1. Introduction
Potassium is one of the three essential nutrients for plant growth and development and is known as the quality element and stress resistance element for plants [
1]. In recent years, with the continuous growth in China’s grain production and the intensity of cultivated land use, crops have taken away a large amount of potassium [
1]. In addition, due to irrational land use by farmers and failure to apply potassium fertilizer, the problem of potassium deficiency in farmland in China has become increasingly serious [
2]. The lack of potassium in cultivated soil not only affects the sustainable use of cultivated land resources but also directly affects the yield and quality of crops. In the long run, it will further affect the sustainable development of regional agriculture. As the form of potassium that can be directly absorbed by plants, soil available potassium (AK) is an important indicator of soil potassium supply capacity [
3] and an indispensable nutrient element that participates in plant biochemical processes throughout the whole life cycle [
2]. Its supply and crop nutrient demand in space are not coordinated, which has been one of the key factors limiting the efficient use of fertilizer [
4]. Therefore, it is of great significance to explore the spatial differentiation characteristics and influencing factors of regional soil AK to aid the scientific formulation of fertilization programs, the protection of ecological security, regional planting planning, the sustainable use of soil, improvements in cultivated land quality, and national food security [
5,
6,
7].
With the development of precision agriculture and soil testing formula technology, research on the heterogeneity of soil AK is also deepening. Many researchers have used a combination of geostatistics and GIS technology to reveal the spatial variability of soil AK and its spatiotemporal variability characteristics [
8] from the perspectives of terrain [
9], land use [
10,
11,
12], watershed [
13], scale [
14], and spatial distribution prediction [
15], and have achieved rich results. The research shows that the spatial heterogeneity of soil AK is closely related to regional natural factors (terrain, climate, soil parent materials) and human factors, which provides a scientific basis for the precise management of the content of soil AK in the region. From the relevant literature, at present, it appears that researchers mostly use correlation analysis, classical regression models, boosted regression tree methods, analysis of variance, canonical redundancy analysis, and other traditional methods to explore the factors affecting the spatial distribution pattern of soil AK, mainly based on the degree of influence of a single factor [
16,
17]. There is still limited research on how the interaction of factors affects the AK distribution in cultivated land. Geodetectors serve as powerful tools for detecting potential spatial correlations between variables [
18]. Originally used for studying endemic diseases and their related geographical impact factors [
19], in recent years they have been applied in environmental [
20], societal [
21], and economic research [
22] and have achieved good research results. Geodetectors could overcome the shortcomings of traditional methods and further comprehensively reveal the explanatory power of influencing factors on the spatial differentiation of soil nutrients. Therefore, this paper uses the geostatistics method, combined with a geographical detector model, to analyze the spatial differentiation characteristics and mechanism of soil AK more comprehensively and objectively. Our research also expands the practical application field of geographical detectors.
The study area is a typical mountainous agricultural region in China, with a relatively low quality of cultivated land and a small per capita cultivated land area. With construction, agricultural structure adjustment, and ecological conversion, the trend of cultivated land reduction is difficult to reverse. Due to the excessive cultivation of land resources, the environment has been damaged: water and soil losses have been serious, and the quality of cultivated land has changed greatly, which has seriously affected food production. This study considers Shiyan, a mountainous area in northwestern Hubei Province, as the research area. Through field investigation, data collection, and sampling analysis by the research team, and on the basis of 3S technology, combined with geostatistics and geographical detectors, the spatial variation in AK in the cultivated land of Shiyan and its influencing factors were studied. The findings of this research can provide a theoretical reference not only for soil improvement in cultivated land but also for improving the quality of cultivated land and promoting agricultural production in the research area.
2. Research Methods and Data Sources
2.1. Overview of the Study Area
Shiyan is located in the northwest mountainous area of the Hubei Province, 109°29′–111°16′ E and 31°30′–33°16′ N, with a total area of about 23,680 km2. The city is high in the north and south and low in the middle, inclined from southwest to northeast, with an altitude of 83–2571 m. The study area belongs to the north subtropical semi-humid climate, with an annual precipitation of about 870 mm and a mean annual temperature of 15.4 °C. The soil parent materials in the area are mainly river and lake alluvial deposits, red sandstone alluvium, quartzite weathering products, argillaceous weathering products, carbonate weathering products, and Quaternary old alluvium. The various soil parent materials staggered. The main soil types are yellow-brown soil and paddy soil. The main crops are corn, wheat, and rice.
2.2. Data Source and Processing
2.2.1. Sampling Point Data
The data were from the sampling results of cultivated land quality evaluation in every county of Shiyan (urban area not included) in 2017, and the sampling points were arranged based on the principles of representativeness and uniformity. We used GPS to locate and collect 0–20 cm cultivated land samples. Each soil sample was mixed with a “plum blossom”- or “S”-shaped procedure according to the size of the plot. Each soil sample was taken away in 1 kg bags for analysis [
23]. A total of 701 samples were taken (
Figure 1), and we recorded surface environmental information such as the cropping system, land use mode, topsoil depth, etc. of the sampled fields in detail. In the later stages, there were more than 40 items in total, including the supplementary soil type and soil parent material information of the relevant sample points of the second soil survey soil map. After the soil sample was air-dried, sieved, and mixed, the pH value of the soil sample was determined by a 2.5:1 water and soil ratio extraction using the pH meter method. The soil organic matter was determined using the oil bath heating potassium dichromate oxidation volumetric method, and the content of AK was determined using the ammonium acetate extraction and flame photometric method.
2.2.2. Impact Factor Data
Based on the existing relevant results [
24,
25], we selected the natural factors (terrain, climate, soil) and human factors that are most closely related to the spatial variability in AK in cultivated soil on this scale. The selected factors were as follows: elevation (X1), slope (X2), mean annual temperature (X3), annual precipitation (X4), soil parent materials (X5), soil type (X6), soil pH (X7), cropping system (X8), and topsoil depth (X9). Among them, the temperature and precipitation data were from the China Meteorological Science Data Sharing Service Network; the DEM data came from the geospatial data cloud, at a resolution of 30 m; and ArcGIS 10.7 software was used to extract the altitude and slope data.
2.2.3. Geodetector
The geodetector model is composed of four sub-models—factor detector, interaction detector, risk detector, and ecological detector—and is a powerful tool for detecting the potential spatial correlation between variables [
18]. It was initially used to study endemic diseases and their related geographical impact factors [
19]; in recent years, it has also been used to study the driving factors of the spatial variability of soil nutrients [
26]. In this paper, a geographical detector was used to quantitatively study the influence of different factors on the spatial distribution of soil AK in the study area. The interaction detector was used to explore the impact of the interaction of different influencing factors on the spatial differentiation of soil AK. The specific calculation method is detailed in [
26].
2.3. Data Preprocessing
SPSS24.0 was used for descriptive statistical analysis and one-way ANOVA. Semi-variance analysis and theoretical model fitting were carried out using GS+9.0. The sample distribution map and ordinary Kriging interpolation were completed in ArcGIS 10.7. According to the requirements of geodetectors for input variables, the data of soil type, soil parent materials, and cropping system were classified according to the category. The data of elevation, slope, temperature, precipitation, soil pH, and topsoil depth were discretized and divided into five categories according to the natural breakpoint method.