A data correction method for multi-source satellite measurement of soil moisture

Abstract

The present invention discloses a data correction method for multi-source satellite measurement of soil moisture. By linearly fitting the soil moisture data retrieved from satellite remote sensing and the measured soil dielectric constant through mathematical methods, the relationship between the soil dielectric constant and the soil moisture data retrieved from satellite remote sensing is established, and the obtained dielectric constant data is substituted into the volumetric water content-dielectric constant relationship function for calculation, so as to obtain accurate soil water content information. The present invention can perform large-scale and more accurate rapid measurement of the soil in the study area, and helps to promote the application expansion of satellite remote sensing retrieved soil moisture products in industries such as agriculture, meteorology, and water conservancy.

Description

Technical Field

[0001] The present invention relates to the technical field of soil moisture measurement, and more particularly to a data correction method for measuring soil moisture by multi-source satellites. Background Art

[0002] Soil moisture is a key parameter in the land surface water cycle. Traditional soil moisture measurement methods include the weighing method, time domain reflectometry, etc. Although the measurement accuracy is high, they are only suitable for small-scale single-point measurements and cannot meet the large-scale detection of soil moisture. The emergence of remote sensing technology makes it possible to observe soil moisture efficiently, at low cost, and in large areas in real time.

[0003] The invention disclosed in the publication number CN114740022A discloses a soil moisture detection method, device and equipment based on multi-source remote sensing technology, including: obtaining satellite parameters of a target area, obtaining soil backscattering coefficient data of the target area, constructing a soil backscattering coefficient simulation data set according to the satellite parameters of the target area, constructing a soil moisture detection model according to the soil backscattering coefficient simulation data set, and inputting the satellite parameters and soil backscattering coefficient data of the target area into the soil moisture detection model to obtain the soil moisture data of the target area. Compared with the prior art, the invention can more comprehensively, accurately and effectively realize the inversion of soil moisture in high vegetation coverage target areas. However, the invention does not consider the influence of environmental parameters on soil moisture content, nor does it consider the matching problem between different satellite remote sensing images and the study area. Summary of the Invention

[0004] Aiming at the deficiencies in the prior art, the present invention provides a data correction method for measuring soil moisture by multi-source satellites, adaptively selects the optimal satellite remote sensing data in combination with environmental parameters and the regional characteristics of the study area, and then respectively establishes relationship models between the soil dielectric constant and the soil inversion data retrieved by satellite remote sensing, and between the soil dielectric constant and the actual soil moisture content data, so as to be able to perform large-scale, more accurate and rapid measurement of the soil in the study area, and contribute to promoting the application expansion of satellite remote sensing retrieved soil moisture products in industries such as agriculture, meteorology, and water conservancy.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A data correction method for measuring soil moisture by multi-source satellites, the data correction method comprising the following steps:

[0007] Collect remote sensing data of n satellites under different environmental parameters at different time periods, and after preprocessing, respectively retrieve the soil moisture inversion data X of different observation points in the study area without correction i,j, where \(i = 1, 2, \ldots, n\); meanwhile, the true soil moisture data \(Y\) of different observation points in the study area is obtained by the artificial drying method. j , the dielectric constant of the soil at different observation points in the study area is measured by a TDR instrument. \(m\) is the number of observation points; according to the true soil moisture data \(Y\). j the volumetric water content of the soil at different observation points in the study area is calculated.

[0008] S2, combining the multiple groups of true soil moisture amounts \(Y\) obtained in step S1. j and the dielectric constant of the soil moisture. the linear relationship function formula between the true soil moisture content \(Y\) and the square root of the dielectric constant at different observation points within the study area is fitted as follows: j and the dielectric constant. is: where \(a\) and \(b\) are dimensionless influence parameters.

[0009] S3, an empirical relationship model between the volumetric water content of the soil at different observation points in the study area. and the dielectric constant of the soil. is constructed as follows:

[0010] S3, combining the soil moisture inversion data \(X\) under different environmental parameters at different time periods obtained in step S1. i,j and the true soil moisture data \(Y\). j , a soil moisture correction model for different satellites is constructed: \(Y\). j \(= c\). i \(X\). i,j 3 \(+ d\). i \(X\). i,j 2 \(+ e\). i \(X\). i,j \(+ f\). i , where \(c\). i , \(d\). i , \(e\). i , \(f\). i are dimensionless influence parameters.

[0011] S4, combining the linear relationship function formula in step S2 and the soil moisture correction model in step S3, the calculation formula for the soil volume water content corresponding to different satellites is obtained:

[0012] S5, collect the remote sensing data of \(n\) satellites at the current time period \(t\), combine the environmental parameters of the previous \(K\) time periods with the current time period \(t\) as the end point, select the optimal satellite remote sensing data, and after preprocessing, invert the soil moisture data \(X\) within the study area.i,j (t), substitute the soil moisture data X i,j (t) into the calculation formula of satellite remote sensing retrieved soil moisture data for the corresponding satellite, and calculate the corrected volumetric soil water content θ i,j (t) within the study area for the current time period t.

[0013] To optimize the above technical solution, the specific measures taken also include:

[0014] Furthermore, in step S1, the remote sensing data of the satellite includes remote sensing image data within the study area collected by Sentinel-2 and Landsat 8.

[0015] Furthermore, in step S1, the preprocessing process includes the following steps:

[0016] Perform radiometric calibration and atmospheric correction on the remote sensing data of Sentinel-2, convert the remotely sensed data after atmospheric correction into corresponding remote sensing image data, and then uniformly set the spatial resolution of each band;

[0017] Perform atmospheric correction on the remote sensing data used by Landsat 8.

[0018] Furthermore, in step S3, the process of constructing the empirical relationship model between the volumetric water content of soil moisture at different observation points within the study area and the dielectric constant of the soil includes the following steps:

[0019] S31, select multiple empirical relationship models between volumetric water content and dielectric constant;

[0020] S32, for each observation point, substitute the volumetric water content of soil moisture at this observation point collected in step S1 and the dielectric constant of the soil into multiple empirical relationship models respectively; calculate the root mean square error corresponding to each model, and select the empirical relationship model with the smallest root mean square error as the empirical relationship model between the volumetric water content of soil moisture at this observation point and the dielectric constant of the soil between.

[0021] Furthermore, the environmental parameters include the evaporation, rainfall, and vegetation density of each observation station in the study area.

[0022] Furthermore, in step S5, the process of selecting the optimal satellite remote sensing data by combining the environmental parameters of the previous K time periods with the current time period as the end point includes the following steps:

[0023] S51, for each soil moisture inversion data X at different observation points in the same time period i,j, the soil volume water content θ at different observation points is calculated by using the soil volume water content calculation formula corresponding to the respective satellite. i,j , calculate the difference between the calculated volume water content of the soil moisture at different observation points and the measured one, and calibrate the accuracy level of the soil moisture inversion data X according to the difference; i,j

[0024] S52, count the accuracy levels of the soil moisture inversion data corresponding to all the observation points in the remote sensing images of different satellites in the same time period, calculate the matching level of the remote sensing image, select the remote sensing image of the satellite with the highest matching level as the positive sample image, and select the remote sensing image of the satellite with the lowest matching level as the negative sample image;

[0025] S53, obtain the environmental parameter sequences of the first K time periods with the remote sensing image acquisition time period as the end point corresponding to all the positive sample images and negative sample images, generate training positive samples and training negative samples, and construct a training data set;

[0026] S54, construct a remote sensing satellite matching model based on a neural network, with the input being the environmental parameter sequences of the first K time periods with the current time period as the end point and the output being the satellite model corresponding to the positive sample image, and import the training data set to train the remote sensing satellite matching model;

[0027] S55, import the environmental parameter sequences of the first K time periods with the current time period as the end point into the remote sensing satellite matching model, and select the optimal satellite remote sensing data corresponding to the current time period.

[0028] The beneficial effects of the present invention are as follows:

[0029] First, for the multi-source satellite soil moisture measurement data correction method of the present invention, relationship models are respectively established between the soil dielectric constant and the soil inversion data retrieved by satellite remote sensing, and between the soil dielectric constant and the actual soil water content data, so that large-scale and more accurate rapid measurement of the soil in the study area can be carried out, which helps to promote the application expansion of the satellite remote sensing retrieved soil moisture products in industries such as agriculture, meteorology, and water conservancy.

[0030] Second, for the multi-source satellite soil moisture measurement data correction method of the present invention, based on the neural network model, the optimal satellite remote sensing data can be adaptively selected in combination with the environmental parameters and the regional characteristics of the study area, further improving the accuracy of the measurement results. Description of the Drawings

[0031] Figure 1 is the flowchart of the multi-source satellite soil moisture measurement data correction method according to the embodiment of the present invention.

[0032] Figure 2Schematic diagram of the pre - processed Sentinel - 2 remote sensing image data in the embodiments of the present invention.

[0033] Figure 3a Schematic diagram of the pre - processed Landsat 8 remote sensing image data in the embodiments of the present invention.

[0034] Figure 3b Schematic diagram of the pre - processed Landsat 8 remote sensing image data in the embodiments of the present invention.

[0035] Figure 4 Schematic diagram of the position distribution of the soil sampling points (observation points) in the embodiments of the present invention.

[0036] Figure 5 Schematic diagram of the soil water content prediction results in the examples. Detailed implementation manners

[0037] Now, the present invention will be further described in detail with reference to the accompanying drawings.

[0038] It should be noted that the terms such as "upper", "lower", "left", "right", "front", "rear", etc. cited in the invention are only for the convenience of clear narration, rather than to limit the scope of implementation of the present invention. The change or adjustment of their relative relationship, without substantial change in the technical content, should also be regarded as the scope of implementation of the present invention.

[0039] See Figure 1 , this embodiment discloses a data correction method for multi - source satellite measurement of soil moisture. The data correction method includes the following steps:

[0040] S1. Under different environmental parameters at different times, collect the remote sensing data of n satellites. After pre - processing, respectively invert to obtain the uncorrected soil moisture inversion data X of different observation points in the study area, where i = 1, 2, …, n; at the same time, use the artificial drying method to obtain the true soil moisture data Y of different observation points in the study area, and use the TDR instrument to measure the dielectric constant of the soil of different observation points in the study area. m is the number of observation points; according to the true soil moisture data Y, calculate the volumetric water content of the soil of different observation points in the study area. i,j , i = 1, 2, …, n; at the same time, use the artificial drying method to obtain the true soil moisture data Y of different observation points in the study area, and use the TDR instrument to measure the dielectric constant of the soil of different observation points in the study area. j , use the TDR instrument to measure the dielectric constant of the soil of different observation points in the study area. m is the number of observation points; according to the true soil moisture data Y j calculate the volumetric water content of the soil of different observation points in the study area.

[0041] S2. Combine the multiple groups of true soil moisture amounts Y obtained in step S1 j and the dielectric constant of the soil moisture to fit and obtain the true soil water content Y of different observation points within the study area j and the dielectric constant The formula of the linear relationship function between the square roots: Where a and b are dimensionless influence parameters.

[0042] S3. Construct the volumetric water content of soil moisture at different observation points within the study area and the empirical relationship model between the dielectric constant of the soil :

[0043] S3. Combine the soil moisture inversion data X i,j under different environmental parameters at different time periods obtained in step S1 j and the true soil moisture data Y j to construct the soil moisture correction model for different satellites: Y i = c i,j X 3 i X i,j 2 + e i X i,j + f i . Where c i , d i , e i , f i are dimensionless influence parameters.

[0044] S4. Combine the linear relationship function formula in step S2 and the soil moisture correction model in step S3 to obtain the calculation formula for the soil volume water content corresponding to different satellites:

[0045] S5. Collect the remote sensing data of n satellites at the current time period t, combine the environmental parameters of the previous K time periods with the current time period t as the end point, select the optimal satellite remote sensing data, and after preprocessing, invert the soil moisture data X i,j (t) within the study area. Substitute the soil moisture data X i,j (t) into the calculation formula for the satellite remote sensing inverted soil moisture data of the corresponding satellite to calculate the corrected soil volume water content θ i,j (t) within the study area at the current time period t.

[0046] This application does not limit the number and type of satellites. As long as the remote sensing data is collected by satellites of the same type, it can be applied to this application. Since this application includes an optimal satellite remote sensing data matching mechanism, even if satellites with poor data effects are selected, it will have no impact on the final soil water content measurement results. However, due to the different remote sensing data parameters and formats of different satellites, it is still necessary to preprocess the remote sensing data of some satellites. In this embodiment, the selected satellites include Sentinel-2 and Landsat 8.

[0047] For the remote sensing data of Sentinel-2, a scene of Sentinel-2 L1C data of the study area was downloaded. Since L1C-level products are atmospheric apparent reflectance products that have undergone orthorectification and geometric precision correction, and no atmospheric correction has been performed, they are subjected to radiation calibration and atmospheric correction to obtain L2A-level products in order to obtain base reflectance data.

[0048] a. Atmospheric correction processing, the processing software is sen2cor, and the data of all resolutions is selected.

[0049] b. Since the processed image format cannot be used by software such as ENVI, SNAP is required for format conversion. In addition, the spatial resolution of each band of Sentinel-2 images is inconsistent, divided into 10 meters, 20 meters and 60 meters. In this study, they were unified to 10 meters resolution. The results are as follows Figure 2 shown.

[0050] For Landsat 8 remote sensing data, a Landsat 8 remote sensing image of the study area was collected, with the transit time being March 26, 2021, and atmospheric correction was performed on it to obtain reflectance data. Figure 3a and Figure 3b The following are schematic diagrams of Landsat 8 remote sensing image data before and after preprocessing. Figure 3a and Figure 3b The text in has no effect on the description of the solution of this embodiment.

[0051] In this embodiment, several observation points are set up along the Jinchuan River Basin. Figure 4 This is a schematic diagram of the location distribution of soil sampling points (observation points) in an embodiment of the present invention. Locations around the river are selected because the soil around the river reduces the influence of weather factors on the moisture content, and the consistency of the flow rate of the same river is better. Specifically, 17 soil samples were collected and their mass moisture content was measured. The soil mass moisture content data is shown in Table 1.

[0052] Table 1

[0053] Serial number Latitude Longitude Water content 1 32.07333 118.7381 0.33722 2 32.08806 118.7403 0.307872 3 32.09694 118.7386 0.268293 4 32.09833 118.7458 0.375456 5 32.08944 118.7344 0.200853 6 32.09 118.7586 0.20478 7 32.09278 118.7408 0.19387 8 32.09194 118.7656 0.267457 9 32.09056 118.775 0.303362 10 32.09139 118.7839 0.240568 11 32.07972 118.7767 0.2678 12 32.08306 118.7681 0.223281 13 32.08306 118.7622 0.231327 14 32.09611 118.7556 0.220865 15 32.10083 118.7522 0.262826 16 32.10806 118.7489 0.237959

[0054] In step S3, the volumetric moisture content of soil moisture at different observation points in the study area is constructed. Dielectric constant of soil The process of modeling the empirical relationship between the two consists of the following steps:

[0055] S31, selecting multiple empirical relationship models between volume water content and dielectric constant;

[0056] S32. For each observation point, substitute the volumetric water content of the soil moisture and the dielectric constant of the soil collected at this observation point in step S1 into multiple empirical relationship models respectively; calculate the root mean square error corresponding to each model, and select the empirical relationship model with the minimum root mean square error as the empirical relationship model between the volumetric water content of the soil moisture and the dielectric constant of the soil at this observation point. and the dielectric constant of the soil respectively into multiple empirical relationship models; calculate the root mean square error corresponding to each model, and select the empirical relationship model with the minimum root mean square error as the empirical relationship model between the volumetric water content of the soil moisture and the dielectric constant of the soil at this observation point. and the dielectric constant of the soil at this observation point.

[0057] The empirical relationship model between the volumetric water content and the dielectric constant is greatly affected by soil texture, etc. Commonly used empirical relationship models currently include the Topp formula, the Herkelrath fitting relationship, etc. Considering the generality of the data correction method, in this embodiment, the most suitable empirical relationship model is selected for different regions of the study area respectively.

[0058] In order to obtain more accurate measurement results, this embodiment also proposes an optimization strategy for selecting remote sensing images in combination with environmental parameters. In this embodiment, the environmental parameters mainly include the evaporation, rainfall and vegetation density of each observation site in the study area. For some special scenarios, other environmental parameters can also be introduced, such as air temperature, etc. The change of soil water content is positively correlated with local precipitation and negatively correlated with evaporation. Since the remote sensing data of bare soil is greatly affected by trees and forests, the influence of interference factors such as trees and forests needs to be fully considered in remote sensing data analysis. In some cases, this part of remote sensing data also needs to be excluded and corrected.

[0059] Exemplarily, in step S5, the process of selecting the optimal satellite remote sensing data in combination with the environmental parameters of the previous K time periods with the current time period as the end point includes the following steps:

[0060] S51. For each soil moisture inversion data X of different observation points in the same time period i,j , calculate the soil volume water content θ of different observation points by using the soil volume water content calculation formula corresponding to the corresponding satellite, calculate the difference between it and the measured volumetric water content of the soil moisture of different observation points i,j , and calibrate the accuracy level of the soil moisture inversion data X according to the difference. For example, assuming that the data of three remote sensing satellites are used, three accuracy levels can be set, and the accuracy of the image data of the three remote sensing satellites can be judged for each observation point. at different observation points, and calibrate the accuracy level of the soil moisture inversion data X according to the difference. For example, assuming that the data of three remote sensing satellites are used, three accuracy levels can be set, and the accuracy of the image data of the three remote sensing satellites can be judged for each observation point. i,j For example, assuming that the data of three remote sensing satellites are used, three accuracy levels can be set, and the accuracy of the image data of the three remote sensing satellites can be judged for each observation point.

[0061] S52. Statistically analyze the accuracy levels of the soil moisture inversion data for all observation points corresponding to the remote sensing images of different satellites in the same time period, calculate the matching degree level of the remote sensing image, select the remote sensing image of the satellite with the highest matching degree level as the positive sample image, and select the remote sensing image of the satellite with the lowest matching degree level as the negative sample image.

[0062] Considering the computational complexity, in this embodiment, only one remote sensing image is selected for inversion. Therefore, it is necessary to comprehensively consider the accuracy levels of each observation point to obtain the matching degree level of the remote sensing image. Since the number of observation points is fixed, for a remote sensing image, the higher the accuracy level of the observation points, the more the number of observation points with high accuracy, and the higher the matching degree level of the remote sensing image.

[0063] S53. Obtain the environmental parameter sequences of the first K time periods ending with the remote sensing image acquisition time period corresponding to all positive sample images and negative sample images, generate training positive samples and training negative samples, and construct a training data set.

[0064] The change of soil water content has temporal characteristics. Therefore, in this embodiment, the environmental parameter sequences of the first K time periods ending with the remote sensing image acquisition time period are selected as the input data. Assuming that one time period is one day, then the corresponding environmental parameter sequences can be selected by backward deduction of 3 - 5 days from the current time as the starting point.

[0065] S54. Based on a neural network, construct a remote sensing satellite matching model. The input is the environmental parameter sequences of the first K time periods ending with the current time period, and the output is the satellite model corresponding to the positive sample image. Import the training data set to train the remote sensing satellite matching model. In this application, the remote sensing satellite matching model can be a support vector machine model, etc.

[0066] S55. Import the environmental parameter sequences of the first K time periods ending with the current time period into the remote sensing satellite matching model, and select the optimal satellite remote sensing data corresponding to the current time period.

[0067] Exemplarily, meteorological conditions have a great impact on the quality of remote sensing images, and the acquisition cycles of different remote sensing satellites are also different. Some satellites acquire remote sensing images once a day, while some satellites can acquire remote sensing images more than a dozen times a day. The latter can integrate and analyze the multiple acquired remote sensing images and delete the remote sensing images with poor quality. However, this does not mean that remote sensing images with good image quality must be used. This is because meteorological conditions are also one of the crucial environmental parameters in soil moisture measurement, and the degree of influence of meteorological conditions on remote sensing images can also be used as a reference factor in soil moisture inversion. For example, although the quality of remote sensing images during rainfall will decline, the impact of rainfall on soil moisture is also quite significant.

[0068] Therefore, in some embodiments, we further studied the degree of influence of environmental parameters on remote sensing images of different satellites, and there is also an associated influence on soil moisture content among environmental parameters. Generally speaking, the vegetation density in a region is relatively stable. When the weather is clear, the evaporation amounts of water in areas with different vegetation densities are different. When it rains, the water retention ability of areas with high vegetation density is stronger. For this reason, we analyzed environmental parameters as a whole.

[0069] Finally, this embodiment improves the process of selecting the optimal satellite remote sensing data to:

[0070] For the acquisition cycles of different remote sensing satellites, a unified image acquisition sub-period is selected; when the meteorological conditions are appropriate within the image acquisition sub-period, the remote sensing image quality of all or most remote sensing satellites meets the requirements of conventional analysis, then the remote sensing images within the image acquisition sub-period are used; when the meteorological conditions within the image acquisition sub-period are not conducive to acquiring clear remote sensing images, for remote sensing satellites with long acquisition cycles, the remote sensing images within the image acquisition sub-period are still selected. For remote sensing satellites with short acquisition cycles, an allowable prediction duration threshold is set according to the environmental parameter sequence, and remote sensing images with higher image quality are selected within the allowable prediction duration threshold. If the meteorological conditions within the allowable prediction duration threshold do not meet the requirements, the remote sensing images within the image acquisition sub-period are still selected.

[0071] The process of setting the allowable prediction duration threshold according to the environmental parameter sequence includes the following steps:

[0072] For any remote sensing satellite with a short acquisition cycle, a certain amount of analysis periods are selected. The analysis periods are within the allowable range of the acquisition task period. For example, if the acquisition task is to measure the soil moisture content in the first ten days of a certain month, the allowable range of the acquisition task period is from the 1st to the 10th of that month. The unified image acquisition sub-period can be set as a certain day among them according to the type of the selected remote sensing satellite. Multiple remote sensing images at different observation points within different analysis periods and the environmental parameter time series data within the corresponding analysis periods are retrieved, and the soil moisture content at different observation points on each remote sensing image is inversely obtained. The difference between the actual soil moisture content and the inversely obtained soil moisture content at different observation points is calculated, and the root mean square error between the actual soil moisture content and the inversely obtained soil moisture content at all observation points is used as the comprehensive error of the inversion data of the remote sensing image. Each environmental parameter time series data is used as an input value, and the acquisition time of the remote sensing image with the smallest root mean square error value is selected as its output value to generate training samples.

[0073] Construct a threshold analysis model based on a neural network and train it using a training sample set. The threshold analysis model is used to process the time-series data of environmental parameters collected in real time, predict the optimal acquisition time for different remote sensing satellites, and select the duration from the minimum value of the optimal acquisition time to the current image acquisition sub-period as the allowable prediction duration threshold.

[0074] Example: According to the results of the Nanjing Meteorological Observatory, the daily evaporation in Nanjing from 8:00 AM on March 23, 2021 to 8:00 AM on March 24, 2021 was 2.6 mm, and there was no rainfall on that day. Assuming that the evaporation is related to the soil mass at a depth of 1 meter, and it is also known that the bulk density of the soil in Nanjing is approximately 1.3 g / cm3, and its reduction coefficient per square meter is 2.6×103 g. The soil moisture content forecast results for March 24 are as Figure 5 shown.

[0075] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A data correction method for multi-source satellite measurement of soil moisture, characterized in that, The data correction method includes the following steps: S1. At different time periods and under different environmental parameters, remotely sensed data of n satellites are collected. After preprocessing, the soil moisture inversion data X of different observation points in the study area are respectively retrieved without correction. i,j , where i = 1, 2, …, n; meanwhile, the true soil moisture data Y of different observation points in the study area are obtained by the artificial drying method. j , and the dielectric constant of the soil at different observation points in the study area is measured by a TDR instrument. , where j = 1, 2, …, m; m is the number of observation points; according to the true soil moisture data Y j , the volumetric water content of the soil at different observation points in the study area is calculated. S2. Combine the multiple groups of true soil moisture amounts Y obtained in step S1 j and the dielectric constant of the soil moisture to fit and obtain the true soil water content Y at different observation points within the study area j and the square root of the dielectric constant to obtain the linear relationship function formula: where a and b are dimensionless influence parameters; S3. Construct the volumetric water content of soil moisture at different observation points within the study area and the dielectric constant of the soil to obtain an empirical relationship model between them: S4. Combine the soil moisture inversion data X under different environmental parameters at different time periods obtained in step S1 i,j and the true soil moisture data Y j to construct a soil moisture correction model for different satellites: Y j = c i X i,j 3 + d i X i,j 2 + e i X i,j + f i , where c i , d i , e i , f i are dimensionless influence parameters; S5. Combining the linear relationship function formula in step S2 and the soil moisture correction model in step S3, the calculation formula for soil volume water content corresponding to different satellites is obtained: S6. Collect the remote sensing data of n satellites in the current time period t, combine the environmental parameters of the previous K time periods with the current time period t as the end point, select the optimal satellite remote sensing data, and after preprocessing, invert to obtain the soil moisture data X within the study area. i,j (t), substitute the soil moisture data X i,j (t) into the calculation formula of the satellite remote sensing inversion soil moisture data of the corresponding satellite, and calculate the corrected soil volume water content θ i,j (t) within the study area in the current time period t; The environmental parameters include the evaporation, rainfall, and vegetation density of each observation site in the study area; In step S6, the process of selecting the optimal satellite remote sensing data by combining the environmental parameters of the previous K time periods with the current time period as the end point includes the following steps: S51. For each soil moisture inversion data X at different observation points in the same time period i,j , the soil volume water content θ of different observation points is calculated by using the soil volume water content calculation formula corresponding to the corresponding satellite i,j . Calculate the difference between the volume moisture content of the soil moisture measured at different observation points , and calibrate the accuracy level of the soil moisture inversion data X according to the difference i,j ; S52. Statistically analyze the accuracy levels of the soil moisture inversion data of all observation points corresponding to the remote sensing images of different satellites in the same time period, calculate the matching degree level of the remote sensing image, select the remote sensing image of the satellite with the highest matching degree level as the positive sample image, and select the remote sensing image of the satellite with the lowest matching degree level as the negative sample image; S53. Obtain the environmental parameter sequences of the previous K time periods with the remote sensing image acquisition time period as the end point corresponding to all positive sample images and negative sample images, generate training positive samples and training negative samples, and construct a training data set; S54. Based on a neural network, construct a remote sensing satellite matching model. The input is the environmental parameter sequence of the previous K time periods with the current time period as the end point, and the output is the satellite model corresponding to the positive sample image. Import the training data set to train the remote sensing satellite matching model; S55. Import the environmental parameter sequence of the previous K time periods with the current time period as the end point into the remote sensing satellite matching model, and select the optimal satellite remote sensing data corresponding to the current time period.

2. The data correction method for multi-source satellite measurement of soil moisture according to claim 1, wherein In step S1, the remote sensing data of the satellite includes the remote sensing image data of the study area collected by Sentinel-2 and Landsat 8.

3. The data correction method for multi-source satellite measurement of soil moisture according to claim 2, wherein In step S1, the preprocessing process includes the following steps: Perform radiometric calibration and atmospheric correction on the remote sensing data of Sentinel-2, convert the remotely sensed data after atmospheric correction into corresponding remote sensing image data, and then uniformly set the spatial resolution of each band; Perform atmospheric correction on the remote sensing data used by Landsat 8.

4. The data correction method for multi-source satellite measurement of soil moisture according to claim 1, characterized in that In step S3, a volumetric water content of soil moisture at different observation points within the study area is constructed and the dielectric constant of the soil The process of establishing an empirical relationship model therebetween includes the following steps: S31. Select multiple empirical relationship models between volumetric water content and dielectric constant; S32. For each observation point, substitute the volumetric water content of the soil moisture and the dielectric constant of the soil collected at this observation point in step S1 into multiple empirical relationship models respectively; calculate the root mean square error corresponding to each model, and select the empirical relationship model with the smallest root mean square error as the empirical relationship model between the volumetric water content of the soil moisture and the dielectric constant of the soil at this observation point. and the dielectric constant of the soil respectively into multiple empirical relationship models; calculate the root mean square error corresponding to each model, and select the empirical relationship model with the smallest root mean square error as the volumetric water content of the soil moisture at this observation point and the dielectric constant of the soil between the empirical relationship models.

5. The data correction method for multi-source satellite measurement of soil moisture according to claim 1, wherein For the acquisition cycles of different remote sensing satellites, select a unified image acquisition sub-time period; when the meteorological conditions within the image acquisition sub-time period make the quality of the remote sensing image meet the requirements of conventional analysis, use the remote sensing image within the image acquisition sub-time period; when the meteorological conditions within the image acquisition sub-time period make the quality of the remote sensing image not meet the requirements of conventional analysis, for remote sensing satellites with a long acquisition cycle, select the remote sensing image within the image acquisition sub-time period, and for remote sensing satellites with a short acquisition cycle, set an allowable pre-estimation duration threshold according to the environmental parameter sequence, and select a remote sensing image with higher image quality within the allowable pre-estimation duration threshold; If the meteorological conditions within the allowable pre-estimation duration threshold do not meet the requirements, select the remote sensing image within the image acquisition sub-time period; A short acquisition cycle means that the number of remote sensing image acquisitions of the remote sensing satellite within the allowable pre-estimation duration threshold is greater than or equal to 2.

6. The data correction method for multi-source satellite measurement of soil moisture according to claim 5, wherein The process of setting the allowable pre-estimation duration threshold according to the environmental parameter sequence includes the following steps: For any remote sensing satellite with a short acquisition cycle, select a certain amount of analysis time periods, and the analysis time periods are within the allowable range of the acquisition task time period; Retrieve multiple remote sensing images at different analysis time periods for different observation points and the time series data of environmental parameters during the corresponding analysis time periods, invert the soil water content at different observation points on each remote sensing image, calculate the difference between the actual soil water content and the inverted soil water content at different observation points, and take the root mean square error of the actual soil water content and the inverted soil water content at all observation points as the comprehensive error of the inversion data of the remote sensing image; Take each time series data of environmental parameters as the input value, and select the acquisition time of the remote sensing image with the smallest root mean square error as its output value to generate training samples; Construct a threshold analysis model based on a neural network, and use the training sample set for training. The threshold analysis model is used to process the time series data of environmental parameters collected in real time, predict the best acquisition time corresponding to different remote sensing satellites, and select the duration from the minimum value of the best acquisition time to the current image acquisition sub-period as the allowable prediction duration threshold.

Images