Elsevier

Science of The Total Environment

Volumes 599–600, 1 December 2017, Pages 2156-2165
Science of The Total Environment

Mapping cation exchange capacity using a Veris-3100 instrument and invVERIS modelling software

https://doi.org/10.1016/j.scitotenv.2017.05.074Get rights and content

Highlights

  • Veris-3100 data firstly inverted by a quasi-3d inversion algorithm.

  • Measured soil CEC strongly correlated with inverted electrical conductivity.

  • A LR model established to predict soil CEC with inverted electrical conductivity.

  • Soil CEC mapped at various depths (0–0.9 m) across a 16-ha field using the LR.

Abstract

The cation exchange capacity (CEC) is one of the most important soil properties as it influences soil's ability to hold essential nutrients. It also acts as an index of structural resilience. In this study, we demonstrate a method for 3-dimensional mapping of CEC across a study field in south-west Spain. We do this by establishing a linear regression (LR) between the calculated true electrical conductivity (σ - mS/m) and measured CEC (cmol(+)/kg) at various depths. We estimate σ by inverting Veris-3100 data (ECa - mS/m) collected along 47 parallel transects spaced 12 m apart. We invert the ECa data acquired from both shallow (0–0.3 m) and deep (0–0.9 m) array configurations, using a quasi-three-dimensional inversion algorithm (invVeris V1.1). The CEC data was acquired at 40 locations and from the topsoil (0–0.3 m), subsurface (0.3–0.6 m) and subsoil (0.6–0.9 m). The best LR between σ and CEC was achieved using S2 inversion algorithm using a damping factor (λ) = 18. The LR (CEC = 1.77 + 0.33 × σ) had a large coefficient of determination (R2 = 0.89). To determine the predictive capability of the LR, we validated the model using a cross-validation. Given the high accuracy (root-mean-square-error [RMSE] = 1.69 cmol(+)/kg), small bias (mean-error [ME] =  0.00 cmol(+)/kg) and large coefficient of determination (R2 = 0.88) and Lin's concordance (0.94), between measured and predicted CEC and at various depths, we conclude we were well able to predict the CEC distribution in topsoil and the subsurface. However, the predictions made in the subsoil were poor due to limited data availability in areas where ECa changed rapidly from small to large values. In this regard, improvements in prediction accuracy can be achieved by collection of ECa in more closely spaced transects, particularly in areas where ECa varies over short spatial scales.

Introduction

The necessity for the development of cost-effective soil mapping methods to enable precision agriculture techniques has long been realized. One of the most important soil properties which need to be mapped is the cation exchange capacity (CEC - cmol(+)/kg). This is because it is a measure of the total capacity of the soil to hold exchangeable cations, which is an indication of nutrient availability. It also indicates how well buffered soil pH might be and therefore it influences soil's reaction to fertilizers and other ameliorants (Hazelton and Murphy, 2016). In addition, the CEC is an index of the shrink-swell potential and therefore structural resilience to tillage (Triantafilis et al., 2009). For example, small CEC (i.e. < 20 cmol(+)/kg) indicates poor shrink-swell potential, whereas intermediate CEC (20–40 cmol(+)/kg) and large CEC (> 40 cmol(+)kg) suggest moderate and good shrink–swell potential, respectively (McKenzie, 1998).

However, the traditional way of measuring CEC is a costly and time-consuming process because laboratory analysis requires determination of major cations through a leaching process (Holmgren et al., 1977, Rayment and Higginson, 1992). Nevertheless, and in order to use the limited CEC data, various studies have shown that soil CEC can be mapped at the field level using classical geostatistical methods (Castrignanò et al., 2000, Asadzadeh et al., 2012). But the major disadvantage of using geostatistical analysis is that a large number of samples are required to map the target soil property, which also needs to be spatially correlated and strongly variable (Webster and Oliver, 1992, Jung et al., 2006). Another way is to make use of an easier to measure and cheaper to acquire soil property. This was the approach of Rashidi and Seilsepour (2008), who developed LR relationship to model CEC from organic carbon.

More recently, the use of ancillary data is being preferred to soil data, because much larger amounts of this type of data can be acquired much more quickly and cheaply. This can improve the accuracy and minimise the bias of prediction as well as provide a finer resolution map of the target soil property. One of the most commonly used sources of ancillary data is the measurement of the apparent electrical conductivity (ECa - mS/m). This is because this type of data is a function of type and concentration of ions in the soil solution, as well as the amount and type of clays, water content and soil temperature (McNeill, 1980).

Bishop and McBratney (2001) were one of the first to determine an environmental correlation between CEC and ECa; but they also found that terrain attributes, aerial photography, crop yield data and bare soil LANDSAT TM were useful ancillary data. In doing this they were able to develop a multiple linear regression (MLR) to map topsoil CEC (0–0.15 m). Similarly, Triantafilis et al. (2009) used a hierarchical spatial regression model to map the average profile (0–2 m) CEC using proximally sensed EM38 and EM31 ECa along with a digitized remote sensed image (Red, Green and Blue bands) and trend surface (Easting and Northing) data. Sudduth et al. (2005) used EM38 and Veris-3100 instruments and observed high and more persistent correlation between CEC and ECa data acquired across various fields. However, in all the above cases the information regarding the spatial variation of CEC with depth was not investigated given these examples either looked at only one depth increment or average CEC in the soil profile.

In order to map soil properties with depth, several authors have attempted to do this by establishing LR or MLR relationships between ECa and various depth increments. However, model errors can lead to loss of accuracy in prediction (Huang et al., 2015). More recently, various studies have shown how ECa can be inverted to make 2-d and 3-d calibration relationships between estimates of the true electrical conductivity (σ - mS/m) and various soil properties, including exchangeable sodium percentage (Huang et al., 2014), salinity (Zare et al., 2015), soil moisture (Huang et al., 2016, Huang et al., 2017a) and organic matter (Huang et al., 2017b). In this paper our interest is in seeing if we can invert Veris-3100 ECa data acquired from shallow (0–0.3 m) and deep (0–0.9 m) arrays, and then to use the estimates of σ at various depth increments to make a LR with CEC measured at the same depths. We validate the LR using a cross-validation technique for different calibration points considering their individual depth sections one at a time and the accuracy, bias and Lin's concordance attained by this approach are discussed.

Section snippets

Study area

The research was conducted at a farm located in the proximity of Alvarado, about 16 km east from Badajoz, south-western Spain. The farm is called “El Carrascal” and its area is around 16 ha. The study site is divided into two separate hedgerow olive orchards, including a rectangular shaped field which is located to the southwest of a larger square shaped field (Fig. 1a). The topography is dominated by gentle hills. In the substrate, limestones predominate over intrusive acidic rocks. According to

Preliminary ECa data analysis

Table 1 shows the summary statistics of ECa measured by the Veris-3100. The mean shallow (0–0.3 m) ECa was 15.4 mS/m with a minimum of 3.2 mS/m and maximum of 58.8 mS/m. The median was slightly smaller (11.9), with the shallow ECa slightly positively skewed (1.2) with a coefficient of variation (CV) of 64.2%. In comparison, the deep (0–0.9 m) ECa had a larger mean (26.7 mS/m) with a minimum of 12.5 and maximum of 66 mS/m. The median deep ECa was again slightly smaller (24.0 mS/m) than the mean with the

Conclusions

A Veris-3100 instrument was used across 47 parallel transects to collect shallow and deep ECa data across a study field in south-western Spain. Using soil samples acquired across 40 locations and various depth increments, including topsoil (0–0.3 m), subsurface (0.3–0.6 m) and subsoil (0.6–0.9 m), to successfully establish a calibration between measured CEC (cmol(+)/kg) and the estimates of true electrical conductivity (σ - mS/m). The latter generated using a quasi-three-dimensional algorithm

Acknowledgments

This research was co-financed by Junta de Extremadura and European Regional Development Fund (ERDF) through GR15050 (Research Group TIC008). The authors are grateful to Carlos Campillo, from CICYTEX (Junta de Extremadura), for providing soil data. Dr. John Triantafilis acknowledges the Organisation for Economic Cooperation and Development for providing him with a Co-operative Research Programme Fellowship to visit Universidad de Extramedura. We acknowledge funding support from two separate

References (39)

  • C. DeGroot-Hedlin et al.

    Occam's inversion to generate smooth, two-dimensional models from magnetotelluric data

    Geophysics

    (1990)
  • EMTOMO LDA

    InvVERIS Version-1.1, Lisbon, Portugal

    (2017)
  • P. Hazelton et al.

    Interpreting Soil Test Results: What do all the Numbers Mean?

    (2016)
  • G.G. Holmgren et al.

    A mechanically controlled variable rate leaching device

    Soil Sci. Soc. Am. J.

    (1977)
  • J. Huang et al.

    Spatial prediction of the exchangeable sodium percentage at multiple depths using electromagnetic inversion modelling

    Soil Use Manag.

    (2014)
  • J. Huang et al.

    An error budget for mapping field-scale soil salinity at various depths using different sources of ancillary data

    Soil Sci. Soc. Am. J.

    (2015)
  • JMP software, SAS Institute Inc.

    SAS Campus Drive, Building T, Cary, NC 27513-2414, USA

    (2014)
  • W.K. Jung et al.

    Spatial characteristics of claypan soil properties in an agricultural field

    Soil Sci. Soc. Am. J.

    (2006)
  • L.I.K. Lin

    A concordance correlation coefficient to evaluate reproducibility

    Biometrics

    (1989)
  • Cited by (18)

    • Machine learning for cation exchange capacity prediction in different land uses

      2022, Catena
      Citation Excerpt :

      However, in a slightly alkaline soils, CEC was increased with application of NPK fertilizers, and the increase in CEC was attributed to the production of much aromatic nature compounds of OM (Brar et al., 2015). The estimation of CEC in the laboratory is very time-consuming and costly (Chapman, 1965; Sparks et al., 1996; Dohrmann, 2006; Elfaki et al., 2015); therefore, several attempts have been made to quantify CEC from easily available soil properties using statistical, proximal sensing, and remote sensing approaches such as linear models (Sulieman et al., 2018; Mishra et al., 2019; Adam et al., 2021), partial least squares (PLS) and stepwise regression (Khaledian et al., 2016a), genetic expression programming (GEP) and multivariate adaptive regression splines (MARS) (Emamgolizadeh et al., 2015), hyperspectral visible near-infrared (Vis–NIR) spectroscopy data (Rehman et al., 2019), Veris-3100 instrument and invVERIS software (Koganti et al., 2017) from ground-based apparent electrical conductivity (ECa), and Sentinel-2A and multispectral satellite images (Vaudour et al., 2019). Nonlinear algorithms particularly random forest (RF) and support vector regression (SVR) have also been applied to quantify CEC.

    • Using depth specific electrical conductivity estimates to improve hydrological simulations in a heterogeneous tile-drained field

      2022, Journal of Hydrology
      Citation Excerpt :

      Geophysical methods, such as electrical conductivity (EC) surveys, are well suited to detect spatial soil heterogeneity at very high resolution (Krüger et al., 2013). Georeferenced soil EC measurements have proven suitable for mapping soil properties such as texture, water content, organic matter, and cation exchange capacity in non-saline soils (Corwin and Lesch, 2004; Heil and Schmidhalter, 2012; Huang et al., 2017; Koganti et al., 2017; Sheets and Hendricks, 1995). Recent advances in electromagnetic induction (EMI) instruments and the development of sophisticated inversion routines have enabled on-the-go measurement of apparent electrical conductivity (ECa) at multiple depths, allowing the determination of depth specific EC estimates.

    • Mapping cation exchange capacity using a quasi-3d joint inversion of EM38 and EM31 data

      2020, Soil and Tillage Research
      Citation Excerpt :

      In this regard, collection of ECa continuously and on close spaced transects would be recommended. Using this approach, Koganti et al. (2017) showed that they could better estimate σ for calibration by accounting for the short scale variation in CEC. This was because they collected ECa from a Veris3100 instrument, which was mobilised and coupled with a GPS and data logging capabilities, on transects spaced only 12 m apart.

    • Digital regolith mapping of clay across the Ashley irrigation area using electromagnetic induction data and inversion modelling

      2019, Geoderma
      Citation Excerpt :

      A similar approach was used across the cotton growing area southeast of Trangie, using reconnaissance EM38 and EM34 ECa, data and to understand clay stratigraphy (Triantafilis and Monteiro Santos, 2009). More recently, quasi-three-dimensional (Q-3D) inversion of DUALEM-1 ECa collected in planar and coplanar modes, has been shown to provide estimates of true electrical conductivity (σ – dS m−1) which can be correlated with topsoil (0–0.3 m), subsurface (0.3–0.6 m), and subsoil (0.6–0.9 m) clay (Koganti et al., 2017). In cotton growing areas of NSW, issues associated with water use efficiency and storage reservoirs are problematic, because leakage has led to isolated cases of point source salinization (Buchanan et al., 2017).

    • An Application of Soil Electrical Conductivity Measurement by Wenner Method in Paddy Field

      2024, International Journal of Engineering Trends and Technology
    View all citing articles on Scopus
    View full text