Spatial variability of soil bulk density and its controlling factors in an agricultural intensive area of Chengdu Plain, Southwest China

LI Shan, LI Qi-quan, WANG Chang-quan, LI Bing, GAO Xue-song, LI Yi-ding, WU De-yong

College of Resources, Sichuan Agricultural University, Chengdu 611130, P.R.China

Abstract Soil bulk density is a basic but important physic soil property related to soil porosity, soil moisture and hydraulic conductivity,which is crucial to soil quality assessment and land use management. In this study, we evaluated the spatial variability of soil bulk density in the 0–20, 20–40, 40–60 and 60–100 cm layers as well as its affecting factors in Southwest China’s agricultural intensive area. Results indicated the mean value of surface soil bulk density (0–20 cm) was 1.26 g cm–3,significantly lower than that of subsoil (20–100 cm). No statistical difference existed among the subsoil with a mean soil bulk density of 1.54 g cm–3. Spatially, soil bulk density played a similar spatial pattern in soil pro file, whereas obvious differences were found in details. The nugget effects for soil bulk density in the 0–20 and 20–40 cm layers were 27.22 and 27.02% while 12.06 and 3.46% in the 40–60 and 60–100 cm layers, respectively, gradually decreasing in the soil pro file,indicating that the spatial variability of soil bulk density above 40 cm was affected by structural and random factors while dominated by structural factors under 40 cm. Soil organic matter was the controlling factor on the spatial variability of soil bulk density in each layer. Land use and elevation were another two dominated factor controlling the spatial variability of soil bulk density in the 0–20 and 40–60 cm layers, respectively. Soil genus was one of the dominated factors controlling the spatial variability of soil bulk below 40 cm.

Keywords: soil bulk density, pro file, spatial variability, controlling factors, Chengdu Plain

1. Introduction

Soils, especially in plain where mechanized farming is prevailing due to the development of modern agriculture,are frail to soil compaction. Compactness of soils in agricultural soil has caused worldwide concern, because it can restrict plant growth and reduce crop yield gross,even cause serve environmental problems (Keller and Håkansson 2010; Lestariningsih and Hairiah 2013; Quraishi and Mouazen 2013; Yanget al.2016). Soil bulk density, a fundamental soil physical property related to soil porosity,soil moisture and hydraulic conductivity (Sequeiraet al.2014; Ileket al.2017), is commonly used to assess soil compactness, which is important to access soil quality and productivity (Lestariningsih and Widianto 2013; Xuet al.2016; Yanget al.2016). Additionally, soil bulk density is an important parameter of soil hydrologic models (Li D Fet al.2016; Walteret al.2016) and the key factor controlling the precision of soil organic carbon storage estimation in different scales (Throopaet al.2012; Walteret al.2016;Xuet al.2016). Due to the complex soil forming process,ecological conditions and human activities, soil bulk density varied with spatial scales and layers, which contributed to the spatial heterogeneity of soil bulk density (Alletto and Coquet 2009; Larkaet al.2014; Wanget al.2014).Thus, understanding the spatial distribution of soil bulk density and the controlling factors can provide valuable information for soil quality improvement and related soil process predictions.

Since the collection of undisturbed soil cores is labor intensive, time-consuming, tedious and expensive direct measurement for spatial variability of soil bulk density in large scale areas is dif ficult to obtain (Wanget al.2014; Yiet al.2016). Geostatistics provides an useful method to describe the spatial variability of soil bulk density across space. Many studies have reported geostatistics method combined with geographic information system (GIS) was successfully applied in the spatial prediction for other soil properties (Marchantet al.2010; Li Q Qet al.2016; Vasuet al.2017), and Ordinary Kriging method is a powerful interpolation tool to minimize the investigation cost which can not only quantify but also reduce the variance of estimate error (Yanget al.2016). Thus, Ordinary Kriging method can be used to interpolate soil bulk density across space to overcome the lack of soil bulk density data.

The spatial variability of soil bulk density is controlled by soil characteristics, environment and anthropogenic activities, which can be divided into two factors including structural factors controlled by the natural soil formation process (e.g., soil types, topography and meteorological factors) and random factors dominated by anthropogenic activities (e.g., tillage, fertility and land use management)(Wanget al.2014; Abegazet al.2016; Li Q Qet al.2016). Recently, a number of authors have estimated the relative importance of structural or random factors on soil bulk density by using semi-variance function, but most researches were restricted to the sampling depth of top 30 cm considering the sampling collection cost (Wanget al.2014; Yanget al.2016). Other studies documented the relevance of sampling depth to soil bulk density prediction,indicating an emergent demands for deeper soil bulk density to better understand the distribution of soil bulk density as well as its in fluencing factors (Voset al.2005; Wanget al.2014). Although previous studies indicated that the spatial variability of soil bulk density could be explained mainly by soil organic carbon, texture and topographic attributes(Beniteset al.2007; Wanget al.2014) and further studies also implied land use could affect the spatial variability of soil bulk density (Sheteet al.2016; Negasaet al.2017), little attention was paid to discern the dominating factor varying with soil layers as well as their relative contributions to the spatial variability of soil bulk density.

The Chengdu Plain in Southwest China, located in the upper Yangtze River, is an aslant fluvial plain with superior natural conditions. Rapid urbanization and industrialization,high population density and intensive agricultural activities in this area have caused a series of severe environmental problems of soils (e.g., heavy metal pollution, non-point source pollution and organic contamination) (Gaoet al.2012; Liuet al.2016; Liet al.2017). However, little is known so far about the spatial variability of soil bulk density and the controlling factors in this area, which can offer useful information for soil environment improvement and the biological cycling of soil carbon, nitrogen and phosphorus.In this paper, we aimed to evaluate the spatial variability of soil bulk density and its controlling factors in each layer in an agricultural intensive area. The specific objectives were to: 1) evaluate the pro file distribution of soil bulk density; 2)estimate the spatial variability of soil bulk density varying with layers; 3) determine the controlling factors of soil bulk density in each layer.

2. Materials and methods

2.1. Study area

This study was conducted in the west area of Chengdu Plain located in Chengdu City including the whole area of Pidu and Wenjiang counties and parts of Dayi, Chongzhou, Xinjing,Dujiangyan and Pengzhou counties with relatively gentle topography (Fig. 1) with the latitude approximately between 30°23´ and 31°1´N and longitude between 103°27´ and 103°55´E. The study area, part of the aslant fluvial plain by Minjiang River, covers a land area of 2 139.1 km2with altitude ranging from 447 to 732 m above mean sea level (Fig. 1-A).The study area is characterized by subtropical monsoon climate with mean annual temperature and total annual precipitation of 16°C and 900 to 1 300 mm, respectively.

2.2. Soil sampling and analysis

Fig. 1 Maps for the location of study area (A) and distribution of sampling sites (B).

Synthetically taking the soil type and land use in the study area into account, soil pro file sites were firstly designed indoors based on the 3 km×3 km grid. If the location of sampling site was not accessible for instance in urban area,in rivers or on roads, we selected an alternative site as nearby as possible within a 1 km radius. Otherwise, we removed this kind of sampling site. A total of 522 soil samplings of 134 soil pro files (Fig. 1-B) were actually collected with mixed sampling method in the April of 2016 before the next crop sowing. According to the depths of soil formation,four sampling depths including 0–20, 20–40, 40–60 and 60–100 cm layers were designed, and the sampling numbers of each layer by turn was 134, 134, 129 and 125,respectively. During the sampling, geographic coordinates and elevations of each sampling site were obtained using a global positioning system (GPS). At each site, undisturbed soil cores were collected in metal cylinders (diameter 5 cm;length 5 cm), by hand using gentle pressure, to measure bulk density in the 0–20, 20–40, 40–60 and 60–100 cm layers,respectively. Two replications were collected considering the cost for sampling collection. Soil cores were dried in an oven to constant weight (105°C) (Lu 2000). About 1 kg disturbed soil samplings were also collected. After rock and roots were removed from each soil sampling, the soils were air-dried and passed through 0.25-mm sieve for determination of soil organic matter by using the potassium dichromate and sulfuric acid method (Lu 2000). Three laboratory replications were conducted for quality controlling.

2.3. Treatment for selected factors affecting the spatial variability of soil bulk density

Soil is a complex synthesis controlled by environment, natural conditions and anthropogenic activities, which can be divided into two aspects as structural and random factors. In this study, we chose elevation, parent materials and soil types as structural factors and land use and soil organic matter as random factors to estimate the effects and relative importance of structural or random factors on the spatial variability of soil bulk density in each layer in the study area.

ElevationTopographic factors could affect soil bulk density through soil forming process. In this study, we chose elevation as one of the topographic factors to estimate the effects of topography on soil bulk density, considering that the terrain in our area is relative gentle which makes topographic factors like slope, aspect and flat curvature stable. And the elevation was derived from the GPS, varying from 443.89 to 700.60 m for soil pro file location in the study area.

Soil organic matterSoil bulk density is generally considered to be closely related with soil organic matter due to the function of soil organic matter on the process of soil conforming (Zinnet al.2005). And the pedotransfer functions of soil bulk density prediction were usually based on soil organic matter (Keller and Håkansson 2010). Thus,given the importance of soil organic matter, in this study,we chose soil organic matter as one of random factors to evaluate the effects on the spatial variability of soil bulk density in each soil layer, considering that soil organic matter is largely in fluenced by anthropogenic activities like fertilization and straw return in agricultural soils. In our study area, the mean value of soil organic matter was estimated at 33.82, 16.79, 10.80 and 9.36 g kg–1in the 0–20, 20–40,40–60 and 60–100 cm layers, respectively.

Parent materialParent material is the key factor to determine the initial status of soil bulk density. The study area is formed in a gray alluvial plain by Mingjiang River.Thus, the dominated parent materials are gray and graybrown alluvium. Four kinds of alluvium including gray alluvium, quaternary old alluvium, purple alluvium and graybrown alluvium are distributed in the study area.

Soil typeSoil type, re flecting the difference of soil forming conditions and forming process, plays a decisive role in initial soil bulk density. In the study area, paddy soil is the dominated soil group, which contains four subgroups and eleven kinds of soil genus, followed by alluvial soil. The details of soil types were presented in Table 1.

Land useLand use affects soil porosity and compaction,which determines water in filtration, groundwater movements and surface-water run-off (Sheteet al.2016). In the studyarea, three land use types including cropping land, garden and agroforestry system are formed during the long term land use process. Cropping system in the study area contains two typical upland and paddy rotation systems including rice-wheat ration and rice-rape rotation. Garden is distributed in the urban and rural ecotone, which is planned for horticultural plants. Agroforestry system combined with vegetables and trees is formed to make better use of land.

Table 1 Details of soil types including soil group, subgroup and soil genus in Chengdu Plain

2.4. Spatial analysis

Geostatistics was widely used in the spatial analysis for regional variables (Li Q Qet al.2016; Yanget al.2016).In this study, we used geostatistics method combined with GIS technology to analyze the spatial variability of soil bulk density, and cross-validation was conducted to estimate the performance of spatial interpolation using mean error between predicted and measured values across sampling sites (Yanget al.2012). Firstly, semi-variance generated in GS+ ver. 7.0 software was used to describe the spatial structure of soil bulk density. Maps for the spatial distribution pattern of soil bulk density were then created with Ordinary Kriging method in the geostatistical module of ArcGIS 10.2 according to the analysis of semi-variance with a higher fitting coef ficient of determination (R2) and lower residual sum of squares (RSS).R2was calculated to re flect the model performance, parameter like range (A)was used to express the spatial autocorrelation extent. The spatial dependency of soil density is classi fied into strong,moderate, and weak type using the nugget effect [C0/(C+C0)]as a criterion (Li Q Qet al.2016). The spatial variability of soil bulk density was affected by structural factors when the nugget effect was under 25%. If the nugget effect was over 75%, random factors were the dominant factors for the spatial variability of soil bulk density. When the nugget effect was between 25 and 75%, the spatial variability of soil bulk density was controlled by both the structural and random factors. Normally, structural factors include soil parent materials, soil type, topography and other natural process of soil conformation while random factors were represented by tillage, fertility, land use and other anthropogenic activities.

2.5. Statistical analysis

General statistics, Pearson correlation, one-way ANOVA and regression analysis were performed in SPSS 19.0 software. Correlation analysis was conducted to determine the statistical relationships between soil bulk density and soil organic matter and elevation. One-way ANOVA with LSD test was performed to examine the difference of soil bulk density among the qualitative factors including parent materials, soil type and land use as well as the difference among soil layers. Stepwise regression analysis with forward selection elimination was further carried out to identify the controlling factors at each soil layer. The probability ofF-value was used to determine whether variables could be kept in regression function. The adjustedR2was computed to re flect the explanation power of each affecting factor as well as its relative importance on the spatial variation of soil bulk density as long as the regression equation was effective.

3. Results

3.1. Descriptive statistics of soil bulk density with layers

As shown in Table 2, soil bulk density in the 0–20 cm layer ranged from 0.92 to 1.54 g cm–3with the mean value of 1.26 g cm–3, which was significantly lower than that in the subsoil(20–100 cm). No obvious difference for bulk density was found among subsoil. The mean value of soil bulk density for subsoil was all approximately 1.54 g cm–3, which was 1.21 times higher than that for surface soil (0–20 cm). The coef ficient of variation (CV, %) of bulk density for surface soil was between 10 and 100% with a moderate spatial variation,while the rest was all below 10%, indicating a gentle variability.

The normal distribution for soil bulk density was examined by the Kolmogorov-Smirnov (K-S) test (Table 2). Results showed theP-value of K-S test for soil bulk density in the 0–20, 40–60 and 60–100 cm layers were all over 0.05,indicating a normal distribution. In constrast, theP-value for bulk density in the 20–40 cm layer was near 0.05, which indicated an approximately normal distribution. Thus,the data could be directly used for the geostatistics and regression analysis.

3.2. Spatial analysis for soil bulk density

Given that conventional statistics just re flect the overall status of soil bulk density without depicting spatial details,geostatistics combined with GIS technology were conducted to explore the spatial structure and spatial distribution across the study area.

Spatial structureThe best- fitted models for soil bulk density at each layer, taking theR2and RSS into account,were shown in Table 3 and Fig. 2. Spherical model was best fitted for soil bulk density in the 20–40 and 40–60 cm layers. In comparison, Exponential and Gaussian model fitted the best for soil bulk density in the 0–20 and 60–100 cm layers. The range of the semi-variance represented the spatial autocorrelation of soil bulk density. The spatial autocorrelation range increased with soil layer from 4.00 to 5.21 km while the nugget effects decreased as the spatial autocorrelation ranges decreasing. The nugget effects for soil bulk density in the 0–20 and 20–40 cm layers were 27.22 and 27.02%, respectively, indicating a moderate spatial dependence controlled by both structural and random factors.In contrast, nugget effects in the 40–60 and 60–100 cm layers were 12.06 and 3.46%, respectively, which impliedsoil bulk density in the layers had strong spatial dependence controlled by structural factors such as parent material,topography and soil type.

Table 2 Descriptive statistics of soil bulk density in Chengdu Plain

Table 3 Semi-variance analysis of soil bulk density in Chengdu Plain1)

Fig. 2 Semi-variance maps for soil bulk density in the 0–20 cm (A), 20–40 cm (B), 40–60 cm (C) and 60–100 cm (D) in Chengdu Plain.

Spatial distributionThe spatial distributions for soil bulk density in each layer, calculated based on the Ordinary Kriging interpolation, were shown in Fig. 3. Firstly, crossvalidation showed that the mean error of measured value and predicted value was 0.0006, 0.0021, 0.0007 and 0.0010 g cm–3in the 0–20, 20–40, 40–60 and 60–100 cm,respectively, indicating that Ordinary Kriging method was able to provide reliable estimates of spatial patterns of soil bulk density across the study area for each soil layer.The spatial distributions for bulk density in the study area presented similar pattern in each layer with higher value in north and lower in south. Nevertheless, significant differences were found in details. Specifically, soil bulk density in the 0–20 cm layer showed obvious spatial differentiation characteristics, decreasing from Northeast to Southwest. Obviously, surface soil bulk density mainly ranging from 1.25 to 1.35 g cm–3was significantly higher than the subsoil (20–100 cm). Soil bulk density in the 20–40 cm layer was mainly between 1.45 and 1.55 g cm–3, which was concentrated in the south and parts of the northern area.In contrast, soil bulk density in the 40–60 and 60–100 cm layers exhibited consistent pattern mainly varying from 1.55 to 1.65 g cm–3and the higher value appeared in north and west while lower in north and east.

3.3. Factors affecting the spatial variation of soil bulk density

Soil organic matter and elevationCorrelation analysis between bulk density and soil organic matter (SOM) and elevation in each layer was shown in Table 4. As shown,there existed significantly negative relationships between SOM and soil bulk density in each layer. It could also be found the correlation coef ficientrdecreased with layer from 0.681 to 0.380. Elevation had significantly positive relationship with soil bulk density in each soil layer. It was worthy that the correlation coef ficientrand thesignificant level between elevation and soil bulk density in the 40–60 cm were obviously larger than others.

Parent materialOne-way ANOVA analysis presented in Table 5 showed that soil bulk density varied with parent materials and layers. As shown, soil bulk density in the 0–20 cm layer for the four parent materials were all significantly lower than that in the 20–40, 40–60 and 60–100 cm layers while no obvious difference was found among subsoil. For each soil layer, no significant difference was tested for soil bulk density among the four parent materials in the 20–40 cm layer. In contrast, there were obvious differences among parent materials for 0–20, 40–60 and 60–100 cm layers. Specifically,there was obvious difference between purple alluvium and gray-brown alluvium of soil bulk density in the 0–20 cm layer. Significant difference between quaternary old allumium and purple alluvium of soil bulk density was tested in the 40–60 cm layer. Soil bulk density in the 60–100 cm layer for purple alluvium was 1.64 g cm–3, substantially higher than that for quaternary old alluvium, gray alluvium and gray-brown alluvium, while no obvious difference was tested among the three parent materials.

Fig. 3 Maps for the spatial distribution of soil bulk density (BD) in the 0–20 cm (A), 20–40 cm (B), 40–60 cm (C) and 60–100 cm(D) layer in Chengdu Plain.

Soil typeAs Table 6 shown, soil bulk density for soil group, subgroup and soil genus in the 0–20 cm layer was significantly lower than that in the 20–40, 40–60 and 60–100 cm, respectively, while no obvious difference was tested among subsoil. In each layer, no significant difference existed between soil group. Similarly, no obvious difference was found among subgroup in each layer expect bulk density for gleyed paddy soil and hydromorphic paddy soil in the 20–40 cm layer. Significant difference was found among soil genus in each layer. These combined results implied that compared with soil group and subgroup, soil genus could be used as the class unit to evaluate the effects of soil type on the spatial variability of soil bulk density.

Land useAs Table 7 shown, land use had significant effect on soil bulk density in the 0–20 cm layer, while no obvious difference was tested among the 20–40, 40–60 and 60–100 cm layers. Specially, in the 0–20 cm layer, soil bulk density for garden, cropland and agro-forestry was presented in a decreasing order of 1.32, 1.24 and 1.23 g cm–3. Even though no obvious difference was found among the layers in the 20–40, 40–60 and 60–100 cm for each land use, soil bulk density for subsoil was all significantly higher than that for surface soil.

Factors affecting the spatial variability of soil bulk density with layersStepwise regression analysis was further carried out to quantitatively describe the controlling factors on soil bulk density in each layer as well as their explanation ability for spatial variability. As shown in Table 8,theP-value of the regression functions were all under 0.01,suggesting that the regression models of soil bulk density in each layer were all effective. For each layer, soil organic matter and land use were the controlling factors affecting the spatial variability of soil bulk density in the 0–20 cm layer, which could jointly explain the 47.4% variability of soil bulk density. Soil bulk density in the 20–40 cm was dominantly controlled by SOM, which could explain the 32.8% variability of soil bulk density. SOM, soil genus and elevation could explain the 21.2% variability of soil bulk density in the 40–60 cm, estimated as the controlling factorsfor the spatial variability of soil bulk density in the soil layer.In contrast, in the 60–100 cm layer, SOM and soil genus were the dominated factors, and these two factors could explain the 21.7% variability of soil bulk density.

Table 4 Correlation analysis between soil bulk density and soil organic matter (SOM) in Chengdu Plain

4. Discussion

4.1. Distribution of soil bulk density with layer

As a fundamental soil physical property, soil bulk density is an important indicator for soil quality degradation. Generally speaking, soil bulk density for fertile plow layer is about 1.0 g cm–3. However, in our current study, the mean value of soil bulk density in the 0–20 cm layer was measured at 1.26 g cm–3(Table 2), which was significantly higher than that of 1.0 g cm–3. This result indicates soils of plow layer in Chengdu Plain are undergoing potential degradation trend.Consistent with the previous study by Hanet al.(2016), our research also found surface soil bulk density (0–20 cm) in our study area was substantially lower than subsoil (20–100 cm) no matter the mean value on the whole study area or under different parent materials, soil types and land use(Tables 2 and 5–7). It is because that surface soil is easily disturbed by human activities and biological factors, such as the input of organic matter, intercropping of crop roots and activities of soil animals especially in an agricultural intensive area, which can loose soil and make the surface soil bulk density significantly lower than subsoil (Sunet al.2008).Report by Sheteet al.(2016) revealed soil bulk density for Gambella and Oromia has increased with increasing soil depth, which was contrary to our study. However, results in our study showed no obvious difference was tested among subsoil (Table 2). Similarly, report conducted in Yellow River Delta by Yaoet al.(2006) also con firmed the variability of soil bulk density in the vertical direction was very small. The two distinctive discrepancy results indicate that soil bulk density exists a high spatial heterogeneity as a result of complex soil forming conditions, ecological processes and anthropogenic activities in different study area, and more efforts are need to better understand the spatial variability of bulk density as well as their affecting factors.

Table 5 Statistics of soil bulk density in different parent materials in Chengdu Plain

Table 6 Statistics of soil bulk densityin different soil types in ChengduPlain

4.2. Mechanism of the spatial variability of soil bulk density

In our study, we found the nugget effect for soil bulk density in the 0–20 and 20–40 cm layers was between 25 and 75%, indicating a moderate spatial dependency affected by both structural and random factors (Table 3). Analogous result was also conducted by Yanget al.(2016), who demonstrated the nugget effect for bulk density was 62.5%. Though similarly moderate spatial dependency was found both in the two researches, the nugget effect for latter was substantially higher than ours (27.22% in the 0–20 cm layer and 27.02% in the 20–40 cm, respectively),mainly because that the sampling depth in Yanget al.(2016) was limited by top 10 cm where soil is more vulnerable to external interference. In contrast, strong spatial dependence for soil bulk density was found both in the 40–60 and 60–100 cm layers where the nugget effects were both below 25% (Table 3), implying the spatial variability in the two layers were controlled by structural factors such as parent material, topography and soil type. The discrepancy of spatial dependency for soil bulk density at different layers results from that the deeper soil is in relatively original state while the upper soil is more vulnerable to be disturbed by external which in turn weak the spatial autocorrelation of the upper soil bulk density.

In agreement with the observation conducted by Yaoet al.(2006) in the Yellow River Delta, our study also found the nugget effects for soil bulk density decreased by turn as soil layer increasing according to the semiviariance analysis (Table 3). This result addressed that the effects of random factors decreased while structural factors played the dominated role as soil layer increasing. In addition, the spatial autocorrelation range for soil bulk density increased from 4.00 to 5.21 km as the nugget effects decreasing. These combined results revealed that the spatial dependency of soil bulk density increased as soil layer decreasing,and the enhanced spatial dependency has weaken the spatial autocorrelation distance.

4.3. Controlling factors affecting the spatial variability of soil bulk density in each layer

Although analysis above has evaluated the relative contribution of structural or random factors on the spatial variability of soil bulk density, more efforts are needed to identify which one of the structural or random factors is the dominated factor in each soil layer. In our study, soil elevation, parent material and soil types including soil group, subgroup and soil genus were characterized as structural factors while soil organic matter and land use were classi fied as random factors.According to the stepwise regression analysis (Table 8), we found soil organic matter was determined as the controlling factor on the spatial variability of soil bulk density in each layer (Table 8). Numerous researches have shown soil organic matter was significantly negatively related with soil bulk density (Denget al.2014; Wanget al.2014; Yanget al.2016). In agreement with the previous researches,our study also found soil bulk density in the study area played significantly negative relationship with soil organic matter in each soil layer (Table 4). Moreover, the correlation coef ficientrdecreased with increasing soil layer, indicating the effects of soil organic matter on bulk density became weaker as soil layer increasing. It is mainly because soil organic matter decreases with increasing soil layer due to the reduced exogenous input of organic material such as litter, roots and animal debris.

Land use combined with soil organic matter could explain 47.4% of the spatial variability for soil bulk density in the 0–20 cm layer (Table 8), indicating land use was another factor controlling the spatial variability of soil bulk density in the layer. Moreover, one-way ANOVA for land use (Table 7)implied no obvious difference existed among land use expect for surface soil bulk density, indicating a reduced role of random factors with increasing soil layer on the spatial variability of bulk density, which was in accordance with semi-variance analysis (Table 3). In agreement with observation by Sunet al.(2016) carried out in the Yili Valley of Xinjiang, our study also found soil bulk density for garden in the 0–20 cm layer was higher than that for cropland and agro-forestry garden in the study area is one of the most important land use for famers to obtain economic income and much fertilizer is applied into soils to make sure the growth of horticultural plants, which could cause soil hardened and consequently increase soil bulk density (Celiket al.2010).In addition, this result indicated that convert of garden to cropland and agro-forestry could reduce surface soil bulk density and slow down soil compactness in the study area.

Elevation, soil genus and soil organic matter were the three factors controlling the spatial variability of soil bulk density in the 40–60 cm (Table 8), which could explain 21.1% of the spatial variability together for soil bulk density in this soil layer. Although the study area is located in the Chengdu Plain with relatively gentle terrain, elevation was found to have significant relationship with soil bulk density in each soil layer (Table 4). Furthermore, elevation was determined as one of the dominated factors controlling the spatial variability of soil bulk density in the 40–60 cm(Table 8). These combined results indicated that we should take the effects of elevation into account when considering the spatial variability of soil bulk density in the study area,especially in the 40–60 cm layer. Soil genus, taking both soil parent material and soil forming process into consideration,was founded to be one of the factors controlling the spatial variability in the 40–60 and 60–100 cm layers (Table 8).Compared with soil group and subgroup, soil genus played a more important role on the spatial variability of soil bulk density (Table 6). Similarly, research by Zhanget al.(2015)also found the interpretation of soil types on soil pH was related to the classification level and the interpretation ability for soil group, subgroup and sol genus was 41.3, 57.3 and 83.7%, respectively. It could be explained by the information that soil genus includes more environment details than soil group, because the classification of soil genus depends on both parent material and the process of soil formulation while the classification of soil group only depends on the process of soil formulation. It is therefore soil genus rather than soil group nor subgroup contributing to the spatial variability ofsoil bulk density in the 40–60 and 60–100 cm layers.

Table 7 Statistics of soil bulk density in different land use types in Chengdu Plain

Table 8 Stepwise regression analysis of soil bulk density in each layer in Chengdu Plain

5. Conclusion

Our study showed that surface soil bulk density (0–20 cm)was significantly higher than subsoil (20–100 cm) and no obvious difference was found among subsoil. Similar spatial pattern was found for soil bulk density in each layer with higher value in north and lower in south on the whole, but obvious difference existed in detail. The spatial variability of soil bulk density in the 0–20 and 20–40 cm layers were synchronously controlled by structural and random factors while structural factors were the dominated factors affecting the spatial variability of soil bulk density in the 40–60 and 60–100 cm layers. Soil organic matter was the controlling factor on the spatial variability of soil bulk density in each layer. Besides soil organic matter, more attention should be given to land use when considering the spatial variability of soil bulk density in surface soil, while soil genus and elevation should also be considered in deeper soil.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (4120124), and the Science Fund of the Education Department of Sichuan Province, China(16ZB0048).