An early warning method for Ulva prolifera green tide disaster based on environmental DNA

Abstract

This invention discloses an early warning method for *Ulva prolifera* green tide disasters based on environmental DNA, belonging to the field of marine ecological disaster monitoring and early warning technology. Monitoring sections and fixed stations are deployed in areas prone to *Ulva prolifera* green tides to periodically collect surface water samples to obtain initial *eDNA* samples, i.e., initial environmental DNA samples. The abundance of *eDNA* for *Ulva prolifera* green tide outbreak genotypes in the water is quantitatively detected, and the rate of change in *eDNA* abundance is calculated. Simultaneous environmental parameters are collected and combined with satellite remote sensing data to construct a joint database of environmental driving factors and green tide biomass. Time series analysis is used to determine the time lead of *eDNA* signals to green tide outbreaks, outputting a green tide risk index. The green tide diffusion path is dynamically simulated to identify risk propagation channels, and early warning information is issued to guide targeted defense. Compared to the lag in traditional remote sensing and on-site observation, which only issue warnings after the algal bloom is visible, *eDNA* signals can be detected before *Ulva prolifera* biomass accumulates on a large scale, achieving early warning during the initial or latent period before a green tide outbreak.

Description

Technical Field

[0001] This invention relates to the field of marine ecological disaster monitoring and early warning technology, specifically to an early warning method for seaweed green tide disasters based on environmental DNA. Background Technology

[0002] The green tide of seaweed in the Yellow Sea is one of the largest and longest-lasting marine algal disasters in the world, seriously affecting the coastal ecological environment, tourism, and marine ranching safety. Current green tide warnings mainly rely on satellite, UAV remote sensing, and on-site observation. Although remote sensing monitoring has high spatial resolution, it has low sensitivity for detecting trace algae in the early stages and can often only be identified after the algal biomass has increased significantly. On-site sampling has high accuracy, but its spatial coverage is limited, making it difficult to achieve large-scale continuous monitoring.

[0003] Environmental DNA (eDNA) technology detects cellular debris and free nucleic acids released by organisms in seawater, reflecting the state of target organisms. Changes in eDNA concentration typically precede changes in visible biomass, exhibiting a significant time lead. Although this technology has been applied in invasive species monitoring and fish community research, there is currently no eDNA-based early warning system for *Ulva prolifera* green tide disasters. Furthermore, in practical applications, it is difficult to dynamically identify and prioritize monitoring key propagation channels along potential green tide spread paths, leading to uneven allocation of early warning resources and hindering precise early control in sensitive areas such as the source areas of *Ulva prolifera* green tides, waterways, and aquaculture zones. To address the problems of early identification difficulties, delayed response, and insufficient spatial coverage resolution in existing green tide monitoring, this patent proposes an early warning method for *Ulva prolifera* green tide disasters based on environmental DNA. By monitoring dynamic changes in genetic signals, it achieves early quantitative identification and graded early warning of green tide risks. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an early warning method for *Ulva prolifera* green tide disaster based on environmental DNA, comprising the following steps: S1. Set up monitoring sections and fixed stations in areas prone to green tides of Ulva prolifera, and collect surface water samples regularly to obtain initial eDNA samples, i.e., initial environmental DNA samples. S2. Process the initial eDNA sample and select primer sequences specific to the genotype of Ulva prolifera green tide outbreak. Quantitatively detect the DNA copy number concentration of the Ulva prolifera green tide outbreak genotype in the water body to calculate the eDNA abundance change rate. S3. Collect environmental parameters from the same period and combine them with satellite remote sensing data to construct a joint database of environmental driving factors and green tide biomass; S4. Standardize the eDNA copy number change rate, environmental driving factors and green tide biomass, establish a multi-factor response model, and use time series analysis to determine the time lead of eDNA signal to green tide outbreak, output green tide risk index, and quantify green tide outbreak risk. S5. Based on historical data, set grading thresholds for eDNA copy number concentration, establish grading early warning standards, and combine them with the green tide risk index. Introduce graph neural networks to dynamically extrapolate the green tide diffusion path and identify risk transmission channels. S6. Using spatial interpolation and GIS visualization technology, generate a spatial distribution map of green tide risk, and at the same time release early warning information to guide targeted deployment.

[0005] Preferably, S1 specifically includes: In the target sea areas where green tides of seaweed are frequent, including the Yellow Sea region, monitoring sections and fixed stations covering different gradient areas of nearshore, coastal and offshore areas will be set up to achieve full spatial coverage and improve the representativeness of the monitoring network through gradient deployment. At the monitoring section and fixed station, surface seawater samples of 2 to 5 liters are collected regularly at a depth of 0 to 2 meters every 3 to 5 days. High-frequency sampling captures early signals and enhances the continuity of the time series. The sampling process continuously covers the entire cycle of the Ulva prolifera green tide, including before, during, and after the outbreak, in order to obtain initial eDNA samples that are spatially and temporally representative. Full-cycle sampling ensures data integrity and supports dynamic risk assessment.

[0006] Preferably, in step S2, the process of quantitatively detecting the eDNA copy number concentration in the water body includes: Seawater samples were filtered through a 0.45-micron pore size membrane to enrich biomass, and total eDNA was extracted from the membrane using a commercial DNA extraction kit. We selected specific gene fragments from the genotypes of green tide outbreaks of Ulva prolifera, designed and synthesized specific primer sequences for quantitative molecular detection; Quantitative PCR or digital PCR technology was used to amplify and detect the extracted eDNA using the specific primers, and the copy number concentration of the eDNA of the Ulva prolifera green tide outbreak genotype in the water was determined, so as to achieve absolute quantification and high-sensitivity monitoring of Ulva prolifera biomass in the water.

[0007] Preferably, in step S2, the process of calculating the rate of change in eDNA abundance includes: Based on the eDNA copy number concentration data measured at different sites and at different time series, the temporal change rate of eDNA abundance was calculated, i.e., the eDNA abundance change rate. The eDNA abundance change rate is the difference in eDNA copy number concentration per unit time, which can effectively quantify instantaneous proliferation potential, compensate for the lag of absolute concentration, and serve as a dynamic biological indicator for coupling modeling and early warning analysis, thereby enhancing the model's sensitivity to early risk identification.

[0008] Preferably, in step S3, the process of establishing a joint database of environmental driving factors and green tide biomass includes: During the collection of eDNA samples, multiple environmental parameters, including water temperature, salinity, nutrient concentration, dissolved oxygen, flow rate, wind speed, and chlorophyll a concentration, were measured simultaneously. By combining multi-source satellite remote sensing data, relevant information on the green tide, such as the coverage area and distribution area of ​​the green tide of Ulva prolifera in the target sea area, is extracted as a macroscopic characterization of the green tide biomass. By integrating the environmental parameters with the green tide biomass information, a spatiotemporally synchronized joint database of environmental driving factors and green tide biomass is constructed.

[0009] Preferably, in step S4, the process of determining the time lead of the eDNA signal to the green tide using time series analysis includes: The data on eDNA abundance change rate, environmental driving factors, and green tide biomass were standardized to eliminate the influence of dimensions and improve data comparability and model convergence efficiency. Based on the standardized data, at least one of the generalized additive model, boosting regression tree, or XGBoost algorithm was selected to establish a multi-factor response model for green tide biomass jointly driven by eDNA and environmental factors, thereby enhancing the fitting ability to complex ecological relationships. Time series analysis was performed on the multi-factor response model to determine whether there was a time lead of the eDNA signal relative to the macroscopic green tide outbreak and its specific length, so as to provide statistical basis for early warning and realize the advanced quantitative identification of green tide outbreak risk.

[0010] Preferably, in step S4, the process of establishing a green tide risk prediction model jointly driven by eDNA and environmental driving factors includes: The time series analysis specifically includes Granger causality test and cross-correlation analysis, which are used to quantify the lead-lag relationship between eDNA signal and green tide outbreak, statistically verify the leading role of eDNA signal, and lay the foundation for early warning. Based on the multi-factor response model and the determined time lead period, a prediction equation is constructed to calculate the green tide risk index, so as to quantify the risk of green tide outbreak in the future period, construct a closed-loop early warning system, and realize the connection from monitoring to decision-making.

[0011] Preferably, in step S5, the process of establishing tiered early warning standards includes: Based on the correlation between eDNA copy number concentration and actual outbreak intensity of Ulva prolifera in historical data, different levels of eDNA copy number concentration thresholds are set to achieve scientific threshold setting based on long-term statistics and improve the reliability of early warning. K-means clustering and / or ROC analysis were used to determine the optimal cutoff points for risk level classification, forming a hierarchical early warning standard that includes safety, concern, warning, and alert. This achieves objectivity and data-driven risk level classification, enhancing the accuracy of early warnings. Safety is classified as Level I, characterized by an algae-free period and an eDNA copy number concentration threshold of <10. 3 (copies / L); the concern level is II, its status is in the nascent stage, and the eDNA copy number concentration threshold is 10. 3 -10 5 (copies / L); the warning level is III, indicating a latent period, with an eDNA copy number concentration threshold of 10. 5 -10 6 (copies / L); Alert level IV, indicating an outbreak, with an eDNA copy number concentration threshold of >10. 6 (copies / L) enables full coverage from algae-free to algae-blooming processes, allowing for more precise tiered responses; When the real-time monitored eDNA copy number concentration crosses a certain threshold in the graded early warning standard, the corresponding risk level early warning information is automatically triggered and released, realizing the linkage between real-time monitoring and automatic early warning, and improving response timeliness.

[0012] Preferably, in step S5, the process of identifying risk propagation channels includes: Based on the aforementioned graded early warning standards, the output green tide risk index is combined with a graph neural network model to achieve a leap from single-point early warning to networked dynamic inference, thereby improving the accuracy of spatial early warning. The graph neural network model uses the green tide risk index as node features and the diffusion dynamics conditions of ocean current field and wind direction as edge weights to construct a graph structure, quantifying the ecological connections and dynamic driving forces between nodes and enhancing the realism of diffusion path simulation. The graph neural network is used to dynamically extrapolate the potential spread paths of Ulva prolifera green tides, analyze the intensity of key paths and the risk aggregation degree of node communities, identify key risk transmission channels and high-risk areas, clarify key areas and paths for prevention and control, and improve the efficiency of emergency resource allocation.

[0013] Preferably, in step S6, the process of issuing early warning information to guide targeted deployment includes: By using inverse distance weighted interpolation or kriging interpolation methods, spatial interpolation is performed on the monitoring and prediction data of discrete stations to achieve seamless coverage of monitoring data from point to area, improve spatial perception capabilities, generate continuous green tide risk spatial distribution maps, form high-precision risk base maps, and support overall situation assessment. In the GIS platform, the spatial distribution map of the green tide risk is fused and overlaid with satellite remote sensing images, and the risk propagation channels are visualized and identified, thereby enhancing the intuitive identification of risk information and the efficiency of spatial positioning. Daily early warning products containing the aforementioned green tide risk spatial distribution map and risk propagation channel information are released through a web interface or decision support system to ensure the real-time nature of early warning information and business continuity, so as to guide targeted prevention and control deployment, achieve hierarchical and precise push, and improve the efficiency of emergency response coordination.

[0014] This invention provides an early warning method for *Ulva prolifera* green tide disasters based on environmental DNA. It has the following beneficial effects: (I) This method for early warning of green tide disasters caused by Ulva prolifera based on environmental DNA involves systematically deploying monitoring stations to regularly collect water eDNA samples and combining them with highly sensitive qPCR / dPCR technology to quantitatively detect the eDNA copy number concentration of Ulva prolifera. Compared with the lag in traditional remote sensing and visual observation, which only alerts after a large amount of Ulva prolifera is visible floating on the sea surface, the eDNA signal can be detected before the biomass of Ulva prolifera accumulates on a large scale. By calculating the rate of change in eDNA abundance, the dynamic trend of population proliferation can be further captured, thereby achieving early warning in the initial or latent period before the outbreak of green tide of Ulva prolifera, thus buying time for prevention and control response.

[0015] (II) This method for early warning of green tide disasters of Ulva prolifera based on environmental DNA constructs a multi-factor response model, integrates eDNA data, real-time environmental parameters and satellite remote sensing information to comprehensively assess the risk of green tide outbreaks, and uses Granger causality test and cross-correlation analysis to statistically confirm the leading role of eDNA signals and quantify their leading period, effectively eliminating false alarms caused by fluctuations of single environmental factors. Based on K-means clustering and ROC analysis of historical data, a graded threshold is set, so that the warning standard has both ecological significance and statistical significance, significantly improving the accuracy and reliability of the warning system.

[0016] (III) This method for early warning of green tide disasters of Ulva prolifera based on environmental DNA introduces graph neural network to construct a spatial propagation model of green tide. The monitoring station is regarded as a node and the ocean current field and wind field are used as boundaries to simulate the transport path of algal clusters under dynamic conditions. Through iterative deduction of message transmission mechanism, the propagation channels with continuous high risk and easy accumulation areas are identified, realizing the leap from point-based early warning to path prediction. This enables the prevention and control measures to be deployed in advance in a targeted manner around key transport channels or sensitive targets, thereby improving the efficiency of resource allocation. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the workflow of an early warning method for Ulva prolifera green tide disaster based on environmental DNA according to the present invention. Figure 2 This is a schematic diagram of the method flow for an early warning method for Ulva prolifera green tide disaster based on environmental DNA according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: an early warning method for *Ulva prolifera* green tide disaster based on environmental DNA, comprising the following steps: S1. Establish monitoring sections and fixed stations in areas prone to Ulva prolifera to regularly collect surface water samples to obtain initial eDNA samples, i.e., initial environmental DNA samples. The monitoring sections and fixed stations include different gradient areas near the coast, along the coast, and offshore. In the target Ulva prolifera green tide areas, including the Yellow Sea, monitoring sections and fixed stations covering different gradient areas near the coast, along the coast, and offshore are established. The gradient deployment achieves full spatial coverage and enhances the representativeness of the monitoring network. At the monitoring sections and fixed stations, 2 to 5 liters of surface seawater samples are collected regularly at a depth of 0 to 2 meters every 3 to 5 days. High-frequency sampling captures early signals and enhances the continuity of the time series. The sampling work continuously covers the entire cycle before, during, and after the Ulva prolifera green tide outbreak to obtain initial eDNA samples with time series representativeness. Full-cycle sampling ensures data integrity and supports dynamic risk assessment. The specific work involves: establishing a spatially gradient monitoring network in the Yellow Sea and target areas prone to Ulva prolifera green tides. This network consists of multiple monitoring sections and fixed stations covering nearshore, coastal, and offshore areas. Nearshore areas will focus on densely populated areas such as estuaries, aquaculture zones, sewage outlets, and ports. Coastal areas will select the intertidal zone to shallow water transition zone (5-20m depth) to cover key areas of green tide accumulation and decomposition. Offshore areas will extend to the endpoint of the green tide drift path (e.g., the central Yellow Sea) to cover the diffusion front. The design of monitoring sections must incorporate historical outbreak trajectories, prioritizing coverage of high-incidence zones and key diffusion nodes. Fixed stations must possess long-term stable monitoring capabilities. Qualitative sampling is conducted using corrosion-resistant sampling equipment and automatic recorders to ensure data continuity. Station density is dynamically adjusted based on gradient differences: one station is deployed every 5-10 kilometers in nearshore areas, one every 15-20 kilometers along the coast, and one every 30-50 kilometers in offshore areas, forming a three-dimensional monitoring network from dense to sparse. Sampling work must adhere to both temporal and spatial standards. Temporally, regular sampling is conducted every 3-5 days, increasing to once every 2-3 days during outbreaks, and returning to the original frequency during recessions. Spatially, surface seawater is collected uniformly at depths of 0-2 meters, using standard water samplers (Niskin bottles) to ensure depth accuracy. The sampling volume per session should be controlled between 2 and 5 liters, balancing sample representativeness (covering the planktonic community) and operational feasibility (avoiding transportation losses). Sampling containers must be sterilized beforehand to prevent exogenous DNA contamination. Samples should be stored at low temperatures (-20°C to 4°C) immediately after collection and filtered within 24 hours. In case of severe weather or equipment failure, the reason for interruption must be recorded and resampling performed to ensure the integrity of the time series. No resampling is required for interruptions ≤12 hours; resampling should be performed once every 12-24 hours; for interruptions >24 hours, samples for the entire period must be collected again. Sampling should be suspended during severe weather (wind speed >10 m / s or wave height >2 m), and the weather code should be recorded. Sampling cycle... The study comprehensively covers the three stages of the Ulva prolifera green tide life cycle: pre-outbreak, mid-outbreak, and post-outbreak. The pre-outbreak stage focuses on capturing early eDNA signals; the mid-outbreak stage monitors peak concentrations and diffusion dynamics; and the post-outbreak stage assesses residual impacts and ecological restoration. Sampling must be continuous for at least two months at each stage to generate continuous time-series data. A standardized sample management process is established: before filtration, the sampling time, location, water depth, and environmental parameters are recorded; after filtration, the filter membrane is aliquoted into sterile centrifuge tubes, labeled with a unique number, and entered into the database; remaining water samples are used for auxiliary analysis. All samples must be simultaneously stored in an ultra-low temperature freezer at -80℃ for long-term archiving for future verification or supplementary research. S2. Initial eDNA samples were processed, and specific primer sequences for the Ulva prolifera green tide outbreak genotype were selected. The eDNA copy number concentration in the water was quantitatively detected to calculate the eDNA abundance change rate. Collected seawater samples were filtered through a 0.45-micron pore size filter membrane to enrich biomass. Total eDNA was extracted from the filter membrane using a commercial DNA extraction kit, efficiently enriching and purifying the target genetic material. Specific primer sequences for the Ulva prolifera green tide outbreak genotype were selected and synthesized for quantitative molecular detection, ensuring highly specific results and effectively avoiding interference from non-target species. Real-time quantitative PCR technology was used to amplify and detect the extracted eDNA using specific primers, accurately determining the abundance and copy number concentration of the Ulva prolifera green tide outbreak genotype eDNA in the water, achieving absolute quantification and high-sensitivity monitoring of Ulva biomass in the water. The specific work involves: immediately passing surface seawater samples through a 0.45-micron pore size filter membrane for vacuum filtration to remove plankton, cell debris, and free nucleic acid biomass. The filter membrane material is either polycarbonate or glass fiber. Polycarbonate membranes have uniform pore size (CV≤5%), making them suitable for low biomass samples, while glass fiber membranes have a high loading capacity (filtering ≥500mL of seawater), making them suitable for high suspended solids samples. Their low DNA adsorption characteristics reduce target molecule loss. The filtration process requires controlled operating conditions to avoid exogenous DNA contamination and is completed in a laminar flow hood. Sterile forceps are used to handle the filter membrane. After filtration, the filter membrane is folded and placed in a sterile centrifuge tube. A commercially available DNA extraction kit (DNeasy PowerWater) is then used. The kit extracts total eDNA using a cell lysis, protein removal, and nucleic acid purification process, efficiently recovering DNA from the filter membrane to obtain a high-purity, low-inhibitory total eDNA solution. For specific gene fragments in *Ulva prolifera*, bioinformatics analysis is used to screen for conserved and unique nucleic acid sequence regions, including ribosomal RNA genes (ITS or 5S). The primers were selected and synthesized based on specific gene fragments of *Ulva prolifera*, including rRNA and chloroplast genes (rbcL or tufA). Primer design ensured a perfect match with the target sequence while avoiding binding to homologous regions of non-target species (other green algae or marine microorganisms). Primer design tools (Primer5) were used to optimize primer length (18-25 bp), GC content (40%-60%), and annealing temperature (55-65℃) to guarantee amplification efficiency and specificity. After primer synthesis, mass spectrometry was used to verify primer sequence accuracy, and gradient PCR was used to test the annealing temperature range to determine the optimal reaction conditions. Quantitative PCR (qPCR) or digital PCR (dPCR) techniques were employed to amplify and detect the extracted eDNA using specific primers. qPCR used a fluorescent dye (SYBR Green) to amplify and detect the eDNA. Green (a method) monitors the accumulation of amplified products in real time and calculates the initial copy number concentration of the target eDNA using the standard curve method; dPCR achieves absolute quantification by dividing the reaction system into microreaction units (droplets) and directly counting the proportion of positive units without relying on standards. It should be noted that both techniques require the setting of negative controls (template-free controls) and positive controls (standards of known concentrations) to exclude contamination and false positives. The final results are expressed as copies / L of seawater. Combined with the conversion of sampling volume and filtered water volume, it accurately reflects the biomass distribution and dynamic changes of Ulva prolifera in the water. The expression for calculating eDNA copy number concentration is as follows: ; ; In the formula: The copy number concentration of *Ulva prolifera*-derived eDNA in the water body represents the number of *Ulva prolifera*-specific eDNA molecules contained in each liter of seawater. The original copy number is determined by qPCR or dPCR. In qPCR, it is obtained by comparing the cycle threshold of the sample with a standard curve; in dPCR, it is calculated by counting positive droplets. The proportion of positive droplets. The total volume of the reaction is [volume]. The volume of a single droplet; The dilution or concentration factor is a factor used to dilute or concentrate the DNA template solution after DNA extraction and before PCR detection. For example, if the extracted DNA stock solution is diluted 10-fold before PCR detection, then... If the undiluted solution is used directly, then ; The filtered seawater volume is the actual volume of the original seawater sample that passes through the filter membrane during eDNA enrichment.

[0020] Furthermore, the process of calculating the eDNA abundance change rate includes: calculating the time-series change rate of eDNA abundance based on eDNA copy number concentration data measured at different sites and at different time series, i.e., the eDNA abundance change rate. This can capture the dynamic change trend of Ulva prolifera populations earlier and improve the foresight of early warning. The calculation of the eDNA abundance change rate is achieved by calculating the difference in eDNA copy number concentration per unit time, which effectively quantifies the instantaneous proliferation potential and makes up for the lag of absolute concentration. In this way, the eDNA abundance change rate can be used as a dynamic biological indicator for coupling modeling and early warning analysis, thereby enhancing the model's sensitivity to early risk identification. The specific work involves: Data cleaning and standardization based on multi-site, long-term eDNA copy number concentration data. Data cleaning includes removing outliers significantly deviating from the mean (more than 3 standard deviations) and filling missing values ​​using linear interpolation. Standardization involves unifying the concentration unit to "copies / L," converting the units based on sampling volume, filtration volume, and elution volume to ensure comparability between different batches. The time series is aligned by interpolating discrete sampling time points into a continuous time series to eliminate the impact of inconsistent sampling intervals on the rate of change calculation. Finally, the time-series change rate of eDNA abundance is obtained by calculating the difference in eDNA copy number concentration between two consecutive sampling time points and dividing it by the corresponding time interval. The rate quantifies the instantaneous change trend of the biomass potential of *Ulva prolifera* in a target sea area within a specific time window. The calculated eDNA abundance change rate is used as a dynamic biological indicator. Compared with the absolute concentration, which only reflects the biomass stock at a certain moment, it captures the initial proliferation activity of *Ulva prolifera* populations in the pre-outbreak stage. Specifically, from the pre-processed time series, two consecutive time points and their corresponding eDNA copy number concentrations are selected sequentially. The difference in eDNA copy number concentration between the two time points is calculated, and the difference in eDNA copy number concentration is divided by the corresponding time interval to obtain the average change rate within the time window, which is the eDNA abundance change rate. Through this process, a change rate value can be calculated for each time interval, forming a new time series that reflects the rate of change in eDNA concentration. The formula for calculating the rate of change in eDNA abundance is as follows: ; In the formula: eDNA abundance change rate, unit: copies / (L·day); For at a certain point in time The measured eDNA copy number concentration (copies / L); For at a certain point in time The measured eDNA copy number concentration (copies / L); The time interval (in days) between two consecutive sampling points. S3. Collect environmental parameters concurrently and establish a joint database of environmental driving factors and green tide biomass by combining satellite remote sensing data. At the same time as the initial eDNA sample is collected, multiple environmental parameters including water temperature, salinity, nutrient concentration, dissolved oxygen, current velocity, wind speed and chlorophyll a concentration are measured simultaneously to achieve spatiotemporal synchronous collection of ecological factors and genetic signals. Combined with multi-source satellite remote sensing data, green tide biomass information, including green tide coverage area and density distribution, is extracted from the target sea area where green tides frequently occur as a macroscopic characterization of green tide biomass, improving the accuracy and timeliness of macroscopic biomass monitoring. The environmental parameters and green tide biomass information are integrated to construct a spatiotemporally synchronous joint database of environmental driving factors and green tide biomass, supporting multi-source data fusion and collaborative analysis. The specific work involves: simultaneously recording multi-dimensional environmental parameters to construct a driving factor database during eDNA sample collection in the target sea area; measuring basic water quality indicators including water temperature, salinity, and dissolved oxygen to reflect the physicochemical properties of the water body; quantifying nutrient concentrations (such as nitrates and phosphates) using chemical analysis methods to assess the eutrophication level of the water body; acquiring flow velocity and wind speed using current and anemometers to capture key dynamic conditions affecting substance diffusion; and simultaneously measuring chlorophyll a concentration to indirectly characterize phytoplankton biomass. All environmental parameter measurements are synchronized with eDNA sampling (error ≤ 1 hour) to ensure spatiotemporal consistency. Data recording adopts a standardized format, including the latitude and longitude of the sampling site, depth, timestamp, and parameter units. Based on multi-source satellite remote sensing data (MODIS, Landsat, Sentinel series), macroscopic information on green tide biomass in the target sea area is extracted. Interference is eliminated through remote sensing image preprocessing (radiative correction, atmospheric correction, geometric correction). The floating algae index (FAI) is used to identify the green tide coverage area, classify the green tide, quantify its coverage area, and evaluate it through texture feature analysis. Green tide density distribution was determined by calculating contrast and correlation based on the gray-level co-occurrence matrix (GLCM) to classify light, moderate, and heavy cover levels. Furthermore, multi-source satellite remote sensing data was cross-validated with field survey data to ensure extraction accuracy (coverage area error ≤10%, density classification consistency ≥85%). This resulted in a spatiotemporally continuous green tide biomass distribution layer, which was incorporated into the database as a macro-ecological indicator. Ground environmental parameters were spatiotemporally aligned and fused with green tide biomass information extracted from multi-source satellite remote sensing data to construct a driving factor-biomass joint database. This database unifies the... The data has spatial resolution (1km×1km grid) and temporal scale (daily scale). Spatial matching is performed through a GIS platform. For data with temporal mismatch (satellite revisit cycle > 1 day), time interpolation is used to fill missing values. At the same time, when integrating the two types of data, the data source, quality level and uncertainty range are labeled to ensure traceability. The resulting joint database contains the following fields: station ID, time, latitude and longitude, water temperature, salinity, nutrients, dissolved oxygen, flow velocity, wind speed, chlorophyll a, green tide coverage area and coverage level, supporting multi-dimensional query and statistical analysis. The formula for calculating the floating algae index is as follows: ; ; In the formula: The floating algae index is used to enhance the signal of floating vegetation (such as seaweed) in remote sensing images; The near-infrared band surface reflectance is corrected; Ulva prolifera has high reflectance in this band. The baseline reflectance is the reflectance that "should have" at the near-infrared position, estimated by linear interpolation based on the red and short-wave infrared bands. The red band surface reflectance is the corrected value; water bodies have strong absorption in this band. The value is the corrected shortwave infrared band surface reflectance. Clean water has extremely strong absorption in this band, and its reflectance is close to 0. It is the center wavelength of the near-infrared band; It is the center wavelength of the red light band; It is the center wavelength of the shortwave infrared band; For clean water bodies: In and The reflectivity of the bands is very low and the values ​​are similar, therefore ; For areas covered by green tides: *Ulva prolifera* in... The band has high reflectivity, and The reflectivity of the band remains very low, resulting in Significantly greater than ,therefore And usually The higher the value, the stronger the green tide signal.

[0021] By counting the total number of all green tide pixels and multiplying it by the actual area of ​​each pixel, we can obtain the total area covered by the green tide. The expression for this is: ; in, For the area covered by the green tide, For the number of pixels in the green tide, Area of ​​a single pixel; Gray-level co-occurrence matrix (GLCM) is a classic method for describing image texture. It selects two features, contrast and correlation, to evaluate the density and uniformity of green tides. ; ; In the formula: For GLCM contrast, For GLCM correlation; This is a gray-level co-occurrence matrix, where the matrix contains elements located at... The element value represents the grayscale value in a certain direction of the image. The pixel and grayscale value The probability of pixels appearing simultaneously; The gray level of the image; The gray level of the image, with values ​​ranging from 0 to... ; , for row mean ( ) and column mean ( ); , for row standard deviation ( ) and standard deviation ( Contrast measures the degree of local change in an image. When the gray values ​​of adjacent pixels in an image differ greatly (i.e., the texture is coarse and the edges are sharp), the contrast is the most significant difference. A high contrast value results in a high contrast value. In green tide density analysis: high-density / aggregated areas have concentrated green tides with uniform internal texture, but clear boundaries and high contrast with the surrounding water, leading to a high contrast value. Low-density / sparse areas have sparse and dispersed green tides, intermingling with the water background, resulting in lower overall contrast and a lower contrast value. Correlation measures the linear dependence of image gray levels, i.e., the uniformity of texture. When the image texture is uniform and regular, the correlation is high. In green tide density analysis: high-density / aggregated areas have uniform internal texture and high pixel gray value similarity, resulting in a high correlation value. Low-density / sparse areas have fragmented and irregular textures, with weak relationships between pixel gray values, resulting in a low correlation value. S4. Standardize the eDNA abundance change rate, environmental driving factors, and green tide biomass to establish a multi-factor response model. Use time series analysis to determine the time lead of eDNA signals to green tides, outputting a green tide risk index to quantify the risk of green tide outbreaks. Standardize the data on eDNA abundance change rate, environmental driving factors, and green tide biomass to eliminate the influence of dimensions, improve data comparability and model convergence efficiency. Based on the standardized data, select at least one of the generalized additive model, boosting regression tree, or XGBoost algorithm to establish a multi-factor response model for green tide biomass jointly driven by eDNA and environmental factors, enhancing the fitting ability to complex ecological relationships. Perform time series analysis on the multi-factor response model to determine whether there is a time lead of eDNA signals relative to the macro-green tide outbreak and its specific length, providing statistical basis for early warning and realizing advanced quantitative identification of green tide outbreak risks. The specific work involves: using the Z-score standardization method to standardize the data on eDNA abundance change rate, environmental driving factors (water temperature, salinity, nutrient concentration, flow velocity, wind speed, etc.), and green tide biomass. Each parameter is converted to a standardized value with a mean of 0 and a standard deviation of 1. The standardized data are stored in a unified database, with the original data range and standardized parameters labeled. The original data is retained to support result verification and uncertainty analysis. Based on the standardized data, at least one of the following algorithms—Generalized Additive Model (GAM), Boosting Regression Tree (BRT), or XGBoost—is used to construct a response model for green tide biomass jointly driven by eDNA and environmental factors. GAM captures the complex relationships between variables through a nonlinear smoothing function, making it suitable for exploratory analysis. BRT and XGBoost, as ensemble learning methods, improve the model's prediction accuracy through multi-tree combinations. With robustness, it is particularly suitable for handling high-dimensional nonlinear data. The model input includes standardized eDNA abundance change rate, environmental driving factors, and green tide biomass, and the output is the probability distribution of green tide biomass. During training, cross-validation is used to optimize hyperparameters, and mean squared error is used to evaluate model performance to ensure that its explanatory power and generalization ability meet the needs of ecological research. Time series analysis is performed on the multi-factor response model. The time correlation characteristics between eDNA signal and green tide biomass are extracted using the sliding window method. The time lead period of eDNA abundance change rate relative to the green tide outbreak is analyzed, that is, the length of time that the eDNA signal precedes the significant increase in green tide biomass. If the eDNA signal shows a significant lead, it indicates that it can be used as an early warning indicator of green tide. If the lead period is unstable or there is no significant correlation, the regulatory role of other driving factors is further analyzed, and an analysis report including the confidence interval of the lead period and the contribution of environmental factors is output. Furthermore, the process of establishing a green tide risk prediction model driven by both eDNA and environmental factors includes: time series analysis, specifically Granger causality test and cross-correlation analysis, to quantify the lead-lag relationship between eDNA signals and green tide events, statistically verify the leading role of eDNA signals, lay the foundation for early warning, and construct a prediction equation based on a multi-factor response model and a determined time lead period to calculate the green tide risk index, thereby quantifying the risk of green tide outbreaks in future periods, constructing a closed-loop early warning system, and realizing the connection from monitoring to decision-making. The specific work involved: using Granger causality tests and cross-correlation analysis in time series analysis to quantify the leading-lag relationship between eDNA signals and green tide events. The Granger causality test was used to statistically determine whether eDNA sequences could effectively explain future information about green tide biomass changes. If introducing historical eDNA data significantly improved the prediction accuracy of green tide biomass, then the eDNA sequence was determined to have a Granger causal relationship with the green tide event, thus statistically confirming its leading role. Cross-correlation analysis calculated the correlation coefficients between eDNA sequences and green tide biomass sequences at different time lags to identify the lag with the strongest correlation between the two. To determine the specific lead time of the eDNA signal relative to the green tide outbreak, two methods are used to complement each other: Granger causality test confirms the statistical significance of the predictive ability, while cross-correlation analysis quantifies the specific length of the lead time. Based on the determination of the eDNA signal's time lead time, a green tide risk prediction equation with a multi-factor response model as its core is constructed. Standardized environmental driving factors and the eDNA abundance change rate during the lead time are used as input variables. The nonlinear mapping relationship is obtained through training with a generalized additive model, boosting regression tree, or XGBoost algorithm, and a comprehensive green tide risk index is output to quantify the probability of a green tide outbreak occurring in the target sea area within a specific future period. The expression for the green tide risk index is as follows: ; In the formula: The green tide risk index quantifies the risk at a specific point in the future. The possibility of a green tide outbreak in the target sea area; It represents a complex nonlinear mapping function trained by machine learning algorithms such as generalized additive models, boosting regression trees, or XGBoost, which encapsulates all the interactions between input variables and risk indices; Indicates the starting point of the leading period Time-varying, standardized eDNA abundance change rate It is the specific lead time of the eDNA signal determined through time series analysis; Represents the current time A series of environmental driving factors were measured and standardized; This represents the current moment, that is, the moment when predictions are made and environmental factor data are collected; This represents the forecast period, i.e., how far into the future the risk is predicted, compared to the lead time of eDNA. Related or set according to management needs; This represents the specific lead time of the eDNA signal determined through cross-correlation analysis; A gradually increasing value indicates an increasing risk of an outbreak, usually caused by positive changes in one or more driving factors; S5. Based on historical data, set grading thresholds for eDNA copy number concentration, establish grading early warning standards, and combine them with the green tide risk index. Introduce graph neural networks to dynamically extrapolate the green tide diffusion path and identify risk transmission channels. The process of establishing a tiered early warning standard includes: based on the correlation between eDNA copy number concentration and the actual outbreak intensity of *Ulva prolifera* green tides in historical data, setting eDNA copy number concentration thresholds for different levels to achieve scientific threshold setting based on long-term statistics, thereby improving the reliability of early warnings; using K-means clustering algorithm and / or ROC analysis to determine the optimal dividing point for classifying risk levels, forming a tiered early warning standard that includes safety, concern, warning, and alert levels, achieving objectivity and data-driven risk level classification, and enhancing the accuracy of early warnings. Safety is Level I, which is the algae-free period, with an eDNA copy number concentration threshold of <10. 3 (copies / L); the concern level is II, its status is in the nascent stage, and the eDNA copy number concentration threshold is 10. 3 -10 5 (copies / L); the warning level is III, indicating a latent period, with an eDNA copy number concentration threshold of 10. 5 -10 6 (copies / L); Alert level IV, indicating an outbreak, with an eDNA copy number concentration threshold of >10. 6 (copies / L) enables full coverage from algae-free to outbreak, with more precise graded response. When the real-time monitored eDNA copy number concentration crosses a certain threshold in the graded early warning standard, it automatically triggers and releases the corresponding risk level early warning information, realizing the linkage between real-time monitoring and automatic early warning, and improving response timeliness. The specific work involves analyzing long-term observation records of eDNA copy number concentration and actual outbreak intensity of *Ulva prolifera* green tides based on multi-year historical data. Statistical analysis is used to establish a quantitative correlation between the two. Historical data is cleaned and standardized to eliminate the influence of seasonal fluctuations and environmental disturbances, ensuring data comparability. Nonlinear regression is employed to fit the mapping relationship between eDNA copy number concentration and outbreak intensity indicators, including green tide biomass and coverage area, clarifying the probability of green tide occurrence and potential impact range corresponding to different concentration ranges. Finally, based on the fitting results and combined with ecological safety thresholds... Based on value theory, multi-level eDNA copy number concentration thresholds were set. From a statistical and practical perspective, K-means clustering and ROC analysis were used to comprehensively determine the optimal cutoff point, classifying the eDNA copy number concentration thresholds into four risk levels: safe, concerning, warning, and alert. Safe indicates extremely low *Ulva prolifera* biomass in the water body, lacking conditions for an outbreak, requiring routine monitoring. Concerning indicates the detection of early proliferation signals, suggesting outbreak potential, requiring enhanced monitoring and attention to changes in environmental factors. Warning indicates the *Ulva prolifera* population has entered a rapid proliferation phase, with a high risk of outbreak, requiring consultation and preparation of control materials and contingency plans. Alert indicates a large-scale green tide outbreak, causing... If the event causes substantial ecological impact, an emergency response should be immediately initiated, including interception and salvage operations. Specifically: K-means clustering algorithm should be used to perform unsupervised classification of historical eDNA copy number concentration data. The elbow rule should be used to optimize the number of clusters, allowing different categories to naturally correspond to potential boundaries of green tide risk levels. ROC (Receiving Analytical Capability) analysis should be used to evaluate the model's ability to distinguish green tide outbreaks at different concentration thresholds. The Youden exponent should be used to maximize the boundary between sensitivity and specificity. Finally, by combining the clustering results and ROC analysis conclusions, a concentration value that has both ecological significance and statistical significance should be selected as the threshold. Early warning thresholds at all levels; Construct an automated early warning system that integrates real-time eDNA monitoring data streams with tiered early warning standards, continuously receives eDNA copy number concentration data, and after data verification and preprocessing, dynamically compares it with preset four-level thresholds. When the monitored value exceeds a certain threshold, an early warning for the corresponding risk level is immediately triggered, and information is released to management departments, research institutions, and the public through multiple channels (SMS, email, and APP push), covering the current risk level, potential affected areas, and suggested countermeasures. At the same time, the time of each early warning trigger, concentration value, and subsequent green tide development are recorded to form a closed-loop feedback mechanism. The process of identifying risk transmission channels includes: based on the graded early warning standards, combining the output green tide risk index and introducing a graph neural network model to achieve a leap from single-point early warning to networked dynamic simulation, thereby improving the accuracy of spatial early warning. The graph neural network model uses the green tide risk index as node features and the diffusion dynamic conditions of ocean current field and wind direction as edge weights to construct a graph structure, quantify the ecological connections and dynamic driving forces between nodes, enhance the realism of diffusion path simulation, and use the graph neural network to dynamically simulate the potential diffusion paths of Ulva prolifera green tide, analyze the intensity of key paths and the risk aggregation degree of node communities, so as to identify key risk transmission channels and high-risk areas, clarify key prevention and control areas and paths, and improve the efficiency of emergency resource allocation. The specific work involves: building a spatial propagation model based on graph neural networks, using the established tiered early warning standard system to dynamically extrapolate the spread of *Ulva prolifera* green tides. The target sea area is abstracted as a graph structure, where each monitoring station is considered a node. The node's feature vector is primarily composed of the green tide risk index output by a multi-factor response model. Standardized information from driving factors such as eDNA signal, water temperature, and nutrients is integrated to quantitatively characterize the current location's green tide potential. The graph's edge structure is constructed based on ocean current field vector data and wind direction and speed observations. The edge weights are determined by the strength of the flow direction correlation between stations, Euclidean distance, and historical diffusion patterns, thereby quantifying different... The probability and intensity of green tide biomass transport between locations were studied. Using the constructed graph structure, the potential diffusion path of Ulva prolifera green tide was dynamically extrapolated through the message passing mechanism of graph neural network. The spatial propagation model aggregated the feature information of adjacent nodes and updated the information according to the edge weights. The spatial transmission process of green tide biomass driven by flow field and wind field was iteratively simulated. The extrapolation of each time step was combined with the latest marine dynamic forecast data to adapt to the dynamic scenario of flow field changes. By performing spatiotemporal analysis on the node state changes of multiple time steps, the critical path that maintains a high risk index was identified, which constitutes the risk propagation channel of green tide diffusion. At the same time, the node community that forms risk clusters in the network was detected to identify high-risk areas. The formula for calculating the critical path strength is as follows: ; In the formula: Critical path strength, i.e., path The total intensity score is used to quantify the significance of a path as a risk propagation channel throughout the entire time window. The higher the score, the greater the likelihood that the path is a major risk channel. A spatial path to be evaluated consists of a series of spatially continuous nodes. constitute; , The time window for spatiotemporal analysis starts from the start time. End time ; For the path A node on; For nodes At any moment The green tide risk index; For nodes At any moment Spatial coherence weights are used to reward nodes that form coherent paths in space and time. The value tends to increase when the path meets the following conditions: high-risk persistence, nodes on the path over a long period of time (from arrive Green Tide Risk Index Both are very high, with spatial coherence; the nodes on the path are spatially continuous, and the high-risk signals are transmitted sequentially along the path (by...). (Capture); by searching all possible paths The one or a few highest-scoring lines can be used to identify the primary channels of risk transmission. The expression for calculating the clustering degree of node community risk is as follows: ; In the formula: The risk clustering degree of the node community, i.e., the community The risk clustering degree comprehensively reflects the average risk level of the region and its clustering ability in the network. The higher the value, the stronger the evidence that the region is considered a high-risk area. This refers to the node communities identified from a graph network using community detection algorithms (Louvain and Leiden algorithms). For communities The number of nodes contained therein; For communities One of the nodes; For nodes The average green tide risk index within the analysis time window; For nodes The weighted degree, which is the sum of the weights of all edges connected to the node, is the centrality of a node in a graph network. Nodes with high degree centrality are the hubs of connections and are more likely to become the gathering places of matter. The value tends to increase under the following conditions: high risk level, average risk index of nodes within the community. High connectivity, high degree of connectivity between nodes within the community in the network ( The large size of the green tide biomass indicates that it is easy for green tide biomass to enter but difficult to leave, thus forming aggregations; using a community detection algorithm, the entire network is divided into several tightly connected communities. Calculate the risk clustering degree of each community. , The community with the highest value is identified as a high-risk area; S6. Using spatial interpolation and GIS visualization technology, generate a spatial distribution map of green tide risks and visualize the risk transmission channels, while releasing early warning information to guide targeted defense.

[0022] The process of issuing early warning information to guide targeted deployment includes: using inverse distance weighted interpolation or Kriging interpolation methods to spatially interpolate monitoring and prediction data of discrete stations, achieving seamless coverage of monitoring data from point to area, improving spatial perception capabilities, generating continuous green tide risk spatial distribution maps, forming high-precision risk base maps to support overall situation assessment, and in the GIS platform, fusing and overlaying green tide risk spatial distribution maps with satellite remote sensing images, and realizing the visualization of risk propagation channels to enhance the intuitive identification and spatial positioning efficiency of risk information. Through the network interface or decision support system, daily early warning products containing green tide risk spatial distribution maps and risk propagation channel information are issued to ensure the real-time nature and business continuity of early warning information, so as to guide targeted prevention and control deployment, achieve hierarchical and precise push, and improve the efficiency of emergency response coordination. The specific work involves: transforming green tide risk index and eDNA data obtained from discrete monitoring stations into continuous spatial distribution information using spatial interpolation methods (inverse distance weighted interpolation or Kriging interpolation). Inverse distance weighted interpolation uses distance as a weight, assuming that spatially close stations have higher similarity, and is suitable for scenarios with uniform station distribution and strong spatial autocorrelation. Kriging interpolation, as a statistical method, not only considers distance relationships but also analyzes the structure and randomness of spatial data through variogram analysis, reflecting the spatial variation characteristics of regionalized variables, and is particularly suitable for... In sea areas exhibiting spatial trends or heterogeneity, an appropriate algorithm is selected during interpolation based on station density, data distribution characteristics, and the marine environment. Interpolation parameters are optimized to control smoothness and accuracy, generating a continuous spatial distribution map of green tide risk covering the entire sea area. In a GIS platform, this continuous green tide risk spatial distribution map is fused with multi-source satellite remote sensing imagery. Before fusion, coordinate system I, resolution matching, and geometric correction must be completed to ensure spatial benchmark consistency. The green tide risk spatial distribution map is overlaid on the remote sensing imagery base map using pseudo-color rendering, with adjustments made to transparency and color. The risk assessment system highlights the distribution range and intensity of different risk areas. For risk propagation channels identified by graph neural network inference, vector lines are used for representation, with line width and color gradients set according to channel intensity to dynamically indicate their spatial direction and evolution trend. High-risk clusters, key monitoring stations, and other environmental auxiliary information are also marked on the map, constructing a comprehensive risk assessment map integrating multi-dimensional information to showcase the current situation and potential future spread paths of the green tide. The processed green tide risk spatial distribution map and propagation channel information are integrated and published through a web interface or decision support system. Based on standardized data formats and web map services, daily early warning products are automatically generated and dynamically updated. In addition to the core risk illustrations, the published content includes structured text information such as risk level statistics, key area descriptions, brief analysis of changing trends, and prevention and control recommendations. It also features multi-level access control and targeted push functions to ensure that early warning information can be delivered to relevant management departments, research institutions, and operational units in real time. Through human-computer interaction, users can query risk details for specific locations and review historical dynamics, providing intuitive spatial decision-making basis for regional collaborative monitoring, emergency resource allocation, and targeted prevention and control deployment.

[0023] Example 2, as Figure 1 , Figure 2 As shown in Example 1, this invention provides an example of an early warning method for Ulva prolifera green tide disaster based on eDNA in 2025, as follows: S1. Set up monitoring sections and fixed stations in areas prone to green algae blooms, and collect surface water samples regularly to obtain initial eDNA samples.

[0024] From May to August 2025, samples were collected at a total of 8 sampling stations in Shandong (Rushan, Qingdao, Rizhao) and Jiangsu (Lianyungang, Sheyang, Dafeng, Rudong, Qidong). The latitude and longitude of each sampling station were accurately recorded, along with the observation of *Ulva prolifera* at the site. 5-10L of water samples were collected from each station, transported to the laboratory under refrigeration, filtered onto a filter membrane, and stored at -80℃.

[0025] S2. Process the initial eDNA sample and select primer sequences specific to the Ulva prolifera green tide outbreak genotype to quantitatively detect the DNA copy number concentration of the Ulva prolifera green tide outbreak genotype in the water body, and calculate the eDNA abundance change rate. Referring to the example, total eDNA was extracted from the filter membrane sample collected in S1, and 5S rDNA primer sequences specific to the Ulva prolifera green tide outbreak genotype were selected. The DNA copy number concentration of the Ulva prolifera green tide outbreak genotype in the water body was quantitatively detected by real-time quantitative PCR, and the eDNA abundance change rate was calculated based on the standard curve.

[0026] S3. Collect environmental parameters concurrently and construct a joint database of environmental driving factors and green tide biomass by combining them with satellite remote sensing data; when collecting initial eDNA samples, collect environmental driving factor data such as water temperature, salinity, nutrient concentration, flow velocity, and wind speed simultaneously. Based on multi-source high-resolution satellite remote sensing data, interpret the daily coverage area and distribution area of ​​the Ulva prolifera green tide, and construct a joint database of environmental driving factors and green tide biomass. The 2025 Ulva prolifera green tide outbreak genotype eDNA detection results show: On May 6, the abundance of the eDNA genotype for the green tide outbreak of *Ulva prolifera* at all sampling stations was above 10. 5 Below copies / L; On May 11, with the rise in seawater temperature, the abundance of the cDNA genotype for the green tide of *Ulva prolifera* increased, reaching its highest point in Rudong, with an abundance of 4.66 × 10⁻⁶. 6 copies / L; On May 17, the highest abundance of the *Ulva prolifera* green tide genotype eDNA shifted northward to Dafeng, with an abundance of 3.36 × 10⁻⁶. 6 copies / L; On May 20, satellite remote sensing discovered floating seaweed for the first time near the shore of a laver farming area in the shallow waters of northern Jiangsu. On May 26, satellite remote sensing detected that the floating seaweed covered an area of ​​more than 5 square kilometers, indicating a large-scale outbreak of seaweed.

[0027] On June 6, the abundance of the cDNA genotype for the green tide outbreak of Ulva prolifera increased significantly, with the abundance of cDNA for the green tide outbreak genotype increasing to 10 at 7 out of 8 observation stations. 6 The highest abundance was found near the Rizhao sea area of ​​Lianyungang, at 1.22 × 10⁻⁶ copies / L.7 copies / L; On June 21, the highest abundance of the cDNA genotype for the green tide of *Ulva prolifera* was observed near the coast of Rizhao, at 1.02 × 10⁻⁶. 6 copies / L, at which point floating seaweed had been detected landing near the coast of Rizhao; On June 29, the highest genotype eDNA abundance during the Ulva prolifera green tide outbreak was 1.21 × 10⁻⁶. 7 The copies / L are still located in the waters near Lianyungang. On July 9, the highest abundance of the cDNA genotype for the Ulva prolifera green tide outbreak was observed in Qingdao, at 4.92 × 10⁻⁶. 6 Copies / L, the amount of floating seaweed coming ashore in the nearshore waters of Qingdao has begun to increase.

[0028] S4. The eDNA copy number change rate, environmental driving factors, and green tide biomass were standardized to establish a multi-factor response model. Time series analysis was used to determine the time lead of eDNA signals to green tide outbreaks, outputting a green tide risk index to quantify the risk of green tide outbreaks. Time series analysis showed that the large-scale outbreak of *Ulva prolifera* green tide in the Yellow Sea in 2025 occurred on May 26th. The significant increase in eDNA abundance of the *Ulva prolifera* green tide genotype in the nearshore waters of the Subei Shoal occurred on May 11th, 9 days earlier than the earliest satellite detection time and 15 days earlier than the large-scale outbreak time. Therefore, eDNA technology can be used for early warning of large-scale *Ulva prolifera* outbreaks. From May 6th to July 9th, the highest abundance point of *Ulva prolifera* gradually shifted northward, with the outbreak center located in Qingdao in early July. At most sampling sites, the peak abundance of the cDNA genotype of Ulva prolifera green tide occurred sequentially from south to north from late June to July (especially from June 6 to July 29), and the time of significant increase in cDNA abundance was earlier than the time of influence of Ulva prolifera green tide.

[0029] S5. Based on historical data, set grading thresholds for eDNA copy number concentration, establish grading early warning standards, and combine them with the green tide risk index. Introduce graph neural networks to dynamically extrapolate the green tide diffusion path and identify risk transmission channels.

[0030] S6. Using spatial interpolation and GIS visualization technology, generate a spatial distribution map of green tide risk, and at the same time release early warning information to guide targeted deployment.

[0031] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An early warning method for *Ulva prolifera* green tide disaster based on environmental DNA, characterized in that, Includes the following steps: S1. Set up monitoring sections and fixed stations in areas prone to green tides of Ulva prolifera, and collect surface water samples regularly to obtain initial eDNA samples, i.e., initial environmental DNA samples. S2. Process the initial eDNA sample and select primer sequences specific to the genotype of Ulva prolifera green tide outbreak. Quantitatively detect the eDNA copy number concentration of the Ulva prolifera green tide outbreak genotype in the water body to calculate the eDNA abundance change rate. S3. Collect environmental parameters from the same period and combine them with satellite remote sensing data to construct a joint database of environmental driving factors and green tide biomass; S4. Standardize the eDNA copy number change rate, environmental driving factors and green tide biomass, establish a multi-factor response model, and use time series analysis to determine the time lead of eDNA signal to green tide outbreak, output green tide risk index, and quantify green tide outbreak risk. S5. Based on historical data, set grading thresholds for eDNA copy number concentration, establish grading early warning standards, and combine them with the green tide risk index. Introduce graph neural networks to dynamically extrapolate the green tide diffusion path and identify risk transmission channels. S6. Using spatial interpolation and GIS visualization technology, generate a spatial distribution map of green tide risk, and at the same time release early warning information to guide targeted deployment.

2. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: S1 specifically includes: In the target sea areas where green tides of seaweed are frequent, including the Yellow Sea region, monitoring sections and fixed stations covering different gradient areas of nearshore, coastal and offshore areas will be set up. At the monitoring section and fixed station, surface seawater samples of 2 to 5 liters are collected regularly at a depth of 0 to 2 meters every 3 to 5 days. The sampling work continuously covered the entire cycle before, during and after the outbreak of the green tide of Ulva prolifera, in order to obtain initial eDNA samples that are spatially and temporally representative.

3. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In step S2, the process of quantitatively detecting the eDNA copy number concentration in the water body includes: Seawater samples were filtered through a 0.45-micron pore size membrane to enrich biomass, and total eDNA was extracted from the membrane using a commercial DNA extraction kit. We selected specific gene fragments from the genotypes of green tide outbreaks of Ulva prolifera, designed and synthesized specific primer sequences for quantitative molecular detection; Quantitative PCR or digital PCR technology was used to amplify and detect the extracted eDNA using the specific primers, and the copy number concentration of the eDNA of the Ulva prolifera green tide outbreak genotype in the water was determined, so as to achieve absolute quantification and high-sensitivity monitoring of Ulva prolifera biomass in the water.

4. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 3, characterized in that: In S2, the process of calculating the rate of change of eDNA abundance includes: Based on the eDNA copy number concentration data measured at different sites and at different time series, the temporal change rate of eDNA abundance was calculated, i.e., the eDNA abundance change rate. The eDNA abundance change rate is the difference in eDNA copy number concentration per unit time, which is used for coupled modeling and early warning analysis.

5. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In S3, the process of establishing a joint database of environmental driving factors and green tide biomass includes: During the collection of eDNA samples, multiple environmental parameters, including water temperature, salinity, nutrient concentration, dissolved oxygen, flow rate, wind speed, and chlorophyll a concentration, were measured simultaneously. By combining multi-source satellite remote sensing data, we extract green tide-related information on the coverage and distribution area of ​​Ulva prolifera in the target sea area, which serves as a macroscopic characterization of green tide biomass. By integrating the environmental parameters with the green tide biomass information, a spatiotemporally synchronized joint database of environmental driving factors and green tide biomass is constructed.

6. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In step S4, the process of determining the time lead of the eDNA signal to the green tide outbreak using time series analysis includes: The data on eDNA abundance change rate, environmental driving factors, and green tide biomass were standardized. Based on the standardized data, at least one of the generalized additive model, boosting regression tree or XGBoost algorithm was selected to establish a multi-factor response model of green tide biomass jointly driven by eDNA and environmental factors. Time series analysis was performed on the multifactor response model to determine whether there was a time lead of the eDNA signal relative to the macroscopic green tide outbreak and its specific length.

7. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In S4, the process of establishing a green tide risk prediction model jointly driven by eDNA and environmental driving factors includes: The time series analysis specifically includes Granger causality test and cross-correlation analysis, used to quantify the leading and lagging relationship between eDNA signal and green tide events; Based on the multi-factor response model and the determined time lead period, a prediction equation is constructed to calculate the green tide risk index in order to quantify the risk of a green tide outbreak in the future period.

8. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In S5, the process of establishing tiered early warning standards includes: Based on the correlation between historical data on eDNA copy number concentration and the actual outbreak intensity of Ulva prolifera green tide, different levels of eDNA copy number concentration thresholds were set. K-means clustering and / or ROC analysis were used to determine the optimal cutoff points for classifying risk levels, forming a tiered early warning standard that includes safety, concern, warning, and alert. Safety is classified as Level I, characterized by an algal-free period and an eDNA copy number concentration threshold of <10. 3 The level of concern is II, indicating a nascent stage, with an eDNA copy number concentration threshold of 10. 3 -10 5 The warning level is Level III, indicating a latent period, with an eDNA copy number concentration threshold of 10. 5 -10 6 The alert level is IV, indicating an outbreak phase, with an eDNA copy number concentration threshold of >10. 6 ; When the real-time monitored eDNA copy number concentration crosses a certain threshold in the graded early warning criteria, the corresponding risk level early warning information will be automatically triggered and issued.

9. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 8, characterized in that: In S5, the process of identifying risk transmission channels includes: Based on the aforementioned graded early warning standards, a graph neural network model is introduced, combining the output green tide risk index. The graph neural network model uses the green tide risk index as node features and the diffusion dynamics of ocean current field and wind direction as edge weights to construct the graph structure. The graph neural network is used to dynamically extrapolate the potential diffusion paths of Ulva prolifera green tides, analyze the intensity of critical paths and the risk clustering degree of node communities, so as to identify key risk propagation channels and high-risk areas.

10. The method for early warning of *Ulva prolifera* green tide disaster based on environmental DNA according to claim 1, characterized in that: In step S6, the process of issuing early warning information to guide targeted deployment includes: Using inverse distance weighted interpolation or Kriging interpolation methods, spatial interpolation is performed on monitoring and forecasting data from discrete stations to generate a continuous spatial distribution map of green tide risk. In the GIS platform, the spatial distribution map of the green tide risk is fused and overlaid with satellite remote sensing images to achieve visual identification of the risk propagation channels; Daily early warning products containing the aforementioned green tide risk spatial distribution map and risk transmission channel information are released through a web interface or decision support system to guide targeted prevention and control measures.

Images