Spatial variability in soil pH and land use as the main influential factor in the red beds of the Nanxiong Basin, China

Soil sample collection

Samples were collected in November 2017 after crops (mainly rice) were fully harvested. A total of 225 samples were gathered from 0–20 cm depth by the nested sampling method at sampling densities. Soil samples were air-dried and passed through a 2 mm sieve for laboratory analysis of soil pH was measured in a 1:2.5 soil:water (DI water) suspension using a PP-50-P11 pH meter (with measurement error ± 0.002) (Liu, Shao & Wang, 2013c) equipped with a calibrated combined glass electrode (Bogunovic et al., 2017a; Bogunovic et al., 2017b). The global positioning system (GPS) (5-m precision) was then used to identify the site’s longitude, latitude and elevation. The actual sampling sites were chosen to represent the main topography, land use, and vegetation types within the range of vision. Site slope and aspect measured with a geological compass, and information on human activities (irrigation, fertilizer use and crop yield) was collected from surveys of the local inhabitants. The distribution of sample points is shown in Fig. 1.

Data analysis

Some basic statistics were calculated, such as the minimum, maximum, and mean values of measurements and their coefficient of variation (CV). The Kolmogorov–Smirnov (K–S) test and correlation analysis of soil pH with topographical variables were performed to analyse data distribution, using the statistical software SPSS 19 (IBM SPSS Statistics for Windows, IBM Corp., Armonk, NY, USA). GS+7 (Gamma Design Software, Plainwell, MI, USA) was used to do the geostatistical analysis. The K–S method was used to evaluate data normality and asymmetry in terms of skewness and kurtosis because these factors have important implications for the performance of the interpolation methods.

A semivariogram is the basic tool of geostatistics (Oliver & Webster, 1986; Goovaerts, 1999; Nasseh et al., 2016). The formula used to calculate the semivariogram is: (1)${\displaystyle \gamma \left(h\right)=\frac{1}{2N\left(h\right)}\sum _{i=1}^{N\left(h\right)}{\left[Z\left(xi\right)-Z\left(xi+h\right)\right]}^2.}$ In Eq. (1), N(h) is the logarithm of the distance when the distance equals h, and Z(x_i) is the value at location x_i; Z(x_i + h) is the value at a distance h from x_i (Yang et al., 2016; Rosemary et al., 2017). Appropriate model functions were fitted to the semivariograms. The semivariograms were used to determine the degree of spatial variability on the basis of the classes of spatial dependence distinguished by Cambardella et al. (1994), strong spatial dependence (C₀/(C₀ + C) >75%), moderate spatial dependence (25% <C₀∕(C₀ + C) <75%) and weak spatial dependence (C₀∕(C₀ + C) <25%). In ArcGIS 9.2 (ESRI 2006. ArcGIS Desktop: Redlands, CA: Environmental Systems Research Institute.), we used kriging interpolation in the geostatistics module to draw the spatial distribution map of soil pH and the trend analysis chart in order to analyse the characteristics of the spatial variability. The main factors controlling spatial variation in soil pH and their influence were analysed using maps of the soil type, slope, aspect, elevation, and land use.

Isotropic semivariogram of soil pH

GS+7.0 software was used to fit the soil pH in the study area to the theoretical model (Table 2). The variogram’s fitting model was selected based on the nugget effect, the coefficient of determination (R²) and the range of variation (Bogunovic et al., 2017a; Bogunovic et al., 2017b). As can be seen from Table 2, the value for nugget (C₀) is 0.12, the value for sill (C₀ + C) is 0.18, the ratio of nugget (C₀) and sill (C₀ + C) is 66.67%, and the determination coefficient (R²) is 0.812. High coefficients of determination indicate that the models fitted the semivariogram well (Jeloudar et al., 2014). The nugget–sill ratio of 66.67%, indicating that the soil pH had a moderate spatial dependence (Cambardella et al., 1994). The spherical model gave the best fit for the variation in soil pH in the study area. The main structural factors were climate, parent material and terrain; these can enhance the spatial dependency of soil pH. In contrast, random factors, which are the result of human activity such as farming and fertilization, can make the spatial dependency of soil pH weaker (Isaaks & Srivastava, 1989). This moderate spatial dependence of soil pH in the red beds implies that the spatial variation of soil pH in the study area is mainly caused by both structural and random factors.

As can be seen in Fig. 5, when the separation distance is more than 161 m, the semivariance fluctuates only slightly, and then stabilizes. This trend might be caused by differences in directional variation. The variance at 250 m implies that the range of the spatial dependence is much wider than the sampling interval. Therefore, the current sampling design was appropriate for this study.

In order to understand the characteristics of the spatial variation in soil pH, the semivariogram was drawn in four directions, E–W (0°), NE–SW (45°), S–N (90°) and SE–NW (135°), using the GS+7.0 software. As shown in Fig. 6, the spatial variation exhibits large differences in different directions, showing the heterogeneity. Table 3 shows that the best-fitting models in the four directions are all spherical. The nugget (C₀) and sill (C₀ + C) values are different and their ratio ranges from 60.24% to 69.23%, indicating moderate variation.

As shown in Fig. 6, the range of the soil pH values from the northeast to the southwest (45°) and from the southeast to the northwest (135°) is significantly smaller than that from east to west (0°) and from north to south (90°), indicating that the variation in the 0° and 90° directions is more complex than those at 45° and 135°.

From east to west (0°), when the separation distance is greater than 161 m, the difference in the semivariance of the soil pH begins to fluctuate, first increasing and afterward decreasing to around 0.0388. The semivariance from north to south (90°) shows the same trend, alternating between high and low, but the degree of fluctuation in the E–W (0°) direction is smaller. When the separation distance is larger than 169 m, the variation of the soil pH in the NE–SW (45°) and SE–NW (135°) directions is more stable near 0.0388, and the degree of variation is not very different. The main reason is that the area is near the badlands hills in the NE–SW and the SE–NW directions; the topography and parent materials are of great influence, and in the SE–NW direction there are more hills and larger undulations. However, in the N–S and E–W directions (0° and 90°, respectively), the soil pH shows high spatial homogeneity because the relief is low and the only land use is farmland in these directions. Taken together, the soil pH in this study area has an obvious spatial heterogeneity, which is suitable for further interpolation analysis.

Analysis of the spatial distribution of soil pH

The effect of trends is a prerequisite for and the basis of prediction by kriging interpolation. The number of parameters that are required for kriging interpolation becomes smaller as the order of the trend effect decreases. Thus, a lower order of the trend effect can reduce error, and many scholars take the lower-order trend among two trends as the trend to be used in conducting prediction by interpolation (Li et al., 2013). Trend analysis can provide a study area sampling point and a three-dimensional perspective with information for the attribute value on the z-axis. The global trend in sampling data can be analysed from different perspectives.

As shown in Fig. 7, soil pH decreases from northeast to southwest, which is consistent with the result of semivariogram analysis. The soil pH values are higher in the northeast and southwest; this pattern can be explained by the difference in land use. In the northeastern and southwestern parts, the land is unused land with a high relief. Arable land is mainly distributed in the northwest, where the relief is low and the land is strongly affected by human activities such as the use of nitrogen fertilizer, which might cause a reduction of the pH value in soil (Yüksek et al., 2009).

Table 4 shows that the pH value of the 0–20 cm soil layer tends to decrease from upper slope to middle slope to downslope for slope below 20°, when slope over 20°, this trend is reversed (P < 0.05).

Spatial distribution pattern of soil pH

Based on the semivariance function model and the spatial distribution trend analyses, the spatial distribution pattern of soil pH in the study area was analysed by interpolation analysis of the 3D map constructed with the GS+7.0 software (Nasseh et al., 2016). Kriging analysis of the 3D map shows that the soil pH varies greatly in the horizontal direction in the study area (Fig. 8); the soil pH is higher in the northeast and the southwest, increases towards the southwest, and decreases towards the northwest. The result of inverse distance weighting interpolation of the 3D map shows that the overall trend for the pH in the study area is consistent with the results from kriging interpolation (Fig. 9).

Influence of slope and position along the slope on the spatial distribution of soil pH

Severe soil erosion can cause a decrease in the pH value (Schindelbeck et al., 2008). Due to the humid monsoon climate and the high erodibility of purple soil, which is caused by its high content of sandy particles, its pH value is generally lower than in the weathering sediments of red beds, which have a pH value higher than 8.

In general, soil pH varies significantly between different slopes and positions along the slope (Henkel, 2003). Thus, the pH of surface soil (0–20 cm) also varies with slope and position, reflecting the geomorphic process.

Soil properties on different slope positions were significantly affected by the degree of soil development and the leaching processes (Tsui, Chen & Hsieh, 2004). The effects of topographic factors on soil pH were discussed in this study. For slopes under 20° in this study, the pH of soil is the highest on the downslope, followed by that in the middle slope, and is the lowest on the upper slope. In case of ploughed vineyard decrease of pH value along the slope catena was noticed by Kenderessy & Lieskovsky (2014), which indicates that calcic carbonates could be leached from the surface layer by excess water from upper positions as a result of accelerated erosion. For slopes over 20°, soils on the footslope which is a concave position has a significantly higher pH than those on other slope positions, similar results were reported by Tsui, Chen & Hsieh (2004) and Huggett (1975), who confirmed that slope, which is involved in the transport and accumulation of solutes, resulted in a higher pH. Thus, to some extent factors affecting soil erosion have an influence on soil pH. However, in this study case in red beds area the concrete mechanisms, how topographical factors affect the pH in different positions, needs further research.

In addition, as we know, topography is a structure factor influencing the spatial variability of soil pH. In our study area, in the E–W and N–S directions (0° and 90°, respectively), the soil pH shows high spatial homogeneity because the relief is low and the only land use is farmland. An important result is that the topography influences soil pH mainly through the slope and indirectly via effects on land use patterns; this is a general conclusion that has rarely been acknowledged elsewhere.

Influence of aspect on the spatial distribution of soil pH

Different slope aspects experience different solar radiation, temperature and water conditions. The vegetation coverage is also different. Therefore, differences in physical, chemical, and biological processes in the topsoil are correlated with different aspect directions, leading to a heterogeneity in pH content and distribution in the topsoil (Vieira et al., 2009; Salehi, Esfandiarpour & Sarshogh, 2011). By combining the aspect distribution map of the study area and the geostatistical analysis module in the ArcGIS software, the spatial distribution map of the soil pH was analysed (Figs. 10 and 11). The result shows that the average pH value varies with aspect of the slope in the study area. The soil pH values on north- and southwest-facing slopes are relatively higher than on slopes of other aspects.

The location of the study area, in the humid red-bed area in south China, is representative for the concentrated distribution of soft rock in red beds. The best fitting models were all spherical, with a high degree of fit for the spatial variability of soil pH, and they were verified in various studies (Wang et al., 2011; Liu, Shao & Wang, 2013c), indicating that the spatial structure of soil pH in the study area was distinct.

Kerry & Oliver (2004) indicated that, as a rough guide, future sampling intervals should be chosen to be less than half the variogram range. According to the results of this study, future sampling intervals for monitoring pH should be 80–100 m.