A rapid mapping method for soil organic matter distribution with 16m resolution using Gaofen-1 satellite

Abstract

The method for rapid mapping of soil organic matter distribution with 16m resolution of Gaofen-1 satellite belongs to the field of satellite remote sensing applications, and specifically relates to a method for rapid mapping of soil organic matter distribution. The present invention is to solve the problem of declining soil fertility and deterioration of cultivated land quality in farmland. The method: obtains remote sensing reflectances of different bands of Gaofen-1 satellite WFV sensor in the target study area; uses remote sensing reflectances of different bands of WFV sensor images and remote sensing reflectance band combinations as input variables of the random forest algorithm, and outputs organic matter (mg / g) in the surface soil of cultivated land, so as to realize pH mapping of the surface soil of cultivated land. Based on the domestic Gaofen-1 satellite WFV sensor, the method of the present invention can realize dynamic monitoring of organic matter in the surface soil of cultivated land. It is verified by actual sample points that the method of the present invention can accurately estimate the organic matter in the surface soil of cultivated land, which is helpful for soil fertility monitoring, precision agriculture and sustainable land use management.

Description

Technical Field

[0001] This invention belongs to the field of satellite remote sensing applications, specifically relating to a rapid mapping method for soil organic matter distribution. Background Technology

[0002] Soil organic matter is a crucial indicator for assessing soil degradation, health, and fertility, and a key factor influencing crop yield. Therefore, rapid and accurate estimation of soil organic matter is urgently needed for smart agriculture and sustainable agricultural land use. Traditional spatial mapping methods, such as the Kriging method, rely on field sampling and testing, often focusing on small-scale, fragmented mapping of cultivated land. This is often unsuitable for large-scale remote sensing mapping because the impact of soil organic matter on agricultural land is influenced by multiple factors, including topography, parent material, climate, vegetation, and agronomic practices, resulting in significant spatial heterogeneity. With the development of Earth observation technology and innovation in artificial intelligence algorithms, satellite remote sensing combined with big data analysis has become an important method for quantitative monitoring and mapping of surface environmental parameters on large scales. Previous digital soil mapping studies have primarily used Landsat and Sentinel-2 satellites, with less emphasis on digital soil mapping based on the domestically developed Gaofen-1 satellite.

[0003] Satellite remote sensing mapping of soil organic matter primarily relies on spectral techniques to analyze the differences in the physicochemical properties of various land features. These specific spectral differences are mainly reflected in the variations in remote sensing reflectance across different spectral bands. Soil organic matter exhibits reflectance spectral characteristics in the visible and near-infrared wavelength ranges, and as soil organic matter content increases, soil reflectance decreases, leading to changes in the shape of the spectral curve. Therefore, in large-scale regional soil organic matter remote sensing mapping, it is necessary to construct an inversion algorithm that combines environmental variables and spectral indices, taking into account changing geographical environments, to improve the accuracy of large-scale regional soil organic matter remote sensing mapping. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problems of declining soil fertility and deterioration of farmland quality, and to provide a rapid mapping method for soil organic matter distribution at a resolution of 16m using the Gaofen-1 satellite.

[0005] The rapid mapping method for soil organic matter distribution at 16m resolution using the Gaofen-1 satellite is as follows:

[0006] Step 1: Obtain raw data

[0007] Acquire WFV sensor image data of the Gaofen-1 satellite (GF-1) in the target study area;

[0008] Step 2: Image Data Preprocessing

[0009] The raw WFV sensor image data of the Gaofen-1 satellite were preprocessed to obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0010] Step 3: Environmental Covariate Data Processing

[0011] Environmental data download and processing: acquire 30m topographic data and climate information. The topographic data is elevation, and the climate information is free raster data of annual average rainfall 1km and annual average temperature 1km. All data are resampled to a spatial resolution of 16m, consistent with the spatial resolution data of WFV sensor imagery.

[0012] The annual average rainfall 1km and annual average temperature 1km free raster data were downloaded from the National Meteorological Center (http: / / data.cma.cn / ), and the 30m topographic data were obtained from USGS (https: / / gdex.cr.usgs.gov / gdex / ), with the elevation extracted in ArcGIS software;

[0013] Step 4: Random Forest Algorithm Inversion

[0014] The input variables for the random forest algorithm are: annual average rainfall, annual average temperature, elevation, spectral index RI2, spectral index RI3, spectral index DI2, spectral index DI3, spectral index NDVI, spectral index RVI, spectral index DVI, spectral index NDI1, spectral index NDI2, spectral index NDI3, spectral index RDVI1, spectral index RDVI2, RDVI3, spectral index GDVI1, and spectral index GDVI3.

[0015] Random Forest algorithm output variable: Organic matter in topsoil of cultivated land, unit mg / kg;

[0016] The spectral index:

[0017] RI2 = Band3 / Band2

[0018] RI3 = Band4 / Band1

[0019] DI2 = Band3 - Band2

[0020] DI3 = Band4 - Band1

[0021] NDVI=(Band4-Band3) / (Band4+Band3)

[0022] RVI = Band4 / Band3

[0023] DVI = Band4 - Band3

[0024] NDI1=(Band3-Band2) / (Band3+Band2)

[0025] NDI2=(Band4-Band1) / (Band4+Band1)

[0026] NDI3=(Band4-Band2) / (Band4+Band2)

[0027]

[0028]

[0029]

[0030]

[0031]

[0032] Band1 represents the remote sensing reflectance of Band 1 of the WFV sensor on the Gaofen-1 satellite;

[0033] Band2 represents the remote sensing reflectance of Band 2 of the WFV sensor on the Gaofen-1 satellite.

[0034] Band3 represents the remote sensing reflectance of Band 3 of the WFV sensor on the Gaofen-1 satellite.

[0035] Band4 represents the remote sensing reflectance of Band 4 of the WFV sensor on the Gaofen-1 satellite.

[0036] Step 5: Remote sensing mapping of soil organic matter. The prediction model is substituted into the WFV sensor image of Gaofen-1 satellite (GF-1) to achieve rapid mapping of regional soil organic matter distribution.

[0037] The preprocessing described in step two includes radiometric calibration, orthorectification, and atmospheric correction.

[0038] The input variables for the random forest algorithm described in step four are constructed based on the measured soil organic matter inversion model. The construction process of the soil organic matter random forest inversion model is as follows:

[0039] a. During the bare soil period in Northeast China (end of April), a total of 482 soil samples with a surface depth of 0-20 cm were collected from different topographic conditions and soil types. The GPS location of the samples was recorded. The collected soil samples were air-dried, ground and sieved. The soil organic matter content was obtained using the potassium dichromate heating method.

[0040] b. Acquire WFV sensor images from the Gaofen-1 satellite (GF-1). After the image data preprocessing in step two, extract the remote sensing reflectance corresponding to the measured soil sample based on the GPS location, and obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0041] c. After step three, obtain environmental covariate data:

[0042] d. Based on measured soil organic matter data and remote sensing reflectance of different bands from Gaofen-1 satellite, a random forest inversion model for soil organic matter was established:

[0043] Randomly selected measured soil organic matter data and corresponding remote sensing reflectance from different bands of the Gaofen-1 satellite were used as the calibration dataset for the model. The linear correlation between remote sensing reflectance in different bands, spectral indices calculated from these indices, environmental covariates, and measured soil organic matter data was analyzed. Variables associated with measured soil organic matter were selected using a significance level of p < 0.05 and a two-sided confidence level of 0.01. These variables included annual average rainfall, annual average temperature, elevation, and spectral indices RI2, RI3, DI2, DI3, NDVI, RVI, DVI, NDI1, NDI2, NDI3, RDVI1, RDVI2, RDVI3, GDVI1, and GDVI3. These selected variables were then used as input variables for a random forest algorithm to establish a remote sensing inversion model for soil organic matter.

[0044] The analysis of linear correlation was performed using SPSS 16.0 software. The Pearson correlation coefficient r is expressed as follows:

[0045]

[0046] Where, x i This is the measured value of soil organic matter, y i It is the remote sensing reflectance of Gaofen-1 satellite in different bands that corresponds to the measured value of soil organic matter, and N is the number of samples;

[0047] e. The soil organic matter remote sensing inversion model is validated for error. The expression for error validation is as follows:

[0048]

[0049] Where RMSE is the root mean square error, N is the number of samples, and y i This is the measured value of soil organic matter, y i 'This is an estimated value of soil organic matter;

[0050] f. Algorithm verification of linear fit of the dataset R 2=0.92, RMSE is 4.24g / kg.

[0051] The random forest algorithm described in step four uses the RandomForest package. The number of decision trees, ntree, is 500, and the number of random variables, mtry, for splitting nodes is 1 / 3 of the input. At this point, the values ​​are relatively small and the errors within the model are basically stable.

[0052] Based on the rapid mapping method for soil organic matter distribution using the 16m resolution of the Gaofen-1 satellite, this invention enables large-scale estimation and spatial mapping of surface soil organic matter in cultivated land, accurately obtaining the spatial distribution of interannual variations in soil organic matter content. Overall, the random forest algorithm for soil organic matter constructed in this invention was trained using 482 measured soil organic matter values ​​covering Northeast China, determining the algorithm's input variables. This algorithm effectively represents variations in different geographical environments, soil types, and elevations. The soil organic matter inversion model developed using this invention has good applicability, reducing the time and cost of on-site sampling compared to traditional methods, and effectively assessing the performance of soil fertility and cultivated land fertility improvement.

[0053] Based on the WFV sensor of the domestically produced Gaofen-1 satellite, the method of this invention can realize dynamic monitoring of organic matter in the topsoil of cultivated land. Verification by actual measurement points shows that the method of this invention can accurately estimate the organic matter in the topsoil of cultivated land, which is helpful for soil fertility monitoring, precision agriculture and sustainable land use management, ensuring high-precision monitoring of soil organic matter content and expanding the application of the domestically produced Gaofen series satellites in agricultural management. Attached Figure Description

[0054] Figure 1 This is the calibration diagram of the soil organic matter inversion model in Northeast China in Experiment 1;

[0055] Figure 2 This is a verification diagram of the soil organic matter inversion model in Northeast China in Experiment 1;

[0056] Figure 3 This is a flowchart illustrating the steps of the rapid mapping method for soil organic matter distribution using the Gaofen-1 satellite at a resolution of 16m, as described in this invention. Detailed Implementation

[0057] The technical solution of the present invention is not limited to the specific embodiments listed below, but also includes any combination of the specific embodiments.

[0058] Specific Implementation Method 1: The rapid mapping method for soil organic matter distribution using the Gaofen-1 satellite at a resolution of 16m is as follows:

[0059] Step 1: Obtain raw data

[0060] Acquire WFV sensor image data of the Gaofen-1 satellite (GF-1) in the target study area;

[0061] Step 2: Image Data Preprocessing

[0062] The raw WFV sensor image data of the Gaofen-1 satellite were preprocessed to obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0063] Step 3: Environmental Covariate Data Processing

[0064] Environmental data download and processing: acquire 30m topographic data and climate information. The topographic data is elevation, and the climate information is free raster data of annual average rainfall 1km and annual average temperature 1km. All data are resampled to a spatial resolution of 16m, consistent with the spatial resolution data of WFV sensor imagery.

[0065] The annual average rainfall 1km and annual average temperature 1km free raster data were downloaded from the National Meteorological Center (http: / / data.cma.cn / ), and the 30m topographic data were obtained from USGS (https: / / gdex.cr.usgs.gov / gdex / ), with the elevation extracted in ArcGIS software;

[0066] Step 4: Random Forest Algorithm Inversion

[0067] The input variables for the random forest algorithm are: annual average rainfall, annual average temperature, elevation, spectral index RI2, spectral index RI3, spectral index DI2, spectral index DI3, spectral index NDVI, spectral index RVI, spectral index DVI, spectral index NDI1, spectral index NDI2, spectral index NDI3, spectral index RDVI1, spectral index RDVI2, RDVI3, spectral index GDVI1, and spectral index GDVI3.

[0068] Random Forest algorithm output variable: Organic matter in topsoil of cultivated land, unit mg / kg;

[0069] The spectral index:

[0070] RI2 = Band3 / Band2

[0071] RI3 = Band4 / Band1

[0072] DI2 = Band3 - Band2

[0073] DI3 = Band4 - Band1

[0074] NDVI=(Band4-Band3) / (Band4+Band3)

[0075] RVI = Band4 / Band3

[0076] DVI = Band4 - Band3

[0077] NDI1=(Band3-Band2) / (Band3+Band2)

[0078] NDI2=(Band4-Band1) / (Band4+Band1)

[0079] NDI3=(Band4-Band2) / (Band4+Band2)

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] Band1 represents the remote sensing reflectance of Band 1 of the WFV sensor on the Gaofen-1 satellite;

[0086] Band2 represents the remote sensing reflectance of Band 2 of the WFV sensor on the Gaofen-1 satellite.

[0087] Band3 represents the remote sensing reflectance of Band 3 of the WFV sensor on the Gaofen-1 satellite.

[0088] Band4 represents the remote sensing reflectance of Band 4 of the WFV sensor on the Gaofen-1 satellite.

[0089] Step 5: Remote sensing mapping of soil organic matter. The prediction model is substituted into the WFV sensor image of Gaofen-1 satellite (GF-1) to achieve rapid mapping of regional soil organic matter distribution.

[0090] Specific Implementation Method Two: This implementation method differs from Specific Implementation Method One in that the preprocessing described in step two includes radiometric calibration, orthorectification, and atmospheric correction. Everything else is the same as in Specific Implementation Method One.

[0091] Specific Implementation Method Three: This implementation method differs from Specific Implementation Method One or Two in that the input variables for the random forest algorithm described in step four are constructed based on a measured soil organic matter inversion model. The construction process of the soil organic matter random forest inversion model is as follows:

[0092] a. During the bare soil period in Northeast China (end of April), a total of 482 soil samples with a surface depth of 0-20 cm were collected from different topographic conditions and soil types. The GPS location of the samples was recorded. The collected soil samples were air-dried, ground and sieved. The soil organic matter content was obtained using the potassium dichromate heating method.

[0093] b. Acquire WFV sensor images from the Gaofen-1 satellite (GF-1). After the image data preprocessing in step two, extract the remote sensing reflectance corresponding to the measured soil sample based on the GPS location, and obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0094] c. After step three, obtain environmental covariate data:

[0095] d. Based on measured soil organic matter data and remote sensing reflectance of different bands from Gaofen-1 satellite, a random forest inversion model for soil organic matter was established:

[0096] Randomly selected measured soil organic matter data and corresponding remote sensing reflectance from different bands of the Gaofen-1 satellite were used as the calibration dataset for the model. The linear correlation between remote sensing reflectance in different bands, spectral indices calculated from these indices, environmental covariates, and measured soil organic matter data was analyzed. Variables associated with measured soil organic matter were selected using a significance level of p < 0.05 and a two-sided confidence level of 0.01. These variables included annual average rainfall, annual average temperature, elevation, and spectral indices RI2, RI3, DI2, DI3, NDVI, RVI, DVI, NDI1, NDI2, NDI3, RDVI1, RDVI2, RDVI3, GDVI1, and GDVI3. These selected variables were then used as input variables for a random forest algorithm to establish a remote sensing inversion model for soil organic matter.

[0097] The analysis of linear correlation was performed using SPSS 16.0 software. The Pearson correlation coefficient r is expressed as follows:

[0098]

[0099] Where, x i This is the measured value of soil organic matter, y i It is the remote sensing reflectance of Gaofen-1 satellite in different bands that corresponds to the measured value of soil organic matter, and N is the number of samples;

[0100] e. The soil organic matter remote sensing inversion model is validated for error. The expression for error validation is as follows:

[0101]

[0102] Where RMSE is the root mean square error, N is the number of samples, and y i This is the measured value of soil organic matter, y i 'This is an estimated value of soil organic matter;

[0103] f. Algorithm verification of linear fit of the dataset R 2 =0.92, RMSE is 4.24 g / kg. Other details are the same as in specific implementation method one or two.

[0104] Specific Implementation Method Four: This implementation method differs from Specific Implementation Methods One to Three in that the random forest algorithm described in step four is used. The random forest inversion algorithm uses the RandomForest package, the number of decision trees (ntree) is 500, and the number of random variables (mtry) for splitting nodes is 1 / 3 of the input. At this point, the values ​​are relatively small and the intra-model error is basically stable. Everything else is the same as in Specific Implementation Methods One to Three.

[0105] The following experiments were used to verify the effectiveness of the invention:

[0106] Experiment 1:

[0107] The rapid mapping method for soil organic matter distribution at 16m resolution using the Gaofen-1 satellite is as follows:

[0108] Step 1: Obtain raw data

[0109] Acquire WFV sensor image data of the Gaofen-1 satellite (GF-1) in the target study area;

[0110] Step 2: Image Data Preprocessing

[0111] The raw WFV sensor image data of the Gaofen-1 satellite were preprocessed to obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0112] Step 3: Environmental Covariate Data Processing

[0113] Environmental data download and processing: acquire 30m topographic data and climate information. The topographic data is elevation, and the climate information is free raster data of annual average rainfall 1km and annual average temperature 1km. All data are resampled to a spatial resolution of 16m, which is consistent with the spatial resolution of WFV sensor image data.

[0114] The annual average rainfall 1km and annual average temperature 1km free raster data were downloaded from the National Meteorological Center (http: / / data.cma.cn / ), and the 30m topographic data were obtained from USGS (https: / / gdex.cr.usgs.gov / gdex / ), with the elevation extracted in ArcGIS software;

[0115] Step 4: Random Forest Algorithm Inversion

[0116] The input variables for the random forest algorithm are: annual average rainfall, annual average temperature, elevation, spectral index RI2, spectral index RI3, spectral index DI2, spectral index DI3, spectral index NDVI, spectral index RVI, spectral index DVI, spectral index NDI1, spectral index NDI2, spectral index NDI3, spectral index RDVI1, spectral index RDVI2, RDVI3, spectral index GDVI1, and spectral index GDVI3.

[0117] Random Forest algorithm output variable: Organic matter in topsoil of cultivated land, unit mg / kg;

[0118] The spectral index:

[0119] RI2 = Band3 / Band2

[0120] RI3 = Band4 / Band1

[0121] DI2 = Band3 - Band2

[0122] DI3 = Band4 - Band1

[0123] NDVI=(Band4-Band3) / (Band4+Band3)

[0124] RVI = Band4 / Band3

[0125] DVI = Band4 - Band3

[0126] NDI1=(Band3-Band2) / (Band3+Band2)

[0127] NDI2=(Band4-Band1) / (Band4+Band1)

[0128] NDI3=(Band4-Band2) / (Band4+Band2)

[0129]

[0130]

[0131]

[0132]

[0133]

[0134] Band1 represents the remote sensing reflectance of Band 1 of the WFV sensor on the Gaofen-1 satellite;

[0135] Band2 represents the remote sensing reflectance of Band 2 of the WFV sensor on the Gaofen-1 satellite.

[0136] Band3 represents the remote sensing reflectance of Band 3 of the WFV sensor on the Gaofen-1 satellite.

[0137] Band4 represents the remote sensing reflectance of Band 4 of the WFV sensor on the Gaofen-1 satellite.

[0138] Step 5: Remote sensing mapping of soil organic matter. The prediction model is substituted into the WFV sensor image of Gaofen-1 satellite (GF-1) to achieve rapid mapping of regional soil organic matter distribution.

[0139] The preprocessing described in step two includes radiometric calibration, orthorectification, and atmospheric correction.

[0140] The input variables for the random forest algorithm described in step four are constructed based on the measured soil organic matter inversion model. The construction process of the soil organic matter random forest inversion model is as follows:

[0141] a. During the bare soil period in Northeast China (end of April), a total of 482 soil samples with a surface depth of 0-20 cm were collected from different topographic conditions and soil types. The GPS location of the samples was recorded. The collected soil samples were air-dried, ground and sieved. The soil organic matter content was obtained using the potassium dichromate heating method.

[0142] b. Acquire WFV sensor images from the Gaofen-1 satellite (GF-1). After the image data preprocessing in step two, extract the remote sensing reflectance corresponding to the measured soil sample based on the GPS location, and obtain the remote sensing reflectance of the Gaofen-1 satellite in different bands.

[0143] c. After step three, obtain environmental covariate data:

[0144] d. Based on measured soil organic matter data and remote sensing reflectance of different bands from Gaofen-1 satellite, a random forest inversion model for soil organic matter was established:

[0145] Randomly selected measured soil organic matter data and corresponding remote sensing reflectance from different bands of the Gaofen-1 satellite were used as the calibration dataset for the model. The linear correlation between remote sensing reflectance in different bands, spectral indices calculated from these indices, environmental covariates, and measured soil organic matter data was analyzed. Variables associated with measured soil organic matter were selected using a significance level of p < 0.05 and a two-sided confidence level of 0.01. These variables included annual average rainfall, annual average temperature, elevation, and the following spectral indices: RI2, RI3, DI2, DI3, NDVI, RVI, DVI, NDI1, NDI2, NDI3, RDVI1, RDVI2, RDVI3, GDVI1, and GDVI3. These selected variables were then used as input variables for a random forest algorithm to establish a remote sensing inversion model for soil organic matter.

[0146] The analysis of linear correlation was performed using SPSS 16.0 software. The Pearson correlation coefficient r is expressed as follows:

[0147]

[0148] Where, x i This is the measured value of soil organic matter, y i It is the remote sensing reflectance of Gaofen-1 satellite in different bands that corresponds to the measured value of soil organic matter, and N is the number of samples;

[0149] e. The soil organic matter remote sensing inversion model is validated for error. The expression for error validation is as follows:

[0150]

[0151] Where RMSE is the root mean square error, N is the number of samples, and y i This is the measured value of soil organic matter, y i 'This is an estimated value of soil organic matter;

[0152] f. Algorithm verification of linear fit of the dataset R 2 =0.81, RMSE is 4.96 g / kg, which meets the internationally accepted quantitative remote sensing inversion accuracy requirements (>45%). Figure 2 .

[0153] The random forest algorithm described in step four uses the RandomForest package for the random forest inversion algorithm. The number of decision trees, ntree, is 500, and the number of random variables, mtry, for splitting nodes is 1 / 3 of the input. At this point, the values ​​are relatively small and the error within the model is basically stable.

Claims

1. A method for rapid mapping soil organic matter distribution using 16m resolution satellite Gaofen-1, characterized in that The method for rapid mapping soil organic matter distribution with 16m resolution using the Gaofen-1 satellite is as follows: Step 1: Get the original data Acquire WFV sensor image data from the Gaofen-1 satellite in the target study area; Step 2: Image data preprocessing The original WFV sensor image data of Gaofen-1 satellite was preprocessed to obtain the remote sensing reflectivity of different bands of Gaofen-1 satellite; Step 3: Environmental covariate data processing Download and process environmental data to obtain 30m terrain data and climate information. The terrain data is elevation, and the climate information is free grid data of annual average rainfall 1km and annual average temperature 1km. All data are resampled to 16m spatial resolution to be consistent with the spatial resolution data of WFV sensor images. The free raster data of annual average rainfall at 1km and annual average temperature at 1km were downloaded from the National Meteorological Center, and the 30m terrain data were obtained from USGS, of which the elevation was extracted in ArcGIS software; Step 4: Random Forest Algorithm Inversion Input variables of the random forest algorithm: average annual rainfall, average annual temperature, elevation, spectral index RI2, spectral index RI3, spectral index DI2, spectral index DI3, spectral index NDVI, spectral index RVI, spectral index DVI, spectral index NDI1, spectral index NDI2, spectral index NDI3, spectral index RDVI1, spectral index RDVI2, RDVI3, spectral index GDVI1 and spectral index GDVI3; Random forest algorithm output variables: organic matter in the topsoil of cultivated land, unit: mg / kg; The spectral index: RI2=Band3 / Band2 RI3=Band4 / Band1 DI2=Band3-Band2 DI3=Band4-Band1 NDVI=(Band4-Band3) / (Band4+Band3) RVI=Band4 / Band3 DVI=Band4-Band3 NDI1=(Band3-Band2) / (Band3+Band2) NDI2=(Band4-Band1) / (Band4+Band1) NDI3=(Band4-Band2) / (Band4+Band2) Band1 represents the remote sensing reflectance of Band 1 of the WFV sensor of the Gaofen-1 satellite; Band2 represents the remote sensing reflectivity of Band 2 of the WFV sensor of the Gaofen-1 satellite; Band3 represents the remote sensing reflectivity of Band 3 of the WFV sensor of the Gaofen-1 satellite; Band4 represents the remote sensing reflectivity of Band 4 of the WFV sensor of the Gaofen-1 satellite; Step 5: Soil organic matter remote sensing mapping: Substitute the prediction model into the WFV sensor image of the Gaofen-1 satellite to achieve rapid mapping of regional soil organic matter distribution.

2. The method for rapid mapping soil organic matter distribution using the Gaofen-1 satellite at a resolution of 16 m according to claim 1, characterized in that The preprocessing described in step 2 includes radiation calibration, orthorectification and atmospheric correction.

3. The method for rapid mapping soil organic matter distribution using the Gaofen-1 satellite at a resolution of 16 m according to claim 1, characterized in that The input variables of the random forest algorithm described in step 4 are constructed based on the measured soil organic matter inversion model. The construction process of the soil organic matter random forest inversion model is as follows: a. During the bare soil period in Northeast China, a total of 482 soil samples with a surface depth of 0-20 cm were collected from different terrain conditions and soil types, and the GPS locations of the samples were recorded. The collected soil samples were air-dried, ground and sieved, and the soil organic matter content was obtained using the potassium dichromate heating method; b. Obtain the WFV sensor image of the Gaofen-1 satellite, and after the image data preprocessing process in step 2, extract the remote sensing reflectivity corresponding to the measured soil sample according to the GPS position, and obtain the remote sensing reflectivity of different bands of the Gaofen-1 satellite; c. After step 3, obtain environmental covariate data: d. Based on the measured soil organic matter data and remote sensing reflectance of different bands of Gaofen-1 satellite, a random forest inversion model of soil organic matter was established: Randomly select measured soil organic matter data and the corresponding remote sensing reflectance of different bands of Gaofen-1 satellite as the correction data set of the model, analyze the linear correlation between remote sensing reflectance of different bands, spectral index calculated by remote sensing reflectance of different bands, environmental covariate data and measured soil organic matter data, and screen out variables related to measured soil organic matter through the significance level p<0.05 and the confidence level of 0.01 on both sides, namely, annual average rainfall, annual average temperature, elevation, spectral index RI2, spectral index RI3, spectral index DI2, spectral index DI3, spectral index NDVI, spectral index RVI, spectral index DVI, spectral index NDI1, spectral index NDI2, spectral index NDI3, spectral index RDVI1, spectral index RDVI2, RDVI3, spectral index GDVI1 and spectral index GDVI3; use the screened variables as input variables of random forest algorithm to establish soil organic matter remote sensing inversion model; The linear correlation of the measured soil organic matter data was realized on the SPSS16.0 software platform, and the Pearson correlation coefficient r was expressed as follows: Among them, x i is the measured value of soil organic matter, y i is the remote sensing reflectance of different bands of Gaofen-1 satellite corresponding to the measured value of soil organic matter, and N is the number of samples; e. Perform error verification on the soil organic matter remote sensing inversion model. The error verification expression is as follows: Where RMSE is the root mean square error, N is the number of samples, and y i is the measured value of soil organic matter, y i ' is the estimated soil organic matter; f. Algorithm test data set linear fit R 2 =0.92, RMSE is 4.24 g / kg.

4. The method for rapid mapping soil organic matter distribution using the Gaofen-1 satellite at a resolution of 16 m according to claim 1, characterized in that The random forest algorithm described in step 4 uses the RandomForest package for the random forest inversion algorithm. The number of decision trees ntree is 500, and the number of random variables for splitting nodes mtry is 1 / 3 of the input. At this time, the value is small and the error within the model is basically stable.

Images