Ecological monitoring system and method based on community-specific species and multi-source intelligence fusion

Abstract

This ecological monitoring system and method, based on the integration of community-specific species and multi-source intelligence, includes a species selection unit that optimizes representative species selection using artificial intelligence and ecological models; a monitoring unit that deploys a multi-scale monitoring network combining drones / unmanned vessels with in-situ sensors; a data acquisition and processing unit that integrates environmental DNA (eDNA) analysis, multi-source remote sensing, and real-time data acquisition using the Internet of Things (IoT), along with edge computing preprocessing; a data analysis and visualization unit that uses machine learning and digital twin technology for ecological health assessment, trend prediction, and visualization analysis; and an alarm unit based on artificial intelligence predictive models. This system enables precise, real-time, and intelligent monitoring and early warning of the dynamics of specific species communities, such as phytoplankton, as well as water health and ecosystem evolution trends, in the Yellow River estuary. This significantly improves monitoring efficiency, coverage, and prediction accuracy, providing strong technical support for estuarine ecological protection and management.

Description

Technical Field

[0001] The present invention relates to the field of ecological monitoring technology, and specifically to a system and method for monitoring, evaluating and warning the health status of an ecosystem based on community-specific indicator species. The system and method are particularly suitable for monitoring the ecological environment of complex aquatic ecosystems such as estuaries and wetlands, such as the Yellow River Estuary area. Background Art

[0002] Ecosystems are the foundation of human survival and development, and their health is directly related to regional sustainable development and ecological security. As unique ecosystems where land and sea interact, estuaries possess high levels of biodiversity and ecological services, but they also face the dual pressures of natural evolution and human activities, making their ecological environment extremely sensitive and fragile. The Yellow River Estuary region is an important wetland ecosystem and biodiversity treasure trove in my country. Maintaining its ecological health is of great significance to regional economic development, biological resource conservation, and the ecological protection and high-quality development of the Yellow River Basin. Therefore, effective monitoring of typical estuary ecosystems such as the Yellow River Estuary, accurate assessment of their health, and timely warning of potential risks are important prerequisites for ecological and environmental protection and management.

[0003] Currently, conventional ecological monitoring methods primarily include fixed-point manual sampling combined with laboratory analysis, automated sensor monitoring at fixed sections or stations, and large-scale analysis using satellite remote sensing imagery. While these methods can provide some information about ecosystems, they face numerous limitations when applied to areas like the Yellow River Estuary, characterized by complex hydrological dynamics, variable environmental factors, and rapid ecological processes. For example, traditional manual sampling monitoring is infrequent, lacks spatial representation, and is time-consuming and labor-intensive, making it difficult to capture ecosystem dynamics and sudden events. While fixed-site sensor monitoring can achieve high-frequency data collection, its limited spatial coverage and limited sensor types and parameters make it difficult to fully reflect the overall state of the ecosystem. While conventional remote sensing technologies can provide large-scale observations, their accuracy is limited in areas with complex water optical properties (such as the highly sediment-rich waters of the Yellow River Estuary), and they struggle to directly obtain information about underwater organisms and community structure parameters.

[0004] Especially in the monitoring of aquatic organisms, for example, phytoplankton, as primary producers in aquatic ecosystems and key links in the food web, changes in their community structure and abundance are important indicators of water nutrient status, pollution levels, and ecosystem health. Traditional phytoplankton monitoring relies primarily on manual identification and counting of collected water samples using a microscope. This method is not only time-consuming and labor-intensive, highly subjective, but also requires extremely high operator expertise, making it difficult to meet the needs of large-scale, efficient, and standardized monitoring.

[0005] In recent years, emerging technologies have brought new opportunities for ecological monitoring. For example, environmental DNA (eDNA) technology, by analyzing DNA fragments released by organisms in environmental samples (such as water and soil), provides a new, non-invasive, and highly sensitive approach for monitoring aquatic biodiversity. Unmanned mobile monitoring platforms, such as unmanned aerial vehicles (UAVs) and unmanned watercraft (USVs), equipped with diverse sensors, can rapidly acquire large-scale, high-resolution ecological and environmental data, significantly improving monitoring efficiency and coverage. The Internet of Things (IoT) enables the interconnection of monitoring equipment and real-time data transmission. Artificial intelligence (AI) and machine learning (ML) algorithms, such as convolutional neural networks (CNNs) and long-short-term memory networks (LSTMs), have shown great potential in processing massive amounts of ecological data, identifying complex patterns, and providing predictions and early warnings. Digital twin technology, by constructing virtual models corresponding to physical entities, provides a new framework for the simulation, analysis, prediction, and optimal management of complex systems. Its application in ecosystem management, particularly wetland management, is also gaining attention.

[0006] However, despite the advantages these emerging technologies offer, challenges remain in their practical ecological monitoring applications. eDNA technology still needs improvement in quantification, standardization, and large-scale automated sampling and analysis, and the accuracy of its monitoring results is susceptible to various environmental factors. Effectively integrating and collaboratively analyzing multi-source, heterogeneous data collected by unmanned monitoring platforms remains a technical challenge. Training AI and ML models relies on high-quality, large-scale labeled data, and the models' generalization and interpretability need to be improved. The application of digital twin technology in ecosystems is still in its exploratory stages, and constructing high-fidelity ecological digital twin models presents significant challenges. More importantly, there is currently a lack of an integrated, intelligent ecological monitoring system that can effectively integrate these advanced technologies and address the specific ecological characteristics of the Yellow River Estuary, from intelligent species selection, multi-dimensional data collection, in-depth analysis and prediction, to precise early warning. For example, Nanjing University has applied eDNA technology to biodiversity monitoring in the Yellow River and developed an online enrichment and automated monitoring system, demonstrating its potential for application in specific regions. However, a more comprehensive monitoring and early warning system that integrates multiple advanced technologies remains a pressing technical challenge.

[0007] Therefore, it is of great theoretical significance and practical application value to develop a system and method that can overcome the shortcomings of existing technologies, integrate multi-source data and advanced intelligent analysis methods, and realize accurate, real-time, intelligent monitoring and early warning of the dynamics and overall health status of specific species communities in complex estuarine ecosystems such as the Yellow River Estuary. Summary of the Invention

[0008] (1) Technical problems to be solved by the present invention

[0009] This invention aims to address a series of key technical bottlenecks faced by existing ecological monitoring technologies when applied to complex estuarine ecosystems such as the Yellow River Estuary. Specifically, existing technologies often have the following problems:

[0010] Limitations and lags in the selection of monitoring objects: Traditional methods often rely on experience or fixed lists when selecting indicator species, making it difficult to make rapid and intelligent optimization adjustments based on dynamic changes in ecosystems and real-time environmental feedback, resulting in insufficient targeting and sensitivity of monitoring.

[0011] The one-sidedness and inefficiency of data acquisition: Conventional monitoring methods (such as fixed-point manual sampling and fixed-site sensing) are insufficient in spatial coverage, temporal continuity and data dimensionality, making it difficult to fully capture the dynamic changes and emergencies of complex ecosystems (especially aquatic communities). Emerging technologies such as environmental DNA (eDNA) and unmanned monitoring platforms have potential, but their systematic integration and coordinated application are still immature.

[0012] The shallowness of data analysis and the lack of predictive capabilities: Existing analyses mostly focus on status description and lack the ability to deeply integrate and intelligently analyze multi-source heterogeneous data. This makes it difficult to accurately assess the health status of ecosystems and reveal their evolution patterns, especially in the early prediction and early warning of ecological risks.

[0013] The monitoring system is not highly intelligent and automated: the existing system has limited automation and intelligence in data collection, processing, analysis, and early warning. Monitoring efficiency, response speed, and decision-making support capabilities need to be improved, making it difficult to meet the needs of refined and proactive management of important ecological areas such as the Yellow River Estuary.

[0014] Therefore, there is an urgent need to develop a comprehensive solution that can integrate emerging monitoring technologies and intelligent analysis methods to achieve accurate, real-time, intelligent monitoring, assessment, and predictive early warning of the dynamics of specific species communities and the overall health of the ecosystem.

[0015] (2) Technical solution of the present invention

[0016] To address these technical issues, this paper provides an ecological monitoring system and method based on the integration of community-specific species and multi-source intelligence. By innovatively integrating cutting-edge technologies such as artificial intelligence, environmental DNA (eDNA), multi-platform collaborative observation, edge computing, and digital twins, this system provides an integrated solution encompassing intelligent species selection, multi-dimensional monitoring, efficient data processing, in-depth analysis and prediction, and precise intelligent early warning.

[0017] According to one aspect of the present invention, there is provided an ecological monitoring system based on specific species in a community, characterized by comprising:

[0018] Species Selection Unit: This unit's core function is to intelligently and dynamically optimize monitoring targets. Based on artificial intelligence algorithms (e.g., reinforcement learning models and machine learning classifiers), it integrates historical ecological data with real-time environmental factors (e.g., hydrological, water quality, and meteorological parameters), and the ecological characteristics and indicative value of target area organisms (e.g., phytoplankton in the Yellow River Estuary). It dynamically selects and identifies one or more specific species (e.g., keystone species, sensitive species, and dominant species) that are most representative and indicative during the current monitoring cycle as core monitoring targets. This process enables adaptive adjustments in response to environmental changes and monitoring objectives (e.g., eutrophication assessment and harmful algal bloom warning).

[0019] Monitoring Unit: This unit is responsible for building a multi-level, multi-scale, three-dimensional monitoring network to comprehensively obtain data on specific species and their habitats. It is configured to scientifically deploy monitoring units within the target area (such as the Yellow River Estuary) based on the monitoring targets and their distribution characteristics determined by the species selection unit:

[0020] Fixed monitoring stations: equipped with a variety of in-situ sensors (such as water quality multi-parameter meters, chlorophyll fluorescence probes, spectral sensors, etc.) to conduct high-frequency, continuous monitoring of basic environmental parameters and biomass-related indicators.

[0021] Mobile monitoring platforms include unmanned aerial vehicles (UAVs) equipped with integrated hyperspectral imagers, thermal infrared cameras, and water quality sensors, which are used to quickly and extensively acquire surface water information. They also include unmanned vessels (USVs) equipped with acoustic detection equipment, multi-parameter water quality profilers, and a self-propelled automatic sampling and filtration system for environmental DNA (eDNA). These USVs are used to obtain underwater topography and vertical water profile parameters, as well as to collect eDNA samples in a mobile, multi-point manner.

[0022] eDNA automatic sampling device / monitoring station: can be deployed at key points or hard-to-reach areas to achieve timed or triggered automatic collection, in situ enrichment, lysis and preservation of eDNA water samples, providing high-quality samples for subsequent high-throughput sequencing analysis.

[0023] Data acquisition and processing unit: This unit is designed to achieve efficient, high-quality acquisition and standardized preprocessing of multi-source heterogeneous data. It is configured as follows:

[0024] By using various sensors (optical, hydrochemical, acoustic, hyperspectral, etc.) and eDNA sampling and analysis technology in the monitoring network, data on the population size, abundance, distribution, physiological status, activity habits of specific species, as well as related environmental parameters such as water quality, hydrology, and meteorology are collected in real time or periodically.

[0025] Integrated edge computing nodes (which can be deployed at monitoring sites or mobile platforms) perform real-time quality control (such as denoising and outlier removal), format conversion, preliminary feature extraction (such as chlorophyll concentration estimation, preliminary eDNA abundance calculation, preliminary calculation of species diversity index), data compression and encryption on the collected raw data to reduce data transmission load, improve response speed and ensure data security.

[0026] A multi-sensor data fusion module is configured, and advanced estimation algorithms (such as Kalman filtering and Bayesian networks) are used to perform spatiotemporal alignment, weighted fusion, and consistency verification on data from different sensors and eDNA analysis on the same monitoring object or parameter, thereby improving the overall accuracy, reliability, and spatiotemporal resolution of monitoring parameters.

[0027] Data Analysis and Visualization Unit: This unit is the core for achieving deep insights into ecosystem status and intelligent prediction of future trends. It is configured as follows:

[0028] Comprehensive analysis of processed multi-source, high-dimensional data is conducted using machine learning models (e.g., deep learning networks such as CNN / LSTM, random forests, and gradient boosting trees) and ecological mechanistic models (e.g., improved, dynamically parameterized logistic growth models). This analysis includes assessing the dynamics of species community structure (e.g., changes in composition, abundance, and diversity), ecosystem health (constructing a comprehensive health index), identifying key drivers, and predicting population trends and potential ecological risks (e.g., future changes in the abundance and spatial distribution of specific phytoplankton species and the probability of harmful algal blooms).

[0029] Using dynamic models based on digital twin technology, a virtual representation system for coupled hydrodynamics, water quality, and ecological processes in target waters (e.g., a specific area at the Yellow River Estuary) is constructed. This system integrates real-time monitoring data for model assimilation and calibration, simulating the dynamic responses of specific species communities and the evolution of ecosystem states under varying natural environmental stresses or human intervention. It also supports the simulation, evaluation, and optimization of multiple management scenarios.

[0030] The analysis results are intuitively displayed in the form of multi-dimensional interactive charts, heat maps, three-dimensional scenes, and a four-dimensional (spatial three-dimensional + time) dynamic visualization platform based on WebGIS, making it easier for users to understand and make decisions.

[0031] Alarm unit: This unit is designed to achieve intelligent, graded, and proactive early warning of ecological risks. It is configured as:

[0032] Based on the ecological health assessment status and trend prediction results output by the data analysis and visualization unit (especially the prediction output of the machine learning model and digital twin model), when the monitoring indicators show significant abnormal changes, or when potential ecological risks are predicted in the short or medium term in the future (such as a sharp decline in the number of specific species, a rapid increase in the abundance of indicative pests (such as specific algae) and may exceed the warning threshold, the ecosystem health index continues to deteriorate, etc.), a multi-level intelligent alarm is triggered.

[0033] The alarm threshold is not fixed, but can be adaptively adjusted based on historical data pattern analysis, ecosystem seasonal rhythms, and dynamic assessment results to improve the sensitivity and specificity of the alarm and reduce false alarms and missed alarms.

[0034] Push warning information to preset user terminals (such as monitoring centers and managers' mobile devices). The information includes risk level, predicted occurrence time and spatial range, and major affected species or parameters. It can also provide preliminary response measures based on the preset knowledge base or decision support rules.

[0035] The present invention also provides an ecological monitoring method based on the above system, whose steps correspond to the functional implementation of each unit of the system. Through the collaborative work of each unit, the entire process from intelligent species selection to monitoring implementation, data processing, analysis and prediction, and intelligent alarm is completed.

[0036] (3) Beneficial effects of the present invention

[0037] Compared with existing technologies, this invention, through the above-mentioned systematic technical solutions, particularly through the integration and innovation of key technologies such as dynamic species selection optimized by artificial intelligence, a three-dimensional monitoring network integrating multiple platforms and technologies, efficient multi-source data processing with integrated edge computing, in-depth ecosystem analysis and trend prediction based on artificial intelligence / machine learning and digital twin technologies, and intelligent adaptive alarms based on prediction results, has achieved the following significant beneficial effects:

[0038] Monitoring accuracy and sensitivity have been significantly improved: by integrating highly sensitive eDNA technology, high-precision in situ / remote sensing sensors, and combining them with artificial intelligence-assisted species identification and abundance estimation algorithms, the composition, abundance and subtle changes of specific species (especially trace, cryptic or morphologically difficult to distinguish species) in the monitoring area can be identified earlier and more accurately, significantly improving the ability to capture early ecosystem degradation signals and key biological events.

[0039] The real-time nature and coverage of monitoring have been greatly enhanced: by utilizing the rapid maneuverability and wide-area coverage characteristics of mobile monitoring platforms such as drones and unmanned boats, combined with Internet of Things technology, real-time or near-real-time transmission of data is achieved, and edge computing nodes are deployed at the front end of the monitoring to perform immediate pre-processing and preliminary analysis of the data. This overcomes the spatial limitations of traditional fixed-point monitoring and the bottleneck of poor timeliness of manual inspections, and realizes dynamic, efficient, and nearly panoramic monitoring of large-scale and complex areas such as the Yellow River Estuary.

[0040] Outstanding intelligent analysis and prediction capabilities: Using advanced machine learning algorithms and ecological mechanism models (especially digital twin technology), deep mining, pattern recognition and dynamic simulation of multi-source, high-dimensional ecological and environmental data can not only accurately assess the current ecological status and diagnose ecological problems, but also scientifically predict the future evolution trend of the ecosystem (such as the probability of occurrence and spatiotemporal dynamics of phytoplankton blooms, and the growth and decline of specific populations), providing a forward-looking scientific basis for the active management of the ecosystem and risk prevention.

[0041] Enhanced early warning capabilities and decision support: Intelligent alarm units, based on AI-powered predictive models and digital twin scenario simulations, dynamically adjust alarm thresholds and issue early, precise, and graded warning signals based on complex historical data patterns, ecological evolution patterns, and predicted future risk states. This significantly improves the timeliness and accuracy of early warnings. Furthermore, the system provides preliminary response recommendations and supports simulation assessments of multiple management scenarios, providing strong technical support for rapid response, scientific decision-making, and adaptive management by management departments.

[0042] Enhanced comprehensiveness and systematicity of monitoring: This system can integrate multi-source heterogeneous data from different levels (biological, chemical, physical), different platforms (ground, water surface, air), and different technical means (sensors, eDNA, remote sensing). Through multi-sensor data fusion technology and comprehensive analysis models, it provides a more comprehensive and systematic understanding of the ecosystem structure, function and process, helps to reveal the complex interactions between various elements of the Yellow River Estuary ecosystem, and supports the overall protection and integrated management of the ecosystem.

[0043] The practicality and potential economic benefits are significant: through the automation of monitoring processes and the intelligence of analysis and decision-making, it is expected that while ensuring or even improving the quality and frequency of monitoring, the manpower, material and time costs required for long-term monitoring will be reduced, the excessive reliance on highly specialized manual interpretation will be reduced, and the overall efficiency and economy of ecological monitoring work will be improved. It is particularly suitable for ecological protection areas, important economic waters and key regulatory areas that require long-term and continuous monitoring.

[0044] In summary, the present invention provides an innovative, highly integrated intelligent ecological monitoring system and method, which can significantly enhance the monitoring, assessment, prediction and early warning capabilities of the dynamics and overall health status of specific species communities in complex estuarine ecosystems such as the Yellow River Estuary, and provide key technical support for the effective protection, scientific management and sustainable development of the regional ecological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 4 is an overall system block diagram of an ecological monitoring system based on community-specific species according to an embodiment of the present invention.

[0046] Figure 2 Detailed block diagram of a species selection unit according to an embodiment of the present invention.

[0047] Figure 3 Detailed block diagram of the data acquisition and processing unit according to an embodiment of the present invention.

[0048] Figure 4 Detailed block diagram of the data analysis and visualization unit according to an embodiment of the present invention.

[0049] Illustration: 100, species selection unit; 101, Yellow River Estuary regional analysis module; 102, feasibility assessment module; 103, decision-making module; 200, monitoring unit; 300, data acquisition and processing unit; 301, data acquisition module; 302, data preprocessing module; 303, data storage and management module; 400, data analysis and visualization unit; 401, data analysis module; 402, visualization module; 500, alarm unit. DETAILED DESCRIPTION

[0050] To make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0051] Summary of the core idea of ​​the present invention:

[0052] This invention proposes an ecological monitoring system and method based on the integration of community-specific species and multi-source intelligence. It aims to construct a three-dimensional monitoring network with multi-scale and multi-technology integration through intelligent selection of indicator species, integrate edge computing and multi-sensor data fusion technology to efficiently acquire and process data, and use advanced analysis methods such as artificial intelligence, machine learning and digital twins to deeply explore the status and evolution laws of ecosystems, and realize accurate and intelligent early warning, in order to solve the bottlenecks of existing technologies in the monitoring of complex ecosystems such as the Yellow River Estuary.

[0053] The specific implementation of the present invention is described in detail below through several progressive embodiments.

[0054] Example 1: Yellow River Estuary Phytoplankton Monitoring System Based on Artificial Intelligence-Enhanced Specific Indicator Species Selection and Basic Monitoring Network

[0055] This embodiment aims to improve the pertinence, preliminary prediction capability and monitoring efficiency of phytoplankton monitoring in the Yellow River Estuary by introducing an artificial intelligence-assisted dynamic species selection mechanism and an ecological model with dynamic parameters.

[0056] 1. Construction and operation of species selection unit (100) (see Figure 2 )

[0057] In this embodiment, the species selection unit (100) is dedicated to realizing intelligent and dynamic optimization of monitoring objects. Its specific configuration and operation process are as follows:

[0058] Construction of the Yellow River Estuary Regional Ecological Background Information Database (corresponding to the Yellow River Estuary Regional Analysis Module 101):

[0059] Data Collection: Systematically collect and integrate historical and recent multi-source heterogeneous data related to the Yellow River Estuary region. This data should include at least:

[0060] Hydrological and hydrodynamic data: for example, historical and real-time runoff, sediment transport, estuary velocity, tidal patterns (tidal level, current), and spatiotemporal distribution data of salinity (surface, bottom layer, and vertical profile) of the Yellow River mainstream.

[0061] Water chemical environmental data: for example, water temperature, pH, dissolved oxygen (DO), turbidity, chemical oxygen demand (COD), biochemical oxygen demand (BOD), and the concentrations of key nutrients (such as nitrate-N, nitrite-N, ammonia nitrogen-N, active phosphate-P, active silicate-Si) and their seasonal and annual variations.

[0062] Biota data: Prioritize historical monitoring data on phytoplankton in the Yellow River Estuary, including but not limited to: species lists, identification records of dominant species, common species, and indicator species (e.g., species sensitive to or tolerant of eutrophication or pollution); quantitative data on cell density, biomass (wet weight, dry weight, or carbon content), and chlorophyll a concentration (Chl-a) for each species; and calculated community structure parameters (e.g., diversity index, evenness index, and dominance index). Relevant information on zooplankton, benthic organisms, and fish should also be collected to understand food web structure.

[0063] Meteorological data: For example, regional rainfall, temperature, light intensity and sunshine hours, wind speed and direction, etc. These factors directly or indirectly affect the physical and chemical properties of water bodies and phytoplankton growth.

[0064] Human activity data: for example, the location and discharge characteristics of major sewage outlets, the distribution and scale of coastal aquaculture areas, port shipping information, scheduling records of large-scale water conservancy projects (such as water and sediment diversion), and the scope of land reclamation activities.

[0065] Data standardization and management: The collected multi-source data are pre-processed by cleaning, deduplication, format unification, and spatiotemporal registration to build a structured "Yellow River Estuary Ecological Background Spatiotemporal Database" to provide a high-quality data foundation for subsequent analysis.

[0066] Screening of indicator species candidate sets and multi-dimensional feasibility assessment (corresponding to feasibility assessment module 102):

[0067] Preliminary screening: Based on historical biomonitoring data from the ecological context database and relevant literature, a group of phytoplankton species with potential indicative value in the Yellow River Estuary ecosystem were initially selected as candidates. The selection criteria included: historical occurrence as dominant or subdominant species; sensitivity (significant increase or decrease in abundance) to specific environmental stresses (such as eutrophication, organic pollution, sudden changes in salinity, and specific toxic substances); importance in ecosystem functions (such as primary production and food web support); and widespread distribution and ease of collection and identification.

[0068] Multi-dimensional feasibility assessment: The candidate indicator species initially screened are comprehensively evaluated and scored from the following dimensions:

[0069] Ecological indicators: The species' response to the target monitoring question (e.g., eutrophication, specific pollution conditions, ecosystem health) is sensitive, specific, stable, and predictable. Whether there is a clear dose-response relationship between its abundance changes and environmental factors.

[0070] Spatial and temporal representativeness: Whether the species has sufficient spatial and temporal distribution breadth and abundance in the target monitoring area of ​​the Yellow River Estuary to represent the ecological status of the entire region or a specific habitat, and whether its seasonal growth and decline patterns are clear.

[0071] Feasibility and cost-effectiveness of monitoring technology: Species can be accurately identified and quantified using conventional monitoring methods (e.g., microscopy and chlorophyll fluorescence). Sampling and analysis are cost-effective and time-efficient. Monitoring has the potential to be enhanced through automated and intelligent monitoring technologies (e.g., eDNA, hyperspectral remote sensing, and image recognition, which will be introduced in subsequent examples).

[0072] Historical data support: Existing historical monitoring data are sufficient to support the understanding and model construction of the ecological habits and environmental response patterns of this species.

[0073] Ecosystem importance: The species's role in the food web of the Yellow River Estuary. Dramatic fluctuations in its abundance can have significant impacts on other trophic levels.

[0074] Quantitative evaluation model: The analytic hierarchy process (AHP), fuzzy comprehensive evaluation or expert scoring method can be used to combine the above evaluation dimensions and their weights to quantitatively score each candidate indicator species and form a ranked list of indicator values.

[0075] Dynamic indicator species decision-making based on hybrid intelligent model (corresponding to decision module 103):

[0076] Constructing a "Knowledge Base of the Ecological Characteristics and Indicative Value of Phytoplankton in the Yellow River Estuary": This is a core expert knowledge system, jointly constructed and maintained by senior experts in phytoplankton ecology, estuarine ecology, and environmental monitoring. The knowledge base stores the following information in a structured or ontological format:

[0077] Detailed physiological and ecological characteristics of common and important phytoplankton species in the Yellow River Estuary: such as optimal growth temperature, salinity, and light range; requirements for major nutrients (such as N / P ratio preference, silicon requirement); life history characteristics (such as dormant spore formation); toxicity information (such as toxic algae species).

[0078] Typical response patterns of various species to different environmental stressors (such as high concentrations of ammonia nitrogen, low dissolved oxygen, petroleum pollutants, and heavy metals): for example, tolerance thresholds, inhibitory concentrations, acute / chronic toxic effects, morphological variations, etc.

[0079] Rules with known indicator relationships: For example, “If the density of a specific cyanobacteria (such as Microcystis) is continuously higher than Xcells / L and the water temperature is higher than Y°C, it is highly likely that the water body is at risk of eutrophication and cyanobacterial blooms.”

[0080] Recommendations for preferred indicator species combinations under different monitoring objectives: For example, for the two different objectives of "early eutrophication diagnosis" and "harmful red tide early warning", the knowledge base will recommend different indicator species combinations and their monitoring priorities.

[0081] Machine learning model selection and training:

[0082] Model selection: You can choose mature machine learning classification or regression algorithms such as decision trees (e.g., C4.5, CART), support vector machines (SVM), naive Bayes classifiers, random forests, or gradient boosted trees (GBT). The selection criteria include data volume, feature dimensionality, and model interpretability requirements.

[0083] Feature Engineering: Environmental factors from historical monitoring data (such as water temperature, salinity, N, P, Si concentrations, and light), phytoplankton abundance data, and species indicator value scores from the feasibility assessment module are used as model input features. This may require feature selection and dimensionality reduction (such as PCA).

[0084] Label construction: Label training samples based on confirmed ecological events in historical data (such as an algal bloom, water quality deterioration / improvement over a period of time) or expert-determined ecological health levels (such as "indicator species A is a good indicator species in this scenario" or "indicator species combination B is not suitable for this scenario").

[0085] Model training and validation: Use historical data to train the selected machine learning model, and evaluate the model's performance indicators such as accuracy, recall rate, F1 score, etc. through cross-validation and holdout methods.

[0086] Hybrid intelligent decision-making process:

[0087] Input current information: Input the real-time data of key environmental factors of the Yellow River Estuary (such as water temperature, salinity, and nutrients from fixed monitoring stations), current monitoring objectives (set by the user, such as "routine health assessment" or "specific pollution event tracking"), and the latest feasibility assessment results into the decision-making module.

[0088] Preliminary screening of the knowledge base and rule reasoning: First, use the rules in the knowledge base to perform preliminary reasoning. For example, based on the current water temperature and salinity range, some candidate species that are not suitable for the environment are eliminated; according to the monitoring objectives, the priority of certain specific indicator species is increased.

[0089] Machine learning model prediction and ranking: The current environmental factor data and the characteristics of candidate species that have been preliminarily screened by the knowledge base are input into the trained machine learning model. The model will predict the indicative effectiveness score of each candidate species or species combination in the current scenario.

[0090] Dynamic optimization and output: The inference results of the comprehensive knowledge base and the prediction scores of the machine learning model may also be combined with some dynamic adjustment strategies (for example, avoiding the selection of exactly the same indicator species combination for multiple consecutive monitoring cycles to increase the breadth of monitoring; or when the confidence of the model prediction is not high, tending to select indicator species with more stable performance in history). The final output is the optimal one or a group of specific phytoplankton indicator species in the current monitoring cycle (for example, determining to monitor species A of diatoms, species B of dinoflagellates, and species C of cyanobacteria), and the recommended monitoring frequency and level of concern can be given.

[0091] Feedback and learning: After the system is operational, actual monitoring results (such as the correlation between the actual abundance changes of these selected indicator species and environmental changes and ecological events) will be fed back to the decision-making module as new data for continuous optimization of knowledge base rules and iterative training of machine learning models, thereby achieving adaptive improvement in the ability to select indicator species.

[0092] Through the above steps, the species selection unit (100) can overcome the limitations of traditional reliance on fixed lists or purely empirical selection of indicator species, and can dynamically and intelligently optimize monitoring objects according to real-time environmental conditions and specific monitoring needs, thereby improving the pertinence and effectiveness of subsequent monitoring work.

[0093] 2. Layout and Operation of Monitoring Unit (200)

[0094] The monitoring unit (200) scientifically plans and deploys a basic monitoring network in the Yellow River Estuary area based on the target phytoplankton indicator species determined by the species selection unit (100) and their known ecological habits (such as suitable habitats, aggregation areas, migration patterns, etc.).

[0095] Monitoring site selection:

[0096] Comprehensive consideration is given to the hydrodynamic characteristics of the Yellow River Estuary (such as the mainstream runoff area, the area with significant tidal influence, and the fresh-salt water mixing front area), the distribution of major pollution sources (such as the estuaries of rivers entering the sea and near sewage outlets), typical habitat types (such as shallows, deep troughs, intertidal zones, and adjacent wetland areas), important functional areas (such as aquaculture areas, core areas and buffer zones of nature reserves, and drinking water source protection areas), and key ecological change areas reflected in historical monitoring data.

[0097] Select several representative monitoring sections and arrange multiple vertical stratified sampling points (surface, middle and bottom layers, depending on the water depth) or comprehensive sampling points on each section.

[0098] Several key points are selected as the construction sites for fixed automatic monitoring stations.

[0099] Fixed monitoring site configuration: automatic water quality monitoring buoys or shore base stations are deployed at selected key points.

[0100] Core sensor configuration: Conventional water quality five-parameter sensors: measuring water temperature, salinity (or conductivity), pH, dissolved oxygen (DO), and turbidity. These reflect the basic physical and chemical properties of water and are key parameters that affect phytoplankton growth.

[0101] Fluorescence chlorophyll a (Chl-a) sensor: Chl-a is a good alternative indicator of phytoplankton biomass. The fluorescence sensor can achieve high-frequency (for example, every 15 minutes or once an hour) online automatic monitoring, reflecting the dynamic changes of the overall phytoplankton biomass in real time.

[0102] Optional enhanced configuration: Based on monitoring needs and costs, a blue-green algae fluorescence sensor (for distinguishing blue algae from other algae), a photosynthetically active radiation (PAR) sensor in water, and an online nutrient analyzer (such as an in-situ analysis module for nitrate, ammonia nitrogen, and phosphate) can be added, but this is more expensive and requires high maintenance.

[0103] Power supply and communication: Automatic monitoring stations are usually powered by solar panels combined with batteries, and transmit monitoring data to the data center in real time or near real time through wireless communication modules such as GPRS / 3G / 4G / 5G or NB-IoT.

[0104] Manual assisted monitoring and sample collection:

[0105] Regular water sampling: Manual water sampling is carried out at all selected monitoring sections and points (including those near automatic monitoring stations) at a preset frequency (for example, routine monitoring once a month, or increasing the sampling frequency to once every two weeks or once a week during specific seasons such as the peak of phytoplankton reproduction in spring and autumn, or during special hydrological events such as water and sediment regulation).

[0106] Sampling specifications: Follow national or industry standards for water sample collection, storage, and transportation. Use plexiglass water samplers or sampling equipment of appropriate depth to collect water samples.

[0107] Sample distribution and processing:

[0108] Qualitative and quantitative analysis samples of phytoplankton: Take a certain volume of water sample (such as 1L), add Lugol's solution or formaldehyde to fix it on site, and bring it back to the laboratory for species identification and cell counting through precipitation, concentration, and microscopic examination (using an optical microscope and, if necessary, a scanning electron microscope), calculate the density and biomass of various indicator species.

[0109] Chlorophyll a sample: Take a certain volume of water sample (e.g., 500 mL to 1 L) and filter it on-site using a filter membrane with a specific pore size (e.g., GF / F membrane). Store the filter membrane in the dark and at a low temperature. Bring it back to the laboratory, extract it with acetone or ethanol, and measure the chlorophyll a concentration using a spectrophotometer or fluorometer. This result can be used to calibrate the readings of the online fluorescence sensor.

[0110] Water chemical analysis samples: Water samples are collected for accurate laboratory determination of nutrients (nitrate nitrogen, nitrite nitrogen, ammonia nitrogen, active phosphate, total nitrogen, total phosphorus, silicate, etc.), COD, BOD and other parameters.

[0111] Function: On the one hand, manual monitoring data provides calibration and verification for automatic monitoring data, ensuring the accuracy of online sensor data; on the other hand, it provides more detailed species composition information and accurate water chemical background, which are difficult to completely replace by current automatic sensors. It is also a high-quality data source required for the AI ​​model training of the species selection unit (100) and the construction of the ecological model of the data analysis unit (400).

[0112] By combining the above-mentioned fixed-site automatic monitoring with manual assisted monitoring, the monitoring unit (200) can obtain basic spatiotemporal data on the community dynamics of specific phytoplankton indicator species and their key environmental driving factors at the Yellow River Estuary.

[0113] 3. Construction and operation of data acquisition and processing unit (300) (see Figure 3 )

[0114] The data acquisition and processing unit (300) is responsible for efficiently and standardizedly collecting, integrating, pre-processing and storing various types of data from the monitoring unit (200).

[0115] Data acquisition module (301):

[0116] Automatic Data Interface: Configured to receive real-time or near-real-time data streams from online sensors (such as five water quality parameters and Chl-a) at fixed monitoring sites via wireless networks. Supports multiple standard communication protocols (such as MQTT, CoAP, and HTTP) and data formats (such as JSON, CSV, and XML).

[0117] Manual data entry interface: Provides user-friendly web forms, mobile apps, or batch import tools for monitoring personnel to enter laboratory analysis data (such as phytoplankton species list, cell density, biomass, detailed water chemistry analysis results, and sample collection metadata such as sampling time, GPS location, water depth, sampling method, etc.) into the system in a standardized manner.

[0118] Data aggregation and preliminary verification: Perform preliminary integrity, format and range verification on the received data to ensure that the data meets the preset specifications.

[0119] Data preprocessing module (302):

[0120] Sensor data cleaning and calibration:

[0121] Data format conversion and timestamp alignment: Convert data from different sources and formats into a standard format and ensure that all data has an accurate and unified timestamp (such as UTC time).

[0122] Sensor drift calibration: Based on the results of regular on-site calibration of online sensors or comparison with laboratory reference methods, the sensor raw readings are corrected to compensate for systematic errors caused by factors such as sensor aging and contamination.

[0123] Outlier Detection and Removal / Remediation: Use statistical methods (such as the 3σ principle and boxplots), time series analysis methods (such as sliding window filtering and differencing), or model-based detection methods to identify and address outliers in the data (e.g., sudden changes caused by instrument failures or transmission errors, or values ​​outside the reasonable physical range). Data points identified as outliers can be removed or repaired using interpolation methods (such as linear interpolation and spline interpolation based on nearby valid data).

[0124] Data interpolation: For short-term data loss caused by equipment maintenance, communication interruption, etc., appropriate interpolation algorithms can be used to fill in the gaps to ensure the continuity of the time series.

[0125] Laboratory Data Standardization:

[0126] Unit consistency: Ensure that all measurements of the same parameter use the same unit of measurement.

[0127] Data format verification and logical consistency test: For example, check whether the species list uses standardized taxonomic codes, whether the values ​​of various parameters are within a reasonable ecological range, and whether there are obvious logical contradictions between different parameters.

[0128] Data quality marking: Each piece of preprocessed data or data set is assigned a corresponding quality control code or quality grade mark to indicate the source, processing process, reliability and other information of the data for subsequent analysis.

[0129] Data storage and management module (303):

[0130] Design of ecological spatiotemporal database: Build a relational database (such as PostgreSQL+PostGIS) or time series database (such as InfluxDB, TimescaleDB) specifically for the ecological monitoring data of the Yellow River Estuary. You can also combine it with a NoSQL database (such as MongoDB) to store unstructured or semi-structured metadata.

[0131] Data model: A well-designed database table structure and data model can efficiently store and manage multiple types of data, including time series data (sensor readings, changes in species abundance), spatial geographic information (monitoring site coordinates, section locations, algal bloom impact range), species classification information, sample metadata, analysis results, etc.

[0132] Functional support: The database should support efficient data addition, deletion, modification and query, complex conditional query, spatiotemporal aggregation analysis, version control, data backup and recovery, user authority management and other functions.

[0133] Metadata management: Establish a comprehensive metadata standard and management system to record in detail the source, collection method, processing flow, quality assessment results, responsible person, timestamp and other information of each data set to ensure the traceability and comprehensibility of the data.

[0134] Through the collaborative work of the above modules, the data acquisition and processing unit (300) can provide a high-quality, standardized data foundation for subsequent data analysis and model building.

[0135] 4. Construction and operation of the data analysis and visualization unit (400) (see Figure 4 )

[0136] The data analysis and visualization unit (400) uses the processed data to perform statistical analysis, ecological model construction and prediction, and presents the results in an intuitive manner.

[0137] Data Analysis Module (401):

[0138] Descriptive statistics and basic ecological analysis:

[0139] Calculate the average abundance, maximum / minimum value, coefficient of variation, frequency of occurrence and other statistics of each monitoring station and each indicator phytoplankton population.

[0140] Phytoplankton community diversity indices (such as the Shannon-Wiener index H'), evenness index (Pielou's J), dominance index (Simpson's D or Berger-Parker index), Margalef species richness index, etc. are calculated to quantify community structure characteristics.

[0141] Analyze the spatiotemporal distribution patterns (such as drawing contour maps and trend surface analysis) and seasonal variation patterns of indicator species abundance, chlorophyll a concentration, and key environmental factors (water temperature, salinity, and nutrients).

[0142] Conduct correlation analysis (such as Pearson or Spearman correlation coefficient) to preliminarily explore the relationship between indicator species abundance and environmental factors.

[0143] Construction and application of improved dynamic parameterized logistic growth model:

[0144] Model selection and theoretical basis: Based on the classic logistic growth model. Its basic form is: dN / dt=rN(1-N / K), or its integrated form N(t)=K / (1+ea-rt), where N(t) is the population size (or biomass, density) at time t, K is the carrying capacity of the environment, r is the instantaneous rate of increase, and a is the integral constant, which is related to the initial population size.

[0145] Core Improvement—Parameter Dynamicization: Recognizing that in a complex and changing environment like the Yellow River Estuary, K and r are not fixed constants but are dynamically affected by multiple environmental factors. One of the key innovations of this embodiment is to express K and r as functions of one or more key real-time environmental factors Et=(E1,t, E2,t,...,Em,t) (e.g., water temperature Tt, salinity St, light intensity It, and specific nutrient concentrations such as DINt and PO4t): Kt=fK(Et) rt=fr(Et)

[0146] Functional form and parameter estimation: The specific mathematical form of the functions fK and fr can be selected based on prior ecological knowledge (for example, the effect of temperature on growth rate may conform to the Arrhenius equation or a bell curve, and the effect of nutrients on environmental carrying capacity may conform to Michaelis-Menten dynamics or the limiting factor law). The unknown parameters in these functions (for example, optimal temperature, half-saturation constant, etc.) will be estimated and calibrated using the population dynamics data of indicator species from historical monitoring data ($N(t)$ series) and synchronized environmental factor data (Et series) through statistical fitting methods such as nonlinear regression, least squares, maximum likelihood estimation, or more complex machine learning regression models (such as support vector regression (SVR), Gaussian process regression (GPR), or small neural networks).

[0147] Model Application and Prediction: Once the model parameters are calibrated, real-time or short-term predicted environmental factor data (Et) can be substituted into Kt=fK(Et) and rt=fr(Et) to obtain the dynamically changing environmental carrying capacity and growth rate. These dynamic parameters, combined with the currently known population size N(t), are then used to solve the differential equation dN / dt=rtN(1-N / Kt) through numerical integration (e.g., Euler or Runge-Kutta methods), or its discrete-time form, to predict the changing trend of the dominant phytoplankton indicator population density in the short term (e.g., within the next 1-3 days or a week).

[0148] Multi-species considerations: For multiple indicator species, an independent dynamic parameterized logistic model can be constructed for each species, or, if data permit, competition or promotion between species can be considered (such as an extended form of the Lotka-Volterra model). However, this will significantly increase the complexity and parameterization difficulty of the model. In this embodiment, priority is given to the accurate construction of a single-species model.

[0149] Model Validation and Uncertainty Assessment: Use an independent dataset (data not used in model training and calibration) to validate the constructed prediction model and evaluate its prediction accuracy (e.g., root mean square error (RMSE) and mean absolute error (MAE)). Analyze the sources of uncertainty in the model's prediction results (e.g., parameter uncertainty, input data uncertainty, and model structure uncertainty).

[0150] Visualization module (402):

[0151] Basic chart display:

[0152] Time series graph: Displays the changing trends of various monitoring parameters (such as Chl-a concentration, density of specific indicator species, water temperature, salinity, and nutrient concentration) over time in the form of line graphs, area graphs, etc., and can overlay and display comparisons of multiple parameters or multiple sites.

[0153] Histograms / bar charts: used to compare parameter values ​​(such as average abundance, maximum value) between different monitoring sites, different time points (such as monthly average, annual average), and different species.

[0154] Pie chart / stacked percentage bar chart: shows the relative abundance (cell density percentage or biomass percentage) of different indicator species or major phyla in the phytoplankton community.

[0155] Scatter plot and regression line: intuitively display the correlation between two variables (such as indicator species abundance and a certain environmental factor) and fit a regression line.

[0156] Spatial visualization (GIS-based):

[0157] Site data overlay display: The real-time or historical average data of each monitoring site (such as Chl-a concentration, indicator species density, and water quality parameters) are marked with symbols of different colors and sizes at the corresponding positions on the electronic map of the Yellow River Estuary, intuitively reflecting the spatial distribution characteristics of the parameters.

[0158] Contour map / spatial interpolation map: Use spatial interpolation methods (such as inverse distance weighted IDW and Kriging) to generate continuous parameter spatial distribution maps (such as chlorophyll a concentration distribution map and salinity distribution map) based on discrete station data to help identify high-value areas, low-value areas, and gradient change areas.

[0159] Heat map: used to show the density or intensity distribution of a parameter in space.

[0160] Visualization of model prediction results:

[0161] The trend of population density changes in the short term predicted by the dynamic parameterized logistic model is displayed together with the historical observation data in the form of a time series graph.

[0162] If the model can output spatial predictions (for example, when combined with a simple advection-diffusion model), the predicted spatial distribution of population density can be overlaid on the map in the form of a contour map or color patch map.

[0163] Interactive Dashboard: Design a comprehensive visualization dashboard that integrates the aforementioned charts and maps, allowing users to interactively explore and analyze data by selecting time ranges, monitoring parameters, geographic regions, etc. The dashboard should clearly display key indicators and short-term trends of the current Yellow River Estuary phytoplankton ecosystem.

[0164] Through the in-depth mining of the data analysis module and the intuitive presentation of the visualization module, this unit can transform the original monitoring data into valuable ecological information and preliminary prediction results.

[0165] 5. Construction and operation of alarm unit (500)

[0166] In this embodiment, the alarm unit (500) mainly issues an alarm based on a preset fixed threshold and the preliminary prediction result of the data analysis unit (400).

[0167] Alarm threshold setting:

[0168] Source: National or local water quality standards / ecological environment quality standards: for example, the limit values ​​for chlorophyll a and specific toxic algae cell density in surface water environmental quality standards.

[0169] Statistical analysis of historical monitoring data: For example, setting a statistically significant high-value warning line based on the percentile (such as the 90th or 95th percentile) of historical data on specific indicator species or chlorophyll a concentrations in the Yellow River Estuary area.

[0170] Ecological research results and expert experience: Combined with literature on threshold studies linking specific phytoplankton indicator species to ecosystem health (e.g., eutrophication level, algal bloom risk), and the empirical judgment of local ecological experts.

[0171] Specific protection targets: If the monitoring area involves sensitive aquaculture areas or nature reserves, it may be necessary to set stricter internal control thresholds.

[0172] Grading thresholds: Multiple levels of alarm thresholds can be set according to the degree of risk, such as "attention level", "warning level", and "alarm level", corresponding to different response measures.

[0173] Parameter selection: Alarm parameters mainly include:

[0174] Real-time monitoring of chlorophyll a concentration.

[0175] Laboratory analysis of cell densities of specific dominant phytoplankton species (particularly species known to form harmful algal blooms, such as certain cyanobacteria and dinoflagellates).

[0176] Other key water quality parameters (such as low dissolved oxygen and abnormal pH fluctuations).

[0177] Alarm triggering mechanism:

[0178] Real-time data exceeding threshold alarm: When the chlorophyll a concentration or other online water quality parameters monitored by the automatic monitoring system in real time exceed the preset alarm threshold for a period of time (for example, 3 consecutive readings or 1-hour average), the system automatically triggers the corresponding level of alarm.

[0179] Laboratory data exceeding threshold alarm: When the manually entered laboratory analysis results (such as the cell density of a specific harmful algae species) exceed the preset threshold, an alarm is triggered.

[0180] Early warning based on model prediction:

[0181] When the improved dynamic parameterized logistic model in the data analysis module (401) predicts that certain key indicators (such as total chlorophyll a concentration or density of specific harmful algae species) have a high probability (for example, the probability of the predicted value exceeding the threshold is > P%, where P can be set) of exceeding the warning threshold in the short term (for example, within the next 24 hours, 48 ​​hours, or 72 hours), the system can issue a "predictive warning" or "attention reminder" in advance even if the current monitoring value has not exceeded the standard. This reflects the system's certain foresight.

[0182] Alarm information generation and push:

[0183] Alarm information content: The alarm information should at least include:

[0184] Alarm time and location (monitoring site name / number, GPS coordinates).

[0185] The name of the parameter that triggers the alarm.

[0186] The current value (or predicted value) that exceeds the standard and the corresponding alarm threshold.

[0187] Alarm level.

[0188] (Optional) A brief description of the phenomenon or preliminary cause analysis (e.g., "The abnormal increase in chlorophyll a concentration may be related to the recent persistent high temperatures and nutrient input").

[0189] Alarm method:

[0190] Local alarm: The software interface of the monitoring center will pop up a striking warning window, change the color of related icons, and emit sound or light prompts.

[0191] Remote notification: Alarm information is sent promptly and accurately to preset recipients (such as ecological environment monitoring managers, emergency response teams, and relevant scientific researchers) via short message service (SMS), email, instant messaging tools (such as WeChat, DingTalk), or specially developed mobile applications (Apps).

[0192] Alarm records: All alarm events should be recorded in detail in the system log, including the alarm trigger time, release time, cause, processing process and results, for subsequent query and analysis.

[0193] Alarm response and release:

[0194] The system should support operations for confirming, processing, and resolving alarm events. Upon receiving an alarm, management personnel should promptly verify the situation (for example, checking whether sensors are functioning properly, reviewing relevant historical data, and arranging on-site review) and take appropriate management or emergency measures based on the actual situation.

[0195] When the monitoring indicator returns to below the threshold, or is manually confirmed to be a false alarm / the problem has been resolved, the alarm status can be lifted manually or automatically (under certain conditions).

[0196] Through the above mechanism, the alarm unit (500) can issue a warning to relevant personnel in a timely manner when an abnormality in the ecological environment is detected or a potential risk is predicted, thereby buying time for effective response measures to be taken.

[0197] Features and limitations of Example 1:

[0198] Features: AI-assisted dynamic species selection improves the targeted nature of monitoring; a parameter-based ecological model is employed, providing preliminary short-term forecasting capabilities; and a combination of automated and manual monitoring ensures basic data quality and coverage. Compared to purely traditional monitoring methods, the system offers improved intelligence and early warning capabilities.

[0199] Limitations: The monitoring method is still mainly point-based monitoring, and the continuity and precision of spatial coverage are limited.

[0200] The identification and quantification of species mainly rely on manual microscopy, which is inefficient and requires high professional skills from personnel.

[0201] The complexity and accuracy of the prediction model are relatively limited, mainly relying on a few key environmental factors, and its ability to simulate complex ecological processes is insufficient.

[0202] The alarm mechanism is mainly based on fixed thresholds, and its adaptability and intelligence need to be further improved.

[0203] The real-time nature and dimensionality of data acquisition need to be enhanced, especially as the response to rapidly changing ecological events may not be timely enough.

[0204] The following embodiments will gradually overcome these limitations by introducing more advanced technical means based on this embodiment.

[0205] (The detailed descriptions of the subsequent embodiments 2 and 3 will follow a similar structure, focusing on the core technology upgrades and functional enhancements of the monitoring unit, data acquisition and processing unit, data analysis and visualization unit, and alarm unit relative to the previous embodiment, for example:)

[0206] Example 2 focuses on the following:

[0207] Monitoring unit: large-scale application of eDNA technology (automatic sampling stations, underway sampling), UAV remote sensing (hyperspectral, thermal infrared), unmanned vessel multi-parameter profiling and underway monitoring.

[0208] Data acquisition and processing unit: deployment and application of edge computing nodes, bioinformatics analysis process of eDNA sequencing data, processing and fusion of multi-source remote sensing data, and more complex multi-sensor data fusion algorithms.

[0209] Data Analysis and Visualization Unit: Community structure analysis, health assessment model, and algal bloom prediction model based on machine learning (such as random forest, LSTM, and CNN); and a four-dimensional interactive visualization platform based on WebGIS.

[0210] Alarm unit: Intelligent graded warning and dynamic alarm threshold adjustment based on AI model prediction results.

[0211] Example 3 focuses on the following:

[0212] Data Analysis and Visualization Unit: Construct a digital twin of the phytoplankton ecosystem in the Yellow River Estuary, including a high-precision three-dimensional geographic environment model, a hydrodynamic model, a water quality model, and a complex phytoplankton ecodynamics model (such as an NPZD-type model that considers multifunctional groups and interactions); achieve two-way data connection between physical entities and virtual models, model assimilation and calibration; support ecological process mechanism analysis, multi-scenario future predictions, management decision optimization, and adaptive management.

[0213] Monitoring unit: The operation strategy of the monitoring network can be dynamically optimized and guided by the simulation results and uncertainty analysis of the digital twin model.

[0214] Alarm unit: Early warning and risk warning based on long-term scenario simulation of digital twins.

[0215] Visualization module: An immersive, interactive digital twin visualization interface that supports “what-if” analysis.

[0216] Table 1: Comparison of monitoring capabilities of various embodiments.

[0217]

[0218] The table above clearly illustrates the technical path and capability improvements of this invention, from a monitoring system with preliminary intelligent features (Example 1), to an advanced monitoring system integrating multiple cutting-edge technologies and possessing powerful predictive capabilities (Example 2), and ultimately to a digital twin system capable of deeply simulating complex ecosystems and supporting adaptive management and decision-making (Example 3). This hierarchical implementation not only demonstrates the completeness and advancement of this invention's technical concept, but also provides diverse options for users with varying application needs and investment capabilities.

[0219] It should be understood that the above embodiments are only preferred examples of the present invention and are not intended to limit the scope of protection of the present invention. Any obvious replacement, modification or improvement made by any person skilled in the art based on the technical content disclosed in the present invention without departing from the spirit and essence of the present invention should be included in the scope of protection of the present invention. For example, the specific algorithm selection, hardware model, parameter setting, etc. within each of the above units can be adjusted and optimized according to the actual application scenario and technological development without affecting the implementation of the core technical solution of the present invention.

Claims

1. An ecological monitoring system based on the integration of community-specific species and multi-source intelligence, characterized by: include: Species selection unit (100): configured to dynamically optimize and select one or more most representative specific species in the monitoring target area as monitoring objects based on an artificial intelligence algorithm, combined with historical ecological data, real-time environmental factor data, and ecological characteristic data of indicator species; the species selection unit (100) is specifically configured to perform the following steps to dynamically optimize and select the specific species: (a) receiving real-time environmental factor data of the monitoring target area and the monitoring target set by the user; (b) using a preset ecological characteristics and indicator value knowledge base storing physiological and ecological characteristics and indicator relationships of phytoplankton species, to preliminarily screen and prioritize candidate indicator phytoplankton species based on the real-time environmental factor data and the monitoring objectives; (c) inputting the characteristics of the initially screened candidate indicator phytoplankton species and the real-time environmental factor data into one or more pre-trained machine learning-based classification and regression models to predict and output an indicator effectiveness score for each candidate phytoplankton species or phytoplankton species combination under the current scenario; as well as (d) combining the ranking results of the knowledge base and the indicator effectiveness scores predicted by the machine learning model, determining and outputting the optimal one or a group of specific phytoplankton species as monitoring objects in the current monitoring period; A monitoring unit (200) is configured to deploy a multi-level, multi-scale monitoring network including fixed monitoring stations, mobile monitoring platforms, and eDNA automatic sampling devices within the monitoring target area according to the distribution characteristics and monitoring requirements of the selected specific phytoplankton species; the mobile monitoring platform includes an unmanned aerial vehicle (UAV) integrated with a hyperspectral imager and a water quality sensor, and / or an unmanned boat equipped with an acoustic detection device, a multi-parameter water quality profile analyzer, and an eDNA underway sampling system; A data acquisition and processing unit (300) is configured to collect data on the population size, abundance, distribution, physiological state, activity habits, and water quality, hydrology, and meteorological related environmental parameters of the specific phytoplankton species in real time or periodically through multi-type sensors and eDNA sampling analysis in the monitoring network, and perform preliminary denoising, quality control, feature extraction, and data fusion processing on the raw data using edge computing nodes deployed at monitoring sites or mobile monitoring platforms to obtain processed multi-source data; A data analysis and visualization unit (400) is configured to use a machine learning model and an ecological model to comprehensively analyze the processed multi-source data to evaluate the dynamics of phytoplankton species community structure, ecosystem health status, and predict population evolution trends, and to visualize the analysis results in the form of multi-dimensional interactive charts, heat maps, or three-dimensional scenes; the machine learning model includes a CNN or LSTM model for predicting the abundance and spatial distribution changes of a specific phytoplankton species in a future period; the ecological model includes a dynamic model based on digital twin technology for constructing a virtual representation system of a specific hydrodynamic-water quality-ecological process coupling in the monitoring target area; Alarm unit (500): configured to trigger a multi-level intelligent alarm based on the prediction results and ecological health assessment status of the data analysis and visualization unit, when a significant abnormal change occurs in the monitoring indicators or a potential ecological risk is predicted, and push warning information and response suggestions to a preset user terminal; the potential ecological risk includes a sharp decline in the number of specific phytoplankton species or the risk of harmful algal blooms, and the threshold of the intelligent alarm is a dynamic threshold that is adaptively adjusted based on historical data pattern analysis and seasonal rhythm characteristics of the ecosystem.

2. The ecological monitoring system based on community-specific species and multi-source intelligence fusion according to claim 1 is characterized by: The artificial intelligence algorithm used by the species selection unit (100) further includes a dynamic indicator phytoplankton species screening model based on reinforcement learning, which can dynamically adjust the weight and type of phytoplankton species selection according to real-time environmental feedback and monitoring targets.

3. The ecological monitoring system based on community-specific species and multi-source intelligence fusion according to claim 1 is characterized by: The eDNA automatic sampling device in the monitoring unit (200) is configured to collect water samples at regular intervals or triggered by specific environmental events, and automatically complete the enrichment, lysis and preservation of eDNA on site to form eDNA samples; the eDNA samples are subsequently combined with high-throughput sequencing technology to perform qualitative and quantitative analysis of target phytoplankton species and community structure analysis, and the high-throughput sequencing technology is macrobarcoding technology or metagenomics technology.

4. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: The mobile monitoring platform in the monitoring unit (200) includes: an unmanned aerial vehicle equipped with a hyperspectral imager and a water quality sensor, which is used to quickly obtain distribution images of specific phytoplankton species and surface water quality parameters in the monitored water area over a large area; and / or an unmanned boat equipped with an acoustic detection device, a multi-parameter water quality profile analyzer, and an eDNA underway sampling system, which is used to obtain underwater topographic information, vertical profile parameters of the water body, and to perform mobile eDNA sample collection.

5. The ecological monitoring system based on community-specific species and multi-source intelligence fusion according to claim 1 is characterized by: The data acquisition and processing unit (300) is further configured with a multi-sensor data fusion module, which uses an estimation algorithm based on Kalman filtering or Bayesian network to perform spatiotemporal alignment, weighted fusion and consistency verification on data on the same monitoring object or environmental parameters obtained from different types of sensors and eDNA analysis, so as to improve the overall accuracy, reliability and spatiotemporal resolution of the monitoring parameters.

6. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: The edge computing node in the data acquisition and processing unit (300) is deployed at a monitoring site or a mobile monitoring platform and is configured to perform real-time quality control, outlier removal, physical unit conversion, basic statistical parameter calculation, and preliminary calculation of specific ecological characteristic parameters on the collected raw sensor data. The specific ecological characteristic parameters include chlorophyll concentration estimated based on spectral data, biological activity frequency identified based on acoustic data, and phytoplankton species diversity index calculated based on eDNA sequence data. The calculation results and the raw data to be stored for a long time are compressed and encrypted to reduce the data transmission bandwidth demand and cloud computing load, and to ensure data security.

7. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: The machine learning model used by the data analysis and visualization unit (400) includes a CNN or LSTM model for predicting changes in abundance and spatial distribution of a specific phytoplankton population in a future period, wherein the specific phytoplankton population includes at least one of diatoms and dinoflagellates, and the future period is 1 to 7 days; the model is trained and predicted using historical biological monitoring data, multi-source environmental factor time series data, and remote sensing image data as input.

8. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: The data analysis and visualization unit (400) uses a dynamic model based on digital twin technology to construct a virtual representation system for the coupling of hydrodynamics, water quality and ecological processes in the monitoring target area. The system can integrate real-time monitoring data for model assimilation and calibration, simulate the dynamic response of specific phytoplankton species communities and the evolution trajectory of ecosystem status under different natural environmental stresses or human activity intervention conditions, and support simulation evaluation and optimization of multiple management scenarios.

9. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: The intelligent alarm threshold of the alarm unit (500) is not fixed, but is adaptively adjusted based on historical data pattern analysis, seasonal rhythm characteristics of the ecosystem, and dynamic evaluation results of the ecosystem health evaluation model output by the data analysis and visualization unit, so as to improve the sensitivity and specificity of the alarm and effectively reduce false alarms and missed alarms.

10. The ecological monitoring system based on community-specific species and multi-source intelligence integration according to claim 1 is characterized by: To realize the system function, the following steps are included: Step 1: using the species selection unit (100), based on an artificial intelligence algorithm, combining historical ecological data, real-time environmental factor data, and ecological characteristic data of indicator phytoplankton species, dynamically optimizing and selecting one or more most representative specific phytoplankton species in the monitoring target area as monitoring objects; Step 2: using the monitoring unit (200), within the monitoring target area, deploying a multi-level, multi-scale monitoring network according to the distribution characteristics and monitoring requirements of the selected specific phytoplankton species, and starting monitoring; Step 3: using the data acquisition and processing unit (300), through the multi-type sensors and eDNA sampling analysis in the monitoring network, to collect and process the relevant data and environmental parameter data of the specific phytoplankton species in real time or periodically, and using the edge computing node to perform preliminary denoising, quality control, feature extraction and data fusion processing on the raw data; Step 4: using the data analysis and visualization unit (400), using machine learning models and ecological models to comprehensively analyze the processed multi-source data, evaluate the dynamics of phytoplankton species community structure and ecosystem health, predict population evolution trends, and visualize the analysis results; Step 5: Using the alarm unit (500), based on the prediction results and the ecological health assessment status, when a significant abnormal change occurs in the monitoring indicators or a potential ecological risk is predicted, a multi-level intelligent alarm is triggered, and early warning information and response suggestions are pushed to a preset user terminal.

Images