A method for assessing the health of aquatic ecosystems

Abstract

This application relates to a method for assessing aquatic ecological health, belonging to the field of environmental monitoring and information technology. The method includes: acquiring water quality data from multiple samples through multi-source acquisition, including: acquiring biotoxicity data based on biological models; acquiring pollutant data based on spectral or mass spectrometry analysis; acquiring physicochemical parameter data of water bodies based on sensors; constructing an aquatic ecological health assessment index system based on the water quality data; constructing a water quality parameter prediction model using machine learning algorithms based on historical and current water quality data; inputting the current water quality data into the water quality parameter prediction model to obtain prediction results for future water quality parameters; and comprehensively assessing the current status and future trends of aquatic ecological health by combining the aquatic ecological health assessment index system and the prediction results of future water quality parameters.

Description

Technical Field

[0001] This application belongs to the field of environmental monitoring and information technology, and specifically relates to a method for assessing the health of aquatic ecosystems. Background Technology

[0002] Aquatic ecological health is fundamental to maintaining biodiversity and sustainable human development. Traditional water quality assessment methods mainly rely on isolated monitoring and evaluation of physicochemical parameters (such as pH, dissolved oxygen, COD, etc.) of a small number of samples, which has obvious limitations: Limited data dimensions: focusing only on some chemical indicators, lacking the ability to screen for comprehensive biological toxicity effects and unknown pollutants in water bodies, and failing to fully reflect the ecological risks of water bodies; Lagging assessment: relying mainly on historical data for current status assessment, lacking the ability to predict future trends, and making early warning difficult; Low throughput and poor efficiency: long cycle from sampling to laboratory analysis, making it difficult to meet the needs of large-scale, high-frequency monitoring, and failing to form valuable "big data" resources; High subjectivity: the determination of indicator weights in traditional index evaluation methods often depends on expert experience, lacking objectivity. Summary of the Invention

[0003] Based on the above analysis, the embodiments of the present invention aim to provide a water ecological health assessment method to solve the technical problems of existing technologies, such as single data dimension, delayed assessment, low throughput, poor efficiency, and strong subjectivity.

[0004] The objective of this invention is achieved as follows: This invention provides a method for assessing the health of aquatic ecosystems, comprising: Water quality data from multiple samples was acquired through multi-source acquisition methods, including: acquiring biotoxicity data based on biological models; acquiring pollutant data based on spectral or mass spectrometry analysis; and acquiring water physicochemical parameter data based on sensors. A water ecological health evaluation index system was constructed based on the aforementioned water quality data. Based on historical and current water quality data, a water quality parameter prediction model is constructed using machine learning algorithms. The current water quality data is input into the water quality parameter prediction model to obtain the prediction results of future water quality parameters. Based on the aforementioned water ecological health evaluation index system and the predicted results of future water quality parameters, a comprehensive assessment of the current status and future trends of water ecological health is conducted.

[0005] Furthermore, the biological model includes primary producers, primary consumers, and secondary consumers and decomposers, and the physicochemical parameters of the water body include temperature, pH, dissolved oxygen, conductivity, turbidity, redox potential, and chlorophyll a.

[0006] Further, the construction of the aquatic ecological health evaluation index system based on the water quality data includes: calculating the water body physicochemical health index: B1=Σ(Wi*Ni), where B1 is the water body physicochemical health index, Wi is the weight of the i-th physicochemical index, Ni is the score of the i-th physicochemical index after standardization and dimensionless processing, Ni is between 0 and 1, and 1 represents the optimal state; calculating the biotoxicity index: B2=Σ(Wj*Tj), where B2 is the biotoxicity index, Wj is the weight of the j-th biotoxicity index, and Tj is the standardized test result of the j-th biotoxicity index; determining α and β based on multiple regression analysis and / or structural equation model, where α and β are the integrated weights of the water body physicochemical health index and the comprehensive biotoxicity index, and α+β=1; calculating the comprehensive aquatic ecological health index: WEHI=α*B1+β*B2 based on the water body physicochemical health index and the comprehensive biotoxicity index, where WEHI is the comprehensive aquatic ecological health index.

[0007] Furthermore, the step of constructing a water quality parameter prediction model based on historical and current water quality data using machine learning algorithms includes: preprocessing the water quality data to construct training samples for the machine learning model; constructing the water quality parameter prediction model using random forest regression algorithm, gradient boosting regression tree algorithm, and / or long short-term memory network regression algorithm; and training and optimizing the water quality parameter prediction model based on the training samples.

[0008] Furthermore, the weights of the physicochemical indicators and the weights of the biotoxicity indicators are determined using a multi-parameter relative weighting method.

[0009] Furthermore, the preprocessing of the water quality data includes: filling missing values ​​using interpolation or data from previous and subsequent time points; identifying and correcting outliers using Z-score or isolated forest algorithms; standardizing parameters with different dimensions using Z-score or Min-Max normalization; and constructing time-series features, including historical lag features, moving statistical features, periodic features, and external collaborative features.

[0010] Furthermore, the training and optimization of the water quality parameter prediction model based on the training samples includes: dividing the time-series features into a training set, a validation set, and a test set in chronological order; using mean squared error or mean absolute error as a loss function to train the model on the training set; tuning the key hyperparameters of the model to find the optimal configuration on the validation set using grid search or Bayesian optimization algorithms; using root mean square error, mean absolute percentage error, and coefficient of determination to quantify the prediction accuracy and generalization ability of the model on the test set; deploying the best trained model to the production environment; and automatically executing preprocessing and feature construction processes when real-time monitoring data flows in, inputting the constructed feature vectors into the deployed model to obtain predicted water quality parameters for future periods.

[0011] Furthermore, the comprehensive assessment of the current status and future trends of water ecological health is conducted by combining the water ecological health evaluation index system and the prediction results of the future water quality parameters, including: quantifying the current health status and predicting the future health status; conducting scenario simulations based on the current and future health status; and generating a comprehensive assessment report and decision-making recommendations based on the simulated scenarios.

[0012] Furthermore, the quantification of the current health status and the prediction of the future health status include: obtaining the current health score and the current health status level based on the current water body physicochemical health index and comprehensive biological toxicity index; and calculating the future health score and the future health status level corresponding to the predicted results of future water quality parameters according to the water ecological health evaluation index system.

[0013] Furthermore, the scenario simulation based on the current and future health status includes: comparing the relevant factors of the current health score with the relevant factors of the future health score to obtain the potential risks reflected by the water body's physicochemical health index and comprehensive biotoxicity index; simulating the improvement effect of the relevant factors of the current and future health scores on the comprehensive water ecological health index, providing a quantitative basis for selecting the optimal treatment plan.

[0014] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: The water ecological health assessment method provided by this invention simultaneously acquires biotoxicity data based on biological models, pollutant data based on spectral or mass spectrometry analysis, and water physicochemical parameter data based on sensors through multi-source acquisition. This method synergistically characterizes the chemical composition, physical state, and biological effects of water bodies, overcoming the shortcomings of traditional water quality assessment methods that rely only on a limited number of physicochemical parameters, have a single data dimension, and are unable to reflect comprehensive ecological risks. By constructing a water ecological health evaluation index system, it unifies and comprehensively quantifies multi-source heterogeneous water quality data, introducing biotoxicity effect indicators and pollutant screening results. This allows the assessment results to simultaneously reflect known pollutants, unknown pollutants, and their comprehensive biological effects, significantly improving the comprehensiveness and scientific rigor of water ecological health assessment. In one step, based on historical and current water quality data, a water quality parameter prediction model is constructed using machine learning algorithms. The current water quality data is then input into the prediction model to obtain prediction results for future water quality parameters. This achieves a shift from simple status quo evaluation to trend prediction and risk warning, effectively overcoming the shortcomings of traditional assessment methods, such as assessment lag and lack of foresight. At the same time, the automated acquisition and modeling of multi-source data significantly improves the throughput and efficiency of water quality monitoring and analysis, supporting the needs of large-scale, high-frequency water ecological monitoring. Furthermore, by constructing an evaluation index system through a data-driven approach, the subjective influence of expert experience on index weights is reduced, thus providing more objective and reliable technical support for refined management and sustainable decision-making in water ecological health. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of the water ecological health assessment method provided in Embodiment 1 of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. It should be noted that, unless otherwise specified, the implementation methods and features in the implementation methods in this disclosure can be combined, separated, interchanged, and / or rearranged. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] Example 1 A specific embodiment of the present invention, such as Figure 1 As shown, a method for assessing the health of aquatic ecosystems is disclosed, comprising the following steps: S1. Obtain water quality data from multiple samples through multi-source acquisition.

[0019] In this embodiment, step S1 specifically includes: S11. Collect biotoxicity data based on biological models; Specifically, the biological model includes primary producers, primary consumers, and secondary consumers and decomposers.

[0020] For example, primary producers include algae such as Scenedesmus obliquus and Chlorella vulgaris; primary consumers include daphnia such as Daphnia macrocarpa and Daphnia reticulata; and secondary consumers and decomposers include zebrafish embryos, bioluminescent bacteria, and large benthic animals (such as chironomid larvae). This multi-trophic-level biological model design can sensitively respond to different types of pollutants (such as neurotoxins, photosynthesis inhibitors, and genotoxic substances), providing comprehensive information on biological effects. For the above biological model, a standardized rapid detection method for 65 biotoxicity indicators is adopted. The detection endpoints include, but are not limited to, mortality rate, inhibition rate, bioluminescence intensity inhibition rate, abnormal motility, embryonic developmental malformation, and changes in gene expression profiles. Through an automated liquid handling workstation and a multi-channel microplate detection system, the throughput of a single test can reach up to 2160 times, realizing efficient and batch acquisition of biotoxicity data.

[0021] S12. Collect pollutant data based on spectral or mass spectrometry analysis; Specifically, for known priority pollutants (such as polycyclic aromatic hydrocarbons, organophosphorus pesticides, antibiotics, and heavy metals), liquid chromatography-tandem mass spectrometry (LC-MS / MS) and gas chromatography-mass spectrometry (GC-MS) combined with isotope internal standard methods are used for precise qualitative and quantitative analysis to ensure data accuracy and comparability. Simultaneously, non-targeted screening techniques based on high-resolution mass spectrometry (HRMS) are used to discover novel and unknown pollutants. Data-dependent acquisition (DDA) and data-independent acquisition (DIA) comprehensively capture the mass spectrometry information of all ionizable compounds in the water sample. The acquired primary and secondary mass spectrometry data are compared with a self-built standard database containing information on 33,017 pollutants to achieve rapid screening of large-scale pollutants. Utilizing retention time, fragment ion differences, and predictive models, the screening results can be further used to identify and distinguish isomers of 8,163 key pollutants, greatly improving the accuracy of pollutant identification.

[0022] S13. Collect sensor-based water physicochemical parameter data.

[0023] Specifically, the physicochemical parameters of the water body include temperature, pH value, dissolved oxygen, conductivity, turbidity, redox potential, and chlorophyll a.

[0024] In this embodiment, the sensor-based physicochemical parameter data acquisition aims to obtain basic, real-time, and continuous physicochemical state parameters of the water body, providing background environmental and dynamic change information for the ecosystem. By deploying an integrated multi-parameter water quality sensor array, key physicochemical parameters such as temperature, pH, dissolved oxygen (DO), conductivity, turbidity, oxidation-reduction potential (ORP), and chlorophyll a are monitored in situ, in real-time, and continuously. Sensor data is transmitted in real-time to a cloud data center via Internet of Things (IoT) technologies (such as LoRa, NB-IoT, or 5G networks), forming a long-term continuous observation dataset, laying the data foundation for subsequent trend analysis and model prediction.

[0025] By employing the three multi-source acquisition methods described above, information on biological effects, chemical substances, and physical states is simultaneously obtained, overcoming the limitations of a single data source. Biotoxicity data directly reflects the potential risks of water bodies to the ecosystem, enabling health assessment to shift from "chemical presence" to "biological effects." The combination of high-throughput biological detection and high-precision mass spectrometry analysis ensures large-scale sample processing capabilities without sacrificing data depth and accuracy. Continuous monitoring by the sensor network provides a data stream for real-time early warning and dynamic models, enabling ecological health assessment to shift from post-event analysis to pre-event prediction.

[0026] S2. Construct an aquatic ecological health evaluation index system based on the water quality data.

[0027] In this embodiment, step S2 specifically includes: S21. Calculate the physicochemical health index of water bodies: B1=Σ(Wi*Ni); Wherein, B1 is the water body physicochemical health index, Wi is the weight of the i-th physicochemical index, Ni is the score of the i-th physicochemical index after standardization and dimensionless processing, Ni is between 0 and 1, and 1 represents the best state.

[0028] Specifically, physicochemical indicators include single indicators formed by physicochemical parameters, composite indicators calculated from multiple physicochemical parameters, and derived indicators. For example, composite indicators include nutrient factors, which are calculated by weighting parameters such as pH, ammonia nitrogen density, nitrate concentration, and chloride concentration after normalization. Derived indicators include physical stability factors such as liquid level change rate (reflecting hydrological stability) and turbidity coefficient (usually characterized by turbidity value).

[0029] S22. Calculate the biotoxicity index: B2=Σ(Wj*Tj); Where B2 is the biotoxicity index, Wj is the weight of the j-th biotoxicity index, and Tj is the standardized result of the j-th biotoxicity test.

[0030] Specifically, the biotoxicity index directly quantifies the comprehensive biological effects of water bodies by integrating the toxicity test results of multiple biological models. Tj is the standardized toxicity unit of the j-th biotoxicity test result. For example, inhibition rate, mortality rate, etc. are converted into a 0-1 scale. Biotoxicity indicators include acute toxicity indicators (such as the 48-hour mortality rate of Daphnia macrocarpa and the 15-minute luminescence inhibition rate of luminescent bacteria), chronic toxicity indicators (such as the 96-hour growth inhibition rate of algae and the embryonic malformation rate of zebrafish), and specific toxicity indicators (such as genotoxicity and endocrine disruption effects).

[0031] In this embodiment, the weights of the physicochemical indicators and the biotoxicity indicators are determined using a multi-parameter relative weighting method. This method objectively assigns weights based on the volatility and conflict of each indicator's data, avoiding human bias.

[0032] In some embodiments, taking into account the importance of organisms at different trophic levels in the ecosystem and their differences in sensitivity to pollutants, the weights can also be determined by combining the analytic hierarchy process (AHP) with the entropy weight method.

[0033] S23. Determine α and β based on multiple regression analysis and / or structural equation modeling; Wherein, α and β are the integrated weights of the water body physicochemical health index and the comprehensive biological toxicity index, respectively, and α+β=1.

[0034] Specifically, multiple regression analysis or structural equation modeling (SEM) is employed, using historical ecological disaster events or health status determined by authoritative experts as the dependent variable, and B1 and B2 as independent variables, to derive the weighting coefficients α and β that best predict ecological health status. This method ensures a strong statistical correlation between the comprehensive index and the actual ecological health status, making the assessment results more scientific and convincing.

[0035] S24. Calculate the comprehensive water ecological health index based on the water body's physicochemical health index and comprehensive biological toxicity index: WEHI=α*B1+β*B2; Among them, WEHI is the comprehensive water ecological health index.

[0036] In this embodiment, biotoxicity, the ultimate effect indicator, is incorporated into the core evaluation system, elevating the assessment result from "water quality chemical compliance" to the level of "ecological health and safety." The use of objective weighting methods based on data characteristics (entropy weighting method, CRITIC method) significantly reduces the interference of subjective human factors on the evaluation results. Preliminary diagnosis can be made by analyzing the differences in scores between criteria layers B1 and B2. For example, a high B1 score and a low B2 score strongly suggest the presence of priority pollutants or emerging pollutants in the water body that are not covered by routine monitoring, thus guiding managers to conduct more in-depth pollution source investigations. This framework is highly flexible and scalable, allowing for dynamic adjustment of specific parameters in the indicator layers according to the characteristics (e.g., rivers, lakes, reservoirs) and main functions of different water bodies, achieving customized assessments.

[0037] S3. Based on historical water quality data and current water quality data, a water quality parameter prediction model is constructed using machine learning algorithms. The current water quality data is input into the water quality parameter prediction model to obtain the prediction results of future water quality parameters.

[0038] In this embodiment, step S3 specifically includes: S31. Preprocess the water quality data to construct training samples for the machine learning model; Specifically, step S31 includes: S311. Use interpolation or data from previous and next time steps to fill in missing values, and use Z-score or isolated forest algorithm to identify and correct outliers. In this embodiment, the original multi-source water quality data needs to undergo rigorous preprocessing in order to construct a training set suitable for machine learning models.

[0039] S312. Perform Z-score standardization or Min-Max normalization on parameters with different dimensions; In this embodiment, Z-score standardization or Min-Max normalization is performed on various parameters with different dimensions (such as temperature, concentration, toxicity units, etc.) to eliminate the influence of dimensions and accelerate model convergence.

[0040] S313. Construct time series features, which include historical lag features, moving statistical features, periodic features, and external collaborative features.

[0041] In this embodiment, for the target prediction parameter (such as the ammonia nitrogen concentration in the next 7 days), the following types of input features need to be constructed: Historical lag characteristics: such as the same parameter value in the past 1 day, 3 days, 7 days, and 30 days.

[0042] Moving statistical characteristics: such as the moving average and moving standard deviation of parameters over the past 7 days.

[0043] Periodic features: For data with obvious periodicity (such as annual or weekly periods), extract features such as "the day of the year" and "the day of the week".

[0044] External synergistic features: Meteorological data (such as temperature, rainfall, and wind speed), hydrological data (such as flow velocity and flow rate), and other highly relevant water quality parameters (such as salinity, turbidity, and dissolved oxygen) are introduced as model inputs to fully utilize the synergistic relationship between variables.

[0045] S32. The water quality parameter prediction model is constructed using the random forest regression algorithm, the gradient boosting regression tree algorithm, and / or the long short-term memory network regression algorithm. In this embodiment, the random forest regression algorithm works by constructing and integrating multiple decision trees, ultimately outputting the mean of all tree predictions. It possesses strong non-linear fitting capabilities and can assess the importance of features. The random forest regression algorithm is particularly suitable for scenarios where the relationship between input features and output targets is unclear, but a strong correlation exists. For example, it can be used to predict composite indicators such as nutrient salt factors or turbidity coefficients. Its built-in feature importance ranking function helps to screen key influencing factors and optimize the model.

[0046] Long Short-Term Memory (LSTM) network regression is a special type of recurrent neural network. Its internal gating mechanism (input gate, forget gate, output gate) effectively learns long-term dependencies in time series data, overcoming the gradient vanishing problem of traditional RNNs. It is the preferred algorithm for processing water quality time series data. Its model structure is naturally suited for learning the dynamic impact of historical water quality data on future trends. It is particularly suitable for predicting parameters with strong autocorrelation and periodicity, such as diurnal variations of dissolved oxygen (DO) and hourly variations of temperature. Its input is a time-ordered sequence of features, and its output is a predicted value for one or more future time steps.

[0047] The gradient boosting regression tree algorithm works by sequentially training a series of decision trees, each attempting to correct the prediction errors of the previous tree. This "boosting" strategy combines them into a powerful final model. Similar to random forests, the gradient boosting regression tree algorithm can handle complex nonlinear relationships and achieves higher prediction accuracy in most cases. It is suitable for precise point prediction of various water quality parameters.

[0048] S33. The water quality parameter prediction model is trained and optimized based on the training samples.

[0049] Specifically, step S33 includes: S331. Divide the time series features into training set, validation set and test set according to the time sequence; In this embodiment, the preprocessed time-series feature data is divided into a training set (e.g., the first 70% of the data), a validation set (the middle 15%), and a test set (the last 15%) according to time order. Random shuffling is strictly prohibited to avoid data leakage.

[0050] S332. On the training set, use mean squared error or mean absolute error as the loss function to train the model. In this embodiment, mean squared error or mean absolute error is used as the loss function, and the model is trained by backpropagation (for LSTM) or fitting (for tree models).

[0051] S333. On the validation set, the key hyperparameters of the model are tuned using grid search or Bayesian optimization algorithms to find the optimal configuration. In this embodiment, the key hyperparameters of the model (such as the number of layers and neurons of LSTM, and the tree depth and number of trees of random forest) are tuned using a validation set through grid search or Bayesian optimization algorithms to find the optimal configuration.

[0052] S334. On the test set, the root mean square error, mean absolute percentage error, and coefficient of determination are used to quantify the model’s prediction accuracy and generalization ability. In this embodiment, the root mean square error, mean absolute percentage error, and coefficient of determination are used on an independent test set to quantify the model's prediction accuracy and generalization ability.

[0053] S335. Deploy the trained best model to the production environment; when real-time monitoring data flows in, the system automatically executes the preprocessing and feature construction process, inputting the constructed feature vector into the deployed model to obtain the predicted water quality parameters for future periods.

[0054] In this embodiment, the trained optimal model is deployed to a production environment (such as a cloud server). When new real-time monitoring data flows in, the system automatically performs the same preprocessing and feature construction process, and then inputs the constructed feature vectors into the deployed model to obtain predicted water quality parameters for specific future periods. These predicted values ​​will serve as key inputs for downstream ecological health assessment and trend judgment.

[0055] In this way, machine learning algorithms can effectively capture the complex nonlinear dynamics and long-term dependencies of water quality parameters, achieving significantly higher prediction accuracy than traditional methods. Simultaneously, they can naturally integrate multi-source, heterogeneous data (physicochemical data, biotoxicity data, and meteorological data), fully unlocking the hidden value within big data. The model can be incrementally trained periodically with the latest data, adapting to slow changes in environmental conditions and maintaining the timeliness of its predictive capabilities. High-precision future water quality predictions enable a shift in environmental management from reactive response to proactive early warning and planning, gaining valuable time for implementing protective measures.

[0056] S4. Based on the water ecological health evaluation index system and the prediction results of the future water quality parameters, conduct a comprehensive assessment of the current status and future trends of water ecological health.

[0057] In this embodiment, step S4 specifically includes: S41. Quantify current health status and predict future health status; In this embodiment, step S41 includes: S411. Obtain the current health score and current health status level based on the current water body's physical and chemical health index and comprehensive biological toxicity index. Specifically, the multi-source water quality data (including physicochemical parameters, pollutant data, and biotoxicity data) at the current time (t0) is input into a pre-constructed water ecological health evaluation index system to calculate the current water ecological health comprehensive index (WEHI_t0) and its two core sub-indices: the water body physicochemical health index (B1_t0) and the comprehensive biotoxicity index (B2_t0). A quantitative current health score (e.g., WEHI_t0=68 / 100) and a health status level (e.g., "sub-healthy") are obtained.

[0058] S412. Based on the aforementioned water ecological health evaluation index system, calculate the future health score and future health status level corresponding to the predicted results of future water quality parameters.

[0059] Specifically, historical and current time-series data on water quality and meteorology are input into a pre-trained machine learning prediction model to obtain predicted values ​​of key water quality parameters at a specific future time point (e.g., t0+1 years). These predicted water quality parameters are then used as input again into the aquatic ecological health evaluation index system to calculate the predicted future health index WEHI_future. A quantified future health score (e.g., WEHI_future = 62 / 100) and its prediction level are obtained.

[0060] S42. Conduct scenario simulations based on current and future health status; In this embodiment, step S42 includes: S421. Compare the relevant factors of the current health score with the relevant factors of the future health score to obtain the potential risks reflected by the water body physicochemical health index and comprehensive biotoxicity index. Specifically, after obtaining preliminary judgment conclusions, the system conducts in-depth causal analysis to support the generation of management plans. Dominant factor analysis: Comparing the changes in sub-indices at the current level (B1_t0, B2_t0) with the predicted future level (B1_future, B2_future). For example, if the prediction shows that the decline in WEHI is mainly driven by B1 (physicochemical index), the core problem lies in the increase in pollutant concentration or the deterioration of physical conditions; if it is mainly driven by B2 (biotoxicity index), it suggests the existence of risks of toxic substances that are not fully recognized.

[0061] S422. Simulate the effects of current and future health score correlation factors on the comprehensive water ecological health index to provide a quantitative basis for selecting the optimal governance solution.

[0062] Specifically, based on the predictive model, the improvement effect of different management measures (such as reducing the emission of a certain pollutant by 30%) on the future WEHI_future value is simulated, providing a quantitative basis for selecting the optimal governance solution.

[0063] S43. Generate a comprehensive evaluation report and decision recommendations based on the simulated scenario.

[0064] Specifically, by integrating all the above analyses, a structured report is generated, which includes: the current health score, level, and main strengths / weaknesses; the future direction, magnitude, and judgment level (such as "serious warning"); the dominant factors leading to the current problems and future trends; and, based on causal diagnosis and scenario simulation, targeted and prioritized control measures recommendations.

[0065] Compared with existing technologies, the water ecological health assessment method provided in this embodiment simultaneously acquires biotoxicity data based on biological models, pollutant data based on spectral or mass spectrometry analysis, and water physicochemical parameter data based on sensors through multi-source acquisition. This synergistically characterizes the chemical composition, physical state, and biological effects of water bodies, overcoming the shortcomings of traditional water quality assessment methods that rely only on a limited number of physicochemical parameters, have a single data dimension, and are unable to reflect comprehensive ecological risks. By constructing a water ecological health evaluation index system, it unifies and comprehensively quantifies multi-source heterogeneous water quality data, introducing biotoxicity effect indicators and pollutant screening results. This enables the assessment results to simultaneously reflect known pollutants, unknown pollutants, and their comprehensive biological effects, significantly improving the comprehensiveness and scientific validity of water ecological health assessment. Furthermore, based on historical and current water quality data, a water quality parameter prediction model is constructed using machine learning algorithms. Current water quality data is then input into the prediction model to obtain predictions of future water quality parameters. This achieves a shift from simple status quo assessment to trend prediction and risk warning, effectively overcoming the shortcomings of traditional assessment methods, such as assessment lag and lack of foresight. Simultaneously, the automated acquisition and modeling of multi-source data significantly improves the throughput and efficiency of water quality monitoring and analysis, supporting large-scale, high-frequency water ecological monitoring needs. Moreover, by constructing an evaluation index system through a data-driven approach, the subjective influence of expert experience on index weights is reduced, thus providing more objective and reliable technical support for refined management and sustainable decision-making regarding water ecological health.

[0066] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the health of aquatic ecosystems, characterized in that, include: Water quality data from multiple samples was acquired through multi-source acquisition methods, including: acquiring biotoxicity data based on biological models; acquiring pollutant data based on spectral or mass spectrometry analysis; and acquiring water physicochemical parameter data based on sensors. A water ecological health evaluation index system was constructed based on the aforementioned water quality data. Based on historical and current water quality data, a water quality parameter prediction model is constructed using machine learning algorithms. The current water quality data is input into the water quality parameter prediction model to obtain the prediction results of future water quality parameters. Based on the aforementioned water ecological health evaluation index system and the predicted results of future water quality parameters, a comprehensive assessment of the current status and future trends of water ecological health is conducted.

2. The water ecological health assessment method according to claim 1, characterized in that, The biological model includes primary producers, primary consumers, and secondary consumers and decomposers. The physicochemical parameters of the water body include temperature, pH, dissolved oxygen, conductivity, turbidity, redox potential, and chlorophyll a.

3. The water ecological health assessment method according to claim 1, characterized in that, The water ecological health evaluation index system constructed based on the water quality data includes: Calculate the water body physicochemical health index: B1=Σ(Wi*Ni), where B1 is the water body physicochemical health index, Wi is the weight of the i-th physicochemical index, Ni is the score of the i-th physicochemical index after standardization and dimensionless processing, Ni is between 0 and 1, and 1 represents the best state; Calculate the biotoxicity index: B2=Σ(Wj*Tj), where B2 is the biotoxicity index, Wj is the weight of the j-th biotoxicity index, and Tj is the standardized test result of the j-th biotoxicity index. α and β are determined based on multiple regression analysis and / or structural equation modeling. α and β are the integrated weights of the water body physicochemical health index and the comprehensive biological toxicity index, respectively, and α+β=1. The comprehensive water ecological health index is calculated based on the water body's physicochemical health index and comprehensive biological toxicity index: WEHI=α*B1+β*B2, where WEHI is the comprehensive water ecological health index.

4. The water ecological health assessment method according to claim 1, characterized in that, The water quality parameter prediction model, constructed using machine learning algorithms based on historical and current water quality data, includes: The water quality data is preprocessed to construct training samples for a machine learning model; the water quality parameter prediction model is constructed using a random forest regression algorithm, a gradient boosting regression tree algorithm, and / or a long short-term memory network regression algorithm; the water quality parameter prediction model is trained and optimized based on the training samples.

5. The water ecological health assessment method according to claim 4, characterized in that, The weights of the physicochemical indicators and the weights of the biotoxicity indicators were determined using a multi-parameter relative weighting method.

6. The water ecological health assessment method according to claim 4, characterized in that, The preprocessing of the water quality data includes: Missing values ​​are handled by interpolation or data from previous and next time steps, and outliers are identified and corrected by Z-score or isolated forest algorithms. Z-score standardization or Min-Max normalization is performed on parameters with different dimensions; Construct time-series features, which include historical lag features, moving statistical features, periodic features, and external collaborative features.

7. The water ecological health assessment method according to claim 4, characterized in that, The training and optimization of the water quality parameter prediction model based on the training samples includes: The temporal features are divided into training set, validation set and test set according to the time sequence. On the training set, mean squared error or mean absolute error is used as the loss function to train the model; On the validation set, the key hyperparameters of the model are tuned using grid search or Bayesian optimization algorithms to find the optimal configuration; On the test set, root mean square error, mean absolute percentage error, and coefficient of determination are used to quantify the model’s prediction accuracy and generalization ability. The trained best model is deployed to the production environment; when real-time monitoring data flows in, the system automatically performs preprocessing and feature construction processes, inputting the constructed feature vectors into the deployed model to obtain predicted values ​​of water quality parameters for future periods.

8. The water ecological health assessment method according to claim 3, characterized in that, The method of comprehensively assessing the current status and future trends of water ecological health by combining the water ecological health evaluation index system and the prediction results of future water quality parameters includes: Quantify current health status and predict future health status; Scenario simulations are conducted based on current and future health status. A comprehensive evaluation report and decision-making recommendations are generated based on the simulated scenario.

9. The water ecological health assessment method according to claim 8, characterized in that, The quantitative assessment of current health status and the prediction of future health status include: The current health score and current health status level are obtained based on the current water body's physicochemical health index and comprehensive biological toxicity index. Based on the aforementioned water ecological health evaluation index system, calculate the future health score and future health status level corresponding to the predicted results of future water quality parameters.

10. The water ecological health assessment method according to claim 9, characterized in that, The scenario simulation based on current and future health status includes: By comparing the factors related to the current health score with those related to the future health score, we can obtain the potential risks reflected by the water body's physicochemical health index and comprehensive biotoxicity index. The simulation of the effects of factors related to current and future health scores on the comprehensive water ecological health index provides a quantitative basis for selecting the optimal governance solution.

Images