Spatial variation of soil nutrients and evaluation of cultivated land quality based on field scale

3.1 Determination of SVSN and message preprocessing

At present, the commonly used methods for soil granule analysis are mainly the pipette method and specific gravity method [11,12]. Both methods have advantages and disadvantages. The overall operation of the former is too intricate; however, the results obtained have high accuracy; The operation of the latter is very simple, but the final result is low. The pipette method is mainly used to analyze the soil granule size content to obtain the spatial variation of many nutrient laws in the soil. The main operation steps are shown in Figure 1.

Figure 1

Flow chart of soil spatial variation analysis.

According to Figure 1, the specific steps are as follows:

1. Step 1: Pretreat the experimental sample;

2. Step 2: Prepare suspension;

3. Step 3: According to the experimental requirements, the suspension containing soil granules is specially prepared by suction pipe, and the granules with different diameters are divided according to the granule size;

4. Step 4: Collect;

5. Step 5: Drying;

6. Step 6: Weigh;

7. Step 7: Calculate the percentage of soil granule content in different granule sizes.

Potassium dichromate method belongs to the volume analysis method, which is mainly used to detect soil organic nutrients. If the soil contains excessive sulfuric acid, the organic carbon in the soil can be oxidized by the oxidizing agent potassium dichromate. The remaining oxidizing agent potassium dichromate can be titrated by the standard ferrous sulfate solution. The organic carbon content can be calculated by the amount of oxidizing agent potassium dichromate consumed, and the organic carbon content can be multiplied by the nutrient coefficient to obtain the organic nutrients in the soil [13,14].

Soil color is mainly affected by different physical and chemical properties. Soil color is determined mainly by Munsell colorimetric method. It has standardized the color of the soil. Its working principle for the soil color solution sampler is shown in Figure 2.

Figure 2

Working principle diagram of soil solution sampler by Munsell Colorimetry.

Field-scale technology plays a very important role in soil sample determination. The results based on field scale evaluation can be extended to medium- and large-scale evaluation, and then, the accuracy of medium- and large-scale evaluation is improved. It mainly uses the field scale to extract the sample message and makes quantitative and qualitative analyses of the physical properties and division properties of the sample. Generally, the analyzed specimen does not need pretreatment. It mainly has the following advantages:

1. It is suitable for fast nondestructive testing.

2. It will not consume chemical reagents and will not pollute the environment.

3. It is suitable for online analysis.

The CLQ on the field scale is closely related to farmers’ behavior, which includes management investment, planting selection, resource utilization, and technology application. In the management of cultivated land, farmers mainly affect land quality, garden land, and other agricultural lands through the choice of planting and management investment behavior (including fertilization, pesticide application, pipette returning to the field, and irrigation). Soil quality evaluation based on field scale can provide farmers with a timely message, which is conducive to the implementation of scientific and reasonable management measures for cultivated land. Nevertheless, due to the wide spectral band, it will lead to serious overlap and low absorption strength. Traditional methods cannot be used to analyze the spectrum through the human senses. Therefore, it is necessary to determine soil nutrients with stoichiometric methods. The specific calculation expression is as follows:

In formula (1), K expresses the measured value of factor exponent, K 1 expresses the upper limit of factor exponent level difference, and K 2 expresses the lower limit of the factor exponent level difference.

The operation steps of the field scale processing method are as follows:

The original signal obtained by the instrument contains information on physical and chemical structures and signals composed of other interfering factors. These signals will affect the accuracy of the final analysis. These noises can be effectively eliminated through stoichiometry. Therefore, the main purpose of on-site scale processing is to minimize non-informational factors through stoichiometry, thus laying a solid foundation for establishing calibration models and predicting unknown samples. In addition, to determine the density of field sampling points, it is necessary to combine the actual situation of the region and use a secondary soil survey to digitize and overlay soil maps, administrative zoning maps, and land use status maps to form an evaluation unit map. Based on the principles of comprehensiveness, balance, and objectivity, using the evaluation unit map as the base map, considering factors such as the number of patches, area, planting system, crop type, and yield level, the number and location of distribution points are determined. At the same time, the distribution method is determined based on the shape and size of the evaluation unit, and a GPS positioning device is used to determine the longitude, latitude, and altitude of the sampling points. It records the farmer information, management information, irrigation conditions, crop yield, and fertilizer usage of the plot where the sampling point is located.

3.2 Build of soil nutritive material message extraction model

In spectral chemometrics, it is the hinge to establish a high-reliability and high-accuracy quantitative correction model. There is an inner relevance among the field scale and quality parameters of the measured specimen. It manages the multispectral and quality parameters through mathematical methods and clarifies the quantitative correction model among them, that is, the correction model. The quality parameters of unknown specimens can be forecasted through the calibration model and spectra of unknown samples. The modeling of calibration model is mainly divided into two different forms, namely:

1. Linear method;

2. Nonlinear method.

Then, the method is used to decompose it, and the main components are analyzed by multiple linear analysis, which realizes the principal component analysis (PCA). It is a new multidimensional statistical method, which is eigenvector shadow and eigenvalue dissociation. The main idea is to reduce the dimension of the sampling spectrum by converting the original variation into some new variable. The correlation degree of the new variables is perpendicular, but not related to each other, which can effectively eliminate the overlap of spectral information. PCA decomposes the spectral array W into the following forms:

In formula (2), E expresses the score array, then E = ( e 1 , e 2 , ⋯ , e n ) ; R s expresses the load array, that is, R s = ( r s 1 , r s 2 , ⋯ , r s n ) . Each point vector is perpendicular, and the load vector is also perpendicular. The length value of the load vector is 1.

Ignoring the shadow of spectral array W on the last few mounts vectors, the PCA dissociation expression W ′ of W can be obtained, that is:

In formula (3), E ′ expresses the shadow score array, R s ′ expresses the shadow load array, and Q g ′ expresses the measured value of projected soil nutrients. In practical application, the values of E ′ , R s ′ , and Q g ′ are significantly lower than E , R s , and Q g , respectively. Therefore, PCA can effectively compress data and obtain a number of feature extraction at the same time. The whole method has the advantages of fast calculation speed and simple operation process. The principal component regression model is obtained by mixed logistic regression through the score vector array Y U , that is:

In formula (4), D B expresses the least square solution and S a expresses the factor exponent.

In the unknown prediction, PCA is used to get the load array and the fractional component, and the principal component regression model is used to get the prediction result of the unknown sample [15,16,17]. When using field-scale techniques to analyze complex mixes such as agricultural products, land, and food products, it is necessary to use the collected samples to build modified models, thus effectively reducing the complexity of the models. Selecting representative calibration set samples can not only reduce the workload but also improve the adaptability and accuracy of the model. In general, an ideal set of corrections must conform to the following conditions:

1. The classification of the calibration center sample basically includes all the chemical components contained in the unknown sample.

2. When this model is used for analysis, the range of consistency change of the calibrated sample group should be larger than that of the unknown sample size.

3. The consistency of the components is consistent in the overall consistent range of variation.

4. Calibration groups must have sufficient samples to accurately determine the mathematical correlation and consistency of spectral variables in statistical processing.

Due to the need to use a large number of soil samples in practical use, the above model is used to automatically correct the samples [18]. Using the spectral shadow method to select characteristic wavelengths can retain the information of spectral samples to the maximum extent and prevent the overlap of information effectively. The surface is dominated by organic matter, minerals, water, and air. Due to the differences in the physical and chemical characteristics of the surface, its transmission, reflection, absorption, and other characteristics are also different. This is a spectral property of the surface. The physical and chemical composition of the surface has a great influence on the spectral characteristics of the surface [19].

Based on the above analysis, combined with the field scale, the soil nutrient message extraction model is constructed. The core of the field scale mainly comes from the labeled data in the source field, which can accurately predict the classification model of the test data category in the target field. On account of the field scale, a soil nutrient message extraction model is constructed, namely:

In formula (5), X r expresses the weight ratio, and C 1 , C 2 , C 3 , and C 4 respectively represent the grade membership of the exponent percentage of soil quality. Its structure is shown in Figure 3.

Figure 3

Soil nutrient message extraction model.

Figure 3 shows that the soil nutrient message extraction model is composed of five functional modules, and the specific functional structure is as follows:

1. Soil nutrient information extraction model: This model is used to extract nutrient information from soil data to inform decision-makers [20].

2. Model function module: As the core part of the model, the function module contains various functions required for information extraction.

3. Data input module: The data input module inputs the original soil data into the model for processing. It mainly includes the following sub-modules: Data management module: responsible for the storage, sorting, and updating of soil data; Data query module: used to find specific soil nutrient information on demand; Data analysis module: Analyzes and interprets the input data, including statistical analysis, prediction, and other functions; Data output module: output the processed soil nutrient information in visual or other easy-to-understand forms.

4. Arc GIS: ArcGIS platform: This software platform is used to configure and manage the model, establish spatial database and attribute database, and extract soil nutrient information. ArcGIS is a powerful geographic information system software, which can realize the integration, query, analysis, and visualization of spatial data, and provide a convenient and quick operating platform for our soil nutrient information extraction model.

5. Spatial database: Geospatial data used to store and manage soil nutrient information. Attribute database: non-spatial attribute data used to store and manage soil nutrient information.

The model established by combining formula (5) and Figure 3 can effectively extract soil nutrient messages.

3.3 Realize the SVSN and the Evaluation of CLQ at the Field Scale

In the deep cultivation of cultivated land, the bottom layer of the pear is damaged at intervals to form alternating structures, which can alleviate the phenomenon of cultivated land soil degradation and help maintain the fertility of cultivated land soil. It can also improve the gas exchange capacity, ensure that the cultivated soil fully absorbs nutrients and increases the yield of crops. The damage principle of the subsoiling shovel to the soil is shown in Figure 4.

Figure 4

Schematic diagram of damage principle of subsoiling shovel.

According to Figure 4, the whole cutting can be regarded as a process in which the soil is divided into different shapes by external forces. The failure mode of soil is also different with different soil depths. Cultivated land can be regarded as soil without boundary, but limited by the simulation model; it is impossible to establish a large soil model. The soil physical and chemical elements damaged by the subsoiling shovel are analyzed based on the simulation of relevant restrains. Taking the soil's physical and chemical elements as the indicators for calculating the CLQ, selecting the membership function of the CLQ indicators, determining its limit value, the membership degree set is obtained. The limit value of the membership function of the specific CLQ indicators is shown in Table 1.

The value of soil exponent is related to land quality, and there is a positive correlation between them. The total phosphorus amount is effective, with a value range of 0–550. The total potassium content includes slow-acting potassium (0–3,000 mg/kg) and fast-acting potassium (0–500 mg/kg). If its membership degree is lower than the lower limit or higher than the upper limit, it has little impact on land quality, then the membership function of the soil exponent can be seen as an ascending half trapezoidal distribution. Its membership function mode is as shown in formula (6):

In formula (6), K ( α ) expresses the membership function of exponent ω ; α expresses the exponent value of CLQ, and α 0 expresses the upper critical value of exponent membership. The membership degree of each exponent is calculated according to formula (3), the weight of each exponent is determined by PCA, and the exponent load array is calculated. Dividing the obtained array by the square root of the corresponding exponent value of the principal component to obtain the corresponding coefficient of each exponent, and the obtained coefficient is the weight of each exponent. The expression for calculating the exponent load array is as follows:

In formula (7), a n m expresses the CLQ exponent. The calculated exponent coefficient is shown in formula (8).

In formula (8), V v expresses the load factor and S s expresses the coefficient corresponding to the CLQ exponent, that is, the exponent weight. After obtaining the weight, combined with the SVSN, the weighted sum method is used to calculate the CLQ exponent. The expression is as follows:

In formula (9), S I expresses the CLQ exponent; s i expresses the weight corresponding to the CLQ exponent; k i expresses the membership value corresponding to the exponent, and P l expresses the stampede grade score. The final S I value range is among [ 0 , 1 ] . The closer the calculated S I value is to 1, the greater the impact of trampling on CLQ. On the contrary, the closer the S I value is to 0, the smaller the impact. So far, the research and design of SVSN and CLQ evaluation based on field scale has been completed.

3.4 Experimental analysis

To verify the impact of SVSN and evaluate CLQ based on a field scale, it designed an experiment. D city was selected as the research object. It is located in northern China, with a flat terrain. The GPS coordinates are located between 37°24′–38°57′ and 112°34′–114°33′E. The climate belongs to a warm temperate semi-humid continental monsoon climate. During the same period of hot and rainy weather, the annual average temperature is 13.5°C, the average precipitation is 650 mm, and the frost-free period is about 220 days. The main crops include wheat, corn, grapes, and watermelon. The soil resources are classified according to the soil classification system, mainly including loess soil, yellow brown soil, brown soil, purple soil, yellow sea soil, gully fill soil, silt, mudflat, and other types. Loess soil is one of the most important soil types, which is widely distributed. Yellow brown soil is mainly distributed in the study area and its surrounding areas. Purple soil is mainly distributed in the southern mountainous areas south of the study area; The Yellow Sea soil is mainly distributed in the plain and Hetao areas; Gully fill soil is mainly distributed in intermountain basins and river valleys in the north and southwest of the study area, while silt and mudflat are mainly distributed in rivers, lakes and coastal areas. There are significant regional differences in soil resources. Based on the administrative division map and land use status map of the research area, a deep understanding of the natural conditions and social development status of the research area will be obtained. It sets the standard as 6 km of the research area × A uniform grid of 6 km, with at least one point arranged in each grid, totaling 80 points. The selection of soil sampling depth and thickness depends on different land use types, plant types, and purposes. Conventional farmland, orchards, vegetable fields, etc., are sampled at a depth of 30 cm and a thickness of 10 cm. The sampling depth for cultivated soil is generally 28 cm, with sampling depth intervals of 10 cm and sampling thickness of 10–15 cm, mainly considering functional areas such as farmland and orchards. The soil fraction preprocessing uses the domain value method combined with the Tyson polygon method to identify and process the specific values of nutrient data, as shown in Figure 5.

Figure 5

Distribution of sampling points.

The soil-saturated hydraulic conductivity, porosity, and soil bulk density are measured by the ring knife method, and the porosity is measured by controlling the water level of soaking the ring knife within the range of 2–3 mm. The capillary porosity can be calculated by the following expression:

In formula (10), H 2 - 3 expresses the weight of the ring knife with soil in 2–3 h of water absorption; H d expresses the weight of the ring knife; H g expresses the weight of dry soil in the ring knife, and H h expresses the volume of the ring knife.

3.5 Analysis method

When analyzing the spatial structure characteristics of soil nutrients, spatial autocorrelation detection and semi-variance function structure were selected for feature analysis. In spatial autocorrelation analysis, the global Moran’s I index was selected for detection, and the semi-variance function was used to measure the variability of random phenomena at different spatial distances. The semi-variance function structure refers to the variation patterns of the semi-variance function within a certain range, including stationarity, trendiness, and periodicity. By viewing and drawing the image of the semi-variance function, characterization and analysis can be carried out. The value of the semi-variance function increased with the increase of spatial distance and showed a stable trend within a certain range. The structural characteristics can be understood by observing the image of the semi-variance function. By analyzing the variability and image characteristics of the semi-variance function, its stationarity can be determined. Periodicity indicates that the semi-variance function exhibits periodic fluctuations within a certain distance range, which can be determined through periodic analysis methods. The analysis of the coefficient of variation B b is the main content of the classical statistical soil nutrient theory. The coefficient of variation B b can be calculated by the following expression:

In formula (11), B ¯ expresses the average value corresponding to the physical properties of soil nutrients; B expresses the standard deviation corresponding to the physical properties of soil nutrients. The coefficient of variation B b is graded according to the degree of variation. The expression belongs to strong variability when the coefficient of variation B b is greater than or equal to 100%. When the coefficient of variation B b is less than or equal to 10%, it belongs to weak variability. When the coefficient of variation B b is 10–100%, it belongs to medium variation.

Let Z x be a random function, the semivariogram assumes that the variance of Z x exists and is limited. The mean value exists and is stable, and the spacing h will affect the function Z ( x i ) . The expression is as follows:

In formula (12), Z x and z ( x i + h ) represent the measured values of regional variation Z x at dimensional positions x and x i + h , respectively; n expresses the number of sampling points when the interval distance is h , which can analyze the SVSN physical properties.

4.1 Soil bulk density results

Soil bulk density will affect soil solute migration characteristics, water holding capacity, and infiltration capacity. Under natural conditions, soil bulk density will be comprehensively affected by farming, biological action, soil-forming process, and other elements, with high dimensional heterogeneity. The soil bulk density in different soil layers is analyzed by semi-variogram, and the analysis results are shown in Table 2.

According to the data in Table 2, the variation characteristics of soil bulk density can be described by the soil nutrient message extraction model under 0–5 cm soil layer. The nugget coefficient test was 21.82%. The variation occurred due to its own structural reasons, and the DA was strong at this time. In the 5–10 cm soil layer, the exponential function model was used to respond to the variation characteristics of bulk density, and the DA was strong at this time. The results of the semivariogram of soil bulk density within 0–20 cm are shown in Figure 6.

Figure 6

Effect of sampling amplitude on coefficient of variation of bulk density: (a) h ₈ = 0−10 cm and (b) h ₈ = 10−20 cm.

As can be seen from Figure 6, with the increase of soil sampling range, the relationship between the variation coefficient of soil bulk density and sampling range is not linear. In the sampling area of 40 m × 20 m, the fluctuation coefficients of both soil layers reach the maximum value. However, when the sampling range increases, the fluctuation coefficient will decrease. Then, with the expansion of the sampling range, the fluctuation coefficient increases gradually. According to previous research results, it can be seen that the fluctuation coefficient will show an upward trend with the increase of sampling range, mainly because more and more variation factors are taken into account with the expansion of the research area.

4.2 Soil porosity results

The soil non-capillary porosity, capillary porosity, and total porosity of different soil layers on the reclamation slope of the waste dump were analyzed by semi-variogram. The analysis results are shown in Table 3.

According to the data in Table 3, in addition to the non-capillary porosity of 0–5 cm soil layer and the capillary porosity of the 5–10 cm soil layer, the DA of the non-capillary porosity of the 5–10 cm soil layer, the capillary porosity of 0–5 cm soil layer and the total porosity of different soil layers were strong, indicating that their own structural reasons are the main elements leading to variation. Gaussian model and exponential model were used to describe the above variation characteristics. The autocorrelation of soil non-capillary porosity in 0–5 cm soil layer and capillary porosity in 5–10 cm soil layer was medium, which can be described by the variation characteristics of exponential model and high-speed model.

4.3 Soil water content results

The SVSN in the reclamation slope of the waste dump was high, and the soil formation and hydrological formation were greatly affected by the variability of soil nutrients. The semi-variogram parameters of water content in different soil layers are shown in Table 4.

From Table 4, the autocorrelation of soil water content in different soil layers was strong, indicating that it is the main cause of variation due to its own structural reasons. The variability characteristics of soil water content can be described by the Gaussian model.

In conclusion, the spatial variation of soil porosity and soil bulk density in different soil layers is the same, indicating that the water conductivity of soil and the number of soil voids are affected by the change in soil bulk density (Figure 7).

Figure 7

Optimal semivariance function of soil salinity: (a) organic matter, (b) total nitrogen, (c) available potassium, and (d) available phosphorus.

The points of soil nutrients in the range of variation in the study area have all approached the theoretical model. The adaptation model of soil total nitrogen and available potassium was an exponential model, while the adaptation model of the other three soil nutrients was a spherical model. The nugget value of the analysis result model reflected the possible degree of randomness within the regionalized variables, which means that the spatial characteristics of the analyzed variables may be highly influenced by random factors such as human interference. In addition, the base value was the largest variation in regional system attributes, caused by the combined effects of structural and random factors, representing the range of variance changes caused by the spatial autocorrelation of regional variables. In the study area, the lump gold coefficients of soil organic matter, available potassium, and total nitrogen were 43.1, 30.8, and 65.2%, respectively. This indicated that the spatial gradient of soil organic matter was relatively poor, and the sudden change in its content range was mainly caused by random factors. Range refers to the spatial autocorrelation scale of regional variables. Within the range of range, points closer in space are more correlated, while points separated by a distance greater than the range do not have a correlation.

Figure 8 shows the spatial distribution variation of cultivated land resources in the study area. This is mainly manifested in two aspects: First, due to differences in geological and climatic conditions, there are a variety of soil types, and there are significant differences between these soil types. This results in different regions having different soils that are suitable for crops and different types of agriculture. Second, there are significant differences in fertility, water storage capacity, and growth capacity of different soils. For example, soil fertility in the loess region is low, but water storage capacity is strong. However, due to the high groundwater level, the drainage capacity of meadow soil is poor and the growth capacity is relatively weak. The yellow soil fertility was low in some areas, and the fertility index was 2. The soil fertility of sandstone in some areas is higher, and the fertility index is 8. In some areas, the meadow soil fertility was moderate, and the fertility index was 5. The clay fertility was also higher in some areas, with a fertility index of 7. The water storage capacity of yellow soil was strong in some areas, and the water separation and retention index was 8. The water storage capacity of sandstone soil in some areas is weak, and the water separation and retention index is 3. In some areas, the water storage capacity of the meadow soil was moderate, and the water separation and retention index was 5. In some areas, the clay storage capacity is strong, and the water retention index is 7. The growth ability of yellow soil was weak in some areas, and the growth index was 3. The growth ability of sandstone soil in some areas is strong, and the growth index is 7. In some areas, the growth ability of the meadow soil was moderate, and the growth index was 5. Some areas of clay growth ability are strong, and the growth index is 8.

Figure 8

Spatial distribution of cultivated land resources.