Method for estimating pH value of surface soil of cultivated land based on MSI

Abstract

The invention discloses a cultivated land surface soil pH value estimation method based on MSI, and relates to the field of satellite remote sensing technology and application thereof, in particular to a cultivated land surface soil pH value estimation method. The invention aims to solve the problems that the existing soil pH value measuring method is limited by the sample size and cost, the timeliness is poor, and the soil pH value distribution is difficult to analyze on the regional or national scale. The method comprises the following steps: (1) acquiring remote sensing reflectivity of different wavebands; (2) taking reflectivity of different wave bands of MSI remote sensing reflectivity, 13 spectral indexes and environment variables as input variables of a random forest regression model; and (3) the output of the random forest regression model is a cultivated land surface soil pH predicted value, and rapid mapping of the cultivated land surface soil pH value is realized. By means of the advantages of low cost, high timeliness, spatial display and non-contact, the spatial distribution diagram of the pH content of the surface soil of the cultivated land in the target area can be obtained only by processing satellite image spectrum data.

Description

Technical Field

[0001] The invention relates to satellite remote sensing technology and its application field, and in particular to a method for estimating pH value of cultivated land surface soil. Background Art

[0002] The pH value of soil indicates the degree of soil acidification and alkalinity. Too high or too low soil pH value will lead to the deterioration of soil physical and chemical properties, making it less permeable to air and water, which in turn affects the absorption of nutrients by plant roots. For cultivated land, the higher the soil pH value, the stronger the soil alkalinity, and the more likely the soil will show compaction. On the one hand, it will weaken the stress resistance of cultivated land soil and make it more difficult to improve alkaline soil. On the other hand, it will affect crop yield and quality, restricting the health of the ecosystem and the sustainable development of the agricultural economy.

[0003] With the acceleration of the industrialization process of human society, the emission of acidic gas SO2, particulate pollutants and the large-scale application of nitrogen fertilizers have gradually aggravated soil acidification. In order to prevent and control the alkalinization of farmland soil, it is necessary to continuously monitor the distribution of soil pH in farmland. Traditional soil pH measurement methods require real-time measurement in farmland or collection of soil samples for laboratory measurement. Although this method can relatively accurately reflect the pH status of cultivated soil, it is limited by sample size and cost, and has poor timeliness, making it difficult to analyze soil pH distribution at a regional or national scale. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problems that the existing soil pH measurement methods are limited by sample size and cost constraints, have poor timeliness, and are difficult to analyze the distribution of soil acidity and alkalinity at a regional or national scale. A method for estimating the pH value of cultivated land surface soil based on MSI is provided.

[0005] The MSI-based method for estimating pH value of cultivated land surface soil is carried out in the following steps:

[0006] (1) Obtain bare soil and cloud-free images of the Sentinel-2 satellite MSI sensor Level 2A sensor, resample them to a spatial resolution of 10 m, and obtain remote sensing reflectance in different bands;

[0007] (2) The reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables were used as input variables of the random forest regression model;

[0008] The reflectances of different bands of the MSI remote sensing reflectance are Band 1, Band 2, Band 3, Band 4, Band 5, Band 6, Band 7, Band 8, Band 9, and Band 10;

[0009] The 13 spectral indices are salinity index (SI), salinity index 1 (SI1), salinity index 3 (SI3), salinity index II (S2), salinity index IX (S9), normalized difference snow index (NDSI), optimized soil adjusted vegetation index (OSAVI), weighted difference vegetation index (WDVI), salinity index I (SI-I), brightness index (BI), soil salinity index 1 (SSI1), salinity index 4 (SI4), and salinity index 2 (SI2);

[0010] The environmental variables are 1km annual average rainfall raster data, 30m digital elevation data of the target area and 1km soil bulk density raster data;

[0011] The calculation methods of the 13 spectral indices are as follows:

[0012] Salinity index = (Band3×Band4) 0.5 ;

[0013] Salinity index 1 = (Band 3 + Band 4) 0.5 ;

[0014] Salinity index 3 = Band3 × Band3 + Band4 × Band4;

[0015] Salinity index II = (Band2×Band4) / (Band2+Band4);

[0016] Salinity index IX = (Band 3 + Band 4 + Band 8) / 2;

[0017] Normalized difference snow index = (Band8 × Band11) / (Band8 + Band11);

[0018] Optimized soil-adjusted vegetation index = (Band4×Band8) / (Band4+Band8+0.16);

[0019] Weighted difference vegetation index = (Band8-Band4×0.5);

[0020] Salinity index I = (Band4 × Band8) / Band3;

[0021] Brightness index = [(Band4×Band4+Band3×Band3+Band2×Band2) / 3] 0.5 ;

[0022] Soil salinity index 1 = (Band2 × Band4) / Band3;

[0023] Salinity index 4 = (Band2×Band4)0.5 ;

[0024] Salinity index 2 = (Band 2 + Band 4) 0.5 ;

[0025] Band1 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 1;

[0026] Band2 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 2;

[0027] Band3 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 3;

[0028] Band4 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 4;

[0029] Band5 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 5;

[0030] Band6 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 6;

[0031] Band7 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 7;

[0032] Band8 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 8;

[0033] Band9 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 9;

[0034] Band10 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 10;

[0035] Band11 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 11;

[0036] The input variables of the random forest regression model, namely the reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables, are constructed based on the measured soil pH inversion model. The construction process is as follows:

[0037] a. Collect soil samples from the surface layer of 0.1-20 cm of cultivated land in Songnen Plain. Considering the different terrain conditions and soil types, the soil samples and water were dissolved in a mass ratio of 1:5, and then measured using a pH meter (PHS-3C);

[0038] b. Download the bare soil cloud-free image of the Sentinel-2 satellite MSI sensor from USGS. The downloaded product is Level-2A, and the band is resampled to 10m spatial resolution;

[0039] c. Obtain the reflectance of different bands of MSI remote sensing reflectance and calculate 13 spectral indices;

[0040] d. Obtain the 1km annual average rainfall raster data of the soil, the 30m digital elevation data of the target area, and the 1km soil bulk density raster data;

[0041] e. Based on the measured pH data and remote sensing reflectance data, a random forest regression model of pH value of cultivated land surface soil was established. The correlation analysis was implemented on the SPSS16.0 software platform. The Pearson correlation coefficient r is expressed as follows:

[0042]

[0043] Among them, x and y are the variables pH value and remote sensing reflectance, and N is the number of samples;

[0044] f. The random forest regression model of pH value of cultivated land surface soil is evaluated based on RMSE and MAE, and the specific calculation is as follows:

[0045]

[0046] Where RMSE is the root mean square error, MAE is the mean absolute percentage error, N is the number of samples, and y i is the measured pH value of the surface soil of cultivated land, y i ' is the predicted pH value of the surface soil of cultivated land;

[0047] (3) The output of the random forest regression model is the predicted pH value of the surface soil of cultivated land, which enables rapid mapping of the pH value of the surface soil of cultivated land.

[0048] The 1km annual average rainfall grid data described in step (2) is downloaded from the National Meteorological Center.

[0049] The 30m digital elevation data of the target area in step (2) is obtained from USGS.

[0050] The 1 km soil bulk density grid data in step (2) was obtained from the National Tibetan Plateau Data Center.

[0051] The random forest regression model in step (2) uses the Random Forest 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.

[0052] Remote sensing technology, with its advantages of low cost, high timeliness, spatial display and non-contact, is used in the quantitative prediction and analysis of soil physical and chemical parameters to guide agricultural production. That is, by simply processing satellite image spectral data, it is possible to obtain the spatial distribution map of the pH content of the surface soil of cultivated land in the target area, providing technical support for farmland soil improvement, improvement of cultivated land quality and development of precision agriculture.

[0053] Based on the present invention, with the help of SNAP, GEE geographic computing cloud platform and online data access function, a random forest regression model can be built through autonomous programming, and the input variables described in the present invention can be input to estimate the pH of the surface soil of cultivated land, and the results can be standardized and quickly mapped.

[0054] The present invention uses 161 measured samples of soil pH in the surface layer of cultivated land covering the Songnen Plain in Northeast China to train the input variables of the random forest regression algorithm (see Appendix Figure 1 ), the measured samples cover a variety of cultivated land soil types and can better represent cultivated land with different soil types, ensuring that the developed inversion algorithm has good applicability (Appendix Figure 2 ), R 2 =0.94, RMSE and MAE were 0.29 and 0.23 respectively. It is helpful to quickly and dynamically monitor the soil acid-base condition in the bare soil period and evaluate the effectiveness of soil alkalinization improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is the sampling distribution map of pH value of surface soil samples in the Songnen Plain of Northeast China in Experiment 1;

[0056] Figure 2 This is the calibration diagram of the pH value inversion model of the cultivated land surface soil in Songnen Plain in Northeast China in Experiment 1;

[0057] Figure 3 This is the verification diagram of the inversion model for pH value of surface soil in the Songnen Plain of Northeast China in Experiment 1. DETAILED DESCRIPTION

[0058] The technical solution of the present invention is not limited to the specific implementation modes listed below, but also includes any combination of the specific implementation modes.

[0059] Specific implementation method 1: The method for estimating the pH value of the cultivated land surface soil based on MSI in this implementation is carried out in the following steps:

[0060] (1) Obtain bare soil and cloud-free images of the Sentinel-2 satellite MSI sensor Level 2A sensor, resample them to a spatial resolution of 10 m, and obtain remote sensing reflectance in different bands;

[0061] (2) The reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables were used as input variables of the random forest regression model;

[0062] The reflectances of different bands of the MSI remote sensing reflectance are Band 1, Band 2, Band 3, Band 4, Band 5, Band 6, Band 7, Band 8, Band 9, and Band 10;

[0063] The 13 spectral indices are salinity index (SI), salinity index 1 (SI1), salinity index 3 (SI3), salinity index II (S2), salinity index IX (S9), normalized difference snow index (NDSI), optimized soil adjusted vegetation index (OSAVI), weighted difference vegetation index (WDVI), salinity index I (SI-I), brightness index (BI), soil salinity index 1 (SSI1), salinity index 4 (SI4), and salinity index 2 (SI2);

[0064] The environmental variables are 1km annual average rainfall raster data, 30m digital elevation data of the target area and 1km soil bulk density raster data;

[0065] The calculation methods of the 13 spectral indices are as follows:

[0066] Salinity index = (Band3×Band4) 0.5 ;

[0067] Salinity index 1 = (Band 3 + Band 4) 0.5 ;

[0068] Salinity index 3 = Band3 × Band3 + Band4 × Band4;

[0069] Salinity index II = (Band2×Band4) / (Band2+Band4);

[0070] Salinity index IX = (Band 3 + Band 4 + Band 8) / 2;

[0071] Normalized difference snow index = (Band8 × Band11) / (Band8 + Band11);

[0072] Optimized soil-adjusted vegetation index = (Band4×Band8) / (Band4+Band8+0.16);

[0073] Weighted difference vegetation index = (Band8-Band4×0.5);

[0074] Salinity index I = (Band4 × Band8) / Band3;

[0075] Brightness index = [(Band4×Band4+Band3×Band3+Band2×Band2) / 3] 0.5 ;

[0076] Soil salinity index 1 = (Band2 × Band4) / Band3;

[0077] Salinity index 4 = (Band2×Band4) 0.5 ;

[0078] Salinity index 2 = (Band 2 + Band 4) 0.5 ;

[0079] Band1 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 1;

[0080] Band2 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 2;

[0081] Band3 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 3;

[0082] Band4 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 4;

[0083] Band5 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 5;

[0084] Band6 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 6;

[0085] Band7 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 7;

[0086] Band8 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 8;

[0087] Band9 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 9;

[0088] Band10 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 10;

[0089] Band11 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 11;

[0090] The input variables of the random forest regression model, namely the reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables, are constructed based on the measured soil pH inversion model. The construction process is as follows:

[0091] a. Collect soil samples from the surface layer of 0.1-20 cm of cultivated land in Songnen Plain. Consider different terrain conditions and soil types when sampling. Dissolve soil samples and water in a mass ratio of 1:5, and then use a pH meter (model PHS-3C) for measurement;

[0092] b. Download the bare soil cloud-free image of the Sentinel-2 satellite MSI sensor from USGS. The downloaded product is Level-2A, and the band is resampled to 10m spatial resolution;

[0093] c. Obtain the reflectance of different bands of MSI remote sensing reflectance and calculate 13 spectral indices;

[0094] d. Obtain the 1km annual average rainfall raster data of the soil, the 30m digital elevation data of the target area, and the 1km soil bulk density raster data;

[0095] e. Based on the measured pH data and remote sensing reflectance data, a random forest regression model of pH value of cultivated land surface soil was established. The correlation analysis was implemented on the SPSS16.0 software platform. The Pearson correlation coefficient r is expressed as follows:

[0096]

[0097] Among them, x and y are the variables pH value and remote sensing reflectance, and N is the number of samples;

[0098] f. The random forest regression model of pH value of cultivated land surface soil is evaluated based on RMSE and MAE, and the specific calculation is as follows:

[0099]

[0100] Where RMSE is the root mean square error, MAE is the mean absolute percentage error, N is the number of samples, and y i is the measured pH value of the surface soil of cultivated land, y i ' is the predicted pH value of the surface soil of cultivated land;

[0101] (3) The output of the random forest regression model is the predicted pH value of the surface soil of cultivated land, which enables rapid mapping of the pH value of the surface soil of cultivated land.

[0102] Specific implementation method 2: This implementation method is different from the specific implementation method 1 in that the 1km annual average rainfall grid data in step (2) is downloaded from the National Meteorological Center. The rest is the same as the specific implementation method 1.

[0103] Specific implementation method 3: This implementation method is different from specific implementation method 1 or 2 in that the 30m digital elevation data of the target area in step (2) is obtained from USGS. The rest is the same as specific implementation method 1 or 2.

[0104] Specific implementation method 4: This implementation method is different from specific implementation methods 1 to 3 in that the 1km soil bulk density grid data in step (2) is obtained from the National Tibetan Plateau Science Data Center. The rest is the same as specific implementation methods 1 to 3.

[0105] Specific implementation method 5: This implementation method is different from specific implementation methods 1 to 4 in that the random forest regression model in step (2) uses the Random Forest package, the number of decision trees (ntree) is 500, and the number of random variables of the split nodes (mtry) is 1 / 3 of the input amount, at which time the value is small and the error in the model is basically stable. The rest is the same as specific implementation methods 1 to 4.

[0106] The following experiments were used to verify the effect of the present invention:

[0107] Experiment 1:

[0108] The MSI-based method for estimating pH value of cultivated land surface soil is carried out in the following steps:

[0109] (1) Obtain bare soil and cloud-free images of the Sentinel-2 satellite MSI sensor Level 2A sensor, resample them to a spatial resolution of 10 m, and obtain remote sensing reflectance in different bands;

[0110] (2) The reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables were used as input variables of the random forest regression model;

[0111] The reflectances of different bands of the MSI remote sensing reflectance are Band 1, Band 2, Band 3, Band 4, Band 5, Band 6, Band 7, Band 8, Band 9, and Band 10;

[0112] The 13 spectral indices are salinity index (SI), salinity index 1 (SI1), salinity index 3 (SI3), salinity index II (S2), salinity index IX (S9), normalized difference snow index (NDSI), optimized soil adjusted vegetation index (OSAVI), weighted difference vegetation index (WDVI), salinity index I (SI-I), brightness index (BI), soil salinity index 1 (SSI1), salinity index 4 (SI4), and salinity index 2 (SI2);

[0113] The environmental variables are 1km annual average rainfall raster data, 30m digital elevation data of the target area and 1km soil bulk density raster data;

[0114] The calculation methods of the 13 spectral indices are as follows:

[0115] Salinity index = (Band3×Band4) 0.5 ;

[0116] Salinity index 1 = (Band 3 + Band 4) 0.5 ;

[0117] Salinity index 3 = Band3 × Band3 + Band4 × Band4;

[0118] Salinity index II = (Band2×Band4) / (Band2+Band4);

[0119] Salinity index IX = (Band 3 + Band 4 + Band 8) / 2;

[0120] Normalized difference snow index = (Band8 × Band11) / (Band8 + Band11);

[0121] Optimized soil-adjusted vegetation index = (Band4×Band8) / (Band4+Band8+0.16);

[0122] Weighted difference vegetation index = (Band8-Band4×0.5);

[0123] Salinity index I = (Band4 × Band8) / Band3;

[0124] Brightness index = [(Band4×Band4+Band3×Band3+Band2×Band2) / 3] 0.5 ;

[0125] Soil salinity index 1 = (Band2 × Band4) / Band3;

[0126] Salinity index 4 = (Band2×Band4) 0.5 ;

[0127] Salinity index 2 = (Band 2 + Band 4) 0.5 ;

[0128] Band1 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 1;

[0129] Band2 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 2;

[0130] Band3 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 3;

[0131] Band4 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 4;

[0132] Band5 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 5;

[0133] Band6 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 6;

[0134] Band7 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 7;

[0135] Band8 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 8;

[0136] Band9 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 9;

[0137] Band10 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 10;

[0138] Band11 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 11;

[0139] The input variables of the random forest regression model, namely the reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables, are constructed based on the measured soil pH inversion model. The construction process is as follows:

[0140] a. Collect soil samples from the surface layer of 0.1-20 cm of cultivated land in Songnen Plain. Considering the different terrain conditions and soil types, the soil samples and water were dissolved in a mass ratio of 1:5, and then measured using a pH meter (PHS-3C);

[0141] b. Download the bare soil cloud-free image of the Sentinel-2 satellite MSI sensor from USGS. The downloaded product is Level-2A, and the band is resampled to 10m spatial resolution;

[0142] c. Obtain the reflectance of different bands of MSI remote sensing reflectance and calculate 13 spectral indices;

[0143] d. Obtain the 1km annual average rainfall raster data of the soil, the 30m digital elevation data of the target area, and the 1km soil bulk density raster data;

[0144] e. Based on the measured pH data and remote sensing reflectance data, a random forest regression model of pH value of cultivated land surface soil was established. The correlation analysis was implemented on the SPSS16.0 software platform. The Pearson correlation coefficient r is expressed as follows:

[0145]

[0146] Among them, x and y are the variables pH value and remote sensing reflectance, and N is the number of samples;

[0147] f. The random forest regression model of pH value of cultivated land surface soil is evaluated based on RMSE and MAE, and the specific calculation is as follows:

[0148]

[0149] Where RMSE is the root mean square error, MAE is the mean absolute percentage error, N is the number of samples, and y i is the measured pH value of the surface soil of cultivated land, y i ' is the predicted pH value of the surface soil of cultivated land;

[0150] (3) The output of the random forest regression model is the predicted pH value of the surface soil of cultivated land, which enables rapid mapping of the pH value of the surface soil of cultivated land.

[0151] The 1km annual average rainfall grid data described in step (2) is downloaded from the National Meteorological Center.

[0152] The 30m digital elevation data of the target area in step (2) is obtained from USGS.

[0153] The 1 km soil bulk density grid data in step (2) was obtained from the National Tibetan Plateau Data Center.

[0154] The random forest regression model in step (2) uses the Random Forest 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.

[0155] In 2023, the pH value of the surface soil of cultivated land is estimated for a total of 67 sampling points. The input variables of the model are the remote sensing reflectance of Band 1 to Band 10, 13 spectral indices and environmental variable data obtained by the Sentinel-2 satellite MSI sensor Level 2A, and the output variable of the model is the pH value of the surface soil of cultivated land.

[0156] The estimated pH value of the cultivated land surface soil was linearly fitted with the measured soil pH value (see Appendix Figure 3 ), the results show that the measured and predicted value fitting points are evenly distributed on both sides of the 1:1 line, R 2 =0.66, RMSE and MAE are 0.66 and 0.56 respectively. The fitting accuracy of soil pH calculated by the method of the present invention and the measured soil pH meets the internationally accepted remote sensing inversion accuracy requirement (<45%).

Claims

1. The method for estimating pH value of cultivated land surface soil based on MSI is characterized by The method for estimating the pH value of the cultivated land surface soil based on MSI is carried out in the following steps: (1) Obtain bare soil and cloud-free images of the Sentinel-2 satellite MSI sensor Level 2A sensor, resample them to a spatial resolution of 10 m, and obtain remote sensing reflectance in different bands; (2) The reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables were used as input variables of the random forest regression model; The reflectances of different bands of the MSI remote sensing reflectance are Band 1, Band 2, Band 3, Band 4, Band 5, Band 6, Band 7, Band 8, Band 9, and Band 10; The 13 spectral indices are salinity index, salinity index 1, salinity index 3, salinity index II, salinity index IX, normalized difference snow index, optimized soil conditioned vegetation index, weighted difference vegetation index, salinity index I, brightness index, soil salinity index 1, salinity index 4, and salinity index 2; The environmental variables are 1km annual average rainfall raster data, 30m digital elevation data of the target area and 1km soil bulk density raster data; The calculation methods of the 13 spectral indices are as follows: Salinity index = (Band3×Band4) 0.5 ; Salinity index 1 = (Band 3 + Band 4) 0.5 ; Salinity index 3 = Band3 × Band3 + Band4 × Band4; Salinity index II = (Band2×Band4) / (Band2+Band4); Salinity index IX = (Band 3 + Band 4 + Band 8) / 2; Normalized difference snow index = (Band8 × Band11) / (Band8 + Band11); Optimized soil-adjusted vegetation index = (Band4×Band8) / (Band4+Band8+0.16); Weighted difference vegetation index = (Band8-Band4×0.5); Salinity index I = (Band4 × Band8) / Band3; Brightness index = [(Band4×Band4+Band3×Band3+Band2×Band2) / 3] 0.5 ; Soil salinity index 1 = (Band2 × Band4) / Band3; Salinity index 4 = (Band2×Band4) 0.5 ; Salinity index 2 = (Band 2 + Band 4) 0.5 ; Band1 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 1; Band2 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 2; Band3 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 3; Band4 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 4; Band5 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 5; Band6 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 6; Band7 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 7; Band8 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 8; Band9 represents the remote sensing reflectivity of the Sentinel-2 satellite MSI band 9; Band10 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 10; Band11 represents the remote sensing reflectivity of Sentinel-2 satellite MSI band 11; The input variables of the random forest regression model, namely the reflectance of different bands of MSI remote sensing reflectance, 13 spectral indices, and environmental variables, are constructed based on the measured soil pH inversion model. The construction process is as follows: a. Collect soil samples from the surface layer of 0.1-20 cm of cultivated land in Songnen Plain, dissolve the soil samples and water in a mass ratio of 1:5, and then measure the pH using a pH meter; b. Download the bare soil cloud-free image of the Sentinel-2 satellite MSI sensor from USGS. The downloaded product is Level-2A, and the band is resampled to 10m spatial resolution; c. Obtain the reflectance of different bands of MSI remote sensing reflectance and calculate 13 spectral indices; d. Obtain the 1km annual average rainfall raster data of the soil, the 30m digital elevation data of the target area, and the 1km soil bulk density raster data; e. Based on the measured pH data and remote sensing reflectance data, a random forest regression model of pH value of cultivated land surface soil was established. The correlation analysis was implemented on the SPSS16.0 software platform. The Pearson correlation coefficient r is expressed as follows: Among them, x and y are the variables pH value and remote sensing reflectance, and N is the number of samples; f. The random forest regression model of pH value of cultivated land surface soil is evaluated based on RMSE and MAE, and the specific calculation is as follows: Where RMSE is the root mean square error, MAE is the mean absolute percentage error, N is the number of samples, and y i is the measured pH value of the surface soil of cultivated land, y i ' is the predicted pH value of the surface soil of cultivated land; (3) The output of the random forest regression model is the predicted pH value of the surface soil of cultivated land, which enables rapid mapping of the pH value of the surface soil of cultivated land.

2. The method for estimating pH value of cultivated land surface soil based on MSI according to claim 1, characterized in that The 1km annual average rainfall grid data described in step (2) is downloaded from the National Meteorological Center.

3. The method for estimating pH value of cultivated land surface soil based on MSI according to claim 1, characterized in that The 30m digital elevation data of the target area in step (2) is obtained from USGS.

4. The method for estimating pH value of cultivated land surface soil based on MSI according to claim 1, characterized in that The 1 km soil bulk density grid data in step (2) was obtained from the National Tibetan Plateau Data Center.

5. The method for estimating pH value of cultivated land surface soil based on MSI according to claim 1, characterized in that The random forest regression model in step (2) uses the Random Forest package, the number of decision trees is 500, and the number of random variables for splitting nodes is 1 / 3 of the input.

Images