El Nino power autonomous prediction method, device and equipment

Abstract

The invention is suitable for the field of meteorological prediction, and provides an El Nino power autonomous prediction method, device and equipment, and the method comprises the steps: obtaining historical reanalysis data and target surface sea temperature observation data; performing space-time matching on the target surface layer sea temperature observation data and historical reanalysis data to construct a target matching data set; pre-training the first neural network model, and establishing an initial mapping relation between the surface sea temperature and the deep sea temperature; carrying out transfer learning on a pre-trained first neural network model to obtain a second neural network model; inputting the target surface layer sea temperature observation data into the second neural network model, and reconstructing target three-dimensional sea temperature data; generating a target initial field based on an assimilation module of the El Nino power autonomous prediction model; and the El Nino power autonomous prediction model is driven by the target initial field to carry out prediction integration, and an El Nino prediction result is output. According to the method, the autonomy of the El Nino prediction is realized.

Description

Technical Field

[0001] This application belongs to the field of meteorological forecasting technology, and in particular relates to an autonomous El Niño dynamic forecasting method, device and equipment. Background Technology

[0002] Since the 1980s, the study of El Niño / La Niña phenomena has been a key focus in ocean and atmospheric science. Current satellite remote sensing technology has limitations in acquiring subsurface to deep ocean temperature data, leading to insufficient accuracy in traditional estimation methods. This problem makes it difficult to achieve high precision in monitoring and predicting El Niño. Current El Niño monitoring and prediction systems mainly rely on reanalysis datasets published abroad, lacking independent and reliable data sources. This dependence exposes my country to data acquisition risks when responding to changes in international relations, affecting the autonomy and security of predictions. Existing prediction models lack sufficient inversion capabilities for deep sea surface temperature, failing to fully utilize target surface sea surface temperature observation data collected by domestic satellites, resulting in reduced model adaptability and accuracy. The data assimilation modules of current El Niño prediction models have limited processing capabilities for new data, making it difficult to effectively integrate data from different sources, especially domestic satellite data, thus affecting the construction of the initial field and the accuracy of prediction results. Summary of the Invention

[0003] This application provides an autonomous El Niño dynamic prediction method, apparatus, and device that can solve the above-mentioned problems.

[0004] In a first aspect, embodiments of this application provide an autonomous El Niño dynamic prediction method, including: Historical reanalysis data and target surface sea surface temperature observation data were acquired; wherein the target surface sea surface temperature observation data was collected by domestic satellites. The target surface sea surface temperature observation data is spatiotemporally matched with the historical reanalysis data to construct a target matching dataset; The first neural network model is pre-trained based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. Based on the target matching dataset, the pre-trained first neural network model is transferred to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The target surface sea surface temperature observation data is input into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data, and the target three-dimensional sea surface temperature data is reconstructed. The assimilation module based on the El Niño dynamic autonomous prediction model assimilates the target's three-dimensional sea surface temperature data into the ocean model to generate the target's initial field. The El Niño dynamic autonomous prediction model is driven by the target initial field to perform prediction integration and output the El Niño prediction result.

[0005] Secondly, embodiments of this application provide an autonomous El Niño dynamic prediction device, comprising: The acquisition unit is used to acquire historical reanalysis data and target surface sea surface temperature observation data; wherein the target surface sea surface temperature observation data is collected by domestic satellites; The first processing unit is used to perform spatiotemporal matching between the target surface sea surface temperature observation data and the historical reanalysis data to construct a target matching dataset. The second processing unit is used to pre-train the first neural network model based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. The third processing unit is used to perform transfer learning on the pre-trained first neural network model based on the target matching dataset to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The fourth processing unit is used to input the target surface sea surface temperature observation data into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data and reconstruct the target three-dimensional sea surface temperature data. The fifth processing unit is used as an assimilation module based on the El Niño dynamic autonomous prediction model to assimilate the target three-dimensional sea surface temperature data into the ocean model and generate the target initial field. The sixth processing unit is used to drive the El Niño dynamic autonomous prediction model to perform prediction integration with the target initial field and output the El Niño prediction result.

[0006] Thirdly, embodiments of this application provide an autonomous El Niño dynamic prediction device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect above.

[0007] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0008] In this embodiment, by combining domestic satellite data and historical reanalysis data, the prediction accuracy of El Niño can be improved, providing more reliable information support for climate research and industries such as agriculture and fisheries. Through real-time data assimilation and neural network transfer learning, it can quickly adapt to new climate conditions, improving the timeliness of predictions. This method achieves autonomous El Niño prediction, reducing reliance on human intervention and enabling continuous monitoring and prediction without human intervention. This prediction model is not only applicable to El Niño event prediction but can also be extended to the study of other climate phenomena, such as La Niña and the monitoring of global climate change. This method provides a new approach for future meteorological forecasting and oceanographic modeling research, promoting the application of artificial intelligence technology in meteorological and oceanographic sciences. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic flowchart of an autonomous El Niño dynamic prediction method provided in the first embodiment of this application; Figure 2 This is a schematic flowchart of steps S1011 to S1013 in an autonomous prediction method for El Niño dynamics provided in the first embodiment of this application; Figure 3 This is a schematic flowchart of S1031~S1032 in an autonomous prediction method for El Niño dynamics provided in the first embodiment of this application; Figure 4 This is a schematic flowchart of S108~S110 in an autonomous prediction method for El Niño dynamics provided in the first embodiment of this application; Figure 5 This is a schematic flowchart of steps S111 to S113 in an autonomous El Niño dynamics prediction method provided in the first embodiment of this application; Figure 6 This is a schematic diagram of the autonomous El Niño dynamic prediction device provided in the second embodiment of this application; Figure 7 This is a schematic diagram of the autonomous El Niño dynamics prediction device provided in the third embodiment of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0013] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0014] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0015] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0016] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0017] Please see Figure 1 , Figure 1This is a schematic flowchart of an autonomous El Niño dynamic prediction method provided in the first embodiment of this application. In this embodiment, the executing entity of the autonomous El Niño dynamic prediction method is a device with autonomous El Niño dynamic prediction capabilities, such as a desktop computer, server, etc. Figure 1 The El Niño dynamics autonomous prediction method shown may include: S101: Acquire historical reanalysis data and target surface sea surface temperature observation data; wherein the target surface sea surface temperature observation data is collected by domestic satellites.

[0018] In this embodiment, historical reanalysis data and target sea surface temperature observation data can be collected in advance. Historical reanalysis data (such as databases provided by NASA, NOAA, etc.) can be downloaded from meteorological databases or meteorological centers. Sea surface temperature observation data acquired from domestic satellites (such as the Gaofen series satellites) are collected and processed to ensure that the data format and time range are consistent with the reanalysis data.

[0019] Historical reanalysis data refers to a series of historical climate data generated by combining numerical weather prediction models with observational data. This data typically includes wind speed, air temperature, ocean temperature, etc., covering multiple periods.

[0020] Target surface sea surface temperature observation data refers to sea surface temperature data acquired through domestically produced satellites. Satellite data typically has high temporal and spatial resolution, reflecting the instantaneous state of the ocean surface.

[0021] For example, ocean surface temperature information obtained through Chinese ocean satellites (such as OceanSat-1) may include daily temperature records, providing information on the temperature distribution of the ocean surface.

[0022] In one embodiment, S101 may include S1011~S1013, such as Figure 2 As shown, S1011~S1013 are as follows: S1011: Obtain GODAS reanalysis data for the first historical time period and obtain surface sea surface temperature observation data for the second historical time period.

[0023] GODAS (GODAE Ocean Data Assimilation System) is an ocean data assimilation system that provides reanalysis data on the state of the global ocean.

[0024] Find publicly available databases of GODAS reanalysis data through the websites of relevant organizations (such as NOAA, NASA, or other meteorological research institutions).

[0025] Determine the first historical period required (e.g., 1990 to 2000).

[0026] Download GODAS reanalysis data based on the selected time period. The data typically includes multi-dimensional information such as temperature, salinity, and flow rate.

[0027] If the data format does not meet the requirements for subsequent processing (such as NetCDF, HDF5, etc.), perform format conversion.

[0028] Sea surface temperature data refers to sea surface temperature data obtained through satellite remote sensing or buoys.

[0029] Use sea surface temperature observation data sources provided by the State Oceanic Administration, NOAA, or relevant research institutions.

[0030] Select a second historical period (e.g., 2001 to 2010).

[0031] Download surface sea surface temperature observation data from the selected data source. The data is usually provided in CSV or NetCDF format.

[0032] After downloading, perform a preliminary check on the completeness and accuracy of the data to ensure there are no obvious missing values ​​or erroneous data.

[0033] S1012: Preprocess the surface sea surface temperature observation data for the second historical time period; wherein, the preprocessing includes, but is not limited to, format unification and anomaly removal.

[0034] The surface sea surface temperature observation data for the second historical time period were preprocessed, including but not limited to format standardization and outlier removal.

[0035] All surface sea surface temperature observation data can be converted to a unified format, such as converting all data to CSV format. Ensure that the naming of the time, longitude, latitude, and temperature fields in the data is consistent for subsequent processing.

[0036] Check if there are missing values ​​(such as NaN) in the data. You can choose to fill them (such as filling with the average value) or remove them.

[0037] Statistical methods (such as Z-score and IQR) are used to detect outliers. For example, if the temperature value of a data point is too high or too low (e.g., exceeding ±3 standard deviations), it is marked as an outlier. Based on the detection results, records marked as outliers are removed to ensure data quality.

[0038] S1013: The preprocessed surface sea surface temperature observation data is interpolated and completed based on the GODAS reanalysis data to obtain the target surface sea surface temperature observation data.

[0039] Interpolation is the process of inferring unknown data points from existing data points. Various interpolation methods can be used, such as linear interpolation, kriging interpolation, and spline interpolation. Choose an interpolation method (such as kriging interpolation) that is suitable for the characteristics of the ocean data.

[0040] The preprocessed surface sea surface temperature observation data were merged with GODAS reanalysis data to obtain the complete spatial distribution.

[0041] Use statistical software (such as scikit-learn in Python or the gstat package in R) to build interpolation models.

[0042] The merged data is interpolated to generate target surface sea surface temperature observation data. The resolution of the interpolation grid can be set to facilitate subsequent analysis.

[0043] In this embodiment, the historical data used is ensured to have high accuracy and consistency, thus laying the foundation for subsequent analysis. By standardizing the format and removing outliers, errors caused by data inconsistencies are reduced, making the subsequent analysis results more reliable. The preprocessed data facilitates further statistical analysis and model training. Using GODAS reanalysis data for interpolation and completion effectively fills in missing parts of the observational data, making the sea surface temperature data more complete. This is crucial for the modeling and prediction of ocean climate models, especially regarding the impact of ocean surface temperature changes on El Niño. This method allows for parameter adjustment and selection of different interpolation methods to adapt to ocean data analysis in different regions and time periods. By incorporating more historical data or different types of climate data, the predictive and generalization capabilities of the model can be further improved.

[0044] S102: Perform spatiotemporal matching between the target surface sea surface temperature observation data and the historical reanalysis data to construct a target matching dataset.

[0045] In this embodiment, the two data sources are matched in time and space to ensure their consistency in time and space.

[0046] Spatiotemporal matching refers to aligning surface sea surface temperature observation data of a target with historical reanalysis data in both time and space. Interpolation methods are typically used to address inconsistencies in spatial resolution.

[0047] Specifically, interpolation algorithms (such as linear interpolation, Kriging interpolation, etc.) can be used to adjust the spatial resolution of the reanalysis data to be the same as that of the satellite observation data. Ensure temporal matching, for example, by unifying all data to daily or monthly averages.

[0048] For example, if historical reanalysis data is updated monthly, while satellite data is daily, a monthly average can be calculated, allowing for comparison between the two.

[0049] S103: Based on the historical reanalysis data, the first neural network model is pre-trained to establish an initial mapping relationship between surface sea temperature and deep sea temperature.

[0050] The device contains a pre-stored neural network (such as a multilayer sensor), which takes surface sea surface temperature data as input and outputs deep sea surface temperature data.

[0051] Train the neural network using historical reanalysis data, selecting an appropriate loss function (such as mean squared error) and optimization algorithm (such as the Adam optimizer). Iterate through multiple iterations to optimize the network weights and obtain a better initial mapping relationship.

[0052] For example, the input surface sea surface temperature data may be the monthly average sea surface temperature data over the past 30 years, and the output is the corresponding deep sea surface temperature data. After 1,000 iterations of training, a preliminary mapping relationship is obtained.

[0053] In one embodiment, S103 may include S1031~S1032, such as Figure 3 As shown, S1031~S1032 are as follows: S1031: Initialize the first neural network model based on the U-Net network architecture.

[0054] Net is a convolutional neural network (CNN) architecture for image segmentation, originally applied to biomedical image processing. Its main characteristic is its symmetrical encoder-decoder structure, which effectively captures contextual information and achieves high-precision segmentation.

[0055] First, import the necessary libraries and define the structure of the U-Net model, including the encoder, bottleneck layer, and decoder. After the model definition is complete, compile it using an appropriate loss function and optimizer. This completes the initialization of the first neural network model based on the U-Net network architecture.

[0056] S1032: Based on the historical reanalysis data, the first neural network model based on the U-Net network architecture is pre-trained to establish an initial mapping relationship between surface sea temperature and deep sea temperature; wherein, the loss function is mean squared error and the optimizer is Adam.

[0057] In this embodiment, historical reanalysis data refers to meteorological and oceanographic data collected by meteorological agencies over many years, including surface sea surface temperature (SST) and deep sea surface temperature (DPT).

[0058] Load historical reanalysis data and divide it into training and validation sets. Standardize the data to facilitate model processing. To improve the model's robustness, data augmentation can be performed, such as rotation, flipping, and scaling.

[0059] The model is trained using training data, and its performance is evaluated on the validation set. Data augmentation can be achieved by passing in a data generator.

[0060] During training, the training loss and validation loss are monitored to avoid overfitting. Callback functions can be used. The loss function is the mean squared error, and the optimizer is Adam.

[0061] After training, use the trained model to make predictions on the validation set or new data. Evaluate the model's predictive performance using metrics such as mean squared error (MSE) and observe its performance on the test set.

[0062] S104: Perform transfer learning on the pre-trained first neural network model based on the target matching dataset to obtain a second neural network model adapted to the target surface sea surface temperature observation data.

[0063] The first neural network model, which has already been trained, is retrained using a small amount of target data to adapt the model to the new dataset.

[0064] Fine-tune the first model using a target matching dataset. Reduce the risk of overfitting by freezing some layers and training only the last few layers. Create a new output layer. For example, if the task is regression (predicting sea surface temperature), add a linear layer; if it's a classification task, add a softmax layer. Choose an appropriate loss function and optimizer, and compile the model. Continue optimizing the network parameters until the model performs satisfactorily on the target dataset.

[0065] In this step, the key to transfer learning lies in leveraging the already trained first neural network model to accelerate and improve the learning efficiency for the new task. By freezing the bottom layer and adding new task-specific layers, transfer learning can be effectively performed on the pre-trained first neural network model to obtain a second neural network model adapted to the target surface sea surface temperature observation data.

[0066] S105: Input the target surface sea surface temperature observation data into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data and reconstruct the target three-dimensional sea surface temperature data.

[0067] The target surface sea surface temperature data is input into the trained second neural network model, which outputs the deep sea surface temperature inversion results.

[0068] Using the model's prediction function, the input surface sea surface temperature data is forward-propagated through a network to obtain the prediction results for deep sea surface temperature.

[0069] Based on deep sea surface temperature data and combined with ocean physical models, the three-dimensional sea surface temperature distribution is reconstructed.

[0070] Suppose that the second neural network model is used as input with surface sea surface temperature data from the most recent month, the output deep sea surface temperature data can be used to display the temperature distribution of the ocean.

[0071] S106: An assimilation module based on the El Niño dynamic autonomous prediction model assimilates the target's three-dimensional sea surface temperature data into the ocean model to generate the target's initial field.

[0072] By combining new observational data (i.e., predicted three-dimensional sea surface temperature data) with existing ocean model status, the model can better reflect the current ocean state.

[0073] Choose an appropriate data assimilation method, such as weighted averaging, Kalman filtering, or three-dimensional variational assimilation (3DVAR). These methods can combine newly acquired three-dimensional sea surface temperature data with the initial field of the ocean model.

[0074] An initial field containing the latest observational information is generated by adjusting the state variables (such as temperature, salinity, and current velocity) of the ocean model through an assimilation algorithm.

[0075] Suppose that a three-dimensional variational assimilation method is used to combine the predicted target three-dimensional sea surface temperature data with the model's previous state to generate a more accurate initial field that reflects the current true state of the ocean.

[0076] In one embodiment, S106 may include: an assimilation module based on an El Niño dynamic autonomous prediction model, which uses the target three-dimensional sea surface temperature data as an observation increment, assimilates the observation increment into the background field of the ocean model using a Newtonian relaxation approximation method, and dynamically adjusts the relaxation coefficient based on the observation error covariance and the background error covariance to generate the target initial field.

[0077] Real-time or historical sea surface temperature (SST) observation data can be collected from sources such as satellite remote sensing, buoys, or weather stations. This data should cover the required time frame and spatial area to provide a basis for subsequent assimilation.

[0078] By running an ocean model (such as the Ocean General Circulation Model) for a period of time, a background field is obtained, including information such as seawater temperature, salinity, and current velocity. The background field is the model's prediction of the current ocean state and is usually obtained through numerical simulation.

[0079] The magnitude of observation error can be assessed based on the source and quality of the observation data. It can be estimated using the difference between historical observation data and actual values.

[0080] For example, observational data over a period of time can be collected and compared with the model output to calculate the variance of the error.

[0081] The uncertainty of the model predictions is evaluated through multiple simulations of the background field (e.g., using different initial conditions or model parameters). The background error covariance matrix is ​​obtained using historical data of the background field and statistical methods.

[0082] An initial relaxation coefficient (λ) is set, typically a value less than 1, to control the influence of the assimilation process. Using the idea of ​​Newton's iteration method, the relaxation coefficient is dynamically adjusted based on the characteristics of the nonlinear equation.

[0083] The observation increment (O), which is the difference between the observed data and the background field, is calculated. Based on the Newton relaxation approximation method, the observation increment is assimilated into the background field to generate a new initial field.

[0084] The relaxation coefficient (λ) is dynamically adjusted based on the current error covariance matrices (R and B) to improve assimilation. For example, a simple rule can be used: increase λ if the current error is large, and decrease λ if the error is small.

[0085] The assimilated ocean state (x_a) serves as the new initial field for the target, which can be used for subsequent ocean model predictions. This initial field will more accurately reflect the current ocean state, aiding in further forecasting and analysis.

[0086] In this embodiment, by effectively assimilating real-time observation data with the model, the accuracy of predictions can be significantly improved, especially in the early stages of an event. The model can be dynamically updated based on the latest observation data, promptly capturing changes in ocean conditions and providing more timely early warning information.

[0087] In one embodiment, before S106, S108-S110 may also be included, such as... Figure 4 As shown, S108~S110 are as follows: S108: Using the initial field of the 20th-century climate state as the driving condition, control the autonomous El Niño dynamic prediction model to perform a 50-year free integration calculation.

[0088] Collect climate data from the 20th century, including information on sea surface temperature (SST), wind fields, and air pressure. This data can come from global climate observation networks (such as NOAA and NASA).

[0089] This data is then organized into the format required for the model to construct the initial climate field for the 20th century. The initial field typically includes variables such as ocean and atmospheric temperature, humidity, and wind speed.

[0090] Using the constructed initial field as input, a 50-year free integration calculation is performed through an El Niño dynamics autonomous prediction model. This process uses numerical integration methods (such as the Euler method or the Runge-Kutta method) to simulate future climate change.

[0091] Assuming that in 1950, the ocean surface temperature was 28°C and the wind speed was 5 m / s, we used these data to construct the initial field and input it into the model for simulation.

[0092] S109: Monitor the energy exchange flux between the atmospheric component and the upper ocean component within the El Niño dynamic autonomous prediction model.

[0093] During model operation, the energy exchange flux between the atmosphere and the upper ocean is calculated. This can be achieved through the principle of energy conservation, involving processes such as ocean surface evaporation and radiative heat transfer.

[0094] Real-time monitoring and recording of changes in energy exchange flux, and storage of relevant data for subsequent analysis.

[0095] Statistical analysis was performed on the monitored energy exchange flux, and the standard deviation of its fluctuations was calculated to determine whether the energy transfer process within the model was stable.

[0096] If the calculated energy exchange flux at a certain moment is 50 W / m², and subsequent monitoring shows that its fluctuation range is within ±5 W / m², then this data should be recorded for further analysis.

[0097] S110: When the standard deviation of the energy exchange flux converges within a preset threshold range, it is determined that the El Niño dynamic autonomous prediction model has reached a physical equilibrium state, the assimilation module is activated, and the target three-dimensional sea surface temperature data is input into the El Niño dynamic autonomous prediction model.

[0098] When the standard deviation of the monitored energy exchange flux fluctuations converges within a preset threshold range (e.g., ±2 W / m²), the model is considered to have reached a state of physical equilibrium.

[0099] Once physical equilibrium is determined to be reached, the assimilation module is activated. At this point, the latest target 3D sea surface temperature data is ready to be input.

[0100] Processed three-dimensional sea surface temperature data (e.g., real-time data obtained through satellite remote sensing) are input into the model so that the model can be updated and the predictions adjusted.

[0101] Assuming the preset standard deviation threshold for fluctuation is ±2 W / m², while the actual monitored standard deviation for fluctuation is 1.5 W / m², once the condition is met, the assimilation module is activated and the latest sea surface temperature data is input.

[0102] In this implementation, 20th-century climatological data is used to provide reliable initial conditions for the model, ensuring its accuracy during free integration. Continuous monitoring of energy exchange fluxes within the model ensures that atmospheric and oceanic energy transfer processes remain within reasonable ranges, avoiding prediction errors caused by physical imbalances. Once the model is detected to have reached physical equilibrium, the assimilation module is activated promptly, inputting the latest sea surface temperature data to ensure the model can quickly adapt to climate change. Through these steps, the model can more accurately predict El Niño events, providing a more reliable basis for climate research and policy making.

[0103] S107: Drive the El Niño dynamic autonomous prediction model to perform prediction integration using the target initial field, and output the El Niño prediction result.

[0104] Using the assimilated initial field as input, the El Niño dynamic model is driven to predict future climate trends.

[0105] The assimilated target initial field is input into the El Niño dynamic model and integrated over time. Time stepping methods of numerical models (such as the Euler method or the Leapfrog method) are typically used for time advancement.

[0106] The model's dynamic equations are used to calculate the ocean state at future points in time, with particular attention to changes in sea surface temperature.

[0107] The model is assumed to be based on the target initial field and to make predictions for the next 12 months, outputting the trend of sea surface temperature changes for each month, especially the prediction of El Niño events.

[0108] In one embodiment, S107 may include: using the target initial field as the initial condition, driving the El Niño dynamic autonomous prediction model to perform prediction integration, outputting a multi-element spatiotemporal evolution sequence, and obtaining the El Niño prediction result; wherein the integration period covers the next 12 months.

[0109] In this embodiment, historical meteorological data is collected, including parameters such as sea surface temperature, air pressure, and wind speed. Commonly used data sources include remote sensing data provided by meteorological agencies such as NOAA and NASA.

[0110] Interpolation is performed on missing parts of the data to ensure data integrity. Data of different magnitudes is standardized to facilitate model processing.

[0111] Choose a specific point in time (e.g., the current date) as the target initial field. Extract relevant meteorological parameters for that point in time from the collected data to construct the target initial field. The target initial field typically includes: ocean temperature distribution (SST), wind field (wind speed and direction), and atmospheric pressure distribution. Choose an appropriate climate model, such as a dynamic model (e.g., CFSv2, GFDL, etc.) or a statistical model (e.g., linear regression, machine learning model, etc.).

[0112] The constructed target initial field is used as input and set as the initial condition of the model.

[0113] The model is integrated using numerical integration methods (such as the Euler method or the Runge-Kutta method) to simulate future climate change. The integration period is set to 12 months, and an appropriate time step (such as 1 day or 1 week) is selected for integration.

[0114] During the prediction process, the model will output the spatiotemporal evolution sequence of multiple climate elements, such as changes in sea surface temperature (SST), changes in atmospheric precipitation, and the evolution of wind fields.

[0115] The multi-element spatiotemporal evolution sequence of the output is analyzed to find the probability and intensity of El Niño phenomenon.

[0116] Suppose we conduct an El Niño forecast on January 1, 2026: We obtain SST, wind speed, and other data from meteorological data from January to December 2025. We extract the SST and wind field data for January 1, 2026, to construct an initial field. We select the CFSv2 model and run it with the target initial field as input. We perform a 12-month forecast, outputting the SST and air pressure changes from January to December 2026. By comparing historical data, we assess the probability and intensity of the 2026 El Niño event.

[0117] In this embodiment, an autonomous prediction method can more efficiently capture changes in the El Niño phenomenon, improving prediction accuracy. It enables timely access to information on future climate change, facilitating the development of countermeasures. The prediction not only focuses on ocean temperature changes but also considers various factors such as atmospheric circulation and precipitation patterns, providing a more comprehensive view of climate change. This multi-dimensional information can help policymakers, agricultural workers, and fisheries managers better address the impacts of climate change. This prediction method can provide data support for climate science research, helping scientists gain a deeper understanding of the mechanisms of the El Niño phenomenon, thereby promoting research and development in related fields.

[0118] In one embodiment, this embodiment may further include S111~S113, such as Figure 5 As shown, S111~S113 are as follows: S111: Obtain historical El Niño observation data.

[0119] First, determine the sources of observational data for historical El Niño events. Common data sources include: NOAA (National Oceanic and Atmospheric Administration) NASA (National Aeronautics and Space Administration) Climate databases from organizations such as IRI (International Institute for Climate and Society).

[0120] The El Niño observation data provided by these agencies typically includes key meteorological parameters such as sea surface temperature (SST), wind speed, and precipitation.

[0121] S112: Perform spatiotemporal matching between the El Niño prediction results and historical El Niño observation data, and calculate the prediction error.

[0122] Use dynamic models (such as ocean-atmosphere coupled models) to predict future El Niño events and generate forecasts, which can be SST values ​​for the next few months.

[0123] Define the time window: For example, set the time range of the forecast results to the same months as the historical data.

[0124] Spatial matching: Ensure that the spatial resolution of the data is consistent. This is usually achieved by matching data with different resolutions through interpolation methods.

[0125] Calculate the prediction error: Use common error metrics, such as root mean square error (RMSE) and mean absolute error (MAE), to quantify it.

[0126] Assuming the predicted SST values ​​are for January to June 2021, and the historical observations for the same period are known, the model's prediction accuracy can be determined by calculating the RMSE of these two sets of data.

[0127] S113: Evaluate the prediction effect based on the prediction error.

[0128] Error analysis is performed based on the calculated error indices (such as RMSE and MAE).

[0129] Judging the quality of prediction: Generally speaking, the smaller the RMSE and MAE, the better the prediction performance.

[0130] Based on historical data and model characteristics, a reasonable threshold can be set, such as an RMSE below a certain value, to indicate good predictive performance. Cross-validation can be used to improve model effectiveness.

[0131] In this embodiment, spatiotemporal matching and error analysis allow for a better understanding of the model's predictive capabilities, thereby improving the accuracy of El Niño predictions. Accurate El Niño predictions can help governments and relevant agencies develop climate response measures and reduce the negative impacts of El Niño.

[0132] In this embodiment, by combining domestic satellite data and historical reanalysis data, the prediction accuracy of El Niño can be improved, providing more reliable information support for climate research and industries such as agriculture and fisheries. Through real-time data assimilation and neural network transfer learning, it can quickly adapt to new climate conditions, improving the timeliness of predictions. This method achieves autonomous El Niño prediction, reducing reliance on human intervention and enabling continuous monitoring and prediction without human intervention. This prediction model is not only applicable to El Niño events but can also be extended to the study of other climate phenomena, such as La Niña and the monitoring of global climate change. This method provides a new approach for future meteorological forecasting and oceanographic modeling research, promoting the application of artificial intelligence technology in meteorology and oceanography.

[0133] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0134] Please see Figure 6 , Figure 6 This is a schematic diagram of the autonomous El Niño dynamic prediction device provided in the second embodiment of this application. The included units are used for execution... Figures 1-5 The steps in the corresponding embodiments. Please refer to the details. Figures 1-5 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 6 The El Niño dynamic autonomous prediction device 6 includes: The acquisition unit 610 is used to acquire historical reanalysis data and target surface sea surface temperature observation data; wherein the target surface sea surface temperature observation data is collected by domestic satellites; The first processing unit 620 is used to perform spatiotemporal matching of the target surface sea surface temperature observation data and the historical reanalysis data to construct a target matching dataset. The second processing unit 630 is used to pre-train the first neural network model based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. The third processing unit 640 is used to perform transfer learning on the pre-trained first neural network model based on the target matching dataset to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The fourth processing unit 650 is used to input the target surface sea surface temperature observation data into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data and reconstruct the target three-dimensional sea surface temperature data. The fifth processing unit 660 is used as an assimilation module based on the El Niño dynamic autonomous prediction model to assimilate the target three-dimensional sea surface temperature data into the ocean model and generate the target initial field. The sixth processing unit 670 is used to drive the El Niño dynamic autonomous prediction model to perform prediction integration with the target initial field and output the El Niño prediction result.

[0135] Furthermore, the acquisition unit is specifically used for: Obtain GODAS reanalysis data for the first historical time period and sea surface temperature observation data for the second historical time period; The surface sea surface temperature observation data for the second historical time period are preprocessed; the preprocessing includes, but is not limited to, format standardization and anomaly removal. The preprocessed surface sea surface temperature observation data is interpolated and completed based on the GODAS reanalysis data to obtain the target surface sea surface temperature observation data.

[0136] Furthermore, the second processing unit is specifically used for: Initialize the first neural network model based on the U-Net network architecture; The first neural network model based on the U-Net network architecture is pre-trained based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature; wherein the loss function is mean squared error and the optimizer is Adam.

[0137] Furthermore, the El Niño dynamic autonomous prediction device also includes: The seventh processing unit is used to control the El Niño dynamic autonomous prediction model to perform a 50-year free integration calculation, using the initial field of the 20th-century climate state as the driving condition. The eighth processing unit is used to monitor the energy exchange flux between the atmospheric component and the upper ocean component within the El Niño dynamic autonomous prediction model. The ninth processing unit is used to determine that the El Niño dynamic autonomous prediction model has reached a physical equilibrium state when the standard deviation of the energy exchange flux converges within a preset threshold range, activate the assimilation module, and input the target three-dimensional sea surface temperature data into the El Niño dynamic autonomous prediction model.

[0138] Furthermore, the fifth processing unit is specifically used for: The assimilation module based on the El Niño dynamic autonomous prediction model uses the target three-dimensional sea surface temperature data as the observation increment, and assimilates the observation increment into the background field of the ocean model using a Newtonian relaxation approximation method. The relaxation coefficient is dynamically adjusted according to the observation error covariance and the background error covariance to generate the target initial field.

[0139] Furthermore, the sixth processing unit is specifically used for: Using the target initial field as the initial condition, the autonomous El Niño dynamic prediction model is driven to perform prediction integration, outputting a multi-element spatiotemporal evolution sequence to obtain the El Niño prediction result; wherein, the integration period covers the next 12 months.

[0140] Furthermore, the El Niño dynamic autonomous prediction device also includes: The tenth processing unit is used to acquire historical El Niño observation data; The eleventh processing unit is used to perform spatiotemporal matching of the El Niño prediction results with historical El Niño observation data and calculate the prediction error; The twelfth processing unit is used to evaluate the prediction effect based on the prediction error.

[0141] Figure 7 This is a schematic diagram of the autonomous El Niño dynamics prediction device provided in the third embodiment of this application. Figure 7 As shown, the El Niño dynamic autonomous prediction device 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70, such as an El Niño dynamic autonomous prediction program. When the processor 70 executes the computer program 72, it implements the steps in the various El Niño dynamic autonomous prediction method embodiments described above, for example... Figure 1 Steps 101 to 107 are shown. Alternatively, when the processor 70 executes the computer program 72, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 6 The functions of modules 610 to 670 are shown.

[0142] For example, the computer program 72 can be divided into one or more modules / units, which are stored in the memory 71 and executed by the processor 70 to complete this application. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 72 in the El Niño dynamic autonomous prediction device 7. For example, the computer program 72 can be divided into an acquisition unit, a first processing unit, a second processing unit, a third processing unit, a fourth processing unit, a fifth processing unit, and a sixth processing unit, with the specific functions of each unit as follows: The acquisition unit is used to acquire historical reanalysis data and target surface sea surface temperature observation data; wherein the target surface sea surface temperature observation data is collected by domestic satellites; The first processing unit is used to perform spatiotemporal matching between the target surface sea surface temperature observation data and the historical reanalysis data to construct a target matching dataset. The second processing unit is used to pre-train the first neural network model based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. The third processing unit is used to perform transfer learning on the pre-trained first neural network model based on the target matching dataset to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The fourth processing unit is used to input the target surface sea surface temperature observation data into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data and reconstruct the target three-dimensional sea surface temperature data. The fifth processing unit is used as an assimilation module based on the El Niño dynamic autonomous prediction model to assimilate the target three-dimensional sea surface temperature data into the ocean model and generate the target initial field. The sixth processing unit is used to drive the El Niño dynamic autonomous prediction model to perform prediction integration with the target initial field and output the El Niño prediction result.

[0143] The autonomous El Niño dynamics prediction device may include, but is not limited to, a processor 70 and a memory 71. Those skilled in the art will understand that... Figure 7 This is merely an example of the El Niño autonomous prediction device 7 and does not constitute a limitation on the El Niño autonomous prediction device 7. It may include more or fewer components than shown, or combine certain components, or different components. For example, the El Niño autonomous prediction device may also include input / output devices, network access devices, buses, etc.

[0144] The processor 70 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.

[0145] The memory 71 can be an internal storage unit of the El Niño autonomous prediction device 7, such as a hard disk or memory. The memory 71 can also be an external storage device of the El Niño autonomous prediction device 7, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the El Niño autonomous prediction device 7. Furthermore, the El Niño autonomous prediction device 7 can include both internal storage units and external storage devices. The memory 71 is used to store the computer program and other programs and data required by the El Niño autonomous prediction device. The memory 71 can also be used to temporarily store data that has been output or will be output.

[0146] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0147] This application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.

[0148] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0149] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.

[0150] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0151] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0152] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0153] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0154] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0155] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An autonomous prediction method for El Niño dynamics, characterized in that, Including the following steps: Historical reanalysis data and target surface sea surface temperature observation data were acquired; wherein the target surface sea surface temperature observation data was collected by domestic satellites. The target surface sea surface temperature observation data is spatiotemporally matched with the historical reanalysis data to construct a target matching dataset; The first neural network model is pre-trained based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. Based on the target matching dataset, the pre-trained first neural network model is transferred to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The target surface sea surface temperature observation data is input into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data, and the target three-dimensional sea surface temperature data is reconstructed. The assimilation module based on the El Niño dynamic autonomous prediction model assimilates the target's three-dimensional sea surface temperature data into the ocean model to generate the target's initial field. The El Niño dynamic autonomous prediction model is driven by the target initial field to perform prediction integration and output the El Niño prediction result.

2. The autonomous prediction method for El Niño dynamics according to claim 1, characterized in that, The acquisition of historical reanalysis data and target surface sea surface temperature observation data includes the following steps: Obtain GODAS reanalysis data for the first historical time period and sea surface temperature observation data for the second historical time period; The surface sea surface temperature observation data for the second historical time period are preprocessed; the preprocessing includes, but is not limited to, format standardization and anomaly removal. The preprocessed surface sea surface temperature observation data is interpolated and completed based on the GODAS reanalysis data to obtain the target surface sea surface temperature observation data.

3. The autonomous prediction method for El Niño dynamics according to claim 1, characterized in that, The step of pre-training the first neural network model based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature includes the following steps: Initialize the first neural network model based on the U-Net network architecture; The first neural network model based on the U-Net network architecture is pre-trained based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature; wherein the loss function is mean squared error and the optimizer is Adam.

4. The autonomous prediction method for El Niño dynamics according to any one of claims 1 to 3, characterized in that, The assimilation module based on the El Niño dynamic autonomous prediction model, before assimilating the target three-dimensional sea surface temperature data into the ocean model and generating the target initial field, also includes the following steps: Using the initial field of the 20th-century climate state as the driving condition, the El Niño dynamic autonomous prediction model is controlled to perform a 50-year free integration calculation. Monitor the energy exchange flux between the atmospheric component and the upper ocean component within the autonomous El Niño dynamics prediction model; When the standard deviation of the energy exchange flux converges within a preset threshold range, it is determined that the El Niño dynamic autonomous prediction model has reached a physical equilibrium state, the assimilation module is activated, and the target three-dimensional sea surface temperature data is input into the El Niño dynamic autonomous prediction model.

5. The autonomous prediction method for El Niño dynamics according to any one of claims 1 to 3, characterized in that, The assimilation module based on the El Niño dynamic autonomous prediction model assimilates the target's three-dimensional sea surface temperature data into the ocean model to generate the target's initial field, including the following steps: The assimilation module based on the El Niño dynamic autonomous prediction model uses the target three-dimensional sea surface temperature data as the observation increment, and assimilates the observation increment into the background field of the ocean model using a Newtonian relaxation approximation method. The relaxation coefficient is dynamically adjusted according to the observation error covariance and the background error covariance to generate the target initial field.

6. The autonomous prediction method for El Niño dynamics according to any one of claims 1 to 3, characterized in that, The process of using the target initial field to drive the autonomous El Niño prediction model to perform prediction integration and output El Niño prediction results includes the following steps: Using the target initial field as the initial condition, the autonomous El Niño dynamic prediction model is driven to perform prediction integration, outputting a multi-element spatiotemporal evolution sequence to obtain the El Niño prediction result; wherein, the integration period covers the next 12 months.

7. The autonomous prediction method for El Niño dynamics according to any one of claims 1 to 3, characterized in that, It also includes the following steps: Obtain historical El Niño observation data; The El Niño prediction results are spatiotemporally matched with historical El Niño observation data to calculate the prediction error; The prediction effect is evaluated based on the prediction error.

8. An autonomous El Niño dynamic prediction device, characterized in that, include: The acquisition unit is used to acquire historical reanalysis data and target surface sea surface temperature observation data; wherein the target surface sea surface temperature observation data is collected by domestic satellites; The first processing unit is used to perform spatiotemporal matching between the target surface sea surface temperature observation data and the historical reanalysis data to construct a target matching dataset. The second processing unit is used to pre-train the first neural network model based on the historical reanalysis data to establish an initial mapping relationship between surface sea temperature and deep sea temperature. The third processing unit is used to perform transfer learning on the pre-trained first neural network model based on the target matching dataset to obtain a second neural network model that is adapted to the target surface sea surface temperature observation data. The fourth processing unit is used to input the target surface sea surface temperature observation data into the second neural network model to obtain the corresponding target deep sea surface temperature inversion data and reconstruct the target three-dimensional sea surface temperature data. The fifth processing unit is used as an assimilation module based on the El Niño dynamic autonomous prediction model to assimilate the target three-dimensional sea surface temperature data into the ocean model and generate the target initial field. The sixth processing unit is used to drive the El Niño dynamic autonomous prediction model to perform prediction integration with the target initial field and output the El Niño prediction result.

9. An autonomous El Niño dynamic prediction device, comprising: A processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the steps of the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images