Global climate information description method, computer equipment and storage medium
The multi-layer interconnected multivariate topology map is generated through recursive subdivision method and mask processing, which solves the problem of describing the earth's climate system in the prior art, and realizes an effective description of curvature representation and scale connection, which is suitable for graph neural network training.
Patent Information
- Application Number
- CN202510378325.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-28
- Publication Date
- 2025-07-11
AI Technical Summary
When describing the earth's climate system, the existing technology cannot effectively represent the curvature of the earth, the non-numerical data of the ocean area, the mixed distribution of different environmental indicators, and the amount of data increases quadratically with the increase of resolution, resulting in difficulty in processing deep learning models.
Recursive segmentation method is used to generate a segmented mesh, connect vertices through directed edges, process non-numerical areas with masks, and generate multi-layer interconnected multivariable topology diagrams, which are suitable for graph neural network training.
Effectively describe the curvature of the earth's surface, remove non-numerical areas of the ocean, represent near, medium and far scale connections, and the sample data volume does not increase with resolution, and is suitable for global climate prediction.
Smart Images
Figure CN120297049A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of the combination of computer data structures and atmospheric and oceanic sciences. Specifically, it relates to a method for describing global climate information, a computer device, and a storage medium. Background Art
[0002] In recent years, many weather large models and ocean large models have emerged for predicting future spatial atmospheric and oceanic physical indicators and derived indicators. For example, the Pangu weather large model, the Fuxi weather large model, the AI-GMOS ocean model, and the GraphCast graph prediction large model. These models mostly use tensor format reanalysis data to describe the physical indicators of the Earth's spatial atmospheric and oceanic environment. Using tensors to describe spatial atmospheric and oceanic environmental variables has the advantages of simple description method, easy processing, and adaptation to various deep learning models. However, this description method also has the following disadvantages:
[0003] 1) The shape of a tensor is planar, while the structure of the Earth is ellipsoidal. Mapping points on the Earth's surface to a tensor usually requires a projection algorithm. For example, the Mercator projection or the equiangular projection. Such projections have serious image distortions, making the sampling points denser in regions with higher latitudes, and the area represented by each point shrinking with the increase in latitude. Although the Mollweide projection represents equal areas, such projections cannot be represented in tensor form.
[0004] 2) Different from the globally covered atmosphere, the ocean accounts for about 70% of the Earth's surface area. If tensors are used to describe the environmental indicators on the ocean surface, about 30% of the area will be non-numerical data, and the distribution is uneven. In addition, the depths of different ocean regions are not the same, ranging from 0m to 11034m. In many ocean regions, the shallow layer is the ocean, and the deep layer is the continental shelf or the oceanic crust. If tensors are used to describe the three-dimensional environmental indicators of the ocean, the proportion of non-numerical data regions will exceed 30%, and the distribution is extremely uneven. Currently, most deep learning models cannot process samples mixed with numerical and non-numerical data.
[0005] 3) The Earth's climate system is a chaotic system. The climate system has mesoscale climate phenomena and global spatial atmospheric and oceanic exchange phenomena, which are affected by short, medium, and long-distance environmental indicators. Tensors can only show local connections and cannot show mesoscale and global connections. If a tensor is regarded as a topological structure, with the position of each sampling point as a node, then the internal nodes are only connected to the adjacent six nodes, and the edge nodes are only connected to the adjacent five, four, or three nodes. The connections and correlations with more distant nodes need to rely on deep learning models for processing and cannot be directly represented by the samples.
[0006] 4) Spaces, the atmosphere, and the ocean usually contain different environmental indicators and use different classification methods. The space has single-layer indicators, including variables such as solar radiation, geomagnetic activity, and ozone content; the atmosphere is classified layer by layer in hectopascals and contains variables such as geopotential, air pressure, and humidity; the ocean is classified layer by layer using sea layer depth and contains variables such as sea temperature, salinity, thermocline depth, and sea level height. If tensors are used to describe all three simultaneously, they are usually scaled to the same resolution and then directly stacked along the channel dimension. This will cause the data volume to increase quadratically with the increase in resolution and cannot be processed by existing large models and computing devices, greatly hindering the development process of using deep learning models to fully fit the global climate change process and create an earth twin model integrating space, sky, and ocean. Summary of the Invention
[0007] Aiming at the defects in the prior art, the purpose of the present invention is to provide a method for describing global climate information, a computer device, and a storage medium. The method of the present invention can effectively describe the earth surface model with positive curvature, remove the non-ocean areas of the ocean-related indicators, represent the near, medium, and far-scale connections of the earth at the same time, and the sample data volume does not increase with the improvement of the sampling resolution.
[0008] To solve the above problems, the technical solution of the present invention is as follows:
[0009] A method for describing global climate information, comprising the following steps:
[0010] Using the recursive subdivision method, each face of the regular polyhedron inscribed in the unit sphere is subdivided to generate more vertices, and each vertex is made to fall on the unit sphere to obtain a subdivided grid;
[0011] For each vertex in the subdivided grid, a finite number of directed incoming edges are created to connect to other vertices according to the distance level and connection density parameters;
[0012] The generated subdivided grid is copied into multiple grid layers to represent different climate characteristics respectively;
[0013] Using a preset mask to process different grid levels, removing the grid vertices in the areas that cannot be numerically represented and removing the edges without vertices;
[0014] For each vertex in each layer of the subdivided grid, it is connected to the vertices in other layers according to the preset connection method;
[0015] The original environmental indicators and position information are used to generate a point feature tensor and an edge feature tensor in a preset manner to obtain a multi-layer interconnected multi-variable topological graph describing global climate information.
[0016] Preferably, in the step of using the recursive subdivision method to subdivide each face of the regular polyhedron inscribed in the unit sphere to generate more vertices and making each vertex fall on the unit sphere to obtain a subdivided grid, the vertices include position information and / or environmental feature information; the edges of the subdivided grid are directed edges, and the directed edges include position information, momentum information and / or environmental feature information; the regular polyhedron inscribed in the unit sphere includes a regular tetrahedron, a regular hexahedron, a regular octahedron, a regular dodecahedron and an icosahedron.
[0017] Preferably, in the step of creating a finite number of directed in-edges to connect with other vertices for each vertex in the subdivided grid according to the distance level and the connection density parameter, the distance level is a sequence composed of multiple real numbers, representing the average straight-line distance of the edges to be connected by the subdivided grid points; the connection density parameter is a sequence composed of multiple real numbers, representing the number of connection edges in each corresponding distance level.
[0018] Preferably, in the step of using a preset mask to process different grid levels, removing the grid vertices in the areas that cannot be numerically represented, and removing the edges without vertices, the preset mask includes, but is not limited to, data representations that exist in a single or combined manner in data structures such as subdivided grids, graphs, matrices, and tensors; the methods of using the preset mask to process the grid layer include, but are not limited to, removing grid points, marking grid points as invalid points, and setting the feature values of grid points to zero, depending on the specific needs of the data in the usage scenario.
[0019] Preferably, in the step of connecting each vertex in each layer of the subdivided grid with vertices in other layers according to a preset connection method, the preset connection method includes directed edge connections between one vertex or multiple vertices in other layers under certain rules; each layer in the subdivided grid can be connected with one or more other layers according to preset rules, and the preset rules for connection between different layers can be the same or different.
[0020] Preferably, in the step of generating a point feature tensor and an edge feature tensor from the original environmental indicators and position information in a preset manner to obtain a multi-layer interconnected multi-variable topological graph describing global climate information, the original environmental indicators include, but are not limited to, spatial, atmospheric, oceanic, and geotopological indicators, and the sources include, but are not limited to, radiosonde satellites, Earth-orbiting satellites, weather balloons, weather aircraft, weather stations, tide stations, Argo floats, survey ships, and deep submergence vehicles. The provided methods include, but are not limited to, raw data, reanalysis data, prediction data, and data obtained by normalizing or otherwise processing the above data to make the model easier to learn; the position information includes, but is not limited to, spherical coordinate position information represented by longitude and latitude, and using the θ angle and Polar coordinate position information represented by angles, Cartesian coordinate position information represented by x, y, and z, and information obtained by normalizing or otherwise processing the above information to make the model easier to learn; the preset method of the generation point and the feature tensor of the edge is an algorithm that maps existing original environmental indicators to each vertex and / or each directed edge in the subdivision grid.
[0021] Further, the present invention also provides a computer device, including a processor and a memory for storing executable instructions of the processor, and the processor is configured to execute the global climate information description method as described above by executing the executable instructions.
[0022] Further, the present invention also provides a computer-readable storage medium, and the computer-readable storage medium is used to store program codes, and the program codes are used to execute the global climate information description method as described above.
[0023] Compared with the prior art, the global climate information description method of the present invention can effectively construct a multi-layer interconnected graph structure to describe spatial, atmospheric, and ocean information, describe the Earth's surface model with positive curvature, remove non-ocean areas of ocean-related indicators, and represent the near, medium, and far-scale correlation relationships of the Earth at the same time. Moreover, the sample data volume does not increase with the improvement of the resolution, and the described graph structure is applicable to the training of graph neural networks and the spatio-temporal prediction of global atmospheric and ocean environmental factors. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] By reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings, other features, objectives, and advantages of the present invention will become more apparent:
[0025] Figure 1 It is a flowchart of the global climate information description method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0026] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those of ordinary skill in the art can make several changes and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.
[0027] Specifically, the present invention provides a global climate information description method, as Figure 1 shown, the method includes the following steps:
[0028] S1: Using the recursive subdivision method, each face of the regular polyhedron inscribed in the unit sphere is subdivided to generate more vertices, and each vertex is made to fall on the unit sphere to obtain a subdivision grid;
[0029] Specifically, the so-called recursive subdivision method divides a closed and bounded plane into multiple planes using certain rules. The multiple divided planes can be further divided using the same method. Each division will retain the original vertices and generate new vertices.
[0030] Furthermore, the vertices contain position information and / or environmental characteristic information.
[0031] In one embodiment, the vertices are represented by the three-dimensional coordinate information of the Cartesian coordinate system where they are located.
[0032] In another embodiment, the information of the vertices contains two parts. One part is the θ angle and angle representation in the polar coordinate system where they are located. The other part is the vector of the mapped earth ocean environmental information, which includes the average sea temperature and average salinity of the region where they are located.
[0033] Furthermore, the regular polyhedra inscribed in the unit sphere include regular tetrahedrons, regular hexahedrons, regular octahedrons, regular dodecahedrons, and regular icosahedrons.
[0034] Furthermore, the edges of the subdivision grid are directed edges. Among them, a directed edge refers to an edge with a clear starting point and ending point, which contains position information and characteristic information. For vertices A and B, directed edge AB and directed edge BA are two different edges. For any two vertices, there can be multiple directed edges in the same direction. A directed edge can be a self-loop, that is, an edge that points to itself. For any vertex, there can be multiple self-loops.
[0035] Furthermore, the directed edges contain position information, momentum information, and / or environmental characteristic information.
[0036] In one embodiment, the directed edge is described by position information, and the position information is the Cartesian coordinate intercept from the starting point to the ending point.
[0037] In another embodiment, the directed edge is described by position information and environmental characteristic information. The position information is the length from the starting point to the ending point, and the environmental characteristic information is the vector of the mapped earth atmospheric environmental information, which includes the average westerly wind speed and average northerly wind speed of the region where it is located.
[0038] In one embodiment of step S1, a regular hexahedron inscribed in the unit sphere is used. Each time of subdivision, the midpoint connection lines of the upper and lower sides and the midpoint connection lines of the left and right sides of the square face are selected to divide the square face into four small squares, and a central focus is obtained. Then, through unit sphere normalization, the midpoints of the four edges and the central focus are scaled to the unit sphere. After 6 times of recursive subdivision, a grid containing 24,576 quadrilateral faces, 24,578 vertices, and 98,304 directed edges is obtained.
[0039] In another embodiment of step S1, a regular icosahedron inscribed in a unit sphere is used. Each time of subdivision, the midpoints of the three sides of an equilateral triangle face are connected to each other, dividing the triangle face into four small triangles, and then the midpoints of the three sides are scaled to the unit sphere through unit sphere normalization. After 6 times of recursive subdivision, a mesh containing 81,920 triangle faces, 40,962 vertices, and 245,760 directed edges is obtained.
[0040] S2: For each vertex in the subdivided mesh, create a finite number of directed incoming edges to connect to other vertices according to the distance level and connection density parameter;
[0041] Among them, the directed incoming edge refers to, for any vertex, a directed edge from other vertices to the current vertex.
[0042] Furthermore, the distance level is a sequence composed of multiple real numbers, representing the average straight-line distance of the edges to be connected by the subdivided grid points. In one embodiment, the distance level is composed of 6 real numbers. The first real number is the diameter of the subdivided grid where it is located, the second real number is twice the first real number, the third real number is twice the second real number, and so on. The sixth real number is 32 times the first real number. Among them, the diameter of the subdivided grid refers to the maximum length of all the edges in the grid.
[0043] Furthermore, the connection density parameter is a sequence composed of multiple real numbers, representing the number of connected edges corresponding to each distance level.
[0044] In one embodiment of step S2, the steps of creating directed incoming edges using a preset distance level and connection density parameter are as follows:
[0045] There is a subdivided grid G, composed of a vertex set V and an edge set E, with a distance level of [1, 3, 5] and a connection density parameter of [6, 12, 24];
[0046] For each vertex v in G, generate a sphere with v as the center and a radius of 1; calculate the vertex set Vs in the subdivided vertex set V that is closest to the sphere and whose distance does not exceed the diameter of G;
[0047] Find 6 vertices with the largest distance from each other in the vertex set Vs, create an edge set of these 6 vertices and vertex v, and add the edge set to the subdivided grid G. If the number of points in Vs is less than 6, then use all the points in Vs to create the edge set;
[0048] Repeat the operations of distance level 3, connection density 12 and distance level 5, connection density 24 for the above two steps to obtain a new subdivided grid G.
[0049] S3: Copy the generated subdivided grid into multiple grid layers, respectively representing different climate characteristics;
[0050] In one embodiment of step S3, the subdivided grid is replicated into three grid layers. One is used to represent geospatial features, where each vertex contains solar radiation, geomagnetic activity, and ozone content, and is represented by a feature vector with a shape of (1,3). One is used to represent atmospheric features, where each vertex contains atmospheric temperature, humidity, northerly wind speed, easterly wind speed, and geopotential at 1, 5, 10, 50, 100, 200, 300, 500, 800, 1000 hPa, and is represented by a feature matrix with a shape of (10,5). One is used to represent ocean features, where each vertex contains ocean temperature, salinity, meridional velocity, and zonal velocity at depths of 1, 5, 10, 50, 100, 150, 200, 300, 400, 500 m, and is represented by a feature matrix with a shape of (10,4).
[0051] In another embodiment of step S3, the subdivided grid is replicated into six grid layers, which are respectively used to represent the physical indicators of the atmospheric diffusion layer, the atmospheric stratosphere, the atmospheric troposphere, the ocean mixed layer (isothermal layer), the ocean thermocline, and the deep ocean. The quantity and level of the physical indicators of each type are related to the climate characteristics of each layer.
[0052] S4: Process different grid levels using a preset mask, remove the grid vertices in areas that cannot be numerically represented, and remove the edges without vertices;
[0053] Wherein, the preset mask is a data set used to block and / or remove data - less areas and / or illegal areas.
[0054] Furthermore, the preset mask includes, but is not limited to, data representations in the form of data structures such as subdivided grids, graphs, matrices, tensors, etc., existing individually or in combination.
[0055] Furthermore, the method of processing the grid layer using the preset mask includes, but is not limited to, methods such as removing grid points, marking grid points as invalid points, setting the feature values of grid points to zero, etc., depending on the specific needs of the data in the usage scenario.
[0056] In one embodiment of step S4, use a land - ocean mask to process the grid representing the sea - surface environment. The land and island areas are set to 0, and the ocean areas are set to 1. For any vertex of a grid layer, draw a circle with this point as the center and a radius of half of the grid diameter. If all the land - ocean masks contained in this circle are 0, then remove this vertex from the grid layer and delete all the directed edges associated with this vertex.
[0057] In another embodiment of step S4, a sea layer mask is used to process the grid representing the three-dimensional marine environment. The continental shelf, seabed continent, and island areas are set to 0, and the ocean area is set to 1. For any vertex of a grid layer, a cube is made with this point as the center and the grid diameter as the side length. If there is more than one 0 in all the sea layer masks contained in this cube, then this vertex is removed from the grid layer, and all the directed edges associated with this vertex are deleted.
[0058] S5: For each vertex in each layer of the subdivided grid, connect it to the vertices in other layers according to a preset connection method;
[0059] Specifically, the preset connection method includes directed edge connections between one vertex or multiple vertices in other layers under certain rules.
[0060] Furthermore, each layer in the subdivided grid can be connected to one or more other layers according to preset rules, and the preset rules for different layer connections can be the same or different.
[0061] It should be noted that in step S5, no directed edges are created to connect the vertices within the same layer itself, and this operation is already included in step S2.
[0062] In one embodiment of step S5, the main purpose is to predict the marine environment, with the spatial environment and atmospheric environment as parameters. The subdivided grid is divided into three layers: space, atmosphere, and ocean. For each vertex in the space grid, create directed edges between this vertex and the corresponding position vertex and its adjacent vertices in the ocean grid; for each vertex in the atmosphere grid, create directed edges between this vertex and the corresponding position vertex and its adjacent vertices in the ocean grid.
[0063] In another embodiment of step S5, the main purpose is to predict the global climate, with the spatial environment as a parameter. The subdivided grid is divided into three layers: space, atmosphere, and ocean. For each vertex in the space grid, create directed edges between this vertex and the corresponding position vertices and their adjacent vertices in the atmosphere grid and the ocean grid; for each vertex in the atmosphere grid, create directed edges between this vertex and the corresponding position vertices and their adjacent vertices in the ocean network; for each vertex in the ocean grid, create directed edges between this vertex and the corresponding position vertices and their adjacent vertices in the atmosphere grid.
[0064] S6: Generate a point feature tensor and an edge feature tensor from the original environmental indicators and location information in a preset manner to obtain a multi-layer interconnected multi-variable topological graph describing global climate information.
[0065] Among them, the topological graph is a mathematical structure that reflects the existence and connectivity of points and is one of the data structures used in graph neural networks. Different from the subdivision grid, the topological graph does not have position information and does not describe the specific positions of points in physical space or coordinate systems. After converting the subdivision grid into a topological graph, the position information of the vertices of the subdivision grid is no more special than other climate environment indicators.
[0066] It should be noted that the subdivision grid is a data structure composed of vertices and directed edges. The topological graph is a data structure composed of points and edges, and is usually described by a feature matrix (feature tensor), an adjacency matrix, an incidence matrix, and / or a weighted matrix.
[0067] Among them, the feature tensor is a tensor representing the features of all points or all edges of the topological graph. In the feature tensor, the feature dimensions of each point or edge are the same, and the tensor obtained by aggregating the features of all points or edges along a new dimension is the feature tensor.
[0068] In one embodiment, the topological graph is a data structure that includes a feature tensor of points and a feature tensor of edges.
[0069] Furthermore, the original environmental indicators include but are not limited to spatial, atmospheric, oceanic, and geotopological indicators, and the sources include but are not limited to radiosonde satellites, Earth-orbiting satellites, weather balloons, weather aircraft, weather stations, tide stations, Argo floats, survey ships, and deep-diving ships. The providing methods include but are not limited to raw data, reanalysis data, prediction data, and data obtained by normalizing or otherwise processing the above data to make the model easier to learn.
[0070] Furthermore, the position information includes but is not limited to spherical coordinate position information represented by longitude and latitude, polar coordinate position information represented by θ angle and angle, Cartesian coordinate position information represented by x, y, and z, and information obtained by normalizing or otherwise processing the above information to make the model easier to learn.
[0071] Furthermore, the preset method for generating the feature tensors of points and edges is an algorithm that maps existing original environmental indicators to each vertex and / or each directed edge in the subdivision grid.
[0072] In an embodiment of step S6, the original environmental indicators are provided by 4,000 ocean Argo floats. The provided environmental variables include sea surface temperature, salinity, meridional velocity, and zonal velocity. Through the aforementioned step S5, a subdivision grid including 40,962 vertices and 245,760 directed edges is obtained. For each vertex, the position information and environmental variables are used as features. The position information is the three-dimensional components of the Cartesian coordinate system, and the environmental variables are obtained by obtaining the environmental variables of the 3 Argo floats closest to the position of this vertex and using bilinear interpolation. For each directed edge, the intercepts of the three-dimensional components of the Cartesian coordinate system of the starting point and the ending point and the length of the directed edge are used as features. The relationship between points and edges is described using an adjacency matrix. Finally, a marine climate environment topology map composed of a (40962,7)-dimensional point feature matrix, a (245760,4)-dimensional edge feature matrix, and a (40962,40962)-dimensional adjacency matrix is obtained.
[0073] According to another aspect of the embodiments of the present application, a computer device is provided, which includes a processor and a memory for storing executable instructions of the processor. The processor is configured to execute the global climate information description method described in the embodiments of the present application by executing the executable instructions. The computer device of the present invention may specifically be a terminal or a server. The terminal may specifically be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The server may be an independent server or a server cluster composed of multiple servers.
[0074] According to still another aspect of the embodiments of the present application, a computer-readable storage medium is provided. The computer-readable storage medium is used to store program codes, and the program codes are used to execute the global climate information description method described in the embodiments of the present application.
[0075] The storage medium may be a ROM (Read Only Memory), or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory), or other types of dynamic storage devices that can store information and instructions. It may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, which is not limited herein.
[0076] The execution device can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in connection with the disclosure of the present application. The execution device can also be a combination that implements computing functions, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.
[0077] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.
Claims
1. A method for describing global climate information, characterized in that, The method includes the following steps: Using the recursive subdivision method, each face of the regular polyhedron inscribed in the unit sphere is subdivided to generate more vertices, and each vertex is made to fall on the unit sphere to obtain a subdivided grid; For each vertex in the subdivided grid, a finite number of directed incoming edges are created to connect to other vertices according to the distance level and connection density parameters; The generated subdivided grid is copied into multiple grid layers, respectively representing different climate characteristics; Using a preset mask to process different grid levels, removing the grid vertices in the areas that cannot be numerically represented, and removing the edges without vertices; For each vertex in each layer of the subdivided grid, it is connected to the vertices in other layers according to a preset connection method; The original environmental indicators and location information are used to generate a point feature tensor and an edge feature tensor in a preset manner to obtain a multi-layer interconnected multi-variable topological graph describing global climate information.
2. The method for describing global climate information according to claim 1, wherein In the step of using the recursive subdivision method, each face of the regular polyhedron inscribed in the unit sphere is subdivided to generate more vertices, and each vertex is made to fall on the unit sphere to obtain a subdivided grid, the vertex includes location information and / or environmental feature information; the edges of the subdivided grid are directed edges, and the directed edges include location information, momentum information and / or environmental feature information; the regular polyhedron inscribed in the unit sphere includes a regular tetrahedron, a regular hexahedron, a regular octahedron, a regular dodecahedron and an icosahedron.
3. The method for describing global climate information according to claim 1, wherein In the step of creating a finite number of directed incoming edges to connect to other vertices for each vertex in the subdivided grid according to the distance level and connection density parameters, the distance level is a sequence composed of multiple real numbers, representing the average straight-line distance of the edges to be connected by the subdivided grid points; the connection density parameter is a sequence composed of multiple real numbers, representing the number of connection edges in each corresponding distance level.
4. The method for describing global climate information according to claim 1, characterized in that, In the step of using a preset mask to process different grid levels, removing the grid vertices in the areas that cannot be numerically represented, and removing the edges without vertices, the preset mask includes but is not limited to data representations existing in a single or combined manner in data structures such as subdivided grids, graphs, matrices, and tensors; the ways of using the preset mask to process the grid layer include but are not limited to removing grid points, marking grid points as invalid points, and setting the grid point feature values to zero, depending on the specific needs of the data in the usage scenario.
5. The method for describing global climate information according to claim 1, wherein In the step of connecting each vertex in each layer of the subdivided grid to the vertices in other layers according to a preset connection method, the preset connection method includes directed edge connections between one vertex or multiple vertices in other layers under certain rules; each layer in the subdivided grid can be connected to one or more other layers according to preset rules, and the preset rules for different layer connections can be the same or different.
6. The method for describing global climate information according to claim 1, wherein In the step of generating the point feature tensor and the edge feature tensor from the original environmental indicators and the location information in a preset manner to obtain the multi-layer interconnected multi-variable topological graph describing the global climate information, the original environmental indicators include, but are not limited to, spatial, atmospheric, oceanic, and geotopological indicators, and the sources include, but are not limited to, radiosondes, Earth-orbiting satellites, weather balloons, weather aircraft, weather stations, tide gauges, Argo floats, survey ships, and deep-diving ships. The provided methods include, but are not limited to, raw data, reanalysis data, prediction data, and data obtained by normalizing or otherwise processing the above data to make the model easier to learn; the location information includes, but is not limited to, spherical coordinate location information represented by latitude and longitude, polar coordinate location information represented by θ and φ angles, Cartesian coordinate location information represented by x, y, and z, and information obtained by normalizing or otherwise processing the above information to make the model easier to learn; the preset method for generating the feature tensors of the points and edges is an algorithm that maps the existing original environmental indicators to each vertex and / or each directed edge in the subdivision grid.
7. A computer device, characterized in that, The computer device includes a processor and a memory for storing executable instructions of the processor, and the processor is configured to execute the global climate information description method according to any one of claims 1-6 by executing the executable instructions.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code, and the program code is used to execute the global climate information description method according to any one of claims 1-6.