County-Scale Spatial Distribution of Soil Nutrients and Driving Factors in Semiarid Loess Plateau Farmland, China

Abstract

Characterized by a topography of thousands of ravines, the Loess Plateau has highly complex spatial variability in terms of soil nutrients. Therefore, it is of considerable importance to study the soil nutrient spatial distribution, driving factors of precise fertilizer management, and the strategic use of soil nutrient resources. In 2017, 242 soil samples were taken from the semiarid Anding district farming region in northern China. The spatial variability and factors influencing soil nutrients were studied using statistical and geostatistical methods. The results showed that the mean soil organic matter (SOM), total nitrogen (TN), available phosphorus (AP), available potassium (AK), and pH values were averaged at 12.64 g·kg⁻¹, 0.84 g·kg⁻¹, 23.20 mg·kg⁻¹, 188.87 mg·kg⁻¹, and 8.60, respectively. The nugget-to-sill ratios for the semi-variograms of SOM, TN, AP, and AK varied from 25.84 to 49.93%, while the coefficients of variation varied from 24.53 to 69.44%, revealing that all four indicators exhibited considerable variability, and that the samples’ geographical variability was produced by a combination of random and structural factors. Overall increasing trends were exhibited from the middle to the northeast and southwest in the distributions of SOM, TN, and AP. The spatial distribution of AK displayed the opposite trend, increasing from the southwest to north and southeast. The texture of the tillage layer was the main factor directly affecting SOM, and explained 8% of its variation. The distribution of TN was mainly influenced by the irrigation method and water source type. AP and AK contents differed significantly between the two parent materials, three textures, and three topography types at the level of p < 0.01. In conclusion, the regional soil fertility was poor, spatial heterogeneity was moderate, and influencing factors were complex, highlighting the need to adopt precise fertilization management and adopting land management measures according to the actual influencing factors of each nutrient, thereby contributing to the enhancement of regional fertility.

1. Introduction

Soil nutrients are vital for supporting plant growth and development and are essential for maintaining the quality of arable land and ensuring food production [1,2]. The effective maintenance and long-term stability of nutrient factors in soil are important for protecting the regional ecological environment and maintaining ecosystem stability [3]. Soil nutrients are influenced by multiple elements, such as climate, parent material, biology, time, and human activities. These variables might result in complicated spatial variability at different scales [4,5]. The soil nutrient heterogeneity directly affects soil productivity, and a detailed understanding of their spatial distribution is crucial for precise nutrient management and appropriate fertilization strategies [6,7]. Therefore, a thorough investigation of the regional soil nutrient variability can serve as the foundation for implementing variable fertilization techniques, which guarantees high crop yield and quality. It is of considerable importance for the scientific development of farm fertilization programs, improved nutrient resource use, and implementation of precise fertilization practices.

Geostatistics, regression trees, neural networks, linear models, and fuzzy systems have been used to analyze soil nutrient variability in recent years, and these methods have proven to be effective tools for understanding nutrient dynamics in the field [8,9,10]. In general, geostatistics are used to forecast nutrient variability patterns in space [11,12]. With the continuous development of geostatistics, methods such as inverse distance weighting, the radial basis function, and ordinary kriging interpolation have been widely used, and their prediction accuracies have been gradually improved [13,14,15].

Recently, researchers have conducted numerous studies on soil nutrient spatial variability using geostatistics. Duffera et al. [16] conducted a study on soil properties’ distribution in the coastal plain of the southeastern United States. Similarly, Zhao et al. [17] investigated the variability and pattern in space of soil nutrients in the Mun River Basin of Thailand, and identified terrain, vegetation, and human disturbance as the primary factors influencing soil nutrient variations in the basin. Several studies have also been carried out in China. Gao et al. [18] collected 555 soil samples in Renshou County, located in the Sichuan Basin of China, and analyzed the spatial heterogeneity of TN, P, and K. Another study by Wu et al. [19] examined the dispersion of soil organic carbon and soil nutrients in the context of Linyi City, which is located in the southeastern region of the central and southern mountains within Shandong Province, China, under diverse land-use classifications. Previous global examinations concerning the spatial heterogeneity of soil nutrients have merged geostatistical methods with geographic information systems (GIS). However, they have predominantly focused on mountains, plains, and basins, and studies of the Loess Plateau area are rare. Some experts have conducted a significant amount of research on soil microbial community structure [20], farmland environmental risk assessment [21], and farming practices [22] in the farmland area of the semiarid Loess Plateau. However, there are fewer studies on the spatial distribution of regional soil nutrients.

Characterized by a topography of thousands of ravines, the semiarid Loess Plateau has highly complex spatial variability in terms of soil nutrients. In recent years, owing to continuous potato crops and improper fertilization, the soil quality of farmland has declined, which, in turn, has affected crop growth and quality. The precise comprehension of soil nutrient spatial variability in the semiarid Loess Plateau holds significant importance for devising efficient strategies for land management. By comprehending how soil nutrients are distributed across different areas, it becomes possible to implement sustainable practices that promote appropriate soil nutrient management. Additionally, this knowledge allows for the precise application of fertilizers, leading to improved soil quality. However, up until now, there has been a lack of extensive investigation regarding the arrangement of soil nutrients in space and the influencing elements at the county level within the semiarid agricultural regions of the Loess Plateau. The Dryland Farming Area of Central Gansu is in northwest China and is located on the semiarid Loess Plateau of China. It is a representative dry rainfed agricultural area that is mainly planted with potatoes and maize. Therefore, the aims of this study were to (i) study the distribution characteristics in space of soil nutrients at the county level in the semiarid area of Loess Plateau using a combination of classical statistics and geostatistics, and (ii) identify the driving factors influencing soil nutrients heterogeneity to provide scientific guidance for precise fertilization management and farmland management measures in the semiarid Loess Plateau.

2. Materials and Methods

2.1. Study Area

The study was conducted in a farmland area between 35°17′ and 36°02′ N and 104°12′ and 105°01′ E, with an altitude of 1700 to 2580 m. The study area was located in the Anding District in Dingxi city of Gansu Province, northwestern China, on the semiarid Loess Plateau (Figure 1), with a total area of 3.6 × 10⁵ hm² and an arable land area of 1.5 × 10⁵ hm². The semiarid western Loess Plateau has a hilly and mountainous landscape [23]. The climate of the region is typical for a temperate continental climate, with an average annual temperature of 6.3 °C and an average annual precipitation of 400 mm. The months with the highest rainfall are July, August, and September. Annual cumulative air temperature >10 °C is 2240 °C, and frost-free period is approximately 130 days, with 2480 h of sunshine per year [23,24]. In this study, the soil types included black loess soils, sierozems, and loess soils. The area is a typical dry-farming, rainfed agricultural area mainly for planting, with potatoes and corn being the main crops.

2.2. Soil Sampling and Analysis

The research article titled “The 30 m annual land cover and its dynamics in China from 1990 to 2019” [25] provided the dataset for Chinese land use and land cover from which the land use remote sensing data were derived. The map of the cultivated land in the Anding District was used as the base map, and the sampling points were all located within the farmland area (Figure 1). Two-hundred-and-forty-two soil sampling sites were established throughout the district, and each site was captured using GPS. Soil samples (0–20 cm) were collected in March 2017 according to the principles of “random-”, “equal volume-”, and “multi-point mixing”. Samples were sent back to the laboratory for weighing, numbering, air-drying, and were then sieved and mixed. Soil nutrient indicators were analyzed using conventional analysis methods. The potassium dichromate volumetric method was utilized to quantify SOM. TN was measured using the Kjeldahl method, AP using the molybdenum antimony anti-colorimetric method with NaHCO₃ (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) extraction, and AK using absorption spectrophotometry method with CH₃COONH₄ (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) extraction. Soil pH was calculated using a pH meter with a water to soil ratio of 2.5:1.

2.3. Statistical and Geostatistical Analysis

The data were statistically analyzed via SPSS (SPSS for Windows, Version 26.0; SPSS, Chicago, IL, USA). The soil nutrient indices in the study area were graded according to the classification standard for soil nutrients during the second China National Soil Survey [26]. The Shapiro–Wilk (S-W) test in SPSS was used to evaluate the normal distribution for each collective. To fulfill the requirements for semi-variance function analysis, data that were not normally distributed were submitted to Box-Cox transformation in Minitab (Version 20.0; State College, PA, USA).

Semi-variograms were produced for each soil variable in the geostatistical study using the GS+ program (Version 9.0; Gamma Delta, Plainwell, MI, USA) as follows:

$$\mathsf{\gamma }\left(h\right)=\frac{1}{2N\left(h\right)}\sum _{i=1}^{N\left(h\right)}{\left[z\left(x_i\right)-z\left(x_i+h\right)\right]}^2$$

(1)

where z(x_i) is the measured value of the soil property at sampled points of x_i, h is the lag distance in meters, r(h) is the variogram between observations z(x_i) and z(x_i + h) at a lag distance of h, and N(h) is the number of pairs of sample points separated by h.

The experimental semi-variogram was fitted with relevant theoretical models, including linear, exponential, Gaussian, and spherical models based on the maximum regression coefficients of determination (R²) and minimal residual sum of squares (RSS) values. Three significant metrics were used for describing the spatial dependency of variables in a semi-variogram plot. A nugget (C₀) is defined as variability on a smaller scale than the sample interval, as well as sampling and analytical error. The partial sill (C) displays the amount of spatial structural variation, and the range denotes the distance at which the semi-variogram stabilizes around a limiting value. The nugget-to-sill ratio (C₀/C₀ + C) represents the geographic dependency of soil parameters, and is frequently used to characterize varied correlation in space. A ratio of less than 25% denotes strong geographic correlation, a ratio of 25% to 75% denotes moderate correlation, and a ratio larger than 75% denotes weak correlation [27]. Based on the calculation of the semi-variance parameters and best-fit models, ordinary kriging interpolation uses the geostatistical analysis module in ArcGIS 10.8 software (ESRI, Redlands, CA, USA) to create a spatial distribution of soil nutrients by interpolating each element.

Pearson correlation analysis was conducted using SPSS software (version 26.0; IBM Inc., Armonk, NY, USA) to analyze the relationships among soil nutrients (i.e., SOM, TN, AP, and AK) and physicochemical properties (i.e., pH and bulk density [BD]), as well as natural conditions and human factors.

One-way analysis of variance (ANOVA) was applied to test the potential differences in soil nutrient variables among soil types, parent materials, textures, topography, irrigation methods, or water source types. The soil types were divided into three types: black loess soils, sierozems, and loess soils; the parent materials were divided into two types: alluvial–pluvial deposits and loess; the topography types were divided into three types: basin, hill, and mountain; the textures of the cultivated layer were divided into three types: light clay, medium loam, and heavy loam; the irrigation methods were divided into four types: furrow, flood, and border irrigation, and none, with three water source types: surface water, groundwater, or none. When there was a significant difference (p < 0.05) in the effect of natural conditions or anthropogenic factors on soil nutrients, the four soil nutrient indices of different soil types, parent materials, topography, texture, irrigation methods, and water sources were compared using the least significant difference (LSD) multiple comparisons at the 5% significance level. ANOVA and multiple comparisons were performed using SPSS. Data are reported as the means ± SD.

A structural equation model (SEM) was used to estimate the main factors influencing SOM variation. SEM is a statistical tool that utilizes multivariate covariance to construct models, which tests the pathways of influence among various variables [28]. Based on the correlation analysis, the factors affecting SOM were selected, and the structural model of this study was established using empirical modeling. This was followed by model fitting, evaluation, and correction to obtain the SEM of the interrelationship between SOM and natural conditions (parent material and tillage texture), human factors (irrigation method and water source type), and soil physicochemical properties (pH and BD). SEM analyses were carried out using AMOS 26.0 (Amos Development Co., Armonk, NY, USA).

3. Results

3.1. Descriptive Statistical Analysis

The means of SOM, TN, AP, and AK content were 12.64 g·kg⁻¹, 0.84 g·kg⁻¹, 23.20 mg·kg⁻¹, and 188.87 mg·kg⁻¹, respectively (

Table 1. The result of descriptive statistics.

Index	Samples	Minimum	Maximum	Mean	SD	CV/%	Skewness	Kurtosis
SOM/(g∙kg−1)	242	7.38	25.70	12.64	3.10	24.53	1.54	3.42
TN/(g∙kg−1)	242	0.42	8.05	0.84	0.50	59.52	12.60	182.00
AP/(mg∙kg−1)	242	0.40	146.00	23.20	16.11	69.44	2.57	13.98
AK/(mg∙kg−1)	242	25.00	586.00	188.87	90.48	47.91	1.27	1.73
pH	242	8.06	8.96	8.60	0.15	1.74	−0.34	−0.41

SOM: soil organic matter; TN: total nitrogen; AP: available phosphorus, AK: available potassium. SD: standard deviation; CV: coefficient of variation.

). The average contents of SOM and TN in this research region were low, while the average content of other nutrients was mid-range, as determined by the categorization standard of soil nutrients used during the second China National Soil Survey. At the experimental site, the average soil pH was 8.60. The coefficient of variation (CV) of SOM, TN, AP, and AK were 24.53%, 59.52%, 69.44%, and 47.91%, respectively, and the CV of pH was 1.74%. The highest CV value was that of AP, while the lowest was that of pH. Based on the skewness and kurtosis indices of variables and the S–W test, the pH values followed a normal distribution. To analyze the spatial distribution, it was necessary to transform raw data with non-normal distribution into data with a normal distribution. In this study, we found that the logarithmic transformation of soil nutrients greatly enhanced their normality.

3.2. Spatial Variability Analysis

The log-transformed data of soil nutrients exhibited a normal distribution. Therefore, the experimental semi-variograms were calculated based on the log-transformed data of SOC, TN, AP, and AK.

Table 2. Semi-variogram models and parameters.

Index	Nugget C0	Sill C0 + C	C0/(C0 + C)/%	Range A/km	Determination Coefficient R2	Residual Square RSS	Theoretical Model
SOM	0.0185	0.0531	34.84	0.2010	0.956	6.424 × 10−5	Spherical model
TN	0.0377	0.0755	49.93	0.2286	0.814	5.215 × 10−4	Gaussian model
AP	0.0920	0.5710	34.55	0.1440	0.905	0.014	Exponential model
AK	0.0658	0.2546	25.84	0.9750	0.905	2.218 × 10−3	Exponential model

summarizes the best-fit models and their parameters that match the experimental semi-variograms. The function curve of soil nutrient content variation is shown in Figure 2. The results show that the fitting curve of the spherical model function for SOM was suitable, that of the Gaussian model function for TN was suitable, and that the models for AP and AK were both exponential models. The determination coefficient (R²) value of the semi-variogram model for TN was 0.814; the values for SOM, AP, and AK exceeded 0.9; and the residual sums of squares (RSS) for the semi-variograms of the four soil nutrients were close to zero. These results indicate that the soil nutrient spatial structure in this study was well-suited to the selected models of semi-variograms. Further analysis showed that SOM, TN, AP, and AK contents were found to be spatially autocorrelated within a specific range. However, the specific performances were different. The nugget variance (C₀) of the soil nutrient variables ranged from 0.0185 to 0.0920, which is less than 0.1, and the nugget values were AP > AK > TN > SOM. Sill variance (C + C₀) was in the range of 0.053 to 0.571. The nugget-to-sill ratios for SOM, TN, AP, and AK were 34.84%, 49.93%, 34.55%, and 25.84%, respectively, which were between 25% and 75%. The soil nutrients’ semi-variograms fluctuated between 144.0 and 975.0 m.

3.3. Spatial Distribution Characteristics

Based on the fitted semi-variogram models, as shown in Figure 3, ordinary kriging was used to map the soil nutrient spatial distributions. The SOM and TN in this area ranged from 8.86 to 20.68 g·kg⁻¹ and 0.56 to 1.40 g·kg⁻¹, respectively. High values of SOM and TN were predominantly distributed in the southwestern and northeastern corners. Meanwhile, low concentrations of these soil nutrients were sporadically scattered in the central-western part of the region (Figure 3A,B). AP content distribution maps showed a large number of patchy distributions, with the high values between 40.00 mg·kg⁻¹ and 59.03 mg·kg⁻¹ distributed in the southwest, and the low values between 3.59 mg·kg⁻¹ and 5.00 mg·kg⁻¹ sporadically scattered in the mid-western and northeast regions (Figure 3C). All three soil nutrient content distributions showed an overall increasing trend from the middle to the northeast and southwest, with patchy distributions in some areas. The spatial distribution of AK displayed opposite characteristics, with the range of 48.25–452.49 mg·kg⁻¹ for the entire area. However, low AK contents (48.25–100 mg·kg⁻¹) were found only in the southwestern area. Meanwhile, high values of AK were mainly concentrated in large areas of the north and southeast (Figure 3D).

3.4. Factors Affecting the Spatial Variability of SOM and Soil Nutrients

Table 3. Differences in SOM and soil nutrients under different conditions.

Influence Factor		Sample Size	SOM/(g∙kg−1)	TN/(g∙kg−1)	AP/(mg∙kg−1)	AK/(mg∙kg−1)
Soil type	Black loessial soil	194	12.55 ± 2.84 a	0.81 ± 0.18 a	23.81 ± 16.73 a	183.54 ± 91.13 a
	Sierozems	2	10.95 ± 0.49 a	0.86 ± 0.15 a	32.85 ± 6.43 a	221.99 ± 18.38 a
	Loessial soil	46	13.09 ± 4.06 a	0.94 ± 1.09 a	20.21 ± 13.14 a	209.96 ± 87.16 a
F-value			0.855	1.245	1.291	1.723
Parent material	Alluvial–pluvial deposits	53	13.95 ± 3.00 a	0.91 ± 0.17 a	30.31 ± 24.33 a	115.26 ± 19.25 b
	Loess parent material	189	12.27 ± 3.04 b	0.82 ± 0.56 a	21.21 ± 12.27 b	209.51 ± 91.84 a
F-value			12.670 ***	1.361	13.925 ***	54.965 ***
Topography type	Basin	17	12.12 ± 2.25 a	0.84 ± 0.18 a	19.94 ± 34.83 ab	119.29 ± 18.81 c
	Hill	92	12.97 ± 3.90 a	0.86 ± 0.79 a	19.15 ± 15.88 b	212.14 ± 91.64 a
	Mountain	133	12.48 ± 2.52 a	0.82 ± 0.16 a	26.43 ± 11.37 a	181.67 ± 89.79 b
F-value			0.932	0.158	6.222 **	9.058 ***
Soil texture	Light clay	18	11.65 ± 0.84 bc	0.82 ± 0.04 a	19.52 ± 10.07 bc	136.61 ± 7.88 bc
	Medium loam	174	12.34 ± 3.11 b	0.82 ± 0.58 a	21.66 ± 12.46 b	212.70 ± 94.67 a
	Heavy loam	50	14.04 ± 3.21 a	0.91 ± 0.03 a	29.89 ± 25.17 a	124.75 ± 37.24 c
F-value			7.147 **	0.633	5.809 **	26.090 ***
Irrigation method	Furrow irrigation	7	12.61 ± 4.22 abc	1.78 ± 2.77 a	34.61 ± 20.05 a	199.29 ± 30.78 ab
	Flood irrigation	20	14.93 ± 3.79 a	0.92 ± 0.21 bcd	25.55 ± 16.61 ab	123.67 ± 50.20 b
	Border irrigation	12	12.59 ± 1.95 bc	0.84 ± 0.20 cd	31.41 ± 10.48 ab	158.06 ± 73.90 ab
	None	203	12.41 ± 2.97 c	0.80 ± 0.18 d	22.09 ± 15.98 b	196.76 ± 93.07 a
F-value			4.116 **	9.901 ***	2.734 *	4.676 **
Water source type	Surface water	7	12.61 ± 4.22 ab	1.78 ± 2.77 a	34.61 ± 20.05 a	199.26 ± 30.77 ab
	Groundwater	32	14.05 ± 3.39 a	0.89 ± 0.21 bc	27.75 ± 14.71 ab	136.56 ± 61.39 b
	None	203	12.42 ± 2.97 b	0.80 ± 0.18 c	22.09 ± 15.98 b	196.76 ± 93.07 a
F-value			3.915 *	14.803 ***	3.593 *	6.443 **

*, **, *** indicate significance at the 0.05, 0.01, and 0.001 levels, respectively. Different lowercase letters represent significant differences among different conditions of influence factors (p < 0.05). Data are means ± SD.

displays the results of an ANOVA test that examined the impacts of soil nutrients in different conditions. There were no significant differences (p > 0.05) between the soil types in terms of SOM, TN, AP, or AK. Different parent materials had an impact on soil nutrients, and the SOM, AP and AK levels of the two types of parent materials were significantly different (p < 0.001). The SOM and AP contents in alluvial–pluvial deposits were significantly higher than those in the loess parent material by 13.69% and 42.90%, respectively. The contents of AK were significantly lower than those in loess parent material by 81.77%. Meanwhile, the parent materials had no significant differences in TN content (p > 0.05). The AP and AK were significantly different among the three topography types, whereas there were no significant differences among the topography types in the SOM and TN. Table 3 shows that the AP content of mountains was significantly higher than that of hills at 38.02%, and the AK content differed significantly among the three terrain types in the following order: hills > mountains > basins. SOM and TN contents in the hills were somewhat higher than those in the basins and mountains. However, the differences were not significant, indicating that the changes in SOM and TN contents in the hills and basins were largely synchronized. The SOM, AP, and AK contents, except for TN, differed significantly among the three textures at the level of p < 0.01. Among them, SOM and AP were significantly higher in heavy loam than in medium loam and light clay, whereas AK was significantly higher in medium loam than in heavy and light clay. Irrigation methods and water source types had significant effects (p < 0.05) on the contents of all four soil nutrients. The SOM content was highest under flood irrigation, with an average content of 14.93 g∙kg⁻¹, and all other soil nutrients were highest under furrow irrigation. The TN content was significantly higher under furrow irrigation than the other three methods. The highest SOM content was 14.05 g∙kg⁻¹ under groundwater irrigation, and the highest TN, AP, and AK contents were found under surface-water irrigation at 1.78, 34.61, and 199.26 mg∙kg⁻¹, respectively.

3.5. Main Factors Affecting SOM Variation

Table 4. Spearman correlation coefficients.

	SOM	TN	AP	AK	pH	BD	ST	PM	TO	TT	IM	WS
SOM	1.00
TN	0.61 **	1.00
AP	0.11	0.15 *	1.00
AK	0.05	−0.03	0.05	1.00
pH	−0.16 *	−0.37 **	−0.25 **	0.29 **	1.00
BD	0.03	0.02	−0.04	−0.08	−0.12	1.00
ST	−0.04	−0.08	−0.08	0.14 *	0.21 **	0.10	1.00
PM	−0.30 **	−0.32 **	−0.16 *	0.49 **	0.33 **	−0.09	0.24 **	1.00
TO	0.001	−0.07	0.46 **	−0.05	−0.33 **	0.13 *	−0.07	0.08	1.00
TT	0.27 **	0.20 **	0.14 *	−0.34 **	−0.14 *	−0.16 *	−0.12	−0.82 **	−0.22 **	1.00
IM	−0.13 *	−0.12	−0.09	0.21 **	0.03	−0.11	−0.01	0.47 **	0.09	−0.32 **	1.00
WS	−0.12	−0.12	−0.10	0.21 **	0.03	−0.12	−0.01	0.45 **	0.08	−0.30 **	0.99 **	1.00

BD: soil bulk density; ST: soil type; PM: parent material, TO: topography; TT: tillage texture; IM: irrigation method; WS: water source. *, ** indicate that the correlation is significant at the 0.05 and 0.01 level (two-tailed), respectively. n = 242.

presents the results of correlation analyses. A highly significant positive correlation between SOM and TN was found in the correlation analysis between soil nutrient indicators (r = 0.61, p < 0.01). The correlation between PH and SOM, TN, AP, and BD were negative (p < 0.05). In line with the ANOVA results, the SOM and AP contents exhibited strong negative correlations with the soil-forming parent material. The SOM and AP contents of the alluvial–pluvial deposits were higher than those of the loess parent material, whereas the opposite was true for AK. The parent material showed a highly significant positive correlation with pH (p < 0.01) and had a negative correlation with BD. The parent material was significantly negatively correlated with the tillage texture, and soils with alluvial–pluvial deposits as the parent material were heavy loam. The content of SOM, TN and AP were strongly and positively correlated with the texture of tillage. The SOM and AP contents of heavy loam were greater than those of medium and light clay soils, whereas the opposite was true for AK. Irrigation methods were strongly positively associated with water source type (p < 0.01). Surface water was mostly used for furrow irrigation and groundwater was predominantly used for flood and border irrigation.

The SEM of the relationship between SOM, natural conditions, and soil showed that the direct effects of soil properties and selected natural and anthropogenic conditions on SOM did not show a strong correlation (Figure 4). In general, the positive effect coefficient of tillage texture on SOM was 0.21 and reached a significant level. SOM, the main index of soil nutrients, was indirectly affected by natural factors through their direct impact on soil properties. pH had a significant negative effect on SOM (−0.16), whereas the positive effect of BD on SOM (0.01) was not significant. These results indicate that the texture of the tillage layer was the main factor directly affecting SOM. A strong correlation was observed between natural factors and anthropogenic conditions. Natural factors had a strong positive impact on pH and a strong negative impact on BD. Combining the direct and indirect effects of each influencing factor on SOM, tillage texture was the dominant factor, explaining 8% of the SOM variability, whereas BD, parent material, irrigation, and water source type had less influence on SOM.

4. Discussion

4.1. General Characteristics Analysis of Soil Nutrient

Soil nutrients are important factors reflecting the quality of arable land and soil fertility [29,30]. The semi-arid Loess Plateau has been one of the primary locations for vegetation restoration and ecological construction in China. This is because of its fragile ecological environment and severe soil erosion caused by the dual influence of natural regional conditions and human activities [23]. To address these problems, it is important to begin by studying the soil properties of the region. Soil nutrients are important factors influencing the growth of vegetation crops in semi-arid loess areas, and a deficiency of soil nutrients or an imbalance between nutrients affects crops’ growth [31]. The classification standard of soil nutrients during the second China National Soil Survey [26] indicated that the average content of SOM and TN in the study area was low and the content of other nutrients were mid-range. This was a sign of the declining soil quality of the farmland from many years of continuous potato crop cultivation in the years leading up to 2017, coupled with improper fertilization. In view of the low SOM and TN contents, the application of organic soil and nitrogen fertilizers should be increased appropriately in future agricultural production and soil planning and utilization. The soil quality can be improved through the strategic and sustainable application of organic and nitrogen fertilizers. The soil nutrients of farmland in the Longzhong dry zone still have potential for improvement.

In the research region, the CV of soil nutrients varied from 24.53% to 69.44%. According to Nielsen’s categorization standards [32], all soil nutrients displayed moderate variability, which belong to medium-intensity variation. This reveals that the soil nutrients in the research region had spatial heterogeneity, which may be related to the complex topographic structure and high habitat heterogeneity in the Loess Plateau area. Therefore, it is necessary to study the causes of soil nutrient heterogeneity in dry agricultural areas of Longzhong using classical statistics and geostatistics.

4.2. Variogram Analysis

Soil nutrient spatial distribution has often shown a high level of spatial heterogeneity [33,34]. Variogram analysis, a typical geostatistical approach in ecology, is useful for defining spatial data. In addition, variogram analysis is crucial for comprehending the link between soil property heterogeneity and pedogenesis [35]. In this study, the regional variance in soil nutrient distribution in China’s semiarid Loess Plateau was described using variogram analysis. The soil nutrient spatial structure of the research region was reasonably intact, and the semi-variance fitted models accurately reflected the soil nutrients properties in space. While TN is a Gaussian model, AP and AK are exponential models, and SOM is a spherical model.

Nugget values (C₀) were less than 0.1 for the semi-variograms of the four soil nutrients. This meant that the spatial variation brought about by random factors like human activities, sampling, and the experimental process, was relatively small. For SOM, TN, AP, and AK, the nugget-to-sill ratios were between 25% and 75%. This demonstrated that the soil nutrients in this research exhibited a medium spatial correlation [27] and that the spatial variability of the samples was driven by a combination of random factors, like irrigation, fertilization, and soil improvement, and structural factors, like climate, topography, and soil type [36]. The amount of spatial dependency in the soil nutrient distribution was assessed using variogram analysis. The investigation revealed that the geographical correlation of the four soil nutrients were strong. This suggests that the effect of random factors on spatial distribution was relatively small, and that their spatial variation was mostly produced by natural factors influencing structural variation, such as parent material, terrain, and climate.

4.3. Spatial Distribution and Impact Factors

A comprehensive understanding of soil nutrient distribution patterns is critical for enhancing soil fertility and fostering sustainable and healthy agricultural production [6,7]. In addition, the impact of several geographical variables on soil nutrient content is complicated. Soil nutrient variation is caused by the combined effect of natural elements like climate, soil parent material, terrain, and soil texture, as well as human activities including fertilization, irrigation application, farming methods, and planting structure [5,37,38]. However, different nutrient factors tend to have different spatial distribution patterns. The environmental variables influencing the geographical distribution of various nutrient factors are also different.

The soil nutrients in the research region displayed distinct spatial distribution patterns at the county scale. The southern and northeastern regions of the study area had high SOM and TN values, while the middle part of the study area had low values. The ANOVA and structural equation modeling results indicated that this distribution of SOM was predominantly related to the soil-forming parent material and texture of the cultivated layer. Parent material has the greatest influence on soil formation, and is often highly associated with soil nutrients [39,40]. The soil-forming parent material directly influences the chemical and mineralogical compositions of soils. Soils are subjected to certain physical, chemical, weathering, and leaching interactions that produce corresponding nutrient changes. Soil texture is an important feature that defines the mechanical composition of soil particles, and can have a substantial impact on soil nutrient cycling and storage [41,42]. SOM is a key component of soil nutrients, and its quantity and quality are important indicators of soil fertility and health [30]. The distribution of TN was mainly influenced by the irrigation method and water source type. The results of this study show that the TN content was substantially higher using surface water and furrow irrigation methods. Combined with the overall low content of SOM and TN in the study area, this might be due to the presence of long-term unsustainable fertilization. This can result in the deterioration of cultivated soil fertility in farmland areas. In addition, the correlation analysis showed that SOM and TN were significantly positively correlated (p < 0.05). Therefore, appropriate amounts of bio-organic fertilizers could be applied during cultivation to promote improved soil fertility. Alternatively, various technical modes, such as green manure planting, might be employed to enhance the healthy growth of farming soils.

The distribution of AP was patchy, severely fragmented, and spatially heterogeneous. The potential reasons for this are as follows. The spatial autocorrelation range of AP was relatively small, and AP only occurred in a small area with strong spatial correlation, resulting in the formation of small patches. Soil P elements exhibited a deposition phenomenon [43]. The spatial distribution of AK showed a more prominent band, with a decreasing trend from east to west. ANOVA showed that the topography was closely connected to the distribution in space features of AP and AK of the study area. This is an important factor affecting soil development because it may change the soil nutrients’ distribution by regulating redistribution in space of precipitation and solar radiation intensity [44]. Across the entire region, the AP content was medium, but its distribution in space was uneven. This may be connected to the amount of phosphorus fertilizer applied during planting. AK was also at a medium level, which might be attributed to the location of study area in the hinterland of the Asian-European continent. This is a typical oasis ecological agricultural area that is less susceptible to the loss of fast-acting potassium from low rainfall, ultimately leading to a high AK content in the area.

Natural factors like soil-forming parent material, soil type, and topographic characteristics impacted soil nutrient, but precipitation, and temperature also played key roles in influencing soil nutrient content [43]. Human activities increasingly influence natural factors and have a direct impact on the soil nutrient content. The amount of chemical fertilizers applied directly affects the nutrient content in the soil, thus causing a substantial variation in soil nutrient heterogeneity [36]. This has shown that the extent of human activity disturbing the distribution of soil nutrients should be a key focus. However, in this study, because of the insufficient amount of factors’ data, the factors influencing the distribution were not taken into account. Therefore, in future studies on the factors influencing soil nutrient heterogeneity in this region, natural elements, such as climate, and human activity factors, such as fertilization, should be considered to establish a more science-based and comprehensive index system. Based on the current situation, soil nutrient distribution should be adapted to local conditions to improve soil fertility levels and support sustainable agricultural growth in the dry agricultural areas of Longzhong.

5. Conclusions

Overall, the SOM and TN contents were low, whereas AP and AK values were moderate. The CV of soil nutrients showed a moderate intensity of variation and all had spatial heterogeneity. All four soil nutrients were spatially autocorrelated with moderate intensity, mainly because of the combined influence of structural and anthropogenic factors. There were similar distribution characteristics for SOM and TN, with different sizes of patches present. However, parent material and tillage texture were the dominant factors affecting SOM; TN was substantially higher when surface water and furrow irrigation methods were used. The distribution of AP and AK have opposite characteristics, and they both differed significantly among parent materials, textures, topographies, and human factors. Therefore, in future agricultural production and soil planning, the amount of organic soil and nitrogen fertilizers applied to the soil should be increased appropriately, and precise fertilization management should be implemented based on soil nutrient heterogeneity and local circumstances. It is of great significance for farmers to promote the quantitative research of soil science and the implementation of precision agriculture encouraged by policy.