Three-dimensional simulation method and system of atmospheric carbon dioxide profile based on multi-source data fusion of sky and ground

Abstract

The present application relates to the field of data processing technology, and provides a method and system for simulating atmospheric CO2 profiles in three dimensions based on multi-source data fusion from the sky and the ground. Based on the atmospheric chemical transport model, the simulation results of CO2 profile distribution are obtained, and multi-source CO2 concentration observation data based on space, air and ground are obtained. Based on the CarbonTracker model data, the air-based CO2 concentration observation data is subjected to sensitivity correction and spatiotemporal scale normalization. The multi-source CO2 concentration observation data based on space, air and ground are processed by machine learning methods respectively, and these data are integrated and fused from the time scale, spatial scale and vertical scale to obtain a multi-source observation fusion data set. The multi-source observation fusion data set is used as the optimization control condition, and the spatiotemporal continuous CO2 profile simulated by the atmospheric chemical transport model is used as the initial value. Based on the high-precision surface modeling HASM method, the optimal estimated value of CO2 concentration in each atmospheric layer is solved, and the three-dimensional fine characterization capability of CO2 concentration in local areas is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an atmospheric CO 2 2 Method and system for three-dimensional simulation of profiles. Background Art

[0002] Carbon dioxide (CO 2 ) is the greenhouse gas most affected by human activities. In recent years, atmospheric CO 2 As the content of carbon dioxide continues to increase, accurate estimation of carbon emissions has become an important issue that needs to be solved urgently. 2 Concentration data have important scientific significance and application value in describing the spatiotemporal distribution characteristics of regional carbon sources and sinks, optimizing assimilation model inputs, and improving the understanding of carbon cycle mechanisms. 2 Accurately characterizing the spatial distribution of carbon is the key to research related to carbon sources and sinks and has attracted widespread attention from scholars.

[0003] However, due to the atmospheric CO 2 The complexity of the temporal and spatial distribution of concentration, high-precision and high-temporal and high-resolution CO 2 There is still a large uncertainty in the results of spatial distribution simulation. From the existing research, domestic and foreign researchers have used methods such as actual observation, data interpolation, and atmospheric transport simulation to estimate atmospheric CO at the site scale, regional scale, and global scale. 2 A large number of simulation studies have been conducted on the spatial distribution of concentrations. Atmospheric models have the ability to simulate CO on an hourly scale. 2 However, the resolution and accuracy of the simulation results depend on the emission inventory dataset in the input source, and cannot characterize the near-surface CO2 generated by local anthropogenic emissions. 2 The concentration changes fluctuate, and there is great uncertainty in regional and local monitoring data; the observation data are more accurate than the model simulation results and can monitor the atmospheric CO caused by emission events. 2 Fluctuation conditions characterize the characteristics of temporal and spatial changes, but due to the limitations of observation methods, the amount of observation data is limited, and there is a problem of discontinuous spatial distribution. At the same time, certain observation errors will occur in the process of instrument calibration and inversion.

[0004] Interpolation and fusion research based on measured data can fill the gaps of the two research methods. However, so far, related research is still mainly focused on the fusion of observation data on the same platform (between satellites), or the fusion of single-platform data (ground observations or satellite-based observations) and model simulation results. It has not been able to fully utilize the advantages of the "sky-air-ground" multi-source observation data set. The fusion results of single-satellite or single-platform observation data cannot meet the growing data needs for high temporal and spatial resolution and high precision.

[0005] Therefore, it is urgent to integrate existing research progress, develop technical methods with innovative ideas, develop multi-platform and cross-scale fusion approaches, and integrate the advantages of multiple sources. Summary of the invention

[0006] The purpose of this application is to provide an atmospheric CO 2 A three-dimensional contour simulation method and system are provided to solve or alleviate the problems existing in the above-mentioned prior art.

[0007] In order to achieve the above objectives, this application provides the following technical solutions:

[0008] This application provides an atmospheric CO 2 2 The three-dimensional profile simulation method includes:

[0009] Based on the atmospheric chemical transport model, CO 2 Profile distribution simulation results;

[0010] Acquire multi-source CO from space, air and ground 2 Concentration observation data;

[0011] Among them, the CO 2 The profile distribution simulation results include CO2 in multiple atmospheric layers in the vertical direction. 2 Profile information; space-based CO 2 Concentration observations include CO2 measurements from tethered balloons, towers, and / or aircraft flights. 2 Concentration data; air-based CO 2 Concentration observation data include multiple global reversible XCO 2 Satellite observations of CO 2 concentration data, and the observation platforms carried by each satellite are different; ground-based CO 2 Concentration observation data include CO2 obtained from background observation stations and carbon column observation stations used to obtain carbon dioxide column concentration. 2 Concentration observation data;

[0012] Based on machine learning method to establish surface CO 2 Concentration, wind speed, pressure profile, humidity profile, temperature profile and the first atmospheric CO 2 The first regression model between the profiles, wherein the first atmospheric CO 2 The profile is related to the ground CO 2 CO concentration observations associated with 2 profile, and the first atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is used as a benchmark to determine the surface CO 2The concentration is the CO obtained from the background observation station 2 concentration;

[0013] Establishing a tethered balloon CO based on machine learning method 2 Profile and tower observation of CO 2 Profile, aircraft flight observation CO 2 Profile, wind speed, pressure profile, humidity profile, temperature profile and the second atmospheric CO 2 The second regression model between the profiles; wherein the second atmospheric CO 2 The profile is consistent with the space-based CO 2 CO concentration observations associated with 2 profile, and the second atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is determined as a benchmark;

[0014] Based on the CarbonTracker model data, the airborne CO 2 The concentration observation data were subjected to sensitivity correction and time and space scale normalization;

[0015] The first CO 2 Profile data, second CO 2 Profile data, pressure profile, temperature profile, humidity profile and the third atmospheric CO 2 The third regression model between the profiles is used to obtain a multi-source observation fusion data set; the first CO 2 The profile data is the first atmospheric CO 2 CO profiles and carbon column observation stations 2 The sum of the concentration observations, the second CO 2 The profile data are the CO2 of tethered balloons, towers and aircraft predicted by the second regression model. 2 The sum of the profiles; the third atmospheric CO 2 The profile is related to the air-based CO 2 CO concentration observations associated with 2 Profile;

[0016] Taking the multi-source observation fusion data set as the optimization control condition, the CO 2 The profile distribution simulation results are the initial values. Based on the high-precision surface modeling HASM method, the CO 2 The best estimate of atmospheric CO concentration 2 3D simulation results of profiles.

[0017] As a preferred solution, the air-based CO2 The concentration observation data is subjected to sensitivity correction, specifically: based on the profile stratification of the CarbonTracker model data, the observation data of each satellite is subjected to stratified expansion processing to obtain the first CO 2 Concentration distribution data, the first CO 2 The profile stratification of the concentration distribution data is consistent with the profile stratification of the CarbonTracker model data;

[0018] The average kernel function and the first CO used in the inversion algorithm for each satellite generated based on the hierarchical expansion process 2 Concentration distribution data, calculate the response difference value of each satellite in the vertical direction;

[0019] The observation data of each satellite and its response difference value are summed to obtain the sensitivity-corrected CO 2 Observation profile.

[0020] As a preferred solution, the spatiotemporal scale normalization process includes:

[0021] Based on the preset reference time, the observation time of each satellite is converted to the reference time;

[0022] The study area is divided into grids according to the preset spatial resolution, and the coverage of each grid is used as the spatial scale benchmark.

[0023] As a preferred solution, the atmospheric chemical transport model is a GEOS-Chem model or a WRF-Chem model.

[0024] As a preferred solution, the multi-source observation fusion data set is used as the optimization control condition, the CO 2 The profile distribution simulation results are the initial values. Based on the high-precision surface modeling HASM method, the CO 2 Best estimates of concentrations, including:

[0025] Construct a HASM optimization controller based on multi-source observation fusion data at point scale;

[0026] According to the research area and the preset simulation resolution, the finite difference method is used to calculate the CO 2 Discretization of the partial differential equations of the concentration surface and calculation to form the HASM numerical simulation solver;

[0027] By the CO 2 The profile distribution simulation results are used as initial values ​​to drive the HASM numerical simulation solver to perform iterative calculations to obtain the CO 2 The best estimate of the concentration.

[0028] As a preferred solution, the number of layers of the atmospheric stratification is 25.

[0029] The present application provides an atmospheric CO 2 2 Profile 3D simulation system, including:

[0030] The first acquisition unit is configured to acquire CO based on an atmospheric chemical transport model. 2 Profile distribution simulation results;

[0031] The second acquisition unit is configured to acquire multi-source CO from space, air and ground sources. 2 Concentration observation data;

[0032] Among them, the CO 2 The profile distribution simulation results include CO2 in multiple atmospheric layers in the vertical direction. 2 Profile information; space-based CO 2 Concentration observations include CO2 measurements from tethered balloons, towers, and / or aircraft flights. 2 Concentration data; air-based CO 2 Concentration observation data include multiple global reversible XCO 2 Satellite observations of CO 2 concentration data, and the observation platforms carried by each satellite are different; ground-based CO 2 Concentration observation data include CO2 obtained from background observation stations and carbon column observation stations used to obtain carbon dioxide column concentration. 2 Concentration observation data;

[0033] The first modeling unit is configured to establish a surface CO 2 Concentration, wind speed, pressure profile, humidity profile, temperature profile and the first atmospheric CO 2 The first regression model between the profiles, wherein the first atmospheric CO 2 The profile is related to the ground CO 2 CO concentration observations associated with 2 profile, and the first atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is used as a benchmark to determine the surface CO 2 The concentration is the CO obtained from the background observation station 2 concentration;

[0034] The second modeling unit is configured to establish a tethered balloon CO based on a machine learning method. 2 Profile and tower observation of CO 2 Profile, aircraft flight observation CO 2Profile, wind speed, pressure profile, humidity profile, temperature profile and the second atmospheric CO 2 The second regression model between the profiles; wherein the second atmospheric CO 2 The profile is consistent with the space-based CO 2 CO concentration observations associated with 2 profile, and the second atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is determined as a benchmark;

[0035] A calibration unit is configured to calibrate the airborne CO based on CarbonTracker model data. 2 The concentration observation data were subjected to sensitivity correction and time and space scale normalization;

[0036] A third modeling unit is configured to establish the first CO based on a machine learning method 2 Profile data, second CO 2 Profile data, pressure profile, temperature profile, humidity profile and the third atmospheric CO 2 The third regression model between the profiles is used to obtain a multi-source observation fusion data set; the first CO 2 The profile data is the first atmospheric CO 2 CO profiles and carbon column observation stations 2 The sum of the concentration observations, the second CO 2 The profile data are the CO2 of tethered balloons, towers and aircraft predicted by the second regression model. 2 The sum of the profiles; the third atmospheric CO 2 The profile is related to the air-based CO 2 CO concentration observations associated with 2 Profile;

[0037] The simulation analysis unit is configured to use the multi-source observation fusion data set as the optimization control condition, the CO 2 The profile distribution simulation results are the initial values. Based on the high-precision surface modeling HASM method, the CO 2 The best estimate of atmospheric CO concentration 2 3D simulation results of profiles.

[0038] The technical solution of the embodiment of the present application has the following beneficial effects:

[0039] Based on the atmospheric chemical transport model, CO 2 Profile distribution simulation results, obtain multi-source CO from space, air and ground 2 Concentration observation data, based on CarbonTracker model data, for air-based CO2 The concentration observation data were subjected to sensitivity correction and time-space scale normalization, and machine learning methods were used to analyze the multi-source CO2 from space, air and ground sources. 2 The concentration observation data are processed and integrated from the time scale, spatial scale and vertical scale to obtain a multi-source observation fusion data set. The spatiotemporal continuous CO concentration in each atmospheric layer is simulated by the atmospheric chemical transport model using the multi-source observation fusion data set as the optimization control condition. 2 The profile is the initial value, and the CO 2 The best estimate of global and regional atmospheric CO concentrations 2 High-precision simulation of the three-dimensional spatial distribution of profiles, while improving the CO 2 Concentration three-dimensional fine characterization capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. Among them:

[0041] Figure 1 A schematic flow chart of a three-dimensional simulation method of atmospheric CO2 profile based on sky-ground multi-source data fusion according to some embodiments of the present application.

[0042] Figure 2 A schematic diagram of the structure of a three-dimensional simulation system for atmospheric CO2 profile based on sky-ground multi-source data fusion according to some embodiments of the present application. DETAILED DESCRIPTION

[0043] At present, the multi-source and multi-scale atmospheric CO 2 The concentration observation data and simulation data are optimized and integrated to generate high-precision and high-resolution atmospheric CO 2 concentration dataset to improve atmospheric CO 2 The ability to characterize spatiotemporal variation characteristics improves atmospheric CO 2 The accuracy of distribution data is a preface and scientific difficulty in the research fields of global change and carbon sources and sinks, and has certain development potential. 2 Data is an effective way to solve the spatiotemporal variation process and observation simulation uncertainty problems faced by carbon sources and sinks. In recent years, multi-source data fusion has become a hot topic in geoscience research. 2 The integration of multi-source data is difficult. 2 The theory and methods of concentration data fusion still need to be further developed and improved.

[0044] Reducing CO from multiple sources 2 The effective means of data uncertainty method is to weigh, match and optimize the combination of multi-source and multi-scale data in a certain way, consider the error characteristics and advantages of different results, and obtain more detailed information than any single data, so as to reduce atmospheric CO 2 The uncertainty of the estimate is reduced, and accurate carbon source and sink data are finally obtained.

[0045] At present, in terms of satellite data fusion, some technologies use the spatial temporal data fusion (STDF) method based on dimensionality reduction Kalman smoothing to achieve the fusion of GOSAT and AIRS satellite data, which mainly solves the problems caused by differences in instrument parameters and measurement errors during fusion. Based on the above technology, some scholars later developed a method using an improved spatial random effect model to illustrate the multivariate nature of fusion. Using OCO-2 and GOSAT observation data, by using vector valued CO 2 The best estimate of the curve and the application of the pressure weight function for data fusion are better than the direct fusion results. In another technical solution, GOSAT and AIRS data are fused based on an improved geostatistical method, and the climate factor is used as an independent variable for regression analysis. The mean square prediction error of the fused data is smaller than that of the single GOSAT data set, and the effect is good. Some other solutions use CO2 data from different satellite sensors such as SCIAMACHY, GOSAT, and OCO-2 to obtain the best prediction error of the fused data. 2 Sensitivity correction, time-matched multi-source satellite data are compared for temporal and spatial consistency, and then normalized. A long-term spatiotemporal continuous grid mapping method is developed based on geostatistical methods to generate a set of global XCO with unified scale. 2 In terms of the fusion of model simulation and observation data, some technical solutions have achieved multivariate CO 2 Observational data fusion, GEOS-Chem model output XCO 2 The XCO with better global scale accuracy was obtained by using the TCCON ground observation data as the driving field and optimization control conditions. 2 Other technical solutions use HASM method fusion based on TanSat calibration data and ground real-time observation data, and cross-validate the fusion results locally. Later, some technologies used HASM to simulate global XCO based on TCCON and GEOS-Chem data. 2 Concentration distribution. 2 Based on the characteristics of spatiotemporal distribution, some scholars have used the HASM method to analyze the XCO of GOSAT in the mesoscale region. 2The data and WRF-Chem model simulated the high-precision XCO in the Beijing-Tianjin-Hebei region 2 Distribute data.

[0046] However, current methods mostly integrate data from the same platform (multiple satellite observations, or multiple ground observation sources), and fail to fully utilize the advantages of multi-source data and achieve cross-scale integration of data from different payload platforms (satellite + ground + flight) and models. This is mainly due to the atmospheric CO 2 The influence of profile stratification differences, observation scales, load sensor errors, and other factors make it impossible to effectively control the uncertainty of the fusion results in spatial distribution simulation. 2 The significant variability of profiles at different scales is important for multi-source and multi-scale CO 2 Data fusion poses huge challenges.

[0047] To this end, an embodiment of the present invention provides a three-dimensional simulation method for atmospheric CO2 profiles based on multi-source data fusion of sky and ground. The method connects point and surface data at spatial scales on the basis of sufficient data support provided by ground-based observations, aerial observations, satellite-based observations and atmospheric model data from different temporal and spatial scales and accuracies. Based on the existing observation data, on the basis of comprehensive comparative analysis of existing model simulation results and remote sensing data, a cross-scale downscaling method considering spatial heterogeneity is adopted to make full use of the complementarity of different data to study new high-resolution atmospheric CO2 profiles. 2 The spatial distribution estimation method of concentration is used to simulate atmospheric CO more accurately. 2 The spatial distribution provides methodological reference.

[0048] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. Each example is provided by way of explanation of the present application and does not limit the present application. In fact, it will be clear to those skilled in the art that modifications and variations may be made in the present application without departing from the scope or spirit of the present application. For example, a feature shown or described as a part of an embodiment may be used in another embodiment to produce yet another embodiment. Therefore, it is desired that the present application includes such modifications and variations within the scope of the appended claims and their equivalents.

[0049] The embodiment of the present invention provides an atmospheric CO2 emission detector based on multi-source data fusion of sky and ground. 2 Three-dimensional contour simulation methods, such as Figure 1 As shown, the method includes:

[0050] Step S101: Obtain CO based on the atmospheric chemical transport model 2 Profile distribution simulation results.

[0051] Atmospheric chemical transport models (also called atmospheric model models or model simulations for short) are a type of numerical models used to simulate the chemical substances (such as CO 2 In this embodiment, the atmospheric chemical transport model can take advantage of the numerical prediction model and use the CO 2 The source emission information, atmospheric transport and diffusion, and ground-atmosphere exchange processes can be used to achieve global and regional CO 2 Simulation of distribution, change and transport to achieve high temporal and spatial resolution of surface CO 2 Sources and sinks, atmospheric CO 2 Simulate the spatiotemporal distribution of concentration and obtain CO 2 Profile distribution simulation results (i.e., CO 2 outline).

[0052] In this embodiment, a common atmospheric chemical transport model can be used to obtain CO according to actual needs. 2 The profile distribution simulation result may be, for example, a GEOS-Chem model, a CAM-Chem model, a WRF-Chem model, a CAMx model, etc. This embodiment does not limit the selection of the atmospheric chemical transport model.

[0053] In this embodiment, CO 2 The profile distribution simulation results include CO2 in multiple atmospheric layers in the vertical direction. 2 Profile information, that is, CO 2 The profile distribution simulation results can reflect the atmospheric CO 2 The concentration changes vertically with the change of altitude or pressure. Atmospheric stratification refers to dividing the atmosphere into multiple layers according to altitude or pressure in the vertical direction. The number and spatial interval of the layers are related to the vertical resolution.

[0054] It should be noted that the output of the atmospheric chemical transport model is CO 2 The profile distribution simulation results are essentially the realization of atmospheric CO 2 The three-dimensional simulation of the profile is only due to the low resolution and accuracy of the simulation results of the atmospheric chemical transport model. 2 The characterization capability is relatively weak, and it is necessary to use the method provided in this application to improve its resolution and accuracy by fusing multi-source data, while reducing its uncertainty in local characterization.

[0055] As a preferred embodiment, the atmospheric chemical transport model uses the GEOS-Chem model or the WRF-Chem model, wherein the GEOS-Chem model is used to calculate the global atmospheric CO 2 The three-dimensional spatial distribution is simulated, and the WRF-Chem model is used to simulate the atmospheric CO2 The three-dimensional spatial distribution is simulated, and the simulation results of both are atmospheric CO 2 Profile data.

[0056] For ease of understanding, the GEOS-Chem model and the WRF-Chem model of this embodiment and related parameters are briefly introduced below.

[0057] (1) GEOS-Chem model

[0058] The GEOS-Chem model is a global three-dimensional atmospheric chemistry model driven by meteorological input from NASA's Goddard Earth Observing System (GEOS). The model code and input data can be obtained from the official website. In this example, it is used for global atmospheric CO 2 Spatially continuous distribution simulation. GEOS-Chem plans to use the V12 version, and the input data uses the model's default GEOS-FP meteorological field data. The model simulates and outputs multiple atmospheric layers in the vertical direction, preferably 25 layers, from 1000hPa to 0.01hPa, and the horizontal resolution is 2° longitude × 2.5° latitude, with a three-hourly grid concentration. Among them, the monthly fossil fuel emissions data used in the GEOS-Chem model comes from the Carbon Dioxide Information Analysis Center (CDIAC inventory), version 2010, and the biomass emissions data uses the GFED3 daily emissions data (daily GFED3 emissions).

[0059] (2) WRF-Chem model

[0060] The WRF-Chem model can use real data to generate atmospheric simulation or idealized conditions. It provides a flexible and efficient computing platform with high-resolution simulation capabilities, and is very suitable for research work at the mesoscale (mesoscale can be, for example, a city, a province, or a country). The WRF-Chem model version used in this embodiment is WRF V3.5, and the meteorological input data is the reanalysis data (ERA-interim) released by the European Centre for Medium-Range Weather Forecasts (ECMWF) and the FNL (Final Reanalysis Data) reanalysis data released by the National Centers for Environmental Prediction (NCEP). The spatial resolution is 1° longitude × 1° latitude, and the relevant meteorological parameters in the simulated area are related to CO 2Concentration distribution. The simulation output is set to multiple atmospheric layers in the vertical direction, preferably 25 layers, and the boundary conditions are updated every six hours. The output is output once an hour, and the output spatial resolution is 1km×1km. The physical parameterization schemes are selected: YSU (Yonsei University, South Korea) boundary layer scheme, Monin-Obukhov (Monin-Obukhov) near-surface scheme, WSM6 (WRF Single-Moment 6-class) microphysics scheme, RRTM (Rapid Radiative Transfer Model) longwave radiation scheme, Goddard shortwave radiation scheme, Noah land surface process, KFETA cumulus scheme.

[0061] Step S102: Acquire multi-source CO from space, air and ground 2 Concentration observation data.

[0062] In this embodiment, the sky-air-ground multi-source data refers to data from multiple different payload platforms such as the sky, air, and ground, where "air" refers to space, that is, satellite observation, "sky" refers to aerospace, that is, aerial observation, including observation methods such as tethered balloons, towers and / or aircraft flight observations, and "ground" refers to ground observations, also known as ground-based observations. The above data are all based on point data in space, so they are also collectively referred to as observation data. Generally speaking, observation data has higher accuracy than model simulation results, but the number of point data is limited, and there is a problem of spatial discontinuity.

[0063] In this embodiment, the space-based CO 2 Concentration observations include CO2 measurements from tethered balloons, towers, and / or aircraft flights. 2 Concentration data; air-based CO 2 Concentration observation data include multiple global reversible XCO 2 Satellite observations of CO 2 concentration data, and the observation platforms carried by each satellite are different; ground-based CO 2 Concentration observation data include CO2 obtained from background observation stations and carbon column observation stations used to obtain carbon dioxide column concentration. 2 Concentration observation data. The following describes the different data sources:

[0064] (1) Satellite observation data

[0065] Satellite observations (air) are used to obtain CO2 from the near-surface to the top of the atmosphere. 2 Profile information, there are currently multiple XCO inversions in the world 2Satellite spectral data, such as SCIAMACHY, GOSAT, OCO-2, TanSat, GOSAT-2, OCO-3, etc., can be observed as early as 2003, which can support nearly 20 years of atmospheric CO 2 Observation simulation. Among them, SCIAMACHY satellite data uses the data set obtained by processing based on the BESD algorithm (V02.01.01), GOSAT and GOSAT-2 use the atmospheric CO2 profile data processed by the NIES algorithm published on the official website (V01.20), OCO-2 and OCO-3 also use the full physical model inversion data (V11r) published on the official website, TanSat currently has no L2 data published, and this embodiment uses the atmospheric CO2 inverted based on the DEEI algorithm. 2 Data processing of profiles.

[0066] (2) Observation data from tethered balloons and aircraft flights

[0067] Airborne observations are used to obtain near-surface CO 2 Profile information. There are many types of aerial observations, depending on the spacecraft, such as tethered balloons and aircraft flight observations.

[0068] Among them, tethered balloon observation is to place sensors on tethered balloons to simultaneously observe atmospheric elements such as temperature and humidity to obtain atmospheric CO at a certain altitude. 2 Profile information, currently there are observation experiments being carried out in many places at home and abroad, but there is less shared data; aircraft flight observations are costly and are mainly used for satellite payload verification and calibration. By placing sensors outside the body and flying a designated route for target objects in a fixed area, atmospheric profile information from the surface to the aircraft's flight altitude is obtained. In recent years, some calibration tests have been conducted for my country's TanSat, which is of great significance to TanSat payload verification and calibration. Whether it is a tethered balloon or an aircraft flight observation, the profile information of the atmosphere near the surface is obtained.

[0069] Considering the great instability of the observation data from tethered balloons and aircraft flights, it is necessary to fill in the profile and evaluate the stability of the data before using it, delete the data that is evaluated as unstable, and merge the aerial observation data that is evaluated as stable with the data from other sources. This can be achieved in the following steps: In terms of atmospheric stratification, the satellite payload detects the entire atmospheric CO from the near surface to the top of the atmosphere. 2 Therefore, the prior CO 2 Profile information, corresponding to the pressure weight function, interpolation to complete the atmospheric CO 2 Profile information, so that the tethered balloon and aircraft flight observation data are consistent with the satellite observation data in the atmospheric layer, and the XCO is obtained by integration2 Simulation results, analysis of XCO 2 The standard deviation of the data results is used to evaluate the stability of the flight observation simulation results and prove the reliability of the method.

[0070] (3) Ground-based observation data

[0071] Ground-based observations (ground) are used to obtain high-precision observation data near the surface or from the surface to a certain height. Depending on the observation instrument, ground-based observations are divided into those used to monitor near-surface atmospheric CO 2 The background observation station of carbon dioxide concentration and the column concentration of carbon dioxide (XCO 2 ) data, as well as observation stations built by meteorological bureaus, universities, research institutes and other research institutions. Among them, WDCGG is a global greenhouse gas background monitoring network. The CO2 data of stations around the world can be obtained from the official website. 2 Concentration data, most of the stations are distributed in the northern hemisphere, with a long observation time span and a rich variety of surface underlying cover types. TCCON is a global carbon column observation network. The stations use Fourier Transform Spectrometer (FTS) to observe data based on unified standards. Its inversion accuracy can reach 0.25%, and the standard deviation of cloud-free measurements within 1 hour is generally 0.1%. In addition, there is the COCCON network developed based on the portable Fourier spectrometer EM27, and the Beijing FTS ground station data that has not joined the TCCON observation, which can be used for multi-platform data fusion to improve data accuracy. Observations at ground observation sites are closer to the true value and are the most accurate observation method, but the number of sites is limited.

[0072] Although data from space-based, air-based, and ground-based sources can be collectively referred to as observation data, the three are different in observation methods, observation instruments, observation time and space scales, observation altitudes, and vertical resolutions. 2 Fusion of concentration observation data is a challenging task. For example, in ground observations, WDCGG only provides near-surface CO 2 TCCON provides complete (from the surface to the top of the atmosphere) CO 2 Column concentration observation results: There are not many ground observation stations, and they need to be planned and built in advance. If a sudden anthropogenic emission event occurs in a local area, the near-surface CO 2If the concentration fluctuates, it is difficult to obtain data in time by relying on these ground observation sites. In comparison, the locations of tethered balloons and aircraft observations are more flexible and can respond to emergencies in a timely manner. However, the observation altitudes of tethered balloons and aircraft observations are extremely uncertain, and observation data of the top layer of the atmosphere cannot be obtained. The advantage of satellite observations is that the observation points cover a wide range and are numerous. However, due to the periodicity and susceptibility to weather of satellites, the time scale of their observations is affected by the satellite cycle and weather conditions, and the CO2 concentration of satellites is also affected by weather conditions. 2 Concentration suffers from insufficient time resolution and missing data.

[0073] Traditional multi-source data fusion methods mostly focus on the fusion of observation data from the same payload platform, such as the fusion between multiple satellites on the same satellite payload platform, and pay more attention to aspects such as the calibration of observation instruments and the matching of observation time and space scales, or the fusion of data from the same payload platform with model simulation data, such as the fusion of satellite observation data with model simulation data. Therefore, although a large amount of data has been accumulated by different payload platforms, there is relatively little research in the industry on how to fuse the observation data from different payload platforms with model simulation data to improve the resolution and accuracy of model simulation data. This embodiment makes an attempt to address this issue.

[0074] Step S103: Establishing surface CO based on machine learning method 2 Concentration, wind speed, pressure profile, humidity profile, temperature profile and the first atmospheric CO 2 The first regression model between the profiles.

[0075] Among them, the first atmospheric CO 2 The profile is related to the ground CO 2 CO concentration observation data (ground observation data) associated with 2 profile, and the first atmospheric CO 2 The atmospheric stratification of the profile is based on CO 2 The atmospheric stratification of the profile distribution simulation results is used as the benchmark for determination of the surface CO 2 The concentration is the CO obtained from the background observation station 2 concentration.

[0076] That is to say, this embodiment first processes the ground observation data, and considers that the ground observation data has the characteristics of temporal continuity, spatial discontinuity and non-uniform observation elevation, and establishes the surface CO 2 2 Concentration, wind speed, pressure profile, humidity profile, temperature profile and atmospheric CO 2 The regression relationship between the profiles is based on the vertical data characteristics of wind speed, pressure profile, humidity profile, temperature profile and other elements, combined with the surface CO2 Concentration (close to surface CO 2 concentration true value), in CO 2 The atmospheric stratification of the profile distribution simulation results is used as a benchmark, and the ground observation data are unified into an atmospheric CO with multiple atmospheric strata (e.g., 25 layers). 2 For ease of description, the atmospheric CO 2 The profile data are called the first atmospheric CO 2 outline.

[0077] For example, the surface CO 2 The concentration, wind speed, pressure profile, humidity profile, and temperature profile are used as features to establish a training data set, and then the first regression model is iteratively trained using the training data set until the model converges. The trained first regression model is saved and the CO 2 The atmospheric stratification benchmark of the profile distribution simulation results is used to obtain the CO 2 The predicted concentration value eventually forms an atmospheric CO with multiple atmospheric layers. 2 Profile data.

[0078] Among them, the machine learning method can be implemented using models such as random forest, ensemble learning XGBoost, etc., and this embodiment does not limit the specific model selection.

[0079] Step S104: Establishing the tethered balloon CO based on the machine learning method 2 Profile and tower observation of CO 2 Profile, aircraft flight observation CO 2 Profile, wind speed, pressure profile, humidity profile, temperature profile and the second atmospheric CO 2 The second regression model between the profiles; among them, the second atmospheric CO 2 The profile is consistent with the space-based CO 2 CO concentration observations associated with 2 profile, and the second atmospheric CO 2 The atmospheric stratification of the profile is based on CO 2 The atmospheric stratification of the profile distribution simulation results is determined as a benchmark.

[0080] After the ground observation data processing is completed, step S104 is mainly used to process the space-based observation data represented by tethered balloons and aircraft flight observations. Considering that space-based observations cannot obtain the complete CO2 from the surface to the top of the atmosphere, 2 The tethered balloon observation height is uncertain. In this embodiment, based on the machine learning method, tethered balloon CO 2 profile, airborne flight experiment data, high tower observation CO 2Profile, surface CO 2 The regression relationship between concentration, wind speed, pressure profile, humidity profile, temperature profile and atmospheric CO2 profile unifies the space-based observation data into atmospheric CO2 profiles with multiple atmospheric layers (e.g. 25 layers). 2 Profile data, referred to as the second atmospheric CO 2 outline.

[0081] Step S105: Based on the CarbonTracker model data, the air-based CO 2 The concentration observation data were subjected to sensitivity correction and time and space scale normalization.

[0082] For air-based CO 2 As for the concentration observation data, different satellite platforms have different design indicators and observation modes. 2 Inversion has different sensitivities (commonly known as XCO 2 The inversion accuracy is different), and the XCO of multiple satellites 2 The products also differ in physical quantities and temporal and spatial scales, so they need to be subjected to sensitivity correction and temporal and spatial scale normalization.

[0083] The CarbonTracker used in this example is a system for simulating and monitoring the global carbon cycle. It was developed by the Earth System Research Laboratory (ESRL) of the National Oceanic and Atmospheric Administration (NOAA) of the United States and aims to provide observations, simulations, and estimates of the global carbon cycle. The CarbonTracker model data is generated by the CarbonTracker system and is used to describe the simulation results of the global carbon cycle. These data usually include carbon dioxide (CO 2 ) concentration, flux and estimation of absorption / emission sources. The data covers the global scope and can provide different temporal and spatial resolutions, including data at global scale, regional scale and local scale, which can meet the needs of sensitivity correction and temporal and spatial scale unification of different satellite platforms.

[0084] Based on CarbonTracker model data, air-based CO 2 The sensitivity correction of the concentration observation data can specifically include the following steps: based on the profile stratification of the CarbonTracker model data, the observation data of each satellite is stratified and expanded to obtain the first CO 2 Concentration distribution data, first CO 2 The profile stratification of the concentration distribution data is consistent with the profile stratification of the CarbonTracker model data; the average kernel function and the first CO 2The concentration distribution data are used to calculate the response difference value of each satellite in the vertical direction; the observation data of each satellite and its response difference value are summed to obtain the sensitivity-corrected CO 2 Observation profile.

[0085] Specifically, the observation data of each satellite can be processed by interpolation or integration to obtain the first CO 2 The concentration distribution data, and the data after layered expansion processing, come with the average kernel function used by the inversion algorithm of each satellite. On this basis, the response difference value of each satellite in the vertical direction can be calculated, and then the observation data of each satellite and its response difference value can be summed to obtain the result of sensitivity correction.

[0086] The spatial scale normalization process may specifically include the following steps: based on a preset reference time (the time may be determined according to the time scale of other data sources, such as 13:00 every day, so that daily scale data can be generated, and of course, it may also be set to other time points or different time intervals), the observation time of each satellite is converted to the reference time; the study area is gridded according to the preset spatial resolution, and the coverage of each grid is used as the spatial scale reference. The specific time conversion method may be, for example, a time interpolation method or a model simulation method, which is not limited in this application.

[0087] Step S106: Establishing the first CO based on the machine learning method 2 Profile data, second CO 2 Profile data, pressure profile, temperature profile, humidity profile and the third atmospheric CO 2 The third regression model between the profiles to obtain a multi-source observation fusion data set; the first CO 2 The profile data is the first atmospheric CO 2 CO profiles and carbon column observation stations 2 The sum of the concentration observation data, the second CO 2 The profile data are the CO2 of tethered balloons, towers and aircrafts predicted by the second regression model. 2 The sum of the profiles; the third atmospheric CO 2 The profile is related to airborne CO 2 CO concentration observations associated with 2 outline.

[0088] The purpose of step S106 is to achieve sky-ground fusion, that is, based on the acquired ground observations (the first CO 2 profile data) and tethered balloon / aircraft flight observations (second CO 2 Atmospheric CO 2The true value of the profile is obtained by comparing it with the satellite observation data (i.e., the atmospheric CO 2 A third regression model was established based on machine learning methods (such as random forest and ensemble learning XGBoost) to extract pressure profile, temperature profile, humidity profile, and sensitivity-corrected atmospheric CO 2 The profile (satellite) is the feature, and the simulation result is the sky-ground fusion result, that is, a multi-source observation fusion data set is obtained.

[0089] Step S107: Using the multi-source observation fusion data set as the optimization control condition, CO 2 The profile distribution simulation results are the initial values. Based on the high-precision surface modeling HASM method, the CO 2 The best estimate of atmospheric CO concentration 2 3D simulation results of profiles.

[0090] It should be noted that the HASM method is a method for modeling any geographical area on the earth's surface. It is based on the principles of differential geometry and optimization control theory. It establishes a high-precision surface modeling model with global approximate data (including remote sensing data and coarse-resolution simulation data of global models) as the driving field and local high-precision data (including monitoring network data and survey sampling data) as the optimization control conditions. Based on a large number of applied research, it has refined the basic laws of earth surface modeling and can obtain high-precision spatial simulation results.

[0091] In this embodiment, the multi-source observation fusion data set is used as the optimization control condition, CO 2 The simulated profile distribution results are initial values. The CO 2 The profile distribution simulation results contain atmospheric stratification information. Therefore, based on the high-precision surface modeling HASM method, the CO 2 The optimal estimate of the concentration can specifically include the following steps: constructing a HASM optimization controller based on multi-source observation fusion data at point scale; combining the finite difference format with the study area and the preset simulation resolution to estimate the CO concentration in each atmospheric layer. 2 The partial differential equations of the concentration surface are discretized and the HASM numerical simulation solver is calculated; the CO 2 The profile distribution simulation results are used as initial values ​​to drive the HASM numerical simulation solver to perform iterative calculations to obtain the CO 2 The best estimate of the concentration.

[0092] In this example, the atmospheric model simulates CO 2 The profiles are surface data at each atmospheric layer, and the atmospheric model simulates CO 2The profile distribution is taken as the initial value, and the multi-source observation fusion data set is taken as the optimization control condition. According to the principle of differential geometry, the CO 2 The initial surface data of the spatial distribution trend of the profile is organically combined with high-precision multi-source observation information to convert the stratified atmospheric CO 2 The concentration data is based on the target research area (also called the calculation area in HASM) and the preset target simulation resolution, combined with the discretization of the partial differential equations of the surface by the finite difference format, and the HASM numerical simulation solver is calculated; the HASM optimization controller is constructed according to the point-scale measured data; the least squares method and multiple iterations are used to continuously adjust the initial surface data, and finally the CO concentration of each layer is obtained. 2 The best estimate of the concentration.

[0093] It should be noted that the target simulation resolution is one of the input parameters of HASM. In order to obtain a higher spatial resolution, the target simulation resolution can be set to be larger than the CO 2 The spatial resolution of the profile distribution simulation results is combined with multi-source observation fusion data to iteratively solve the CO 2 The optimal estimate of the concentration of local CO 2 Detailed depiction of concentration in three-dimensional space.

[0094] Furthermore, the HASM optimization controller first generates simulated atmospheric CO based on multi-source observation fusion data. 2 The profile simulation optimizes the control rules and 2 CO simulated by the model under the profile simulation optimization control rule 2 The profile is used as the initial value to drive the HASM numerical simulation solver to perform iterative calculations, integrating point data (i.e. multi-source observation fusion data, "sky-air-ground" integrated data set) and surface data (atmospheric model simulation results, i.e. CO 2 profiles), constructing a multi-source and multi-scale three-dimensional atmospheric CO 2 The concentration fusion analysis system optimizes atmospheric CO 2 Three-dimensional spatial distribution surface data, obtain high-resolution, high-precision, and spatiotemporal continuous results, HASM obtains CO 2 The simulation results of the optimal estimate of concentration can not only have the accuracy of the observed data, but also take into account the CO2 in the area without observation value. 2 Spatial distribution, realizing multi-source and multi-scale three-dimensional atmospheric CO 2 The profiles are effectively integrated to enhance the CO 2 The degree of detail in describing the spatial distribution.

[0095] In summary, this embodiment takes into account the fact that there are abundant observation data nowadays, but there are differences between them, which makes it impossible to maximize the advantages of multi-source data and the problem that multi-source data cannot be directly applied to data monitoring. Multi-source data fusion is performed for satellite, aircraft flight, and ground-based observations to design an atmospheric CO 2 The three-dimensional fusion method of profiles can improve data fusion accuracy and reduce error transmission, and can be applied to fields such as global climate change monitoring. This method can preprocess and fuse data from different platforms, and perform atmospheric CO fusion of multi-source data based on the HASM theory. 2 Three-dimensional simulation of concentration profile to obtain CO 2 High-precision spatiotemporal distribution of profile data.

[0096] Exemplary Systems

[0097] The present application provides an atmospheric CO 2 2 Profile 3D simulation system, such as Figure 2 As shown, the system includes: a first acquisition unit 201, a second acquisition unit 202, a first modeling unit 203, a second modeling unit 204, a correction unit 205, a third modeling unit 206, and a simulation analysis unit 207. Among them:

[0098] The first acquisition unit 201 is configured to acquire CO based on the atmospheric chemical transport model. 2 Profile distribution simulation results;

[0099] The second acquisition unit 202 is configured to acquire multi-source CO from space, air, and ground. 2 Concentration observation data;

[0100] Among them, the CO 2 The profile distribution simulation results include CO2 in multiple atmospheric layers in the vertical direction. 2 Profile information; space-based CO 2 Concentration observations include CO2 measurements from tethered balloons, towers, and / or aircraft flights. 2 Concentration data; air-based CO 2 Concentration observation data include multiple global reversible XCO 2 Satellite observations of CO 2 concentration data, and the observation platforms carried by each satellite are different; ground-based CO 2 Concentration observation data include CO2 obtained from background observation stations and carbon column observation stations used to obtain carbon dioxide column concentration. 2 Concentration observation data;

[0101] The first modeling unit 203 is configured to establish a surface CO 2Concentration, wind speed, pressure profile, humidity profile, temperature profile and the first atmospheric CO 2 The first regression model between the profiles, wherein the first atmospheric CO 2 The profile is related to the ground CO 2 CO concentration observations associated with 2 profile, and the first atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is used as a benchmark to determine the surface CO 2 The concentration is the CO obtained from the background observation station 2 concentration;

[0102] The second modeling unit 204 is configured to establish a tethered balloon CO based on a machine learning method. 2 Profile and tower observation of CO 2 Profile, aircraft flight observation CO 2 Profile, wind speed, pressure profile, humidity profile, temperature profile and the second atmospheric CO 2 The second regression model between the profiles; wherein the second atmospheric CO 2 The profile is consistent with the space-based CO 2 CO concentration observations associated with 2 profile, and the second atmospheric CO 2 The atmospheric stratification of the profile is based on the CO 2 The atmospheric stratification of the profile distribution simulation results is determined as a benchmark;

[0103] The calibration unit 205 is configured to calibrate the air-based CO based on the CarbonTracker model data. 2 The concentration observation data were subjected to sensitivity correction and time and space scale normalization;

[0104] The third modeling unit 206 is configured to establish the first CO based on the machine learning method 2 Profile data, second CO 2 Profile data, pressure profile, temperature profile, humidity profile and the third atmospheric CO 2 The third regression model between the profiles is used to obtain a multi-source observation fusion data set; the first CO 2 The profile data is the first atmospheric CO 2 CO profiles and carbon column observation stations 2 The sum of the concentration observations, the second CO 2 The profile data are the CO2 of tethered balloons, towers and aircraft predicted by the second regression model. 2 The sum of the profiles; the third atmospheric CO 2 The profile is related to the air-based CO 2CO concentration observations associated with 2 Profile;

[0105] The simulation analysis unit 207 is configured to use the multi-source observation fusion data set as the optimization control condition, the CO 2 The profile distribution simulation results are the initial values. Based on the high-precision surface modeling HASM method, the CO 2 The best estimate of atmospheric CO concentration 2 3D simulation results of profiles.

[0106] The atmospheric CO 2 2 The three-dimensional profile simulation system can realize the atmospheric CO 2 profile simulation based on the fusion of multi-source data of sky and ground provided in any of the above embodiments. 2 The steps and processes of the contour three-dimensional simulation method achieve the same technical effect and will not be repeated here.

[0107] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A three-dimensional simulation method of atmospheric CO2 profile based on multi-source data fusion of sky and ground, characterized in that: include: Based on the atmospheric chemical transport model, obtain the simulation results of CO2 profile distribution; Obtain multi-source CO2 concentration observation data based on space, air and ground; The CO2 profile distribution simulation results include CO2 profile information of multiple atmospheric layers in the vertical direction; the space-based CO2 concentration observation data include CO2 concentration data obtained by tethered balloons, high towers, and / or aircraft flight observations; the air-based CO2 concentration observation data include CO2 concentration data obtained by multiple satellite observations around the world that can invert XCO2, and the observation platforms carried by each satellite are different; the ground-based CO2 concentration observation data include CO2 concentration observation data obtained by background observation stations and carbon column observation stations for obtaining carbon dioxide column concentrations; A first regression model is established between the surface CO2 concentration, wind speed, pressure profile, humidity profile, temperature profile and the first atmospheric CO2 profile based on a machine learning method, wherein the first atmospheric CO2 profile is a CO2 profile associated with ground-based CO2 concentration observation data, and the atmospheric stratification of the first atmospheric CO2 profile is determined based on the atmospheric stratification of the CO2 profile distribution simulation result, and the surface CO2 concentration is the CO2 concentration obtained by the background observation station; A second regression model is established based on a machine learning method between the tethered balloon CO2 profile, the tower observation CO2 profile, the aircraft flight observation CO2 profile, the wind speed, the pressure profile, the humidity profile, the temperature profile and the second atmospheric CO2 profile; wherein the second atmospheric CO2 profile is a CO2 profile associated with the space-based CO2 concentration observation data, and the atmospheric stratification of the second atmospheric CO2 profile is determined based on the atmospheric stratification of the CO2 profile distribution simulation result; Based on the CarbonTracker model data, the airborne CO2 concentration observation data is subjected to sensitivity correction and time and space scale normalization processing; A third regression model is established between the first CO2 profile data, the second CO2 profile data, the pressure profile, the temperature profile, the humidity profile and the third atmospheric CO2 profile based on a machine learning method to obtain a multi-source observation fusion data set; the first CO2 profile data is the sum of the first atmospheric CO2 profile and the CO2 concentration observation data obtained by the carbon column observation station, and the second CO2 profile data is the sum of the tethered balloon, tower and aircraft flight CO2 profiles predicted by the second regression model; the third atmospheric CO2 profile is a CO2 profile associated with the air-based CO2 concentration observation data; Taking the multi-source observation fusion data set as the optimization control condition and the CO2 profile distribution simulation result as the initial value, based on the high-precision surface modeling HASM method, the optimal estimation value of the CO2 concentration of each atmospheric layer is solved, thereby obtaining the three-dimensional simulation result of the atmospheric CO2 profile.

2. The method for three-dimensional simulation of atmospheric CO2 profile based on multi-source data fusion of sky and ground according to claim 1 is characterized in that: Based on the CarbonTracker model data, the air-based CO2 concentration observation data is subjected to sensitivity correction, specifically: Based on the profile stratification of the CarbonTracker model data, the observation data of each satellite is subjected to layered expansion processing to obtain first CO2 concentration distribution data, wherein the profile stratification of the first CO2 concentration distribution data is consistent with the profile stratification of the CarbonTracker model data; Based on the average kernel function and the first CO2 concentration distribution data used in the inversion algorithm of each satellite generated by the hierarchical expansion process, the response difference value of each satellite in the vertical direction is calculated respectively; The observation data of each satellite and its response difference value are summed to obtain the sensitivity-corrected CO2 observation profile.

3. The method for three-dimensional simulation of atmospheric CO2 profile based on multi-source data fusion of sky and ground according to claim 2 is characterized in that: The spatiotemporal scale normalization process includes: Based on the preset reference time, the observation time of each satellite is converted to the reference time; The study area is divided into grids according to the preset spatial resolution, and the coverage of each grid is used as the spatial scale benchmark.

4. The method for three-dimensional simulation of atmospheric CO2 profile based on multi-source data fusion of sky and ground according to claim 1, characterized in that: The atmospheric chemical transport model is a GEOS-Chem model or a WRF-Chem model.

5. The method for three-dimensional simulation of atmospheric CO2 profile based on multi-source data fusion of sky and ground according to claim 1, characterized in that: Taking the multi-source observation fusion data set as the optimization control condition and the CO2 profile distribution simulation result as the initial value, the optimal estimation value of the CO2 concentration of each atmospheric layer is solved based on the high-precision surface modeling HASM method, including: Construct a HASM optimization controller based on multi-source observation fusion data at point scale; According to the research area and the preset simulation resolution, the HASM numerical simulation solver is calculated by combining the finite difference format to discretize the partial differential equations of the CO2 concentration surface of each atmospheric layer; The CO2 profile distribution simulation result is used as an initial value to drive the HASM numerical simulation solver to perform iterative calculations, and the optimal estimated value of the CO2 concentration in each atmospheric layer is obtained.

6. The method for 3D simulation of atmospheric CO2 profile based on multi-source data fusion of sky and ground according to any one of claims 1 to 5, characterized in that: The number of the atmospheric stratification layers is 25.

7. A three-dimensional simulation system of atmospheric CO2 profile based on multi-source data fusion of sky and ground, characterized in that: include: A first acquisition unit is configured to acquire a CO2 profile distribution simulation result based on an atmospheric chemical transport model; A second acquisition unit is configured to acquire multi-source CO2 concentration observation data based on space, air and ground; The CO2 profile distribution simulation results include CO2 profile information of multiple atmospheric layers in the vertical direction; the space-based CO2 concentration observation data include CO2 concentration data obtained by tethered balloons, high towers, and / or aircraft flight observations; the air-based CO2 concentration observation data include CO2 concentration data obtained by multiple satellite observations around the world that can invert XCO2, and the observation platforms carried by each satellite are different; the ground-based CO2 concentration observation data include CO2 concentration observation data obtained by background observation stations and carbon column observation stations for obtaining carbon dioxide column concentrations; a first modeling unit configured to establish a first regression model between surface CO2 concentration, wind speed, pressure profile, humidity profile, temperature profile and a first atmospheric CO2 profile based on a machine learning method, wherein the first atmospheric CO2 profile is a CO2 profile associated with ground-based CO2 concentration observation data, and the atmospheric stratification of the first atmospheric CO2 profile is determined based on the atmospheric stratification of the CO2 profile distribution simulation result, and the surface CO2 concentration is the CO2 concentration obtained by the background observation station; The second modeling unit is configured to establish a second regression model between the tethered balloon CO2 profile, the tower observation CO2 profile, the aircraft flight observation CO2 profile, the wind speed, the pressure profile, the humidity profile, the temperature profile and the second atmospheric CO2 profile based on a machine learning method; wherein the second atmospheric CO2 profile is a CO2 profile associated with the space-based CO2 concentration observation data, and the atmospheric stratification of the second atmospheric CO2 profile is determined based on the atmospheric stratification of the CO2 profile distribution simulation result; A correction unit configured to perform sensitivity correction and spatiotemporal scale normalization processing on the air-based CO2 concentration observation data based on CarbonTracker model data; The third modeling unit is configured to establish a third regression model between the first CO2 profile data, the second CO2 profile data, the pressure profile, the temperature profile, the humidity profile and the third atmospheric CO2 profile based on a machine learning method to obtain a multi-source observation fusion data set; the first CO2 profile data is the sum of the first atmospheric CO2 profile and the CO2 concentration observation data obtained by the carbon column observation station, and the second CO2 profile data is the sum of the tethered balloon, the tower and the aircraft flight CO2 profiles predicted by the second regression model; the third atmospheric CO2 profile is a CO2 profile associated with the air-based CO2 concentration observation data; The simulation analysis unit is configured to use the multi-source observation fusion data set as the optimization control condition and the CO2 profile distribution simulation result as the initial value, and solve the optimal estimation value of the CO2 concentration of each atmospheric layer based on the high-precision surface modeling HASM method, so as to obtain the three-dimensional simulation result of the atmospheric CO2 profile.

8. The system according to claim 7, characterized in that The correction unit is further configured as follows: Based on the profile stratification of the CarbonTracker model data, the observation data of each satellite is subjected to layered expansion processing to obtain first CO2 concentration distribution data, wherein the profile stratification of the first CO2 concentration distribution data is consistent with the profile stratification of the CarbonTracker model data; Based on the average kernel function and the first CO2 concentration distribution data used in the inversion algorithm of each satellite generated by the hierarchical expansion process, the response difference value of each satellite in the vertical direction is calculated respectively; The observation data of each satellite and its response difference value are summed to obtain the sensitivity-corrected CO2 observation profile.

9. The system according to claim 8, characterized in that The correction unit is further configured as follows: Based on the preset reference time, the observation time of each satellite is converted to the reference time; The study area is divided into grids according to the preset spatial resolution, and the coverage of each grid is used as the spatial scale benchmark.

10. The system according to claim 7, characterized in that The simulation analysis unit is further configured as follows: Construct a HASM optimization controller based on multi-source observation fusion data at point scale; According to the research area and the preset simulation resolution, the HASM numerical simulation solver is calculated by combining the finite difference format to discretize the partial differential equations of the CO2 concentration surface of each atmospheric layer; The CO2 profile distribution simulation result is used as an initial value to drive the HASM numerical simulation solver to perform iterative calculations, and the optimal estimated value of the CO2 concentration in each atmospheric layer is obtained.

Images