Arctic atmosphere coupling forecasting method capable of automatically fusing sea ice concentration and thickness

Abstract

The invention provides a north pole atmosphere coupling forecasting method capable of automatically fusing sea ice concentration and thickness, which belongs to the technical field of meteorology, and comprises the following steps: firstly, acquiring GFS background field data and correcting by applying an ice-sea heat exchange correction equation; and then sea ice concentration and thickness information is extracted, and physical consistency verification is carried out based on a thermodynamic equilibrium equation. Performing multi-element fusion through an Arctic ice sea assimilation model, calculating the sea ice surface temperature by applying a polar region sea ice heat flux equation, and establishing a sea ice concentration thickness boundary layer interaction equation set to describe the modulation effect of sea ice on a boundary layer; the method comprises the core steps of performing deep fusion on various sea ice features by applying a feature fusion model, adaptively adjusting the weight based on an ice-gas interface balance index, generating an optimized mode initial field and boundary conditions, and finally setting a parameter scheme to execute an integral program to obtain a forecasting result. The technical problem of low forecasting accuracy caused by insufficient physical consistency of sea ice information is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorology, and in particular relates to an Arctic atmosphere coupled forecasting method that automatically integrates sea ice density and thickness. Background Art

[0002] Numerical weather forecasting in the Arctic region is a key area of ​​meteorological research, crucial for polar expeditions, navigation planning, and climate change studies. Traditional Arctic weather forecasts rely primarily on background field data provided by the Global Forecast System (GFS), assimilated with limited observational data, and then fed into numerical models. Existing techniques typically use sea ice concentration as a boundary condition input into the model, simulating its impact on the atmospheric boundary layer using simplified sea ice parameterization schemes.

[0003] However, traditional techniques have significant drawbacks when processing sea ice information. For one thing, existing methods typically only consider sea ice concentration, ignoring the crucial role of sea ice thickness. This results in models failing to accurately describe sea ice heat capacity and its modulation of the atmospheric boundary layer. Furthermore, even when both sea ice concentration and thickness information are incorporated, physical consistency between the two is difficult to ensure, as they are often provided by different data sources and lack effective physical constraints.

[0004] This technical problem of insufficient physical consistency of sea ice density and thickness information leads to significant deviations in the simulation of the atmospheric boundary layer structure in the Arctic region. In particular, in areas where sea ice changes rapidly, errors in the calculation of boundary layer turbulent fluxes accumulate, which in turn affects the accuracy of large-scale circulation simulations, ultimately resulting in insufficient accuracy in weather forecasts in the Arctic region, which cannot meet the high-precision requirements of applications such as polar scientific research and route planning. Summary of the Invention

[0005] In view of this, the present invention provides an Arctic atmosphere coupling forecasting method that automatically integrates sea ice density and thickness, which can solve the technical problem in the prior art that the accuracy of atmospheric coupling forecasting in the Arctic region is low due to insufficient physical consistency of sea ice density and thickness information.

[0006] The present invention is implemented as follows: The present invention provides an Arctic atmospheric coupling forecasting method that automatically integrates sea ice density and thickness, including obtaining and correcting global GFS background field data, extracting and verifying sea ice density and sea ice thickness information, performing multi-factor fusion to generate a background field, applying the polar sea ice heat flux equation to calculate the sea ice surface temperature, establishing a sea ice density thickness boundary layer interaction equation group to calculate the boundary layer structure, applying a feature fusion model to deeply fuse sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure, adaptively adjusting the weight of each element in the feature fusion model output vector based on the ice-air interface balance index, generating an optimized model initial field and boundary conditions, executing an integration program to obtain forecast results and performing post-processing.

[0007] Among them, the step of obtaining and correcting global GFS background field data specifically involves obtaining global GFS background field data, which contains 41 layers in the vertical direction, a time interval of 3 hours, and a single forecast duration of 168 hours, and applying the ice-sea heat exchange correction equation to preliminarily correct the boundary layer parameters of the sea ice area in the GFS background field data.

[0008] Among them, the step of extracting and verifying sea ice density and sea ice thickness information specifically involves extracting sea ice density and sea ice thickness information from the GFS background field data, performing physical consistency verification on the sea ice density and sea ice thickness information based on the sea ice thermodynamic equilibrium equation, and rewriting the verified sea ice density and sea ice thickness information into an intermediate file that can be recognized by the model pre-processing program.

[0009] Among them, the step of executing multi-factor fusion to generate a background field specifically includes setting a model pre-processing parameter file, executing a terrain processing program, a format rewriting program and an interpolation program, and fusing meteorological elements, the sea ice density and the sea ice thickness information through the Arctic ice sea assimilation model to generate a model background field with higher physical consistency.

[0010] Among them, the step of applying the feature fusion model to deeply fuse the sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure specifically involves applying the feature fusion model to deeply fuse the sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure, and adaptively adjusting the weight of each element in the output vector of the feature fusion model based on the ice-air interface balance index.

[0011] Among them, the step of executing the integration program to obtain the forecast results and performing post-processing specifically includes setting the physical parameterization scheme and the sea ice parameterization scheme, using the model initial field and the boundary conditions to execute the integration program to obtain the weather forecast results for the next 7 days, and using the ensemble forecasting technology to evaluate the uncertainty impact of the sea ice concentration and the sea ice thickness on the weather forecast results.

[0012] Among them, the step of executing the integration program to obtain the forecast results and performing post-processing also includes post-processing the meteorological forecast results, comparing historical observation data to calculate the forecast skill score, and generating standardized forecast products according to the application needs of the Arctic region.

[0013] Among them, the GFS background field data refers to the gridded forecast data generated by the global forecast system, which includes multiple layer elements of the atmosphere, ocean, land surface and sea ice. The GFS background field data provides initial information for subsequent processing.

[0014] Among them, the ice-air interface balance index is used to evaluate the balance state of heat, momentum and water vapor exchange during the interaction between sea ice and the atmosphere. The ice-air interface balance index is obtained by calculating the product of the sea ice density and the sea ice thickness and dividing it by the surface net radiation flux. When the ice-air interface balance index is less than 0.3, it indicates a weak exchange process; when it is greater than 0.7, it indicates a strong exchange process; and between the two, it indicates a moderate exchange process.

[0015] Among them, the feature fusion model is a deep learning model based on multi-head self-attention mechanism and spatiotemporal graph convolution, which is used to optimize the fusion of the sea ice density, the sea ice thickness, the sea ice surface temperature and the sea ice modulated boundary layer structure.

[0016] The present invention achieves a deep fusion of sea ice information and atmospheric forecasts through multi-step systematic processing. The method first performs a physical consistency check on the sea ice density and thickness information based on the sea ice thermodynamic balance equation, and then uses the Arctic ice-sea assimilation model to perform a multi-factor fusion of the verified sea ice information and meteorological elements to generate a model background field with higher physical consistency. By establishing a set of sea ice density and thickness boundary layer interaction equations, the present invention achieves an accurate description of the sea ice modulated boundary layer structure, and applies a feature fusion model to deeply fuse the sea ice density, thickness, surface temperature and boundary layer structure, and adaptively adjusts the feature weights based on the ice-air interface balance index, thereby solving the problem of insufficient physical consistency of sea ice information in traditional methods. The feature fusion model considers both prediction accuracy and physical consistency constraints during the training process, ensuring that the fusion results are both in line with physical laws and have high-precision characteristics. The present invention successfully solves the technical problem of low accuracy of atmospheric coupling forecasts in the Arctic region due to insufficient physical consistency of sea ice density and thickness information, and provides a more reliable method for weather forecasting in the Arctic region. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of the method of the present invention.

[0018] Figure 2 This is the Taylor diagram of the forecast elements in Example 2.

[0019] Figure 3 This is the forecast result map of various elements in the Arctic region on August 11, 2021 in Example 2, including 4 sub-maps, namely (a) 2-meter temperature forecast map, (b) 2-meter dew point temperature forecast map, (c) sea level pressure forecast map, and (d) 10-meter wind speed forecast map. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0021] like Figure 1 FIG. 1 is a flow chart of an Arctic atmospheric coupled forecasting method for automatically integrating sea ice density and thickness provided by the present invention. The method includes the following steps: S01. Use the API interface to call the Global Meteorological Center server to obtain GFS background field data, which includes multiple pressure layers from the surface to the upper air. Apply the Arctic Ocean heat exchange correction equation to make preliminary corrections to the boundary layer parameters in the sea ice area and improve the description of the boundary layer parameters in the polar region. S02. Extract sea ice density and thickness information from the GFS background field data, perform physical consistency verification based on the sea ice thermodynamic equilibrium equation, and adjust the sea ice thickness through iterative methods when the deviation between the theoretical value and the sea ice thickness is too large. The verified data is rewritten into an intermediate file that can be recognized by the model pre-processing program. S03. Set the model pre-processing parameter file, adopt the projection method and resolution setting suitable for the Arctic region, execute the terrain processing program, format rewriting program and interpolation program, and use the Arctic ice sea assimilation model using four-dimensional variational assimilation technology to perform multi-factor fusion to generate a model background field with higher physical consistency; S04. Configure the sea ice thermodynamic condition parameters and the sea-air momentum exchange coefficient in the model parameters, apply the polar sea ice heat flux equation to calculate the sea ice surface temperature, and comprehensively consider various radiation and heat flux processes to determine the temperature state of the ice-air interface; S05. Establish a set of interaction equations for sea ice density and thickness boundary layer, taking into account the modulation effect of sea ice on the atmospheric boundary layer structure, calculate sea ice surface roughness, thermal stability parameters, boundary layer turbulent fluxes, boundary layer height, and temperature and humidity gradients, and generate the sea ice modulated boundary layer structure; S06. Apply the feature fusion model to deeply fuse sea ice density, thickness, surface temperature, and boundary layer structure, adaptively adjust feature weights based on the ice-air interface balance index, optimize the multi-source feature fusion process, and generate high-quality model initial fields and boundary conditions. S07. Set up the physical parameterization scheme and the sea ice parameterization scheme, use the initial field and boundary conditions generated in the previous step to execute the integration procedure to obtain the meteorological forecast results, and use the ensemble forecasting technique to evaluate the uncertainty impact of the sea ice parameters on the forecast results; S08. Post-process the meteorological forecast results, including error correction and product generation, calculate the forecast skill score by comparing with historical observation data, and generate standardized forecast products based on the application needs of the Arctic region.

[0022] Among them, GFS background field data refers to the gridded forecast data generated by the global forecast system that includes multiple layer elements of the atmosphere, ocean, land surface and sea ice. The GFS background field data provides initial information for subsequent processing.

[0023] Among them, sea ice density refers to the percentage of area covered by sea ice in a unit sea area, and the value range is 0 to 1, where 0 means no sea ice and 1 means complete sea ice coverage. The sea ice density affects the sea-air heat exchange process.

[0024] The sea ice thickness refers to the vertical distance from the bottom to the top of the sea ice, measured in meters. The sea ice thickness affects the thermal inertia and conduction characteristics of the sea ice.

[0025] Among them, the ice-sea heat exchange correction equation is used to correct the boundary layer parameter deviation of the standard atmospheric model under polar sea ice conditions. The input includes the standard calculated boundary layer turbulent flux, sea ice impact factor, sea ice density and sea ice thickness, and the output is the corrected boundary layer turbulent flux. The corrected boundary layer turbulent flux is used to optimize the model initial conditions.

[0026] Among them, the sea ice thermodynamic equilibrium equation is the basic equation that describes the heat change and phase change process of sea ice. The input parameters include sea ice thermal conductivity, sea ice salinity, the sea ice surface temperature, the sea ice bottom temperature and the sea ice internal temperature profile. The output parameters are the sea ice thickness change rate and the internal temperature distribution state. The sea ice thickness change rate and the internal temperature distribution state are used to verify the physical consistency of the sea ice density and the sea ice thickness.

[0027] Among them, the polar sea ice heat flux equation is used to calculate the sea ice surface temperature and the energy exchange process at the ice-air interface, accurately describing the thermodynamic equilibrium state of the Arctic sea ice region. Based on the principle of energy balance, this equation comprehensively considers multiple physical processes, including solar shortwave radiation, atmospheric longwave radiation, sea ice surface emitted radiation, and sensible heat, latent heat, and conductive heat exchange. Inputs include physical quantities such as sea ice albedo, downward shortwave radiation flux, sea ice emissivity, downward longwave radiation flux, Stefan-Boltzmann constant, near-surface wind speed, near-surface air temperature, near-surface specific humidity, saturation specific humidity, sea ice thermal conductivity, seawater freezing temperature, and sea ice thickness. The output is the sea ice surface temperature, a key physical parameter obtained by solving the energy balance equation that directly affects the heat exchange process at the ice-air interface.

[0028] The sea ice concentration-thickness boundary layer interaction equations are used to describe the modulation of sea ice on the structure of the atmospheric boundary layer, establish a quantitative relationship between sea ice properties and atmospheric boundary layer parameters, and implement coupled sea ice-atmosphere simulations. This system of equations considers the influence of ice ridge structure on surface roughness and the regulation of atmospheric thermodynamic stability on turbulent exchange processes. Inputs include physical quantities such as sea ice thickness, sea ice concentration, gravitational acceleration, virtual potential temperature and its vertical difference, height difference, wind speed difference, air density, drag coefficient, near-surface wind speed, air constant-pressure specific heat capacity, sensible heat exchange coefficient, and near-surface air temperature. Outputs include sea ice surface roughness, boundary layer height, temperature and humidity gradient distribution within the boundary layer, and boundary layer turbulent fluxes, including momentum flux, sensible heat flux, and latent heat flux. These parameters serve as fundamental data for describing sea ice-atmosphere interaction processes and constructing boundary layer parameterization schemes.

[0029] Among them, the ice-air interface balance index is used to evaluate the balance state of heat, momentum and water vapor exchange during the interaction between sea ice and the atmosphere. The ice-air interface balance index is obtained by calculating the product of the sea ice density and the sea ice thickness and dividing it by the surface net radiation flux. When the ice-air interface balance index is less than 0.3, it indicates a weak exchange process; when it is greater than 0.7, it indicates a strong exchange process; and between the two, it indicates a moderate exchange process. The ice-air interface balance index is used to adjust the weight of each element in the output vector of the feature fusion model.

[0030] Among them, the feature fusion model is a deep learning model based on the multi-head self-attention mechanism and spatiotemporal graph convolution, which is used to optimize the fusion of the sea ice density, the sea ice thickness, the sea ice surface temperature and the sea ice modulated boundary layer structure; the specific structure of the feature fusion model is that the input layer receives the feature data such as the sea ice density, the sea ice thickness, the sea ice surface temperature and the sea ice modulated boundary layer structure, and captures the correlation between different variables through the multi-head self-attention layer, and then extracts the physical relationship pattern in the local area through the spatiotemporal graph convolution layer, followed by the multi-scale feature fusion layer to integrate the feature representations of different spatial scales, and finally the fully connected layer outputs the optimized feature representation. The entire network adopts a residual connection structure to ensure stable gradient propagation and complete retention of feature information.

[0031] Among them, the steps of establishing the training data set of the feature fusion model specifically include collecting historical Arctic exploration satellite data and field observation data, normalizing them according to spatial resolution and temporal resolution, extracting key features such as the sea ice density, the sea ice thickness, the sea ice surface temperature, the near-surface air temperature and the near-surface wind speed, constructing spatiotemporal sequence sample pairs, and screening and eliminating unreasonable data according to the principle of physical conservation, and finally obtaining a training data set containing multi-scale spatiotemporal features, in which the validation set and test set account for 20% and 10% of the total data volume, respectively.

[0032] Among them, the steps of feature fusion model training specifically include first initializing the network parameters and using pre-trained physical knowledge to guide the weight distribution, then using the small batch stochastic gradient descent method to optimize the network parameters, the loss function considers both prediction accuracy and physical consistency constraints, and the learning rate adopts the cosine annealing strategy. During the training process, the weight distribution of the feature fusion layer is dynamically adjusted according to the ice-air interface balance index to enhance the expression ability of the index abnormal area, and avoid overfitting through the early stopping method, and finally a model with both physical interpretability and predictive accuracy is obtained.

[0033] The near-surface wind speed refers to the horizontal wind speed measured at a height of 10 meters from the ground surface, in meters per second. The near-surface wind speed is obtained from the GFS background field data.

[0034] The net surface radiation flux refers to the difference between the solar radiation received by the surface and the long-wave radiation emitted by the surface, and its unit is watts per square meter. The net surface radiation flux is calculated by the polar sea ice heat flux equation.

[0035] The snow cover thickness refers to the thickness of the snow layer covering the sea ice surface, measured in meters. The snow cover thickness is obtained from the GFS background field data.

[0036] The model initial field refers to the starting state field where the numerical model starts the integral calculation, including physical quantities such as wind, temperature, humidity, and air pressure of each atmospheric layer. The model initial field is generated by the feature fusion model.

[0037] The boundary conditions refer to the side boundary and bottom boundary conditions required in the numerical model integration process, and the boundary conditions are generated by the feature fusion model.

[0038] Among them, the meteorological forecast results refer to the forecast of future weather conditions obtained by model integration calculation, including factors such as temperature, humidity, wind direction and speed, air pressure and precipitation. The meteorological forecast results are used for subsequent forecast product generation.

[0039] Among them, the Arctic ice sea assimilation model is used to fuse multi-source data of meteorological elements, the sea ice density and the sea ice thickness information, and realize multi-variable and multi-scale collaborative updating based on four-dimensional variational assimilation technology. The model background field with higher physical consistency is output by the Arctic ice sea assimilation model.

[0040] The near-surface air temperature refers to the atmospheric temperature measured at a height of 2 meters from the ground surface, in degrees Celsius. The near-surface air temperature is obtained from the GFS background field data.

[0041] The relative humidity refers to the percentage of the water vapor content in the air to the saturated water vapor content at the same temperature, and the relative humidity is obtained from the GFS background field data.

[0042] The downward shortwave radiation flux refers to the solar radiation energy reaching the earth's surface, with the unit of watts per square meter. The downward shortwave radiation flux is obtained from the GFS background field data.

[0043] The downward longwave radiation flux refers to the longwave radiation energy emitted downward by the atmosphere and reaching the earth's surface, with the unit being watts per square meter. The downward longwave radiation flux is obtained from the GFS background field data.

[0044] The specific implementation of the above steps is described in detail below. Step S01 is implemented by invoking the Global Weather Service (GFS) server via an API interface to obtain GFS background field data. This data covers the entire global geographic area and vertically includes 41 non-uniformly distributed pressure layers from the surface to approximately 0.01 hPa. The horizontal resolution is 0.25° × 0.25°, the temporal resolution is 3 hours, and the single forecast duration is 168 hours. After acquiring the data, the boundary layer parameters in the sea ice region are initially corrected using the ice-sea heat exchange correction equation. This correction equation is based on the Monnin-Obukhov similarity theory and considers the influence of sea ice characteristics on boundary layer stability and flux exchange. Sea ice concentration and thickness are introduced as key parameters. The correction coefficient is 0.85 for sea ice concentrations greater than 0.7, 0.92 for sea ice concentrations between 0.3 and 0.7, and 0.98 for sea ice concentrations less than 0.3. The purpose of this step is to obtain high-quality background field data and preliminarily improve the description of polar boundary layer parameters, laying the foundation for subsequent processing.

[0045] The specific implementation method of step S02 is to extract sea ice density and sea ice thickness information from the GFS background field data, and the extraction adopts a bilinear interpolation algorithm to ensure spatial continuity. Then, the extracted sea ice density and thickness information are physically checked for consistency based on the sea ice thermodynamic equilibrium equation. The verification process first calculates the theoretical sea ice thickness value under thermodynamic equilibrium conditions. When the relative deviation between the actual sea ice thickness and the theoretical value exceeds 20%, the sea ice thickness is adjusted by an iterative method to satisfy the thermodynamic constraints. The coordination of sea ice density and sea ice thickness has also been tested. For newly formed thin ice areas, the positive correlation between sea ice density and thickness should be satisfied, that is, when the sea ice thickness is less than 0.3m, the sea ice density should not exceed 0.85. The verified sea ice density and thickness data are rewritten in binary format into an intermediate file that can be recognized by the model pre-processing program, and a lossless compression algorithm is used to reduce storage space. The purpose of this step is to ensure the physical consistency of the sea ice parameters and avoid non-physical solutions during the model integration process.

[0046] The specific implementation of step S03 involves setting the model pre-processing parameter file, including key parameters such as model resolution, number of vertical layers, projection method, and computational domain. For the Arctic region, a polar stereographic projection is used to minimize grid distortion. The horizontal resolution is set to 15 km, the vertical layer is set to 57 layers, and the top layer height is set to 10 hPa. When executing the terrain processing program, a Gaussian filter is used to smooth the terrain with a filter radius of four grid points to eliminate numerical noise. The format rewriting program converts all input data into a model-specific binary format, and the interpolation program uses a conformal interpolation algorithm to ensure that physical quantities are conserved during the spatial interpolation process. The Arctic ice sea assimilation model uses a four-dimensional variational assimilation technique, introducing a background error covariance matrix to describe the correlation between different variables. The covariance structure is calculated using the NMC method. Considering the specificity of the sea ice region, a flow-related length scale adjustment algorithm is introduced to enable the assimilation process to better preserve the frontal characteristics of the sea ice edge. This step generates a model background field with higher physical consistency through multi-factor fusion, providing high-quality initial conditions for subsequent model integration.

[0047] The specific implementation of step S04 is to configure the sea ice thermodynamic condition parameters and the sea-air momentum exchange coefficient in the model parameters. The sea ice thermodynamic parameters include the sea ice thermal conductivity set to 2.0W / (m·K), the sea ice albedo dynamically adjusted according to the sea ice thickness, with a value between 0.5 and 0.7 when the thickness is less than 0.5m, and a value between 0.7 and 0.85 when the thickness is greater than 0.5m. The sea-air momentum exchange coefficient adopts the modified Charnock relationship, and the roughness length parameter takes a value of 0.0012 when the sea ice density is greater than 0.6 and a value of 0.0018 when it is less than 0.6. The polar sea ice heat flux equation is used to calculate the sea ice surface temperature. This equation is based on the energy balance principle and takes into account multiple physical processes such as solar radiation, atmospheric long-wave radiation, sea ice surface emitted radiation, and sensible and latent heat exchange. The sea ice surface temperature is obtained by solving the nonlinear equations using the Newton iteration method. During the calculation, when the sea ice thickness is less than 0.1m, the effect of seawater heat flux penetrating the sea ice is taken into account, and a heat permeability coefficient of 0.3 is introduced; when the sea ice thickness is greater than 1.5m, the vertical stratification is no less than 5 layers to accurately describe the internal temperature profile. The purpose of this step is to accurately describe the heat exchange process between the sea ice thickness and the atmosphere, providing key input for the boundary layer parameterization. Among them, the Charnock relationship is a dimensionless relationship in fluid mechanics that is used to describe the characteristics of liquid surface waves. It was proposed by British fluid mechanic John Charnock in 1955 and is mainly used in the fields of marine meteorology and ocean engineering.

[0048] The specific implementation of step S05 is to establish a set of interactive equations for sea ice density and thickness boundary layer, which is based on the modified boundary layer theory and takes into account the modulation effect of sea ice on the structure of the atmospheric boundary layer. First, the surface roughness of the sea ice is calculated. The roughness length is related to the sea ice density, ice ridge height and distribution. The ice ridge height is associated with the sea ice thickness through an empirical relationship, with a typical value between 0.01m and 0.05m. Then, the thermal stability parameter is calculated and characterized by the Richardson number. When the Richardson number is greater than 0.25, the boundary layer is in a strongly stable state, and when the Richardson number is less than 0, it is in an unstable state. The turbulent flux is calculated based on the stability-corrected K theory, and the boundary layer height is determined by the critical Richardson number method. Usually, the stable boundary layer height on sea ice is 100m to 300m. The temperature gradient and humidity gradient are calculated using the discrete difference method. The vertical resolution is encrypted at the bottom of the boundary layer, and the thickness of the first layer is 5m. The sea ice modulated boundary layer structure generated in this step contains key parameters such as boundary layer height, temperature gradient and humidity gradient, which are used as input for subsequent feature fusion models to accurately describe the impact of sea ice on boundary layer processes.

[0049] The specific implementation of step S06 involves applying a feature fusion model to deeply fuse sea ice density, sea ice thickness, sea ice surface temperature, and sea ice-modulated boundary layer structure. The feature fusion model first normalizes the input features using the Z-score standardization method. It then uses a multi-head self-attention mechanism to capture the spatiotemporal correlations between different variables. The number of attention heads is set to 8, and the hidden layer dimension is 256. The spatiotemporal graph convolution layer uses a three-layer stacked structure with a 3×3 convolution kernel size. The number of output channels per layer is 128, 64, and 32, and the activation function uses LeakyReLU. The multi-scale feature fusion layer integrates feature representations at different spatial scales using a pyramid pooling network, with pooling scales of 1×1, 2×2, 4×4, and 8×8, respectively. The weight of each element in the feature fusion model's output vector is adaptively adjusted based on the ice-air interface balance index. When the index is less than 0.3, the weight of the sea ice concentration feature is increased by 30%; when the index is greater than 0.7, the weight of the sea ice surface temperature feature is increased by 25%. When the index is between 0.3 and 0.7, the weights of all features remain balanced. This step optimizes the multi-source feature fusion process through deep learning methods, generating more physically consistent model initial fields and boundary conditions, providing high-quality input for model integration.

[0050] The specific implementation of step S07 involves setting physical parameterization schemes and sea ice parameterization schemes. Physical parameterization schemes include the RRTMG long-short wave scheme for radiation, the YSU scheme for boundary layer, the Kain-Fritsch scheme for cumulus convection, and the WSM6 scheme for microphysics. Sea ice parameterization schemes include the use of multilayer nonlinear diffusion equations for heat conduction, consideration of salinity effects for phase transitions, and elastic-plastic rheological methods for dynamic processes. The integration procedure is executed using the model initial fields and boundary conditions generated in step S06, with a time step of 45 seconds, an integration period of 168 hours, and an output frequency of 1 hour. Ensemble forecasting techniques are used to assess the uncertainty impact of sea ice concentration and sea ice thickness on meteorological forecast results. Ensemble members are generated by perturbing sea ice concentration by ±10% and sea ice thickness by ±20%. Ensemble forecast results describe forecast uncertainty using a probability distribution function. When the ensemble dispersion exceeds a preset threshold, the system automatically marks the region as high uncertainty. The threshold is set to a standard deviation greater than 1.5 times the climatological mean. The purpose of this step is to obtain weather forecasts for the next seven days and quantify forecast uncertainties. Specifically, the RRTMG long- and short-wave radiation scheme is a fast and accurate atmospheric radiative transfer model. By accounting for the absorption spectra of various gas components and aerosol effects, it calculates the radiative flux and heating rate of the atmosphere, balancing computational accuracy and efficiency. The YSU boundary layer scheme, a nonlocal K-theory closed model, explicitly addresses turbulent mixing and top entrainment within the atmospheric boundary layer, improving the simulation of heat and water vapor exchange between the nighttime boundary layer and the free atmosphere. The Kain-Fritsch cumulus convection scheme, based on a Lagrangian mass flux framework, simulates convective updrafts and downdrafts, accounting for cloud-environment mixing and precipitation processes, making it particularly suitable for precipitation simulations at mesoscale grids. The WSM6 microphysics scheme, a complex single-moment parameterization scheme, tracks the formation, transformation, and deposition of six types of hydrometeors (water vapor, cloud water, rain, cloud ice, snow, and graupel), providing a comprehensive description of cold and warm cloud processes and the microphysical processes of mixed-phase clouds.

[0051] The specific implementation of step S08 involves post-processing the meteorological forecast results, including systematic error correction, downscaling, and product generation. Systematic error correction utilizes a quantile mapping method, constructing an error correction model based on the statistical relationship between historical forecasts and observations. The training sample length is at least 365 days. Forecast skill scores are calculated by comparing historical observational data. Score metrics include root mean square error (RMS), bias, correlation coefficient, and forecast skill score. For Arctic region forecast skill score thresholds require a 72-hour forecast skill of no less than 0.7 for the 500hPa geopotential height field and a 48-hour forecast skill of no less than 0.65 for surface temperature. Standardized forecast products are generated based on application needs in the Arctic region. Product types include general meteorological element forecast maps, sea ice concentration change forecast maps, extreme weather event probability forecast maps, and navigation index forecast maps. Product resolution is dynamically adjusted based on user needs, with a standard resolution of 0.1° × 0.1°. Product formats support various formats, including GRIB2, NetCDF, and GeoTIFF. The purpose of this step is to convert raw forecast results into user-friendly forecast products and assess forecast quality through skill scores.

[0052] The detailed structure of the feature fusion model employs a deep neural network framework and consists of the following main components: The input layer receives normalized feature data such as sea ice density, sea ice thickness, sea ice surface temperature, and sea ice-modulated boundary layer structure, with feature dimensions of [batch size, time step, spatial height, spatial width, number of feature channels]. The multi-head self-attention layer consists of eight attention heads, each of which independently learns the correlation patterns between different features. The input features are first linearly transformed to generate query, key, and value matrices, then attention weights are calculated. Finally, the outputs of each head are concatenated and linearly transformed to obtain the output features. The spatiotemporal graph convolution layer consists of three stacked graph convolutional networks. Each layer combines spatial and temporal convolutions, with a spatial convolution kernel size of 3×3 and a temporal convolution kernel size of 3. Batch normalization and LeakyReLU activation functions are used between layers. The multi-scale feature fusion layer integrates information from different spatial scales using a pyramid pooling network, performing 1×1, 2×2, 4×4, and 8×8 pooling operations, respectively. The pooled results are then upsampled back to the original resolution and concatenated. The fully connected layer maps the fused features to the output space, using a residual connection structure to ensure stable gradient propagation and complete preservation of feature information. The final output layer dimensions are consistent with the format of the initial field and boundary conditions. The entire network is trained with a batch size of 64, an initial learning rate of 0.001, a cosine annealing strategy, and 300 training rounds.

[0053] The detailed steps for constructing the training dataset for the feature fusion model include: First, Arctic satellite data are collected from multiple sources, including AMSR2 sea ice concentration products with a resolution of 6.25 km; CryoSat-2 and SMOS joint inversions of sea ice thickness with a resolution of 25 km; and MODIS sea ice surface temperature products with a resolution of 1 km. Field data from polar research stations, buoys, and ships are also collected, including boundary layer temperature profiles, humidity profiles, and wind speed profiles. The collected data are resampled to a 15 km grid based on spatial resolution and a daily temporal resolution, using bilinear interpolation to ensure conservation of physical quantities. Key features such as sea ice concentration, sea ice thickness, sea ice surface temperature, near-surface air temperature, and near-surface wind speed are extracted from the processed data. A sliding window sample pair with a 30-day time window is constructed, with the input being the features from the first 27 days and the output being the features from the last three days. Irrational data was screened and eliminated based on the principles of physical conservation. This included checking whether sea ice density was within the range of 0 to 1, whether sea ice thickness was positive and the concentration of thin ice areas did not exceed a threshold, and whether surface temperatures met thermodynamic constraints. Samples that did not meet these criteria were removed. The resulting training dataset contained multi-scale spatiotemporal features, with a total sample size of no less than 10,000. The validation set and test set accounted for 20% and 10% of the total data, respectively, to ensure the model's good generalization capabilities.

[0054] It should be noted that the model preprocessing program receives the GFS background field data obtained through the API in step S01. This data covers the entire globe, includes 41 non-uniformly distributed pressure layers in the vertical direction, has a horizontal resolution of 0.25° × 0.25°, and a temporal resolution of 3 hours. The program first applies the ice-sea heat exchange correction equation to perform a preliminary correction of the boundary layer parameters in the sea ice region, applying different correction factors (0.85, 0.92, and 0.98) depending on the sea ice concentration range (>0.7, 0.3-0.7, and <0.3). In step S02, the program extracts sea ice concentration and thickness information from the GFS background field and performs a physical consistency check using the sea ice thermodynamic equilibrium equation. If the actual sea ice thickness deviates from the theoretical value by more than 20%, it is iteratively adjusted to meet thermodynamic constraints. For areas of newly formed thin ice (thickness <0.3 m), the sea ice concentration is limited to no more than 0.85. The verified data is rewritten into a binary intermediate file for subsequent processing.

[0055] It should be noted that the terrain processing program is responsible for processing the terrain data required for the model and is an important part of the model pre-processing in step S03. For the Arctic region, the program uses polar stereographic projection to reduce grid deformation and sets the horizontal resolution to 15km. During the terrain processing process, the program applies a Gaussian filter to smooth the terrain, and the filter radius is set to 4 grid point distances in order to eliminate small-scale terrain features that may cause numerical noise. This processing ensures that the terrain data meets the stability requirements of the numerical model and avoids computational instability caused by excessive terrain gradients. The smooth terrain field output by the terrain processing program will directly affect the boundary layer parameterization and surface process simulation in the model, which is particularly important for the simulation of the Arctic sea ice region.

[0056] It should be noted that the format rewriting program uniformly converts all types of input data into a model-specific binary format to ensure that the data structure is consistent with the internal representation of the model. The interpolation program uses a conformal interpolation algorithm to ensure that the physical quantities are conserved during the spatial interpolation process, which is crucial for accurately describing the physical characteristics of the sea ice edge area. The Arctic ice sea assimilation model uses four-dimensional variational assimilation technology, introduces a background error covariance matrix to describe the correlation between different variables, and the covariance structure is calculated using the NMC method. At the same time, in view of the particularity of the sea ice area, the program introduces a flow-related length scale adjustment algorithm to enable the assimilation process to better maintain the frontal characteristics of the sea ice edge. The collaborative work of these programs ensures the fusion processing of multi-source data and generates a model background field with higher physical consistency.

[0057] It should be noted that the integration program uses the model initial fields and boundary conditions generated in step S06 and performs time-integrated calculations based on the specified physical parameterization schemes (RRTMG long-shortwave scheme for radiation, modified YSU scheme for boundary layer, Kain-Fritsch scheme for cumulus convection, and WSM6 scheme for microphysics) and sea ice parameterization schemes (multilayer nonlinear diffusion equations for heat conduction, consideration of salinity effects for phase transitions, and elastic-plastic rheological methods for dynamic processes). The integration program has a time step of 45 seconds, an integration period of 168 hours (7 days), and an output frequency of 1 hour. The program also implements an ensemble forecast technique, generating 21 ensemble members by perturbing sea ice concentration by ±10% and sea ice thickness by ±20%. This is used to assess the impact of sea ice parameters on meteorological forecast uncertainty. Ensemble forecast results are characterized by probability distribution functions (PDFs). When the ensemble dispersion exceeds a preset threshold (standard deviation greater than 1.5 times the climatological mean), the system automatically marks the region as high uncertainty.

[0058] Specifically, the present invention is based on the physical mechanisms of the sea ice-atmosphere coupled system and the theory of multi-source data fusion. Its core lies in the construction of a complete sea ice information processing chain and sea ice-atmosphere interaction model, achieving full process optimization from data acquisition, physical consistency verification, to multi-factor fusion.

[0059] First, the present invention uses the ice-sea heat exchange correction equation to perform a preliminary correction of the boundary layer parameters in the sea ice region of the GFS background field data, resolving the systematic bias of the Standard Atmospheric Model in the polar regions. A key innovation lies in applying the sea ice thermodynamic equilibrium equation to perform a physical consistency check on the sea ice concentration and thickness information, ensuring that these two key parameters conform to the fundamental laws of thermodynamics. This step overcomes the physical contradiction caused by the inconsistent sources of sea ice concentration and thickness information in traditional methods, laying a solid foundation for subsequent processing.

[0060] Secondly, the present invention establishes a polar sea ice heat flux equation and a sea ice concentration-thickness boundary layer interaction equation system to calculate sea ice surface temperature and sea ice-modulated boundary layer structure, respectively. These two equation systems fully account for the combined influence of multiple physical factors, including sea ice concentration and thickness, near-surface wind speed, surface net radiation flux, and snow cover thickness. They accurately describe the energy and momentum exchange processes at the sea ice-atmosphere interface, thereby providing boundary conditions that are more realistic.

[0061] The most innovative feature is the feature fusion model proposed in this paper. Based on a multi-head self-attention mechanism and spatiotemporal graph convolution, it is capable of learning the complex nonlinear relationships between sea ice density, thickness, surface temperature, and boundary layer structure. This model addresses the problem of insufficient model generalization under varying sea ice conditions by introducing an ice-air interface balance index as an adaptive weight adjustment mechanism. During training, the loss function simultaneously considers both prediction accuracy and physical consistency constraints, ensuring that the model output conforms to historical observations while not violating fundamental physical laws. This theoretically guarantees the physical consistency of sea ice information and the reliability of forecast results.

[0062] In summary, this invention establishes a complete physical link of sea ice density, thickness and atmospheric coupling through systematic sea ice information processing and deep learning-driven feature fusion, effectively solving the problem of insufficient physical consistency of sea ice information in traditional methods, and providing theoretical and technical support for improving the accuracy of weather forecasts in the Arctic region.

[0063] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0064] The specific implementation of step S01 is to call the Global Weather Service (GFS) server via an API interface to obtain GFS background field data. This data covers the entire global geographic area and vertically includes 41 non-uniformly distributed pressure layers from the surface to approximately 0.01 hPa. The horizontal resolution is 0.25° × 0.25°, the temporal resolution is 3 hours, and the single forecast duration is 168 hours. After obtaining the data, the ice-sea heat exchange correction equation is applied to perform a preliminary correction of the boundary layer parameters in the sea ice area. The ice-sea heat exchange correction equation is specifically expressed as follows: ; ; Where, is the corrected boundary layer turbulent flux, in units of ; is the standard calculated boundary layer turbulent flux, in units of ; is the correction factor, dimensionless; is the sea ice impact factor, ranging from 0.05 to 0.15, dimensionless; is the sea ice density, dimensionless, ranging from 0 to 1; is the sea ice thickness in meters. The purpose of this step is to obtain high-quality background field data and preliminarily improve the description of the polar boundary layer parameters, laying the foundation for subsequent processing.

[0065] The specific implementation of step S02 is to extract sea ice density and thickness information from the GFS background field data, using a bilinear interpolation algorithm to ensure spatial continuity. The extracted sea ice density and thickness information is then physically verified for consistency based on the sea ice thermodynamic equilibrium equation. The sea ice thermodynamic equilibrium equation is specifically expressed as follows: ; ; ; ; Where, is the sea ice density, which is 917 ; is the specific heat capacity of sea ice, which is 2067 ; is the internal temperature of sea ice, in K; is the time, the unit is s; is the vertical coordinate, in m; is the thermal conductivity of sea ice, in units of ; is the sea ice thickness, in m; is the heat flux at the bottom of the sea ice, in units of ; is the heat flux on top of sea ice, in ; is the latent heat of seawater freezing, which is 3.35× J / kg; is the thermal conductivity of pure ice, which is 2.03 ; is the salinity influence coefficient, which is 0.13; is the sea ice salinity, in ‰; is the sea ice surface temperature, in K; is the freezing temperature of seawater, in K.

[0066] Based on the above equation, a physical consistency check is performed. When the relative deviation between the actual sea ice thickness and the theoretical value exceeds 20%, the sea ice thickness is adjusted through an iterative method to meet the thermodynamic constraints. The coordination between sea ice density and sea ice thickness is also tested. For the newly formed thin ice area, the relationship is defined as: ; Where, is the maximum reasonable value for sea ice density in thin ice areas. The verified sea ice density and thickness data are rewritten in binary format into an intermediate file recognizable by the model preprocessor. This step ensures the physical consistency of the sea ice parameters and avoids unphysical solutions during the model integration process.

[0067] The specific implementation method of step S03 is to set the model pre-processing parameter file, including key parameters such as model resolution, vertical layer number, projection mode and calculation domain range. For the Arctic region, polar stereographic projection is used to reduce grid deformation, and the horizontal resolution is set to 15km, 57 layers in the vertical direction, and the top layer height is 10hPa. When executing the terrain processing program, Gaussian filtering is used to smooth the terrain, and the filter radius is 4 grid point distances to eliminate numerical noise. The format rewriting program converts all types of input data into a model-specific binary format, and the interpolation program uses a conformal interpolation algorithm to ensure that the physical quantity is conserved during the spatial interpolation process. The Arctic ice sea assimilation model adopts four-dimensional variational assimilation technology, and the basic expression is as follows: ; ; ; Where, is the cost function; is the analysis field state vector; is the background field state vector; is the background error covariance matrix; is the observation operator; is the observation vector; is the observation error covariance matrix; is the flow-related length scale in km; is the benchmark length scale, which is 50 km; is the adjustment coefficient, which is set to 2.5; is the sea ice concentration gradient, is the maximum value of the sea ice concentration gradient. This step generates a model background field with higher physical consistency by fusing multiple factors, providing high-quality initial conditions for subsequent model integration.

[0068] The specific implementation of step S04 is to configure the sea ice thermodynamic condition parameters and the sea-air momentum exchange coefficient in the model parameters. The sea ice thermodynamic parameters include the sea ice thermal conductivity set to 2.0 , the sea ice albedo is dynamically adjusted according to the sea ice thickness, and the specific expression is: ; Where, is the sea ice albedo, dimensionless; is the sea ice thickness in meters.

[0069] The air-sea momentum exchange coefficient adopts the modified Charnock relationship, which is expressed as: ; Where, is the roughness length parameter, unit is m; is the sea ice density, dimensionless.

[0070] The polar sea ice heat flux equation is used to calculate the sea ice surface temperature. The equation is expressed as follows: ; ; ; ; Where, is the sea ice albedo, dimensionless; is the downward shortwave radiation flux, in units of ; is the sea ice emissivity, which is 0.97 and dimensionless; is the downward longwave radiation flux, in units of ; is the Stefan-Boltzmann constant, which is 5.67× ; is the sea ice surface temperature, in K; is the sensible heat flux, in units of ; is the latent heat flux, in units of ; is the conduction heat flux, in units of ; is the air density in units of ; is the specific heat capacity of air at constant pressure, which is 1004 ; is the sensible heat exchange coefficient, dimensionless; is the near-surface wind speed in m / s; is the near-surface air temperature in K; is the latent heat of water vaporization, which is 2.5× J / kg; is the latent heat exchange coefficient, dimensionless; is the near-surface specific humidity, dimensionless; is the saturation specific humidity, dimensionless; is the thermal conductivity of sea ice, in units of ; is the freezing temperature of seawater, in K; is the sea ice thickness, in m; is the heat permeability coefficient, which is 0.3 when the sea ice thickness is less than 0.1 m, otherwise it is 0; is the attenuation coefficient, which is dimensionless and takes a value of 7. The purpose of this step is to accurately describe the heat exchange process between sea ice thickness and the atmosphere, providing key input for boundary layer parameterization.

[0071] The specific implementation of step S05 is to establish a set of sea ice density and thickness boundary layer interaction equations. This set of equations is based on the modified boundary layer theory and takes into account the modulation effect of sea ice on the atmospheric boundary layer structure. First, the surface roughness of the sea ice is calculated. The specific expression is: ; ; ; Where, is the surface roughness of sea ice, in m; is the ice ridge coverage ratio, dimensionless; is the height of the ice ridge, in m; For reference ice ridge height, the value is taken as 1m; is the sea ice thickness in meters.

[0072] Then calculate the thermal stability parameter, the expression is: ; Where, is the Richardson number, dimensionless; is the acceleration due to gravity, which is 9.8 ; is the virtual potential temperature, unit is K; is the virtual temperature difference, unit is K; is the height difference, in m; is the wind speed difference, in m / s.

[0073] Based on the above stability, the boundary layer turbulent flux is calculated: ; ; ; Where, is the momentum flux, with units of ; is the air density in units of ; is the drag coefficient, dimensionless; is the near-surface wind speed in m / s; is the stability correction function, dimensionless; is the sensible heat flux, in units of ; is the specific heat capacity of air at constant pressure, in units of ; is the sensible heat exchange coefficient, dimensionless; is the near-surface air temperature in K; is the sea ice surface temperature in K.

[0074] The boundary layer height is determined by the critical Richardson number method: ; Where, is the boundary layer height, in m; is the height of the jth floor, in meters; For height Richardson number at ; is the critical Richardson number, which is 0.25.

[0075] The temperature gradient and humidity gradient are calculated using the discrete difference method: ; ; Where, is the temperature gradient, in K / m; is the temperature difference, in K; is the height difference, in m; is the humidity gradient, in kg / (kg·m); The sea ice modulated boundary layer structure generated in this step contains key parameters such as boundary layer height, temperature gradient, and humidity gradient. It is used as input for the subsequent feature fusion model to accurately describe the impact of sea ice on boundary layer processes.

[0076] The specific implementation of step S06 is to use the feature fusion model to deeply fuse the sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure. The feature fusion model first normalizes the input features using the Z-score normalization method: ; Where, is the normalized feature; is the original feature; is the characteristic mean; is the characteristic standard deviation.

[0077] The weight of each element in the feature fusion model output vector is adaptively adjusted based on the ice-air interface balance index. The ice-air interface balance index is calculated as follows: ; Where, is the ice-air interface equilibrium index, in units of ; is the sea ice density, dimensionless; is the sea ice thickness, in m; is the net surface radiation flux in units of .

[0078] The weight adjustment method is: ; Where, is the adjusted weight; is the original weight; is the feature index, where 1 represents sea ice density and 3 represents sea ice surface temperature. This step optimizes the multi-source feature fusion process through deep learning methods, generating model initial fields and boundary conditions with higher physical consistency, providing high-quality input for model integration.

[0079] The specific implementation method of step S07 is to set the physical parameterization scheme and the sea ice parameterization scheme. The physical parameterization scheme selection includes the use of the RRTMG long and short wave scheme for the radiation scheme, the modified YSU scheme for the boundary layer scheme, the Kain-Fritsch scheme for the cumulus convection scheme, and the WSM6 scheme for the microphysics scheme. The sea ice parameterization scheme selection includes the use of multi-layer nonlinear diffusion equations for heat conduction, the consideration of salinity effects in the phase change process, and the use of elastic-plastic rheology methods for the dynamic process. The integration program is executed using the model initial field and boundary conditions generated in step S06, with the time step set to 45 seconds, the integration time to 168 hours, and the output frequency to 1 hour. The ensemble forecasting technology is used to evaluate the uncertainty impact of sea ice density and sea ice thickness on meteorological forecast results, and the ensemble member perturbation method is as follows: ; ; ; Where, is the sea ice density of the jth ensemble member; To control the forecast sea ice concentration; is the sea ice concentration disturbance coefficient; is the sea ice thickness of the jth ensemble member; To control the predicted sea ice thickness; is the sea ice thickness disturbance coefficient; Represents normal distribution. The ensemble forecast results are described by probability distribution function to describe the forecast uncertainty. The criteria for determining high uncertainty areas are: ; Where, is the standard deviation of the ensemble forecast result; The purpose of this step is to obtain the weather forecast results for the next 7 days and quantify the forecast uncertainty.

[0080] The specific implementation of step S08 is to post-process the weather forecast results, including systematic error correction, downscaling and product generation. The systematic error correction adopts the quantile mapping method, and the specific expression is: ; Where, is the forecast result after correction; is the inverse function of the observed probability distribution; is the forecast probability distribution function; is the original forecast result. The forecast skill score is calculated by comparing the historical observation data. The scoring indicators include root mean square error, bias, correlation coefficient and forecast skill score. The calculation method is as follows: ; ; ; Where, is the root mean square error; is the forecast value; is the observed value; is the sample size; for deviation; is the correlation coefficient; is the forecast mean; is the observed mean; Score skills; This is a reference forecast. The skill score threshold for Arctic forecasts requires a 72-hour forecast skill of no less than 0.7 for the 500hPa geopotential height field and a 48-hour forecast skill of no less than 0.65 for surface temperature. Standardized forecast products are generated based on Arctic application needs. These include conventional meteorological element forecast maps, sea ice concentration change forecast maps, extreme weather event probability forecast maps, and navigation index forecast maps. This step aims to transform raw forecast results into user-friendly forecast products and assess forecast quality through skill scores.

[0081] The detailed structure of the feature fusion model utilizes a deep neural network framework and includes the following main components: The input layer receives normalized feature data such as sea ice concentration, sea ice thickness, sea ice surface temperature, and ice-modulated boundary layer structure. The detailed steps for establishing the training dataset for the feature fusion model include: First, Arctic exploration satellite data are collected from multiple sources, including the AMSR2 sea ice concentration product with a resolution of 6.25 km; the CryoSat-2 and SMOS joint inversion sea ice thickness product with a resolution of 25 km; and the MODIS sea ice surface temperature product with a resolution of 1 km. Field data from polar research stations, buoys, and ship observations are also collected, including boundary layer temperature profiles, humidity profiles, and wind speed profiles. The collected data are resampled to a 15 km spatial grid and a daily temporal resolution, using bilinear interpolation to ensure conservation of physical quantities.

[0082] Key features such as sea ice concentration, sea ice thickness, sea ice surface temperature, near-surface air temperature, and near-surface wind speed were extracted from the processed data. A sliding window sample pair with a 30-day time window was constructed, with features from the first 27 days as input and features from the last three days as output. Irrational data was screened and removed based on the principle of physical conservation. This included verifying that sea ice concentration was within the range of 0 to 1, that sea ice thickness was positive and that the concentration in thin ice areas did not exceed a threshold, and that surface temperature met thermodynamic constraints. Samples that did not meet these criteria were removed. The resulting training dataset contained multi-scale spatiotemporal features, totaling at least 10,000 samples, with the validation set and test set comprising 20% ​​and 10% of the total data, respectively.

[0083] Optionally, for the sea ice thermodynamic balance equation, consider sea ice salinity The value is 12‰ in new ice and 4‰ in multi-year ice. Affected by atmospheric forcing, the bottom temperature of sea ice Taking the freezing point of seawater as an example, for seawater with a salinity of 35‰, the freezing point is approximately -1.8°C. The physical consistency check process first solves the sea ice thermodynamic equilibrium equation to obtain the theoretical sea ice thickness. This value is then compared with the actual sea ice thickness. If the relative deviation exceeds a threshold of 20%, the sea ice thickness is adjusted through an iterative method. Each iteration is adjusted by 50% of the deviation, with a maximum of 10 iterations. The convergence criterion is a relative deviation of less than 5%.

[0084] Optionally, in four-dimensional variational assimilation, the length scale is smaller in regions with larger gradients, helping to preserve the frontal characteristics of the sea ice edge; in regions with smaller gradients, the length scale is larger, facilitating the wide dissemination of information. Four-dimensional variational assimilation is solved using the conjugate gradient method with a maximum of 100 iterations. The convergence criterion is that the gradient norm decreases to 0.01 of the initial value or the gradient change is less than 0.1% over three consecutive iterations.

[0085] Alternatively, the polar sea ice heat flux equation is based on the energy balance principle and considers multiple physical processes, including solar radiation, atmospheric longwave radiation, radiation emitted by the sea ice surface, and sensible, latent, and conductive heat exchange. The left side of the equation represents the energy balance on the sea ice surface, and the terms are: represents the solar shortwave radiation absorbed by sea ice, represents the atmospheric longwave radiation absorbed by sea ice, represents the longwave radiation emitted from the sea ice surface, is the sensible heat exchange coefficient, which is 0.0013 under neutral conditions. is the latent heat exchange coefficient, which is 0.0013 under neutral conditions. This equation is solved by Newton iteration method, and the initial guess value is , the iterative process is: ; ; ; Where, For the The surface temperature of the iteration, is the energy balance residual, is the derivative of the residual with respect to the surface temperature. The iterative termination condition is Or the maximum number of iterations, 20, is reached.

[0086] Optionally, in the boundary layer interaction equations for sea ice density and thickness: the boundary layer height is determined by the critical Richardson number method, that is, searching upward from the surface, when a certain layer of Richardson number For the first time, it is greater than or equal to the critical Richardson number (value is 0.25), the layer height is the boundary layer height . Temperature gradient and humidity gradient The discrete difference method is used for calculation. The vertical resolution is encrypted at the bottom of the boundary layer. The thickness of the first layer is 5m, which gradually increases with the height and does not exceed 50m.

[0087] Optional, ice-air interface balance index, when the index When it is less than 0.3, it indicates that the exchange process is weak, and the weight of the enhanced sea ice density characteristic increases by 30%; when the index When it is greater than 0.7, it indicates that the exchange process is strong, and the weight of the enhanced sea ice surface temperature characteristic is increased by 25%; when the index When the index is between 0.3 and 0.7, the weights of the various features remain balanced. The threshold setting of the ice-air interface balance index is determined based on historical data analysis. Statistics show that areas with an index less than 0.3 generally correspond to thin ice or ice edge areas, where changes in sea ice density are more sensitive to atmospheric influences; areas with an index greater than 0.7 generally correspond to thick ice areas, where changes in sea ice surface temperature have a more significant impact on the atmosphere.

[0088] Optionally, the terrain processing program can be expressed mathematically as follows: Gaussian filter smoothing: ; ; Where, is the height of the smoothed terrain; is the original terrain height; is the Gaussian filter kernel; is the filter radius, which is equal to the distance between 4 grid points; is the standard deviation of the Gaussian distribution, which is .

[0089] Polar Stereographic Coordinate Transformation: ; Where, is the rectangular coordinate of the projection plane; is the radius of the Earth; is latitude; is the longitude; is the central longitude.

[0090] Optionally, the mathematical expressions of the format rewriting procedure and the interpolation procedure are as follows: Shape-preserving interpolation algorithm: ; ; Where, The interpolation result of the target point; For the The value of a known point; is the weight coefficient; is the distance from the target point to the known point.

[0091] The four-dimensional variational assimilation technique has been described previously and will not be described in detail here.

[0092] Alternatively, the mathematical expression of the integration procedure is as follows: Momentum equation (horizontal direction): ; Continuity equation: ; Thermodynamic equation: ; Water vapor conservation equation: ; Where, is the horizontal wind speed vector; For time; is the vertical speed; is the vertical coordinate; is the air density; is the air pressure; is the Coriolis parameter; is the vertical unit vector; is the friction force; is temperature; is the specific volume; is the specific heat at constant pressure; is the heating rate; is specific humidity; is the water vapor source and sink term.

[0093] Optionally, the integration procedure also includes ensemble forecast perturbations and forecast uncertainty criteria: Optionally, the integration procedure also includes the multi-layer nonlinear diffusion equation for heat conduction in the sea ice parameterization: ; ; Optionally, the integration procedure also includes the dynamic equations of the elastic-plastic rheological method: ; ; Where, is the sea ice temperature; is the thermal conductivity of sea ice; is the sea ice salinity; is the thermal conductivity of pure ice; is the salinity influence coefficient; is the temperature dependence coefficient; is the freezing temperature; is the sea ice velocity vector; is the sea ice thickness; is the stress tensor; is the atmospheric stress; is ocean stress; is the Coriolis parameter; is the acceleration due to gravity; is the sea level height; is the shear viscosity; is the bulk viscosity; is the strain rate tensor; is the trace of the strain rate tensor; is the unit tensor; is the internal strength of sea ice.

[0094] It should be noted that the four solutions in step S07 are prior arts. To more comprehensively describe this embodiment, mathematical descriptions of the four solutions are provided below.

[0095] The radiation flux calculation is based on the fast radiative transfer model, whose basic equation is as follows: ; Where, The wavelength is The radiation intensity, in units of ; is the optical path length, in units of ; is the extinction coefficient, in units of ; is the emission coefficient, in units of .

[0096] The atmospheric shortwave radiation transfer equation can be expressed as: ; Where, is the optical thickness, dimensionless; is the cosine of the zenith angle, dimensionless; is the azimuth, in units of ; is the source function, and The units are the same.

[0097] The source function takes into account both scattering and thermal emission: ; Where, is the single scattering albedo, dimensionless, ranging from ; is the scattering phase function, dimensionless; is the Planck function, with the same unit as same; is the temperature in units of .

[0098] The Planck function is expressed as: ; Where, is Planck's constant, which is ; is the speed of light, and its value is ; is the Boltzmann constant, which is ; is the wavelength, in units of ; is the temperature in units of .

[0099] In the RRTMG scheme, the radiation calculation adopts the correlation coefficient method (Correlated-k Distribution), and the full spectrum radiation transmission is calculated by group summation: ; Where, is the total radiation flux, in units of ; For the Group weight coefficient, dimensionless; For the Group radiation flux, in units of ; is the number of groups. In the RRTMG scheme, there are 14 groups for short waves and 16 groups for long waves.

[0100] In the modified YSU scheme, the vertical distribution expression of turbulent flux is: ; Where, is a physical quantity The turbulent flux, whose units depend on the nature of; is the turbulent diffusion coefficient, in units of ; is a physical quantity vertical gradient; is a non-local term, and its unit is the same as same.

[0101] The turbulent diffusion coefficient is calculated as: ; Where, is the proportionality coefficient, dimensionless, with a value of 0.4 for momentum, and 0.4 for heat and water vapor; is the mixing length in units of ; is the turbulent kinetic energy, in units of ; is the stability correction function, dimensionless.

[0102] The nonlocal term expression is: ; Where, is the non-local term coefficient, dimensionless, and has a value range of ; is the surface turbulent flux, with the same unit as same; is the boundary layer height in units of ; is the height in units of ; is the power exponent, dimensionless, and its value is .

[0103] To adapt to the sea ice environment, the boundary layer height calculation formula of the modified YSU scheme in the sea ice area is: ; Where, is the friction speed, in units of ; is the Coriolis parameter, in units of ; is the Monin-Obukhov length, in ; is a dimensionless constant with a value of 0.4 on sea ice and 0.7 on the sea surface.

[0104] The friction velocity is calculated as: ; Where, is the drag coefficient, dimensionless; is the wind speed at the reference height, in units of .

[0105] The Monin-Obukhov length is calculated as: ; Where, is the virtual temperature, unit is ; is the Karman constant, dimensionless, with a value of 0.4; is the acceleration due to gravity, in units of , the value is ; is the virtual temperature flux, in units of .

[0106] The stability correction function is expressed as: ; Where, is the Richardson number, dimensionless, and is calculated as: ; Where, is the virtual temperature difference, in units of ; is the height difference in units of ; and are the east-west and north-south wind speed differences, in units of .

[0107] In the Kain-Fritsch scheme, the cloud base mass flux calculation expression is: ; Where, is the cloud base mass flux, in units of ; is the cloud base air density in units of ; is the vertical velocity of the cloud base, in units of ; is the cloud base area ratio, dimensionless, and has a value range of .

[0108] The parameterized expression of cloud base vertical velocity is: ; Where, is the effective potential energy of convection, in units of ; is the proportional coefficient, in units of , the value is .

[0109] The CAPE calculation formula is: ; Where, is the free convection height in units of ; is the equilibrium height, in units of ; is the acceleration due to gravity, in units of ; is the virtual temperature of the rising air mass, in units of ; is the virtual temperature of the environment, in units of ; is the vertical height element, the unit is .

[0110] The formula for calculating the convection adjustment time scale is: ; Where, To adjust the time scale, the unit is ; is the effective potential energy of convection, in units of ; is the CAPE consumption rate, in units of , the calculation formula is: ; The thermodynamic equation for the rising air in a cloud is: ; ; Where, is the cloud potential temperature, in units of ; is the height in units of ; is the latent heat of water vaporization, in ; is the specific heat at constant pressure, in units of ; is the Exenat function, dimensionless; is the liquid water ratio, dimensionless; is the involvement rate, in units of ; is the updraft mass flux, in units of ; is the ambient temperature, in units of ; is the total water ratio, dimensionless; is the ratio of ambient water to total water, dimensionless.

[0111] The parameterized expression of the involvement rate is: ; Where, is the proportionality coefficient, dimensionless, and its value is 0.03; is the cloud radius in units of .

[0112] In the WSM6 scheme, the basic kinetic equation for water phase change is: ; ; ; ; ; ; Where, is the water vapor ratio, dimensionless; is the cloud-water ratio, dimensionless; is the rainwater ratio, dimensionless; is the cloud ice ratio, dimensionless; is the snow ratio, dimensionless; is the graupel ratio, dimensionless; is the time, the unit is ; is the cloud water condensation rate, in units of ; is the water vapor deposition rate, in units of ; is the rainwater condensation rate, in units of ; is the evaporation rate, in units of ; is the melting rate, in units of ; is the automatic conversion rate, in units of ; is the collision rate, in units of ; is the cloud ice rate for rainwater collection, in units of ; is the rate of rain freezing into snow, in units of ; is the rate of rain freezing into graupel, in units of ; is the snow collection cloud ice rate, in units of ; is the advection term, with units of .

[0113] The calculation formula for cloud water condensation rate is: ; Where, is the saturated water vapor ratio, dimensionless; is the latent heat of water vaporization, in ; is the specific heat at constant pressure, in units of ; is the water vapor gas constant, in units of ; is the temperature in units of ; is the time step in units of .

[0114] The automatic conversion rate (cloud water to rain water) is calculated as follows: ; Where, is the conversion rate coefficient, in units of , the value is ; is the critical cloud water content, dimensionless, and is taken as ; is the Heaviside step function, dimensionless.

[0115] The collision rate (raindrops collecting cloud droplets) is calculated as: ; Where, is the collision coefficient, in units of , the value is .

[0116] The parameterized expression of falling speed is: ; ; ; Where, , , are the falling speeds of raindrops, snow and graupel, respectively, in units of ; is the air density in units of ; , , are the diameters of raindrops, snow and graupel respectively, in units of ; , , is the coefficient, and its unit is , respectively take values ​​of , , ; , , is a dimensionless coefficient, and its values ​​are , , ; , , is the power index, dimensionless, and takes the values , , .

[0117] To better understand and implement the present invention, Example 2 of a specific application scenario is provided below. This example details the physical consistency verification of sea ice density and thickness information, the multi-factor integration of the Arctic ice-sea assimilation model, the configuration of sea ice thermodynamic condition parameters, and the deep fusion processing of the feature fusion model, thereby significantly improving the accuracy of Arctic regional weather forecasts. The forecasting method includes a data download program, a sea ice thickness preprocessing program, a data preprocessing program, a model calculation program, a post-processing program, and an overall control program.

[0118] In this embodiment, researchers selected the route of China's 12th Arctic scientific expedition north of 60°N (12UTC on July 21, 2021 to 23UTC on September 15, 2021) for testing and verification. First, step S01 is implemented through the data download program. The wget download function of the Linux server is used to obtain the global GFS background field data from the server of the National Centers for Environmental Prediction of the United States. The data has 41 vertical layers, a horizontal resolution of 0.25°×0.25°, a time interval of 3 hours, and a single forecast duration of 168 hours. The program provides data download detection and restart functions. When a download interruption is detected, the download process is automatically restarted to ensure data integrity. The program operation information is saved in the download_err file. As shown in Table 1: Table 1 Main parameters of GFS background field data

[0119] After data download is complete, the sea ice thickness preprocessor implements step S02, extracting sea ice density and thickness information from the GFS background field and performing a physical consistency check based on the sea ice thermodynamic equilibrium equation. First, the sea ice thickness from the GFS background field is extracted and converted into NETCDF format using wgrib2 software. Then, an NCL script reads the processed sea ice thickness data and rewrites it into an intermediate file recognizable by WPS. During the physical consistency check, when the sea ice thickness is less than 0.3 m, the corresponding sea ice density is constrained to no more than 0.85, consistent with the physical characteristics of thin new ice areas. When the relative deviation between the actual sea ice thickness and the theoretical value of thermodynamic equilibrium exceeds 20%, the sea ice thickness is adjusted iteratively to meet the thermodynamic constraints. The verified sea ice density and thickness data are rewritten in binary format into an intermediate file recognizable by the model preprocessor. A lossless compression algorithm is used to reduce storage space. In order to ensure that the results of the sea ice thickness preprocessing program are consistent with the time names of other binary files, a time loop function is designed in the preprocessing program to ensure that the generated intermediate file information can be viewed using the rd_intermediate.exe program that comes with WPS.

[0120] The data preprocessing program implements step S03, integrating meteorological elements, sea ice concentration, and sea ice thickness information through the Arctic Ice and Sea Assimilation Model to generate a more physically consistent model background field. According to the Polar WRF model's operating requirements, the namelist.wps file is set. For the Arctic region, a polar stereographic projection is used to minimize grid distortion. The horizontal resolution is set to 15 km, the vertical stratification is set to 57 layers, and the top layer is set to 10 hPa. The terrain processing program geogrid.exe, the format rewriting program ungrib.exe, and the interpolation program metgrid.exe are executed sequentially. During the terrain processing program, a Gaussian filter is used to smooth the terrain with a filter radius of four grid points to eliminate numerical noise. The format rewriting program converts all input data into a model-specific binary format. The interpolation program uses a conformal interpolation algorithm to ensure that physical quantities are conserved during spatial interpolation. Before running metgrid.exe, sea ice thickness information must be specified in METGRID.TBL and the ICEDEPTH option must be added to the fg_name parameter in namelist.wps. The Arctic ice-sea assimilation model uses four-dimensional variational assimilation technology, introduces a background error covariance matrix to describe the correlation between different variables, and calculates the covariance structure using the NMC method. Considering the specificity of the sea ice region, a flow-related length scale adjustment algorithm is introduced to ensure that the assimilation process can better preserve the frontal characteristics of the sea ice edge. The files generated by the data preprocessing results are shown in Table 2: Table 2 Main parameters of data preprocessing and result files

[0121] The model calculation program implements steps S04 to S07, sequentially performing sea ice thermodynamic condition parameter configuration, sea ice modulated boundary layer structure calculation, feature fusion model depth fusion processing, and model integral calculation. First, real.exe is used to convert background field data into the initial fields wrfinput, side boundary wrfbdy, and bottom boundary wrflowinp required by the WRF model. Due to an interpolation bug in the Polar WRF version of real.exe, WRFV3.9.1 was used for processing. The model parameters configure sea ice thermodynamic condition parameters and the air-sea momentum exchange coefficient: the sea ice thermal conductivity is set to 2.0 W / (m·K), and the sea ice albedo is dynamically adjusted based on thickness, ranging from 0.5-0.7 for thicknesses less than 0.5 m and from 0.7-0.85 for thicknesses greater than 0.5 m. The air-sea momentum exchange coefficient uses a modified Charnock relation, and the roughness length parameter is set to 0.0012 for sea ice concentrations greater than 0.6 and 0.0018 for concentrations less than 0.6. The polar sea ice heat flux equation is used to calculate the sea ice surface temperature. This equation is based on the principle of energy balance and takes into account multiple physical processes such as solar radiation, atmospheric long-wave radiation, sea ice surface emitted radiation, and sensible and latent heat exchange. The nonlinear equation group is solved by the Newton iteration method.

[0122] A set of interaction equations for sea ice density and thickness boundary layer is established to calculate the sea ice modulated boundary layer structure. This set of equations is based on the modified boundary layer theory and takes into account the modulating effect of sea ice on the atmospheric boundary layer. First, the sea ice surface roughness is calculated, then the thermal stability parameters are calculated, the Richardson number is used to characterize the stability state, and the turbulent flux is calculated based on the stability-corrected K theory. A feature fusion model is applied to deeply fuse the sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure, and the weight of each element in the feature fusion model output vector is adaptively adjusted based on the ice-air interface balance index. When the index is less than 0.3, the weight of the enhanced sea ice density feature is increased by 30%; when the index is greater than 0.7, the weight of the enhanced sea ice surface temperature feature is increased by 25%. The specific configurations of the physical parameterization scheme and the sea ice parameterization scheme are shown in Table 3: Table 3 Key parameter settings of the model calculation program

[0123] The post-processing program implements step S08, post-processing the meteorological forecast results, calculating the forecast skill score by comparing them with historical observational data, and generating standardized forecast products based on the application requirements of the Arctic region. The wrfout files obtained by model integration are interpolated and reformatted using ARWpost software, converting them into a binary format readable by GrADs. These binary files are then converted into standard NetCDF format using a Fortran program. Systematic error correction uses a quantile mapping method, constructing an error correction model based on the statistical relationship between historical forecasts and observations, with a training sample length of at least 365 days. Forecast skill scores are calculated by comparing historical observational data. Scoring metrics include root mean square error, bias, correlation coefficient, and forecast skill score. The forecast skill score thresholds for the Arctic region require a 72-hour forecast skill of no less than 0.7 for the 500hPa geopotential height field and a 48-hour forecast skill of no less than 0.65 for the surface temperature. The resulting forecast products include maps for conventional meteorological elements such as 2-meter air temperature, 2-meter dew point, sea level pressure, and 10-meter wind speed, as well as maps for sea ice concentration changes, extreme weather event probability, and navigation index. The products have a resolution of 0.1° × 0.1° and support multiple formats, including NetCDF, GrADS, and GeoTIFF.

[0124] The entire forecast process is connected in series through the overall control program, and each program is executed in the order of steps S01 to S08. The operation status of each program can be viewed in real time through the msg.run file. In order to evaluate the forecast effect of this method, the route of China's 12th Arctic scientific expedition north of 60°N (12UTC on July 21, 2021 to 23UTC on September 15) was selected for testing and comparison. The model forecast variables include dozens of variables such as wind, temperature, relative humidity, air pressure, precipitation, etc., and the single forecast duration is 7 days. The initial field, side boundary and bottom boundary data are updated once every 24 hours, and finally 57 groups of forecast products are obtained. The first 24 hours of each group of products are extracted and spliced ​​to form a 24-hour forecast result, and so on to form a 48-hour and 72-hour forecast product. The researchers also drew a Taylor diagram of the 24-hour forecast elements, which can simultaneously display three statistical indicators: correlation coefficient, standard deviation ratio and root mean square error, so as to more comprehensively evaluate the model forecast performance. As Figure 2As shown, in the 24-hour forecast, the model performs best for sea level pressure, with similar performance for 2-meter air temperature and 2-meter dew point. Wind speed at 10 meters is the worst. This is directly related to the optimization of sea ice surface temperature and boundary layer structure in steps S04 to S06, demonstrating the superiority of the proposed method in simulating thermodynamic processes. The researchers designed two sets of comparative experiments: a baseline experiment (Exp-A) that did not incorporate the fusion of sea ice concentration and thickness, and an improved experiment (Exp-B) that incorporated the proposed method. After incorporating sea ice concentration and thickness distribution, the model significantly improves its forecast performance for 2-meter air temperature and 2-meter dew point temperature. The root mean square error (RMS) for 24- to 72-hour 2-meter air temperature is reduced by 5.2-10.3% compared to Exp-A, and the root mean square error for 24- to 72-hour 2-meter dew point is reduced by 8.6-14.7% compared to Exp-A. The forecast errors for sea level pressure and 10-meter wind speed remain relatively unchanged, with improvements of less than 1.2% for both. Furthermore, it can be seen that the error in sea level pressure accumulates rapidly with increasing forecast time, with the error in a 72-hour forecast being 3.4 times that of a 24-hour forecast. The errors in 2-meter air temperature and 2-meter dew point gradually increase, while the error in 10-meter wind speed remains relatively unchanged as the forecast time increases.

[0125] Finally, the results were plotted using GrADs software. Figure 3 The forecast results of various elements in the Arctic region are shown.

[0126] This result demonstrates that the present invention effectively improves the forecasting capability of the boundary layer temperature field by integrating sea ice density and thickness, particularly through deep fusion processing using the multi-factor fusion of the Arctic Ice Sea Assimilation Model and the feature fusion model. Traditional Arctic weather forecasting methods have three main problems: first, they ignore the influence of sea ice density and thickness, resulting in inaccurate boundary layer parameterization and description of the heat exchange process; second, they fail to consider the complex relationship between sea ice thickness and atmospheric heat exchange, ignoring the coupling effect between sea ice surface temperature and thickness; and third, they lack a physical consistency verification mechanism, which can lead to physical contradictions between sea ice parameters. The core technological innovations of this invention include: first, physical consistency verification of sea ice concentration and thickness information based on the sea ice thermodynamic equilibrium equation in step S02 to ensure the physical rationality of the data; second, multi-factor fusion of meteorological elements, sea ice concentration, and sea ice thickness information through the Arctic Ice-Sea Assimilation Model in step S03; third, accurate calculation of sea ice surface temperature using the polar sea ice heat flux equation in step S04, effectively describing the heat exchange process between sea ice thickness and the atmosphere; fourth, establishment of a set of boundary layer interaction equations for sea ice concentration and thickness in step S05 to calculate the sea ice modulated boundary layer structure; and fifth, deep fusion of multi-source data using a feature fusion model in step S06, adaptively adjusting the feature fusion model output weights based on the ice-air interface balance index. These innovative technologies have been validated in the practical application of the 12th Arctic Scientific Expedition, significantly improving the forecast accuracy of near-surface air temperature and dew point temperature. The root mean square error of the 72-hour forecast of 2-meter air temperature and 2-meter dew point temperature has been reduced by 10.3% and 14.7%, respectively.

[0127] This invention successfully solves technical problems in traditional methods such as insufficient fusion of multi-source data and inaccurate description of sea ice-atmosphere interaction processes, providing a more reliable weather forecast basis for Arctic scientific expeditions and polar navigation.

[0128] It should be noted that the variables involved in the present invention are explained in detail as shown in Tables 4, 5 and 6 below.

[0129] Table 4 Variable Explanation Table (Part 1)

[0130] Table 5 Variable Explanation Table (Part II)

[0131] Table 6 Variable Explanation Table (Part 3)

[0132] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. An Arctic atmospheric coupled forecasting method that automatically integrates sea ice density and thickness, characterized by: It includes obtaining and correcting global GFS background field data, extracting and verifying sea ice density and sea ice thickness information, performing multi-factor fusion to generate background field, applying polar sea ice heat flux equation to calculate sea ice surface temperature, establishing sea ice density thickness boundary layer interaction equation group to calculate boundary layer structure, applying feature fusion model to deeply fuse sea ice density, sea ice thickness, sea ice surface temperature and sea ice modulated boundary layer structure, adaptively adjusting the weight of each element in the feature fusion model output vector based on the ice-air interface balance index, generating optimized model initial field and boundary conditions, and executing integration procedure to obtain forecast results.

2. The method according to claim 1, characterized in that The GFS background field data contains 41 layers in the vertical direction, with a time interval of 3 hours and a single forecast duration of 168 hours. The ice-sea heat exchange correction equation is used to perform a preliminary correction on the boundary layer parameters of the sea ice area in the GFS background field data.

3. The method according to claim 2, characterized in that The step of extracting and verifying the sea ice density and sea ice thickness information specifically involves extracting the sea ice density and sea ice thickness information from the GFS background field data, performing a physical consistency check on the sea ice density and sea ice thickness information based on the sea ice thermodynamic equilibrium equation, and rewriting the verified sea ice density and sea ice thickness information into an intermediate file that can be recognized by a model pre-processing program.

4. The method according to claim 3, characterized in that The step of executing multi-factor fusion to generate a background field specifically includes setting a model pre-processing parameter file, executing a terrain processing program, a format rewriting program, and an interpolation program, and fusing meteorological elements, the sea ice density, and the sea ice thickness information through an Arctic ice-sea assimilation model to generate a model background field with higher physical consistency.

5. The method according to claim 4, characterized in that The deep fusion specifically applies a feature fusion model to deeply fuse the sea ice density, the sea ice thickness, the sea ice surface temperature and the sea ice modulated boundary layer structure, and adaptively adjusts the weight of each element in the output vector of the feature fusion model based on the ice-air interface balance index.

6. The method according to claim 5, characterized in that The forecast results are weather forecast results for the next 7 days.

7. The method according to claim 6, characterized in that It also includes post-processing the weather forecast results, calculating the forecast skill score by comparing with historical observation data, and generating standardized forecast products based on the application needs of the Arctic region.

8. The method according to claim 5, characterized in that The GFS background field data refers to the gridded forecast data generated by the global forecast system that includes multiple layer elements of the atmosphere, ocean, land surface and sea ice.

9. The method according to claim 8, characterized in that The ice-air interface balance index is used to assess the balance state of heat, momentum and water vapor exchange during the interaction between sea ice and the atmosphere.

10. The method according to claim 9, characterized in that The ice-air interface balance index is obtained by calculating the product of the sea ice density and the sea ice thickness and dividing it by the surface net radiation flux. When the ice-air interface balance index is less than 0.3, it indicates a weak exchange process; when it is greater than 0.7, it indicates a strong exchange process; and between the two, it indicates a moderate exchange process.

Images