Spatial heterogeneity of soil nutrients in Yili River Valley

Study area

The Yili River Valley Basin is located on the western slope of the Tianshan Mountains southwest of the Zhungeer Basin, at latitude 42°14′–44°50′ and longitude 80°09′–84°56′. The Yili River Valley Basin is a rift basin separated by the south and north Tianshan mountains. These mountains surround the basin on the east, south, and north sides. From west to east, the elevation and rainfall gradually increase, the area has a typical temperate continental climate with annual average precipitation was 417.6 mm, an annual average temperature of 10.4 degrees Celsius, and 2,898.4 h of annual sunshine on average (Sun, Li & Yang, 2012). This area is about 5.53 × 10⁴ km² in size; the distance from east to west is 360 km, and the length from south to north is 275 km (Yang et al., 2010). The sampling points were: 43°51′–44°9′N and 81°32′–81°28′E, in the north slope of Yining City (Fig. 1). The natural vegetation is temperate grassland vegetation and semi-desert grassland vegetation, and the soil type is mainly gray alluvial soil, boggy soil, chernozem soil, sierozem soil and meadow soil.

Sample collection and analysis

We chose representative samples in the Yili River Valley in May 2015, associated with regional DEM, land use types, climate, soil and other natural conditions. The self-made sampler (10 cm diameter and 50 cm height) was used to collect topsoil samples (0–20 cm). Each sample point was positioned by GPS, and factors such as latitude, longitude and elevation were recorded. One hundred twenty-seven topsoil samples at depths of 0–20 cm were collected (Fig. 1) including cultivated land samples (nine), garden samples (18), woodland samples (12), and grassland samples (88). The organic carbon concentration of the soil (g•kg⁻¹) was determined using the K₂Cr₂O₇-external heating method; total nitrogen concentration (g•kg⁻¹) was determined using the perchloric acid-sulphuric acid digestion method; total phosphorus in the soil was found using the acid-soluble and molybdenum antimony ratio colorimetric method; total potassium in the soil was determined using the acid digestion and flame photometric method; soil pH was determined using the potential method; SBD was determined using the ring shear testing method; water content was determined using oven drying method (Liu et al., 2016); and the Mastersizer 2000 particle size analyzer (Malvern Instruments, Malvern, England) was used to measure soil particle composition (Zhang et al., 2016).

Classical statistical description of the physical and chemical properties of the soil

The coefficient of variation (CV) values were 68%, 57.76%, 12.62%, 28.99%, 13.51% and 49.74% for SOC, STN, STP, sand, silt and SWC, respectively, that ranged from 12.62% to 68.00%, indicating a moderate degree of variation. CV values were 9.17%, 8.82% and 1.89% for SBD, STK and pH, respectively, ranging from 1.89% to 9.17%, corresponding to a weak variability. SOC was between 2.04 g·kg⁻¹ and 34.79 g·kg⁻¹ with a mean value of 6.99 g·kg⁻¹, which was lower than the Chinese national average value of 17.53 g·kg⁻¹ (Li et al., 2010). STN values were between 0.24 g·kg⁻¹ and 3.52 g·kg⁻¹ with an average value of 0.79 g·kg⁻¹, which was also lower than the Chinese national average value of 1.54 g·kg⁻¹ (Li et al., 2010) (Table 1).

A non-parameter and single sample test of K-S showed that, SOC, SWC and STK values followed a normal distribution; STP, STK, sand, silt, SBD and pH values followed log-normal distribution (Table 1).

Correlation analyses

The Pearson’s correlation coefficient was calculated to characterize the relationship between soil nutrients (SOC, STN, STP and STK) and selected properties (sand, silt, SWC, SBD and pH). SOC, STN, STP and STK had a significant correlation with silt content and a negative correlation with both sand and SBD. SWC was significantly positively correlated with SOC and STN, but had no relation to STP and STK. pH was significantly negatively correlated with SOC and STN, but had no relation to STP and STK (Table 2).

Relationship between soil nutrients and topographic factors

Topography is one of the main factors influencing soil properties (Moore et al., 1993). Therefore, we performed a correlation analysis between topographic factors (gradient, altitude and slope) and soil nutrients. The slope (°) was transformed into the sine value representing the extent of the eastern slope and the cosine value representing the extent of the northern slope (Lian et al., 2009). Pearson’s correlation coefficient was calculated for each variable to characterize the relationship between soil nutrients and selected topographic factors. The results showed that SOC and STN had a strong positive correlation with Cos a and elevation (p < 0.01); STP had a positive correlation with Cos a (p < 0.05). Gradient and Sin a were not significantly correlated with other soil nutrients (Table 3).

Spatial variability analysis of soil nutrients

We calculated the semi-variogram for all studied variables to characterize the spatial variation of soil nutrients and selected the best models based on the determination coefficient R². The results showed that the reduced sum of squares (RSS) of all nutrients was lower. In addition, the R² of STP was small (0.34), while other R² values ranged from 0.58 to 0.97, which could reflect the spatial heterogeneity of the topsoil nutrients. In most cases, the variograms of the soil nutrients were expressed well by a logarithmic model (STP) or an exponential model (Table 4).

Nuggets of STK, STP and STN were 1.08, 0.01 and 0.01, respectively. The nugget of STK was 0.004, which indicated a possible error caused by sampling, the experimental process, or the small minimum sample size. The Cs/Sill ratios for the SOC and STN were more than 75%, showing a strong spatial correlation. The Cs/Sill ratios for the STK and STP were between 25% and 75%, which showed that the STK and STP had a medium spatial correlation. The ranges of SOC, STN and STK were all similar, about 61.0 km, while the variation of STP was 37.6 km (Table 4).

Spatial interpolation of soil nutrients using an Ordinary Kriging interpolation

The maps of SOC, STN, STP, and STK using the ordinary kriging interpolation method were drawn. SOC was generally higher in the north than the south with the high-values between 7.5 g·kg⁻¹ and 35 g·kg⁻¹ distributed in the northwest and northeast, and the low-values between 2 g·kg⁻¹ and 4.7 g·kg⁻¹ distributed in the middle and southwest. The value of STN was higher in the north between 0.9 g·kg⁻¹ and 3.5 g·kg⁻¹ than in the central and southwest, where it had a lower value between 0.24 g·kg⁻¹ and 0.56 g·kg⁻¹. STP was higher in the northwest and southeast (0.74–1.1 g·kg⁻¹), and lower in the northeast and southwest (0.38–0.69 g·kg⁻¹). The STK was high in the northwest and northeast (14.09–15.92 g·kg⁻¹), and low in the central and southeast (5.21–12.03 g·kg⁻¹) (Fig. 2).

Spatial variation characteristics of soil nutrients

Classical statistics showed that in addition to the STK of weak variability, the variation coefficients of SOC, STN and STP were between 12.62–68%, indicating a moderate degree of variation (Table 1). These results were similar to that of other studies (Hu et al., 2007; Wang et al., 2014). Huang et al. (2015) indicated that SOC, TN, TP had moderate variability in a Taklimakan desert highway shelterbelt caused by soil parent materials. A classical statistical analysis can help explain the general status of soil nutrients but cannot effectively reflect their spatial characteristics. Therefore, the semivariance function was used to analyze the spatial variability of soil nutrients. Many studies showed that the spatial variability of soil properties was controlled by intrinsic factors such as soil parent material and soil texture and extrinsic factors such as fertilizer and irrigation application (Yang et al., 2013). The strong correlations of SOC (96.1%) and STN (98.48%) were indicated from the same soil parent material. The moderate spatial correlation of STP (52.31%) and STK (57.22%) indicated that they came from both intrinsic and extrinsic factors (Table 2). In typical soil of karst forest land, Zhang et al. (2013b) found SOC and STN showed moderate spatial correlation. In the oasis of the Manas River Watershed, Maimaiti, Abuduwaili & Ge (2013) found that STN and SOM showed moderate spatial correlation. Our study results differed from previous research (Zhang et al., 2013b; Maimaiti, Abuduwaili & Ge, 2013). These differences may be because our samples are from grasslands, which were often less impacted by human activity and influenced more by natural activities.

The range is an essential parameter for measuring spatial heterogeneity, which is important for sampling space and scope validity. The variable range of SOC, STN and STK were all similar, about 61.0 km, which meant that they had the same ecological process, while the variation of STP was 37.6 km, which showed that the ecological processes of STP were similar. If the sampling grid could map soil nutrient heterogeneity, the sampling grid should be less than the variable range, so the sampling grid of SOC, STN and STK should be less than 60 km and the sampling grid of STP should be less than 30 km.

Spatial interpolation analysis of soil nutrients

The distributions of SOC and STN were high in the north and low in the south (Fig. 2). Those distributions were mainly influenced by natural factors (topography, elevate and soil texture). Our investigation found that SOC and STN strongly correlate positively with Cos a (p < 0.01). We realized that radiation was influenced by the slope—the greater the Cos a, the less the acceptance of solar radiation. In general, the sunny slope was dry, the soil organic matter decomposed rapidly, whereas the shady slope was wet, the accumulation of organic matter was more, and the soil nutrients were relatively high (Lian et al., 2009). Climatic factors also influenced SOC and STN under natural conditions (Sun et al., 2017). SOC and STN have a strong positive correlation with SWC (Table 2). The unique terrain determined less precipitation in plain and more precipitation in the Yili River valley mountainous areas. The increase of precipitation made the vegetation grow vigorously and promoted organic matter accumulation. Sun et al. (2017) argued a significant positive correlation between SOC and precipitation, the same as STN.

We found that SOC and STN have a strong positive correlation with elevation (Table 3). SOC and STN increased with altitude from south to north (Table 3, p < 0.01), a result also found by Yang et al. (2010), who hypothesized that SOC and STN increased with elevation. As elevation increases, microbial activity decreases, and the dynamic decomposition rate of animal and plant residues also decreases (Sun, Li & Yang, 2012), causing SOC and STN to increase. STP was high in the northwest and southeast and low in the northeast and southwest. STK was high in the northwest and northeast and low in the central and southeast. STP and STK have no relevance to SWC, pH and elevation, but in a natural environment, STP and STK were influenced by soil parent material more than SOC and STN (Liu, Shao & Wang, 2012).

SOC, STN, STP and STK had a significant correlation with silt content and a negative correlation with sand content and SBD (Table 2). This finding was in agreement with the results of Saggar, who suggested that soil nutrients were often positively associated with fine soil particles (Saggar, Yeates & Shepherd, 2001). SBD is a key physical property of soil that affects the transport of water and solutes, and is essential to assessing nutrient reserves (Yang et al., 2015). SBD generally increased, which means that the process of soil fertility declined (Table 2). Soil texture impacts the input of SOC, STN, STP and STK by influencing vegetation productivity via moisture availability and soil fertility (Wang et al., 2010).