Use principal component analysis to extract the comprehensive correlation weights of candidate indicators, calculate the dynamic sensitivity weights of candidate indicators based on permutation entropy algorithm, and determine the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights.

4. The watershed ecological environment risk early warning method according to claim 3, characterized in that, The process of extracting the comprehensive correlation weights of candidate indicators using principal component analysis, calculating the dynamic sensitivity weights of candidate indicators based on the permutation entropy algorithm, and determining the key risk indicators and their weights based on the comprehensive correlation weights and dynamic sensitivity weights includes the following steps: S231. Based on the standardized candidate indicators, construct the covariance matrix between the indicators, solve the matrix eigenvalues ​​and eigenvectors, obtain each principal component and contribution rate, and determine the principal components based on the cumulative contribution rate. S232. Calculate the comprehensive correlation weight of the candidate indicators based on the percentage of the absolute values ​​of the loadings of the candidate indicators in the principal components; calculate the permutation entropy of the candidate indicators based on the preset embedding dimension and delay time, and perform normalization processing to obtain the dynamic sensitivity weight of the candidate indicators. S233. Using the weighted average method, combined with the comprehensive correlation weight and dynamic sensitivity weight of the candidate indicators, calculate the comprehensive weight of the candidate indicators. Based on the ranking results of the comprehensive weight of the candidate indicators, screen key risk indicators and determine the weight of the key risk indicators.

5. The watershed ecological environment risk early warning method according to claim 1, characterized in that, The method of determining the time delay of high-frequency time series data of key risk indicators based on the time delay embedding theorem and combined with the autocorrelation function method, and determining the embedding dimension using the spurious nearest neighbor method, includes the following steps: S311. Calculate the autocorrelation coefficient of the high-frequency time series data of key risk indicators, and select the time interval when the autocorrelation coefficient first drops to the preset initial value as the time delay to ensure that the correlation between adjacent points in the phase space is reasonable. S312. Starting from the initial dimension, gradually increase the dimension. When the proportion of false nearest neighbors drops to a preset percentage, use that dimension as the optimal embedding dimension to ensure the dynamic characteristics of the ecological environment are fully reconstructed in the phase space.

6. The watershed ecological environment risk early warning method according to claim 1, characterized in that, The process of obtaining key risk indicator data for the historical stable ecological state of the target watershed, constructing a phase space point set for the corresponding period, and analyzing the phase space point set using the kernel density estimation method to fit an attractor distribution model of the ecosystem in the phase space under historical equilibrium states includes: S331. Obtain key risk indicator data of the historical ecological stability period of the target watershed, construct the phase space point set of the corresponding period, calculate the average distance of each point to all other points, and remove outliers in the phase space point set based on the average distance. S332. Based on the distribution characteristics of the phase space point set, the Gaussian kernel function is selected as the basic function for kernel density estimation, and the cross-validation method is used to calculate the estimation error corresponding to different bandwidths. The optimal bandwidth for kernel density estimation is determined based on the calculation results of the estimation error. S333. Based on the selected Gaussian kernel function and optimal bandwidth, the kernel density of the phase space point set of the historical equilibrium state after outlier removal is estimated. Based on the calculation results of the kernel density estimation, the region in the phase space with a density value higher than a set threshold is defined as the attractor core region. The parameters of the kernel density function are saved to form a complete attractor distribution model of the historical equilibrium state.

7. The watershed ecological environment risk early warning method according to claim 1, characterized in that, The process of determining the early warning level of the target watershed's ecological environment based on the critical distance index and combined with the phase transition classification threshold, and visualizing the real-time early warning map and response decisions for the target watershed, includes the following steps: S41. Obtain the critical distance index records corresponding to different ecological risk levels in the target watershed in history, combine them with the phase transition characteristics of the ecosystem from stability to imbalance, and determine the phase transition classification threshold through statistical analysis. S42. Based on the comparison between the current critical distance index and the phase transition classification threshold of the target watershed, determine the early warning level of the ecological environment of the target watershed; S43. Identify areas with clustered critical distance index values ​​through spatial analysis, and locate ecologically fragile and sensitive zones by combining phase change thermal layers; construct a real-time early warning map of the target watershed, and generate corresponding measures by calling the preset decision library according to the early warning level.

8. The watershed ecological environment risk early warning method according to claim 7, characterized in that, The phase transition grading thresholds include a first-level threshold, a second-level threshold, and a third-level threshold, with the first-level threshold, the second-level threshold, and the third-level threshold increasing sequentially. The warning levels include safe, attention, warning, and emergency. When the current critical distance index of the target watershed is less than or equal to the first-level threshold, the early warning level of the ecological environment of the current target watershed is determined to be safe. When the current critical distance index of the target watershed is greater than the first-level threshold and less than or equal to the second-level threshold, the warning level of the ecological environment of the current target watershed is determined to be of concern. When the current critical distance index of the target watershed is greater than the secondary threshold and less than or equal to the tertiary threshold, the warning level of the ecological environment of the current target watershed is determined to be a warning. When the current critical distance index of the target watershed is greater than the level 3 threshold, the early warning level of the ecological environment of the current target watershed is determined to be emergency.

Images