A multi-source carbon observation joint assimilation method and system based on priority screening and data fusion

Abstract

The application discloses a multi-source carbon observation joint assimilation method and system based on priority screening and data fusion, relates to the technical field of atmospheric data assimilation and carbon cycle monitoring, and comprises the following steps: obtaining multi-source observation data and performing preprocessing, selecting a joint assimilation scheme based on the preprocessed data, establishing an observation data type priority rule library, performing observation data detection and classification within a grid, selecting observation data based on priority, and performing data assimilation analysis to obtain a carbon flux scale factor when a priority assimilation scheme is selected, performing multi-source observation data classification fusion and error re-calibration, constructing an extended observation operator and a cost function, performing adaptive covariance localization based on distance, and performing collaborative assimilation analysis to obtain an optimal state analysis value and update a posterior CO2 concentration field when a mixed assimilation scheme is selected. The application can realize efficient and high-precision surface carbon flux inversion.

Description

Technical Field

[0001] This application relates to the field of atmospheric data assimilation and carbon cycle monitoring, and in particular to a method and system for joint assimilation of multi-source carbon observation based on priority screening and data fusion. Background Technology

[0002] Accurate estimation of global and regional terrestrial ecosystem carbon sources and sinks is crucial for addressing climate change. The "top-down" approach based on atmospheric inversion systems, which couples atmospheric transport models with data assimilation algorithms to optimize surface carbon fluxes using atmospheric CO2 concentration observations, has become the mainstream method in international carbon cycle research. With the development of carbon satellite remote sensing technology, satellites such as GOSAT and OCO-2 have provided high-density XCO2 column concentration data at a global scale, offering opportunities to improve inversion resolution.

[0003] However, multi-source carbon observation data vary significantly in terms of accuracy, spatiotemporal coverage, and error characteristics: ground-based observations offer high accuracy (approximately 0.1-0.5 ppm) but have sparse stations; GOSAT data are stable with low random error but lack spatial coverage; OCO-2 data have broad coverage and high density but exhibit systematic bias in ocean regions and involve a massive amount of data. Existing co-assimilation techniques mainly suffer from the following drawbacks: 1. Data conflicts and computational redundancy: Traditional methods simply overlay all observations, leading to direct conflicts between high-precision ground-based data and a large amount of satellite data within the same grid, causing oscillations in the analysis field; at the same time, the massive amount of OCO-2 data increases the computational load dramatically.

[0004] 2. Insufficient information utilization: Simple prioritization (such as using only ground-based data) will discard the independent information contained in the unselected satellite data; while direct averaging will mask the statistical characteristics of various observations.

[0005] 3. Coarse spatial representativeness: Traditional assimilation treats observations within the localization radius equally, without continuous variation with distance, resulting in the same influence of distant and nearby observations, introducing representativeness error.

[0006] Therefore, there is an urgent need for a joint assimilation method that can effectively handle conflicts between multiple data sources, maximize the use of various types of observation information, and take into account both computational efficiency and spatial refinement. Summary of the Invention

[0007] The purpose of this application is to provide a multi-source carbon observation joint assimilation method and system based on priority screening and data fusion. It can solve problems such as multi-source data conflict, low computational efficiency, insufficient information utilization and coarse spatial representativeness through priority assimilation scheme and hybrid assimilation scheme, and achieve high-precision and high-efficiency inversion of surface carbon flux at a global 2°×2.5° grid and daily scale.

[0008] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a multi-source carbon observation joint assimilation method based on priority screening and data fusion, including: Acquire multi-source observation data; the multi-source observation data includes: ground-based data, GOSAT data and OCO-2 data in several grids; the ground-based data, GOSAT data and OCO-2 data are all observation data obtained after obtaining the background field based on the GEOS-Chem v12 atmospheric transport model.

[0009] The multi-source observation data is preprocessed to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening.

[0010] Based on the preprocessed data, a joint assimilation scheme is selected; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window and localization base radius.

[0011] When a preferred assimilation scheme is selected, the highest priority observation data is selected in each grid based on a preset priority rule; the priority rule is: ground-based data > GOSAT data > OCO-2 data.

[0012] The highest priority observation data is input into the 4D-LETKF (four-dimensional local ensemble transform Kalman filter) assimilation algorithm module to obtain the carbon flux scaling factor.

[0013] When a hybrid assimilation scheme is selected, the observation data in each grid are grouped by type, and within each group, a weighted average is calculated based on the inverse square of the original observation error to obtain several representative observations and their corresponding errors; at the same time, the weighted average position of each representative observation is calculated.

[0014] An extended observation operator is constructed, and an extended cost function is constructed based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator.

[0015] For each representative observation, calculate the spherical distance between its position and the center point of the grid.

[0016] Based on the spherical distance, adaptive localization weights are calculated using a Gaussian function.

[0017] The adjusted error covariance matrix is ​​obtained by multiplying the adaptive localization weights by the inverse of the observation error covariance matrix.

[0018] Based on each representative observation and the corresponding adjusted error covariance matrix, the extended cost function is solved to obtain the optimal state analysis value; the updated carbon flux scaling factor is determined according to the optimal state analysis value, and the updated carbon flux scaling factor is multiplied by the prior flux to obtain the posterior flux, which drives the forward modeling of the GEOS-Chemv12 atmospheric transport model to update the posterior CO2 concentration field.

[0019] Optionally, the expression representing the observed value and the corresponding error is: ; in, Representative observations; , k =1 represents the foundation. k =2 represents GOSAT. k =3 represents OCO-2; For the first i One observation data; For the first k The number of data types observed; For the first i Errors in the observed data; This represents the observation error.

[0020] Optionally, the expression for the extended cost function is: ; in, To expand the cost function; The optimal state analysis value to be solved; This refers to the background state at the initial moment of the assimilation window; The background error covariance matrix; To extend the observation operator to act on the state x The obtained number k Similar to simulated observation; Representative observations; , k =1 represents the foundation. k =2 represents GOSAT. k =3 represents OCO-2; For the first k The class represents the inverse of the observation error covariance matrix.

[0021] Optionally, the formula for calculating the spherical distance is: ; ; ; in, The distance is the spherical distance. A , N The Earth reference ellipsoid at latitudes P The radius of curvature of the meridian and the radius of curvature of the lateral meridian at that point; P The geographic latitude of the grid center point; dP , dR These represent the latitude and longitude differences between the observation point and the grid center point, respectively. R po , R eq These are the Earth's polar radius and equatorial radius, respectively. e The eccentricity of the Earth's ellipsoid.

[0022] Optionally, the formula for calculating the adaptive localization weight is: ; in, For adaptive localization weights; The distance is the spherical distance. The localization radius.

[0023] Optionally, the expression for the adjusted error covariance matrix is: ; in, This is the adjusted error covariance matrix; For adaptive localization weights; For the first k The class represents the inverse of the observation error covariance matrix.

[0024] Secondly, this application provides a multi-source carbon observation joint assimilation system based on priority screening and data fusion. The system is used to implement the aforementioned multi-source carbon observation joint assimilation method based on priority screening and data fusion. The system includes: The multi-source observation data acquisition module is used to acquire multi-source observation data; the multi-source observation data includes: ground-based data, GOSAT data and OCO-2 data in several grids; the ground-based data, GOSAT data and OCO-2 data are all observation data obtained after obtaining the background field based on the GEOS-Chem v12 atmospheric transport model.

[0025] The preprocessing module is used to preprocess the multi-source observation data to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening.

[0026] The scheme selection module is used to select a joint assimilation scheme based on the preprocessed data; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window and localization base radius.

[0027] The priority assimilation submodule is used to select the highest priority observation data in each grid based on a preset priority rule when a priority assimilation scheme is selected; the priority rule is: ground-based data > GOSAT data > OCO-2 data.

[0028] The input module is used to input the highest priority observation data into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor.

[0029] The hybrid assimilation submodule is used to group the observation data in each grid by type when a hybrid assimilation scheme is selected. Within each group, the data is fused by weighted average of the inverse square of the original observation error to obtain several representative observations and their corresponding errors. At the same time, the weighted average position of each representative observation is calculated.

[0030] An extended cost function construction module is used to construct an extended observation operator and construct an extended cost function based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator.

[0031] The first calculation module is used to calculate the spherical distance between the position of each representative observation and the center point of the grid.

[0032] An adaptive localized weight calculation module is used to calculate adaptive localized weights based on the spherical distance using a Gaussian function.

[0033] The second calculation module is used to multiply the adaptive localization weights with the inverse of the observation error covariance matrix to obtain the adjusted error covariance matrix.

[0034] The solution update module is used to solve the extended cost function based on each representative observation and the corresponding adjusted error covariance matrix to obtain the optimal state analysis value; determine the updated carbon flux scaling factor based on the optimal state analysis value, and multiply the updated carbon flux scaling factor by the prior flux to obtain the posterior flux, drive the forward modeling of the GEOS-Chem v12 atmospheric transport model, and update the posterior CO2 concentration field.

[0035] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-source carbon observation joint assimilation method based on priority screening and data fusion described above.

[0036] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned multi-source carbon observation joint assimilation method based on priority screening and data fusion.

[0037] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned multi-source carbon observation joint assimilation method based on priority screening and data fusion.

[0038] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a multi-source carbon observation joint assimilation method and system based on priority screening and data fusion. By acquiring multi-source observation data and performing preprocessing, it solves the problems of quality defects, duplication and redundancy, and disordered observation times in multi-source observation data. It achieves accurate purification and standardized screening of raw data, eliminates invalid and interfering data, and significantly improves the data foundation accuracy and reliability of subsequent assimilation operations. The method involves selecting the highest priority observation data within each grid based on preset priority rules when choosing a priority assimilation scheme. This highest priority observation data is then input into the 4D-LETKF assimilation algorithm. The method module obtains the carbon flux scaling factor; it solves the problem that single high-priority data cannot accurately invert carbon flux parameters, and realizes high-precision data assimilation based on the mature 4D-LETKF algorithm, quickly outputting reliable carbon flux scaling factors to support the initial carbon flux inversion; when selecting the hybrid assimilation scheme, the observation data in each grid are grouped by type, and within each group, the data is fused by weighted average based on the inverse square of the original observation error to obtain several representative observations and their corresponding errors; at the same time, the weighted average position of each representative observation is calculated, which solves the problem that the priority assimilation scheme cannot utilize multiple types of redundant observation data and data utilization. To address the issue of low efficiency, this paper achieves error-weighted fusion and precise correction of observation locations for similar types of observation data, maximizing the extraction of effective information from multi-source data and generating grid observation data with both accuracy and representativeness. By constructing an extended observation operator and building an extended cost function based on the extended observation operator and the representative observation values, the paper solves the problems of inconsistent dimensions and poor adaptability of assimilation cost functions for various types of carbon observation data, realizing the construction of an extended observation operator adapted to both ground-based and satellite data, and the construction of a dedicated extended cost function. Furthermore, by calculating the spherical distance between each representative observation value and the grid center point, adaptive localization weights are calculated using a Gaussian function. Multiplying the adaptive localization weights by the inverse of the observation error covariance matrix yields the adjusted error covariance matrix. This solves the problems of a fixed observation error covariance matrix failing to adapt to spatial differences and inaccurate assimilation error correction, enabling adaptive dynamic adjustment of the observation error covariance matrix and significantly improving the scientific rigor and accuracy of assimilation error correction. The extended cost function is solved based on each representative observation and the corresponding adjusted error covariance matrix to obtain the optimal state analysis value. The updated carbon flux scaling factor is multiplied by the prior flux to obtain the posterior flux, driving the forward modeling of the GEOS-Chem v12 atmospheric transport model and updating the posterior CO2 concentration field. This solves the problems of traditional assimilation models failing to simultaneously consider multi-source fusion data and spatial correction parameters, and insufficient accuracy in updating the posterior carbon flux and concentration field. It achieves optimal solution of the extended cost function and accurate iterative updates of carbon flux and CO2 concentration fields, completing efficient joint assimilation of multi-source carbon observation data.This application effectively addresses many pain points of traditional carbon observation assimilation techniques, such as single data source, low utilization rate, data conflicts, poor spatial adaptability, rigid error correction, and insufficient accuracy of assimilation results. It realizes efficient, accurate, and intelligent joint assimilation of multi-source heterogeneous carbon observation data, significantly improving the accuracy, stability, and spatiotemporal adaptability of carbon flux and atmospheric CO2 concentration field inversion, and providing high-precision technical support for regional and global carbon budget accounting, carbon cycle simulation, and accurate analysis of carbon sources and sinks. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is an application environment diagram of a multi-source carbon observation joint assimilation method based on priority screening and data fusion in one embodiment of this application; Figure 2 A flowchart illustrating a multi-source carbon observation joint assimilation method based on priority screening and data fusion, provided as an embodiment of this application; Figure 3 A schematic diagram illustrating the priority determination process of a priority assimilation scheme provided in an embodiment of this application; Figure 4 A schematic diagram of the adaptive localization fusion method for multi-source carbon observation data provided in an embodiment of this application; Figure 5 This is a schematic diagram illustrating data fusion and error recalibration of a hybrid assimilation scheme provided in an embodiment of this application; Figure 6 A schematic diagram of an adaptive localization weighting function curve based on spherical distance provided in an embodiment of this application; Figure 7 A bar chart comparing the posterior uncertainty reduction rate (URR) of preferred assimilation and mixed assimilation schemes provided in an embodiment of this application; Figure 8 A comparison of the spatial distribution of global posterior and prior carbon fluxes in 2018, retrieved from the assimilation system under a hybrid assimilation scheme provided in an embodiment of this application. Figure 9 A spatial distribution comparison of the global posterior carbon flux in 2018, the CO2 concentration simulated by the prior carbon flux, and the difference between the two, provided in an embodiment of this application under a hybrid assimilation scheme. Figure 10A schematic diagram of the functional modules of a multi-source carbon observation joint assimilation system based on priority screening and data fusion is provided in an embodiment of this application; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] This invention aims to provide a multi-source carbon observation joint assimilation method and system based on priority screening and data fusion. By using two complementary assimilation strategies—a priority assimilation scheme and a hybrid assimilation scheme—it solves problems such as multi-source data conflicts, low computational efficiency, insufficient information utilization, and coarse spatial representativeness, and achieves high-precision and high-efficiency inversion of surface carbon flux at a global 2°×2.5° grid and daily scale.

[0043] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Example 1: The multi-source carbon observation joint assimilation method based on priority screening and data fusion provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send acquired multi-source observation data to server 104. This multi-source observation data includes: ground-based data, GOSAT data, and OCO-2 data within several grids; the ground-based data, GOSAT data, and OCO-2 data are all based on GEOS-Chem. After obtaining the background field observation data using the v12 atmospheric transport model, server 104 receives the multi-source observation data and preprocesses it to obtain preprocessed data. The preprocessing includes quality screening, weight reduction, and local time screening. Based on the preprocessed data, a joint assimilation scheme is selected. The joint assimilation scheme includes a priority assimilation scheme and a hybrid assimilation scheme. The priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window, and localized base radius. When the priority assimilation scheme is selected, the highest priority observation data is selected within each grid based on a preset priority rule. The priority rule is: ground-based data > GOSAT data > OCO-2 data. The highest priority observation data is input into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor. When the hybrid assimilation scheme is selected, the observation data within each grid is processed according to... The data is grouped by type, and within each group, a weighted average of the inverse square of the original observation error is used to obtain several representative observations and their corresponding errors. Simultaneously, the weighted average position of each representative observation is calculated. An extended observation operator is constructed, and an extended cost function is built based on the extended observation operator and the representative observations. The extended observation operator includes a ground-based mapping operator and a satellite column concentration mapping operator. For each representative observation, the spherical distance between its position and the grid center point is calculated. Based on the spherical distance, an adaptive localization weight is calculated using a Gaussian function. The adaptive localization weight is multiplied by the inverse of the observation error covariance matrix to obtain an adjusted error covariance matrix. Based on each representative observation and the corresponding adjusted error covariance matrix, the extended cost function is solved to obtain the optimal state analysis value. The updated carbon flux scaling factor is multiplied by the prior flux to obtain the posterior flux, which drives the forward modeling of the GEOS-Chem v12 atmospheric transport model, updating the posterior CO2 concentration field. Server 104 can feed back the obtained optimal state analysis value and posterior CO2 concentration field to terminal 102.In addition, in some embodiments, the multi-source carbon observation joint assimilation method based on priority screening and data fusion can also be implemented by the server 104 or the terminal 102 separately. For example, the terminal 102 can directly perform multi-source carbon observation joint assimilation based on priority screening and data fusion on the multi-source observation data, or the server 104 can obtain multi-source observation data from the data storage system and perform multi-source carbon observation joint assimilation based on priority screening and data fusion on the multi-source observation data.

[0045] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, and tablets. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0046] Example 2: In one exemplary embodiment, such as Figure 2 As shown, a multi-source carbon observation joint assimilation method based on priority screening and data fusion is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included: S1: Acquire multi-source observation data; the multi-source observation data includes: ground-based data, GOSAT data and OCO-2 data in several grids; the ground-based data, GOSAT data and OCO-2 data are all observation data obtained after obtaining the background field based on the GEOS-Chem v12 atmospheric transport model.

[0047] S2: Preprocess the multi-source observation data to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening.

[0048] S3: Based on the preprocessed data, select a joint assimilation scheme; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window and localization base radius.

[0049] S4: When selecting a preferred assimilation scheme, the highest priority observation data is selected in each grid based on a preset priority rule; the priority rule is: ground-based data > GOSAT data > OCO-2 data.

[0050] S5: Input the highest priority observation data into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor.

[0051] S6: When a hybrid assimilation scheme is selected, the observation data in each grid are grouped by type, and within each group, the weighted average of the inverse square of the original observation error is used to obtain several representative observations and their corresponding errors; at the same time, the weighted average position of each representative observation is calculated.

[0052] S7: Construct an extended observation operator, and construct an extended cost function based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator.

[0053] S8: For each representative observation, calculate the spherical distance between its position and the center point of the grid.

[0054] S9: Calculate the adaptive localization weights using a Gaussian function based on the spherical distance.

[0055] S10: Multiply the adaptive localization weights by the inverse of the observation error covariance matrix to obtain the adjusted error covariance matrix.

[0056] S11: Based on each representative observation and the corresponding adjusted error covariance matrix, solve the extended cost function to obtain the optimal state analysis value; determine the updated carbon flux scaling factor according to the optimal state analysis value, and multiply the updated carbon flux scaling factor by the prior flux to obtain the posterior flux, drive the forward modeling of the GEOS-Chem v12 atmospheric transport model, and update the posterior CO2 concentration field.

[0057] By implementing steps S1 to S11 above, this application can achieve efficient and high-precision surface carbon flux inversion for multi-source heterogeneous carbon observation data such as ground-based CO2 observation, GOSAT satellite XCO2 and OCO-2 satellite XCO2, through two complementary joint assimilation strategies: priority screening and data fusion-expanded cost function.

[0058] Option 1: Prioritized Assimilation Scheme (Priority Screening).

[0059] This scheme uses a preset priority rule to select only the highest priority type of observation data for assimilation within each grid, thereby avoiding data conflicts and reducing computational load.

[0060] Step 1: Establish a priority rule base for observation data types.

[0061] Priority, from highest to lowest, is as follows: First priority: Continuous ground-based CO2 concentration observation data (ObsPack, filtered from 12:00 to 16:00 local time).

[0062] Second priority: First satellite (GOSAT) XCO2 column concentration data (quality marker = 0, warning level ≤ 7).

[0063] Third priority: XCO2 column concentration data from the second satellite (OCO-2) (mass marker = 0, warning level ≤ 15, weighted average after 1 second to 10 seconds).

[0064] Step 2: Detection and classification of observation data within the grid.

[0065] Within the assimilation window (7 days), all observations within the localized radius (2000km) of the current grid center are extracted and classified into the corresponding priority queues according to their type.

[0066] Step 3: Priority-based selection of observation data.

[0067] like Figure 3 As shown, the queue is checked sequentially from high priority to low priority: if the first priority is not empty, only ground-based observations are selected; otherwise, if the second priority is not empty, only GOSAT observations are selected; otherwise, OCO-2 observations are selected; if all are empty, the grid is skipped.

[0068] Step 4: Data assimilation analysis.

[0069] Input the selected observations into the 4D-LETKF analysis module to update the state variables (carbon flux scaling factor). λ The output analysis field is shown, and unselected observations are retained as independent validation data.

[0070] Option 2: Hybrid assimilation scheme (data fusion + extended cost function + adaptive localization).

[0071] This scheme fuses and compresses all observations into representative observations, constructs an extended cost function while utilizing multi-source information, and adaptively adjusts observation weights based on distance to achieve refined assimilation.

[0072] like Figure 4 As shown, it includes the following steps: Step 1: Classification, fusion and error recalibration of multi-source observation data.

[0073] The observations within the grid were divided into three groups according to type. Within each group, a weighted average was calculated based on the reciprocal of the square of the original observation error, such as... Figure 5 As shown, three representative observations and their errors are obtained: ; in, Representative observations; , k =1 represents the foundation. k =2 represents GOSAT. k=3 represents OCO-2; at the same time, the location representing the observation is calculated (each observation location is weighted and averaged with the same weight). For the first i One observation data; For the first k The number of data types observed; For the first i Errors in the observed data; This represents the observation error.

[0074] Step 2: Construct the extended observation operator and cost function.

[0075] Extended observation operator Includes ground-based mapping operator and satellite column concentration mapping operator The total cost function is: ; in, To expand the cost function; The optimal state analysis value to be solved; The background (prior) state at the initial moment of the assimilation window; The background error covariance matrix (obtained from the background perturbation estimation of the set members); To extend the observation operator to act on the state x The obtained number k Similar to simulated observation; For the first k The class represents the inverse of the observation error covariance matrix.

[0076] The optimal state is solved by minimizing the cost function and utilizing three types of observation information. .

[0077] Step 3: Distance-based adaptive localization.

[0078] For each representative observation, calculate the spherical distance between its position and the center point of the grid. : ; ; ; in, The distance is the spherical distance. A , N These represent the Earth reference ellipsoid (WGS84) at latitudes. P The radius of curvature of the meridian and the radius of curvature of the lateral meridian at that point; P The geographic latitude of the grid center point; dP , dRThese represent the latitude and longitude differences (in radians) between the observation point and the grid center point, respectively. R po , R eq These are the Earth's polar radius and equatorial radius (unit: meters); e The eccentricity of the Earth's ellipsoid.

[0079] The formula for calculating the adaptive localization weight is as follows: ; in, For adaptive localization weights; In this embodiment, the localization radius is taken as the localization radius. km. The adaptive localization weighting function curve based on spherical distance is shown in the figure. Figure 6 As shown.

[0080] It should be noted that, hour, (The observation point is the center and has the highest weight); km time, ; km time, The weights decrease continuously and smoothly with distance, avoiding sudden hard truncation within localized boundaries.

[0081] Multiplying the weights by the inverse of the observation error covariance matrix yields the adjusted error covariance matrix: ; in, This is the adjusted error covariance matrix.

[0082] Step 4: Co-assimilation analysis.

[0083] Input the three representative observations and their adjusted error covariance matrices into the 4D-LETKF kernel and substitute them into the standard analysis formula for solution. Since the number of observations is reduced to a maximum of three, the computational workload is minimal.

[0084] Among them, solving the extended cost function The obtained optimal state analysis value , and carbon flux scaling factor λ The relationship is explained as follows: The optimal state analysis value of the assimilation system in this application The carbon flux scaling factor component of each set member in each grid ( λ bio , λ ocn (These are scale factors corresponding to terrestrial ecosystems and oceanic prior fluxes, respectively), which constitute "solving the optimal state analysis value". In this application, it refers to "solving the scale factor". λ These two are different ways of representing the same variable: It is the state vector that is used as the solution object in the 4D-LETKF algorithm. λ yes The geoscientific names of the components directly related to carbon flux. Solving the cost function to obtain the optimal... After that, directly from Read the corresponding grid and the corresponding set member. λ bio , λ ocn The a posteriori natural flux is obtained by multiplying the two components by the prior natural flux, without the need for additional conversion steps.

[0085] In addition, the GEOS-Chem v12 atmospheric transport model has a horizontal resolution of 2°×2.5°, a vertical layer of 47, and a time step of 1 hour. The 4D-LETKF assimilation algorithm module has an assimilation window of 7 days, a localization base radius of 2000km, and a set of 48 members; the upper limit of the adaptive expansion factor of the background error covariance is 2.0.

[0086] In summary, this application has the following beneficial effects: 1. Flexibility and Compatibility: This application provides two solutions, which users can choose according to their computing resources and accuracy requirements. The preferred assimilation solution is suitable for rapid operational use, while the hybrid assimilation solution is suitable for high-precision scientific research.

[0087] 2. Completely avoid data conflicts: The priority assimilation scheme uses only one type of observation in each grid, eliminating contradictions between multiple data sources at the source; the hybrid assimilation scheme, through the fusion and expansion of the cost function, enables various types of observations to work synergistically under a unified framework, avoiding direct conflicts.

[0088] 3. Significantly improved computational efficiency: The priority assimilation scheme reduces computation by more than 40% (especially for ground-based grids). The hybrid assimilation scheme compresses the number of observations per grid to a maximum of 3, reducing computation time by more than 90% compared to full assimilation.

[0089] 4. Significantly improved inversion accuracy: Compared with single-foundation, GOSAT, and OCO-2 assimilation, the RMSE of the priority assimilation scheme is reduced by 10%, 16%, and 21%, respectively; compared with the priority assimilation scheme, the RMSE of the hybrid assimilation scheme is further reduced by 11%; the uncertainty reduction rate URR is increased from 4.6% to 10.1% (more than doubled).

[0090] 5. Refined Spatial Analysis: The Gaussian distance weight in the hybrid assimilation scheme makes the influence of observation continuously decay with distance, with clear physical meaning, avoiding hard truncation abrupt changes, and making the spatial distribution of posterior carbon flux more reasonable.

[0091] 6. Maximize information utilization: The hybrid assimilation scheme retains the high precision of the ground, the stability of GOSAT, and the high resolution of OCO-2 information simultaneously, with no information loss.

[0092] Example 3: This embodiment integrates two joint assimilation schemes based on the GEOS-Chem v12 atmospheric transport model and the 4D-LETKF assimilation algorithm module, and verifies them using global data from 2016 as an example.

[0093] 1. Data preparation.

[0094] Foundation data: ObsPack v7.0, filtered obs_flag=1, local time 12:00-16:00 hourly average, assigned MDM values ​​0.75-7.5ppm.

[0095] GOSAT data: ACOS V9, XCO2_quality_flag=0, warning level ≤7.

[0096] OCO-2 data: V10r, quality mark 0, warning level ≤15, weighted average weight reduction over 1-10 seconds.

[0097] Background flux: CT2019B (land + ocean, 3 hours), GCP-GridFED fossil fuels (month → hour), GFAS wildfires (day → hour).

[0098] 2. Implementation of Option 1: Prioritize assimilation.

[0099] Set the parameters ASSIM_MODE=priority and PRIORITY_LIST=surface,gosat,oco2 in the system configuration file.

[0100] Taking a grid in the eastern region as an example: During the assimilation window of July 1-7, 2016, 3 ground-based stations, 15 GOSATs, and 220 OCO-2s were detected. Three ground-based observations were prioritized for 4D-LETKF analysis, and λ was updated. Posterior flux showed that the carbon sink in this area increased by 12% compared to the prior. Unused satellite data were retained for independent validation.

[0101] Taking a grid in the central Pacific Ocean as an example: without a ground-based system, there are 4 GOSAT flare observations and 90 OCO-2 observations. The system selects GOSAT observations, and the marine carbon sink has increased slightly.

[0102] Verification: For the 20% of site data that did not participate in assimilation, the preferred assimilation scheme had an RMSE of 1.63ppm, which is lower than surface-only (1.79ppm), GOSAT-only (2.10ppm), and OCO-2-only (2.42ppm).

[0103] 3. Implementation of Option Two: Hybrid Assimilation.

[0104] Set ASSIM_MODE=hybrid in the system configuration file.

[0105] Taking a grid in North America as an example: there are 2 foundations (σ≈0.6ppm), 20 GOSAT (σ≈1.4ppm), and 180 OCO-2 (σ≈1.2ppm) within the grid.

[0106] Fusion: Calculating three representative observations: ppm, ppm; ppm, ppm; ppm, ppm; Distance calculation: The distance from the ground station to the center is 65km (w≈0.999), the distance from the GOSAT representative point to the center is 210km (w≈0.995), and the distance from the OCO-2 representative point to the center is 950km (w≈0.89).

[0107] Solving the extended cost function yields the optimal scaling factor λ = 0.92 (prior = 1.0), with a posterior carbon sink enhancement of 8%.

[0108] Verification: such as Figure 7 As shown, the URR of the hybrid assimilation scheme reached 10.1%, which is more than double that of the preferred assimilation scheme (4.6%), indicating that multi-source information fusion can significantly reduce inversion uncertainty.

[0109] Among them, the definition ; , These are the posterior and prior carbon flux uncertainties, respectively (calculated from the standard deviations of the analysis set perturbation and the background set perturbation).

[0110] 4. Global inversion results.

[0111] Optimize global carbon flux from 2016 to 2018 based on a hybrid assimilation scheme: The global annual average carbon flux to terrestrial ecosystems is -2.75 PgC / yr, to the oceans -3.56 PgC / yr, to the land net exchange -0.89 PgC / yr, and the atmospheric CO2 growth rate is 5.03 PgC / yr (consistent with NOAA observations).

[0112] The average annual carbon flux of terrestrial ecosystems in China is -0.25 PgC / yr.

[0113] The spatial distribution of posterior carbon fluxes shows an increase in carbon sinks in eastern North America, most of Russia, and northeastern China; and an increase in carbon sources in southeastern South America and parts of Europe. The spatial distribution of global posterior and prior carbon fluxes in 2018, derived from the assimilation system under the mixed assimilation scheme, is shown below. Figure 8 As shown, the spatial distribution of the 2018 global posterior carbon flux retrieved by the assimilation system under the mixed assimilation scheme, the CO2 concentration simulated by the prior carbon flux, and the difference between the two are as follows: Figure 9 As shown.

[0114] 5. Recommendations for Option Selection

[0115] Operational rapid forecasting: Prioritize assimilation schemes (low computational cost, high stability).

[0116] Scientific research and high-precision assessment: Hybrid assimilation scheme (highest accuracy and most efficient use of information).

[0117] Example 4: Based on the same inventive concept, this application also provides a multi-source carbon observation joint assimilation system based on priority screening and data fusion for implementing the multi-source carbon observation joint assimilation method based on priority screening and data fusion described above. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more embodiments of the multi-source carbon observation joint assimilation system based on priority screening and data fusion provided below can be found in the limitations of the multi-source carbon observation joint assimilation method based on priority screening and data fusion described above, and will not be repeated here.

[0118] In one exemplary embodiment, such as Figure 10 As shown, a multi-source carbon observation joint assimilation system based on priority screening and data fusion is provided, including: The multi-source observation data acquisition module is used to acquire multi-source observation data; the multi-source observation data includes: ground-based data, GOSAT data and OCO-2 data in several grids; the ground-based data, GOSAT data and OCO-2 data are all observation data obtained after obtaining the background field based on the GEOS-Chem v12 atmospheric transport model.

[0119] The preprocessing module is used to preprocess the multi-source observation data to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening.

[0120] The scheme selection module is used to select a joint assimilation scheme based on the preprocessed data; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window and localization base radius.

[0121] The priority assimilation submodule is used to select the highest priority observation data in each grid based on a preset priority rule when a priority assimilation scheme is selected; the priority rule is: ground-based data > GOSAT data > OCO-2 data.

[0122] The input module is used to input the highest priority observation data into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor.

[0123] The hybrid assimilation submodule is used to group the observation data in each grid by type when a hybrid assimilation scheme is selected. Within each group, the data is fused by weighted average of the inverse square of the original observation error to obtain several representative observations and their corresponding errors. At the same time, the weighted average position of each representative observation is calculated.

[0124] An extended cost function construction module is used to construct an extended observation operator and construct an extended cost function based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator.

[0125] The first calculation module is used to calculate the spherical distance between the position of each representative observation and the center point of the grid.

[0126] An adaptive localized weight calculation module is used to calculate adaptive localized weights based on the spherical distance using a Gaussian function.

[0127] The second calculation module is used to multiply the adaptive localization weights with the inverse of the observation error covariance matrix to obtain the adjusted error covariance matrix.

[0128] The solution update module is used to solve the extended cost function based on each representative observation and the corresponding adjusted error covariance matrix to obtain the optimal state analysis value; determine the updated carbon flux scaling factor based on the optimal state analysis value, and multiply the updated carbon flux scaling factor by the prior flux to obtain the posterior flux, drive the forward modeling of the GEOS-Chem v12 atmospheric transport model, and update the posterior CO2 concentration field.

[0129] Example 5: In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 11As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores multi-source observation data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a multi-source carbon observation co-assimilation method based on priority screening and data fusion.

[0130] Figure 11 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0131] Example 6: In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0132] Example 7: In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0133] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.

[0134] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0135] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0136] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A multi-source carbon observation joint assimilation method based on priority screening and data fusion, characterized in that, include: Acquire multi-source observation data; The multi-source observation data includes: ground-based data, GOSAT data, and OCO-2 data within several grids; the ground-based data, GOSAT data, and OCO-2 data are all observation data obtained after acquiring the background field based on the GEOS-Chem v12 atmospheric transport model; The multi-source observation data is preprocessed to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening. Based on the preprocessed data, a joint assimilation scheme is selected; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window, and localization base radius; When selecting a preferred assimilation scheme, the highest priority observation data is selected within each grid based on a preset priority rule; the priority rule is: ground-based data > GOSAT data > OCO-2 data. The highest priority observation data is input into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor; When a hybrid assimilation scheme is selected, the observation data in each grid are grouped by type, and within each group, the weighted average of the inverse square of the original observation error is used to obtain several representative observations and their corresponding errors; at the same time, the weighted average position of each representative observation is calculated. An extended observation operator is constructed, and an extended cost function is constructed based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator; For each representative observation, calculate the spherical distance between its position and the center point of the grid; Based on the spherical distance, adaptive localization weights are calculated using a Gaussian function; Multiply the adaptive localization weights by the inverse of the observation error covariance matrix to obtain the adjusted error covariance matrix; Based on each representative observation and the corresponding adjusted error covariance matrix, the extended cost function is solved to obtain the optimal state analysis value; the updated carbon flux scaling factor is determined according to the optimal state analysis value, and the updated carbon flux scaling factor is multiplied by the prior flux to obtain the posterior flux, which drives the forward modeling of the GEOS-Chem v12 atmospheric transport model to update the posterior CO2 concentration field.

2. The multi-source carbon observation joint assimilation method based on priority screening and data fusion according to claim 1, characterized in that, The expression representing the observed value and the corresponding error is: ； in, Representative observations; , k =1 represents the foundation. k =2 represents GOSAT. k =3 represents OCO-2; For the first i One observation data; For the first k The number of data types observed; For the first i Errors in the observed data; This represents the observation error.

3. The multi-source carbon observation joint assimilation method based on priority screening and data fusion according to claim 1, characterized in that, The expression for the extended cost function is: ; in, To expand the cost function; The optimal state analysis value to be solved; This refers to the background state at the initial moment of the assimilation window; The background error covariance matrix; To extend the observation operator to act on the state x The obtained number k Similar to simulated observation; Representative observations; , k =1 represents the foundation. k =2 represents GOSAT. k =3 represents OCO-2; For the first k The class represents the inverse of the observation error covariance matrix.

4. The multi-source carbon observation joint assimilation method based on priority screening and data fusion according to claim 1, characterized in that, The formula for calculating the spherical distance is: ； ； ； in, The distance is the spherical distance. A , N The Earth reference ellipsoid at latitudes P The radius of curvature of the meridian and the radius of curvature of the lateral meridian at that point; P The geographic latitude of the grid center point; dP , dR These represent the latitude and longitude differences between the observation point and the grid center point, respectively. R po , R eq These are the Earth's polar radius and equatorial radius, respectively. e The eccentricity of the Earth's ellipsoid.

5. The multi-source carbon observation joint assimilation method based on priority screening and data fusion according to claim 1, characterized in that, The formula for calculating the adaptive localization weight is as follows: ； in, For adaptive localization weights; The distance is the spherical distance. The localization radius.

6. The multi-source carbon observation joint assimilation method based on priority screening and data fusion according to claim 1, characterized in that, The expression for the adjusted error covariance matrix is: ； in, This is the adjusted error covariance matrix; For adaptive localization weights; For the first k The class represents the inverse of the observation error covariance matrix.

7. A multi-source carbon observation joint assimilation system based on priority screening and data fusion, characterized in that, The system is used to implement the multi-source carbon observation joint assimilation method based on priority screening and data fusion as described in any one of claims 1-6. The system comprises: A multi-source observation data acquisition module is used to acquire multi-source observation data; the multi-source observation data includes: ground-based data, GOSAT data and OCO-2 data within several grids; the ground-based data, GOSAT data and OCO-2 data are all observation data obtained after obtaining the background field based on the GEOS-Chem v12 atmospheric transport model; The preprocessing module is used to preprocess the multi-source observation data to obtain preprocessed data; the preprocessing includes: quality screening, deduplication, and local time screening; The scheme selection module is used to select a joint assimilation scheme based on the preprocessed data; the joint assimilation scheme includes: a priority assimilation scheme and a hybrid assimilation scheme; the priority assimilation scheme and the hybrid assimilation scheme share the 4D-LETKF assimilation algorithm module, assimilation window and localization base radius; The priority assimilation submodule is used to select the highest priority observation data in each grid based on a preset priority rule when a priority assimilation scheme is selected; the priority rule is: ground-based data > GOSAT data > OCO-2 data. The input module is used to input the highest priority observation data into the 4D-LETKF assimilation algorithm module to obtain the carbon flux scaling factor; The hybrid assimilation submodule is used to group the observation data in each grid by type when a hybrid assimilation scheme is selected. Within each group, the data is fused by weighted average of the inverse square of the original observation error to obtain several representative observations and their corresponding errors. At the same time, the weighted average position of each representative observation is calculated. An extended cost function construction module is used to construct an extended observation operator and construct an extended cost function based on the extended observation operator and the representative observation value; the extended observation operator includes: a ground-based mapping operator and a satellite column concentration mapping operator; The first calculation module is used to calculate the spherical distance between the position of each representative observation and the center point of the grid. An adaptive localization weight calculation module is used to calculate adaptive localization weights based on the spherical distance using a Gaussian function. The second calculation module is used to multiply the adaptive localization weights with the inverse matrix of the observation error covariance to obtain the adjusted error covariance matrix. The solution update module is used to solve the extended cost function based on each representative observation and the corresponding adjusted error covariance matrix to obtain the optimal state analysis value; determine the updated carbon flux scaling factor based on the optimal state analysis value, and multiply the updated carbon flux scaling factor by the prior flux to obtain the posterior flux, drive the forward modeling of the GEOS-Chem v12 atmospheric transport model, and update the posterior CO2 concentration field.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the multi-source carbon observation co-assimilation method based on priority screening and data fusion as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the multi-source carbon observation co-assimilation method based on priority screening and data fusion as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the multi-source carbon observation co-assimilation method based on priority screening and data fusion as described in any one of claims 1-6.

Images