Regional high-resolution atmosphere reanalysis data construction method and system based on dynamic downscaling and application

Abstract

The invention discloses a regional high-resolution atmosphere reanalysis data construction method based on dynamic downscaling, and aims to improve the depiction capability of medium and small-scale weather processes. According to the method, global reanalysis data is used as a background field, multi-source observation data and high-resolution land utilization information are fused, multi-time and short-period simulation is executed by constructing a multi-layer nested WRF region mode and combining an assimilation strategy combining analysis Nudging and observation Nudging, and a reanalysis data set with hour-level and kilometer-level resolution is constructed step by step. Through comparison and verification of multi-source observation data, the constructed data set is obviously superior to ERA5 in the aspects of ground meteorological elements and vertical profile structure reduction, and has higher regional adaptability and simulation precision. The method is suitable for a plurality of high-resolution meteorological application scenes such as regional artificial intelligence meteorological large models, urban fine meteorological services, high-influence weather disaster early warning, regional climate change evaluation, energy meteorological resource evaluation and the like.

Description

Technical Field

[0001] This invention belongs to the field of atmospheric science and relates to a method for constructing atmospheric reanalysis data. More specifically, it relates to a method, system, and application for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling. Background Technology

[0002] Atmospheric reanalysis data are continuous four-dimensional meteorological fields reconstructed from multi-source historical observation data through numerical weather prediction models and data assimilation systems. They are of great value in climate monitoring, environmental simulation, and extreme event replay. While current mainstream global reanalysis products, such as ERA5, MERRA-2, and JRA-55, have achieved high levels of variable richness and spatiotemporal coverage, their spatial resolution is generally between 25 and 125 kilometers, which is insufficient to meet the needs for detailed characterization of small-scale atmospheric processes in complex terrain and densely urbanized areas.

[0003] To overcome the limitations of global reanalysis products in terms of regional accuracy, dynamic downscaling techniques have been widely adopted. This method constructs regional numerical weather models, using global reanalysis data as initial and boundary conditions, and combines higher-resolution geographic information and physical parameter settings to remodel the atmospheric state of a specific region, thereby obtaining more detailed meteorological information. However, traditional dynamic downscaling faces the problem of continuously accumulating boundary condition errors during long-term simulations. Without effective dynamic constraints from observational data, the simulation results often gradually deviate from the true atmospheric state, affecting the physical consistency and reliability of the regional simulation results.

[0004] Furthermore, land use type, as a key surface input in regional models, directly affects important physical processes such as boundary layer thermal structure, underlying surface flux exchange, local circulation evolution, and even precipitation generation. Most existing regional simulation studies still use general-purpose land use data such as MODIS, which suffer from outdated data and struggle to reflect actual surface changes such as rapid urbanization and agricultural pattern evolution. In regions with strong surface heterogeneity, this bias has a more pronounced impact on simulation results. With the continuous development and release of high-resolution remote sensing data, effectively reclassifying it and mapping it to the land use type system supported by regional models has become an important direction for improving the accuracy of model surface inputs.

[0005] In summary, current global reanalysis data have significant limitations in regional accuracy. While traditional dynamic downscaling has potential for improvement, it still requires refinement in areas such as model boundary error control, surface input updates, and observational constraint mechanisms. Therefore, there is an urgent need to establish a method for constructing regional high-resolution reanalysis data that integrates high-precision underlying surface information with multi-source observational constraints and possesses stable dynamic simulation capabilities, in order to meet the demands of modern meteorological services for refined acquisition of small- and medium-scale atmospheric information. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing global reanalysis products in terms of regional accuracy, as well as the problems of boundary error accumulation and lag in surface input updates in traditional dynamic downscaling methods. It proposes a method for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling. By fusing multi-source observational data, introducing updated high-resolution land use information, and employing an assimilation strategy combining analytical and observational nudging, a stable, detailed, and physically consistent regional atmospheric reanalysis data production workflow is established.

[0007] The specific technical solution to achieve the purpose of the invention is:

[0008] A method for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling is described below:

[0009] Step 1: Obtain the fifth-generation global reanalysis data (ERA5) released by the European Centre for Medium-Range Weather Forecasts (ECMWF) and the global high-resolution land cover data product developed under the leadership of the European Space Agency; download global communication system data from the research and development data center website of the National Center for Atmospheric Research (NCAR) in the United States, and obtain observation data from automatic weather stations in the target area.

[0010] Step 2: Reclassify the global high-resolution land cover data products, and perform time alignment and format unification on the global communication system data and automatic weather station observation data, and fuse them to generate a multi-source observation data set;

[0011] Step 3: Construct a WRF model with a multi-layered nested grid structure. Set the horizontal resolution of the inner grid to the spatial resolution of the target reanalysis dataset to ensure that the simulation area covers the target area. Start geogrid.exe in the WPS module of the WRF model to generate a geographic topographic grid field of the simulation area by combining the reclassified land use type data. Start ungrib.exe and metgrid.exe in the WPS module to convert the hourly ERA5 data into the grid feature field required by the WRF model.

[0012] Step 4: Start the OBSGRID module in WRF mode to perform quality control and standardization processing on the multi-source observation dataset, and fuse the automatic weather station observation data with the model grid feature field to output a grid feature field containing observation information.

[0013] Step 5: Enable the real module in WRF mode to generate the initial field and boundary conditions required for simulation; enable wrf.exe to perform short-term regional simulations at multiple reporting times to reduce the accumulation of model errors; during the simulation, adopt an assimilation strategy that combines analytical nudging and observational nudging to maintain consistency between the simulated field and the large-scale background field while enhancing the capture and observational constraints of small-scale regional features.

[0014] Step Six: Repeat steps Three to Five to complete the continuous simulation of the target time period; discard the initial simulation results of each simulation period, specifically discarding the simulation results of the first hour of each simulation period to reduce the impact of physical imbalances in the initial stage on the simulation accuracy; transform the target variable and adjust the vertical structure of the simulation results to form an atmospheric reanalysis dataset that conforms to a unified standard; the constructed atmospheric reanalysis dataset has hourly temporal resolution, kilometer-level spatial resolution, and contains three-dimensional grid data of multiple meteorological elements, covering the specified historical time period of the target area.

[0015] In one specific implementation, step six may be followed by step seven: conducting statistical tests on meteorological elements on the constructed atmospheric reanalysis dataset, including: long-term statistical performance evaluation and stability test of surface meteorological elements and layer-by-layer error analysis of atmospheric vertical profiles.

[0016] Furthermore, in step two, the global high-resolution land cover products are reclassified. Specifically, the original resolution remote sensing land cover image is converted using a spatial resampling method, the target dataset is reclassified and mapped to the MODIS classification system, and the mapped dataset is spatially cropped to obtain a suitable high-resolution land use type dataset.

[0017] Furthermore, step five employs an assimilation strategy combining analytical nudging and observational nudging. Observational nudging uses the standardized multi-source fused observational data from step four as input, imposing observational constraints on the simulation process. Its key assimilation parameters are set based on sensitivity test results, including the observational assimilation time window, the update frequency of the observational data, and its spatial influence range, to improve the simulation results' responsiveness to real observations. Analytical nudging is continuously enabled throughout the simulation period to maintain consistency between the simulated background field and the large-scale reanalysis background. The remaining control parameters for each assimilation module are executed according to the recommended configuration of the WRF mode.

[0018] Further, the surface meteorological elements mentioned in step seven include: air temperature at 2 meters altitude, relative humidity at 2 meters altitude, wind speed component at 10 meters altitude, and surface air pressure. These surface meteorological variables are evaluated using a combination of mean absolute error, root mean square error, and Nash efficiency coefficient. The atmospheric vertical profile includes temperature, relative humidity, and wind speed. This atmospheric vertical profile is evaluated using a combination of mean deviation and root mean square error. The evaluation indicators are expressed as follows:

[0019] average deviation

[0020] Mean Absolute Error

[0021] Root mean square error

[0022] Nash efficiency coefficient

[0023] Among them, A i For analysis values, O i For the corresponding observations, is the observed average, and n is the total number of samples.

[0024] The present invention also provides a system for constructing regional high-resolution atmospheric reanalysis data, the system comprising: a data acquisition module, a data preprocessing module, a downscaling modeling module, an observation processing module, a numerical simulation module, a product construction module, and an accuracy evaluation module;

[0025] The data acquisition module is used to acquire data including global reanalysis data, land cover data products, global communication system data, and meteorological observation data of the target area.

[0026] The data preprocessing module is used to reclassify land cover products and perform time alignment, format unification, and data fusion of global communication system data and meteorological observation data of the target area;

[0027] The downscaling modeling module is used to establish a pattern grid with a multi-layered nested structure, and generate a grid feature field by combining reclassified land use type data and global reanalysis data.

[0028] The observation processing module is used to perform quality control and standardized format conversion on the fused multi-source observation dataset, and to fuse the meteorological observation data of the target area with the gridded element field;

[0029] The numerical simulation module is used to generate the initial field and boundary conditions of the model, execute short-term simulation tasks at multiple reporting times, and combine observation nudging and analysis nudging strategies during the simulation process to output hourly simulation results.

[0030] The product construction module is used to stitch together the output data from each simulation period, remove the initial unstable phase, perform variable transformation and vertical structure reconstruction, and form a unified and standardized atmospheric reanalysis data dataset.

[0031] The accuracy assessment module is used to perform statistical error analysis on the reanalysis dataset, including time series testing of surface meteorological elements and layer-by-layer error assessment of atmospheric vertical structure.

[0032] This invention also provides the application of the above-mentioned method or system in constructing high-resolution atmospheric reanalysis data, fine simulation and forecasting of small and medium-scale weather systems, regional high-resolution artificial intelligence meteorological models, urban fine meteorological services, early warning of high-impact weather disasters, regional climate change assessment, and energy and meteorological resource assessment.

[0033] The beneficial effects of this invention include: It proposes a method for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling, effectively solving the problems of insufficient spatial resolution, error accumulation in traditional dynamic downscaling, and lag in surface input information in current mainstream global atmospheric reanalysis data. Compared to existing coarse-resolution global reanalysis products, this invention employs a multi-layer nested WRF model, with the innermost layer achieving kilometer-level resolution, significantly improving the simulation capability for small-scale meteorological processes in urban agglomerations, complex terrains, and heterogeneous surface regions. By introducing high-resolution remote sensing land cover data, it overcomes the problem of lag in traditional MODIS data updates, effectively enhancing the regional model's responsiveness to actual underlying surface changes. Furthermore, this invention integrates two assimilation strategies in the regional simulation process: analytical nudging and observational nudging. The former dynamically introduces multi-source observational data at high temporal frequency to continuously constrain the simulation process, while the latter maintains the consistency between the model field and the large-scale circulation structure of ERA5. The reanalysis system is constructed using a multi-time-short-period simulation method, avoiding the problem of continuous accumulation of boundary errors in traditional long-term simulations. Comparative evaluation with actual measurements showed that the constructed reanalysis dataset outperformed ERA5 in both surface meteorological element fitting and vertical profile structure reconstruction. Specifically, the MAE of U10m and V10m decreased by 21.6% and 26.1%, respectively, while the NSE increased by 33.3% and 40.1%; the MAE of T2m and RH2m decreased by 35.8% and 36.3%, respectively, while the NSE increased by 56.7% and 57.5%; the MAE of PSFC decreased from 6.541 hPa to 1.429 hPa, a reduction of 78.2%, while the NSE improved from negative to 0.953. In terms of vertical structure simulation, the constructed dataset performed particularly well below the boundary layer. The RMSE of temperature at 1000 hPa decreased from 1.98 K to 0.93 K, an improvement of 53%; the relative humidity deviation decreased by nearly 50%, and the RMSE decreased by approximately 45%; the RMSE of wind speed at 975 hPa decreased by approximately 42%. The above results fully demonstrate the comprehensive advantages of this invention in terms of simulation accuracy and regional adaptability. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of a method for constructing regional high-resolution atmospheric reanalysis data in an embodiment of the present invention.

[0036] Figure 2This is a schematic diagram of the WRF pattern nested mesh structure constructed in an embodiment of the present invention.

[0037] Figure 3 This is a monthly-scale time series plot of the reanalysis data described in this invention and ERA5 on the U10m and V10m variables, showing MAE and NSE.

[0038] Figure 4 This is a monthly-scale time series plot of the reanalysis data described in this invention and ERA5 on the T2m, RH2m, and PSFC variables, showing MAE and NSE.

[0039] Figure 5 This is a comparison chart of the reanalysis data described in this invention and the statistical indicators of ERA5 based on automatic weather station observations.

[0040] Figure 6 This is a comparison chart of the vertical profile errors of temperature, relative humidity, and wind speed between the reanalysis data described in this invention and ERA5 data based on radiosonde observations. Detailed Implementation

[0041] The invention will be further described in detail below with reference to the specific embodiments and accompanying drawings. Except for the contents specifically mentioned below, the processes, conditions, and experimental methods for implementing the invention are all common knowledge and general knowledge in the art, and the invention does not have any particular limitations.

[0042] The Chinese meanings of some of the abbreviations involved in this invention are explained below:

[0043] ERA5: Refers to the fifth generation of global reanalysis data products released by the European Centre for Medium-Range Weather Forecasts (ECMWF).

[0044] MERRA-2: or "Modern-Era Retrospective analysis for Research and Applications, Version 2", was released by NASA.

[0045] JRA-55: This refers to the 55th year of global reanalysis data released by the Japan Meteorological Agency.

[0046] MODIS stands for "Moderate Resolution Imaging Spectroradiometer".

[0047] WRF stands for "Weather Research and Forecasting Model".

[0048] WPS: or "WRF Preprocessing System", is used to generate the geographic information required for simulation and perform meteorological field interpolation preprocessing. It is an essential process before the model starts in WRF applications.

[0049] OBSGRID: The observation data gridding module in the WRF model is responsible for spatial interpolation and quality control of discrete observation point data, and integrates it into the model background field. It supports data preprocessing using the observation nudging method.

[0050] OBS_DOMAIN: refers to the standard input file format required for observation nudging in WRF mode;

[0051] Bias: Average bias, a statistical indicator that measures the systematic deviation between simulation results and observed values. A value close to zero indicates that the simulation results have no significant systematic bias overall, a positive value indicates that the simulation results are generally too high, and a negative value indicates that the simulation results are generally too low.

[0052] MAE: Mean Absolute Error, is a commonly used indicator to measure the degree of deviation between simulation results and observed values. The smaller the value, the higher the accuracy.

[0053] RMSE: Root Mean Square Error is another statistical measure of error, reflecting the overall dispersion of the error. It is often used to evaluate the stability and accuracy of model simulations.

[0054] NSE: Nash-Sutcliffe Efficiency, is an important indicator for evaluating the simulation ability of a model. The closer its value is to 1, the stronger the model's ability to fit the observed trend of change.

[0055] U10m / V10m: These represent the east-west wind speed component and the north-south wind speed component at a height of 10 meters, respectively.

[0056] T2m: Temperature at a height of 2 meters;

[0057] RH2m: Relative humidity at a height of 2 meters;

[0058] PSFC: Ground Pressure;

[0059] YRD1km: Atmospheric reanalysis dataset generated in this embodiment of the invention, used to refer to the regional high-resolution reanalysis dataset constructed by this method;

[0060] This invention provides a method for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling, aiming to improve the ability to characterize small- and medium-scale weather processes. This method uses global reanalysis data as the background field, integrates multi-source observational data and high-resolution land use information, constructs a multi-layered nested WRF regional model, and combines an assimilation strategy that integrates analytical nudging and observational nudging to perform multi-temporal short-duration simulations, gradually constructing hourly and kilometer-level resolution reanalysis datasets. Verification through multi-source observational data shows that the constructed dataset significantly outperforms ERA5 in terms of surface meteorological elements and vertical profile structure reconstruction, exhibiting stronger regional adaptability and simulation accuracy. This method is applicable to various high-resolution meteorological applications, including regional AI-powered meteorological models, refined urban meteorological services, high-impact weather disaster early warning, regional climate change assessment, and energy meteorological resource assessment.

[0061] The specific steps of this method are as follows:

[0062] Step 1: Obtain global reanalysis data, land cover data products, global communication system data, and meteorological observation data for the target area;

[0063] Step 2: Reclassify the land cover data products and integrate the preprocessed global communication system data with the meteorological observation data of the target area to generate a multi-source observation dataset;

[0064] Step 3: Set up a multi-layer nested grid WRF mode, combine it with the reclassified land use type data to generate a simulated regional geographic topography grid field, and transform global reanalysis data into a grid feature field;

[0065] Step 4: Perform quality control and format standardization on the multi-source observation dataset, and fuse the preprocessed meteorological observation data of the target area with the gridded feature field to output a gridded feature field containing observation information.

[0066] Step 5: Generate the initial field and boundary conditions for simulation, perform regional simulation, and output the simulation results hourly.

[0067] Step 6: Repeat steps 3 to 5 to perform continuous simulations of the target area over a specified period, post-process the simulation results, and construct an atmospheric reanalysis dataset.

[0068] Example

[0069] like Figure 1 As shown in this embodiment, a method for constructing high-resolution atmospheric reanalysis data based on dynamic downscaling is provided, including the following steps:

[0070] Step 1: Obtain the fifth-generation global reanalysis data (ERA5) released by the European Centre for Medium-Range Weather Forecasts (ECMWF) and the global high-resolution land cover data product developed under the leadership of the European Space Agency (ESA); download global communication system data from the research and development data center website of the National Center for Atmospheric Research (NCAR) in the United States, and obtain automatic weather station observation data in the target area.

[0071] Step two involves converting the original resolution remote sensing land cover imagery of the global high-resolution land cover data using spatial resampling methods, reclassifying and mapping the target dataset to the MODIS classification system, and spatially cropping the mapped dataset to obtain a suitable high-resolution land use type dataset. Furthermore, the global communication system data and automatic weather station observation data are time-aligned and format-unified, and fused to generate a multi-source observation dataset; in one specific implementation, the relevant data are organized into hourly little_r format.

[0072] Step 3: Design a three-layer nested grid WRF model. The outer grid has a horizontal resolution of 9km, the middle grid has a horizontal resolution of 3km, and the inner grid has a horizontal resolution of 1km, covering the target area. Start geogrid.exe in the WPS module for the WRF model to generate a geographic topographic grid field for the simulated area using the updated land use type data. Start ungrib.exe and metgrid.exe in the WPS module to convert the hourly ERA5 data into the grid feature field required by the WRF model.

[0073] Figure 2 This is a schematic diagram of the nested mesh structure of the WRF pattern constructed in this invention, with the inner mesh covering the target area.

[0074] Step 4: Enable the OBSGRID module in WRF mode to perform quality control and standardization processing on the multi-source observation dataset, output multi-source observation data files in OBS_DOMAIN format, and fuse the automatic weather station observation data with the model grid feature field to output a grid feature field containing observation information.

[0075] Step 5: Using the real module in WRF mode, generate the initial field and boundary conditions required for the simulation based on the updated grid feature field; enable wrf.exe, and conduct regional simulations lasting 6 hours each time, starting at 00:00, 06:00, 12:00, and 18:00 UTC daily, to reduce model error accumulation. During the simulation, an assimilation strategy combining analytical nudging and observational nudging is adopted. Observational nudging uses multi-source fused observation data in OBS_DOMAIN format as input to constrain the simulation process. Key parameter settings obtained through sensitivity experiments include: an assimilation time window of 3 hours, an update frequency of every 6 minutes, and a horizontal influence radius of 30 kilometers; analytical nudging is continuously enabled throughout the simulation period, with an update frequency of every 1 hour, and the remaining parameters use the default settings in WRF mode, outputting hourly simulation results.

[0076] Step Six: Repeat steps Three through Five to complete the hourly regional simulation for the specified historical target time period. Discard the results of the first hour in each 6-hour simulation segment to reduce the impact of physical imbalances during the initialization phase on simulation accuracy. Perform target variable transformation and vertical structure processing on the simulation results to construct an atmospheric reanalysis dataset with a time resolution of 1 hour, a spatial resolution of 1 km, and 32 vertical layers. The vertical layers cover the range from 10 hPa to 1000 hPa, and the layer settings are consistent with ERA5.

[0077] Step 7: Conduct statistical tests on the constructed atmospheric reanalysis dataset for meteorological elements, including: long-term statistical performance evaluation and stability testing of surface meteorological elements, and layer-by-layer error analysis of the atmospheric vertical profile. Surface meteorological elements include: air temperature at 2 meters altitude, relative humidity at 2 meters altitude, wind speed components at 10 meters altitude (including east-west and north-south wind speed components), and surface air pressure, evaluated using mean absolute error, root mean square error, and Nash efficiency coefficient. The atmospheric vertical profile includes temperature, relative humidity, and wind speed, evaluated using mean deviation and root mean square error. The evaluation indicators are as follows:

[0078] average deviation

[0079] Mean Absolute Error

[0080] Root mean square error

[0081] Nash efficiency coefficient

[0082] Among them, A i For analysis values, O i For the corresponding observations, is the observed average, and n is the total number of samples.

[0083] Figure 3 For the monthly time series plots of MAE and NSE of the reanalysis data described in this invention and ERA5 on the U10m and V10m variables, YRD1km shows better simulation accuracy in both metrics, especially with a significant reduction in error during the nighttime period. Figure 5 Statistical comparison shows that, compared to ERA5, YRD1km reduced MAE by 21.6% in U10m and V10m by 26.1%, while NSE increased by 33.3% and 40.1%, respectively.

[0084] Figure 4 For the monthly time series plots of MAE and NSE on T2m, RH2m, and PSFC variables described in this invention using reanalysis data and ERA5, the statistical index curve of YRD1km is smoother and more stable, resulting in a 35.8% decrease in MAE and a 56.7% increase in NSE on T2m; and a 36.3% decrease in MAE and a 57.5% increase in NSE on RH2m. Regarding PSFC, ERA5 fails to accurately reflect pressure gradient changes caused by complex terrain, resulting in a high MAE of 6.541 hPa and a negative NSE; while YRD1km reduces MAE to 1.429 hPa, a decrease of 78.2%, and increases NSE to 0.953 (see...). Figure 5 ).

[0085] Figure 6 This is a comparison chart of the vertical profile errors of temperature, relative humidity, and wind speed based on radiosonde observation data described in this invention and ERA5. The overall vertical distribution accuracy of YRD1km is better than ERA5, especially in the boundary layer. At the 1000 hPa altitude, the RMSE of the temperature profile decreased from 1.98 K in ERA5 to 0.93 K in YRD1km, an improvement of 53%; the deviation in relative humidity decreased by approximately 50%, and the RMSE decreased by approximately 45%. Regarding wind speed error, YRD1km reduced the RMSE by approximately 42% in the near-surface layer (975 hPa).

[0086] The scope of protection of this invention is not limited to the above embodiments. Any variations and advantages that can be conceived by those skilled in the art without departing from the spirit and scope of the inventive concept are included in this invention and are protected by the appended claims.

Claims

1. A method for constructing regional high-resolution atmospheric reanalysis data based on dynamic downscaling, characterized in that, include: Step 1: Obtain global reanalysis data, land cover data products, global communication system data, and meteorological observation data for the target area; Step 2: Reclassify the land cover data products and integrate the preprocessed global communication system data with the meteorological observation data of the target area to generate a multi-source observation dataset; Step 3: Set up a multi-layer nested grid WRF mode, combine it with the reclassified land use type data to generate a simulated regional geographic topography grid field, and transform global reanalysis data into a grid feature field; Step 4: Perform quality control and format standardization on the multi-source observation dataset, and fuse the preprocessed meteorological observation data of the target area with the gridded feature field to output a gridded feature field containing observation information. Step 5: Generate the initial field and boundary conditions for simulation, perform regional simulation, and output the simulation results hourly. Step 6: Repeat steps 3 to 5 to perform continuous simulations of the target area over a specified period, post-process the simulation results, and construct an atmospheric reanalysis dataset.

2. The method as described in claim 1, characterized in that, Step six also includes conducting statistical tests on meteorological elements of the constructed atmospheric reanalysis dataset, including: time series statistical performance evaluation and stability test of surface meteorological elements, and layer-by-layer error analysis of atmospheric vertical profiles.

3. The method as described in claim 1, characterized in that, In step one, the global reanalysis data comes from ERA5, the fifth-generation global reanalysis data released by the European Centre for Medium-Range Weather Forecasts; the land cover data products come from high-resolution global land cover data products developed under the leadership of the European Space Agency; the global communication system data is downloaded from the research and development data center website of the National Center for Atmospheric Research in the United States; and the meteorological observation data of the target area comes from automatic weather stations in the region where the target area is located.

4. The method as described in claim 1, characterized in that, In step two, reclassification is performed according to the MODIS standard: the original resolution remote sensing land cover image is transformed by spatial resampling method, the target dataset is reclassified and mapped, and converted to the MODIS classification system. The mapped dataset is then spatially cropped according to the WRF model simulation area to obtain a suitable high-resolution land use type dataset. And / or, The data from the global communication system and the meteorological observation data of the target area are time-aligned, converted and organized into a predetermined format, and then fused.

5. The method as described in claim 1, characterized in that, In step three, a multi-layer nested mesh design is adopted, and the horizontal resolution of the inner mesh is set to the spatial resolution of the target reanalysis dataset to ensure that the simulation area covers the target area; And / or, The WPS module in the WRF model is used to generate a geographic topographic grid field of the simulated area by combining the reclassified land use type data. And / or, The WPS module in WRF mode converts ERA5 hourly data into the grid feature field required by WRF mode.

6. The method as described in claim 1, characterized in that, In step five, the regional simulation is conducted using multiple fixed reporting times and within a preset time period. And / or, An assimilation strategy combining analytical nudging and observational nudging is adopted. The observational nudging takes multi-source fused observational data in a standardized format as input, and its key assimilation parameters are set according to the results of sensitivity experiments, including the observational assimilation time window, the update frequency of the observational data, and its spatial influence range. The analytical nudging is continuously enabled throughout the simulation period, and the remaining control parameters of each assimilation module are executed according to the recommended configuration of WRF mode.

7. The method as described in claim 1, characterized in that, In step six, the simulation results for the first hour of each simulation period are discarded, and the target variable is transformed and the vertical structure is processed. And / or, The atmospheric reanalysis dataset has hourly temporal resolution and kilometer-level spatial resolution, and contains three-dimensional grid data of multiple meteorological elements.

8. The method as described in claim 2, characterized in that, The surface meteorological elements include: air temperature at 2 meters altitude, relative humidity at 2 meters altitude, wind speed component at 10 meters altitude, and surface air pressure. These surface meteorological variables are evaluated using mean absolute error, root mean square error, and Nash efficiency coefficient. The atmospheric vertical profile includes temperature, relative humidity, and wind speed. This atmospheric vertical profile is evaluated using mean deviation and root mean square error. The evaluation indicators are expressed as follows: average deviation Mean Absolute Error Root mean square error Nash efficiency coefficient Among them, A i For analysis values, O i For the corresponding observations, is the observed average, and n is the total number of samples.

9. A system for constructing regional high-resolution atmospheric reanalysis data, characterized in that, The system is used to implement the method as described in any one of claims 1-8, comprising: a data acquisition module, a data preprocessing module, a downscaling modeling module, an observation processing module, a numerical simulation module, a product construction module, and an accuracy evaluation module; The data acquisition module is used to acquire data including global reanalysis data, land cover data products, global communication system data, and meteorological observation data of the target area. The data preprocessing module is used to reclassify land cover products and perform time alignment, format unification, and data fusion of global communication system data and meteorological observation data of the target area; The downscaling modeling module is used to establish a pattern grid with a multi-layered nested structure, and generate a grid feature field by combining reclassified land use type data and global reanalysis data. The observation processing module is used to perform quality control and standardized format conversion on the fused multi-source observation dataset, and to fuse the meteorological observation data of the target area with the gridded element field; The numerical simulation module is used to generate the initial field and boundary conditions of the model, execute simulation tasks at multiple reporting times, and combine observation nudging and analysis nudging strategies during the simulation process to output hourly simulation results. The product construction module is used to stitch together the output data from each simulation period, remove the initial unstable phase, perform variable transformation and vertical structure reconstruction, and form a unified and standardized atmospheric reanalysis data dataset. The accuracy assessment module is used to perform statistical error analysis on the reanalysis dataset, including time series testing of surface meteorological elements and layer-by-layer error assessment of atmospheric vertical structure.

10. The method described in any one of claims 1-8, or the system described in claim 9, is used in the construction of high-resolution atmospheric reanalysis data, fine simulation and forecasting of small and medium-scale weather systems, regional high-resolution artificial intelligence meteorological models, urban fine meteorological services, early warning of high-impact weather disasters, regional climate change assessment, and energy and meteorological resource assessment.

Images