Method for Rapidly Mapping the pH of the Surface Soil of Cultivated Land by WFV

Abstract

A method for rapid spatial mapping of the pH value of the arable land surface soil by WFV. The present invention belongs to the field of satellite remote sensing technology and its applications, and relates to a method for rapid spatial mapping of the pH value of the arable land surface soil. This method: obtains the images of the WFV sensor of the GF-1 satellite in the target study area, and obtains the surface reflectance of different bands after preprocessing; uses the covariate data of different geographical and climatic environments determined by the present invention as the input variables of the random forest algorithm, and the output variable is the pH value of the arable land surface soil, so as to realize the mapping of the pH value of the arable land surface soil. The method of the present invention expands the application of the domestic GF-1 satellite in digital soil mapping, can reduce manpower, material resources and financial resources, and can monitor the soil acidity and alkalinity of agricultural arable land without destroying the arable land target by on-site sampling.

Description

Technical Field

[0001] The present invention belongs to the field of satellite remote sensing technology and its applications, and particularly relates to a method for rapidly mapping the pH of the surface soil of cultivated land in space. Background Art

[0002] Soil organic matter is one of the important indicators for evaluating soil degradation, soil health, and soil fertility, and is a key factor affecting crop yields. Therefore, in the aspects of smart agriculture and sustainable utilization of agricultural land, it is urgent to rapidly and accurately estimate soil organic matter. Traditional spatial mapping methods represented by geostatistics (such as the Kriging method) require field sampling and testing, and are often concentrated on the spatial mapping of fragmented cultivated land in small-scale areas. For remote sensing mapping in large-scale areas, they are often not applicable because the influence of soil organic matter in agricultural land is affected by multiple factors such as topography, parent material, climate, vegetation, and agronomic measures, and the spatial heterogeneity is relatively large. With the development of earth observation technology and the innovation of artificial intelligence algorithms, the combination of satellite remote sensing observation means and big data analysis methods has become an important method for quantitatively monitoring and mapping surface environmental parameters in large-scale areas. In previous digital soil mapping studies, Landsat series satellites and Sentinel-2 series satellites were mostly used, while there was less digital soil mapping based on the domestic high-resolution No. 1 satellite.

[0003] Satellite remote sensing mapping of soil organic matter mainly relies on spectral technology to analyze the differences in the physical and chemical properties of different ground objects, and the specific spectral characteristic differences are mainly reflected in the differences in the remote sensing reflectance of different spectral bands. Soil organic matter has a reflectance spectral characteristic in the visible and near-infrared wavelength ranges, and as the content of soil organic matter increases, the soil reflectance decreases, resulting in changes in the shape of the spectral curve, and it is impossible to observe the dynamic changes in the soil acidity and alkalinity of farmland cultivated land. Therefore, in the satellite remote sensing mapping of soil organic matter in large-scale areas, it is necessary to construct an inversion algorithm that combines environmental variables and spectral indices based on considering the changing geographical environment and other backgrounds to improve the accuracy of satellite remote sensing mapping of soil organic matter in large-scale areas. Summary of the Invention

[0004] The purpose of the present invention is to solve the technical problem that as the content of soil organic matter increases, the soil reflectance decreases, resulting in changes in the shape of the spectral curve, and it is impossible to observe the dynamic changes in the soil acidity and alkalinity of farmland cultivated land, and provides a method for rapidly mapping the pH of the surface soil of cultivated land by WFV.

[0005] The method for rapidly mapping the pH of the surface soil of cultivated land by WFV is as follows:

[0006] Step 1: Obtain the data of the WFV sensor image of the high-resolution No. 1 satellite (abbreviated as WFV image);

[0007] Step 2: Preprocess the images of the WFV sensor of GF-1 satellite. The preprocessing includes orthorectification, radiometric calibration, and atmospheric correction;

[0008] Step 3: Obtain geoclimatic environmental covariate data, including 1-km spatial resolution soil texture remote sensing products, 1-km spatial resolution annual average temperature data, 1-km spatial resolution annual average rainfall data, 30-m spatial resolution DEM elevation data, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0009] The 1-km spatial resolution soil texture is clay, sand, or silt;

[0010] Step 4: Resample the spatial resolution of the geoclimatic environmental covariate data to 16 m;

[0011] Step 5: Random forest algorithm:

[0012] Input variables of the random forest algorithm: clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0013] Output variable of the random forest algorithm: soil pH;

[0014] The spectral index

[0015] RI2 = band3 / band2

[0016] RI3 = band4 / band1

[0017] DI2 = band3 - band2

[0018] DI3 = band4 - band1

[0019] NDVI = (band4 - band3) / (band4 + band3)

[0020] RVI = band4 / band3

[0021] DVI = band4 - band3

[0022] NDI1 = (band3 - band2) / (band3 + band2)

[0023] NDI2 = (band4 - band1) / (band4 + band1)

[0024] NDI3 = (band4 - band2) / (band4 + band2)

[0025]

[0026] band1 represents the surface reflectance of band 1 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0027] band2 represents the surface reflectance of band 2 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0028] band3 represents the surface reflectance of band 3 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0029] band4 represents the surface reflectance of band 4 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0030] Step 6: Substitute the model into the pre - processed WFV sensor image of GF - 1 satellite to realize the remote sensing mapping of the spatial distribution of regional soil pH.

[0031] The atmospheric correction process of the image data described in Step 2 uses FLAASH.

[0032] The input variables of the random forest algorithm described in Step 5 are obtained based on the large - scale collection of soil samples and the training of soil pH testing. The process is as follows:

[0033] a. In the Northeast Black Soil Region, a total of 161 soil samples of the surface layer of 0 - 20 cm are collected, and the longitude and latitude information of each soil sample is recorded. Then, the collected soil samples are air - dried, ground, and sieved, and the soil pH content is obtained using the glass electrode method;

[0034] b. Obtain the WFV sensor image of GF - 1 satellite. After the image data pre - processing process described in Step 2, according to the longitude and latitude information of each soil sample, the surface reflectance of different bands of the WFV sensor image of GF - 1 satellite corresponding to each soil sample is extracted;

[0035] c. The measured soil pH and the corresponding surface reflectance of different bands are used as the calibration data set of the model;

[0036] The surface reflectance of different bands is the surface reflectance of band 1 after the preprocessing of the WFV sensor image of GF-1 satellite with bands 1 - 4;

[0037] d. Through steps three and four, obtain the geo - climatic environmental covariate data with the same spatial resolution as the WFV sensor image of GF - 1 satellite:

[0038] e. Based on the measured soil pH and the surface reflectance of different bands of the WFV sensor image of GF - 1 satellite, establish a random forest inversion model for soil pH:

[0039] Analyze the linear correlation between the surface reflectance of different bands, the spectral indices calculated from the surface reflectance of different bands, the geo - climatic environmental covariate data and the measured soil pH, screen out the variables related to the measured soil pH, and finally determine them as clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2;

[0040] Take the screened variables as the input variables of the random forest algorithm and establish a remote sensing inversion model for soil pH;

[0041] The linear correlation analysis is implemented on the SPSS16.0 software platform, with a significance level p < 0.05 and a two - sided confidence level of 0.01. The Pearson correlation coefficient r is expressed as follows:

[0042]

[0043] Among them, x i is the measured value of soil pH, y i is the surface reflectance of the WFV sensor image of GF - 1 satellite after pre - processing matching the measured value of soil pH, and N is the number of samples;

[0044] f. For the random forest algorithm, conduct error verification. The expression of error verification is as follows:

[0045]

[0046] Among them, RMSE is the root mean square error, N is the number of samples, y i is the measured value of soil pH, y i ' is the estimated value of soil pH;

[0047] g. The linear fitting R of the algorithm test data set 2= 0.94, and the RMSE is 0.29.

[0048] The random forest algorithm described in step five is characterized by using the RandomForest package, where the number of decision trees ntree is 500, and the number of random variables for splitting nodes mtry is 1 / 3 of the input quantity. At this time, the value is small and the error within the model is basically stable.

[0049] This method can quickly and quantitatively complete the spatial mapping of the pH of the arable land surface soil in different regions and at different times, and accurately obtain the changes in soil pH. The input variables of the random forest determined by the method of the present invention are constructed based on the measured soil pH, and the measured soil pH can represent the changes under different soil types and agronomic measures, and is applicable to the spatial mapping of soil pH in Northeast China. The method of the present invention expands the application of the domestic Gaofen-1 satellite in digital soil mapping, can reduce manpower, material resources and financial resources, and can achieve the monitoring of the soil acidity and alkalinity of agricultural arable land without destroying the arable land by field sampling. Description of the Drawings

[0050] Figure 1 is the calibration diagram of the soil pH inversion model in Experiment 1;

[0051] Figure 2 is the verification diagram of the soil pH inversion model in Experiment 1;

[0052] Figure 3 is the step schematic diagram of the method for rapid spatial mapping of the pH of the arable land surface soil of the WFV of the present invention. Detailed Embodiments

[0053] The technical solution of the present invention is not limited to the following specific embodiments listed, and also includes any combination between the specific embodiments.

[0054] Specific Embodiment 1: The method for rapid spatial mapping of the pH of the arable land surface soil of the WFV in this embodiment is as follows:

[0055] Step 1, obtain the data of the WFV sensor image of the Gaofen-1 satellite (abbreviated as WFV image);

[0056] Step 2, preprocess the WFV sensor image of the Gaofen-1 satellite. The preprocessing includes orthorectification, radiometric calibration and atmospheric correction;

[0057] Step 3: Obtain geoclimatic environmental covariate data, including 1-km spatial resolution soil texture remote sensing products, 1-km spatial resolution annual average temperature data, 1-km spatial resolution annual average rainfall data, 30-m spatial resolution DEM elevation data, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0058] The 1-km spatial resolution soil texture is clay, sand, or silt;

[0059] Step 4: Resample the spatial resolution of the geoclimatic environmental covariate data to 16 m;

[0060] Step 5: Random forest algorithm:

[0061] Input variables of the random forest algorithm: clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0062] Output variable of the random forest algorithm: soil pH;

[0063] The spectral index

[0064] RI2 = band3 / band2

[0065] RI3 = band4 / band1

[0066] DI2 = band3 - band2

[0067] DI3 = band4 - band1

[0068] NDVI = (band4 - band3) / (band4 + band3)

[0069] RVI = band4 / band3

[0070] DVI = band4 - band3

[0071] NDI1 = (band3 - band2) / (band3 + band2)

[0072] NDI2 = (band4 - band1) / (band4 + band1)

[0073] NDI3 = (band4 - band2) / (band4 + band2)

[0074]

[0075] band1 represents the surface reflectance of band 1 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0076] band2 represents the surface reflectance of band 2 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0077] band3 represents the surface reflectance of band 3 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0078] band4 represents the surface reflectance of band 4 after the pre - processing of the WFV sensor image of GF - 1 satellite;

[0079] Step 6: Substitute the model into the pre - processed WFV sensor image of GF - 1 satellite to realize the remote sensing mapping of the spatial distribution of regional soil pH.

[0080] Specific Embodiment 2: The difference between this embodiment and Specific Embodiment 1 is that the FLAASH is used for the atmospheric correction processing of the image data in Step 2. Others are the same as Specific Embodiment 1.

[0081] Specific Embodiment 3: The difference between this embodiment and Specific Embodiment 1 or 2 is that the input variables of the random forest algorithm in Step 5 are obtained by training based on the large - scale collection of soil samples and the measurement of soil pH. The process is as follows:

[0082] a. In the black soil area of Northeast China, a total of 161 soil samples of the surface layer of 0 - 20 cm are collected, and the longitude and latitude information of each soil sample is recorded. Then, the collected soil samples are air - dried, ground, and sieved, and the soil pH content is obtained using the glass electrode method.

[0083] b. Obtain the WFV sensor image of GF - 1 satellite. After the image data pre - processing process described in Step 2, according to the longitude and latitude information of each soil sample, the surface reflectance of different bands of the WFV sensor image of GF - 1 satellite corresponding to each soil sample is extracted.

[0084] c. The measured soil pH and the corresponding surface reflectance of different bands are used as the calibration data set of the model.

[0085] The surface reflectance of different bands is the surface reflectance of band 1 after the preprocessing of the WFV sensor image of GF-1 satellite with bands 1 - 4;

[0086] d. After steps three and four, obtain the geo - climatic environmental covariate data with the same spatial resolution as the WFV sensor image of GF-1 satellite:

[0087] e. Based on the measured soil pH and the surface reflectance of different bands of the WFV sensor image of GF-1 satellite, establish a random forest inversion model for soil pH:

[0088] Analyze the linear correlation between the surface reflectance of different bands, the spectral indices calculated from the surface reflectance of different bands, the geo - climatic environmental covariate data and the measured soil pH, screen out the variables related to the measured soil pH, and finally determine them as clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2;

[0089] Take the screened variables as the input variables of the random forest algorithm and establish a remote sensing inversion model for soil pH;

[0090] The linear correlation analysis is implemented on the SPSS16.0 software platform, with a significance level p < 0.05 and a two - sided confidence level of 0.01. The Pearson correlation coefficient r is expressed as follows:

[0091]

[0092] Among them, x i is the measured value of soil pH, y i is the surface reflectance of the WFV sensor image of GF-1 satellite after preprocessing matching the measured value of soil pH, and N is the number of samples;

[0093] f. For the random forest algorithm, perform error verification. The expression for error verification is as follows:

[0094]

[0095] Among them, RMSE is the root mean square error, N is the number of samples, y i is the measured value of soil pH, and y i ' is the estimated value of soil pH;

[0096] g. The linear fitting R of the algorithm test dataset 2= 0.94, and the RMSE is 0.29. Others are the same as in the first or second specific implementation manner.

[0097] Specific implementation manner four: The difference between this implementation manner and one of the first to third specific implementation manners is the random forest algorithm described in step five. The RandomForest package is used, the number of decision trees ntree is 500, and the number of random variables for splitting nodes mtry is 1 / 3 of the input quantity. At this time, the value is small and the internal error of the model is basically stable. Others are the same as one of the first to third specific implementation manners.

[0098] The following experiment is used to verify the effect of the present invention:

[0099] Experiment one:

[0100] The method for rapid spatial mapping of the pH value of the surface soil of cultivated land by WFV is as follows:

[0101] Step one: Obtain the data of the WFV sensor image of the GF-1 satellite (abbreviated as WFV image);

[0102] Step two: Preprocess the WFV sensor image of the GF-1 satellite. The preprocessing includes orthorectification, radiometric calibration, and atmospheric correction;

[0103] Step three: Obtain the geographical and climatic environmental covariate data, including the remote sensing product of soil texture with a spatial resolution of 1 km, the annual average temperature data with a spatial resolution of 1 km, the annual average rainfall data with a spatial resolution of 1 km, the DEM elevation data with a spatial resolution of 30 m, the spectral index RI2, spectral index, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0104] The soil texture with a spatial resolution of 1 km is clay, sand, or silt;

[0105] Step four: Resample the spatial resolution of the geographical and climatic environmental covariate data to 16 m;

[0106] Step five: Random forest algorithm:

[0107] Input variables of the random forest algorithm: clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1, and spectral index GDVI2;

[0108] Output variable of the random forest algorithm: soil pH;

[0109] The said spectral index

[0110] RI2 = band3 / band2

[0111] RI3 = band4 / band1

[0112] DI2 = band3 - band2

[0113] DI3 = band4 - band1

[0114] NDVI = (band4 - band3) / (band4 + band3)

[0115] RVI = band4 / band3

[0116] DVI = band4 - band3

[0117] NDI1 = (band3 - band2) / (band3 + band2)

[0118] NDI2 = (band4 - band1) / (band4 + band1)

[0119] NDI3 = (band4 - band2) / (band4 + band2)

[0120]

[0121] band1 represents the surface reflectance of band 1 after preprocessing of the WFV sensor image of Gaofen-1 satellite;

[0122] band2 represents the surface reflectance of band 2 after preprocessing of the WFV sensor image of Gaofen-1 satellite;

[0123] band3 represents the surface reflectance of band 3 after preprocessing of the WFV sensor image of Gaofen-1 satellite;

[0124] band4 represents the surface reflectance of band 4 after preprocessing of the WFV sensor image of Gaofen-1 satellite;

[0125] Step 6: Substitute the model into the preprocessed GF-1 satellite WFV sensor images to achieve remote sensing mapping of the spatial distribution of regional soil pH.

[0126] The FLAASH is used for the atmospheric correction of the image data described in Step 2.

[0127] The input variables of the random forest algorithm described in Step 5 are obtained based on the training of soil samples collected on a large scale and the indoor experimental measurement of soil pH. The process is as follows:

[0128] a. A total of 68 surface soil samples of 0-20 cm were collected in the black soil area of Northeast China during the bare soil period from April 20th to 29th, 2023. The longitude and latitude information of each soil sample was recorded. After air-drying, grinding and sieving the collected soil samples in the laboratory, the soil pH content was obtained using the glass electrode method.

[0129] b. Obtain the GF-1 satellite WFV sensor images. After the image data preprocessing process described in Step 2, according to the longitude and latitude information of each soil sample, the surface reflectance of different bands of the GF-1 satellite WFV sensor images corresponding to each soil sample was extracted.

[0130] c. The measured soil pH and the corresponding surface reflectance of different bands are used as the calibration data set of the model.

[0131] The surface reflectance of different bands is the surface reflectance of Band 1 after the preprocessing of the GF-1 satellite WFV sensor images of Band 1-Band 4.

[0132] d. After Step 3 and Step 4, obtain the geo-climatic environmental covariate data with the same spatial resolution as the GF-1 satellite WFV sensor images:

[0133] e. Based on the measured soil pH and the surface reflectance of different bands of the GF-1 satellite WFV sensor images, establish a random forest inversion model for soil pH:

[0134] Analyze the linear correlation between the surface reflectance of different bands, the spectral indices calculated from the surface reflectance of different bands, the geo-climatic environmental covariate data and the measured soil pH, and screen out the variables related to the measured soil pH. Finally, they are determined to be clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2.

[0135] Use the filtered variables as the input variables of the random forest algorithm to establish a remote sensing inversion model for soil pH;

[0136] The linear correlation analysis is implemented on the SPSS 16.0 software platform. The significance level p < 0.05 and the confidence level is two-sided 0.01. The Pearson correlation coefficient r is expressed as follows:

[0137]

[0138] where x i is the measured value of soil pH, y i is the surface reflectance of the GF-1 satellite WFV sensor image after preprocessing matching the measured value of soil pH, and N is the number of samples;

[0139] f. For the random forest algorithm described above, perform error verification. The expression for error verification is as follows:

[0140]

[0141] where RMSE is the root mean square error, N is the number of samples, y i is the measured value of soil pH, and y i ' is the estimated value of soil pH;

[0142] g. The linear fitting R 2 of the algorithm test dataset is 0.70, and the RMSE is 0.63, meeting the internationally accepted quantitative remote sensing inversion accuracy requirements (R 2 > 0.45), as shown in Figure 2 .

[0143] For the random forest algorithm described in step five, use 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 quantity. At this time, the value is relatively small and the internal error of the model is basically stable.

Claims

1. WFV's method for rapid spatial mapping of pH in the topsoil of cultivated land is characterized by The WFV method for rapid spatial mapping of pH in the cultivated surface soil is as follows: Step 1: Obtain the WFV sensor image data of the Gaofen-1 satellite; Step 2: Preprocess the WFV sensor image of the Gaofen-1 satellite, including orthorectification, radiometric calibration and atmospheric correction; Step 3, obtain geographic climate environment covariate data, including 1km spatial resolution soil texture remote sensing products, 1km spatial resolution annual average temperature data, 1km spatial resolution annual average rainfall data, 30m spatial resolution DEM elevation data, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2; The soil texture of the 1km spatial resolution is clay, sand or silt; Step 4: The spatial resolution of geographic climate and environmental covariate data was resampled to 16 m; Step 5: Random Forest Algorithm: Input variables of random forest algorithm: clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2; Random forest algorithm output variables: soil pH; 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 surface reflectance of band 1 after preprocessing of the WFV sensor image of the Gaofen-1 satellite; band2 represents the surface reflectance of band 2 after preprocessing of the WFV sensor image of the Gaofen-1 satellite; band3 represents the surface reflectance of band 3 after preprocessing of the WFV sensor image of the Gaofen-1 satellite; band4 represents the surface reflectance of band 4 after preprocessing of the WFV sensor image of the Gaofen-1 satellite; Step 6: Substitute the model into the preprocessed GF-1 satellite WFV sensor image to achieve remote sensing mapping of regional soil pH spatial distribution.

2. The WFV method for rapid spatial mapping of pH of cultivated surface soil according to claim 1, characterized in that The atmospheric correction process of the image data in step 2 adopts FLAASH.

3. The WFV method for rapid spatial mapping of pH of cultivated surface soil according to claim 1, characterized in that The input variables of the random forest algorithm described in step 5 are obtained based on a large-scale collection of soil samples and soil pH testing training. The process is as follows: a. In the black soil region of Northeast China, 68 to 161 soil samples with a surface depth of 0-20 cm were collected, and the longitude and latitude information of each soil sample was recorded. The collected soil samples were then air-dried, ground and sieved, and the soil pH content was obtained using the glass electrode method; b. Obtaining the Gaofen-1 satellite WFV sensor image, after the image data preprocessing process described in step 2, extracting the surface reflectance of the Gaofen-1 satellite WFV sensor image of different bands corresponding to each soil sample according to the latitude and longitude information of each soil sample; c. The measured soil pH and the corresponding surface reflectance in different bands are used as the calibration data set of the model; The surface reflectance in different bands is the surface reflectance in band 1 after preprocessing of the WFV sensor image of the Gaofen-1 satellite from band 1 to band 4; d. After steps 3 and 4, obtain geographic climate environment covariate data consistent with the spatial resolution of the Gaofen-1 satellite WFV sensor image: e. Based on the measured soil pH and the surface reflectance of different bands of the WFV sensor image of the Gaofen-1 satellite, a random forest inversion model of soil pH was established: The linear correlation between the surface reflectance in different bands, the spectral index calculated from the surface reflectance in different bands, the geographic climate environment covariate data and the measured soil pH was analyzed, and the variables related to the measured soil pH were screened out, and finally determined to be clay, sand, silt, 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, spectral index RDVI3, spectral index GDVI1 and spectral index GDVI2; The screened variables were used as input variables of the random forest algorithm to establish a soil pH remote sensing inversion model; The linear correlation between the geographic climate environment covariate data and the measured soil pH was realized on the SPSS16.0 software platform, with a significance level of p<0.05 and a confidence level of 0.01 on both sides. The Pearson correlation coefficient r is expressed as follows: Among them, x i is the measured soil pH value, y i is the surface reflectance of the WFV sensor image of the Gaofen-1 satellite after preprocessing to match the measured soil pH value, and N is the number of samples; f. The random forest algorithm is subjected to error verification, and the expression of error verification is as follows: Where RMSE is the root mean square error, N is the number of samples, and y i is the measured soil pH value, y i ' is the estimated soil pH; g. Algorithm test data set linear fit R 2 =0.70, RMSE is 0.

63.

4. The WFV method for rapid spatial mapping of pH of cultivated land surface soil according to claim 1, characterized in that The random forest algorithm described in step 5 uses the RandomForest package, 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