Method for improving urban green space biogenic emissions based on high-resolution global model

Abstract

The application provides a method for improving urban green space biological source emission based on high-resolution global model, aiming at the problem of lack of urban green space emission in current numerical model, and proposes a technical scheme for optimizing biological source emission list by using high-resolution land cover data. The technologies include a weakly supervised global land cover mapping method, which can fuse the spectral information of the image itself, use the features extracted by a visual large model to draw a high-resolution land cover image, a vegetation functional type matching method for converting high-resolution land cover data into vegetation functional type, and a high-resolution numerical model input file preparation, which converts the above files into an input format suitable for high-resolution earth system model, and carries out numerical simulation to estimate global biological source emission, especially urban green space emission.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bio-source emission inventory making, in particular, and particularly relates to a method for improving urban green bio-source emission based on high-resolution global model. BACKGROUND

[0002] Atmospheric pollution is a major factor affecting human health and the ecological system, and typical atmospheric pollutants include ozone and fine particles. As an important precursor, one of the important sources of volatile organic compounds is bio-source, which accounts for 90% of the total global volatile organic compound emissions. Biogenic volatile organic compounds (BVOC) mainly include isoprene, terpene, etc., and have strong photochemical activity, which can promote the generation of ozone and secondary organic aerosols, and induce atmospheric pollution.

[0003] Forests are the main source of BVOC, which has been widely recognized in past studies. However, the concern for BVOC emissions from vegetation in cities is relatively small, while these BVOC emissions in cities are crucial to atmospheric pollution. Taking ozone as an example, the chemical reaction of its precursor requires the simultaneous presence of volatile organic compounds and nitrogen oxides. The BVOC emissions in primary forests are large, but often lack nitrogen oxide sources (such as factories, motor vehicles); but the nitrogen oxide emissions in cities are relatively large, and the BVOC emissions from vegetation in cities can react very effectively with nitrogen oxides to generate ozone. One of the important reasons for ignoring BVOC emissions from urban vegetation (or urban green) in the past is that the spatial resolution of land cover type data is low, making it difficult to identify vegetation in cities. Making land cover data relies on remote sensing data, such as the commonly used medium-resolution imaging spectrometer (MODIS) satellite data with a spatial resolution of 500 meters. This resolution can identify large areas of forest and other vegetation, but has low recognition ability for urban vegetation. Urban vegetation is often scattered in different areas, such as streets and residential areas. Therefore, bio-source emissions based on MODIS satellite land cover data cannot include the emissions of urban green.

[0004] To solve this problem, improving the spatial resolution of land cover is one of the important prerequisites. With the advancement of remote sensing technology, the Landsat series (30 meters) and the Sentinel series (10 meters) satellites provide higher spatial resolution. Currently, from the perspective of spatial resolution, 10-meter high-resolution data products include the European Space Agency's 10-meter land cover data (such as ESA WorldCover10m) and Tsinghua University's 10-meter land cover data (FROM-GLC10). Although these data have a spatial resolution of 10 meters, trees are classified as a single type (Trees). However, different vegetation types have significant differences in emission factors. On the basis of improving resolution, it is necessary to identify different types of vegetation. Recently, a study based on the European Space Agency's Sentinel-2 satellite data produced global 30-meter land cover data (GLC_FCS30-2020; global 30-m landcover classification with a fine classification system in 2020), including 30 land cover types (dividing trees into several categories, such as broadleaf forest and coniferous forest), but the identification of vegetation within cities is still limited (for example, in Beijing, Figure 2 c). Based on this, it is necessary to use high spatial resolution (10 meters) Sentinel-2 data, combined with its own spectral feature extraction, and visual models that can identify long-range dependencies in images (such as Vision Transformer), to produce 10-meter resolution, 30-class fine land cover type data, which will strongly support high-precision urban green space biogenic emission estimation and subsequent atmospheric pollution numerical simulation.

[0005] In addition to land cover, the application of high-resolution numerical models is a necessary means to improve the accuracy of biogenic emissions and their impact on atmospheric pollution assessment. On the one hand, biogenic emissions are closely related to extreme weather such as high-temperature heatwaves. High-resolution numerical models are expected to improve the simulation of extreme weather such as high-temperature heatwaves, better capture the process of heatwaves promoting biogenic emissions, and more reasonably simulate the impact of biogenic emissions on atmospheric pollutant concentrations. In our previous work, we have implemented high-resolution Earth System (25-kilometer atmosphere) simulation based on domestic many-core heterogeneous supercomputing, and achieved improvement in ozone simulation (Patent No.: ZL 2023 1 1654525.2). This patent will be based on the previously optimized high-resolution Earth System model, improving the resolution of land cover data, and high-precision land cover data can be efficiently applied in high-resolution models to improve the accuracy of biogenic emission simulation, especially in urban areas. SUMMARY

[0006] In order to make up for the deficiency of the prior art, the present application provides a method for improving urban green space biological source emissions based on high-resolution global model, which is a technical solution for optimizing biological source emission inventory by using high-resolution land cover data in view of the problem of lack of urban green space emissions in the current numerical model. These technologies include a weakly supervised global land cover mapping method, which can fuse the spectral information of the image itself and use the features extracted by the visual large model to draw a high-resolution land cover image; a vegetation function type (PFT) matching method for converting high-resolution land cover data into vegetation function type; and high-resolution numerical model input file preparation, which converts the above files into an input format suitable for high-resolution earth system models and carries out numerical simulation to estimate global biological source emissions, especially urban green space emissions. The present application can effectively capture urban green space emissions and improve the accuracy of urban green space emission inventory and atmospheric pollution simulation.

[0007] The present application is realized by the following technical solutions: a method for improving urban green space biological source emissions based on high-resolution global model, specifically comprising the following steps:

[0008] Step S1, data preprocessing:

[0009] Six multispectral images of Sentinel-2 are selected, including three visible light and three infrared bands. After selecting the data, the infrared band data is interpolated to 10-meter resolution using the Google Earth Engine platform, and the visible light band is kept at the same resolution, so as to obtain 10-meter resolution Sentinel image data;

[0010] Step S2, pixel feature extraction:

[0011] For each 6-band pixel of Sentinel-2, its own and surrounding neighborhood multispectral features are extracted, and then a pre-trained Vision Transformer based on previous remote sensing data is used to extract high-level features of the same image. Specifically, the following steps are included:

[0012] Step S2-1, when extracting pixel features, a sliding window method is used to focus on the spectral and texture details of the image; for each pixel, its own and surrounding cross region are extracted, and 6-band spectral information of a total of 9 pixels is extracted; then the obtained spectral information is connected to form a long sequence with a length of 54 of 9 pixels x 6 bands;

[0013] Step S2-2, feature extraction using a pre-trained deep learning model, using the visual self-attention mechanism model encoder provided by the Vision Transformer to extract high-level semantic features in the remote sensing image; the encoder captures a wide range of image context and long-range dependencies in the image; each image is cropped into multiple slices with overlapping parts, and then inference calculation of the visual self-attention mechanism model encoder is performed on each slice to obtain feature maps with the same spatial size; finally, according to the above cropping mode, the feature maps of these slices are spliced back to the original image size;

[0014] Step S3, feature fusion:

[0015] A feature fusion method is used to balance the local, low-level features from the pixel feature extraction branch and the global, high-level features extracted from the deep learning model to obtain the optimal clustering result, which specifically includes the following steps: merging the spectral data from the pixel feature extraction branch and the data from the deep learning model branch to form a feature vector with a length of 64, and performing a standardization operation on the feature matrix to construct an input feature that can be directly used for clustering algorithms;

[0016] Step S4, clustering calculation and 10-meter land cover data image production:

[0017] After feature extraction, clustering analysis is performed based on the clustering algorithm to generate 300-class clustering results, analyze the relationship between the 30-meter GLC_FCS30-2020 resolution product and the clustering results, and find the corresponding land cover type for each cluster; further, based on the data identified as green land in the 10-meter resolution ESAWorldCover data, the previously generated 10-meter resolution land cover classification results are rechecked, and if the classification results of some areas in the generated 10-meter data are not marked as green land, the spectral features of the area and its surroundings are rejudged and re-assigned to the category belonging to green land; automatic clustering merging is completed to produce 10-meter high-resolution land cover data (WS-GLM1030-PSET; weakly-supervised 10-m 30-class global land cover mapping method with pixel spectral extraction for tree identification), stored as a tagged image file format TIFF;

[0018] Step S5, integration of 10-meter land cover data to 500-meter resolution

[0019] The land cover data in TIFF format has a resolution of 10 meters. According to the needs, it is further integrated into multiple resolutions for model application. The data is integrated into a 500-meter grid, and the integration method specifically includes the following steps: for each 500-meter grid, clarify the 10-meter resolution grid contained in this grid, and record it in percentage form according to the land cover type to which these grids belong.

[0020] Step S6, matching of land cover data to vegetation functional type:

[0021] The 500-meter land cover data is matched to the vegetation functional type required by the numerical model, which specifically includes the following steps: first, read the 500-meter land cover data, and correspond to the 16 vegetation functional types used in numerical simulation;

[0022] Step S7, vegetation functional type comparison:

[0023] The high-resolution vegetation functional type completed in step S6 is compared with the traditional MODIS and GLC_FCS30-2020, ESAWorldCover produced vegetation functional type data, to evaluate the improvement of the newly produced high-resolution, refined land cover type data in identifying vegetation, especially urban green space;

[0024] Step S8, interpolation of 500-meter vegetation functional type to numerical model resolution

[0025] After completing the 500-meter vegetation functional type processing, these data are further interpolated into a 25-kilometer grid of spectral elements (SpectralElement). The NCL tool is used to convert the newly generated PFT data into NetCDF format compatible with the original input file;

[0026] Step S9, production of high-resolution earth system model SW-HRESM input file:

[0027] Based on Python and the climate data operation software CDO, a program is written to extract the PFT data in the newly generated NetCDF file, replace the corresponding variables in the original input file of SW-HRESM, and ensure that the variable name PCT_PFT, type double, dimension, and unit (%) are consistent with the original file;

[0028] Step S10, production of high-resolution biogenic emission inventory:

[0029] Using the high-resolution earth system model SW-HRESM and its biogenic gas and aerosol emission estimation model MEGAN, the input file generated in step S9 is input into the model for test run, and then SW-HRESM is run to generate a global high-resolution biogenic emission inventory with improved accuracy.

[0030] Step S11, atmospheric pollution simulation and subsequent analysis:

[0031] Based on the newly generated bio-source emission list, high-resolution earth system model simulation is carried out, and then a program is written to evaluate the influence of global bio-source emissions, especially urban green space emissions, on atmospheric pollutants.

[0032] As a preferred solution, 6 are selected from the spectral image of the Shenzhou-2 in step S1, and the total data reaches 87TB; the visible light is red, green and blue, and the resolution is 10 meters; the resolution of the three infrareds is 10-20 meters.

[0033] As a preferred solution, the data processing software in step S9 also includes NCO.

[0034] Compared with the prior art, the weakly supervised global land cover mapping method has the following beneficial effects: the weakly supervised global land cover mapping method is proposed, the spectral information of the Sentinel-2 (satellite data) image itself and the image features extracted by the visual large model (Vision Transformer; ViT) are fused, a set of high-resolution land cover data (WS-GLM1030-PSET) is made, the accuracy of land cover recognition is effectively improved, and the storage and processing pressure of data is reduced. Using the high-resolution land cover data, the PFT is improved, the simulation of BVOC emission in the global model SW-HRESM is optimized, the bio-source emission of urban green space is more accurately estimated, and the model can also better simulate the generation of ozone by improving the BVOC emission of urban green space. It is of great significance for formulating effective ozone pollution control strategies and improving urban air quality.

[0035] Additional aspects and advantages of the application will become apparent from the following description with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0036] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings, in which:

[0037] Figure 1 Process diagram for feature extraction based on pixel neighborhood and visual Transformer

[0038] Figure 2 Land cover mapping based on Sentinel image, MODIS, GLC_FCS30-2020, WS-GLM1030-PSET: taking Beijing (No. 190 grid) as an example. DETAILED DESCRIPTION

[0039] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0041] The following is combined Figures 1 to 2 This paper details an embodiment of a method for improving urban green space biosource emissions based on a high-resolution global model. A weakly supervised global land cover mapping method is proposed, which improves the accuracy of land cover identification by fusing spectral information from Sentinel-2 imagery with high-level features extracted by a Vision Transformer. Utilizing details from multispectral data and global semantic information from a deep learning model, the edge detection performance of land cover mapping is improved, enabling the model to generate more accurate land cover boundaries. This reduces the burden of data storage and processing and improves identification results, producing global 10-meter land cover data (WS-GLM1030-PSET).

[0042] The 10-meter land cover data WS-GLM1030-PSET was upscaled to 500 meters for comparison and evaluation with the commonly used MODIS satellite data.

[0043] 500-meter land cover data was converted into 25-kilometer spectral data, and PFT was calculated based on the land cover data. This data was then provided as input to the High Resolution Earth System Model (SW-HRESM). This enabled the High Resolution Earth System Model to estimate biosource emissions from urban green spaces, particularly emissions from urban green spaces.

[0044] like Figures 1 to 2 As shown, this invention proposes a method for improving biosource emissions from urban green spaces based on high-resolution global models, specifically including the following steps:

[0045] Step S1, Data Preprocessing:

[0046] Six bands (total data of 87 TB) were selected from the multispectral images of Sentinel-2, including three visible light (red, green, and blue; RGB; resolution of 10 meters) and three infrared bands (resolution of 10-20 meters), which can effectively improve the recognition accuracy and reduce the data storage and processing pressure. After selecting the data, the Google Earth Engine (GEE) platform was used for preprocessing, and the infrared band data was interpolated to 10-meter resolution, consistent with the resolution of the visible light band, to obtain 10-meter resolution Sentinel image data;

[0047] Step S2, pixel feature extraction:

[0048] For each Sentinel-2 pixel, 6 bands of its own and its surrounding neighborhood were extracted, and these multispectral features with detailed texture helped the model generate accurate feature boundaries. Second, the Vision Transformer pre-trained based on previous remote sensing data was used to extract high-level features of the same image; this deep learning model based on Vision Transformer can capture long-distance dependencies in the image. It helps the model to obtain coarse-grained semantic information from the image. The specific steps include:

[0049] Step S2-1, the process of feature extraction is shown in Figure 1 When extracting pixel features, the sliding window method is used to focus on the spectral and texture details of the image; multispectral information can enrich feature description and improve the accuracy and robustness of the model. For each pixel, 6-band spectral information of itself and its surrounding cross-region, a total of 9 pixels, is extracted; then the obtained spectral information is connected to form a long sequence of 54 length; this sliding window design can improve the performance of edge detection and enable the model to produce clear and sharp edges.

[0050] Step S2-2, a pre-trained deep learning model is used for feature extraction, and the Vision Transformer provides a visual self-attention mechanism model encoder to extract high-level semantic features in remote sensing images; this encoder captures a wide range of image contexts and long-distance dependencies in the image; each image is cropped into multiple slices with overlapping parts to meet the memory requirements of the encoder calculation and the continuity of the edges between the slices; then based on the visual self-attention mechanism model encoder, each slice is processed into a feature map with the same spatial size; finally, according to the above cropping mode, the feature maps of these slices are spliced back to the original image size;

[0051] Step S3, feature fusion:

[0052] The feature fusion method is used to balance the local and low-level features from the pixel feature extraction branch and the global and high-level features extracted from the deep learning model to obtain the optimal clustering result, and specifically includes the following steps: merging the spectral data from the pixel feature extraction branch and the data from the deep learning model branch to form a feature vector with a length of 64, and performing a standardization operation on the feature matrix to construct an input feature that can be directly used for clustering algorithm;

[0053] Step S4, clustering calculation and 10-meter land cover data image production:

[0054] After completing the feature extraction, a clustering analysis is performed based on a clustering algorithm (K-means) to generate 300-class clustering results, the relationship between the slightly lower resolution product (30 meters, GLC_FCS30-2020) and the clustering results is analyzed, and a corresponding land cover type is found for each cluster; and based on the data in the 10-meter resolution ESAWorldCover data predicted as green land category, the 10-meter resolution land cover classification result generated previously is rechecked, if the classification result of some area in the generated 10-meter data is not marked as green land, the area and its surrounding spectral features are re-judged and re-assigned to the category belonging to green land, automatic clustering merging is completed, and 10-meter high-resolution land cover data (WS-GLM1030-PSET) is produced, which is stored in a tag image file format TIFF; Beijing is selected as an example to show the comparison between the high-resolution land cover data produced by the Sentinel-2 and the present technology, and MODIS and GLC_FCS30-2020 are used as controls, as shown in Figure 2 , which shows the improvement of the present technology product in identifying urban green land.

[0055] Step S5, integration of 10-meter land cover data to 500-meter resolution

[0056] The land cover data in TIFF format has a resolution of 10 meters, and according to the needs, it is further integrated into multiple resolutions for model application; in order to be compatible with commonly used MODIS satellite data, the data is integrated into a 500-meter grid, and the integration method specifically includes the following steps: for each 500-meter grid, clarify the 10-meter resolution cells contained in this grid, and record the percentage of the land cover type to which these cells belong; for example, a 500-meter grid contains 2500 10-meter cells, of which 250 are deciduous broad-leaved forest, i.e. the proportion of deciduous broad-leaved forest in this grid is recorded as 10%. The data is in a comma-separated value (CSV) file.

[0057] Step S6, matching processing of land cover data to vegetation function type:

[0058] The 500-meter land cover data is matched to the vegetation functional types required by the numerical model, which includes the following steps: first, read the 500-meter land cover data corresponding to the 16 vegetation functional types used in numerical simulation; for example, among the commonly used 16 vegetation functional types, temperate broadleaf deciduous forest (Broadleaf deciduous temperate trees) is one of the types, which is matched from Open deciduous broadleaved forest and Closed deciduous broadleaved forest in the data according to the latitude zone;

[0059] Step S7, vegetation functional type comparison:

[0060] The high-resolution vegetation functional types completed in step S6 are compared with the traditional MODIS and GLC_FCS30-2020, ESA WorldCover produced vegetation functional type data, to evaluate the improvement of the newly produced high-resolution, refined land cover type data on vegetation, especially urban green space. Taking Beijing urban area as an example, the WS-GLM1030-PSET land cover image can capture more details of urban green space. In the MODIS data, due to the limitation of 500-meter resolution, many scattered or small-scale urban green spaces are difficult to accurately identify, which leads to the underestimation of urban green space area and distribution.

[0061] Step S8, interpolation of 500-meter vegetation functional types to numerical model resolution (25 kilometers)

[0062] After completing the 500-meter vegetation functional type processing, further interpolate these data to the 25-kilometer grid of the spectral element, and use the NCL tool to convert the newly generated PFT data to the NetCDF format compatible with the original input file;

[0063] Step S9, making high-resolution earth system model (SW-HRESM) input file:

[0064] Based on Python and climate data operation software CDO, a program is written to extract PFT data from the newly generated NetCDF file, replace the corresponding variables (variable name PCT_PFT) in the original input file of SW-HRESM, and ensure that the variable name, type (double), dimension and unit (%) are consistent with the original file;

[0065] Step S10, making high-resolution biogenic emission inventory:

[0066] The input file generated in step S9 is input into the high-resolution earth system model SW-HRESM and its land module, i.e., the biogenic gas and aerosol emission estimation model (MEGAN), to perform a test run to ensure that the MEGAN model can correctly read and use the new data. Then, the SW-HRESM is run to generate a global high-resolution biogenic emission inventory with improved accuracy, in particular, an emission inventory of volatile organic compounds from global urban green spaces;

[0067] Step S11, atmospheric pollution simulation and subsequent analysis:

[0068] Based on the newly generated biogenic emission inventory, a high-resolution earth system model simulation is performed, and a program is written to evaluate the impact of global biogenic emissions, in particular, urban green space emissions, on atmospheric pollutants (such as ozone and fine particulate matter). For example, the NCAR Command Language (NCL) is used to interpolate the model results (convert the spectral element grid to a latitude-longitude grid), extract atmospheric pollutants, and draw time series graphs and spatial distribution graphs.

[0069] In the description of the present application, the term "a plurality of" refers to two or more, unless otherwise explicitly limited, and the terms "upper", "lower", and the like indicate the orientation or positional relationship shown in the drawings based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application; the terms "connection", "installation", "fixation" and the like should be understood broadly, for example, "connection" can be fixed connection, can also be detachable connection, or integral connection; can be directly connected, or indirectly connected through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0070] In the description of the present application, the terms "one embodiment", "some embodiments", "a specific embodiment" and the like are intended to mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0071] The above is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for improving urban green space biogenic emissions based on a high-resolution global model, characterized in that, Specifically comprising the following steps: Step S1, select 6 from the multispectral image of Sentinel-2, including three visible light and three infrared bands, interpolate the infrared band data to 10 meter resolution, so as to obtain 10 meter resolution Sentinel image data; Step S2, feature extraction, specifically comprising: Step S2-1, when extracting pixel features, a sliding window method is used, for each pixel, 6 band spectral information of itself and surrounding cross region, a total of 9 pixels, is extracted, the obtained spectral information is connected to form a long sequence of 54 length as spectral data; Step S2-2, a pre-trained deep learning model is used for feature extraction, a visual self-attention mechanism model encoder provided by Vision Transformer is used to extract high-level semantic features in remote sensing image; Each image is cut into multiple slices with overlapping parts, and then each slice is processed into a feature map with the same spatial size based on the visual self-attention mechanism model encoder; Finally, the feature maps of these slices are spliced back to the original image size as the depth data; Step S3, merging the spectral data obtained in step S2-1 and the depth data obtained in step S2-2 to form a feature vector with a length of 64, and performing standardization operation on the feature matrix; Step S4, clustering analysis is performed to generate 300 class clustering results, the relationship between the resolution product of 30 meter GLC_FCS30-2020 and the clustering results is analyzed, and the corresponding land cover type is found for each cluster; Based on the data predicted as green land category in 10 meter resolution ESAWorldCover data, the previously generated 10 meter resolution land cover classification results are rechecked, if the classification results of some areas in the generated 10 meter data are not marked as green land, according to the spectral features of the area and its surrounding area, it is rejudged and redistributed as the category belonging to green land; Complete automatic clustering and merging to produce 10 meter high resolution land cover data WS-GLM1030-PSET, stored as tag image file format TIFF; Step S5, integrating 10 meter high resolution land cover data into 500 meter grid, for each 500 meter grid, clarify the 10 meter resolution grid contained in this grid, according to the land cover type to which these grids belong, record in percentage form; Step S6, reading 500 meter land cover data, corresponding to 16 vegetation function types used in numerical simulation; Step S7, comparing the high resolution vegetation function type completed in step S6 with the traditional MODIS and the vegetation function type data made by GLC_FCS30-2020 and ESAWorldCover; Step S8, after completing the 500 meter vegetation function type processing, interpolate these data into 25 kilometer grid, use NCL tool to convert the newly generated PFT data into NetCDF format compatible with the original input file; Step S9, a program is written based on Python and climate data operation software CDO to extract PFT data in the newly generated NetCDF file, replace the corresponding variables of the original input file of SW-HRESM, and ensure that the variable name, type, dimension and unit are consistent with the original file; Step S10, the newly generated input file in step S9 is input into the MEGAN model by using the high-resolution earth system model SW-HRESM and the biological source gas and aerosol emission estimation model MEGAN of the land module of SW-HRESM, a test run is performed, and then SW-HRESM is run to generate a global high-resolution biological source emission inventory; Step S11, based on the newly generated biological source emission inventory, a high-resolution earth system model simulation is carried out, and a program is written to evaluate the influence of urban green space emission on atmospheric pollutants.

2. The method for improving the urban green space biogenic emission based on high-resolution global model according to claim 1, characterized in that In the step S1, 6 of the 13 spectral images of the Shenzhou-2 satellite are selected, and the total data of the 6 spectral images reaches 87 TB; the visible light is red, green and blue, and the resolution is 10 meters; the resolution of the three infrareds is 10-20 meters.

3. The method for improving the biogenic emission of urban green space based on high-resolution global model according to claim 1, characterized in that In the step S9, the data processing software further includes NCO.

Images