Chlorophyll value determination method and apparatus, computer device, and storage medium

Abstract

The application provides a chlorophyll value determination method and device, computer equipment and storage medium, applied to the field of marine numerical simulation, comprising obtaining the parameter data of a target sea area corresponding to a first time period; taking the parameter data of the target sea area corresponding to the first time period as input, determining the chlorophyll concentration simulation data of the target sea area corresponding to a second time period by using a marine biogeochemical coupling model; using the trained intelligent network model, based on the chlorophyll concentration observation data of the target sea area in the second time period, correcting the chlorophyll concentration simulation data to generate corrected chlorophyll concentration simulation data; performing vertical interpolation processing on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data in the second time period. The application realizes the optimization processing of the chlorophyll concentration simulation data by using the intelligent network model and the chlorophyll concentration observation data, and improves the optimization efficiency and optimization accuracy.

Description

Technical Field

[0001] This application relates to the field of marine numerical simulation technology, and more specifically, to a method, apparatus, computer equipment, and storage medium for determining chlorophyll values. Background Technology

[0002] Marine physical-biogeochemical coupled numerical models (referred to as marine biogeochemical models) are one of the main methods for understanding and analyzing the spatiotemporal distribution characteristics of marine chlorophyll. The core of these models is to construct a cycle process model that includes elements such as nutrients, phytoplankton, zooplankton, and detritus, and incorporate a series of functions such as feeding, predation, metabolism, decomposition, death, and sedimentation. Combined with empirical and semi-empirical parameters, these models can quantitatively simulate relevant elements such as nutrient cycling, phytoplankton growth and reproduction, dissolved oxygen, and seawater carbon dioxide content in the ocean. Ultimately, they can achieve a quantitative output of chlorophyll concentration in the ocean surface, which is an important means to understand the spatiotemporal variation of chlorophyll concentration and explore the mechanism of marine biogeochemical cycles.

[0003] To improve the accuracy of chlorophyll numerical simulations, the optimization of the aforementioned marine biogeochemical models mainly includes two types of improvement methods: First, adjusting and optimizing key parameters in the marine biogeochemical cycle process, such as the phytoplankton growth half-saturation constant, the temperature-limiting parameters for biological growth and mortality, and light adaptation parameters; second, using data assimilation methods, which involve fusing observational data and iteratively correcting the initial values ​​of the model to improve the simulation results and enhance simulation accuracy. Typical methods include ensemble Kalman filter (EnKF), three-dimensional variational (3D-Var), and four-dimensional variational (4D-Var).

[0004] In the aforementioned process, the difficulty in setting the core parameters for numerical simulation leads to significant challenges in improving simulation accuracy. Specifically, the coupled marine physical-biogeochemical numerical models involve complex functional formulas and rely on dozens, or even hundreds, of empirical and semi-empirical parameters closely related to the characteristics of the marine ecological environment. However, there are significant differences between the marine ecological environments of ocean basins and nearshore areas. Related technologies generally employ single-parameter settings, failing to fully consider the environmental specificities of different sea areas. This results in limitations in the matching of parameters with the actual marine environment, directly affecting the accuracy of chlorophyll concentration simulation results.

[0005] Meanwhile, the optimization methods in related technologies still have many shortcomings in terms of systematically and holistically improving simulation accuracy. For example, traditional parameter optimization only adjusts individual key parameters and lacks a systematic consideration of the interaction between many parameters. The optimization dimension is singular and cannot achieve the global optimum of parameter combination. Data assimilation methods (such as EnKF, 3D-Var, 4D-Var, etc.) have problems of high computational complexity and low efficiency when dealing with high-dimensional state variables. They are also highly dependent on observation data, and the assimilation effect is limited in scenarios where the spatiotemporal coverage of observation data is insufficient. Summary of the Invention

[0006] In view of this, this application provides a method, apparatus, computer device and storage medium for determining chlorophyll values.

[0007] Specifically, this application is implemented through the following technical solution:

[0008] In a first aspect, embodiments of this application provide a method for determining chlorophyll values, including: Obtain parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data; Using the parameter data of the first time period corresponding to the target sea area as input, the simulated chlorophyll concentration data of the second time period corresponding to the target sea area is determined by the marine biogeochemical coupling model; wherein the second time period is later than the first time period; Using the trained intelligent network model, the chlorophyll concentration simulation data is corrected based on the chlorophyll concentration observation data of the second time period in the target sea area, and the corrected chlorophyll concentration simulation data is generated. Vertical interpolation is performed on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period.

[0009] In one optional implementation, the intelligent network model is trained according to the following steps: Acquire training samples, wherein the training samples include chlorophyll concentration simulation samples and chlorophyll concentration observation samples at multiple sampling times corresponding to the sample area. The chlorophyll concentration simulation samples are determined using the marine biogeochemical coupling model, and the chlorophyll concentration observation samples are chlorophyll concentration calibration data of the sample area at the sampling time determined based on the observation dataset. The training samples are input into the neural network model to be trained for processing, and chlorophyll prediction data is output; wherein the neural network model to be trained is a U-shaped symmetric network structure. Based on the chlorophyll prediction data and the chlorophyll concentration observation samples, the loss value is determined; and the network parameters of the neural network model to be trained are adjusted based on the loss value to obtain the adjusted neural network model. The adjusted neural network model is used as a new neural network model to be trained. The process returns to the step of inputting the training samples into the neural network model to be trained for processing, until the training cutoff condition is met, and the intelligent network model is obtained.

[0010] In an optional embodiment, the method further includes: performing data preprocessing on the simulated chlorophyll concentration samples and the observed chlorophyll concentration samples in the training samples to generate processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples, wherein the data preprocessing includes logarithmic normalization, spatial grid alignment, and default value labeling; the processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples are used to train the neural network model to be trained.

[0011] In one optional implementation, the intelligent network model includes: an encoder, a bottleneck layer, and a decoder, wherein the bottleneck layer is connected to the encoder and the decoder respectively, and the encoder and the decoder have a symmetrical structure; The encoder includes n feature encoding modules connected in sequence. The first feature encoding module is used to perform feature encoding processing on the input data and input the generated first encoded feature data to the second feature encoding module. The ith feature encoding module is used to perform feature encoding processing on the (i-1)th encoded feature data, where i is a positive integer greater than 1 and less than or equal to n, and n is a positive integer greater than 1. The decoder includes n feature decoding modules connected in sequence. The first feature decoding module is used to decode the feature data output by the bottleneck layer and the nth encoded feature data output by the nth feature encoding module to generate the first decoded feature data. The jth feature decoding module is used to decode the (j-1)th decoded feature data and the (n-j+1)th encoded feature data to generate the jth decoded feature data. Here, j is a positive integer greater than 1 and less than or equal to n.

[0012] In an optional implementation, the method further includes: determining a mask image corresponding to the sample region, wherein the region corresponding to the first pixel value in the mask image is an effective region where chlorophyll concentration exists, and the region corresponding to the second pixel value is an invalid region where chlorophyll concentration does not exist; The step of determining the loss value based on the chlorophyll prediction data and the chlorophyll concentration observation samples includes: determining the loss value based on the mask image, the chlorophyll prediction data, and the chlorophyll concentration observation samples.

[0013] In an optional implementation, the method further includes: taking the chlorophyll concentration distribution data of the second time period as input, and using a marine biogeochemical coupling model to determine the simulated chlorophyll concentration data of the next time period corresponding to the target sea area.

[0014] In one optional embodiment, the environmental parameters include at least one of the following: nutrient concentration, seawater temperature, salinity, and topographic data; the boundary condition parameters include at least one of the following: atmospheric wind field heat flux on the upper surface, offshore tides, and runoff input.

[0015] Secondly, embodiments of this application also provide a chlorophyll value determination device, comprising: The acquisition module is used to acquire parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data; The determination module is used to determine the simulated chlorophyll concentration data of the target sea area in a second time period using parameter data of a first time period corresponding to the target sea area as input and a marine biogeochemical coupling model; wherein the second time period is later than the first time period. The first generation module is used to use a trained intelligent network model to correct the chlorophyll concentration simulation data based on the chlorophyll concentration observation data of the second time period in the target sea area, and generate corrected chlorophyll concentration simulation data. The second generation module is used to perform vertical interpolation processing on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period.

[0016] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for determining chlorophyll values.

[0017] This application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for determining chlorophyll values.

[0018] This application uses parameter data from the first time period corresponding to the target sea area as input. After determining the simulated chlorophyll concentration data for the second time period corresponding to the target sea area using a marine biogeochemical coupling model, it uses a trained intelligent network model to correct the simulated chlorophyll concentration data based on the observed chlorophyll concentration data of the second time period in the target sea area, generating corrected simulated chlorophyll concentration data. This achieves the optimization processing of simulated chlorophyll concentration data using an intelligent network model and observed chlorophyll concentration data, alleviating the problems of lack of systematicity and unstable optimization in related technologies that rely on manually adjusting individual parameters, as well as the high computational complexity and dependence on dense observation data in data assimilation methods, thus improving optimization efficiency and accuracy. Furthermore, subsequent vertical interpolation processing of the corrected simulated chlorophyll concentration data can generate chlorophyll concentration distribution data for the second time period with relatively accurate results. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for determining chlorophyll values ​​according to an exemplary embodiment of this application; Figure 2 This is a flowchart illustrating the intelligent network model in a chlorophyll value determination method according to an exemplary embodiment of this application. Figure 3 This is a schematic diagram of a chlorophyll value determination device shown in an exemplary embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in this application. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended foregoing.

[0021] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0023] To facilitate understanding of this embodiment, a detailed description of the chlorophyll value determination method disclosed in this application embodiment will be provided first. The execution subject of the chlorophyll value determination method provided in this application embodiment is generally a computer device with a certain computing capability. This computer device includes, for example, a terminal device, a server, or other processing devices. The terminal device can be a user equipment (UE), a mobile device, a user terminal, a computing device, etc. In some possible implementations, the chlorophyll value determination method can be implemented by a processor calling computer-readable instructions stored in memory.

[0024] See Figure 1 The diagram shows a flowchart of a method for determining chlorophyll values ​​provided in an embodiment of this application. The method includes steps S101 to S104, wherein: S101. Obtain parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data.

[0025] The parameter data for the target sea area includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data. The environmental parameters include, but are not limited to, nutrient concentration, seawater temperature, salinity, and topographic data. Boundary condition parameters include, but are not limited to, upper surface atmospheric wind heat flux, offshore tides, and runoff input. The current chlorophyll concentration distribution data can be the chlorophyll concentration distribution data for the target sea area within the first time period. This data can be simulated chlorophyll data determined using a marine biogeochemical coupling model, or it can be chlorophyll distribution data obtained by optimizing the simulated chlorophyll data output from the marine biogeochemical coupling model using an intelligent network model.

[0026] The target sea area can be any sea area to be detected, and the first time period can be any historical time period. This application does not make any specific restrictions on this.

[0027] S102. Using the parameter data of the first time period corresponding to the target sea area as input, the simulated chlorophyll concentration data of the second time period corresponding to the target sea area is determined by the marine biogeochemical coupling model; wherein the second time period is later than the first time period.

[0028] During implementation, the parameter data of the first time period corresponding to the target sea area are input into the marine biogeochemical coupling model (i.e., the marine biogeochemical coupling mode) for simulation processing, so that the marine biogeochemical coupling model outputs the simulated chlorophyll concentration data of the second time period corresponding to the target sea area, where the second time period is later than the first time period.

[0029] For example, when starting the simulation, the initial environmental parameters of the target sea area (such as initial nutrient concentration, seawater temperature and salinity, topographic data, etc.), boundary condition parameters (such as atmospheric wind field heat flux on the upper surface, offshore tides, runoff input, etc.), and initial chlorophyll concentration can be used as inputs to start the corresponding marine biogeochemical coupling mode, carry out synchronous numerical simulation of chlorophyll and multiple elements according to the daily time cycle, and output spatiotemporal distribution data of chlorophyll concentration (i.e., chlorophyll concentration simulation data), and can also output auxiliary simulation data such as nutrients, temperature, and flow field at the same time.

[0030] S103. Using the trained intelligent network model, based on the chlorophyll concentration observation data of the second time period in the target sea area, the simulated chlorophyll concentration data is corrected to generate corrected simulated chlorophyll concentration data.

[0031] After the marine biogeochemical coupling model outputs the simulated chlorophyll concentration data for the second time period corresponding to the target sea area, the trained intelligent network model can be used to correct the simulated chlorophyll concentration data.

[0032] During implementation, chlorophyll concentration observation data for the second time period in the target sea area can be acquired. This second time period's chlorophyll concentration observation data and simulated chlorophyll concentration data are then input into an intelligent network model for processing, generating corrected simulated chlorophyll concentration data. The chlorophyll concentration observation data can be obtained through satellite remote sensing data.

[0033] The chlorophyll concentration observation data in this application can be the observation results of the chlorophyll concentration in the target horizontal layer of the target sea area, and the chlorophyll concentration simulation data can also be the simulation results of the chlorophyll concentration in the target horizontal layer of the target sea area.

[0034] During implementation, to eliminate spatial differences between simulated and observed data, as well as the large range of chlorophyll concentration values, the simulated chlorophyll concentration data can be processed by spatial grid unification and log10 logarithmic standardization to obtain processed simulated chlorophyll concentration data; similarly, the observed chlorophyll concentration data can be processed by spatial grid unification and log10 logarithmic standardization to obtain processed observed chlorophyll concentration data. This allows the processed simulated and observed chlorophyll concentration data to be input into the intelligent network model to generate corrected simulated chlorophyll concentration data.

[0035] The intelligent network model is a trained deep learning model, and its network structure can be configured according to business requirements. This application uses a U-shaped U-Net neural network as the intelligent network model.

[0036] Optionally, firstly, the simulated chlorophyll concentration data output from the marine biogeochemical coupling model is subjected to log10 logarithmic normalization to obtain processed simulated chlorophyll concentration data; and the observed chlorophyll concentration data is subjected to spatial grid unification and log10 logarithmic normalization to obtain processed observed chlorophyll concentration data. Secondly, the processed simulated chlorophyll concentration data and the processed observed chlorophyll concentration data are input into a pre-trained dedicated U-Net intelligent network model. This U-Net intelligent network model extracts the systematic deviations and spatial correlation features between the simulated and observed chlorophyll concentration data through an encoder, and reconstructs and outputs a high-resolution chlorophyll concentration correction field through a decoder, completing the global systematic deviation correction of the original simulation results and obtaining the optimal chlorophyll correction result for the current time period, i.e., the corrected simulated chlorophyll concentration data. Compared to traditional data assimilation methods, which rely on high-density observations, have high computational complexity, and poor adaptability to sparse observation scenarios, this application relies on the strong spatial feature learning capabilities of U-Net to achieve efficient and accurate correction even under sparse observation conditions. The correction results have both spatial continuity and physical rationality, providing reliable data support for subsequent feedback iterations.

[0037] S104. Perform vertical interpolation on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period.

[0038] Since the corrected chlorophyll concentration simulation data is chlorophyll concentration data of a specific horizontal layer in the target sea area, vertical interpolation can also be performed on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period. This chlorophyll concentration distribution data reflects the chlorophyll concentration distribution in the vertical direction of the target sea area.

[0039] This application uses parameter data from the first time period corresponding to the target sea area as input. After determining the simulated chlorophyll concentration data for the second time period corresponding to the target sea area using a marine biogeochemical coupling model, it uses a trained intelligent network model to correct the simulated chlorophyll concentration data based on the observed chlorophyll concentration data of the second time period in the target sea area, generating corrected simulated chlorophyll concentration data. This achieves the optimization processing of simulated chlorophyll concentration data using an intelligent network model and observed chlorophyll concentration data, alleviating the problems of lack of systematicity and unstable optimization in related technologies that rely on manually adjusting individual parameters, as well as the high computational complexity and dependence on dense observation data in data assimilation methods, thus improving optimization efficiency and accuracy. Furthermore, subsequent vertical interpolation processing of the corrected simulated chlorophyll concentration data can generate chlorophyll concentration distribution data for the second time period with relatively accurate results.

[0040] Compared with the lack of systematicity in "manually adjusting individual parameters" in related technologies, and the limitations of "data assimilation methods (such as EnKF, 3D-Var, 4D-Var)" which have high computational complexity and rely on dense observation data, this application utilizes the powerful spatiotemporal feature extraction and nonlinear mapping capabilities of intelligent network models to achieve multi-element correlation and systematic deviation correction; and uses satellite remote sensing observation data (which may include sparse data such as cloud cover and inversion accuracy) to carry out global deviation correction, which significantly improves computational efficiency compared with traditional data assimilation methods, while the spatial resolution and accuracy of the correction results are better.

[0041] Optionally, the method further includes: using the chlorophyll concentration distribution data of the second time period as input, and using a marine biogeochemical coupling model to determine the simulated chlorophyll concentration data of the next time period corresponding to the target sea area.

[0042] After obtaining the chlorophyll concentration distribution data for the second time period, this application can also use the chlorophyll concentration distribution data for the second time period as input to determine the simulated chlorophyll concentration data for the next time period corresponding to the target sea area using a marine biogeochemical coupling model, thus forming a fully closed-loop optimization feedback mechanism of "simulation-intelligent correction-result feedback-iterative simulation".

[0043] Studies have found that related technologies lack optimization feedback mechanisms, resulting in a one-off simulation process that fails to achieve continuous optimization. This application, however, utilizes the aforementioned feedback mechanism to reuse the corrected optimal results (i.e., chlorophyll concentration distribution data from the second time period) in the next time cycle. This allows the simulation parameters of the marine biogeography model to gradually adapt to the target marine environment, continuously reducing simulation bias and improving the accuracy of chlorophyll concentration simulation.

[0044] It is evident that this application uses numerical simulation of a marine biogeochemical model to obtain basic data, and completes intelligent correction of chlorophyll simulation results through an intelligent network model. The corrected chlorophyll field is used as the initial field of the model for the next time period for feedback iteration, forming a fully closed-loop optimization mechanism. This achieves deep coupling between simulation and intelligent correction, continuously corrects systematic biases in the model, and overcomes the limitations of existing data intelligent optimization technology that is disconnected from simulation and cannot continuously improve accuracy, thus ensuring the accuracy of chlorophyll concentration simulation data.

[0045] In practice, the intelligent network model can be trained according to the following steps: Step a1: Obtain training samples, wherein the training samples include chlorophyll concentration simulation samples and chlorophyll concentration observation samples at multiple sampling times corresponding to the sample area. The chlorophyll concentration simulation samples are determined using the marine biogeochemical coupling model, and the chlorophyll concentration observation samples are chlorophyll concentration calibration data of the sample area at the sampling time determined based on the observation dataset. Step a2: Input the training samples into the neural network model to be trained for processing, and output chlorophyll prediction data; wherein the neural network model to be trained is a U-shaped symmetric network structure; Step a3: Determine the loss value based on the chlorophyll prediction data and the chlorophyll concentration observation samples; and adjust the network parameters of the neural network model to be trained based on the loss value to obtain the adjusted neural network model; Step a4: Using the adjusted neural network model as a new neural network model to be trained, return to the step of inputting the training samples into the neural network model to be trained for processing, until the training cutoff condition is met, and obtain the intelligent network model.

[0046] In practice, training samples can be constructed first. These training samples include simulated chlorophyll concentration samples and observed chlorophyll concentration samples at multiple sampling times corresponding to the sample area. The simulated chlorophyll concentration samples are training data determined using the marine biogeochemical coupling model. Specifically, the chlorophyll concentration data at multiple sampling times corresponding to the sample area output by the marine biogeochemical coupling model are obtained as simulated chlorophyll concentration samples. Optionally, the temporal resolution of the sample data can be set to 1 day, and the spatial resolution can be set to 1 / 24°.

[0047] Optionally, training samples can be constructed by acquiring spatiotemporal sequence datasets of multiple elements such as chlorophyll concentration, nutrients, seawater temperature, salinity, and flow field output from the marine biogeochemical coupling model. That is, the training samples include: simulated chlorophyll concentration samples, training data such as nutrients, seawater temperature, salinity, and flow field, as well as observed chlorophyll concentration samples.

[0048] Chlorophyll concentration observation samples can serve as calibration data for chlorophyll concentration at the sampling time in a sample area determined based on the observation dataset. For example, chlorophyll concentration observation samples can be obtained from the v6.0 version of the global satellite remote sensing chlorophyll a concentration L3-level gridded product publicly released by the European Space Agency (ESA) Ocean Colour Climate Change Initiative (OC-CCI). This product is a long-term global ocean color dataset fused from multiple satellite sources, with a temporal resolution of 1 day and a spatial resolution of 4 km. It features global coverage, good spatiotemporal consistency, and verified accuracy, and is widely used internationally. It can be used as a calibration benchmark, i.e., true value input, for ocean chlorophyll concentration simulation results.

[0049] The temporal and spatial resolutions of the training samples are merely illustrative examples. In practice, the temporal and spatial resolutions of the training data (i.e., simulated chlorophyll concentration samples) and calibration data (i.e., observed chlorophyll concentration samples) can be appropriately adjusted and matched according to the model's own grid resolution and data resolution.

[0050] The training samples are then input into the neural network model to be trained for processing, and chlorophyll prediction data is output; the neural network model to be trained is a U-shaped symmetric network structure.

[0051] Optionally, the method further includes: performing data preprocessing on the simulated chlorophyll concentration samples and the observed chlorophyll concentration samples in the training samples to generate processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples, wherein the data preprocessing includes logarithmic normalization, spatial grid alignment, and default value labeling; the processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples are used to train the neural network model to be trained.

[0052] Studies have found that there are differences in the original spatial grid division between simulated chlorophyll concentration samples and observed chlorophyll concentration samples. In order to alleviate the above differences, this application can perform spatial grid alignment processing on simulated chlorophyll concentration samples and observed chlorophyll concentration samples, that is, convert all input data into a standardized spatiotemporal grid format, unify the spatial benchmark for model training and accuracy verification, and avoid calculation errors caused by spatial grid mismatch.

[0053] To address the issues of large numerical ranges in chlorophyll concentration and the tendency for direct training with raw data to overly sensitive to high-value regions while neglecting subtle changes in low-value regions, this application performs log10 logarithmic standardization and default value labeling on both simulated and observed chlorophyll concentration samples. Specifically, positive chlorophyll concentration values ​​with physical meaning undergo a base-10 logarithmic transformation, while negative and zero values, lacking practical physical meaning, are directly labeled as default values ​​(NaN) and retained. Through this preprocessing, the logarithmically transformed data distribution more closely approximates a normal distribution, balancing the training weights of high and low value regions, significantly improving the model's feature learning efficiency and overall stability across the entire concentration range, and ensuring effective model training.

[0054] This application integrates data from marine biogeography model simulations and ESA OC-CCI v6.0 satellite remote sensing chlorophyll observation data, converting all data into a standardized spatiotemporal grid format to eliminate grid mismatch errors. It also performs log10 logarithmic standardization on chlorophyll concentration, marks negative and zero values ​​as default values, and balances the training weights in high and low value regions to provide a unified and stable input for the model.

[0055] Next, based on chlorophyll prediction data and chlorophyll concentration observation samples, a loss value is determined, where the loss function can be set according to requirements. Then, the network parameters of the neural network model to be trained can be adjusted based on the loss value to obtain an adjusted neural network model. This adjusted neural network model is then used as a new neural network model to be trained, and the process returns to the step of inputting training samples into the neural network model to be trained for processing, until the training cutoff condition is met. After multiple rounds of iterative training, an intelligent network model is obtained.

[0056] Optionally, the method further includes: determining a mask image corresponding to the sample region, wherein the region corresponding to the first pixel value in the mask image is a valid region where chlorophyll concentration exists, and the region corresponding to the second pixel value is an invalid region where chlorophyll concentration does not exist. For example, the pixel value of the valid region where chlorophyll concentration exists in the mask image is 1, and the pixel value of the invalid region where chlorophyll concentration does not exist is 0.

[0057] Then, the loss value can be determined based on the mask image, the chlorophyll prediction data, and the chlorophyll concentration observation samples. That is, the loss value can be calculated using only the pixel values ​​of the effective region through the mask image, without using the pixel values ​​of the invalid region, so as to eliminate the interference of missing values ​​on gradient updates, ensure the consistency between simulation results and remote sensing observations, and ensure the training accuracy of the neural network model.

[0058] For example, during training, an adaptive learning rate decay strategy can be employed. If the validation set loss does not decrease for several consecutive epochs (e.g., 5), the learning rate is automatically halved to achieve fine-tuning of the model. Simultaneously, gradient pruning is used to prevent gradient explosion. For example, the batch size for model training can be set to 4, with a total of 51 training epochs. After each iteration, the average loss of the training and validation sets is calculated synchronously. The model prediction accuracy can be quantitatively evaluated using two statistical metrics: root mean square error (RMSE) and bias, to verify the reliability of the results.

[0059] This application constructs a U-shaped symmetric deep learning model adapted to the spatial distribution characteristics of marine chlorophyll, consisting of a four-level downsampling encoder, a bottleneck layer, a symmetric decoder, and skip connections. It employs a chlorophyll concentration error loss function with effective region masking, calculating only effective pixel errors to eliminate interference from default values. Furthermore, it combines adaptive learning rate decay and gradient clipping strategies to complete model training, using simulated chlorophyll as input and remotely sensed chlorophyll as the learning target, thereby achieving high-precision and high-efficiency spatial correction of the simulation results.

[0060] This application, through multi-source data fusion preprocessing and the design of an effective region mask loss function for the UNet model, ensures consistency between chlorophyll simulation results and remote sensing observations while taking into account the regular constraints of marine environmental characteristics such as land-sea boundaries and coastlines, thus avoiding the problem of parameters deviating from reality and being unusable. The method proposed in this application can be flexibly adapted to different marine environmental characteristics, such as nearshore and open ocean, broadening the applicability of marine biogeographical models. It has good adaptability and applicability in chlorophyll numerical simulation applications in various complex marine environments.

[0061] Optionally, the intelligent network model includes an encoder, a bottleneck layer, and a decoder, wherein the bottleneck layer is connected to the encoder and the decoder respectively, and the encoder and the decoder have a symmetrical structure.

[0062] The encoder includes n feature encoding modules connected in sequence. The first feature encoding module is used to perform feature encoding processing on the input data and input the generated first encoded feature data into the second feature encoding module. The i-th feature encoding module is used to perform feature encoding processing on the (i-1)-th encoded feature data, where i is a positive integer greater than 1 and less than or equal to n, and n is a positive integer greater than 1.

[0063] The decoder includes n feature decoding modules connected in sequence. The first feature decoding module is used to decode the feature data output by the bottleneck layer and the nth encoded feature data output by the nth feature encoding module to generate the first decoded feature data. The jth feature decoding module is used to decode the (j-1)th decoded feature data and the (n-j+1)th encoded feature data to generate the jth decoded feature data. Here, j is a positive integer greater than 1 and less than or equal to n.

[0064] This application utilizes the U-Net deep learning model to construct an intelligent network model adapted to the spatial distribution characteristics of marine chlorophyll. This model establishes a precise mapping relationship using high-resolution chlorophyll simulation data as input and satellite remote sensing observation data of chlorophyll concentration as the learning target. Leveraging U-Net's strong spatial feature extraction capabilities, it fits the systematic deviations and spatial structure differences between simulated and observed values, achieving efficient and high-precision correction of marine ecological model chlorophyll simulation results. The intelligent network model adopts a typical U-shaped symmetrical network structure, comprising an encoder, a bottleneck layer, and a decoder. The bottleneck layer is connected to both the encoder and decoder, which are symmetrically structured. The encoder and decoder achieve cross-layer feature fusion through skip connections, balancing spatial feature extraction and high-resolution reconstruction capabilities.

[0065] In implementation, the encoder comprises n feature encoding modules connected sequentially. Assuming n is 4, the first feature encoding module performs feature encoding on the input data and inputs the generated first encoded feature data to the second feature encoding module; similarly, the second feature encoding module performs feature encoding on the first encoded feature data, and the fourth feature encoding module performs feature encoding on the third encoded feature data. The width and height of the first encoded feature data are greater than the width and height of the second encoded feature data, and the number of feature channels in the first encoded feature data is less than the number of feature channels in the second encoded feature data. For example, the first encoded feature data could be 32×W×H, and the second encoded feature data could be 64×W / 2×H / 2.

[0066] The decoder consists of four feature decoding modules connected in sequence. Taking n=4 as an example, the first feature decoding module decodes the feature data output from the bottleneck layer and the fourth encoded feature data output from the fourth feature encoding module to generate the first decoded feature data; the second feature decoding module decodes the first decoded feature data and the third encoded feature data to generate the second decoded feature data; similarly, the third feature decoding module decodes the second decoded feature data and the second encoded feature data to generate the third decoded feature data; and the fourth feature decoding module decodes the third decoded feature data and the first encoded feature data to generate the fourth decoded feature data.

[0067] See Figure 2 The diagram shown illustrates the processing of the intelligent network model. Figure 2 In the intelligent network model, n is 4, and the following is combined with Figure 2 The processing procedure of the intelligent network model is illustrated with an example.

[0068] In this application, the encoder can adopt a four-level downsampling structure, where each convolutional module contains two 3×3 convolutional layers, a batch normalization layer, and a linear activation function. The convolutional operation uses a padding size of 1 to keep the feature map spatial size unchanged. After each convolutional module, the feature map is downsampled by a factor of 2 through a 2×2 max pooling layer, so that the number of feature channels increases sequentially, for example, from the initial 32 to 64, 128, and 256, gradually expanding the receptive field and extracting high-level semantic features.

[0069] See Figure 2 As shown, taking a four-module feature encoding approach as an example, the input data is 1×W×H. This input data is fed into the first feature encoding module, which outputs the first encoded feature data as 32×H×W. The first feature encoding module includes a convolutional module containing two 3×3 convolutional layers, a batch normalization layer, and a linear activation function. The convolution operation uses padding of 1 to maintain the feature map's spatial dimensions (i.e., the height H and width W of the feature map are unchanged). The second feature encoding module utilizes the convolutional module and... The max-pooling layer performs convolution and downsampling on the first encoded feature data, outputting a second encoded feature data of 64×W / 2×H / 2. The third feature encoding module uses a convolution module and a max-pooling layer to perform convolution and downsampling on the second encoded feature data, outputting a third encoded feature data of 128×W / 4×H / 4. The fourth feature encoding module uses a convolution module and a max-pooling layer to perform convolution and downsampling on the third encoded feature data, outputting a fourth encoded feature data of 256×W / 8×H / 8. Figure 2The orange arrows indicate the execution of convolution operations in the convolution module, namely two 3×3 convolutional layers, a batch normalization layer, and a linear activation function (i.e., Conv2d(3×3)+BatchNorm+ReLU×2); the green arrows indicate the execution of convolution operations in the convolution module and downsampling operations in the max pooling layer (i.e., Conv2d(3×3)+BatchNorm+ReLU×2, Max.pooling(2×2)(downsampling)); the dark blue arrows indicate the execution of convolution operations in the convolution module, upsampling operations in bilinear interpolation, and 1×1 convolution operations (i.e., Conv2d(3×3)+BatchNorm+ReLU×2, Bilinear(upsampling), Conv2d(1×1)); the light blue arrows indicate the execution of 1×1 convolution operations (i.e., Conv2d(1×1)).

[0070] A bottleneck layer is set after the encoder. In the bottleneck layer, the number of feature channels of the fourth encoded feature data is expanded to 512×W / 16×H / 16 feature data by performing two convolution operations.

[0071] The decoder and encoder have a symmetrical structure. Each decoding stage first performs a 2x upsampling through bilinear interpolation. After adjusting the channel dimension through 1×1 convolution, it is aligned with the feature dimension of the corresponding level of the encoder. The upsampled feature map is then concatenated with the feature map of the corresponding level of the encoder along the channel dimension. The concatenated features are further extracted and enhanced through a convolution module consisting of two 3×3 convolution layers, a batch normalization layer, and a linear activation function, thereby improving the model's ability to express spatial structural information.

[0072] See Figure 2 As shown, taking four feature decoding modules as an example, the first feature decoding module decodes the feature data output from the bottleneck layer and the fourth encoded feature data output from the fourth feature encoding module to generate the first decoded feature data. Specifically, the first feature decoding module first performs bilinear interpolation on the feature data output from the bottleneck layer to achieve a 2x upsampling, and then uses a 1×1 convolution to adjust the channel dimension so that the dimension of the dimension-adjusted feature data is consistent with the dimension of the fourth encoded feature data; then, the dimension-adjusted feature data and the fourth encoded feature data are concatenated, and the convolution module is used to perform feature processing on the concatenated features to generate the first decoded feature data.

[0073] The second feature encoding module first performs bilinear interpolation on the first decoded feature data to achieve a 2x upsampling, then uses a 1×1 convolution to adjust the channel dimensions so that the dimension of the dimension-adjusted feature data is consistent with the dimension of the third encoded feature data. Next, the dimension-adjusted feature data is concatenated with the third encoded feature data, and the convolution module performs feature processing on the concatenated features to generate the second decoded feature data. Similarly, the third and fourth decoded feature data can be generated.

[0074] In the final output layer of the network, a 1×1 convolution is used to linearly map the last layer feature map of the decoder (i.e., the fourth decoded feature data), converting the 32-dimensional feature channels into the target number of output channels (e.g., 1×H×W). The intelligent network model of this application directly outputs the log-odds value of the prediction result, instead of using the sigmoid activation function inside the network, so as to combine it with the loss function for unified calculation during training, thereby improving the stability and convergence efficiency of model training.

[0075] The method provided in this application is illustrated below. In specific implementation, in the first stage, chlorophyll numerical simulation can be performed using a marine biogeochemical model within the current time period to obtain simulated chlorophyll concentration data. Specifically, using the initial environmental parameters of the target sea area (such as initial nutrient concentration, seawater temperature and salinity, topographic data, etc.), boundary conditions (such as surface atmospheric wind field heat flux, offshore tides, runoff input, etc.), and initial chlorophyll concentration distribution data as input, the corresponding marine biogeochemical coupling model is initiated. Chlorophyll and multi-element synchronous numerical simulations are conducted according to a daily time period, outputting spatiotemporal distribution data of chlorophyll concentration (i.e., simulated chlorophyll concentration data), and simultaneously outputting auxiliary simulation data such as nutrients, temperature, and flow field. Then, the simulated chlorophyll concentration data is log10 normalized to generate standardized chlorophyll concentration basic simulation data (i.e., processed simulated chlorophyll concentration data).

[0076] In the second stage, the simulation results are intelligently optimized and corrected. Specifically, the processed chlorophyll concentration simulation data output from the first stage, along with the satellite remote sensing chlorophyll observation data (i.e., the processed chlorophyll concentration observation data) preprocessed by log10 standardization and spatial grid unification, are input into the trained dedicated U-Net intelligent network model. Optionally, marine environmental characteristic data (such as nearshore-ocean attributes, land-sea masking, etc.) can also be input into the intelligent network model. The intelligent network model extracts the systematic deviations and spatial correlation features between the simulation data and the observation data through the encoder, and reconstructs and outputs a high-resolution chlorophyll concentration correction field through the decoder. This completes the global systematic deviation correction of the original simulation results, obtaining the optimal chlorophyll correction result for the current time period (i.e., the corrected chlorophyll concentration simulation data).

[0077] In the third stage, the optimization results are fed back to the marine biogeochemical model to participate in the numerical simulation for the next time period. Specifically, the optimal chlorophyll concentration correction results obtained in the second stage (i.e., the corrected chlorophyll concentration simulation data) are subjected to vertical interpolation of the model grid to generate chlorophyll concentration distribution data for the current time period. This chlorophyll concentration distribution data can then be used as the initial field for the numerical simulation of the next time period, and the marine biogeochemical coupled model for the corresponding sea area in the next time period is activated, forming a fully closed-loop optimization mechanism of "simulation-intelligent correction-result feedback-iterative simulation".

[0078] This application leverages the advantages of U-shaped convolutional neural networks (U-Net), such as lightweight design, end-to-end training, strong spatial feature extraction capabilities, good adaptability to small samples, and high computational efficiency, to deeply integrate its strong spatiotemporal feature learning and nonlinear mapping capabilities with marine biogeographical models. This results in the construction of a novel integrated chlorophyll optimization method that uses observational data as the learning target, is driven by both simulated and observational data, features U-Net intelligent correction, and incorporates cyclical feedback of results.

[0079] Unlike existing intelligent optimization technologies that rely on data-driven prediction, this application designs a three-stage closed-loop process. The optimal chlorophyll concentration result, corrected by the UNet model, is used as the initial field for the next time-cycle numerical simulation and fed back to the marine biogeochemical model, forming a fully closed-loop optimization process of "simulation → intelligent correction → result feedback → iterative simulation." This mechanism allows simulation biases to be gradually corrected through iterations, achieving continuous improvement in the accuracy of chlorophyll concentration simulation and enhancing the continuity and adaptability of simulation optimization.

[0080] Corresponding to the embodiments of the aforementioned chlorophyll value determination method, this application also provides embodiments of a chlorophyll value determination device. Figure 3 A schematic diagram of the chlorophyll value determination device provided in this application specifically includes: The acquisition module 301 is used to acquire parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters and current chlorophyll concentration distribution data; The determination module 302 is used to determine the simulated chlorophyll concentration data of the second time period corresponding to the target sea area using the parameter data of the first time period corresponding to the target sea area as input and the marine biogeochemical coupling model; wherein the second time period is later than the first time period. The first generation module 303 is used to use a trained intelligent network model to correct the chlorophyll concentration simulation data based on the chlorophyll concentration observation data of the second time period in the target sea area, and generate corrected chlorophyll concentration simulation data. The second generation module 304 is used to perform vertical interpolation processing on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period.

[0081] In an optional implementation, the system further includes a training module 305 for training the intelligent network model according to the following steps: Acquire training samples, wherein the training samples include chlorophyll concentration simulation samples and chlorophyll concentration observation samples at multiple sampling times corresponding to the sample area. The chlorophyll concentration simulation samples are determined using the marine biogeochemical coupling model, and the chlorophyll concentration observation samples are chlorophyll concentration calibration data of the sample area at the sampling time determined based on the observation dataset. The training samples are input into the neural network model to be trained for processing, and chlorophyll prediction data is output; wherein the neural network model to be trained is a U-shaped symmetric network structure. Based on the chlorophyll prediction data and the chlorophyll concentration observation samples, the loss value is determined; and the network parameters of the neural network model to be trained are adjusted based on the loss value to obtain the adjusted neural network model. The adjusted neural network model is used as a new neural network model to be trained. The process returns to the step of inputting the training samples into the neural network model to be trained for processing, until the training cutoff condition is met, and the intelligent network model is obtained.

[0082] In an optional implementation, the training module 305 is further configured to: perform data preprocessing on the simulated chlorophyll concentration samples and the observed chlorophyll concentration samples in the training samples to generate processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples, wherein the data preprocessing includes logarithmic normalization, spatial grid alignment, and default value labeling; the processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples are used to train the neural network model to be trained.

[0083] In one optional implementation, the intelligent network model includes: an encoder, a bottleneck layer, and a decoder, wherein the bottleneck layer is connected to the encoder and the decoder respectively, and the encoder and the decoder have a symmetrical structure; The encoder includes n feature encoding modules connected in sequence. The first feature encoding module is used to perform feature encoding processing on the input data and input the generated first encoded feature data to the second feature encoding module. The ith feature encoding module is used to perform feature encoding processing on the (i-1)th encoded feature data, where i is a positive integer greater than 1 and less than or equal to n, and n is a positive integer greater than 1. The decoder includes n feature decoding modules connected in sequence. The first feature decoding module is used to decode the feature data output by the bottleneck layer and the nth encoded feature data output by the nth feature encoding module to generate the first decoded feature data. The jth feature decoding module is used to decode the (j-1)th decoded feature data and the (n-j+1)th encoded feature data to generate the jth decoded feature data. Here, j is a positive integer greater than 1 and less than or equal to n.

[0084] In an optional implementation, the training module 305 is further configured to: determine a mask image corresponding to the sample region, wherein the region corresponding to the first pixel value in the mask image is an effective region with chlorophyll concentration and the region corresponding to the second pixel value is an invalid region without chlorophyll concentration. When determining the loss value based on the chlorophyll prediction data and the chlorophyll concentration observation samples, the training module 305 is used to: determine the loss value based on the mask image, the chlorophyll prediction data, and the chlorophyll concentration observation samples.

[0085] In an optional implementation, the second generation module 304 is further configured to: take the chlorophyll concentration distribution data of the second time period as input, and use the marine biogeochemical coupling model to determine the simulated chlorophyll concentration data of the next time period corresponding to the target sea area.

[0086] In one optional embodiment, the environmental parameters include at least one of the following: nutrient concentration, seawater temperature, salinity, and topographic data; the boundary condition parameters include at least one of the following: atmospheric wind field heat flux on the upper surface, offshore tides, and runoff input.

[0087] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0088] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0089] This application also provides a computer-readable storage medium storing a computer program that can be used to execute the chlorophyll value determination method described in the above embodiments.

[0090] This application also provides a computer device, see [link to relevant documentation] Figure 4 The diagram shown illustrates the structure of the computer device provided in this application. At the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for various operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the chlorophyll value determination method described in the above embodiments. Of course, besides software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0091] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs) containing computer-usable program code. The form of a computer program product implemented on ROM, optical memory, etc.

[0092] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes. In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0095] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0096] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, and optical disc read-only memory (CD-ROM). ROM, digital multifunction optical disc (DVD) or other optical storage, magnetic cassette tape, magnetic magnetic disk storage or other magnetic storage devices or any other non-transfer media, may be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0097] It should also be noted that 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 limitation, 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.

[0098] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0099] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0100] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0101] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0102] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0103] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CDs. ROM and DVD ROM disk. The processor and memory can be supplemented by or integrated into dedicated logic circuitry.

[0104] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0105] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0106] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0107] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for determining chlorophyll content, characterized in that, The method includes: Obtain parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data; Using the parameter data of the first time period corresponding to the target sea area as input, the simulated chlorophyll concentration data of the second time period corresponding to the target sea area is determined by the marine biogeochemical coupling model; wherein the second time period is later than the first time period; Using the trained intelligent network model, the chlorophyll concentration simulation data is corrected based on the chlorophyll concentration observation data of the second time period in the target sea area, and the corrected chlorophyll concentration simulation data is generated. Vertical interpolation is performed on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period. The intelligent network model is trained according to the following steps: Acquire training samples, wherein the training samples include chlorophyll concentration simulation samples and chlorophyll concentration observation samples at multiple sampling times corresponding to the sample area. The chlorophyll concentration simulation samples are determined using the marine biogeochemical coupling model, and the chlorophyll concentration observation samples are chlorophyll concentration calibration data of the sample area at the sampling time determined based on the observation dataset. The training samples are input into the neural network model to be trained for processing, and chlorophyll prediction data is output; wherein the neural network model to be trained is a U-shaped symmetric network structure. Based on the chlorophyll prediction data and the chlorophyll concentration observation samples, the loss value is determined; and the network parameters of the neural network model to be trained are adjusted based on the loss value to obtain the adjusted neural network model. The adjusted neural network model is used as a new neural network model to be trained. The process returns to the step of inputting the training samples into the neural network model to be trained for processing, until the training cutoff condition is met, and the intelligent network model is obtained.

2. The method according to claim 1, characterized in that, The method further includes: The simulated chlorophyll concentration samples and observed chlorophyll concentration samples in the training samples are preprocessed to generate processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples. The data preprocessing includes logarithmic normalization, spatial grid alignment, and default value labeling. The processed simulated chlorophyll concentration samples and processed observed chlorophyll concentration samples are used to train the neural network model to be trained.

3. The method according to claim 1, characterized in that, The intelligent network model includes an encoder, a bottleneck layer, and a decoder. The bottleneck layer is connected to the encoder and the decoder respectively, and the encoder and the decoder have a symmetrical structure. The encoder includes n feature encoding modules connected in sequence. The first feature encoding module is used to perform feature encoding processing on the input data and input the generated first encoded feature data to the second feature encoding module. The ith feature encoding module is used to perform feature encoding processing on the (i-1)th encoded feature data, where i is a positive integer greater than 1 and less than or equal to n, and n is a positive integer greater than 1. The decoder includes n feature decoding modules connected in sequence. The first feature decoding module is used to decode the feature data output by the bottleneck layer and the nth encoded feature data output by the nth feature encoding module to generate the first decoded feature data. The jth feature decoding module is used to decode the (j-1)th decoded feature data and the (n-j+1)th encoded feature data to generate the jth decoded feature data. Here, j is a positive integer greater than 1 and less than or equal to n.

4. The method according to claim 1, characterized in that, The method further includes: Determine the mask image corresponding to the sample region, wherein the region corresponding to the first pixel value in the mask image is the effective region where chlorophyll concentration exists, and the region corresponding to the second pixel value is the invalid region where chlorophyll concentration does not exist; The step of determining the loss value based on the chlorophyll prediction data and the chlorophyll concentration observation samples includes: The loss value is determined based on the mask image, the chlorophyll prediction data, and the chlorophyll concentration observation samples.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: Using the chlorophyll concentration distribution data of the second time period as input, the simulated chlorophyll concentration data of the next time period corresponding to the target sea area is determined by the marine biogeochemical coupling model.

6. The method according to any one of claims 1-4, characterized in that, The environmental parameters include at least one of the following: nutrient concentration, seawater temperature, salinity, and topographic data; The boundary condition parameters include at least one of the following: atmospheric wind field heat flux on the upper surface, ocean tide, and runoff input.

7. A chlorophyll value determination device, characterized in that, The device includes: The acquisition module is used to acquire parameter data for the first time period corresponding to the target sea area, wherein the parameter data includes environmental parameter data, boundary condition parameters, and current chlorophyll concentration distribution data; The determination module is used to determine the simulated chlorophyll concentration data of the target sea area in a second time period using the parameter data of the first time period corresponding to the target sea area as input and the marine biogeochemical coupling model; wherein the second time period is later than the first time period. The first generation module is used to use a trained intelligent network model to correct the chlorophyll concentration simulation data based on the chlorophyll concentration observation data of the second time period in the target sea area, and generate corrected chlorophyll concentration simulation data. The second generation module is used to perform vertical interpolation processing on the corrected chlorophyll concentration simulation data to generate chlorophyll concentration distribution data for the second time period. The training module is used to train the intelligent network model according to the following steps: Acquire training samples, wherein the training samples include chlorophyll concentration simulation samples and chlorophyll concentration observation samples at multiple sampling times corresponding to the sample area. The chlorophyll concentration simulation samples are determined using the marine biogeochemical coupling model, and the chlorophyll concentration observation samples are chlorophyll concentration calibration data of the sample area at the sampling time determined based on the observation dataset. The training samples are input into the neural network model to be trained for processing, and chlorophyll prediction data is output; wherein the neural network model to be trained is a U-shaped symmetric network structure. Based on the chlorophyll prediction data and the chlorophyll concentration observation samples, the loss value is determined; and the network parameters of the neural network model to be trained are adjusted based on the loss value to obtain the adjusted neural network model. The adjusted neural network model is used as a new neural network model to be trained. The process returns to the step of inputting the training samples into the neural network model to be trained for processing, until the training cutoff condition is met, and the intelligent network model is obtained.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor performs the steps of the method according to any one of claims 1-6.

Images