A method for fusing all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total data

Abstract

The application relates to an all-weather high-temporal-spatial resolution multi-source remote sensing atmospheric water vapor total amount data fusion method, which comprises the following steps: obtaining original temporal-spatial image data of a geographical region to be studied; constructing a nonlinear fusion model between dependent variables and independent variables under clear sky conditions; and constructing all-weather atmospheric water vapor total amount data of the geographical region to be studied by using the original temporal-spatial image data and the nonlinear fusion model. The application can obtain atmospheric water vapor total amount data with a preset spatial resolution and in a continuous time and space, so that a user can obtain the spatial distribution details of the all-weather atmospheric water vapor total amount, and more abundant information can be provided for atmospheric environment monitoring and water cycle process research. Compared with the prior art, the atmospheric water vapor total amount data after fusion can be not affected by the ground signal and effectively cope with the high-temporal-spatial dynamic change characteristics of water vapor, the precision of the atmospheric water vapor total amount data after fusion is improved compared with the original data, and the spatial distribution trend of the original atmospheric water vapor total amount data is retained.

Description

Technical Field

[0001] This invention relates to the field of meteorological monitoring, and in particular to a method for fusing all-weather, high spatiotemporal resolution, multi-source remote sensing atmospheric water vapor total data. Background Technology

[0002] Water vapor is a key parameter in water cycle and climate change research. Total precipitable water (TPW), as a physical quantity measuring the spatiotemporal distribution characteristics of atmospheric water vapor, requires accurate measurement at global and regional scales. This is fundamental to our understanding of the water cycle, energy balance, and climate change. Currently, methods for observing TPW both domestically and internationally include in-situ observation and satellite remote sensing. In-situ observation, including Global Position System (GPS) observation and radiosonde observation, typically offers high accuracy but has limited spatial coverage and is usually used as validation data. Satellite remote sensing is an effective approach for large-scale TPW observation. Generally, infrared remote sensing offers high accuracy but only provides daytime data; thermal infrared observation provides data both day and night but with lower accuracy. Both methods are affected by weather conditions and cannot acquire sub-cloud data. Passive microwave remote sensing has the ability to retrieve TPW under all-weather conditions, but its spatial resolution is lower than that of infrared remote sensing. In summary, current remote sensing telemetry (TPW) data have their respective advantages and disadvantages. Observing the distribution of global or regional TPW data is limited and challenged by spatiotemporal continuity and resolution, leading to spatiotemporal discontinuities or low spatial resolution when studying local hydrological processes or atmospheric environments. This makes it difficult to accurately understand local water cycle processes and atmospheric environments. Therefore, making full use of existing data and developing algorithms that can acquire high-resolution spatiotemporally continuous TPW data is of great significance for meeting the needs of regional hydrological process and atmospheric environment characteristic monitoring research.

[0003] To accurately reconstruct high-resolution TPW in cloud-covered areas, appropriate auxiliary variables, fusion models, and fusion schemes must be selected. Compared to parameters such as surface reflectance and surface temperature, TPW exhibits rapid spatiotemporal variation characteristics. When fusing microwave and optical data to reconstruct high-resolution TPW in cloud-covered areas, the selection of fusion variables needs to consider the spatiotemporal variation characteristics of atmospheric water vapor. The method of directly fusing microwave brightness temperature and optical remote sensing data to reconstruct high-resolution surface temperature is not entirely applicable to high-resolution TPW reconstruction under clouds. This is mainly because the microwave band, which is sensitive to water vapor, is also affected by the high surface emissivity of the microwave band. When using only microwave brightness temperature (such as 23GHz band brightness temperature), which is sensitive to water vapor, to fuse with optical data to reconstruct high-resolution TPW under clouds, surface signals will be introduced to some extent, negatively impacting the fusion effect. Furthermore, under cloud conditions, when using only high spatial resolution auxiliary information such as DEM and NDVI to reconstruct high-resolution TPW in combination with the fusion model, there will be insufficient reconstruction of high-resolution TPW information. This is mainly because DEM and NDVI can only reflect the correlation with TPW on a large spatial and temporal scale, and cannot capture the rapid changes in TPW. Summary of the Invention

[0004] This application provides a method for fusing all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total data. The technical solution is as follows:

[0005] In a first aspect, this application provides a method for fusing all-weather, high spatiotemporal resolution, multi-source remote sensing atmospheric water vapor total data, the method comprising:

[0006] Based on the infrared observation of total atmospheric water vapor, surface elevation data, normalized vegetation index, microwave observation of total atmospheric water vapor, first brightness temperature, second brightness temperature and third brightness temperature, and microwave observation of precipitation in the geographical area to be studied, the original spatiotemporal image data of the geographical area to be studied are obtained.

[0007] Quality control is performed on the infrared observation atmospheric water vapor total amount information and microwave observation atmospheric water vapor total amount information at the same grid location in the original spatiotemporal image data to obtain sample spatiotemporal image data of the geographical area to be studied.

[0008] Using the total atmospheric water vapor information from infrared observations that meet clear sky conditions in the sample spatiotemporal image data as the dependent variable, and other data that meet clear sky conditions in the sample spatiotemporal data as the independent variable, a nonlinear fusion model between the dependent variable and the independent variable under the clear sky conditions is constructed.

[0009] Using the microwave precipitation data as a precipitation mask, the original spatiotemporal image is divided into datasets in precipitation and non-precipitation areas. By combining the nonlinear fusion model of the precipitation and non-precipitation areas, high-resolution TPWs of the precipitation and non-precipitation areas are generated. The two are then stitched together using the precipitation mask to construct the all-weather total atmospheric water vapor data of the geographical area under study.

[0010] Secondly, this application provides an all-weather, high spatiotemporal resolution, multi-source remote sensing atmospheric water vapor total fusion data acquisition device, the device comprising:

[0011] The acquisition module is used to obtain the original spatiotemporal image data of the geographical area under study based on the infrared observation of total atmospheric water vapor, surface elevation data, normalized vegetation index, microwave observation of total atmospheric water vapor, first brightness temperature, second brightness temperature, third brightness temperature and microwave observation of precipitation of the geographical area under study.

[0012] The filtering module is used to perform quality control on the infrared observation atmospheric water vapor total amount information and microwave observation atmospheric water vapor total amount information at the same grid position in the original spatiotemporal image data to obtain sample spatiotemporal image data of the geographical area to be studied.

[0013] The first processing module is used to take the total atmospheric water vapor information of infrared observations that meet the clear sky conditions in the sample spatiotemporal image data as the dependent variable and other data that meet the clear sky conditions in the sample spatiotemporal data as the independent variable, and respectively construct a nonlinear fusion model between the dependent variable and the independent variable under the clear sky conditions, which is applied to precipitation and non-precipitation areas.

[0014] The second processing module is used to divide the original spatiotemporal image into datasets in precipitation and non-precipitation areas using the microwave precipitation data as a precipitation mask. It then combines the nonlinear fusion models of the precipitation and non-precipitation areas to produce high-resolution TPWs for both areas. Finally, it stitches the two datasets together using the precipitation mask to construct all-weather atmospheric water vapor total data for the geographical area under study.

[0015] Thirdly, this application provides a computer-readable storage medium including a computer program that, when executed on a device, causes the device to perform the method described in the first aspect above.

[0016] Fourthly, this application provides a chip system including a chip coupled to a memory for executing a computer program stored in the memory to perform the method described in the first aspect above.

[0017] The beneficial effects of the technical solution provided in this application include at least the following:

[0018] The method proposed in this application can acquire spatiotemporally continuous TPW data with a spatial resolution of 0.05°×0.05°, enabling users to obtain detailed spatial distribution of TPW data throughout the day and providing richer information for atmospheric environmental monitoring and water cycle process research. Compared with existing methods for acquiring high-resolution spatiotemporally continuous water vapor, the method provided in this application is unaffected by surface signals and effectively addresses the high spatiotemporal dynamics of water vapor. The fused water vapor data shows improved accuracy compared to the original data, while preserving the spatial distribution trend of the original water vapor data. Attached Figure Description

[0019] Figure 1 A schematic flowchart illustrating a method for fusing all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor totality data, provided in an embodiment of the present invention;

[0020] Figure 2 A schematic block diagram of an all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total fusion data acquisition device provided in an embodiment of the present invention;

[0021] Figure 3 A schematic architecture diagram of a method for fusing total atmospheric water vapor data from multiple sources using high spatiotemporal resolution in all-weather remote sensing, provided in an embodiment of the present invention;

[0022] Figure 4 A schematic diagram illustrating the effect of a method for fusing all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total data provided in an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram illustrating the effect of a method for fusing total atmospheric water vapor data from multiple sources using high spatiotemporal resolution in all weather conditions, as provided in an embodiment of the present invention. Specific implementation methods

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, Embodiment 1 of the present invention provides a method for fusing all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total data, including the following steps;

[0026] S1. Based on the infrared observation of total atmospheric water vapor, surface elevation data, normalized vegetation index, microwave observation of total atmospheric water vapor, first brightness temperature, second brightness temperature, third brightness temperature and microwave observation of precipitation in the geographical area to be studied, the original spatiotemporal image data of the geographical area to be studied are obtained.

[0027] S2. Perform quality control on the infrared observation atmospheric water vapor total amount information and microwave observation atmospheric water vapor total amount information at the same grid position in the original spatiotemporal image data to obtain sample spatiotemporal image data of the geographical area to be studied.

[0028] S3. Using the total atmospheric water vapor information of infrared observations that meet the clear sky conditions in the sample spatiotemporal image data as the dependent variable, and other data that meet the clear sky conditions in the sample spatiotemporal data as the independent variable, a nonlinear fusion model between the dependent variable and the independent variable under the clear sky conditions is constructed and applied to precipitation and non-precipitation areas respectively.

[0029] S4. Using the microwave precipitation data as a precipitation mask, the original spatiotemporal image data is divided into datasets in precipitation areas and non-precipitation areas. Combining the nonlinear fusion models of the precipitation areas and non-precipitation areas, high-resolution TPWs of the precipitation areas and non-precipitation areas are generated. The two are stitched together using the precipitation mask to construct the all-weather atmospheric water vapor total data of the geographical area to be studied.

[0030] Based on the above embodiments, S1 further includes;

[0031] S11. The infrared observation of total atmospheric water vapor in the geographical area to be studied, the surface elevation data and the normalized vegetation index are resampled into the latitude and longitude grid image of the geographical area to be studied. The infrared observation of total atmospheric water vapor is obtained by observing the geographical area to be studied using an infrared radiometer.

[0032] S12. After converting the date in the microwave observation atmospheric water vapor total information to Julian date, the microwave observation atmospheric water vapor total information, the first brightness temperature, the second brightness temperature, the third brightness temperature, and the microwave precipitation are interpolated into the iso-latitude and longitude grid image to obtain the original spatiotemporal image data of the geographical area to be studied. The microwave observation atmospheric water vapor total information and precipitation information are obtained by observing the geographical area to be studied using a passive microwave radiometer. The first brightness temperature is the 18 GHz brightness temperature, the second brightness temperature is the 23 GHz brightness temperature, and the third brightness temperature is the 89 GHz brightness temperature.

[0033] Based on the above embodiments, S11 further includes:

[0034] The longitude, latitude, infrared atmospheric water vapor totality information, surface elevation data, and normalized vegetation index of the pixels in the image of the geographical area under study are used to formulate the first formula. Second Formula And the third formula Resampled to an equal latitude and longitude grid image with a preset spatial resolution, where Row is the row number of the i-th pixel under equal latitude and longitude projection, Col is the column number of the i-th pixel under equal latitude and longitude projection, and Lat... i It is the longitude of the i-th pixel, Lon i R is the latitude of the i-th pixel, and R is the spatial resolution of the isotropic grid image. top It is the latitude of the top left corner of the latitude and longitude grid in the isotropic latitude and longitude grid image. left It is the longitude of the top left corner of the equal latitude and longitude grid in the equal latitude and longitude grid image, v i Value is the value of the i-th pixel, and Value is the value of the resampled Row row and Col column. N is the total number of effective pixels in the Row row and Col column of the original data resampled to the same latitude and longitude grid image.

[0035] Based on the above embodiments, it further includes: if there is a data gap in the normalized vegetation index in the time series, the normalized vegetation index of the missing date is filled with the NDVI data of the previous day.

[0036] Based on the above embodiments, S2 further includes:

[0037] The location of the pixel in the original spatiotemporal image data that meets the clear sky condition based on the infrared observation atmospheric water vapor total information is used as the mask of the preset clear sky condition for the fused data. The infrared observation atmospheric water vapor total information, surface elevation data, normalized vegetation index, microwave observation atmospheric water vapor total information, first brightness temperature, second brightness temperature and third brightness temperature within the first preset date area that meet the clear sky condition are selected as the spatiotemporal image intermediate data.

[0038] The intermediate data of the spatiotemporal image is input into the following fourth formula: Delta = NIR TPW -MW TPW And in the fifth formula, Valid = {Delta}, the constraint condition for Delta is: Delta ≥ (MEANDelta - STDDelta)

[0039] Delta ≤ (MEANDelta + STDDelta), where NIR TPW These are infrared observations of total atmospheric water vapor at the same grid location, in MW. TPWThis refers to the total atmospheric water vapor information observed via microwave at the same grid location. Delta is the difference between the total atmospheric water vapor information observed via infrared and microwave at the same grid location. Delta It is the mean of Delta, STD Delta Valid is the standard deviation of Delta, and Valid is the pixel position corresponding to Delta that meets the fusion constraint requirements. Valid serves as the criterion for judging valid data. After filtering the independent and dependent variable data used to construct the fusion model, the sample spatiotemporal image data of the geographical area to be studied are obtained.

[0040] Based on the above embodiments, S3 further includes:

[0041] The total atmospheric water vapor content observed by infrared observations in the sample spatiotemporal image data that meets the clear sky conditions is used as the dependent variable, and the total atmospheric water vapor content observed by microwave observations in the sample spatiotemporal data that meets the clear sky conditions (MW) is used as the dependent variable. TPW The surface elevation data (Elevation), the Normalized Difference Vegetation Index (NDVI), the first brightness temperature (18 GHz BT), the second brightness temperature (23 GHz BT), the third brightness temperature (89 GHz BT), longitude (Lat), latitude (Lon), and Julian date (JD) are used as independent variables, and the following sixth formula is used to calculate the independent variables.

[0042] NIR TPW =f{MW TPW ,18GHzBT,23GHzBT,89GHzBT,Elevation,NDVI,Lat,Lon,JD}and the

[0043] Seven Formulas NIR TPW =f{MW TPW The nonlinear fusion model is constructed using the random forest algorithm, 18GHzBT,23GHzBT,ELevation,NDVI,Lat,Lon,JD}.

[0044] Based on the above embodiments, S4 further includes:

[0045] The portion of the total atmospheric water vapor observed by microwaves in the second preset date region of the original spatiotemporal image data that is greater than zero is used as an all-weather mask. The invalid data of the microwave observation precipitation product is used as a non-precipitation mask. The total atmospheric water vapor observed by microwaves, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, the third brightness temperature, longitude, latitude, and Julian date that meet the non-precipitation conditions in the preset date are extracted and input into the sixth formula to obtain the total atmospheric water vapor in the non-precipitation region.

[0046] Using the effective data of the microwave observation precipitation product as a precipitation mask, the total atmospheric water vapor from the microwave observation, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, longitude, latitude, and Julian date that meet the precipitation conditions in the preset date are extracted and input into the seventh formula to obtain the total atmospheric water vapor in the precipitation area.

[0047] Based on the non-precipitation mask and the precipitation mask, the total atmospheric water vapor in the non-precipitation area and the total atmospheric water vapor in the precipitation area are spliced ​​together to obtain the all-weather total atmospheric water vapor data of the geographical area to be studied.

[0048] It should be understood that microwave precipitation products are used to construct masks for both precipitation and non-precipitation areas. The independent variables in the fusion model for precipitation areas do not include 89GHz BT, while the independent variables in the fusion model for non-precipitation areas do include 89GHz BT. Microwave total atmospheric water vapor (MW TPW) data and other dependent variable data are used through the aforementioned two masks and two fusion models to produce high-resolution total atmospheric water vapor data for both precipitation and non-precipitation areas, respectively.

[0049] It should be noted that current methods for fusion of optical and microwave remote sensing data mainly fall into several categories, including spatial statistical downscaling methods, time series analysis methods, spatiotemporal adaptive methods, and machine learning methods. However, most downscaling or fusion methods are not specifically designed for the characteristics of TPW (Transient Term Warp). The main problems with these methods include: 1) TPW is a quantity with high spatiotemporal dynamic characteristics. The auxiliary data used in spatial statistical downscaling methods cannot accurately represent the detailed spatial features of TPW. When the spatiotemporal distribution of TPW changes drastically, these auxiliary variables used for downscaling cannot reconstruct the actual spatial distribution characteristics of TPW. 2) When reconstructing high-resolution data in cloud-covered areas, time series analysis methods use transformation functions established under clear-sky conditions. In many cases, the temporal variation characteristics of surface or atmospheric parameters under clouds differ significantly from those under clear skies. This leads to situations where the function transformation relationship established under clear-sky conditions cannot accurately reconstruct high-resolution surface or atmospheric parameters under clouds. 3) Spatiotemporal adaptive methods utilize spatially matched coarse-resolution and high-resolution data from two time points before and after the time to be estimated to achieve high-resolution reconstruction of coarse-resolution data. This method is not applicable when high-resolution data is lacking, such as in areas with persistent cloud cover. 4) Learning-based methods operate as a "black box" during model training, making it difficult to summarize and generalize the principles of downscaling.

[0050] To address the aforementioned issues and considering the advantages and disadvantages of various current fusion methods, this application, based on a random forest model, incorporates existing microwave brightness temperature, topographic, and geomorphological information. It introduces newly developed all-weather microwave TPW products less susceptible to surface signal interference and high-resolution 89GHz high-frequency microwave brightness temperature information containing rich spatial information from TPW data. Simultaneously, it adds auxiliary information such as time and location to construct a high-resolution TPW fusion algorithm that integrates microwave and optical remote sensing observations. This ultimately yields all-weather TPW data with a spatial resolution of 0.05°×0.05°, while simultaneously improving the accuracy of the fused data compared to the original data.

[0051] like Figure 3 As shown, the data used in this technical solution mainly comes from the passive microwave radiometer's TPW (MW TPW) and 18GHz BT, 23GHz BT, 89GHz BT, precipitation (RR), the TPW (NIRTPW) acquired based on the near-infrared band of the infrared radiometer, and auxiliary data composed of the Julian date (JD) of microwave remote sensing data, the longitude (Lon), latitude (Lat) of the infrared radiometer, the elevation, and the normalized difference vegetation index (NDVI) of the optical sensor. Then, under clear sky conditions, a random forest (RF) model and variables such as NIR TPW, MW TPW, 18GHz BT, 23GHz BT, 89GHz BT, Lon, Lat, JD, NDVI, and elevation are used to establish fusion models applied to precipitation and non-precipitation areas on a daily basis. Finally, using microwave precipitation data as a precipitation mask, the datasets were divided into precipitation and non-precipitation independent variable datasets. The precipitation and non-precipitation independent variables were applied to the precipitation and non-precipitation regional fusion models, respectively, to obtain high spatial resolution TPWs for both precipitation and non-precipitation regions. The TPWs of the two regions were then mosaicked to generate an all-weather high spatial resolution spatially continuous TPW. The main improvement of this invention is the introduction of weather-independent MW TPWs and 89GHz high-frequency brightness temperature containing water vapor information. Combined with the high-resolution information of NIR TPWs, a parameterization scheme with high spatiotemporal dynamic characteristics was constructed, achieving the acquisition of high-resolution all-weather TPWs. Furthermore, the fusion result has higher accuracy than the original data. The specific details of the technical solution are as follows:

[0052] (1) Spatiotemporal matching of fused data

[0053] In terms of spatiotemporal matching of data in the fusion method, the high spatial resolution NIR TPW, Elevation and NDVI pixels are traversed to extract the latitude and longitude information and data value of the pixels. Based on formula (1-3), they are resampled into the same latitude and longitude grid image with a spatial resolution of 0.05°×0.05°.

[0054]

[0055]

[0056]

[0057] Where Row and Col are the row and column numbers of the i-th pixel under the same latitude and longitude projection, Lati and Loni are the latitude and longitude of the i-th pixel, Lattop and Lonleft are the latitude and longitude of the top left corner of the same latitude and longitude grid, R is the spatial resolution of the same latitude and longitude grid image, vi is the value of the i-th pixel, ValueRow,Col is the resampled value of Row row and Col column, and N is the total number of effective pixels in Row row and Col column of the original data resampled to the same latitude and longitude projection.

[0058] The year, month, and day times of the microwave TPW were converted to the number of days since January 1, 1979, denoted as the Julian Date (JD). The low-resolution microwave TPW, vertically and horizontally polarized 18GHz BT, 23GHz BT, 89GHz BT, and microwave precipitation were interpolated to 0.05° × 0.05° using bilinear interpolation. Microwave precipitation will be used in the third step as a precipitation mask to divide the independent variable data into precipitation and non-precipitation regions. Because NDVI is a composite product of 16-day maximum values, there are missing data in the time series. The missing NDVI data within the 16 days were filled in using the NDVI data from the previous day, obtaining NDVI data with a complete time series, thus completing the matching of NDVI data with other data in the time series. At this point, the spatiotemporal resolution of all data required for fusion was unified to 0.05° / day.

[0059] (2) Construction of fusion model under clear sky conditions

[0060] Due to differences in the sensors used to acquire MW TPW and NIR TPW, as well as differences in orbital parameters such as scan width and intersection time, the inversion methods differ, leading to discrepancies between the two results. Therefore, before constructing the fusion model under clear sky conditions, quality control of the input data should be performed, selecting pixels with smaller differences between the two TPW data sets as valid data for building the fusion model. After quality control, a fusion model was built using near-infrared TPW (NIR TPW) as the dependent variable and microwave TPW (MWTPW), 18GHz BT, 23GHz BT, 89GHz BT, elevation, NDVI, latitude (Lat), longitude (Lon), and JD as independent variables, employing a random forest algorithm. Specific details are as follows:

[0061] 1) Quality Control. For TPW fusion on date N, the pixel locations under clear sky conditions in the NIR TPW are used as masks for the clear sky conditions in the fusion data. Near-infrared TPW (NIR TPW), microwave TPW (MW TPW), 18GHz BT, 23GHz BT, 89GHz BT, Elevation, NDVI, Lat, Lon, and JD under clear sky conditions from date N-3 to N+3 are selected to construct the fusion model. To ensure the stability and generalization ability of the constructed fusion model, quality control is performed on the data used to construct the fusion model. The quality control formula is as follows:

[0062] Delta = NIR TPW -MW TPW Formula (4)

[0063] Valid={Delta} Formula (5)

[0064] Constraint for Delta: Delta ≥ (MEANDelta - STDDelta)

[0065] Delta ≤ (MEANDelta + STDDelta)

[0066] NIR TPW MW TPW These represent near-infrared and microwave TPW values ​​at the same grid location, respectively. Delta is the difference between the two TPW values. Delta STD Delta These are the mean and standard deviation of Delta, respectively. Here, one standard deviation around the mean is used as a constraint. Valid is the pixel position corresponding to Delta that satisfies the fusion constraint requirements. Valid is selected as the criterion for judging valid data, and the independent and dependent variable data used to construct the fusion model are filtered.

[0067] 2) Constructing a fusion model under clear sky conditions. After quality control of the fused data, a fusion model with nonlinear relationships was constructed using NIR TPW as the dependent variable and MW TPW, 18GHz BT, 23GHz BT, 89GHz BT, Elevation, NDVI, Lat, Lon, and JD as independent variables. The specific formula is as follows:

[0068] NIR TPW =f{MW TPW ,18GHzBT,23GHzBT,89GHzBT,Elevation,NDVI,Lat,Lon,JD}Formula (6)

[0069] NIR TPW =f{MWTPW ,18GHzBT,23GHzBT,ELevation,NDVI,Lat,Lon,JD} Formula (7)

[0070] Where f represents the nonlinear relationship between the independent and dependent variables. Because the random forest algorithm has the advantage of better solving nonlinear problems compared with traditional statistical algorithms, the random forest algorithm is chosen to construct the fusion model with nonlinear relationship. Considering the sensitivity of 89GHz BT to clouds, the use of 89GHz BT in precipitation areas will affect the fusion effect. Therefore, different fusion models are constructed in precipitation and non-precipitation areas. The fusion model constructed by formula (6) with 89GHz BT is selected for fusion in non-precipitation areas, and the fusion model constructed by formula (7) without 89GHz BT is selected for fusion in precipitation areas.

[0071] (3) High-resolution fusion water vapor production under all weather conditions

[0072] Based on the assumption that the fusion model constructed under clear skies still holds true under all-weather conditions, the clear skies model is applied to all-weather conditions to produce all-weather high-resolution fused TPW. The portion of MW TPW greater than zero on date N is used as the all-weather mask, and the invalid pixels of the microwave precipitation product are used as the non-precipitation mask. Microwave TPW (MWTPW), 18GHz BT, 23GHz BT, 89GHz BT, Elevation, NDVI, Lat, Lon, and JD data under non-precipitation conditions on date N are extracted and input into the fusion model constructed by formula (6) in step 2 to produce TPW for the 0.05°×0.05° non-precipitation area. In the precipitation area, the valid pixels of the microwave precipitation product are used as the precipitation mask. Microwave TPW (MWTPW), 18GHz BT, 23GHz BT, Elevation, NDVI, Lat, Lon, and JD data under precipitation conditions on date N are extracted and input into the fusion model constructed by formula (7) in step 2 to produce TPW for the precipitation area. The TPW obtained from non-precipitation areas and precipitation areas is mosaicked to obtain a 0.05°×0.05° all-weather TPW.

[0073] Based on the non-precipitation mask and the precipitation mask, the total atmospheric water vapor in the non-precipitation area and the total atmospheric water vapor in the precipitation area are stitched together to obtain all-weather high-resolution total atmospheric water vapor data for the geographical area under study.

[0074] The embodiments described above in this application can acquire spatiotemporally continuous TPW data with a spatial resolution of 0.05°×0.05°, enabling users to obtain detailed spatial distribution of TPW data throughout the day and providing richer information for atmospheric environmental monitoring and water cycle process research. Compared with existing methods for acquiring high-resolution spatiotemporally continuous water vapor, the method provided in this application is unaffected by surface signals and effectively addresses the high spatiotemporal dynamics of water vapor. The fused water vapor data shows improved accuracy compared to the original data while preserving the spatial distribution trend of the original water vapor data.

[0075] The effects of applying the method in this application in practice are explained below.

[0076] like Figure 4-5 As shown, the TPW data of AMSR2 and NIRTPW data of MYD05 were fused in the region of 54°N~3°N, 73°E~136°E in 2016.

[0077] Figure 4 This paper compares the spatial distribution of MYD05, AMSR2, and the fused TPW within the region of 54°N–3°N and 73°E–136°E on April 29, 2016. In the figure, a and b represent the overall and local distributions of MYD05 TPW, respectively; c and d represent the overall and local distributions of AMSR2 TPW, respectively; and e and f represent the overall and local distributions of the fused TPW, respectively. The comparison shows that the fused TPW data has richer distribution details than AMSR2 TPW and more comprehensive spatial coverage than the MYD05 NIR TPW resampled to 0.05°, thus better displaying the spatial distribution of TPW. Furthermore, this invention also validated the fused TPW data at 0.05° resolution using TPW observed by GPS from the China Land Status Network. Figure 5 As shown. Figure 5 This is a verification comparison between MYD05, AMSR2, and the fused TPW data and the GPS TPW data from the land surface network in the region of 54°N~3°N, 73°E~136°E in 2016. Among them, a. MYD05 TPW; b. AMSR2 TPW; c. fused TPW. As can be seen from the figure, the fused water vapor data shows improvements in both correlation and root mean square error (RMSE) compared to the original data.

[0078] like Figure 2 As shown, this application provides an all-weather, high spatiotemporal resolution multi-source remote sensing atmospheric water vapor total fusion data acquisition device, the device comprising:

[0079] The acquisition module is used to obtain the original spatiotemporal image data of the geographical area under study based on the infrared observation of total atmospheric water vapor, surface elevation data, normalized vegetation index, microwave observation of total atmospheric water vapor, first brightness temperature, second brightness temperature, third brightness temperature and microwave observation of precipitation of the geographical area under study.

[0080] The filtering module is used to perform quality control on the infrared observation atmospheric water vapor total amount information and microwave observation atmospheric water vapor total amount information at the same grid position in the original spatiotemporal image data to obtain sample spatiotemporal image data of the geographical area to be studied.

[0081] The first processing module is used to take the total atmospheric water vapor information of infrared observations that meet the clear sky conditions in the sample spatiotemporal image data as the dependent variable and other data that meet the clear sky conditions in the sample spatiotemporal data as the independent variable, and respectively construct a nonlinear fusion model between the dependent variable and the independent variable under the clear sky conditions, which is applied to precipitation and non-precipitation areas.

[0082] The second processing module is used to divide the original spatiotemporal image into datasets in precipitation and non-precipitation areas using the microwave precipitation data as a precipitation mask. It then combines the nonlinear fusion models of the precipitation and non-precipitation areas to produce high-resolution TPWs for both areas. Finally, it stitches the two datasets together using the precipitation mask to construct all-weather atmospheric water vapor total data for the geographical area under study.

[0083] Based on the above embodiments, the acquisition module is specifically used to resample the infrared observation atmospheric water vapor total information of the geographical area to be studied, the surface elevation data, and the normalized vegetation index into the latitude and longitude grid image of the geographical area to be studied. The infrared observation atmospheric water vapor total information is obtained by observing the geographical area to be studied using an infrared radiometer.

[0084] After converting the date in the total atmospheric water vapor information from microwave observations to Julian days, the total atmospheric water vapor information from microwave observations, the first brightness temperature, the second brightness temperature, the third brightness temperature, and the microwave precipitation are interpolated into the isotropic grid image to obtain the original spatiotemporal image data of the geographical area to be studied. The total atmospheric water vapor information and precipitation information from microwave observations are obtained from observations of the geographical area to be studied using a passive microwave radiometer. The first brightness temperature is the 18 GHz brightness temperature, the second brightness temperature is the 23 GHz brightness temperature, and the third brightness temperature is the 89 GHz brightness temperature.

[0085] Based on the above embodiments, the acquisition module is specifically used to acquire the longitude, latitude, infrared atmospheric water vapor totality information, surface elevation data, and normalized vegetation index of the pixels of the image of the geographical area to be studied using a first formula. Second Formula And the third formula Resampled to an equal latitude and longitude grid image with a preset spatial resolution, where Row is the row number of the i-th pixel under equal latitude and longitude projection, Col is the column number of the i-th pixel under equal latitude and longitude projection, and Lat... i It is the longitude of the i-th pixel, Lon i R is the latitude of the i-th pixel, and R is the spatial resolution of the isotropic grid image. top It is the latitude of the top left corner of the latitude and longitude grid in the isotropic latitude and longitude grid image. left It is the longitude of the top left corner of the equal latitude and longitude grid in the equal latitude and longitude grid image, v i Value is the value of the i-th pixel, and Value is the value of the resampled Row row and Col column. N is the total number of effective pixels in the Row row and Col column of the original data resampled to the same latitude and longitude grid image.

[0086] The filtering module is specifically used to take the location of the pixel in the original spatiotemporal image data where the infrared observation atmospheric water vapor total information meets the clear sky condition as the mask of the preset clear sky condition of the fused data, and select the infrared observation atmospheric water vapor total information, surface elevation data, normalized vegetation index, microwave observation atmospheric water vapor total information, first brightness temperature, second brightness temperature and third brightness temperature in the first preset date area that meet the clear sky condition as the spatiotemporal image intermediate data;

[0087] The intermediate data of the spatiotemporal image is input into the following fourth formula: Delta = NIR TPW -MW TPW And in the fifth formula, Valid = {Delta}, the constraint condition for Delta is: Delta ≥ (MEANDelta - STDDelta)

[0088] Delta≤(MEAN Delta +STD Delta ), where NIR TPW These are infrared observations of total atmospheric water vapor at the same grid location, in MW. TPW This refers to the total atmospheric water vapor information observed via microwave at the same grid location. Delta is the difference between the total atmospheric water vapor information observed via infrared and microwave at the same grid location. Delta It is the mean of Delta, STD DeltaValid is the standard deviation of Delta, and Valid is the pixel position corresponding to Delta that meets the fusion constraint requirements. Valid serves as the criterion for judging valid data. After filtering the independent and dependent variable data used to construct the fusion model, the sample spatiotemporal image data of the geographical area to be studied are obtained.

[0089] Based on the above embodiments, the first processing module is specifically used to take the infrared observation atmospheric water vapor total amount information that meets the clear sky conditions in the sample spatiotemporal image data as the dependent variable, and the microwave observation atmospheric water vapor total amount (MW) that meets the clear sky conditions in the sample spatiotemporal image data as the dependent variable. TPW The surface elevation data (Elevation), the Normalized Difference Vegetation Index (NDVI), the first brightness temperature (18 GHz BT), the second brightness temperature (23 GHz BT), the third brightness temperature (89 GHz BT), longitude (Lat), latitude (Lon), and Julian date (JD) are used as independent variables, and the following sixth formula is used to calculate the independent variables.

[0090] NIR TPW =f{MW TPW ,18GHzBT,23GHzBT,89GHzBT,Elevation,NDVI,Lat,Lon,JD}and the seventh formula NIR TPW =f{MW TPW The nonlinear fusion model is constructed using the random forest algorithm, 18GHzBT,23GHzBT,ELevation,NDVI,Lat,Lon,JD}.

[0091] Based on the above embodiments, the second processing module is specifically used to take the portion of the total atmospheric water vapor observed by microwave observation in the second preset date region of the original spatiotemporal image data that is greater than zero as an all-weather mask, and take the invalid data of the microwave observation precipitation product as a non-precipitation mask. The module extracts the total atmospheric water vapor observed by microwave observation, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, the third brightness temperature, longitude, latitude and Julian date that meet the non-precipitation conditions in the preset date and inputs them into the sixth formula to obtain the total atmospheric water vapor in the non-precipitation region.

[0092] Using the effective data of the microwave observation precipitation product as a precipitation mask, the total atmospheric water vapor from the microwave observation, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, longitude, latitude, and Julian date that meet the precipitation conditions in the preset date are extracted and input into the seventh formula to obtain the total atmospheric water vapor in the precipitation area.

[0093] Based on the non-precipitation mask and the precipitation mask, the total atmospheric water vapor in the non-precipitation area and the total atmospheric water vapor in the precipitation area are spliced ​​together to obtain the all-weather total atmospheric water vapor data of the geographical area to be studied.

[0094] Furthermore, this application provides a computer-readable storage medium including a computer program that, when executed on a device, causes the device to perform the methods described in the above embodiments.

[0095] This application also provides a chip system including a chip coupled to a memory for executing a computer program stored in the memory to perform the methods described in the above embodiments.

[0096] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An all-weather high-temporal and spatial resolution multi-source remote sensing atmospheric water vapor total data fusion method, characterized in that, The method comprises: The method comprises: The first brightness temperature is 18GHz brightness temperature, the second brightness temperature is 23GHz brightness temperature, and the third brightness temperature is 89GHz brightness temperature. The method comprises: The method comprises: Specifically, infrared observed atmospheric water vapor total amount information meeting the clear sky condition in the sample spatio-temporal image data is taken as the dependent variable, microwave observed atmospheric water vapor total amount information meeting the clear sky condition in the sample spatio-temporal image data is taken as the independent variable, and the non-linear fusion model is constructed by the sixth formula , the surface elevation data Elevation, the normalized vegetation index NDVI, the first brightness temperature 18GHz BT, the second brightness temperature 23GHz BT, the third brightness temperature 89GHz BT, the longitude Lat, the latitude Lon and the Julian date JD as the independent variables, through the following sixth formula and the seventh formula , the random forest algorithm is used to construct the non-linear fusion model. The method comprises:

2. The method of claim 1, wherein, The method comprises: The method comprises: The method comprises:

3. The method of claim 2, wherein, The method comprises: The longitude, latitude, infrared observation total atmospheric water vapor information, surface elevation data and normalized vegetation index of the pixels of the image of the geographical region to be studied are resampled into a preset spatial resolution equilatitude and longitude grid image by a first formula , a second formula and a third formula , wherein Row is the row number corresponding to the i-th pixel under equilatitude and longitude projection, Col is the column number corresponding to the i-th pixel under equilatitude and longitude projection, Lat i is the longitude of the i-th pixel, Lon i is the latitude of the i-th pixel, R is the spatial resolution of the equilatitude and longitude grid image, Lat top is the latitude of the upper left corner of the equilatitude and longitude grid image, Lon left is the longitude of the upper left corner of the equilatitude and longitude grid image, v i is the value of the i-th pixel, Value is the resampled value of Row row Col column, and N is the total amount of effective pixels of Row row Col resampled from the original data to the equilatitude and longitude grid image.

4. The method of claim 2, wherein, If the normalized vegetation index has data missing in the time sequence, the normalized vegetation index of the missing date is filled with the NDVI data of the previous day. ​ 5. The method of claim 1, wherein, The quality control is performed on the infrared observation atmospheric water vapor total amount information and the microwave observation atmospheric water vapor total amount information at the same grid position in the original spatio-temporal image data, to obtain sample spatio-temporal image data of the geographical region to be researched. The position of the pixel in the original spatio-temporal image data, at which the infrared observation atmospheric water vapor total amount information satisfies the clear-sky condition, is taken as a mask of the preset clear-sky condition of the fusion data, and the infrared observation atmospheric water vapor total amount information, the surface elevation data, the normalized vegetation index, the microwave observation atmospheric water vapor total amount information, the first brightness temperature, the second brightness temperature and the third brightness temperature in the first preset date region, which satisfy the clear-sky condition, are selected as intermediate data in the spatio-temporal image. inputting the spatio-temporal image intermediate data into the following fourth formula and the fifth formula , wherein, is the infrared observed atmospheric water vapor total information at the same grid position, is the microwave observed atmospheric water vapor total information at the same grid position, is the difference between the infrared observed atmospheric water vapor total information and the microwave observed atmospheric water vapor total information at the same grid position, is the mean value of , is the standard deviation of , is the pixel position corresponding to satisfying the fusion constraint requirement, as the judgment basis of the effective data, the sample spatio-temporal image data of the geographical area to be studied is obtained after screening the independent variable and dependent variable data used for constructing the fusion model.

6. The method of claim 5, wherein, The original spatio-temporal image data and the nonlinear fusion model are used to construct all-weather atmospheric water vapor total amount data of the geographical region to be researched, specifically including: The part of the microwave observation atmospheric water vapor total amount in the second preset date region, which is greater than zero, is taken as an all-weather mask, and the invalid data of the microwave observation precipitation is taken as a non-rainfall mask, and the microwave observation atmospheric water vapor total amount, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, the third brightness temperature, the longitude, the latitude and the Julian date in the second preset date, which satisfy the non-rainfall condition, are input into the sixth formula to obtain atmospheric water vapor total amount in a non-rainfall region. The valid data of the microwave observation precipitation is taken as a rainfall mask, and the microwave observation atmospheric water vapor total amount, the surface elevation data, the normalized vegetation index, the first brightness temperature, the second brightness temperature, the longitude, the latitude and the Julian date in the preset date, which satisfy the rainfall condition, are input into the seventh formula to obtain atmospheric water vapor total amount in a rainfall region. According to the non-rainfall mask and the rainfall mask, the atmospheric water vapor total amount in the non-rainfall region and the atmospheric water vapor total amount in the rainfall region are spliced to obtain all-weather atmospheric water vapor total amount data of the geographical region to be researched.

7. An all-weather high-temporal and spatial resolution multi-source remote sensing atmospheric water vapor total amount fusion data acquisition device, characterized in that, The device comprises: The acquisition module is configured to obtain original spatio-temporal image data of the geographical region to be researched based on infrared observation atmospheric water vapor total amount information, surface elevation data, a normalized vegetation index, microwave observation atmospheric water vapor total amount information, a first brightness temperature, a second brightness temperature, a third brightness temperature and microwave observation precipitation of the geographical region to be researched. The first brightness temperature is an 18GHz brightness temperature, the second brightness temperature is a 23GHz brightness temperature, and the third brightness temperature is a 89GHz brightness temperature. The screening module is configured to perform quality control on the infrared observation atmospheric water vapor total amount information and the microwave observation atmospheric water vapor total amount information at the same grid position in the original spatio-temporal image data, to obtain sample spatio-temporal image data of the geographical region to be researched. The first processing module is configured to take infrared observation atmospheric water vapor total amount information meeting the clear sky condition in the sample spatio-temporal image data as a dependent variable, and take other data meeting the clear sky condition in the sample spatio-temporal image data as an independent variable, and construct a nonlinear fusion model between the dependent variable and the independent variable under the clear sky condition, which is applied to a precipitation area and a non-precipitation area. Specifically, the infrared observed atmospheric water vapor total amount information in the sample spatio-temporal image data meeting the clear sky condition is taken as the dependent variable, the microwave observed atmospheric water vapor total amount information in the sample spatio-temporal image data meeting the clear sky condition is taken as the independent variable, and the nonlinear fusion model is constructed by the sixth formula , the surface elevation data Elevation, the normalized vegetation index NDVI, the first brightness temperature 18GHz BT, the second brightness temperature 23GHz BT, the third brightness temperature 89GHz BT, the longitude Lat, the latitude Lon and the Julian date JD are taken as the independent variables, and the nonlinear fusion model is constructed by the sixth formula and the seventh formula , using the random forest algorithm. The second processing module is configured to divide original spatio-temporal image data into data sets in the precipitation area and the non-precipitation area by taking the microwave observation precipitation data as a precipitation mask, and combine the nonlinear fusion model in the precipitation area and the non-precipitation area to produce high-resolution atmospheric water vapor total amount in the precipitation area and the non-precipitation area, and splice the two by using the precipitation mask to construct all-weather atmospheric water vapor total amount data of the geographic area to be studied.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a computer program, which, when running on a device, causes the device to perform the method of any one of claims 1 to 6.

9. A chip system, characterized by The chip is coupled with a memory, and is configured to execute a computer program stored in the memory to perform the method of any one of claims 1 to 6.

Images