Intermittent river ecosystem degradation diagnosis and regulation method and system and storage medium
By using eDNA processing and a hierarchical Bayesian causal network model, the problem of intermittent river ecosystem assessment bias was solved, enabling accurate diagnosis and ecological regulation of major stressors, and providing intelligent means of ecosystem monitoring and regulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HOHAI UNIV
- Filing Date
- 2026-03-02
- Publication Date
- 2026-06-30
AI Technical Summary
Existing technologies are insufficient for accurate diagnosis and stress tracing of intermittent river ecosystems, and the assessment methods are incompatible, resulting in assessment bias and poor data reliability. They are unable to effectively monitor biological community information during dry periods and lack early warning and control measures.
By combining eDNA processing with a hierarchical Bayesian causal network model, and simultaneously collecting water samples during wet periods and sediment samples during dry and wet periods, a causal inference model was constructed. This model, combined with prior ecological knowledge and a data-driven hierarchical Bayesian causal network, was used for the diagnosis and regulation of ecosystem degradation.
It enables intelligent diagnosis of intermittent river ecosystems, identifies major stressors and their impact mechanisms, provides visualized reports and ecological regulation suggestions, shortens the decision-making cycle, and improves the reliability and objectivity of the assessment.
Smart Images
Figure CN121743847B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of environmental monitoring and ecological assessment technology, and in particular to a method, system and storage medium for diagnosing and regulating the degradation of intermittent river ecosystems. Background Technology
[0002] Intermittent rivers are one of the most widespread river types globally, accounting for over 50% of the world's total river length. These rivers do not have continuous flow throughout the year, but rather experience periodic drying and reflow. Widely distributed in arid and semi-arid regions, they are a core component of regional ecosystems. Their key characteristics include dramatic hydrological fluctuations, strong spatial and temporal variability in flow and water level, a sedimentary bed composed primarily of sand and pebbles, and a biological community centered on drought-tolerant organisms, such as benthic animals with dormancy strategies and drought-resistant algae. Within a year, intermittent rivers experience four phases: a flowing period, a non-flowing period, a drying period, and a reflow period. This dramatic change in hydrological phases makes intermittent river ecosystems more vulnerable and susceptible to human disturbances (such as insufficient ecological flow and pollution), leading to prolonged drying periods, decreased biological recovery capacity, and ultimately disrupting the natural "stress-response-recovery" balance of the ecosystem, resulting in a certain degree of degradation of the river ecosystem.
[0003] Global attention to intermittent rivers is growing, but targeted early warning methods for ecosystem degradation remain lacking. Specifically, existing river and lake ecosystem health assessment methods are primarily designed for perennial rivers and are ill-suited to intermittent river scenarios: First, there are gaps in dry-season monitoring; traditional methods rely on water samples, failing to obtain information on biological communities during dry periods, leading to a disconnect between wet and dry season data and insufficient assessment completeness. Second, attribution is ambiguous, failing to effectively distinguish between ecological changes caused by natural hydrological fluctuations and health degradation caused by human disturbance. Third, assessment indicators and methods are incompatible; the general Index of Biological Integrity (IBI) does not consider the specific drought-resistant biological groups in intermittent rivers, resulting in assessment bias, and simple weighted averages do not account for the cascading effects between stress, hydrology, and biology at the causal mechanism level. Fourth, monitoring equipment is ill-suited to the environment of intermittent rivers with high sediment content and drastic water level fluctuations, resulting in poor data reliability. Therefore, developing a technology that can adapt to the dynamic characteristics of intermittent rivers and intelligently diagnose the root causes of their ecological problems has become an urgent need. Summary of the Invention
[0004] The purpose of this invention is to provide a method, system, and storage medium for the diagnosis and regulation of intermittent river ecosystem degradation, so as to solve the problem that existing technologies are unable to achieve accurate diagnosis, stress source tracing, and scenario simulation of intermittent river ecosystem degradation mechanisms.
[0005] This invention provides a method for diagnosing and regulating intermittent river ecosystem degradation, comprising the following steps:
[0006] Step 1: Based on the intermittent river catchment area and the characteristics of alternating wet and dry periods, sampling points covering different hydrological zones and land use types are set up. Water samples during the wet period and sediment samples during the wet and dry periods are collected simultaneously, along with hydrological parameters, habitat data, and spatiotemporal location information.
[0007] Step 2 involves processing the collected water samples from wet periods and sediment samples from dry and wet periods with eDNA and performing scenario-based biological information analysis, including eDNA extraction, purification and macrobarcode sequencing, screening for drought-resistant biological groups, calculating the IBI feature parameters specific to at least two different biological groups in the intermittent river, and forming a biological response feature vector.
[0008] Step 3: Integrate eDNA scenario-based bioinformatics analysis data, hydrological parameters, habitat data, and historical monitoring data to construct a comprehensive dataset. After outlier removal, missing value imputation, and standardization, extract core feature variables.
[0009] Step four: Input the biological response feature vector along with the hydrological parameters and habitat data into a pre-constructed causal inference model for diagnostic analysis, construct and train the causal inference model; wherein, the causal inference model is a hierarchical Bayesian causal network model that integrates ecological prior knowledge and data-driven approaches, and the causal inference model includes:
[0010] The first layer, the ecological relationship knowledge graph layer, is a static semantic network constructed based on ecological principles, describing the inherent causal and correlational relationships between stressors, environmental factors, and biological indicators.
[0011] The second layer is the dynamic Bayesian causal network layer. Its network structure is initialized by the knowledge graph layer and uses historical data to learn the conditional probability distribution between nodes. It is used to perform probabilistic reasoning based on real-time observation data to identify the most likely stress sources and causal paths.
[0012] The third layer, the interpretability output and scenario simulation layer, is used to generate a visual diagnostic report that includes the contribution of stressors and causal paths, and to provide ecological response prediction simulation based on adjusting the state of stressors.
[0013] Step 5: Input the watershed data to be diagnosed into the causal inference model, output the integrated ecological health level of dry and wet seasons, key driving factors and their contributions, and generate a visual report containing ecological regulation recommendations.
[0014] Furthermore, in step one, 3-5 repeated sampling points are set up in each ecological functional area, with a sampling point spacing of not less than 500 meters; the water sample collection depth is 0cm-30cm during the wet season and the sediment sample collection depth is 0cm-5cm during the dry season; after collection, the samples are transported in a light-proof refrigerated environment at 0℃-4℃ and pre-processed within 24 hours.
[0015] The hydrological parameters include flow rate, peak flow time within the year, and duration of flow interruption; the habitat parameters include habitat physical characteristics, water body characteristics, and other environmental characteristics; wherein, the habitat physical characteristics include river channel physical characteristics and riparian zone conditions; the water body characteristics include hydraulic characteristics and water body physicochemical parameters; the other environmental characteristics include altitude, temperature, and precipitation, or can be collected in real time through existing in-situ sand and water prevention sensor arrays.
[0016] Furthermore, in step two, the eDNA processing and scenario-based bioinformatics analysis involve eDNA extraction using -80℃ freeze-drying and microwave-assisted extraction methods, with sediment sample extraction parameters of 500W-800W power and 15min-20min time; macrobarcoding sequencing uses primers designed for barcode regions specific to drought-resistant biological groups, and species annotation is completed based on NCBI, BOLD databases, and the intermittent river local species list; multi-group biological community data are amplified by multiplex PCR using eDNA macrobarcoding technology and specific primers for different biological groups, and at least two types of biological community data from fish, benthic macroinvertebrates, and attached algae are obtained in an integrated manner.
[0017] The specific IBI characteristic parameters include resistance parameters and resilience parameters. The resistance parameter reflects the organism's ability to resist environmental stress during drought, while the resilience parameter reflects the organism's ability to restore its population and community structure after rewetting. The weight ratio of the resistance parameter and resilience parameter for each species is calculated using the entropy method to obtain the specific IBI characteristic parameters for the species.
[0018] Furthermore, the ability to resist environmental stress during drought is the ability to survive under conditions of no water or low flow, including the proportion of dormant bodies, which is the percentage of dormant eggs or cysts in the sediment relative to the total number of individuals of the species; the ability to restore population and community structure after rewetting is the speed and extent of recovery from drought stress, including aerial dispersal ability, which includes the percentage of species with flying adults and the drift dispersal rate; the drift dispersal rate is the density of individuals of the species drifting downstream from the upstream perennial flow section after reflow.
[0019] Furthermore, in step three, outliers are removed using a box plot method with an IQR of 1.5; missing values are filled using K-nearest neighbor interpolation with k=5; and Z-score standardization is used to eliminate dimensional differences.
[0020] Furthermore, in step four, the ecological relationship knowledge graph layer of the causal inference model includes one or more of the relationships between stressors, environmental factors and biological indicators, such as inhibition, promotion and causation, which are used to constrain and initialize the network structure of the dynamic Bayesian causal network layer to prevent causal connections that violate ecological principles.
[0021] The reasoning process of the dynamic Bayesian causal network layer includes: inputting real-time observation data as evidence into the network nodes, using Bayesian inference algorithms to calculate the posterior probability of each pressure source node and system health status node, and outputting the diagnostic conclusion of the dominant pressure type and its confidence level in probabilistic form.
[0022] The interpretable output and the interpretable output of the scenario simulation layer are based on the reasoning results of the dynamic Bayesian causal network layer, generating a visualized and interpretable diagnostic report, including a quantitative decomposition of the types of major stressors and their contributions, and a key stress-biological response causal path diagram.
[0023] Furthermore, in step five, the simulation regulation prediction results are output, and the ecological health level is divided into five levels: excellent, good, average, poor, and very poor based on the exclusive IBI value. The key driving factors are distinguished between natural factors and human factors, and the contribution of each factor is quantified. The ecological regulation recommendations include ecological flow replenishment schemes and dry season sediment pollution control measures.
[0024] In a second aspect, the present invention provides a system for diagnosing and regulating intermittent river ecosystem degradation, used in the intermittent river ecosystem degradation diagnosis and regulation method described in the first aspect, comprising:
[0025] An integrated data acquisition terminal is used to simultaneously collect water samples during wet periods and sediment samples during dry and wet periods, and to collect hydrological parameters, habitat image data and spatiotemporal positioning information.
[0026] The data communication and preprocessing module is used for data transmission, storage, and preprocessing of raw data;
[0027] The cloud-based intelligent analysis platform is communicatively connected to the integrated data acquisition terminal. The cloud-based intelligent analysis platform includes: a bioinformatics analysis unit for processing eDNA data and calculating multi-group-specific IBI feature vectors; a causal inference model engine for running hierarchical Bayesian causal network models; a knowledge graph management unit for storing and updating intermittent river-specific ecological relationship knowledge graphs; and a scenario simulation and visualization unit for generating diagnostic reports and simulation results.
[0028] Furthermore, the integrated data acquisition terminal integrates an eDNA water sample filtration device, a multi-parameter hydrological data sensor, and an image acquisition module, and is equipped with a data synchronization and positioning module to ensure that all acquired data have a unified time and space reference.
[0029] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect.
[0030] The beneficial effects of this invention are as follows: By conducting in-depth collaborative analysis and causal inference of multi-group biological data, this invention overcomes the shortcomings of traditional methods in terms of fuzzy attribution. It can identify the main stressors and their influencing mechanisms in a probabilistic form. Its automated and intelligent analysis process reduces over-reliance on expert experience, shortens the cycle from data to decision, and makes the conclusions more repeatable and objective. The built-in scenario simulation function can predict the ecological response under different management measures, providing a powerful quantitative decision support tool for the formulation and comparison of river and lake protection and restoration schemes. Attached Figure Description
[0031] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.
[0032] Figure 1 This is a schematic diagram of the workflow of the intermittent river ecosystem degradation diagnosis and control method of the present invention;
[0033] Figure 2 This is a schematic diagram of the three-layer structure of the causal inference model of the present invention. Detailed Implementation
[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The technical solutions provided by various embodiments of this invention will be described in detail below with reference to the accompanying drawings.
[0035] Please see Figure 1 This invention provides a method for diagnosing and regulating intermittent river ecosystem degradation, comprising the following steps:
[0036] Step 1, Defining the monitoring scope and collecting dual samples: Based on the intermittent river catchment area and the characteristics of alternating wet and dry periods, sampling points covering different hydrological zones and land use types are set up. Water samples during the wet period and sediment samples during the wet and dry periods are collected simultaneously, along with hydrological parameters, habitat data and spatiotemporal location information.
[0037] Sampling points were set up in accordance with the intermittent wet and dry characteristics of rivers to ensure coverage of all hydrological zones, land use types and key habitats. 3-5 duplicate points were set up in each ecological functional zone, with a distance of not less than 500 meters between points. The water sample collection depth was 0cm-30cm during the wet period and the sediment sample collection depth was 0cm-5cm during the dry period. After collection, the samples were transported in a dark refrigerated environment at 0℃-4℃ and pre-processed within 24 hours.
[0038] Hydrological parameters include flow rate, peak flow time within the year, and duration of flow interruption. Habitat parameters include physical characteristics of the habitat, water body characteristics, and other environmental parameters. Physical characteristics of the habitat include river channel physical features and riparian zone conditions; water body characteristics include hydraulic and physicochemical parameters; other environmental parameters include altitude, temperature, and precipitation, or can be collected in real-time using existing in-situ sand and water control sensor arrays. River channel physical features include water connectivity, channel slope, and cross-sectional width-to-depth ratio. Riparian zone conditions include riparian vegetation cover, riparian land use, human disturbance, and riparian width. Hydraulic parameters include flow velocity, water depth, water surface width, and Froude number. Physicochemical parameters of the water body include dissolved oxygen, conductivity, total phosphorus, total nitrogen, ammonia nitrogen, nitrate nitrogen, phosphate, sulfate, biochemical oxygen demand, water temperature, and turbidity.
[0039] Step 2, Environmental DNA (eDNA) processing and scenario-based bioinformatics analysis: eDNA extraction, purification and macrobarcoding of the collected dual samples are performed to screen drought-resistant biological groups, calculate the IBI characteristic parameters of at least two different biological groups in the intermittent river, and form a biological response feature vector.
[0040] Specifically, eDNA extraction employed a -80℃ freeze-drying method combined with microwave-assisted extraction. Sediment sample extraction parameters were 500W-800W power and 15-20 minutes. Metabarcoding sequencing used primers designed for barcode regions specific to drought-tolerant biological groups. Species annotation was completed based on the National Center for Biotechnology Information (NCBI), the Barcode of Life Data Systems (BOLD), and the Local Species List of Intermittent Rivers. Multi-group biological community data were processed using eDNA metabarcoding technology and multiplex polymerase chain reaction (PCR) with specific primers for different biological groups. ChainReaction (PCR) amplification was used to acquire biocommunity data for at least two of the following: fish, benthic macroinvertebrates, and attached algae. Bioinformatics analysis was performed to calculate the specific IBI characteristic parameters for each species, including resistance and resilience parameters. Resistance parameters reflect an organism's ability to withstand environmental stress during drought, i.e., its ability to survive under conditions of no water or low flow (e.g., dormant body proportion: the percentage of dormant eggs / cysts in the sediment relative to the total number of individuals (including dormant bodies) of that species). Resilience parameters reflect an organism's ability to recover its population and community structure after rewetting, i.e., the speed and extent of recovery from drought stress (e.g., aerial dispersal ability: the percentage of flying adults (e.g., Diptera, Trichoptera) and drift dispersal rate: the density of individuals of a species drifting downstream from the upstream perennial flow section after reflow (ind. / m²). 2 •d) The weighted proportions of resistance and resilience parameters for each species are calculated using the entropy method to obtain its unique IBI characteristic parameters. The specific calculation method is as follows:
[0041] Specific IBI feature parameters: .
[0042] In the formula: The weight of the j-th parameter; Let be the standardized score of the j-th parameter; m is the number of parameters.
[0043] .
[0044] In the formula, The information entropy of each parameter; The difference between each parameter is represented by the degree of difference; the greater the difference, the higher the weight. n is the number of samples. Let be the proportion of the standardized score of the j-th parameter in the i-th sample. To avoid ln0, At that time, take ; The standardized values of each parameter in the i-th sample are obtained using the Z-score standardization method.
[0045] Step 3, Multi-source data integration and preprocessing: Integrate eDNA analysis data, hydrological parameters, habitat data and historical monitoring data to construct a comprehensive dataset. After outlier removal, missing value imputation and standardization, extract core feature variables.
[0046] Multi-source data includes in-situ monitored hydrological parameter data, land use data extracted by remote sensing, historical wet and dry period monitoring data, and habitat data such as human activity data. Data preprocessing uses box plot method to remove outliers (IQR=1.5); K-nearest neighbor interpolation method to fill missing values (k=5); and Z-score standardization to eliminate dimensional differences.
[0047] Specifically, the box plot method includes: First, calculating the interquartile range (IQR): IQR = Q3 - Q1. Q3 is the upper quartile, which is the value at the 75th percentile after the data is sorted in ascending order; Q1 is the lower quartile, which is the value at the 25th percentile after the data is sorted in ascending order. Then, determining the outlier thresholds: upper boundary = Q3 + 1.5 × IQR, lower boundary = Q1 - 1.5 × IQR; finally, traversing the data sequence, identifying data values greater than the upper boundary or less than the lower boundary as outliers, deleting them from the dataset or marking them for separate processing.
[0048] The K-nearest neighbor interpolation method includes: first, determining the Euclidean distance (applicable to continuous data), calculating the similarity between the missing sample and all other complete samples; then, finding the 5 nearest samples without missing values to each sample with a missing value; finally, assigning weights based on the distance between the nearest samples, with greater weights for closer samples; and calculating the imputation value for the missing position by weighted averaging. If the missing feature is a categorical variable, the category with the highest frequency among the nearest samples can be used for imputation.
[0049] Z-score standardization involves: after outlier removal and missing value imputation, calculating the mean and standard deviation of a certain indicator data, and then standardizing it using the following formula:
[0050]
[0051] Where Z represents the standardized data value; Y represents the original data; µ represents the average value of a certain indicator; and σ represents the standard deviation of a certain indicator.
[0052] Step 4, Hierarchical Bayesian Causal Network Model Construction and Training: The biological response feature vectors, along with hydrological and habitat data, are input into a pre-constructed causal inference model for diagnostic analysis. The biological response feature vectors are parameters or indicators that significantly respond to changes in stress sources. The causal inference model is a hierarchical Bayesian causal network model that integrates prior ecological knowledge with data-driven approaches. Please refer to [link to relevant documentation]. Figure 2 The model includes:
[0053] The first layer, the ecological relationship knowledge graph layer, is a static semantic network constructed based on ecological principles. It describes the inherent causal and correlational relationships between stressor environmental factors and biological indicators. This graph serves as prior ecological knowledge, defining the basic assumptions and constraints that allow connections between nodes in the model, ensuring that subsequent inferences conform to ecological principles. The relationships between stressors, environmental factors, and biological indicators in the first layer of the causal inference model include one or more of the following: inhibition, promotion, and causation. This layer is used to constrain and initialize the network structure of the second layer, the dynamic Bayesian causal network, preventing causal connections that violate ecological principles.
[0054] Specifically, stressors include anthropogenic stresses: agricultural non-point source pollution, industrial wastewater discharge, insufficient ecological flow, and riparian degradation. Natural stresses include increased drought intensity and flood pulses. Environmental factors include hydrological factors: water body connectivity, dry season duration, and flow pulses during refill periods. Water quality factors include: ammonia nitrogen concentration, dissolved oxygen, and turbidity. Habitat factors include: sediment complexity, vegetation cover, and sediment water content. Bioindicators include multi-group IBI characteristics: number of benthic EPT species, proportion of drought-tolerant groups, proportion of native fish species, reproductive strategies, algal diatom quotient, and proportion of pollution-tolerant species. EPT species refer to Ephemeroptera, Plecoptera, and Trichoptera. Bioindicators include: community resilience index and dry-wet alternation response rate. Static semantic networks include: insufficient ecological flow → leading to prolonged dry season → promoting the proportion of drought-tolerant species; increased drought intensity → leading to low water connectivity → inhibiting the number of benthic animal EPT species; agricultural non-point source pollution → leading to increased ammonia nitrogen concentration → inhibiting the number of benthic animal EPT species; and riparian zone destruction → leading to reduced vegetation cover → inhibiting the proportion of drought-tolerant species.
[0055] The second layer is the dynamic Bayesian causal network layer. Its network structure is initialized by the knowledge graph layer and uses historical data to learn the conditional probability distribution between nodes. It is used to perform probabilistic reasoning based on real-time observation data to identify the most likely stress sources and causal paths.
[0056] The reasoning process of the second layer of the causal inference model, the dynamic Bayesian causal network, includes: inputting real-time observation data as evidence into network nodes, which include stressor nodes, environmental state nodes, IBI index nodes for various biological groups, and system health state nodes. The conditional probability distribution between nodes is trained and dynamically updated using historical monitoring data and machine learning. This layer receives real-time observation data. Based on the Bayesian inference algorithm, the posterior probability of observing the current combination of biological and environmental states under different stressor scenarios is calculated. The diagnostic conclusion of the dominant stress type and its confidence level are output in probabilistic form, thereby identifying the most likely stress type and its influencing path. The specific process is as follows:
[0057] Initializing the network structure based on the knowledge graph: Based on the knowledge graph of the first layer, the confirmed strong causal relationships are transformed into directed edges in the Bayesian network to construct the initial network topology.
[0058] Training Conditional Probability Distributions with Historical Data: Utilizing long-term historical monitoring data, including stressors, environmental factors, and biological indicators, a Bayesian estimation method is employed to learn the Conditional Probability Table (CPT) for each node in the network under the state of its parent node. For example, the probability distribution of the benthic animal EPT index is learned under different combinations of water connectivity (high / medium / low) and ammonia nitrogen concentration (high / medium / low) of the parent node.
[0059] Probabilistic inference based on real-time observation data: When new observation data (feature vectors and environmental data) is input, the model uses the observed values as evidence to input into the corresponding network nodes. It employs precise inference algorithms, such as the connection tree algorithm, or approximate inference algorithms, such as Markov chain Monte Carlo, to calculate the posterior probability distribution of all unobserved nodes, especially those representing unknown pressure sources and system health status. The system selects the combination of pressure sources with the highest posterior probability and its impact path as the most likely diagnostic conclusion. For example, the inference output might be: "Diagnosis: Insufficient ecological flow dominates, confidence level 88%; Path: Insufficient ecological flow → prolonged dry season → increased proportion of drought-tolerant groups → decreased biodiversity."
[0060] The third layer, the interpretability output and scenario simulation layer, is used to generate a visual diagnostic report that includes the contribution of stressors and causal paths, and to provide ecological response prediction simulations based on adjustments to stressor states.
[0061] The interpretability output of the third layer of the causal inference model is based on the reasoning results of the second layer, generating a visualized and interpretable diagnostic report. This report includes a quantitative decomposition of the types and contributions of major stressors and a key stress-biological response causal path diagram. The "What-if" scenario simulation function is specifically implemented as follows: it receives virtual intervention settings from users regarding the state of specific stressor nodes (human intervention in the existing environment), inputs this intervention as new evidence into the dynamic Bayesian causal network layer, recalculates the state probability distribution of downstream biological indicator nodes, and thus predicts the ecological change trend after the implementation of management measures. The specific process is as follows:
[0062] Visualized report generation: The system automatically generates diagnostic reports, using radar charts to display the health scores of various dimensions such as water quality, organisms, and habitats, using directed graphs to clearly show the inferred main causal paths, and using bar charts to show the percentage contribution of each stressor. The contribution is the ratio of the confidence level of a certain driving factor to the sum of the confidence levels of all driving factors.
[0063] Interactive scenario simulation: Users can manually adjust the status of a stress source node in a visual interface. For example, dragging the virtual slider for ammonia nitrogen concentration from high to low will be used as new intervention evidence by the system to rerun Bayesian network inference. This will quickly calculate the probability distribution of improvement in the status of each biological indicator node after assuming ammonia nitrogen is controlled, thereby generating a prediction report: if the ammonia nitrogen concentration is reduced to Class I water standards, the probability of the benthic animal EPT index recovering to a good level within 3-6 months is 75%.
[0064] Step 5, output the simulation regulation prediction results: input the watershed data to be diagnosed into the causal inference model, output the integrated ecological health level of dry and wet periods, key driving factors and their contributions, and generate a visual report containing ecological flow regulation recommendations.
[0065] The simulation and regulation prediction results are output by inputting the watershed data to be diagnosed into the causal inference model. The model outputs the integrated ecological health level, key driving factors, and their contributions during wet and dry periods, generating a visualized report containing ecological flow regulation recommendations. Based on the dedicated IBI values, the ecological health level is divided into five levels: Excellent (0.80, 1.00), Good (0.60, 0.80), Average (0.40, 0.60), Poor (0.20, 0.40), and Very Poor (0, 0.20). Key driving factors are differentiated between natural factors such as hydrological fluctuations and flow interruption duration, and anthropogenic factors such as pollution emissions and land use, with the contribution of each factor quantified. The contribution is the ratio of the confidence level of a driving factor to the sum of the confidence levels of all driving factors. Ecological regulation recommendations include optimization paths such as ecological flow replenishment schemes and sediment pollution control measures during dry periods.
[0066] The final diagnosis and simulation report of the intelligent diagnosis and control system for intermittent river ecosystem degradation provided by this invention is pushed to managers and decision-makers through a web interface or mobile APP, providing direct and quantitative scientific basis for them to formulate precise river and lake management, pollution control or ecological restoration plans.
[0067] The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention.
Claims
1. A method for diagnosing and regulating intermittent river ecosystem degradation, characterized in that, Includes the following steps: Step 1: Based on the intermittent river catchment area and the characteristics of alternating wet and dry periods, sampling points covering different hydrological zones and land use types are set up. Water samples during the wet period and sediment samples during the wet and dry periods are collected simultaneously, along with hydrological parameters, habitat data, and spatiotemporal location information. Step 2 involves processing the collected water samples from wet periods and sediment samples from dry and wet periods with eDNA and performing scenario-based biological information analysis, including eDNA extraction, purification and macrobarcode sequencing, screening for drought-resistant biological groups, calculating the IBI feature parameters specific to at least two different biological groups in the intermittent river, and forming a biological response feature vector. Step 3: Integrate eDNA scenario-based bioinformatics analysis data, hydrological parameters, habitat data, and historical monitoring data to construct a comprehensive dataset. After outlier removal, missing value imputation, and standardization, extract core feature variables. Step four: Input the biological response feature vector along with the hydrological parameters and habitat data into a pre-constructed causal inference model for diagnostic analysis, construct and train the causal inference model; wherein, the causal inference model is a hierarchical Bayesian causal network model that integrates ecological prior knowledge and data-driven approaches, and the causal inference model includes: The first layer, the ecological relationship knowledge graph layer, is a static semantic network constructed based on ecological principles, describing the inherent causal and correlational relationships between stressors, environmental factors, and biological indicators. The second layer is the dynamic Bayesian causal network layer. Its network structure is initialized by the knowledge graph layer and uses historical data to learn the conditional probability distribution between nodes. It is used to perform probabilistic reasoning based on real-time observation data to identify the most likely stress sources and causal paths. The third layer, the interpretability output and scenario simulation layer, is used to generate a visual diagnostic report that includes the contribution of stressors and causal paths, and to provide ecological response prediction simulation based on adjusting the state of stressors. Step 5: Input the watershed data to be diagnosed into the causal inference model, output the integrated ecological health level of dry and wet seasons, key driving factors and their contributions, and generate a visual report containing ecological regulation recommendations.
2. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 1, characterized in that, In step one, 3-5 repeated sampling points are set up in each ecological functional zone, with a sampling point spacing of not less than 500 meters; the water sample collection depth is 0cm-30cm during the wet season and the sediment sample collection depth is 0cm-5cm during the dry season; after the samples are collected, they are transported in the dark at 0℃-4℃ and pre-processed within 24 hours. The hydrological parameters include flow rate, peak flow time within the year, and duration of flow interruption; the habitat parameters include habitat physical characteristics, water body characteristics, and other environmental characteristics; wherein, the habitat physical characteristics include river channel physical characteristics and riparian zone conditions; the water body characteristics include hydraulic characteristics and water body physicochemical parameters; the other environmental characteristics include altitude, temperature, and precipitation, or can be collected in real time through existing in-situ sand and water prevention sensor arrays.
3. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 1, characterized in that, In step two, the eDNA processing and scenario-based bioinformatics analysis involve eDNA extraction using -80℃ freeze-drying and microwave-assisted extraction. The sediment sample extraction parameters are 500W-800W power and 15min-20min time. Macrobarcoding sequencing uses primers designed for barcode regions specific to drought-resistant biological groups. Species annotation is completed based on NCBI, BOLD databases, and the local species list of intermittent rivers. Multi-group biological community data are amplified by multiplex PCR using eDNA macrobarcoding technology and specific primers for different biological groups, thus acquiring biological community data for at least two of the following groups: fish, benthic macroinvertebrates, and attached algae. The specific IBI characteristic parameters include resistance parameters and resilience parameters. The resistance parameter reflects the organism's ability to resist environmental stress during drought, while the resilience parameter reflects the organism's ability to restore its population and community structure after rewetting. The weight ratio of the resistance parameter and resilience parameter for each species is calculated using the entropy method to obtain the specific IBI characteristic parameters for the species.
4. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 3, characterized in that, The ability to resist environmental stress during drought is the ability to survive under conditions of no water or low flow, including the proportion of dormant bodies, which is the percentage of dormant eggs or cysts in the sediment relative to the total number of individuals of the species; the ability to restore population and community structure after rewetting is the speed and extent of recovery from drought stress, including aerial dispersal ability, which includes the percentage of species with flying adults and the drift dispersal rate; the drift dispersal rate is the density of individuals of the species drifting downstream from the upstream perennial flow section after reflow.
5. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 1, characterized in that, In step three, outliers are removed using a box plot method with an IQR of 1.5; missing values are filled using K-nearest neighbor interpolation with k=5; and Z-score standardization is used to eliminate dimensional differences.
6. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 1, characterized in that, In step four, the ecological relationship knowledge graph layer of the causal inference model includes one or more of the relationships between stressors, environmental factors and biological indicators, such as inhibition, promotion and causation, which are used to constrain and initialize the network structure of the dynamic Bayesian causal network layer to prevent causal connections that violate ecological principles. The reasoning process of the dynamic Bayesian causal network layer includes: inputting real-time observation data as evidence into the network nodes, using Bayesian inference algorithms to calculate the posterior probability of each pressure source node and system health status node, and outputting the diagnostic conclusion of the dominant pressure type and its confidence level in probabilistic form. The interpretable output and the interpretable output of the scenario simulation layer are based on the reasoning results of the dynamic Bayesian causal network layer, generating a visualized and interpretable diagnostic report, including a quantitative decomposition of the types of major stressors and their contributions, and a key stress-biological response causal path diagram.
7. The method for diagnosing and regulating intermittent river ecosystem degradation according to claim 1, characterized in that, In step five, the simulation and regulation prediction results are output, and the ecological health level is divided into five levels: excellent, good, average, poor, and very poor based on the exclusive IBI value. The key driving factors are distinguished between natural and human factors, and the contribution of each factor is quantified. The ecological regulation recommendations include ecological flow replenishment schemes and dry season sediment pollution control measures.
8. A system for diagnosing and regulating intermittent river ecosystem degradation, used to implement the method for diagnosing and regulating intermittent river ecosystem degradation as described in any one of claims 1-6, characterized in that, include: An integrated data acquisition terminal is used to simultaneously collect water samples during wet periods and sediment samples during dry and wet periods, and to collect hydrological parameters, habitat image data and spatiotemporal positioning information. The data communication and preprocessing module is used for data transmission, storage, and preprocessing of raw data; The cloud-based intelligent analysis platform is communicatively connected to the integrated data acquisition terminal. The cloud-based intelligent analysis platform includes: a bioinformatics analysis unit for processing eDNA data and calculating multi-group-specific IBI feature vectors; a causal inference model engine for running hierarchical Bayesian causal network models; a knowledge graph management unit for storing and updating intermittent river-specific ecological relationship knowledge graphs; and a scenario simulation and visualization unit for generating diagnostic reports and simulation results.
9. The intermittent river ecosystem degradation diagnosis and control system according to claim 8, characterized in that, The integrated data acquisition terminal integrates an eDNA water sample filtration device, a multi-parameter hydrological data sensor, and an image acquisition module, and is equipped with a data synchronization and positioning module to ensure that all acquired data have a unified time and space reference.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.