Three-dimensional stochastic simulations of soil clay and its response to sampling density

https://doi.org/10.1016/j.compag.2017.08.031Get rights and content

Highlights

  • The paper simulated soil clay content in three dimensions and revealed its response to sampling density.

  • Considering the root mean square error accurate plot, standard error map, and quartile deviation map, SD1 had the best effect of three-dimensional stochastic simulation, followed by SD4.

  • For soil clay, three-dimensional sampling can be appropriate to reduce the number of samples required in the lower horizon in order to reduce the sampling workload.

Abstract

Clay is an active component in the mechanical composition of soil. The quantitative study of the spatial distribution of soil clay content is crucial to soil microecological research and agricultural or environmental management. The main purpose of the paper was to simulate soil clay content in three dimensions and reveal its response to sampling density based on sequential Gaussian simulation. The results showed the following: (1) With a reduction in samples, especially in the A horizon, spatial correlation was relatively enhanced and randomness weakened. (2) The spatial distribution of soil clay showed soil had high clay content in the mid-eastern region of the Haidian District, Beijing, and clay content was generally low in the other areas; (3) With a decrease of sampling density, the simulated spatial distribution of clay became gradually more homogeneous. The stochastic simulation results for two kinds of sampling densities, i.e., SD1 and SD4 were closer to the original measured values; the general distribution was discrete and could more accurately reflect the local volatility of the original data distribution; (4) Considering the root mean square error (RMSE), accurate plot, standard error map, and quartile deviation map, SD1 had the best effect of three-dimensional stochastic simulation, followed by SD4; (5) For soil clay, three-dimensional sampling can be applicable to reduce samples required in the lower horizon in order to reduce the sampling workload.

Introduction

The clay content of soil has a significant effect on soil C mineralization, organic matter decomposition, and migration of soil microorganisms (Wang et al., 2003, He et al., 2009). Soil physics tells us that soil colloid dispersion decreases with increasing clay content because of increased aggregate strength (Kjaergaard et al., 2002). In addition, the clay content of soil is an important input parameter for models of soil water and salt transport, or for carbon and nitrogen circulation models such as Hydrus and De-Nitrification-Decomposition (DNDC) models; in addition, clay content is an important indicator used in soil pedotransfer functions such as the estimation of soil water capacity (Bagheri et al., 2005). Therefore, it is necessary to understand the distribution of soil clay content. Some effort has been devoted to the assessment of soil textures on a regional scale through one-dimensional simulation (Li et al., 1997, Li et al., 1999, Zhang and Li, 2008, Zhang et al., 2013, Zhang et al., 2014) and two-dimensional simulation at field (Weigand et al., 2001). However, the three-dimensional spatial distribution simulation of the clay content of soil, which may play a significant role in soil management and agricultural production, is still rare in the literature.

In recent years, some studies on three-dimensional spatial simulation of soil properties, such as soil strength (Castrignano et al., 2002), soil nitrate-nitrogen (Van Meirvenne et al., 2003), soil organic matter or soil carbon storage (Minasny et al., 2006, Minasny et al., 2014, Meersmans et al., 2008, Mishra et al., 2009), soil texture horizons (He et al., 2009) and soil salinity (Li et al., 2013a, Li et al., 2013b) have been carried out based on the three-dimensional geostatistics and profile distribution function method. The application of the conditional stochastic simulation method to soil science and other disciplines has attracted increased attention after the defects of the kriging method were overcome so that the spatial structure of the original variables can be reproduced very well, and the uncertainty of the prediction results of simulation were quantifiable. Scale (e.g., sampling magnitude, sampling size, and sampling support) is frequently used in the study of geography recently, which is the spatio-temporal dimension of the study object or process, the spatio-temporal unit for information collection and processing, or the change pattern determined by spatio-temporal scope (Chen and Cai, 2010). The spatial variability of soil properties is the scale function, and autocorrelation of the same variable is very different at different scales. However, the response of a three-dimensional simulation of the clay content in soil to sample density, which will provide a basis for the optimization of sampling plan and reducing sampling cost, has still rarely been reported.

In this study, a size 55 m × 60 m field was selected as the study area. The following research topics were examined in the article under five sampling densities.

  • (1)

    Three-dimensional spatial variability of soil clay and its response to different sampling densities through nested variogram analysis and sequential Gaussian simulation (sGsim).

  • (2)

    The simulation effect and its response to sampling density under the same simulation method (i.e., sGsim) through comparison of three-dimensional simulation accuracy, local and spatial uncertainty.

Section snippets

Soil sampling and data processing

The 55 m × 60 m size study area was in the Haidian District, Beijing, and was divided into a grid of 156 subplots of a size of 5 m × 5 m. Soil samples were collected to a 1-m soil depth in the 100 randomly selected grid points (Fig. 1). Each sample hole was re-packed using soil from the natural soil horizons (i.e. A horizon 0–30, B horizon 30–60, C horizon 60–100 cm). A total of 300 samples were obtained from the A, B, and C horizons (i.e., 100 for each horizon). The soil clay content was determined by

Exploratory data analysis for different sampling densities

Statistical characteristics were calculated and the probability distribution was tested through the Kolmogorov-Smirnov (K-S) method (Table 1).

Table 1 showed that the mean, standard deviation and coefficient of variation consistency of soil clay for five sampling densities were basically the same. Based on the general evaluation to the coefficient of variation (CV) value, it is called weak variability when CV < 10% and medium variability when 10% < CV < 100%. CV values from SD1 to SD5 were 29.89%,

Discussions

At present, the three-dimensional simulation research of soil particle composition was mainly focused on methods, and how to provide prediction accuracy and reduce uncertainty (Chen et al., 2015, Benedetto et al., 2012). There is less research on the response to the sampling scheme. Based on the sGsim method, this paper discussed a three-dimensional simulation of soil clay and its response to five changes in sampling density. The results of variogram analysis show that the macro-structural

Conclusions

Three-dimensional simulation of soil clay under five different sampling densities was realized based on the SGsim method. The research multi-perspectively revealed the response of different sampling densities to three-dimensional spatial simulation results from the nested variogram, simulation accuracy, as well as local and spatial uncertainty. The following conclusions may be drawn:

  • (1)

    Soil clay showed a moderate spatial correlation in the study area. With a reduction in the sampling number,

Acknowledgements

This work was supported by the National Key Research and Development Program of China (2016YFD0300801) and the National Natural Science Foundation of China (41471186; 41571217). We acknowledge all the reviewers and editors of the journal for their valuable comments, suggestions, and revisions on this paper. We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

References (41)

  • S. Zhang et al.

    Spatial prediction of soil organic matter using terrain indices and categorical variables as auxiliary information

    Geoderma

    (2012)
  • S. Zhang et al.

    Prediction of soil moisture scarcity using sequential gaussian simulation in an arid region of china

    Geoderma

    (2017)
  • S.W. Zhang et al.

    Spatial interpolation of soil texture using compositional kriging and regression kriging with consideration of the characteristics of compositional data and environment variables

    J. Integrat. Agricult.

    (2013)
  • Agterberg, F.P., 2001. Multifractal simulation of geochemical map patterns//Merriam DF, Davis JC. Geologic Modeling and...
  • F. Bagheri et al.

    Estimation of available water capacity in clay soil by using pedo-transfer functions

    J. Agricult. Sci.

    (2005)
  • D.D. Benedetto et al.

    Integrating geophysical and geostatistical techniques to map the spatial variation of clay

    Geoderma

    (2012)
  • M. Bianchi et al.

    Sgems: a free and versatile tool for three-dimensional geostatistical applications

    Groundwater

    (2009)
  • H. Bourennan et al.

    Uncertainty assessment of soil water content spatial patterns using geostatistical simulations: an empirical comparison of a simulation accounting for single attribute and a simulation accounting for secondary information

    Ecol. Model.

    (2007)
  • C.A. Cambardella et al.

    Field-scale variability of soil properties in central iowa soils

    Soil Sci. Soc. Am. J.

    (1994)
  • X.R. Chai et al.

    Random simulation of soil organic matter using elevation as auxiliary information

    Transact. Chin. Soc. Agricult. Eng.

    (2008)
  • Cited by (0)

    View full text