Satellite remote sensing modeling method for regional scale soil profile salt content

Abstract

The application discloses a kind of regional scale whole section soil profile salt's satellite remote sensing partition modeling method, the method comprises: obtaining the vegetation index and soil salt index of remote sensing image;According to the soil profile of different depth, the partition factor of partition is determined, and the vegetation index and soil salt index of the remote sensing image are classified to corresponding partition according to partition factor;The vegetation index and soil salt index of remote sensing image in each partition are carried out multiple linear regression to obtain segmented linear function as final model.The application can obtain soil profile salinization information with high precision, provide advanced technical means for development and utilization of saline soil resources, and has positive significance under the background of global population rapid growth.

Description

Technical Field

[0001] This invention relates to the field of satellite remote sensing technology, and more specifically, to a satellite remote sensing zoning modeling method for regional-scale whole-section soil profile salinity. Background Technology

[0002] Saline soils are a widespread soil resource in arid and semi-arid regions. Although these soils are characterized by high salinity, low fertility, limited biodiversity, and low plant productivity (Bui, 2013; Cassel et al., 2015; Chi et al., 2018; Ma et al., 2018; Wang et al., 2018), the rapid population growth has made it difficult for existing arable land to meet the needs of the rapidly growing population and the demands of sustainable social development. As a potential reserve of arable land, saline soils have been extensively reclaimed (Peng et al., 2019; Gutierrez and Johnson, 2010; Metternicht and Zinck, 2003; Yang and Yu, 2017). The remediation of saline soils requires significant amounts of freshwater, and this consumption increases with increasing soil salinity (Corwin et al., 2007). The water consumption required for saline-alkali soil improvement is calculated based on leaching requirements. A leaching requirement refers to the amount of water needed per unit area of ​​soil to leach salts in order to meet desalination standards. These standards are related to the thickness of the desalination soil layer, which is typically 0-60cm, 0-80cm, or 0-100cm (Corwin et al., 2007; Letey et al., 2011). Saline-alkali soils need to be rationally reclaimed based on leaching requirement data to improve the efficiency of limited water resource utilization and balance the needs of ecological and agricultural water use. Therefore, profile soil salinity data is a crucial parameter for calculating saline-alkali soil leaching requirements and an important reference for planning the reclamation of saline-alkali soil resources.

[0003] Traditional soil salinity surveys rely on excavating soil profiles, which is time-consuming, labor-intensive, and inefficient. For large-area surveys, due to the strong spatial variability of soil salinization, relying solely on limited manual sampling points is insufficient to accurately reflect the dynamic spatial distribution of soil salinity (Allbed et al., 2014; Barbouchi et al., 2015; Harti et al., 2016). Satellite remote sensing technology, a rapid, inexpensive, real-time, and dynamic means of monitoring soil properties (Farifteh et al., 2006; Peng et al., 2019), has been widely applied to soil salinization monitoring. Currently, most reports focus on the quantitative monitoring and spatiotemporal variability of surface soil salinity (Fernández-Buces et al, 2006; Weng et al, 2010; Sidike et al, 2014; Scudiero et al, 2015; Muller and Niekerk, 2016; Vermeulen and Niekerk, 2017; Ivushkin et al, 2018; Taghadosi et al, 2019). Only a very few scholars have conducted satellite remote sensing monitoring of soil salinization at different soil layers (20-40cm, 40-60cm, 60-80cm, and 80-100cm) (Taghizadeh-Mehrjardi et al, 2014; Wu et al, 2019). In addition to soil salinity, Mendesetal (2019) used satellite remote sensing data and indoor hyperspectral data to quantitatively study eight physicochemical indicators, including clay content, CEC, and Chroma, in 80-100 cm soil. The model's R-value was [missing information]. 2 Between 0.05 and 0.74. To date, there are very few reports on satellite remote sensing monitoring of soil salinization in whole-section profiles (0-40cm, 0-60cm, 0-80cm and 0-100cm) at depths equal to or greater than 40cm.

[0004] Topsoil salinization exhibits strong spatiotemporal variability and significant differentiation along the vertical profile, leading to considerable uncertainty in monitoring results. Even within the same year, monitoring results can vary considerably between different months (Cho et al., 2018; Herrero and (2015; Harti et al., 2016; Gutierrez and Johnson., 2010) makes it difficult to provide land and resources management departments with stable and reliable decision-making basis for the development and utilization of saline soil resources. However, for arid areas, due to scarce precipitation and long time intervals between adjacent precipitation events, especially in arid areas where single precipitation is usually less than 20 mm, the depth of salt migration in the profile is very limited, and this migration phenomenon usually only occurs in the topsoil. Therefore, the salinization information of soil profiles at a certain depth has relatively weaker spatiotemporal variability than that of topsoil, and the monitoring results are relatively more stable (Zhang et al., 2014). Therefore, how to obtain high-precision soil profile salinization information is a key scientific problem that must be solved in the process of developing and utilizing saline soil resources against the backdrop of rapid global population growth. Summary of the Invention

[0005] To address the shortcomings in the accuracy of soil profile salinization information acquisition, this invention provides a satellite remote sensing zoning modeling method for whole-segment soil profile salinity at a regional scale, effectively solving the technical problems mentioned in the background art.

[0006] The specific technical solution of the present invention is as follows:

[0007] According to the first technical solution of the present invention, a satellite remote sensing zoning modeling method for regional-scale whole-section soil profile salinity is provided, the method comprising:

[0008] Acquire vegetation index and soil salinity index from remote sensing images;

[0009] Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors.

[0010] Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model.

[0011] Furthermore, the vegetation index includes:

[0012] The green light normalization index NG, NG = G / (NIR + R + G);

[0013] Red light normalization index NR, NR=R / (NIR+R+G);

[0014] The near-infrared normalized index NNIR is defined as NNIR = NIR / (NIR + R + G).

[0015] Ratio vegetation index (RVI), RVI = NIR / R;

[0016] Greenness ratio vegetation index GRVI, GRVI = NIR / G;

[0017] Difference Vegetation Index (DVI), DVI = NIR – R;

[0018] Greenness difference vegetation index GDVI, GDVI = NIR – G;

[0019] Normalized Difference Vegetation Index (NDVI): NDVI = (NIR – R) / (NIR + R);

[0020] Green Normalized Difference Vegetation Index (GNDVI), GNDVI = (NIR – G) / (NIR + G);

[0021] Soil-adjusted vegetation index (SAVI): SAVI = (1+L)*(NIR–R) / (NIR+R+L)];

[0022] The soil greenness-modifying vegetation index GSAVI is calculated as follows: GSAVI = (1+L)*[(NIR–G) / (NIR+G+L)];

[0023] Optimize the soil to adjust the vegetation index OSAVI, OSAVI = (NIR – R) / (NIR + R + 0.16);

[0024] Green light optimizes soil and adjusts vegetation index GOSAVI, GOSAVI = (NIR – G) / (NIR + G + 0.16);

[0025] Modified Soil Adjusted Vegetation Index 2MSAVI2

[0026] MSAVI2=0.5*[2*(NIR+1)-SQRT((2*NIR+1)2-8*(NIR-R))];

[0027] Re-regulation vegetation index RDVI, RDVI = SQRT(NDVI*DVI);

[0028] The three-band maximum gradient difference vegetation index (TGDVI) is given by: TGDVI = (NIR – R) / (λ) NIR -λR)-(SWIR–NIR) / (λ SWIR –λ NIR );

[0029] The comprehensive spectral response index, COSRI, is given by: COSRI = (B+G) / (R+NIR)*(NIR–R) / (NIR+R);

[0030] Normalized Difference Moisture Index (NDWI): NDWI = (NIR – SWIR) / (NIR + SWIR);

[0031] Greenness index (GVI)

[0032] GVI=-0.2848*B-0.2435*G+0.5436*R+0.7243*NIR+0.084*SWIR1-0.18*SWIR2;

[0033] The canopy response salinity index (CRSI) is given by: CRSI = SQRT[(NIR*R – G*B) / (NIR*R + G*B)];

[0034] Enhanced Vegetation Index (EVI): EVI = G * [(NIR – R) / (NIR + C1 * R - C2 * B + L)];

[0035] The atmospheric impedance index for green waves is GARI, where GARI = {NIR - [G + γ*(B – R)]} / {NIR + [G + γ*(B – R)]};

[0036] Generalized Difference Vegetation Index (GDVI), GDVI = (NIR) n –R n ) / (NIR n +R n );

[0037] Nonlinear vegetation index NLI, NLI = (NIR) 2 -R) / (NIR 2 +R);

[0038] The soil salinity index includes:

[0039] The salinization index SI-T, SI-T = (R / NIR) * 100;

[0040] Luminance index BI, BI = SQRT(R) 2 +NIR 2 );

[0041] Normalized Difference Salinity Index (NDSI): NDSI = (R - NIR) / (R + NIR);

[0042] Salinity index SI, SI = SQRT(B*R);

[0043] Salinity index 1 SI1, SI1 = SQRT(G*R);

[0044] Salinity index 2 SI2, SI2=SQRT(G2+R2+NIR2);

[0045] The salinity index is SI3, where SI3 = SQRT(G2 + R2).

[0046] Salinization index 1 S1, S1 = B / R;

[0047] Salinization index 2 S2, S2=(BR) / (B+R);

[0048] Salinization index 3 S3, S3=(G*R) / B;

[0049] Salinization index 5 S5, S5=(B*R) / G;

[0050] Salinization index 6 S6, S6=(R*NIR) / G;

[0051] Intensity index 1 Int1, Int1 = (G + R) / 2;

[0052] Intensity index 2 Int2, Int2=(G+R+NIR) / 2;

[0053] Among the various types,

[0054] G represents the reflectance received by the satellite's multispectral sensor in the green band.

[0055] NIR represents the reflectance received by a satellite multispectral sensor in the near-infrared band.

[0056] R represents the reflectance received by the satellite's multispectral sensor in the red band.

[0057] SQRT stands for square root extraction.

[0058] λ SWIR This indicates the center wavelength of the satellite's multispectral sensor in the shortwave infrared band.

[0059] λ NIR This indicates the center wavelength of the satellite's multispectral sensor in the near-infrared band.

[0060] SWIR1 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 1 band.

[0061] SWIR2 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 2 band.

[0062] B represents the reflectivity received by the satellite's multispectral sensor in the red band.

[0063] Furthermore, the step of determining the zoning factors for zones based on soil profiles at different depths, and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zones according to the zoning factors, includes:

[0064] There are four pre-defined partitions: Partition 1, Partition 2, Partition 3, and Partition 4.

[0065] When the soil profile depth is 0-60cm:

[0066] The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.224361, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition.

[0067] The zoning factors for the second zone are satellite remote sensing ground temperature (LST) and digital elevation data (DEM). When LST > 48.61 & DEM <= 1035, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone.

[0068] The zoning factors for the third zone are satellite remote sensing ground temperature (LST) and digital elevation data (DEM). When LST <= 48.61 & DEM <= 1035, the vegetation index and soil salinity index of the remote sensing image are classified into the third zone.

[0069] The zoning factors for the fourth zone are digital elevation data (DEM) and normalized difference vegetation index (NDVI). When DEM > 1035 & NDVI <= 0.224361, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone.

[0070] Furthermore, the step of determining the zoning factors for zones based on soil profiles at different depths, and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zones according to the zoning factors, includes:

[0071] When the soil profile depth is 0-80cm:

[0072] The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition.

[0073] The zoning factors for the second zone are satellite remote sensing ground temperature (LST), digital elevation data (DEM), and normalized difference vegetation index (NDVI). When LST > 52.7477 & DEM > 1022 & NDVI <= 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone.

[0074] The partitioning factor for the third partition is the digital elevation data (DEM). When the DEM <= 1022, the vegetation index and soil salinity index of the remote sensing image are classified into the third partition.

[0075] The zoning factors for the fourth zone are satellite remote sensing ground temperature (LST) and normalized difference vegetation index (NDVI). When LST <= 52.7477 & NDVI <= 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone.

[0076] Furthermore, the step of determining the zoning factors for zones based on soil profiles at different depths, and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zones according to the zoning factors, includes:

[0077] When the soil profile depth is 0-100cm:

[0078] The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition.

[0079] The zoning factors for the second zone are satellite remote sensing ground temperature (LST), digital elevation data (DEM), and normalized difference vegetation index (NDVI). When LST > 52.7477 & DEM > 1022 & NDVI <= 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone.

[0080] The partitioning factor for the third partition is the digital elevation data (DEM). When the DEM <= 1022, the vegetation index and soil salinity index of the remote sensing image are classified into the third partition.

[0081] The zoning factors for the fourth zone are satellite remote sensing ground temperature (LST) and normalized difference vegetation index (NDVI). When LST <= 52.7477 & NDVI <= 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone.

[0082] Furthermore, when the soil profile depth is 0-60cm:

[0083] The regression model for the first partition is expressed as: EC = 34.51461 - 12.6 15_13NDVI - 33 15_14NR - 0.013DEM + 2.1 15_11BI - 0.18 2015TGDVI + 0.04 15_13LST;

[0084] The regression model for the second partition is expressed as: EC = -208.734453 + 366 15_14NR + 26.3 15_11BI + 0.72 15_13LST + 0.058DEM - 13.5 15_13NDVI;

[0085] The regression model for the third partition is expressed as: EC = 75.630928 - 39.2 15_11BI - 7.8 15_13NDVI - 0.006DEM - 4 15_14NR;

[0086] The regression model for the fourth partition is expressed as: EC = 101.650755 + 228 15_14NR + 3.762015TGDVI - 0.145DEM - 25.5 15_00GARI.

[0087] Furthermore, when the soil profile depth is 0-80cm:

[0088] The regression model for the first partition is expressed as: EC = -109.575883 + 443 20151026NR - 0.022DEM + 3 15_11BI - 3.9 15_00GARI;

[0089] The regression model for the second partition is expressed as: EC = 71.659178 - 0.8115_07LST + 2020151026NR - 0.003DEM;

[0090] The regression model for the third partition is expressed as: EC = 55.34712 - 178.1 15_14NDVI + 3.462015TGDVI - 0.36 15_07LST - 0.008DEM + 0.8 15_11BI + 18 20151026NR;

[0091] The regression model for the fourth partition is expressed as: EC = 77.466334 - 91 15_14NDVI - 2.892015TGDVI + 20.7 15_00GARI - 0.044DEM.

[0092] Furthermore, when the soil profile depth is 0-60cm:

[0093] The regression model for the first partition is expressed as: EC = -101.20565 + 377 20151026NR - 0.011DEM - 2.1 15_00GARI + 1 15_11BI;

[0094] The regression model for the second partition is expressed as: EC = 234.33144 - 0.153DEM - 0.915_07LST;

[0095] The regression model for the third partition is expressed as: EC = 36.85798 - 91.1 15_14NDVI - 3.45 15_02TGDVI - 0.008DEM + 21 20151026NR;

[0096] The regression model for the fourth partition is expressed as: EC = 154.16555 - 70.7 15_14NDVI - 0.12 DEM - 0.02 15_07LST.

[0097] According to a second technical solution of the present invention, a satellite remote sensing zoning modeling device for regional-scale whole-section soil profile salinity is provided, comprising a processor configured to:

[0098] Acquire vegetation index and soil salinity index from remote sensing images;

[0099] Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors.

[0100] Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model.

[0101] According to a third technical solution of the present invention, a computer-readable storage medium is provided, characterized in that it stores computer-readable instructions thereon, which, when executed by a computer's processor, cause the computer to perform the method described above.

[0102] The satellite remote sensing zoning modeling method for regional-scale whole-section soil profile salinity disclosed in this invention can obtain soil profile salinization information with high precision, providing advanced technical means for the development and utilization of saline soil resources, and has positive significance in the context of rapid global population growth. Attached Figure Description

[0103] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0104] Figure 1 A flowchart is shown for a satellite remote sensing zoning modeling method for regional-scale whole-segment soil profile salinity according to an embodiment of the present invention.

[0105] Figure 2 A schematic diagram showing the geographical location of the study area and the distribution of sampling points according to an embodiment of the present invention is shown.

[0106] Figure 3a A salt distribution map of a soil profile at a depth of 0-60 cm according to an embodiment of the present invention is shown.

[0107] Figure 3b A salt distribution map of a soil profile at a depth of 0-60 cm according to an embodiment of the present invention is shown.

[0108] Figure 3c A salt distribution map of a soil profile at a depth of 0-60 cm according to an embodiment of the present invention is shown. Detailed Implementation

[0109] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0110] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0111] The invention will now be further described with reference to the accompanying drawings.

[0112] This invention provides a satellite remote sensing method for regional-scale modeling of soil profile salinity across an entire area. The method employs the Cubist approach. Cubist originates from a machine learning method for predictive modeling tools, similar to a classification regression tree, which constructs a linear regression model at different leaves. Therefore, the regression model it constructs is a piecewise linear model. The Cubist model tree is transformed into a series of rules, each generating a related, easily interpretable linear model. This involves simplifying the set of rules derived from the path from the root node to each leaf. Each Cubist rule is a linear model based on an If condition. If the predictor variables associated with an observation satisfy the condition set, the linear model is used to predict that observation dataset. When any observation and its associated predictor variables satisfy multiple rules, the average of the predictions is used as the final predicted value.

[0113] Please see Figure 1 The diagram shown is a flowchart of the method proposed in this invention. The method includes steps S101-103, as detailed below:

[0114] Step S101: Obtain the vegetation index and soil salinity index of the remote sensing image.

[0115] Step S102: Determine the partitioning factor of the partition based on the soil profile at different depths, and classify the vegetation index and soil salinity index of the remote sensing image into the corresponding partition based on the partitioning factor.

[0116] Step S103: Perform multiple linear regression on the vegetation index and soil salinity index of the remote sensing images in each partition to obtain piecewise linear functions as the final model.

[0117] In step S101, the vegetation index and soil salinity index of the remote sensing image on the sampling day are calculated. The detailed calculation formula is shown in Table 1. At the same time, the original band information, vegetation index and soil salinity index are classified.

[0118] Table 1.

[0119]

[0120]

[0121] In the formula, G represents the reflectance received by the satellite multispectral sensor in the green band.

[0122] NIR represents the reflectance received by a satellite multispectral sensor in the near-infrared band.

[0123] R represents the reflectance received by the satellite's multispectral sensor in the red band.

[0124] SQRT stands for square root extraction.

[0125] λ SWIR This indicates the center wavelength of the satellite's multispectral sensor in the shortwave infrared band.

[0126] λ NIR This indicates the center wavelength of the satellite's multispectral sensor in the near-infrared band.

[0127] SWIR1 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 1 band.

[0128] SWIR2 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 2 band.

[0129] B represents the reflectance received by the satellite's multispectral sensor in the red band.

[0130] In step S102, three different depth profiles are selected, namely 0-60cm, 0-80cm, and 0-100cm. For each different depth profile, relevant zoning factors are determined based on the different depth profiles. The vegetation index and soil salinity index of the remote sensing image are then zoned according to preset conditions, as shown in Table 2.

[0131] Table 2.

[0132]

[0133] In step S103, please refer to Table 3. Multiple linear regression is performed based on the data in the partitions formed in Table 2. Finally, the rules and the multiple linear regression model are combined to obtain a piecewise linear function as the final model.

[0134] Table 3.

[0135]

[0136] The following embodiments of the present invention will further illustrate the feasibility and progressiveness of the present invention with specific experimental data.

[0137] The study area selected in this embodiment is located in Xinjiang, China, specifically the Kongtailike alluvial fan in Wensu County, Aksu Prefecture, bordered by the Tianshan Mountains to the north and Alar City to the south (Fig. 1). This study area is a typical desert-oasis transition zone in Xinjiang, belonging to a piedmont alluvial fan, located at 40°40′-41°32′N, 80°36′-81°41′E, with an altitude of 1000-1400m. The terrain is higher in the north and lower in the south and west and lower in the east, with a north-south length of 110km and an east-west width of 80km, forming a fan shape with an area of ​​5600km². 2 It has a continental warm temperate arid climate with an average annual precipitation of 46.4-64.5 mm, concentrated in June, July, and August. The average annual evaporation is as high as 1992.0-2863.4 mm, with an evaporation-to-precipitation ratio of approximately 40:1, making it a typical extremely arid region. It also has abundant light and heat resources, with an average annual total solar radiation of 544-590 kJ·cm². -2 The area receives 2855-2967 hours of sunshine annually, with a frost-free period of 205-219 days (Peng et al, 2019). A north-south provincial highway, S215, runs through the entire study area, while only a few agricultural farm roads run east-west. Land use types include newly reclaimed farmland and desert. Most of the newly reclaimed farmland is in the salinization improvement stage, with only a small portion already cultivated, primarily for cotton. The natural vegetation in the desert is mostly halophytes, such as tamarisk, saltwort, salsa, reeds, and camel thorn, with vegetation cover ranging from 0-100% (Jiang et al, 2019). Vegetation cover shows an increasing trend from south to north. Multiple years of surface surveys indicate that soil salinization initially increases and then decreases from south to north, while from west to east, it initially increases, then decreases, and then increases again (Peng et al, 2019). In recent years, due to the rapid increase in population, a large area of ​​saline soil in the southern part of the study area has been indiscriminately reclaimed into arable land, which requires a large amount of freshwater resources for improvement and irrigation, exacerbating the contradiction between ecological water use and agricultural water use, and endangering the sustainable development of agriculture and the ecological environment in the region.

[0138] Apparent conductivity was measured using an EM38-MK2 geodetic conductivity meter from October 26th to October 28th, 2015, during which all weather conditions were sunny and without rainfall. Considering both vegetation cover and salinity, 30 quadrats with different vegetation cover and salinity levels were established in the study area. When selecting quadrats, areas with relatively uniform vegetation cover and salinity were chosen to avoid heterogeneity issues that could degrade the overall data quality. Based on the spatial resolution of Landsat 8, each quadrat was set to 100m × 100m, approximately 3 pixels × 3 pixels. A grid sampling method was used, with 5 straight lines collected in both north-south and east-west directions for each quadrat, resulting in approximately 1000 measurement points per quadrat, with a point spacing of approximately 1.0m (see [link to sampling method]). Figure 2 ).

[0139] After collecting apparent conductivity data in each quadrat, the quadrat was divided into three regions: high, medium, and low apparent conductivity. One soil column profile (0-60cm, 0-80cm, and 0-100cm) was collected from each region to avoid the concentration of sample data in a narrow range during modeling, which could affect the model's range and reduce its generalizability. A 38mm diameter soil auger was used for sampling, and a total of 270 profile samples at different depths were collected throughout the study area to construct an inversion model between conductivity and apparent conductivity. After converting apparent conductivity to actual conductivity using the inversion model, the average conductivity of all measured points within the same pixel was taken as the final conductivity of that pixel.

[0140] After the collected soil samples were brought back to the laboratory and plant debris, roots and stones were removed, the samples were air-dried in the laboratory and ground through a 2 mm sieve. After drying at 105℃ and cooling naturally, the conductivity was measured using a soil-to-water ratio of 1:5 (Peng et al., 2019).

[0141] To establish a classification standard for desert soil salinization, the soil salinization status and salt-tolerant vegetation growth status in the study area were investigated. Soil salinization status was primarily assessed based on electrical conductivity, while salt-tolerant vegetation growth status was primarily assessed based on vegetation cover and plant diversity. Based on the survey results and referencing the natural soil salinization classification standards of Zhang et al. (2015) and Fernández-Buces et al. (2006), the classification standard for this study area was defined as five levels: non-salinized, slightly salinized, moderately salinized, severely salinized, and saline soil. During the survey, five typical plots were investigated for each salinization level, and five topsoil samples (0-20 cm) were collected from each plot. Vegetation cover and plant species were also investigated. After air-drying, grinding, and sieving, the electrical conductivity of the soil samples was measured using a soil-to-water ratio of 1:5. The desert soil salinization classification standard was established by combining the data on salt-tolerant vegetation growth and topsoil electrical conductivity, as shown in Table 1.

[0142] Table 1 Classification criteria for desert soil salinization

[0143] grade <![CDATA[EC(dSm -1 )]]> Vegetation coverage (%) non-salting ≤7.5 80-100 Mild salinization 7.5-15 40-80 Moderate salinization 15-30 10-40 Severe salinization 30-60 5-10 extreme salinization ≥60 0-5

[0144] In this embodiment, the selected remote sensing data are satellite image data and DEM data of three sensors, Landsat5, Landsat7 and Landsat8. Two scenes of images are required to completely cover the study area. The row and column numbers of the images are 146 / 31 and 146 / 32. A total of 163 scenes of images with cloud cover less than 10% and no snow and ice cover from 1990 to 2015 are collected. Among them, the data selected from August 1990 to March 1999 and from June 2003 to January 2013 are Landsat5, with a total of 92 scenes; the data selected from April 1999 to May 2003 are Landsat7, with a total of 28 scenes; the data selected from February 2013 to October 2015 are Landsat8, with a total of 43 scenes. The Landsat5, Landsat7 and Landsat8 images are downloaded from the website provided by the United States Geological Survey (USGS), and the download URL is https: / / glovis.usgs.gov / . The resolution of the DEM data is 30m, and the download URL is https: / / earthexplorer.usgs.gov / . The processing process of the satellite images is radiometric correction, atmospheric correction, geometric correction and mosaicking. The DEM is filled with depressions. The processed satellite images are further processed for vegetation index, salinity index calculation, surface temperature inversion, and synthesis of maximum, minimum and average values of these parameters over many years.

[0145] Based on the remote sensing data of the study area obtained by the method described above, and then according to the Figure 1 flow method shown in the figure, zoning modeling is carried out, and the performance of the obtained regression model is detected. In this embodiment, four indicators, R2, RMSE, MAE and RPD, are used for evaluation. Generally speaking, the smaller the RMSE and MAE, and the higher the R2 and RPD, the higher the accuracy of the model and the more reliable the performance; on the contrary, it means that the accuracy of the model is lower and the performance is less reliable (Peng et al., 2016). When RPD>2, it means that the model has a high-precision prediction ability. When 1.4<RPD<2, it means that the model only has the ability to distinguish high and low values. When RPD<1.4, it indicates that the model does not have the prediction ability (Nocita et al., 2013; Gomez et al., 2013; Chang and Laird., 2002).

[0146] Using 30 sample areas collected from October 26th to 28th, 2015 (see Figure 2 Using apparent conductivity data from 450 profiles at different depths and indoor conductivity measurements as data sources, three regression models for measured conductivity were constructed using 1m (ECav1m) or 0.5m (ECav0.5m) coils in vertical mode and their apparent conductivity as independent variables (see Table 2). In this experiment, soil samples from five different depths were collected from each of the 30 quadrats, with 90 samples from each depth. During the modeling process, the 90 samples from each profile were sorted according to conductivity, and then sampled at equal intervals. The ratio of the number of samples in the modeling set to the prediction set was 2:1, i.e., 60 samples in the modeling set and 30 samples in the prediction set.

[0147] Significance tests showed that all models in the modeling set reached a highly significant level. However, experiments revealed that the R-squared values ​​of soil models at different depth profiles varied. 2 There are some differences: 0.68-0.79 in the 0-20cm soil profile, 0.75-0.80 in the 0-40cm profile, 0.79-0.86 in the 0-60cm profile, 0.81-0.89 in the 0-80cm profile, and 0.82-0.90 in the 0-100cm profile. It can also be seen that R increases with increasing profile depth. 2 The radius (R) shows an increasing trend, from 0-20cm. 2 The salinity is relatively low in the 0-100cm range and relatively high in the 0-100cm range. This is because ECav1m and ECav0.5m data reflect the overall average salinity information of the 0-100cm and 0-75cm soil profiles, respectively. As depth increases, the salinity information in the profile shows a higher degree of agreement with the information from ECav and ECah, thus exhibiting a higher R... 2 An increasing trend. Among the three models constructed with different independent variables, ECav0.5m was not used alone to construct a 0-100cm profile conductivity model because it can only detect salinity information in the 0-75cm range. Comparing the modeling accuracy of the three models, the bivariate models of ECav and ECah are significantly better than the univariate models, with higher R-values. 2 The R-value is 0.79-0.93, while the R-values ​​of the ECav and ECah models are... 2 The values ​​were 0.68–0.82 and 0.70–0.92, respectively. Therefore, subsequent inversions of soil profile electrical conductivity were performed using bivariate models of ECav and ECah.

[0148] Table 2 Regression model of apparent conductivity and measured conductivity (dS m) -1 (n=60)

[0149]

[0150] The test results of the model using 30 independent samples are shown in Table 3. The prediction results show that, compared with the modeling set, R0 2 The decrease was only slight, indicating that the model is very stable. Meanwhile, the RMSE ranged from 2.31 to 10.70 dS m. -1 Furthermore, the RMSE showed a significant decreasing trend with increasing profile depth. This is mainly attributed to the obvious surface accumulation phenomenon in Xinjiang saline-alkali soils, where most salts accumulate in the 0-20cm depth, and the salt content decreases significantly with increasing profile depth. Among the three models, the RMSE of the bivariate model was significantly lower than that of the univariate model, followed by the ECav0.5m model, while the ECav1m model had the highest RMSE. This is consistent with the modeling set R... 2 The changing patterns are completely consistent.

[0151] The application of electromagnetic induction technology in soil salinization monitoring began in the 1960s (Doolittle and Brevik, 2014). To date, numerous studies have been conducted globally, but most have focused on topsoil or a specific layer within a soil profile (Lesch et al., 2005; Yao and Yang, 2010; Taghizadeh-Mehrjardi et al., 2014; Ding and Yu, 2014; Triantafilis et al., 2013; Triantafilis and Monteiro Santos, 2013; Scudiero et al., 2014; Scudiero et al., 2015; Corwin and Lesch, 2014; Wu et al., 2014), while studies targeting profiles at different depths are relatively few. Compared with similar studies, this study, firstly, starts from the basic principles of electromagnetic induction, enabling a high degree of agreement between soil profile salinity information and apparent conductivity, effectively improving model accuracy. In particular, the salinity information at 0-60cm and 0-80cm depths shows a high degree of matching with ECav0.5m. 2 The accuracy can reach over 0.89. Secondly, this study simultaneously used ECav1m and ECav0.5m information, which significantly improved the model accuracy. Finally, the survey period was concentrated between October 26-28, 2015, with a short cycle and falling within the dry season, effectively avoiding interference from precipitation and temperature. However, there are also some shortcomings. Due to the large workload and heavy tasks of the ground survey, the number of single-depth profile samples collected in each sample area was relatively small, insufficient to construct a sample area-scale model. Therefore, a global modeling approach was adopted. Corwin and Lesch (2014) showed that the accuracy of field-scale models is significantly higher than that of regional-scale models. Therefore, if the number of sample areas is increased and modeling is done using sample areas as the basic unit, higher model accuracy can be obtained.

[0152] Table 3 Comparison of measured conductivity and predicted conductivity (dS m)-1 (n=30)

[0153]

[0154] Conductivity data for different depth profiles in 30 sample areas were obtained using an inversion model. The specific statistics are shown in Table 4. The conductivity of the different depth profiles ranged from 4.14 to 68.07 dS / m. -1 The average value is 22.19-41.55 dS m. -1 According to the salinization grading standard, the 0-60, 0-80, and 0-100 cm depth profiles are classified as moderately salinized, while the 0-20 and 0-40 cm depths are classified as severely salinized; the apparent conductivity of ECav1m is 134.69-1120.23 dS / m. -1 The ECav0.5m range is 143.35-1446.54 dS m. -1 Comparing the conductivity of profiles at different depths reveals a significant decreasing trend with increasing profile depth, with conductivity ranging from 9.54 to 68.07 dS / m in the 0-20 cm depth range. -1 The average value is 41.55 dS m -1 The values ​​for 0-100cm were 4.14-39.02 and 22.19 dS m, respectively. -1 The apparent electrical conductivity of ECav0.5m and ECav1m was also statistically analyzed. The results showed that the apparent electrical conductivity of ECav0.5m was higher than that of ECav1m, corroborating each other and indicating a significant surface accumulation of soil salinization in this study area. Furthermore, the coefficient of variation showed a significant increasing trend with depth, increasing from 33.27% in 0-20cm to 37.28% in 0-100cm. This suggests that soil profiles with similar surface electrical conductivity may exhibit significant differences in electrical conductivity at deeper layers as depth increases. Therefore, the relationship between surface soil electrical conductivity and deep soil electrical conductivity is uncertain; some soils with high surface electrical conductivity also have relatively high deep soil electrical conductivity, while others have low deep soil electrical conductivity.

[0155] Table 4. Descriptive statistics of soil profile characteristics (dS m) -1 )

[0156]

[0157] This paper extracts 55 surface parameters, including NDVI, EVI, and SI, from geostationary satellite imagery and analyzes their correlation with soil electrical conductivity. Autocorrelation tests were also performed. Based on this, considering the workload, only six surface parameters with the best correlation—NR, TGDVI, NDVI, LST, GARI, and TC1—were selected for time-series analysis. TGDVI, NDVI, NR, and GARI primarily reflect the dynamic changes in vegetation cover. Among these four parameters, TGDVI, NDVI, and GARI show a positive correlation with vegetation cover. Considering the interference of accidental factors such as grazing, pests, and drought on vegetation cover, their multi-year maximum values ​​were calculated. NR shows a negative correlation with vegetation cover, and its multi-year minimum value was calculated similarly. TC1 and LST primarily reflect the dynamic changes in brightness and temperature in bare soil areas, so their multi-year maximum values ​​were also calculated. Table 5 shows the correlation between electrical conductivity at different depth profiles and the optimal time scale for surface parameters. Table 5 shows that with increasing soil profile depth, the time period corresponding to the highest correlation coefficient between electrical conductivity and the maximum or minimum values ​​of surface parameters shows a significant increasing trend: 5-10 years for 0-60 cm, and 7-18 years for 0-80 cm and 0-100 cm. With increasing profile depth, the correlation between surface parameters and electrical conductivity shows a significant weakening trend, mainly because satellite remote sensing can only detect information from the soil surface layer. Among these six parameters, although the correlation with electrical conductivity reached a highly significant level, NR, TGDVI, and NDVI were superior to LST, GARI, and TC1, with NR showing the strongest correlation and TC1 the weakest.

[0158] Table 5. Maximum correlation coefficient between extreme values ​​of parameters synthesized at different durations and soil profile electrical conductivity.

[0159]

[0160] Table 6 shows the Cubist subset partitioning rules. According to Table 7, the partitioning of the 0-60cm, 0-80cm, and 0-100cm subsets is mainly based on the values ​​of NDVI, DEM, and LST. Each profile is divided into four subsets, but the partitioning rules differ. It can also be observed that NR, TGDVI, and TC1 did not participate in subset partitioning, indicating that these three factors did not differ significantly in the study area, or their differences were similar to, but less significant than, the selected partitioning factors. Furthermore, comparing the subset partitioning factors of soil profiles at different depths shows that the more soil profile depths, the more subset partitioning factors are involved, and the subset partitioning of deeper soil profiles is often constrained by multiple factors. Among the selected partitioning factors, NDVI participated in the partitioning of all soil profiles, DEM participated in the partitioning of all soil profiles except the 0-20cm soil profile, and LST only participated in the partitioning of the 0-60cm, 0-80cm, and 0-100cm soil profiles.

[0161] Table 6. Zoning factors and rules for soil profiles at different depths.

[0162]

[0163]

[0164] Table 7 lists the contribution rates of the factors involved in the zoning. In the 0-60cm soil profile, DEM had the highest contribution rate, followed by LST, while NDVI had the lowest. The 0-80cm and 0-100cm soil profiles showed the same pattern: NDVI had the highest contribution rate, followed by LST, and DEM had the lowest. Among all three factors involved in the zoning, no trend of increasing or decreasing contribution rate with increasing soil profile depth was observed. This is mainly due to two factors: firstly, soil salinity exhibits strong spatial variability in the profile; and secondly, the time scales of the same factor selected for different depth profiles are not entirely consistent.

[0165] Table 7. Zonal Contribution Rate of Surface Parameters from Soil Profiles at Different Depths

[0166] Soil profile NDVI LST DEM 0-60cm 44％ 56％ 84％ 0-80cm 70％ 50％ 44％ 0-100cm 68％ 49％ 46％

[0167] Table 8 shows the Cubist model results for conductivity at different depth profiles. The Cubist model at different depth profiles achieves high R-values ​​on the modeling set. 2 There were no significant differences in dS m values ​​between 0.88 and 0.90, and the RMSE ranged from 2.65 to 4.28 dS m. -1 MAE in 2.03-3.33 dS m -1 RMSE and MAE show a decreasing trend with increasing profile depth, mainly due to the decrease in electrical conductivity with increasing profile depth. In the prediction set, compared to the modeling set, the RMSE values ​​for the 0-20cm, 0-40cm, and 0-100cm soil profiles are significantly higher. 2 There was no significant decrease, but slight decreases were observed in the 0-60cm and 0-80cm ranges, from 0.90 and 0.89 to 0.83 and 0.80 respectively. RMSE ranged from 2.97 to 5.08 dS m. -1 MAE is in the range of 2.28-3.92 dS m -1 All soil profiles showed an RPD greater than 2.0, with the highest being 3.00 in the 0-40cm section and the lowest being 2.10 in the 0-80cm section. According to Williams (2001)'s evaluation criteria, the 0-20, 0-40, 0-60, and 0-100cm soil profiles all met the requirements of RPD ≥ 2.0 and 0.82 ≤ R. 2The condition ≤0.90 indicates that the corresponding profile model has good predictive ability, while the 0-80cm soil profile can only satisfy RPD≥1.5 and 0.66≤R 2 Under the condition ≤0.81, the model only has a rough predictive ability. To verify the model's predictive performance, the average conductivity distribution maps of different depth profiles in the study area generated by the model were specifically calculated. The average conductivity values ​​for five profiles from 0-20 cm to 0-100 cm were 40.91, 31.96, 26.75, 24.36, and 21.92 dS / m, respectively. -1 The results are very close to the average value of the actual sampling points, indicating that the model has good reliability.

[0168] Table 8. Performance of cube models for soil profiles at different depths.

[0169]

[0170] According to the soil salinization classification standard, the electrical conductivity of soil profiles at different depths in the study area was divided into five levels: non-salinized, slightly salinized, moderately salinized, severely salinized, and saline soil. The salinization status of soil profiles at different depths is shown in [reference needed]. Figure 3a -c. According to Figure 3aAs shown by -c, the degree of salinization decreases significantly with increasing soil profile depth. For example, areas with severe salinization or saline soil in the 0-20cm profile decrease to moderate or slight salinization in the 0-100cm profile. This is consistent with the strong surface accumulation characteristic of soil salinization in this area, and also indirectly demonstrates the high reliability of the results. There is a trend of increased salinization from north to south and from west to east. This is mainly related to the topography of the study area, which is high in the north and low in the south and west and low in the east. Salt accumulates in relatively low-lying areas under the influence of surface runoff, leading to severe salinization in the southeast over a long period. The degree of salinization in vegetation-covered areas and around water systems is significantly lower than in other areas, which is completely consistent with the results of our multiple field investigations. Field survey results show that areas with high vegetation cover are generally located in non-salinized or slightly salinized areas, while sparse vegetation is generally located in moderately salinized areas. In severely salinized and saline soil areas, vegetation struggles to grow, resulting in mostly bare soil. Areas around water systems are typically slightly or moderately salinized, but their impact is limited to within 50 meters. High-coverage shrubs and herbaceous plants, such as tamarisk, saltwort, and reeds, are common around water systems. In the southern part of the study area, there are large areas of newly reclaimed farmland, all of which have undergone varying degrees of improvement. However, the salinity of the top 0-20cm soil layer in some of these newly reclaimed farmlands is significantly higher than in surrounding areas, classifying them as saline soil. This area mainly consists of farmland reclaimed 1-2 years ago. Summer flooding is used for salinization improvement, but the area has extremely high evaporation rates. In autumn and winter, water-soluble salts accumulate in the topsoil layer with the evaporation of water, forming significant secondary salinization. A 5-10mm white salt crust is visible during the surface survey.

[0171] Figure 3a -c shows the area and percentage of different salinization levels in soil profiles at different depths. (Source: [Insert Source Here]) Figure 3a As can be seen from -c, in the soil profiles at three different depths, the 0-60cm, 0-80cm, and 0-100cm profiles were mainly characterized by moderate salinization, with areas and percentages of 1099.19 km², respectively. 2 (49.09%), 1422.5km 2 (63.54%) and 1902.18km 2(84.96%). The percentage of non-saline soil was extremely low, ranging from only 0.1% to 3.04%, and its area and percentage showed a significant upward trend with increasing profile depth. However, compared to the 0-60cm soil profile, the area and percentage of non-saline soil decreased significantly in the 0-80cm section, from 1.83% to 0.13%. According to the ground survey results, there was a distinct salt accumulation layer in the 60-80cm vegetation cover area, with a significantly higher electrical conductivity than the adjacent soil layers. The percentage of slightly saline soil in the 0-20cm and 0-40cm profiles was extremely low, at only 1.03% and 2.13%, respectively. The percentages in the 0-60cm, 0-80cm, and 0-100cm profiles were roughly the same, all around 11%. The percentage of moderately salinized soils increased significantly with increasing profile depth, from 13.77% to 84.96%, while the percentage of severely salinized soils showed the opposite trend, decreasing from 84.27% to 0.13%.

[0172] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A satellite remote sensing zoning modeling method for salinity of a whole-section soil profile at a regional scale, characterized in that, The method includes: Acquire vegetation index and soil salinity index from remote sensing images; Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors. Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model. When the soil profile depth is 0-60cm, the step of determining the zoning factors for different soil profile depths, and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zoning based on the zoning factors, includes: There are four pre-defined partitions: Partition 1, Partition 2, Partition 3, and Partition 4. The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.224361, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition. The zoning factors for the second zone are satellite remote sensing ground temperature (LST) and digital elevation data (DEM). If LST > 48.61 & DEM <= 1035, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone. The zoning factors for the third zone are satellite remote sensing ground temperature (LST) and digital elevation data (DEM). If LST <= 48.61 & DEM <= 1035, the vegetation index and soil salinity index of the remote sensing image are classified into the third zone. The zoning factors for the fourth zone are digital elevation data (DEM) and normalized difference vegetation index (NDVI). When DEM > 1035 & NDVI <= 0.224361, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone. When the soil profile depth is 0-60cm: The regression model for the first partition is expressed as: EC = 34.51461 - 12.6×15_13NDVI - 33×15_14NR - 0.013×DEM + 2.1×15_11BI - 0.18×2015TGDVI + 0.04×15_13LST; The regression model for the second partition is expressed as: EC = -208.734453 + 366×15_14NR + 26.3×15_11BI + 0.72×15_13LST + 0.058×DEM - 13.5×15_13NDVI; The regression model for the third partition is expressed as: EC = 75.630928 - 39.2×15_11BI - 7.8×15_13NDVI - 0.006×DEM - 4×15_14NR; The regression model for the fourth partition is expressed as: EC = 101.650755 + 228×15_14NR + 3.76×2015TGDVI - 0.145×DEM - 25.5×15_00GARI; NR is the red light normalization index, BI is the brightness index, TGDVI is the three-band maximum gradient difference vegetation index, and GARI is the green wave atmospheric impedance index.

2. The method according to claim 1, characterized in that, The vegetation index includes: The green light normalization index NG is defined as NG = G / (NIR + R + G). Red light normalization index NR, NR=R / (NIR+R+G); The near-infrared normalized index NNIR is defined as NNIR = NIR / (NIR + R + G). Ratio vegetation index (RVI), RVI = NIR / R; Greenness ratio, vegetation index GRVI, GRVI = NIR / G; Difference Vegetation Index (DVI), DVI = NIR–R; Greenness difference vegetation index GDVI, GDVI=NIR–G; Normalized Difference Vegetation Index (NDVI): NDVI = (NIR – R) / (NIR + R); Green Normalized Difference Vegetation Index (GNDVI), GNDVI = (NIR–G) / (NIR+G); Soil-adjusted vegetation index (SAVI), SAVI = 1.5 * (NIR – R) / (NIR + R + 0.5); The soil greenness moderating vegetation index GSAVI is calculated as follows: GSAVI = (1.5)*[(NIR–G) / (NIR+G+0.5)]; Optimize the soil-adjusted vegetation index OSAVI, OSAVI = (NIR – R) / (NIR + R + 0.16); Green light optimizes soil and adjusts vegetation index GOSAVI, GOSAVI = (NIR–G) / (NIR+G+0.16); Modified Soil Adjusted Vegetation Index (MSAVI2) MSAVI2=0.5*[2*(NIR+1)-SQRT((2*NIR+1) 2 -8*(NIR-R))]; The normalized vegetation index RDVI is calculated as: RDVI = SQRT(NDVI*DVI). The three-band maximum gradient difference vegetation index (TGDVI) is calculated as: TGDVI = (NIR–R) / (λ). NIR -λ R )-(SWIR–NIR) / (λ SWIR –λ NIR ); The comprehensive spectral response index, COSRI, is calculated as: COSRI = (B+G) / (R+NIR)*(NIR–R) / (NIR+R). Normalized Difference Moisture Index (NDWI): NDWI = (NIR – SWIR) / (NIR + SWIR); Greenness index (GVI) GVI=-0.2848*B-0.2435*G+0.5436*R+0.7243*NIR+0.084*SWIR1-0.18*SWIR2; The canopy response salinity index (CRSI) is given by: CRSI = SQRT[(NIR*R–G*B) / (NIR*R+G*B)]; Enhanced vegetation index (EVI): EVI = G*[(NIR–R) / (NIR+6*R-7.5*B+1)]; The atmospheric impedance index for green waves is GARI, where GARI = {NIR - [G + 1.7 * (B – R)]} / {NIR + [G + 1.7 * (B – R)]}. Generalized Difference Vegetation Index (GDVI), GDVI = NIR–R; Nonlinear vegetation index NLI, NLI = (NIR) 2 -R) / (NIR 2 +R); The soil salinity index includes: Salinization index SI-T, SI-T = (R / NIR)*100; Luminance index BI, BI = SQRT(R) 2 +NIR 2 ); Normalized Difference Salinity Index (NDSI): NDSI = (R - NIR) / (R + NIR); Salinity index SI, SI = SQRT(B*R); Salinity index 1 SI1, SI1 = SQRT(G*R); Salinity index 2 SI2, SI2=SQRT(G 2 +R 2 +NIR 2 ); Salinity index 3 SI3, SI3=SQRT(G) 2 +R 2 ); Salinization index 1 S1, S1=B / R; Salinization index 2 S2, S2 = (BR) / (B+R); Salinization index 3 S3, S3 = (G*R) / B; Salinization index 5 S5, S5 = (B*R) / G; Salinization index 6 S6, S6 = (R*NIR) / G; Intensity index 1 Int1, Int1 = (G+R) / 2; Intensity index 2 Int2, Int2 = (G+R+NIR) / 2; In the formula, G represents the reflectance received by the satellite multispectral sensor in the green band. NIR represents the reflectance received by a satellite multispectral sensor in the near-infrared band. R represents the reflectance received by the satellite's multispectral sensor in the red band. SWIR represents the reflectance received by a satellite multispectral sensor in the shortwave infrared band. SQRT stands for square root extraction. λ SWIR This indicates the center wavelength of the satellite's multispectral sensor in the shortwave infrared band. λ NIR This indicates the center wavelength of the satellite's multispectral sensor in the near-infrared band. λ R This indicates the center wavelength of the satellite's multispectral sensor in the red band. SWIR1 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 1 band. SWIR2 represents the reflectance received by the satellite's multispectral sensor in the shortwave infrared 2 band. B represents the reflectance received by the satellite's multispectral sensor in the blue band.

3. A satellite remote sensing zoning modeling method for salinity of a whole-section soil profile at a regional scale, characterized in that, The method includes: Acquire vegetation index and soil salinity index from remote sensing images; Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors. Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model. When the soil profile depth is 0-80cm, the step of determining the zoning factor for each zone based on soil profiles at different depths, and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zones based on the zoning factor, includes: Four partitions are preset: Partition 1, Partition 2, Partition 3, and Partition 4. The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition. The zoning factors for the second zone are satellite remote sensing ground temperature (LST), digital elevation data (DEM), and normalized difference vegetation index (NDVI). When LST > 52.7477 & DEM > 1022 & NDVI <= 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone. The partitioning factor for the third partition is the digital elevation data (DEM). If the DEM <= 1022, the vegetation index and soil salinity index of the remote sensing image are classified into the third partition. The zoning factors for the fourth zone are satellite remote sensing ground temperature (LST) and normalized difference vegetation index (NDVI). When LST <= 52.7477 & NDVI <= 0.177036, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone. When the soil profile depth is 0-80cm: The regression model for the first partition is expressed as: EC = -109.575883 + 443×20151026NR - 0.022×DEM + 3×15_11BI - 3.9×15_00GARI; The regression model for the second partition is expressed as: EC = 71.659178 - 0.81×15_07LST + 2020151026NR - 0.003 DEM; The regression model for the third partition is expressed as: EC = 55.34712 - 178.1×15_14NDVI + 3.46×2015TGDVI - 0.36×15_07LST - 0.008×DEM + 0.8×15_11BI + 18×20151026NR; The regression model for the fourth partition is expressed as: EC = 77.466334 - 91×15_14NDVI - 2.89×2015TGDVI + 20.7×15_00GARI - 0.044×DEM; NR is the red light normalization index, BI is the brightness index, TGDVI is the three-band maximum gradient difference vegetation index, and GARI is the green wave atmospheric impedance index.

4. A satellite remote sensing zoning modeling method for salinity of a whole-section soil profile at a regional scale, characterized in that, The method includes: Acquire vegetation index and soil salinity index from remote sensing images; Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors. Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model. When the soil profile depth is 0-100cm, the process of determining the zoning factors for different soil profile depths and classifying the vegetation index and soil salinity index of the remote sensing image into the corresponding zoning based on the zoning factors includes: There are four pre-defined partitions: Partition 1, Partition 2, Partition 3, and Partition 4. The partitioning factor for the first partition is the Normalized Difference Vegetation Index (NDVI). When NDVI > 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the first partition. The zoning factors for the second zone are satellite remote sensing ground temperature (LST), digital elevation data (DEM), and normalized difference vegetation index (NDVI). When LST > 52.7477 & DEM > 1022 & NDVI <= 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the second zone. The partitioning factor for the third partition is the digital elevation data (DEM). If the DEM <= 1022, the vegetation index and soil salinity index of the remote sensing image are classified into the third partition. The zoning factors for the fourth zone are satellite remote sensing ground temperature (LST) and normalized difference vegetation index (NDVI). When LST <= 52.7477 & NDVI <= 0.214, the vegetation index and soil salinity index of the remote sensing image are classified into the fourth zone. When the soil profile depth is 0-100cm: The regression model for the first partition is expressed as: EC = -101.20565 + 377×20151026NR - 0.011×DEM - 2.1×15_00GARI + 1×15_11BI; The regression model for the second partition is expressed as: EC = 234.33144 - 0.153×DEM - 0.9×15_07LST; The regression model for the third partition is expressed as: EC = 36.85798 - 91.1×15_14NDVI - 3.45×15_02TGDVI - 0.008 DEM + 21×20151026NR; The regression model for the fourth partition is expressed as: EC = 154.16555 - 70.7×15_14NDVI - 0.12×DEM - 0.02×15_07LST; NR is the red light normalization index, BI is the brightness index, TGDVI is the three-band maximum gradient difference vegetation index, and GARI is the green wave atmospheric impedance index.

5. A satellite remote sensing zoning modeling device for regional-scale whole-segment soil profile salinity, used to implement the method described in any one of claims 1-4, characterized in that, Includes a processor, the processor being configured to: Acquire vegetation index and soil salinity index from remote sensing images; Based on soil profiles at different depths, the zoning factors of each zone are determined, and the vegetation index and soil salinity index of the remote sensing image are classified into the corresponding zones according to the zoning factors. Multiple linear regression was performed on the vegetation index and soil salinity index of the remote sensing images in each zone to obtain piecewise linear functions as the final model.

6. A computer-readable storage medium, characterized in that, It stores computer-readable instructions that, when executed by the processor of a computer, cause the computer to perform the method described in any one of claims 1-4.

Images