PERPUSTAKAANSULTAN ABDUL SAMAD

UNIVERSITI PUTRA MALAYSIA

PENERBITAN PEGAWAI

Application of Geostatist ical Tools to QuantifySpatial Variabil i ty of Selected Soil ChemicalProperties from a Gultivated Tropical Peat

S. K. Balasundram, M. H. A. Husni and O. H.Ahmed

PFP20

Joumal ofAgronomy 7 (1): 82-87,2008ISSN 1812-5379O 2008 Asiar.r Network for Scientific Information

Application of Geostatistical Tools to Quantify Spatial Variability of Selected Soil ChemicalProperties from a Cultivated Tropical Pent

rS K. Balasundram, 2M H A. Husni and rO.H. Ahmed'Department of Agriculture Technology,

zDepartment of Land Managemen! Faculty of Agriculture, Universiti Putra N'Ialaysia,

- ̂ 43400 Serdang, Selangor, Malaysia

'Department of Crop Science, Facuhy of Agriculture and Food Sciences,Universitr Putra Malaysia, 970118 Bintulu, Sarawak, Malaysia

Abstract: Quantification of spatial variability is a vital prerequisite for precrsion agriculture. This study wasaimed at quantifying the spatial variability of selected chemical properties in a tropical peat cultivated withpineapple. A 1-ha study plot was estab[shed in a commercial pineapple plantation in Simpang Reugam, Johor.Georet'erencer-l topsoil sampies 1n = 6c)) were obtained systematically from 8'18 rn spacings in the x and ydnectrorl rcspectrvely. These samples were tested for total C, extractable P, K, Cu, Zn arrd B. Soil data were firstexplored usLng umvariate statistics, including normality check, non-spatial outlier detection and datatransformation, This was ibllowed by variography and kriging analvses to quantrly the spatial variability ofchemical properties. Results revealed a high degree of spatial variabrlity in the majorrty of chemical properties,which exhibited non-normal drstributiorrs with CVs rangrng from 12 to 54Vo. A1l propertres exhibited a definablespatial structr.ue, which were described by either spherical or exponential models. Carbon, P and B showedstrong spatial dependence. The majority olproperties had a short effective ra:rge. Surface maps of chemicalproperties clearly showed spatial clutering of test values. Excepting K, all other properties showed acceptableaccurac1/ of interpolated vaiues. These combured data suggest the need for a site-specific approach inmanagrng tropical peat cultrvated with pineapple, partrcularly with regard to nutrient managenrent.

Key rvords: Precision agriculture, spatial variabilitv, tropical peat, prneapple

INTRODUCTION Where:

h = The separation distance between locattonGeostatrstics are based on the concepts of \ or x,+h

regionalized variables, random functrons and stationarity 1 or 4*r : The measured values for the regionalized(Trangmar et al., 1985) and include two fundamental variable at location4 orx,-ncomponents (Isaaks and Srivastava, 1989): (i) spatial n(h) = The number of pails at any separationcontinuity analysis and (ii) interpolatron. Spatial distance hcontinuity analysis is most commonly performed lstngvariography lvhrle urteqtolation is often carried out usu'rg In plactice, the semivariogram is modeled r.sutgkriging. Vanography uses sentivariograms to characterize several authonzed rnodels (Oliver, i987; Isaaks andandmodel the spatial variance of data, whilekriging Srivastava, 1989) such as Gaussiarl spherical anduses the modeled variance to estrmate values between exponontial. These models are then fitted to thesatttples. semivariogram data. Key leahres of a semivariogram

The sentivaliogram, rvhich basically measures tire model are described by tfuee parameters, narnely nugge!reductron 1n variance between sample points as sill and range. Nugget is a measr.re of the anto1nt ofseparation distance decreases, can be estimated by the variance imposed by errors in sampltng, measurementfollowing formula(Burgess andWebster, 1980): and other unexplained source(s) of variance. Sill refers

to the total vertical scale of ilre variogram whereby the

r(h)--05n(h))l[',-u.,]' (r) ::il11'T::'::'1ffi"::J]s'il;J t'ffJ'..:H;::

CorreqrondingAuthor: S.K. Balasundram, Departrnent ofAgnculture Technology, Facdb/ ofAgriculnre,Uruversiti Putra Malaysia, 43400 Serdang, Selangcrr, Malaysra

82

J. Agron., 7 (1) : 82-87, 2008

distance that reflects a cutoff between spatial dependence

and spatial independence. This implies that at separabon

distances greater than the range, sampled pornts cease to

be spatrally correlated (i.e,, random).A sequel to variography is interpolation, a technique

aimed at estrmatmg regionalzed variables at unsampled

locatioru usmg the structual properties of the

semivinriogram and the initial data set (Trangmar el a/.,

1985) Several studres have compared kriging and

altemative methods of interpolation such as inverse

drstance rveighting and cubic splines, Use of inverse

distance weighting is advocated for data exhrbiturg

short-range vanability (Cooke et a1.,7993) whereas spline-

based interpolation is recommended for data sets that

show an abn-rpt change in values across a short distance(Voltz and Webster, 199)). \\rhelan e/ a/. (1996) forurd that

rnverse distance weighring is more effective ttran krrgurg

rvhen interpolatrng based on a small nrrmber of

obs erv atioru col lected at m oderate intensity.

Geostatistical analyses prov ide quantitative

information about spatial variability, wllch is a prrncipal

requirement for Site Specific Crop Management (SSCM).

SSCN{ rs a fundamental component of Prectston

Agnculture (PA), which embodies a holisttc farm

management sbategy where farm operators can adjust

input ue and cultrvation methods, includhg seedfertilizer, pesticide and water applicatrorq variety selection"

planturg, trllage and hawesting, to match varying soil,crop and other field attributes (Robert, 1999) PA

contlasts wrth conventional agriculture that is based onr.u-riibrm treatment(s) across a field.

This studf is part of an ongoing research program todiagnose site-specific strategies suitable for pineapple

management on tropical peat. The objective of tlls sh:dywas to quantify the spatial varrability of selected chemicalpropertics in a tropical peat culhvated with pineapple.

MATERIALS AND METHODS

This shrdy rvas conducted in a commercial pineappleplantation located in Simpang Rengam, Johor. Pineapple

rn thrs plantation is cultrvated on a deep peat (classified

as Saprig), which had an average pH of 3.1 and electricalconductivity of 454 pS cm-' and its cultivation is basedon a 2-year cropplng cycle that includes a 6-month fallowpenod.

A l-ha study plot was demarcated based on

crop variety (i e , Gandul) The study plot comprised95 planting beds with each bedmeasuring 0.61100 m

and spaced at 0.9 m between one and the other.The study plot had a plant spacing of 0.3 m and a net

stand density of 55,000 plants ha-'. Topsoil samples

(0-25 cm depth) were obtained systematically lrom

60 georeferenced points. Sampling points were spaced

8 m in the x direction (inter-bed) and 18 m rr the y

drection (intra-bed). Soil samples were tested for total C,

extractable P, K, Cu, Zn and B usrng standard laboratory

procedures.

Soil data were first subject to Exploratory Data

Analys is (EDA) involving ur.rivariate statistics, includng

normality check and non-spatial outlier detection. Where

necessary, non-normal data were fa:rsformed using the

appropriate ftrrction. Following the EDA, sorl data were

analyzed using variography and kriging techniques. An

rsotropic semivariogram was constructed to determine the

spatial structure and quantily spatial atfibutes such as

nugget, sill and eflective range. These attributes were

used to perform point kriging. Variography and krrging

were computed usLng GS+ Version 5.1.1 (Gamma Software

Design, P1arnwel1, MI). Measured and kriged values were

mapped using Sr.rfer Version 7 O (Golden Software Co.,

Golden Co).

Kriged values were cross-validated based on the

criteria proposed by Delhomme (1978) and Dowd (1984).

Firstly, the interpolated Mean Error (ME) should be close

to zero. The ME is calculated as follows:

vE: r /n! t-(x,)-z(x,) l (2)

Where:The number of sample points

2(x,) = The predicted value of the variable at point 4

z (+) = The measr-red value of the variable at point x,

Secondly, the Mean Squared Error (MSE) should be

less than the sample variance. The MSE is given by:

MSE = l /n l t i t x , t -z tx , )1 '1 (3 )

Thirdly, the ratio of theoretical and calculated

variance, called the Standardized Mean Squared Error(SMSE), should be approxrmately close to one. The SMSE

is given by:

SMSE = t ln-t i t?x,)- z(x) l ' to ' (4)

Where:

o' = The theoretrcal variance

83

J. Agron., 7 ( l) : 82-87, 2008

RESLILTS AND DISCUSSION

flnivnriate statistics: The majori[, of chemical properties,

with the exception of P and Zq exhibited non-normaldistributions as detemined by the Shapiro-Wi1k statrstrc(Table 1). Among these properties, C and K weresignificantly skewed (2xSES = 0.64) while B wassrgnficantly kurtotrc (2x SEK : 1.26). The coefficients ofskervness and kr.rtosis describe the shape of the sample

dislrrbution A positive skew indicates asymmeby in thedistribution rn'ith the higher values tailing to the right anda negatrve skew represents lower values tailing left.Kurtosis describes the relative size of the dislribution'stails. A positive hrrtosis indicates that the distributron rspeaked and a negative kurtosis indicates a relatively flatdistnbution. Both these coefficients depict the conformityof the data to a normal dishrbution. Transformation uurg1og,o was only performed on Cr:, which initially showed asigmfrcant positive skew. After being 1og-trarsformed, Cuconfomred to a normal distribution. The other non-normally distributed properties did not respond favorablyto translbrmatron. i.e., thev remained non-normal aftertransfonnation.

The Coefficient of Variatron (CV), whrch is the mtro ofstandard deviatron to the mean, for the properties

measured ranged lrom 129'o for C to 54o/o for B (Table 1),ndicating low to moderate variability in the data.

The skerved distribution of C is possibly due to ther.rneven bumrng of plant biomass prior to replantrng whiletlut of K can be attributed to leaching as a result of microtopograpllcal variation. The relief within the study plotwas <1 m.

Spatial structure and attributes: Semivariograms ofthe chemical properties are given in Fig. 1. Thesemivariograms were computed based on an active lag of55 m and a lag class interval of 6 nr.

A11 properties exhibited a definable spatial structtre.C--arbon P and Cu were described by a spherical model,while K, Zn and B subscribed to an exponential modei.The properries that exhibited strong spatial dependencewere C, P and B with a nugget to sill ratio of 0.04,

'7.41 utd

213%, respectively This physically mears that theexplaurable proporlron of the total variation in C, P and Bis 99.96, 92.59 and 97.8t"o,i, respectively, while theremaining variation is attrrbutable to random sources. Theother propertres showed a moderate spatial dependence.

The ma.lority of properties had a short EffectiveRange (ER) i.e., <70 m. Both K and Zn, however, hadrnoderate ER values The practical significance of ER isthat sampling points separated at drstances greater thantheERwillno longer exiribit spatral correlation. At this

Table 1: Univariate statistics for chmical propedies

CVVriables nr Mem Median (90) SkewnmJ Kurtosis2 Normal

C P/o) 60 43.00 45.50 11.77 -0.970 0.069 0.879**P (pg g-') 58 33.15 33.20 27.34 0.629 1.026 0.945*K (pgg_') 58 159.91 139,05 46.05 0.735 -0.079 0.937*Cu(pgg- ' ) 58 1.36 1,00'71.42 0.918 -0.362 0.96f fZn(VSt ' ) 60 35.36 34.30 35.37 0.358 -0.547 0,97d"B (Lrgq ' ) 57 1.08 1.00 54.21 -0.107 -1.450 0.915*r: Couts < than 60 indicate that non-spatial outlims wre removed frm the

data sel Non-spatial outliqs wme detected using the exbeme sudfltized

deviate (ESD) method,2: Significmt if the absolute value of skewness

or kurtosis is >2 times its stmdrd enor. The standard error of

skemess=(6/n)05 while the stmdad mor of U*or1, = 124ln)0ti3:Estimated using the Shapiro-Wilk test. lf the test statistic W is significant(p<0.05) thm the distribution is not normal. @or Cll, the W was computedaJis dda trmformation using logls); {: Significmt at the 0.05 probability

level; **: Signi{icant at the 0.01 probability level; ns: nm-significant

Table ?: Cross validation statistics ofkiged values for chetnical properties

Smple

Variables vriance ME l\{SE SMSE

c (9/") 10.82 0.001P (pg g- ' ) 8210 -0.290

K (pC S-') 542r.e'7Cu (pgg-') 0.95Zn (pg3- ' ) 156.35

-0.660 4878.000 .0100.080

9.767 l . l l

0.'l '7159.9.1

0.920.880.920.831.0.1

B (us s rJ 0. I 8 0.004

jr.mcture, it is worth noting that the semrvanogram does

not provide any informatron for distances shorler than the

minimum spacing between samples Sampling designsaimed at delineating spatial structures usu€r11y employseparation distances that are lesser than the ER. Flatmanand Yfantls (1984) recommended that sampies be spacedbetween 0.25 and 0.5 of the ER. Based on this, it is clearthat sample spacurg for C, P, Cu and B should be closer

than that of K and Zn.

Spatial variability: The distribution and pattem ofbothmeasu'ed and kriged values for each chemical propertyare represented as surface maps (Fig. 2).

Generally, all properties exhibrted spatial clustering oftest values across the study plot. The surface mapssuggest a linear association between Cu and C, Cu and K

and Cu and Zn. Based on measured values, significantcorrelations at the 59lo probability level were registeredfor C\r and C (r: 0.49), Cu and K (r = -0.35) and Cu and Zn(r : 0.40). The interaction of Cu and C is expected becauseCu rs typically sorped onto C-rich organic surfaces(Fotlr 1991).

Cross-validatron statistics showed that all properties,with the exception of K, satisfied the interpolahonaccuracy criteria (Tab1e 2). From a managementperspective, accurately kiged values can provide acheap and fast means of estimating a partrcularvariable (in this case, chemical property), withouthavlrg to perfonn field sampling and laboratory analyses.

84

€B

H

o

Ee. F

. :a

o o.El

i 0.54U

a0.n

8 4896

.t

.z 326/

6

- 0.32

5 0.21E

30 . l t

(i) TotalC (ydModcl: Sphatcal; Spatlal depodrm: StrorgNuggct:0,01; SiI = 24.9; Effoctivc rugp:43.5 m

0.00 t3.75 27.50 4t.25

Sepration distanoc (n)

(iii) FxEactsble K (pg g-)

ff Extnrtablo P (pg g-)Modcl Slhcricel; Speirel dcpcodaoccr StongNugg€t = 6.5; Sill = E7.t; Efrectiw ruge = 26.9 m

55.00

Model: E:<pomtiod; Spatial depodme: ModwteNug8sr - 3750; SiX = 10565; Eftadve rugc : 109.6 n

13.15 27.50 41.25 55.00

Scpration distmc€ (m)

(v) htractablc zn (pgg-)Modcl hponmtial; Spstiat dapmdoce: ly{o&ntaNugBcr = 123; Sitr = 245; Eftcriv€ rmgp = 167.2 m

0.00 13.75 ?'1.50 41.25 55

S@6rsdon dtrtsDDc (n)

Infcrpatcd bqs€d @ Cubrdclh :r a/, (194), r*erc:

0.00 13.75 27.50 4r.25 55.00

Sepmtlon dlatene (m)

(ig Exnacbblc Cu Ge g-)lrtodcl: Spboicd; Sprtial depcodaacc: ModratcNuggot:0.5It Sill: l.0t; Effnctivo mge = 562 m

0.000.00 13.75 27.50 41.25 55.00

Scpffidon dilt8ilcc (m)

(vi) Extactable B (pg g-)Modcl Expoacuthl; Spatial dcpadrosc: SbongNugger = 0.01; Silf = 0.47; Effeqtire mge = 23.9 D

t3.75 27.50 4125

Sc?rr.don distbcc (m)

spatially variable. A previous investigation showed thatpineapple yields were also spatitrlly variable across thesame study site (Baiasundran et al., 2005). As such, rtappears tllat unifonn agronomic management based onconventional strategies may not be suitable to optimrzecrop productron. Instead, a site-specific approach such as

1632

0+-0.00

NugEef Si[< 0250.25<tuSect Sill< 0.75Nu€gs* Silk 0.75

StoDg spltial depcndrDccModcnto rpatial de,pundonmlbak epatial dsp€nd€nco

Frg. 1: Spatral structure 'and attnbutes of chemical properttes

Horvever, tlis is only practical for variables that aretemporally stable, i.e., test values that do not changesigrlficantly over time.

General discussion: This study demonstrated thatselected chemical propefties of tropical peat were

85

J.Ag*r., 7 (l): 32-37, 2(fr8

*4*--rai 1 3

l llb

e3 \t* vr.

%: _,.n

$*{

te.

, I b

X,g;.- **q."lt ,1 , 1bE

i . f

Fig. 2: Spatua1 r'"ari abilrty of dremicel pcperties (based on meastued er:d laiged values)

Crop l"{ar:agemerd. Zmr:ng (CM4 may offer a bettetor.rtcr:me. CI\M refers to qr.n.trtitahve trsufrnC of areaswith zunilar poduction potentralflrnritdicn in crder tofacrlitate optimlrn ctop anrl soil m magerL ent as a fi-urcli onof local l'ariahility. Crop managan ent zotreq if desigredaccwately ard efficierdly, crr potenfally lead toincreased crap prodr.utivriy and quality (Ealasr:ndram,?003)

CONCLUSION

A lrrgir degree of spa[al vuiabilitSr was obserued insele cte d chemic al poperh es frc'rn a I ha cr-r1ti',:ate d kop c atped. The mqority of dremical 1:rqpaties exhibitedncn-nc'nn el S stributi crrs wiih CV s r ea'rging from 12 to J4%. Allcfu mic al p opedre s er Ltbite d a definable spati al strtrcD_re,

wlidr were des,;ribed by eitha qpherical cr expcnenlialmodels Proprties such as C, P a$d E showed stcngqpatial dqpaidenre. The maj wity of poptrties had a shsteffective ra4gq which irulicates the needto emplo5r closersampl e qp acing in crder to ac c or-mt fcr sp aii al correl a[ cn.Surface mrys cf ttre chemicd popertieq which cc,nprisedmeaswed and luiged volueq clearly shnwed qpdialclustenngof testvshns. Wrththe excepticn of K, all otherpropetties showed acceptable accwary of intapolatedvalues, whid: cen be used to estimate elemenialconcerdration in a cosL dfec[l-e andtim+ saving martrrer.These comtlned data zuggeslttre ruedfct a site-specificapproach in mura6jng tropical peat cultivated wrthpine a1:ple, p uti cr:l ady with re gar d to rr,lklert.mE.rrEgemerjt.

E$ r

t iu a

-Sr '* ' t '

86

J. Agron., 7 (1): 82-87,2008

ACKNOWLEDGMENTS

The authors are grateful to Peninsular Planlations

Sdn Bhd for logistical supporl and technical collaboratron.Appreciation is also extended to N4r, Juraidr Jaafar for

assisting rnith field work and Ms. Sarimah Hashim forhelping with laboratory analyses.

R-EF.ER-ENCES

Balasr.ndram, S.K , 2003 Stlategies for precision oilpalm management in South Sumatera, Indonesta.Ph D Thesis, Unrversity of Minnesota, St. Paul,

Murnesota, USA., pp: 203.Balasundram, S K., M.H.A Husm and O.H. Ahmed, 2005,

Precrsion Pineapple Management on Tropical Peat I.Spatral Valiability of Yield. In: Proc. Soils 2005

Advances in Soil Science lor Sustarrable FoodProduction, The, C.B.S et al. (Eds.). Sr:ngai Petaru,

Kedah Malaysian Society of Soil Science (MSSS),pp: 90-93.

Burgess, T.M. and R. Webster, 1980. Optrmalinterpolatron and isarithmic mapping of soripropefi.res. L The semivariogram and punctualk lg ing . L So i l Sc i . , 31 : 315-331 ,

Cambardella, C.A., T.B. Moormarl J.M. Novak,T B Parkin, D L Karleq R.F. Turco andA.E. Konopka, 1994. Field-scale variability of soilpropertres m cental Iowa soils. Soil Sci. Soc. Am. J.,5 8 l 5 O 1 - 1 5 1 1

Cooke, R.A., S. Mostaghimi and JB Campbell, 1993.Assessment of methods for interpolating steady-state infiltrability. Trars ASAE., 36. 1333 -13 41.

Delhomme, J.P., 1978. Kriging r.n the hydrosciences.

Adv. Res., 1.251-266Dowd, P A, 1984. Cited i t .Paz, A., M.T. Taboada and

M.J. Gomez, 1996. Spatial variability in topsoil

micronutrient contents in a one-hectare croplandplot. Comrnr:n. Soil Sci. PlantAnal.,21 (3-4):4'79-503.

Flatman, G.T. and A.A. Yfantis, 1984. Geostatistical

strategy for soil sampling: The survey and the

cersus. Environ. Monit. Assess., 4: 335-349.

Foth, H.D., 1991. Fundamentals of Soi l Science.8thEdnJohn Wiley and Sors, Inc., New York, pp: 360.

Isaaks, E.H. and R.M. Srivastava, 1989 An Introductionto Applied Geostatistics Oxford University Press,

New York, pp: 561 .Oliver, M.A., i 987. Geostatistics and its applicatron to soil

science. Soil Use Manage., '7:

206-211 .

Rober! P.C., 1999 Precision Agriculture. Research Needs

and Status in the USA. In: Society of Chemical

Industry, Stafford, J.V. (Ed.). Proceeding Preciston

Agriculture, Vol 1. (SCI), Londorl Unrted Kurgdom,pp: 19-33.

Trangmar, B.B., R.S. Yost and G. Uehara, 1985.Application of geostatistics to spatial studies of soil

properties. Adv. Agron., 38. 45-94.Yoltz, M. and R. Webster, 1990. A comparrson of

kigrng, cubic splines and classificatron for predrctrngsoil properties from sample information. J. Soil Sci.,

4l:4'13-490.Whelan, B.M., A.B. McBrabrey and R.A.V. Rossel, 1996.

Spatial Prediction for Precision Agriculture. In:Proceeding 3rd Intemational Precision AgricultureConference, Minneapolis, Robert, M.N. e/ a/. (Eds.).

ASA-CSSA-SSSA, Madison, WI, pp, 331-342.

87