Advertisement

Spatial analysis of soil properties using GIS based geostatistics models

  • Pravat Kumar ShitEmail author
  • Gouri Sankar Bhunia
  • Ramkrishna Maiti
Short Communication

Abstract

Accurate assessment of the spatial variability of soil properties is key component of the agriculture ecosystem and environment modeling. The main objective of the present study is to measure the soil properties and their spatial variability. A combination of conventional analytical methods and geostatistical methods were used to analyze the data for spatial variability. In November 2014 a total of 32 soil samples were collected in the field through random sampling in Medinipur Sadar block of Paschim Medinipur district in West Bengal (India). Soil properties of pH, electric conductivity (EC), phosphorus (P), potassium (K), and organic carbon (OC) were estimated using the standard analytical methods. A classical ordinary kriging (OK) interpolation was used for direct visualization of soil properties. The spatial distribution of EC, pH, and OC in soil are influenced by structural factors, such as climate, parent material, topography, soil properties and other natural factors. The semivariograms of the six soil properties were fit with exponential curve and root mean square error value is near about zero (0). Finally, spatial distribution and correlation between OC and other soil properties is shown by overlay of maps in GIS environment. The present study suggest that the OK interpolation can directly reveal the spatial distribution of soil properties and the sample distance in this study is sufficient for interpolation.

Keywords

Soil properties Organic carbon GIS Ordinary kriging Semivariograms 

Introduction

Sustainable land management requires reliable information on the spatial distribution of soil properties affecting both landscape process and services (Lin et al. 2005; Shibu et al. 2006). In conventional soil survey soil properties are recorded at representative sites and assigned to entire mapping unit, which are delineated using both physiographic and geopedologic approaches. Although soil surveyors are very well aware of the spatial variability of soil properties, conventionally prepared soil maps do not reflect it as soil units are limited by boundaries (Heuvelink and Webster 2001). But in nature the soil properties are highly variable spatially (Burrough 1993) and for accurate estimation of soil properties these continuous variability should be considered. The traditional method of soil analysis and interpretation are laborious, time consuming, hence becoming expensive. Geostatistical techniques (kriging) are widely recognized as an important spatial interpolation technique in land resource inventories (Hengl et al. 2004; Bhunia et al. 2016).

Geostatistical methods quantify spatial distribution and variability based on the spatial scale of the study area, distance between sampling points and spatial pattern of modeling semivariograms. They have been widely applied to evaluate spatial correlation in soils and to analyze the spatial variability of soil properties, such as soil physical, chemical and biological properties (Fromm et al. 1993; Wigginton et al. 2000; Vieira et al. 2007; Zheng et al. 2009; Liu et al. 2014).

In India majority of soil maps were prepared by conventional methods and a very little work has been done so far to use the modern spatial prediction techniques in this regard (Saha et al. 2012; Pal et al. 2014; Behera and Shukla 2015; Tripathi et al. 2015; Bhunia et al. 2016). The accurate estimation of spatial distribution of soil properties [soil pH, organic carbon (OC), electrical conductivity, phosphorous, potassium, etc.] is important in precision agriculture and is one of the bases for decision and policy makers to make plans and strategies. So, research in environmental monitoring, modeling and precision agriculture need good quality and inexpensive soil data. The aim of this paper is to evaluate the potential for measuring soil properties [electric conductivity (EC), pH, K, P, and OC] and its spatial variability using geostatistical methods.

Materials and methods

Study area

The present study was conducted in Medinipur Sadar block of Paschim Medinipur district in West Bengal (India) extending between 22°23′45″N–22°32′50″N latitude and 87°05′40″E–87°31′01″E longitude covering an area of 353 km2 (Bhunia et al. 2016). The geomorphology of the block is undulating surface topography predominantly covered by laterite soil. The sand, silt and clay contents of the soil are 45, 35 and 20 %, respectively. The average annual precipitation in the study area is 1,800 mm. In addition, 70 % of the annual rainfall is this region occurs between June and August. During these months the rainfalls is intense and causes extensive erosion. The study area is under two main land use types cover agricultural land (56.91 %) and forest land (24.76 %).

Soil sampling and measurement

In November 2014 a total of 32 soil samples were collected in the field through random sampling from whole of the study area. A portable global positioning system (GPS) was used to collect each sample site. The undisturbed soil samples at depths of 0–20 cm, was collected with five soil cores each, and well mixed into a composite soil sample. Soil samples were air dried and passed through a 2 mm sieve for laboratory analysis of soil texture. Soil pH, EC, phosphorus (P), potassium (K), and soil OC were determined using the standard analytical methods (Table 1).
Table 1

Analytical methods of estimation of different physio-chemical parameters of soil

Parameters

Methods

Electrical conductivity (EC)

Conductivity bridge method (Richards 1954)

Soil pH

Digital pH meter

Potassium (K)

Flame spectrophotometer (Jackson 1958)

Organic carbon (OC)

Walkley–Black wet oxidation method (Nelson and Sommers 1996; Bao 2000)

Phosphorus (P)

Spectrophotometer (Bremner 1996)

Geostatistical methods

Geostatistical method is a spatial distribution and variability analysis method that was developed from classical statistics. The ordinary kriging (OK) interpolation method was used for prediction of the values of the unmeasured sites (un-samples locations) x 0 by assuming the z*(x 0) equals the line sum of the known measured value (field measured value). Kriging process is calculated by the following equation (Wang 1999):
$$Z^{*} (x_{0} ) = \sum\limits_{i = 1}^{n} {\lambda_{i} z(x_{i} )}$$
(1)
Where z*(x 0) is the predicted value at position x 0, Z(x i ) the known value at sampling site x i , λ i the weighting coefficient of the measured site and n is the number of sites within the neighborhood searched for the interpolation.
Semivariograms were used as the basic tool to examine the spatial distribution structure of the soil properties. Based on the regionalized variable theory and intrinsic hypotheses (Nielsen and Wendroth 2003), a semivariogram is expressed as:
$$\gamma (h) = \frac{1}{2N(h)}\sum\limits_{i = 1}^{N(h)} {[Z(x_{i} ) - Z(x_{i} + h)]^{2} }$$
(2)
where γ(h) is the semivariance, h the lag distance, Z the parameter of the soil property, N(h) the number of pairs of locations separated by a lag distance h, Z(x i ), and Z(x i  + h) are values of Z at positions x i and x i  + h (Wang and Shao 2013). The empirical semivariograms obtained from the data were fitted by theoretical semivariogram models to produce geostatistical parameters, including nugget variance (C 0), structured variance (C 1), sill variance (C 0 + C 1), and distance parameter (λ). The nugget/sill ratio, C 0/(C 0 + C 1), was calculated to characterize the spatial dependency of the values. In general, a nugget/sill ratio <25 % indicates strong spatial dependency and >75 % indicates weak spatial dependency; otherwise, the spatial dependency is moderate (Cambardella et al. 1994).

Cross-validation

Cross-validation technique was adopted for evaluating and comparing the performance of OK interpolation method. The sample points were arbitrarily divided into two datasets, with one estimate mean value against measured mean were used to validate the model. The root mean square error (RMSE) is error based measures to evaluate the accuracy of interpolation methods.
$${\text{RMSE}} = \sqrt {\frac{{\sum\nolimits_{i = 1}^{N} {(0_{i} - S_{i} )^{2} } }}{N}}$$
(3)
Where 0 i is observed value and S i is the predicted value, N is the Number of samples.

Result and discussion

Geostatistical analysis

Table 2 shows the soil properties where variable characteristics was generated from semivariogram model. C 0 is the nugget variance; C is the structural variance, and C 0 + C represents the degree of spatial variability, which affected by both structural and stochastic factors (Fig. 1). The higher ratio indicates that the spatial variability is primarily caused by stochastic factors, such as fertilization, farming measures, cropping systems and other human activities. The lower ratio suggests that structural factors, such as climate, parent material, topography, soil properties and other natural factors, play a significant role in spatial variability. The value of <0.25, 0.25–0.75, and >0.75 can show strong, moderate and weak spatial autocorrelation in soil properties, respectively. As shown in Table 2, the C 0/C 0 + C ratio values for P, K, pH, OC and EC were 0.08, 0.05, 0.28, 0.56, and 0.04, respectively. The nugget/sill ratio were between 25 and 75 % of OC, indicating moderate spatial correlation with 2,219 m ranges and impact of stochastic and structural factors. The C 0/C 0 + C ratio were less 25 % in the four soil properties such as P, K, pH, and EC indicates a strong spatial correlation. The spatial correlation was apparent in the 925–2,018 m ranges and was affected by structural factors.
Table 2

Geo-statistical parameters of the fitted semivariogram models for soil properties and cross-validation statistics

Soil property

Model

Nugget (C 0)

Sill (C 0 + C)

Range (m)

Nugget/sill

R 2

RMSE

P (ppm)

Exponential

0.26

3.309

1,100

0.08

0.58

0.480

K (ppm)

Exponential

0.24

4.662

1,131

0.05

0.53

0.590

pH

Exponential

0.06

0.218

2,018

0.28

0.44

0.365

Organic carbon (%)

Exponential

0.54

0.970

2,219

0.56

0.39

0.858

EC (µs)

Exponential

0.04

1.040

925

0.04

0.61

0.381

R 2 coefficient of determination, RMSE root mean square error

Fig. 1

Semivariogram parameters of best-fitted theoretical model to predict soil properties, a EC, b K, c P, d soil pH, and e SOC

The semivariance function model fits exponential curve for each soil properties. The exponential curve gradually increased with increasing spatial distance before stabilizing. All five soil properties have coefficients of determination R 2 values of 0.39–0.61 and a small RMSE. The R 2 was calculated to measure the goodness of fit. The R 2 of all variables, except for pH, and OC were >0.5, indicating a good fit (Table 2). The pH and OC had a moderate fit, with R 2 values of 0.44 and 0.39, respectively. The RMSE were all approximately near to 0, but the theoretical model for K and OC showed RMSE values of 0.590 and 0.858, respectively. These results indicate that the theoretical model was an adequate representation of the spatial structural properties of soil.

Spatial distribution of soil properties

Figure 2a–d shows the spatial distribution of soil properties of EC, pH, K, P and OC using OK interpolation method. An OK technique was used to switch point soil samples into continuous fields of soil properties. The spatial correlation map of soil properties (EC, pH, K, and P) and measured OC was produced, compared and analyzed for the results. Spatial variability maps among the soil pH, EC, K, P and predicted OC was prepared using ArcGIS to represent the dependence of OC (Fig. 2). Concentration of OC was observed in the north-western part of the study area. Soil pH found to be an important factor in the analysis of the soil inorganic carbon (SIC), but its contribution towards organic carbon cannot be overruled. The spatial map of pH is generated from the measured pH value from the collected samples in the study sites. To the central and northeastern portion of the study site, higher pH is concentrated.
Fig. 2

Spatial relationship between SOC and soil properties, a EC and SOC, b soil pH and SOC, c K and SOC, and d P and SOC

The value of P was highly concentrated in the eastern and south western part, whereas the higher OC was observed in the northwestern corner of the study site. Spatial distribution of P did not varied extremely with organic carbon; however, well reflected changes were observed in spatial distribution for OC and pH. The EC varied with organic carbon and some portion showed concentrated EC. The K value showed some pockets of concentration. However, regression analysis was preformed between soil properties and predicted OC. The statistical regression among the predicted OC and EC, pH, K, and P of R 2 values were 0.32, 0.45, 0.38 and 0.41, respectively. These results suggest that the certain management practices, e.g., minimum tillage, cover crops, and rotations, are to be used to recover OC of the topsoil in the study area. The greater amount of soil OC was perhaps due to the maximum concentration of root mass, waste material and secrete root exudates, growing physical steadiness and microbial activity (Holeplass et al. 2004; Kukal et al. 2008). In our analysis, it was found that storage of OC in soil was mainly influenced by structural factors, such as climate, parent material, topography, soil properties and other natural factors played a significant role in spatial variability.

Conclusion

Understanding the spatial distribution and accurate mapping of soil properties at large scale are essential for precision farming, environmental monitoring, and modeling. This study showed that geostatistical models were fitted for six soil properties, namely phosphorus (P), potassium (K), pH, electrical conductivity (EC), and organic carbon (OC). Cross-validation of variogram models through OK showed that spatial prediction of soil properties is better than assuming the mean of the observed values at any unmeasured location. Finally, six prediction maps were developed using best fit semivariogram models with OK. Our results suggest that the ordinary kriging interpolation can directly reveal the spatial distribution of soil properties and the sample distance in this study is sufficient for interpolation. However, future studies are needed to clarify the spatial variability on the larger scale and better understand the factors controlling spatial variability of soil properties.

References

  1. Bao SD (2000) Soil and agricultural chemistry analysis. China Agriculture Press, Beijing, p 495Google Scholar
  2. Behera SK, Shukla AK (2015) Spatial distribution of surface soil acidity, electrical conductivity, soil organic carbon content and exchangeable potassium, calcium and magnesium in some cropped acid soils of India. Land Degrad Dev 26:71–79CrossRefGoogle Scholar
  3. Bhunia GS, Shit PK, Maiti R (2016) Comparison of GIS-based interpolation methods for spatial distribution of soil organic carbon (SOC). J Saudi Soc Agric Sci. doi: 10.1016/j.jssas.2016.02.001 Google Scholar
  4. Bremner JM (1996) Nitrogen-total. In: Sparks DL (ed) Methods of soil analysis—part 3—chemical methods. Soil Science Society of America, American Society of Agronomy, Madison, pp 1085–1121Google Scholar
  5. Burrough P (1993) Soil variability: a late 20th century view. Soils Fertil 56:529–562Google Scholar
  6. Cambardella CA, Moorman TB, Novak JM, Parkin TB, Karlen DL, Turco RF, Konopka AE (1994) Field-scale variability of soil properties in central Iowa soils. Soil Sci Soc Am J 58:1501–1511CrossRefGoogle Scholar
  7. Fromm H, Winter K, Filser J (1993) The influence of soil type and cultivation system on the spatial distributions of the soil fauna and microorganisms and their interactions. Geoderma 60:109–118CrossRefGoogle Scholar
  8. Hengl T, Rossiter DG, Stein A (2004) Soil sampling strategies for spatial prediction by correlation with auxiliary maps. Aust J Soil Res 41(8):1403–1422CrossRefGoogle Scholar
  9. Heuvelink GBM, Webster R (2001) Modeling soil variation: past, present, and future. Geoderma 100(3–4):269–301CrossRefGoogle Scholar
  10. Holeplass H, Singh BR, Lal R (2004) Carbon sequestration in soil aggregates under different crop rotation and nitrogen fertilization in an inceptisol in southeastern Norway. Nutr Cycl Agroecosyst 70:167–177CrossRefGoogle Scholar
  11. Jackson ML (1958) Soil chemical analysis. Prentice-Hall, Englewood Cliffs, pp 111–133Google Scholar
  12. Kukal SS, Kaur M, Bawa SS (2008) Erodibility of sandy loam aggregates in relation to their size and initial moisture content under different land uses in semi-arid tropics of India. Arid Land Res Manag 22:216–227CrossRefGoogle Scholar
  13. Lin H, Wheeler D, Bell J, Wilding L (2005) Assessment of soil spatial variability at multiple scales. Ecol Model 182:271–290CrossRefGoogle Scholar
  14. Liu L, Wang H, Dai W, Lei X, Yang X, Li X (2014) Spatial variability of soil organic carbon in the forestlands of northeast China. J For Res 25(4):867–876CrossRefGoogle Scholar
  15. Nelson DW, Sommers LE (1996) Total carbon, organic carbon, and organic matter. In: Sparks DL, Page AL et al (eds) Methods of soil analysis, part 3. Chemical methods, vol 5. Soil Science Society of America Book Series, Wisconsin, pp 961–1010Google Scholar
  16. Nielsen DR, Wendroth O (2003) Spatial and temporal statistics—sampling field soils and their vegetation. Catena Verlag GMBH, ReiskirchenGoogle Scholar
  17. Pal S, Manna S, Aich A, Chattopadhyay B, Mukhopadhyay SK (2014) Assessment of the spatio-temporal distribution of soil properties in East Kolkata wetland ecosystem (A Ramsar site: 1208). J Earth Syst Sci 123(4):729–740CrossRefGoogle Scholar
  18. Richards LA (1954) Diagnosis and improvement of saline and alkali soils. Agriculture handbook 60. US Department of Agriculture, Washington, DC, p 160Google Scholar
  19. Saha D, Kukal SS, Bawa SS (2012) Soil organic carbon stock and fractions in relation to land use and soil depth in the degraded Shiwaliks hills of lower Himalayas. Land Degrad Dev. doi: 10.1002/ldr.2151 Google Scholar
  20. Shibu M, Leffelaar P, Van Keulen H, Aggarwal P (2006) Quantitative description of Soil organic carbon matter dynamics—a review of approaches with reference to rice-based cropping systems. Geoderma 137:1–18CrossRefGoogle Scholar
  21. Tripathi R, Nayak AK, Shahid M, Raja R, Panda BB, Mohanty S, Kumar A, Lal B, Gautam P, Sahoo RN (2015) Characterizing spatial variability of soil properties in salt affected coastal India using geostatistics and kriging. Arab J Geosci. doi: 10.1007/s12517-015-2003-4 Google Scholar
  22. Vieira VA, Mello CR, Lima JM (2007) Spatial variability of soil physical attributes in small watershed. Ciencia e Agrotecnologia 31(5):1477–1485 (in Portuguese) CrossRefGoogle Scholar
  23. Wang ZQ (1999) Geostatistics and its applications in ecology. Beijing Science Press, BeijingGoogle Scholar
  24. Wang YQ, Shao MA (2013) Spatial variability of soil physical properties in a region of the Loess Plateau of PR China subject to wind and water erosion. Land Degrad Dev 24(3):296–304CrossRefGoogle Scholar
  25. Wigginton JD, Lockaby BG, Trettin CC, Nelson EA, Kolka RK, Wisniewski J (2000) Soil organic matter formation and sequestration across a forested floodplain chronosequence. In: Proceeding of workshop held at Clemson University, 12–14 April 1999, pp 141–155Google Scholar
  26. Zheng Z, Zhang FR, Ma FY (2009) spatiotemporal changes in soil salinity in a drip irrigated field. Geoderma 149:243–248CrossRefGoogle Scholar

Copyright information

© Springer International Publishing Switzerland 2016

Authors and Affiliations

  • Pravat Kumar Shit
    • 1
    Email author
  • Gouri Sankar Bhunia
    • 2
  • Ramkrishna Maiti
    • 3
  1. 1.Department of GeographyRaja N.L.Khan Women’s CollegeMedinipurIndia
  2. 2.Bihar Remote Sensing Application CenterPatnaIndia
  3. 3.Department of Geography and Environment ManagementVidyasagar UniversityMedinipurIndia

Personalised recommendations