Enteromorpha green tide marine environment assessment and prediction method based on deep learning

Abstract

The invention relates to the technical field of marine environment monitoring and prediction, and particularly provides an enteromorpha green tide marine environment assessment and prediction method based on deep learning. The method comprises the steps that multi-source marine environment data are collected and preprocessed; constructing a spatial feature project and a time feature; a suitability scoring model based on the growth cycle is established, and applicability marking is carried out in combination with sea area division; according to the method, the two-way space convolution GRU prediction model is constructed, training optimization is performed on the two-way space convolution GRU prediction model so as to evaluate and predict the future marine environment and output the prediction result, and the scientificity and the foresight of green tide management and prevention and control are improved by constructing the two-way space convolution GRU prediction model.

Description

Technical Field

[0001] This invention relates to the field of marine environmental monitoring and prediction technology, and in particular to a deep learning-based method for assessing and predicting marine environmental conditions related to seaweed green tides. Background Technology

[0002] Green tides of *Ulva prolifera* are among the largest and most widespread algal blooms globally. Since their first outbreak in the Yellow Sea of ​​China in 2007, they have caused large-scale green tide disasters almost every year from late spring to summer. The Yellow Sea's unique hydrogeographical conditions and the impact of human activities make it a high-risk area for frequent green tides. Statistics show that the maximum coverage area of ​​green tides can reach tens of thousands of square kilometers, severely damaging marine ecosystems and causing huge economic losses to fishery resources, coastal infrastructure, and tourism. Therefore, effective monitoring and accurate prediction of *Ulva prolifera* green tides have become a key issue in marine environmental management and disaster prevention and control.

[0003] The current monitoring and prediction of Ulva prolifera green tides mainly rely on the following three types of technical means: (1) Remote sensing monitoring technology. Remote sensing is currently the most widely used monitoring method. It mainly relies on satellite image data such as MODIS and GOCI, and identifies Ulva prolifera areas through vegetation indices (such as NDVI and FAI). The advantage of this method is that it has a wide monitoring range and can provide large-scale, near-real-time green tide distribution information. However, its main limitation is that it can only conduct post-event observations and it is difficult to achieve early warning. At the same time, it is greatly affected by cloud cover and weather conditions, and the spatial resolution of satellite images is low, making it difficult to accurately identify nearshore details. (2) Numerical ocean model. Numerical model is a simulation method based on ocean physical processes. It usually considers the influence of dynamic factors such as ocean currents and wind fields on the drift and spread of Ulva prolifera. Its advantage is that it is based on physical mechanisms and has a strong theoretical foundation, and can simulate the movement path of Ulva prolifera. However, this type of model has problems such as computational complexity, parameter sensitivity, and low prediction accuracy, and generally ignores the biological mechanisms and ecological processes required for Ulva prolifera growth. (3) Statistical prediction methods: Statistical models rely on historical monitoring data and use traditional machine learning methods such as regression analysis, support vector machine (SVM), and random forest to predict the probability of green tides. These methods are computationally efficient and relatively simple to implement. However, because they ignore the nonlinear coupling relationship and time-dependent characteristics between environmental factors, their predictive ability is weak, and they are particularly difficult to adapt to rapid environmental changes and extreme events.

[0004] A thorough analysis of the aforementioned technologies reveals several key issues with existing methods for monitoring and predicting *Ulva prolifera* (green algae) outbreaks. First, there are significant shortcomings in modeling the biological mechanisms. Most methods neglect the complete life cycle of *Ulva prolifera* from germination and rapid growth to decline, failing to accurately reflect its ecological dynamics and leading to discrepancies between predicted and actual outbreak processes. Second, current models lack sufficient consideration of spatial heterogeneity. *Ulva prolifera* growth is significantly influenced by spatial variables such as topography, water depth, shore distance, and light intensity; omitting these factors can easily result in regional misjudgments. Furthermore, the ability to integrate multiple factors is limited. Environmental factors such as temperature, nutrients, light, salinity, and dissolved oxygen all play crucial roles in *Ulva prolifera* growth, but existing methods often struggle to effectively integrate and jointly model these variables, limiting the model's interpretability and generalization ability. In addition, most methods have limited capabilities in time-series modeling. The lack of effective capture of historical environmental change trends and seasonal patterns makes it difficult for prediction models to accurately grasp long-term change patterns. More critically, the current mainstream methods have low prediction accuracy, often failing to exceed 70%, thus failing to meet the precise early warning requirements of practical applications. At the same time, its forecast timeliness is poor, often only providing short-term forecasts of 1 to 3 days, lacking a time window for advance deployment and prevention. Summary of the Invention

[0005] In view of this, the present invention provides a deep learning-based method for assessing and predicting the marine environment of green tides caused by Ulva prolifera, in order to improve the scientific nature and foresight of green tide management and control.

[0006] In a first aspect, the present invention provides a deep learning-based method for assessing and predicting marine environmental conditions related to *Ulva prolifera* green tides, the method comprising: Step 1: Collect multi-source marine environmental data and perform preprocessing; Step 2: Based on Step 1, construct spatial feature engineering and temporal features; Step 3: Based on Step 2, establish a suitability scoring model based on the growth cycle, and combine it with sea area division to perform suitability labeling; Step 4: Using the method from Step 3, construct a bidirectional spatial convolutional GRU prediction model, train and optimize it to assess and predict the future marine environment, and output the prediction results.

[0007] Optionally, step 1 includes: Multi-source marine environmental data includes sea surface temperature data, marine optical data, nutrient concentration data, dissolved oxygen concentration, seawater salinity data, and the distribution area of ​​Ulva lactuca blooms in the ocean; among which, marine optical data includes photosynthetically active radiation and water transparency. Preprocessing includes outlier detection and handling, employing... The rules include: removing extreme values; filling missing values ​​using bilinear interpolation; normalizing data by normalizing all numerical variables to [0,1]; aligning time to a daily or monthly scale; and spatial interpolation using bilinear interpolation to map the data to a regular grid.

[0008] Optionally, step 2 includes: First, extract multi-dimensional spatial geographic features, including: geographic coordinates (latitude) ,longitude ); Distance from the center of the Yellow Sea The shortest distance from the coastline Seasonal codes: Winter = 0, Spring = 1, Summer = 2, Autumn = 3; Let the latitude of the location point be... Longitude is The spatial features include: a. Geographic coordinates: (Latitude= Longitude = ); b. Based on the distance from the center of the Yellow Sea: The coordinates of the center of the Yellow Sea are set as follows: The spherical distance between two points is calculated using the Haversine semi-versine function, and its expression is: ; Where R is the Earth's radius; c. Shortest distance from the coastline Its expression is: ; in, C represents the set of points on the coastline. The spherical distance; minimum value The nearest coastline; The temporal feature uses seasonal coding features as the input variable for the time dimension; the seasonal coding is winter=0, spring=1, summer=2, autumn=3; the temporal feature discretizes the four seasons into four numerical categories so that the model can identify the regulation of Ulva prolifera growth by different seasons. Its mathematical expression is as follows: Let m be the month of the current time, then the season code is... Determined by the following function: .

[0009] Optionally, step 3 includes: The suitability scoring model is divided into 12 growth stages based on months, and combines the basic suitability score with the temperature response coefficient to construct a suitability scoring mechanism that reflects the growth process throughout the year. Define suitability function Where m represents the month, ranging from 1 to 12; T represents the average sea surface temperature for the month; L represents the average solar radiation intensity for the month; N represents the average nitrate concentration for the month; and P represents the average phosphate concentration for the month. Monthly parameters include basic suitability score. Temperature factor Light factor Nutrient factors If the environmental mean vector includes temperature, light, nitrogen, and phosphorus, then the temperature vector... Lighting vector Nitrogen salt vector ; Phosphate vector ; Suitability Function for: ; The suitability function output value is used for suitability labeling, when Greater than or equal to the threshold If the condition is met, mark it as appropriate; otherwise, mark it as inappropriate.

[0010] Optionally, step 4 includes: First, a feature fusion layer is constructed to integrate heterogeneous inputs. The feature fusion layer includes environmental features, spatial features, and suitability scores based on a suitability scoring model. The above features are vectorized and dimension-unified through a multilayer perceptron (MLP) to ensure the integration of heterogeneous information, i.e., feature fusion. Subsequently, by embedding multi-scale convolutional layers to capture spatial dependencies at different scales, convolutional kernels of sizes 3×3, 5×5 and 7×7 are applied in parallel to extract local fine-grained changes and spatial trends, and their outputs are concatenated to form a multi-scale representation. The core component of the bidirectional spatial convolutional GRU prediction model is the bidirectional spatial convolutional GRU network. In the bidirectional spatial convolutional GRU network, the time series is processed simultaneously by forward and backward sequences. The forward sequence encodes historical dependencies, while the backward sequence integrates future information. The outputs of the two sequences are concatenated to capture the complete temporal dependencies. Convolutional operations are used to replace the fully connected layers in the bidirectional spatial convolutional GRU network to preserve spatial structure information during gating and state updates. To further enhance spatial differentiation, a spatial attention mechanism is introduced; attention weights are applied to the output of the bidirectional spatial convolutional GRU network to generate a spatial matrix. Finally, the output layer is designed with a dual-function head: a segmentation head and a prediction head. The segmentation head uses convolutional layers to generate a spatial distribution map for visualizing the suitability distribution. The prediction head integrates convolutional layers, ReLU layers, and fully connected layers to generate time series predictions.

[0011] Optionally, the core update of the basic unit of the bidirectional spatial convolutional GRU network is expressed as follows: ; in, Indicates an update to the door. This indicates that the door is being reset. This represents a candidate hidden state. Indicates the final hidden state. Indicates time t The input feature vector, This represents the hidden state at the previous time step. W and U This represents the weight parameters that can be trained. b Indicates the bias term. This represents element-wise multiplication. This represents the activation function Sigmoid; To further incorporate spatial structure information, the bidirectional spatial convolutional GRU network is improved in two aspects: a) Multi-scale spatial convolutional embedding: This method embeds input features using multi-scale convolution. The expression for extracting spatial perception information at different scales is as follows: ; in, Indicates the kernel size as Two-dimensional convolution operation, K Indicates the scale number; b. Spatial attention gating mechanism, introducing positional attention gating. To adjust the contribution of the spatial region to the updated state, its expression is: ; ; Final output This indicates the memory state of the fused spatial attention at the current time step.

[0012] In a second aspect, embodiments of the present invention provide a computer-readable storage medium comprising a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to execute the deep learning-based marine environmental assessment and prediction method for *Ulva prolifera* green tides, as described in the first aspect or any possible implementation thereof.

[0013] Thirdly, embodiments of the present invention provide an electronic device, including: one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the device, cause the device to perform the deep learning-based marine environmental assessment and prediction method for *Ulva prolifera* green tides as described in the first aspect or any possible implementation of the first aspect.

[0014] The technical solution provided by this invention includes collecting multi-source marine environmental data and preprocessing it; constructing spatial feature engineering and temporal features; establishing a suitability scoring model based on growth cycle and combining it with sea area division for suitability labeling; constructing a bidirectional spatial convolutional GRU prediction model and training and optimizing it to assess and predict the future marine environment, and outputting the prediction results. This method improves the scientificity and foresight of green tide management and control by constructing a bidirectional spatial convolutional GRU prediction model. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic diagram of a deep learning-based marine environmental assessment and prediction method for Ulva prolifera green tides provided in an embodiment of the present invention; Figure 2a A schematic diagram of geographic coordinate features provided in an embodiment of the present invention; Figure 2b A schematic diagram illustrating the distance characteristics from the center of the Yellow Sea provided in an embodiment of the present invention; Figure 3 The annual growth cycle adaptive change curve of *Ulva prolifera* provided in the embodiments of the present invention; Figure 4 A time distribution diagram of the annual growth stages of *Ulva prolifera* provided in an embodiment of the present invention; Figure 5a The temperature change trend provided in the embodiments of the present invention; Figure 5b The dissolved oxygen variation trend provided in the embodiments of the present invention; Figure 5c The trend of nutrient salt variation provided in the embodiments of the present invention; Figure 5d The phosphorus variation trend provided in the embodiments of the present invention; Figure 5e The nitrogen variation trend provided in the embodiments of the present invention; Figure 6 An architecture diagram of the bidirectional spatial convolutional GRU prediction model provided in an embodiment of the present invention; Figure 7 A schematic diagram of a basic unit of a spatial convolutional GRU network provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the spatial gridding of the suitability of *Ulva prolifera* provided in an embodiment of the present invention; Figure 9 A schematic diagram illustrating the results of the suitability assessment for the growth of *Ulva prolifera* provided in an embodiment of the present invention; Figure 10 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention are also intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0021] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0022] This invention provides a deep learning-based method for assessing and predicting marine environmental conditions related to *Ulva prolifera* green tides, such as... Figure 1 As shown, the method includes: Step 1: Collect multi-source marine environmental data and perform preprocessing.

[0023] In this embodiment of the invention, step 1 includes: Multi-source marine environmental data includes sea surface temperature data, marine optical data, nutrient concentration data (e.g., nitrates, phosphates), dissolved oxygen concentration, seawater salinity data, and the distribution area of ​​Ulva lactuca blooms in the ocean; among which, marine optical data includes photosynthetically active radiation and water transparency. Preprocessing includes outlier detection and handling, employing... The rules include: removing extreme values; filling missing values ​​using bilinear interpolation; normalizing data by normalizing all numerical variables to [0,1]; aligning time to a daily or monthly scale; and spatial interpolation using bilinear interpolation to map the data to a regular grid.

[0024] Step 2: Based on Step 1, construct spatial feature engineering and temporal features.

[0025] In this embodiment of the invention, step 2 includes: First, extract multi-dimensional spatial geographic features, including: geographic coordinates (latitude) ,longitude ),like Figure 2a As shown; distance from the center of the Yellow Sea ,like Figure 2b As shown; the shortest distance to the coastline Seasonal codes: Winter = 0, Spring = 1, Summer = 2, Autumn = 3; Let the latitude of the location point be... Longitude is The spatial features include: a. Geographic coordinates: (Latitude= Longitude = ); b. Based on the distance from the center of the Yellow Sea: The coordinates of the center of the Yellow Sea are set as follows: The spherical distance between two points is calculated using the Haversine semi-versine function, and its expression is: ; Where R is the Earth's radius (approximately 6371 km). c. Shortest distance from the coastline Its expression is: ; in, C represents the set of points on the coastline. The spherical distance; minimum value The nearest coastline; To reflect the impact of seasonal changes during the growth of *Ulva prolifera*, seasonal coding features were used as input variables for the time dimension. As shown in Table 1, the seasonal codes are winter=0, spring=1, summer=2, and autumn=3. The time features discretize the four seasons into four numerical categories so that the model can identify the regulation of *Ulva prolifera* growth by different seasons. Its mathematical expression is as follows: Let m be the month of the current time, then the season code is... Determined by the following function: .

[0026] After feature extraction is completed, feature evaluation is performed, which includes: a. Calculate the correlation coefficients between each characteristic and the suitability score of *Ulva prolifera*; b. Use principal component analysis (PCA) or mutual information (MI) methods to assess the explanatory power of features for the target variable; c. Sort features by importance and filter those with lower contribution to ensure that the input features retain the main environmental driving factors while reducing redundancy and noise.

[0027] Table 1 Seasonal Codes .

[0028] In embodiments of the present invention, such as Figure 1 As shown, the time series data combines time series information of multiple environmental factors and inputs it; the evaluation is an assessment of the presence or absence of *Ulva prolifera*.

[0029] Step 3: Based on Step 2, establish a suitability scoring model based on the growth cycle, and combine it with sea area division to perform suitability labeling.

[0030] In this embodiment of the invention, applicability is marked by combining sea area division, and the environmental conditions of the research sea area are divided into two categories: suitable and unsuitable.

[0031] In this embodiment of the invention, step 3 includes: To effectively improve the biological accuracy of *Ulva prolifera* growth prediction, a suitability scoring model based on the growth cycle was established. The suitability scoring model is divided into 12 growth stages, using months as the unit, and combines a basic suitability score with a temperature response coefficient to construct a suitability scoring mechanism that reflects the entire year's growth process. The growth of *Ulva prolifera* exhibits significant seasonality, undergoing a complete cycle of dormancy, germination, growth, flourishing, and decline throughout the year, depending on environmental conditions (temperature, light, nutrients, etc.). Therefore, this invention proposes a piecewise function divided by month to provide differentiated basic growth suitability and potential for different growth stages.

[0032] Define suitability function Where m represents the month, ranging from 1 to 12; T represents the average sea surface temperature for the month; L represents the average solar radiation intensity for the month; N represents the average nitrate concentration for the month; and P represents the average phosphate concentration for the month. Monthly parameters include a base suitability score. Temperature Factor Light Factor Nutrition Factor If the environmental mean vector includes temperature, light, nitrogen, and phosphorus, then the temperature vector... Lighting vector Nitrogen salt vector ; Phosphate vector ; Suitability Function for: ; The suitability function output value is used for suitability labeling, when Greater than or equal to the threshold If the condition is met, mark it as appropriate; otherwise, mark it as inappropriate.

[0033] The above parameters The value is set based on the measured data, as shown in Table 2.

[0034] Table 2. Values ​​of each factor at different times .

[0035] To improve the temporal modeling capability of *Ulva prolifera* growth suitability, this invention constructs a continuous suitability scoring model covering 12 months based on its life cycle pattern, such as... Figure 3 and Figure 4As shown, June is the peak period, with a suitability score of 0.74; December is the dormant period, with a suitability score of 0.02. This model, with biological characteristics at its core, dynamically integrates temperature factors to adjust the monthly score, thereby enhancing its sensitivity to seasonal changes and environmental responses. Furthermore, this method has good scalability and can be used in conjunction with time series models to achieve continuous prediction and spatiotemporal modeling of the *Ulva prolifera* growth process, further improving the ecological consistency and accuracy of the predictions.

[0036] The method of this invention exhibits significant advantages in ecological modeling, such as... Figures 5a to 5e As shown, on the one hand, by introducing biological driving mechanisms, the model can more accurately reflect the changing patterns of Ulva growth intensity at different times throughout the year, thus improving the ecological rationality of the prediction results; on the other hand, this method has good adaptability and can be flexibly embedded in a multi-factor modeling framework to jointly construct an ecological-environment coupled prediction system with environmental variables, providing more scientific decision support for green tide prevention and ecological early warning.

[0037] The suitability scoring model proposed in this invention can not only be used independently for assessing the ecological suitability of *Ulva prolifera*, but also serve as prior input for deep learning models. By incorporating the suitability score into the subsequent bidirectional spatial convolutional GRU network, the model can be provided with time-seasonal constraints that conform to biological mechanisms, avoiding ecologically unreasonable predictions that may result from the model relying solely on data fitting.

[0038] Step 4: Using the method from Step 3, construct a bidirectional spatial convolutional GRU prediction model, train and optimize it to assess and predict the future marine environment, and output the prediction results.

[0039] In this embodiment of the invention, step 4 includes: This invention designs a bidirectional spatial convolutional GRU (BSC-GRU) prediction model, the structure of which is as follows: Figure 6 As shown, this structure not only integrates multi-source environmental and spatial features, but also embeds the obtained biological suitability score as a prior feature into the model input. Through multi-scale convolution and spatial attention mechanisms, BSC-GRU can maintain the consistency of prediction results in response to the life cycle patterns of Ulva prolifera while learning complex spatiotemporal dependencies.

[0040] The BSC-GRU architecture of this invention integrates multiple functional modules to enhance spatiotemporal modeling capabilities. First, a feature fusion layer is constructed to integrate heterogeneous inputs. The feature fusion layer includes environmental features (e.g., sea surface temperature, nitrate, phosphate, salinity, dissolved oxygen, and light intensity), spatial features (e.g., latitude, longitude, and distance from the coast), and suitability scores based on a suitability scoring model. These features are vectorized and dimension-unified using a multilayer perceptron (MLP) to ensure the integration of heterogeneous information, i.e., feature fusion. Subsequently, by embedding multi-scale convolutional layers to capture spatial dependencies at different scales, convolutional kernels of sizes 3×3, 5×5, and 7×7 are applied in parallel to extract local fine-grained variations and spatial trends, and their outputs are concatenated to form a multi-scale representation. This design enables the model to simultaneously consider small-scale local dynamics and large-scale background processes, thereby improving its adaptability to complex marine environments.

[0041] The core component of the bidirectional spatial convolutional GRU prediction model is the bidirectional spatial convolutional GRU network. In this network, time series data are processed simultaneously by forward and backward sequences. The forward sequence encodes historical dependencies, while the backward sequence integrates future information. The outputs from both sequences are concatenated to capture the complete temporal dependencies. Figure 6 The input sequence includes a time series and an assessment of the presence of seaweed; convolution operations are used to replace the fully connected layers in the bidirectional spatial convolutional GRU network to preserve spatial structure information during gating and state updates; To further enhance spatial differentiation, a spatial attention mechanism is introduced. Attention weights are applied to the output of a bidirectional spatial convolutional GRU network to generate a spatial matrix. This matrix adaptively emphasizes key regions, such as high-risk nearshore areas and regions of abrupt environmental change. By redistributing attention across spatial domains, this mechanism improves the predicted ecological relevance and robustness. Figure 6 As shown, the input is a C×H×W original feature map, where C is the number of channels (depth of the original feature map), H is the height (number of rows of the original feature map), and W is the width (number of columns of the original feature map). Feature mapping represents operations (such as convolution, pooling, or linear transformation) on the input original feature map to obtain intermediate features required for spatial attention calculation. Weight calculation represents calculating attention weights in the spatial dimension, that is, calculating a weight for each spatial location (h, w) to represent the importance of that location in the overall features. The attention weights are represented by grids representing different spatial locations, and different colors representing different weights; darker colored grids represent areas with higher importance and greater weights, while lighter colored areas represent areas with lower importance and smaller weights. Weighted fusion represents performing element-wise weighted operations on the input original feature map and the calculated attention weights, and then fusing them to obtain a new output feature map C×H×W. The output image size specification C×H×W means that the output feature map has the same size as the original input feature map, but after the spatial attention mechanism, the feature value at each position has been adjusted according to the attention weight.

[0042] Finally, the output layer features a dual-function head: a segmentation head and a prediction head. The segmentation head uses convolutional layers to generate a spatial distribution map for visualizing suitability distribution. The prediction head integrates convolutional layers, ReLU layers, and fully connected layers to generate time-series predictions and classifies the future marine environment as suitable or unsuitable. These output designs collectively provide spatially defined and temporally continuous predictions, offering comprehensive support for ecological monitoring and management. Figure 6 As shown, the segmentation head uses a 3×3 convolutional layer with a kernel size of 3 to extract spatial features from the input features, capturing local spatial information. The ReLU activation function provides a non-linear mapping, setting negative values ​​to 0 and retaining positive values ​​to increase the model's expressive power. A 1×1 convolutional layer with a kernel size of 1 performs feature mapping, mapping the convolution result to the category dimension required for the segmentation task. The prediction head uses adaptive pooling to unify input feature maps of different sizes into a fixed-size output, ensuring consistent input dimensions for subsequent fully connected layers. The flattening layer flattens multi-dimensional feature maps into one-dimensional vectors for easier input to the fully connected layer. The linear layer (appearing for the first time) is the fully connected layer, mapping the input features to a new feature space. The ReLU activation function provides a non-linear mapping, setting negative values ​​to 0 and retaining positive values ​​to increase the model's expressive power. The linear layer (appearing for the second time) outputs the prediction result, specifically the time series regression value. The prediction result is judged as appropriate or inappropriate.

[0043] In embodiments of the present invention, such as Figure 7 As shown, the core update of the basic unit of the bidirectional spatial convolutional GRU network is expressed as follows: ; in, Indicates an update to the door. This indicates that the door is being reset. This represents a candidate hidden state. Indicates the final hidden state. Indicates time t The input feature vector (containing rasterized environmental elements such as SST, salinity, etc.) This represents the hidden state at the previous time step. W and U This represents the weight parameters that can be trained. b Indicates the bias term. This represents element-wise multiplication. This represents the activation function Sigmoid; To further incorporate spatial structure information, the bidirectional spatial convolutional GRU network is improved in two aspects: a. Multi-scale Spatial Convolutional Embedding: This method uses multi-scale convolutions to embed input features. The expression for extracting spatial perception information at different scales is as follows: ; in, Indicates the kernel size as Two-dimensional convolution operation, K Indicates the scale number; b. Spatial Attention Gating (SAG) mechanism, introducing a positional attention gate. To adjust the contribution of the spatial region to the updated state, its expression is: ; ; Final output This indicates the memory state of the fused spatial attention at the current time step.

[0044] This improved structure significantly enhances the model's ability to model regional heterogeneity and temporal variability. In practical ocean multivariate prediction tasks, compared to the traditional GRU, the BSC-GRU can better measure the root mean square error (RMSE) and... Significant improvements were achieved in all areas, especially in complex sea areas (such as the Yellow Sea coast).

[0045] In the model training and optimization process of this invention, the Mean Squared Error (MSE) is first selected as the loss function to measure the difference between the model's predicted values ​​and the actual observed values. MSE squares the prediction bias, assigning higher penalty weights to larger errors, thereby prompting the model to converge to the true values ​​more quickly during training and exhibiting good performance in terms of numerical stability and optimization controllability.

[0046] Regarding parameter updates, this invention preferably employs the Adam adaptive optimization algorithm, which combines the advantages of momentum gradient descent and RMSProp. It achieves adaptive learning rate adjustment of parameters through dynamic estimation of the first and second moments, effectively improving convergence speed and reducing oscillations in non-convex optimization spaces. To further enhance training stability and generalization performance, this invention introduces a learning rate scheduling strategy. The learning rate is dynamically adjusted based on the validation set performance. When the validation set error does not improve within a preset number of epochs, the learning rate is automatically reduced, prompting the model to perform more refined parameter searches in later stages. Simultaneously, a gradient pruning mechanism is introduced during backpropagation to constrain the gradient norm, preventing numerical instability caused by gradient explosion.

[0047] Furthermore, this invention introduces an early stopping mechanism, which automatically terminates training and retains the optimal model parameters when the validation set performance does not significantly improve within several consecutive training cycles. This avoids the risk of overfitting in the later stages of training and significantly reduces the waste of computational resources. Through the synergistic effect of the above-mentioned loss function design, optimization algorithm selection, learning rate scheduling, gradient pruning, and early stopping mechanism, the model constructed by this invention exhibits significant advantages in convergence speed, prediction accuracy, and generalization ability, and can meet the high accuracy and high stability requirements of complex spatiotemporal prediction tasks.

[0048] In the grid prediction stage, biological suitability factors and deep learning prediction results work together. The former serves as a priori modulator, while the latter represents the spatiotemporal dynamic learning outcome. Their fusion further enhances the ecological rationality and accuracy of the prediction. First, a spatial gridding method is used to systematically divide the target sea area to achieve high-precision, full-coverage spatial prediction. Specifically, the target sea area is divided into regular grids according to latitude and longitude coordinates, such as... Figure 8 As shown, the spatial resolution of each grid cell can be flexibly set according to application requirements, such as 0.01°×0.01° or higher, in order to achieve a balance between computational cost and prediction accuracy.

[0049] In the prediction process, spatial features (including geographical location, sea depth, distance from shore, and lighting conditions) and temporal features (including seasonal encoding and historical environmental factor trends) are first extracted for each grid cell. These features are then input into a trained bidirectional spatial convolutional GRU prediction model to obtain the predicted value for the next time step for the corresponding grid cell. This method ensures the spatial continuity and consistency of the prediction results while taking into account the influence of time series trends and local environmental differences.

[0050] To ensure the forecasting capability across long time spans, this invention supports the introduction of a multi-step rolling forecasting mechanism in spatial grid forecasting. That is, after completing the first forecasting step, the forecast result is used as one of the inputs for the next step, iterating until the required time length is reached, thus achieving full-cycle forecasting from short-term (1–3 days) to medium- to long-term (weeks or even months). During this process, the model can simultaneously capture the dynamic processes of *Ulva prolifera*, such as spatial migration, local outbreaks, and decline, meeting various practical application needs such as marine ecological monitoring, emergency response, and resource allocation.

[0051] Furthermore, the gridded prediction results can generate high-resolution spatial distribution maps, providing decision-making departments with intuitive and actionable spatial information support. This method significantly enhances the model's applicability and visualization value in practical ocean management while maintaining prediction accuracy.

[0052] This invention provides results visualization and analysis, presenting the model's predictions in an intuitive and easy-to-understand format. This facilitates researchers, management departments, and related industries in quickly obtaining key information and supporting decision-making. The system supports multiple visualization output methods and incorporates a statistical analysis module for result interpretation and risk warning, such as... Figure 9 As shown.

[0053] The bi-spatial convGRU (BSC-GRU) constructed in this invention combines multi-scale spatial convolution with spatial attention gating mechanisms to achieve fusion modeling of environmental features, spatial features, and biological suitability factors, thereby enabling high-precision spatiotemporal prediction of the growth process of *Ulva prolifera*. This method not only overcomes the shortcomings of traditional models in terms of spatial dependence, temporal coupling, scale awareness, and regional differentiation modeling, but also introduces biological prior constraints to ensure the ecological rationality of the prediction results. It accurately reflects the biological characteristics and seasonal variation patterns of *Ulva prolifera*, effectively integrates multi-source marine environmental data and spatial geographic information, achieves continuous prediction capabilities from short-term to long-term, and provides visualized prediction results and risk assessments to assist in prevention and control decisions.

[0054] Compared with existing technologies, this invention has the following advantages: it enhances the ability to capture the dynamics of the life cycle of Ulva prolifera through bidirectional temporal modeling; it introduces a multi-scale convolutional structure to effectively take into account both local small-scale changes and large-scale trends; it utilizes a spatial attention mechanism to highlight the prediction effect in key areas and improve the model's generalization ability; it combines a biological suitability model to ensure that the prediction results are ecologically reasonable; and it supports continuous prediction from the short term to the long term, providing a scientific basis for the prevention and control of green tides.

[0055] The technical solution provided by this invention includes collecting multi-source marine environmental data and preprocessing it; constructing spatial feature engineering and temporal features; establishing a suitability scoring model based on growth cycle and combining it with sea area division for suitability labeling; constructing a bidirectional spatial convolutional GRU prediction model and training and optimizing it to assess and predict the future marine environment, and outputting the prediction results. This method improves the scientificity and foresight of green tide management and control by constructing a bidirectional spatial convolutional GRU prediction model.

[0056] The various steps in the embodiments of the present invention can be performed by an electronic device. This electronic device includes, but is not limited to, tablet computers, portable PCs, and desktop computers.

[0057] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is running, it controls the electronic device containing the computer-readable storage medium to execute the above-described embodiment of the deep learning-based method for assessing and predicting marine environmental conditions related to seaweed green tides.

[0058] Figure 10 A schematic diagram of an electronic device provided in an embodiment of the present invention, such as... Figure 10 As shown, the electronic device 21 includes a processor 211, a memory 212, and a computer program 213 stored in the memory 212 and executable on the processor 211. When the computer program 213 is executed by the processor 211, it implements the deep learning-based marine environmental assessment and prediction method for Ulva prolifera green tide in the embodiment. To avoid repetition, it will not be described in detail here.

[0059] Electronic device 21 includes, but is not limited to, processor 211 and memory 212. Those skilled in the art will understand that... Figure 10 This is merely an example of electronic device 21 and does not constitute a limitation on electronic device 21. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0060] The processor 211 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0061] The memory 212 can be an internal storage unit of the electronic device 21, such as a hard disk or RAM of the electronic device 21. The memory 212 can also be an external storage device of the electronic device 21, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or FlashCard equipped on the electronic device 21. Furthermore, the memory 212 can include both internal and external storage units of the electronic device 21. The memory 212 is used to store computer programs and other programs and data required by network devices. The memory 212 can also be used to temporarily store data that has been output or will be output.

[0062] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A deep learning-based method for assessing and predicting marine environmental conditions related to *Ulva prolifera* green tides, characterized in that... The method includes: Step 1: Collect multi-source marine environmental data and perform preprocessing; Step 2: Based on Step 1, construct spatial feature engineering and temporal features; Step 3: Based on Step 2, establish a suitability scoring model based on the growth cycle, and combine it with sea area division to perform suitability labeling; Step 4: Using the method from Step 3, construct a bidirectional spatial convolutional GRU prediction model, train and optimize it to assess and predict the future marine environment, and output the prediction results.

2. The method according to claim 1, characterized in that, Step 1 includes: Multi-source marine environmental data includes sea surface temperature data, marine optical data, nutrient concentration data, dissolved oxygen concentration, seawater salinity data, and the distribution area of ​​Ulva lactuca blooms in the ocean; among which, marine optical data includes photosynthetically active radiation and water transparency. Preprocessing includes outlier detection and handling, employing... The rules include: removing extreme values; filling missing values ​​using bilinear interpolation; normalizing data by normalizing all numerical variables to [0,1]; aligning time to a daily or monthly scale; and spatial interpolation using bilinear interpolation to map the data to a regular grid.

3. The method according to claim 2, characterized in that, Step 2 includes: First, extract multidimensional spatial geographic features, including: geographic coordinates (latitude) ,longitude ); Distance from the center of the Yellow Sea The shortest distance from the coastline Seasonal codes: Winter = 0, Spring = 1, Summer = 2, Autumn = 3; Let the latitude of the location point be... Longitude is The spatial features include: a. Geographic coordinates: (Latitude= Longitude = ); b. Based on the distance from the center of the Yellow Sea: The coordinates of the center of the Yellow Sea are set as follows: The spherical distance between two points is calculated using the Haversine semi-versine function, and its expression is: ； Where R is the Earth's radius; c. Shortest distance from the coastline Its expression is: ； in, C represents the set of points on the coastline. The spherical distance; minimum value The nearest coastline; The temporal feature uses seasonal coding features as the input variable for the time dimension; the seasonal coding is winter=0, spring=1, summer=2, autumn=3; the temporal feature discretizes the four seasons into four numerical categories so that the model can identify the regulation of Ulva prolifera growth by different seasons. Its mathematical expression is as follows: Let m be the month of the current time, then the season code is... Determined by the following function: 。 4. The method according to claim 3, characterized in that, Step 3 includes: The suitability scoring model is divided into 12 growth stages based on months, and combines the basic suitability score with the temperature response coefficient to construct a suitability scoring mechanism that reflects the growth process throughout the year. Define suitability function Where m represents the month, ranging from 1 to 12; T represents the average sea surface temperature for the month; L represents the average solar radiation intensity for the month; N represents the average nitrate concentration for the month; and P represents the average phosphate concentration for the month. Monthly parameters include basic suitability score. Temperature factor Light factor Nutrient factors If the environmental mean vector includes temperature, light, nitrogen, and phosphorus, then the temperature vector... Lighting vector Nitrogen salt vector ; Phosphate vector ; Suitability Function for: ； The suitability function output value is used for suitability labeling, when Greater than or equal to the threshold If the condition is met, mark it as appropriate; otherwise, mark it as inappropriate.

5. The method according to claim 4, characterized in that, Step 4 includes: First, a feature fusion layer is constructed to integrate heterogeneous inputs. The feature fusion layer includes environmental features, spatial features, and suitability scores based on a suitability scoring model. The above features are vectorized and dimension-unified through a multilayer perceptron (MLP) to ensure the integration of heterogeneous information, i.e., feature fusion. Subsequently, by embedding multi-scale convolutional layers to capture spatial dependencies at different scales, convolutional kernels of sizes 3×3, 5×5 and 7×7 are applied in parallel to extract local fine-grained changes and spatial trends, and their outputs are concatenated to form a multi-scale representation. The core component of the bidirectional spatial convolutional GRU prediction model is the bidirectional spatial convolutional GRU network. In the bidirectional spatial convolutional GRU network, the time series is processed simultaneously by forward and backward sequences. The forward sequence encodes historical dependencies, while the backward sequence integrates future information. The outputs of the two sequences are concatenated to capture the complete temporal dependencies. Convolutional operations are used to replace the fully connected layers in the bidirectional spatial convolutional GRU network to preserve spatial structure information during gating and state updates. To further enhance spatial differentiation, a spatial attention mechanism is introduced; attention weights are applied to the output of the bidirectional spatial convolutional GRU network to generate a spatial matrix. Finally, the output layer is designed with a dual-function head: a segmentation head and a prediction head. The segmentation head uses convolutional layers to generate a spatial distribution map for visualizing the suitability distribution. The prediction head integrates convolutional layers, ReLU layers, and fully connected layers to generate time series predictions.

6. The method according to claim 5, characterized in that, The core update of the basic unit of the bidirectional spatial convolutional GRU network is expressed as follows: ； in, Indicates an update to the door. This indicates that the door is being reset. This represents a candidate hidden state. Indicates the final hidden state. Indicates time t The input feature vector, This represents the hidden state at the previous time step. W and U This represents the weight parameters that can be trained. b Indicates the bias term. This represents element-wise multiplication. This represents the activation function Sigmoid; To further incorporate spatial structure information, the bidirectional spatial convolutional GRU network is improved in two aspects: a) Multi-scale spatial convolutional embedding: This method embeds input features using multi-scale convolution. The expression for extracting spatial perception information at different scales is as follows: ； in, Indicates the kernel size as Two-dimensional convolution operation, K Indicates the scale number; b. Spatial attention gating mechanism, introducing positional attention gating. To adjust the contribution of the spatial region to the updated state, its expression is: ； ； Final output This indicates the memory state of the fused spatial attention at the current time step.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the deep learning-based marine environmental assessment and prediction method for *Ulva prolifera* green tides as described in any one of claims 1 to 6.

8. An electronic device, characterized in that, include: One or more processors; Memory; And one or more computer programs, wherein the one or more computer programs are stored in the memory, the one or more computer programs including instructions that, when executed by the device, cause the device to perform the deep learning-based marine environmental assessment and prediction method for Ulva prolifera green tides as described in any one of claims 1 to 6.

Images