International Journal of Agricultural Science and Research (IJASR) ISSN 2250-0057 Vol.

3, Issue 1, Mar 2013, 129-142 TJPRC Pvt. Ltd.

DELINEATION OF COMPACT ZONES THROUGH SPATIAL VARIABILITY ANALYSIS OF SOIL PHYSICAL PROPERTIES
PRAGATI PRAMANIK & PRAMILA AGGARWAL Indian Agricultural Research Institute, New Delhi, India

ABSTRACT
An experiment was conducted to carry out two dimensional spatial variability analysis of penetration resistance (PR), soil water content (SWC) and bulk density (BD) of two texturally different soils - sandy loam soil and loam soil of IARI farm. Soil samples at 54 locations at a grid interval of 33.3 m x 50 m, covering a total area of 12 hectare of IARI main farm were collected from surface 0-15 cm and sub surface 15-30 cm soil layers .Observations on PR (at field capacity soil water content) , BD and SWC were recorded twice to obtain two states of soil compaction -first time was before ploughing and second time after ploughing of field for sowing of rabi/winter crop. Average PR and BD values of 15-30 cm layer were reduced after ploughing still they remained above their critical limits which confirmed the presence of permanent plow pan between 15-30 cm. Before plowing major and minor range of PR, SWC and BD varied between 160178 m. After ploughing, the major range of these parameters reduced by 30 cm. Reduction in minor range was more than 100 m .The results suggested that plowing increased short range variation and hence the range of spatial dependence reduced. Nugget ratio for PR was low which showed strong spatial dependence of parameter within its range whereas both SWC and BD had low to medium nugget ratio indicating strong to medium spatial dependence of these parameters. The prediction maps of PR and BD both before and after ploughing also indicated the presence of plow pan (PR >2000kPa and BD >1.6Mgm-3 for sandy loam and 1.56Mgm-3 for loam) in the sub surface. Deep ploughing should be recommended for the area where plough pan existed even after plowing.

KEYWORDS: Geostatistics, Penetration Resistance, Spatial Variability, Krigging INTRODUCTION

As a consequence of progressive increase in mechanization of farm operations, the problem of subsurface soil compaction is very severe in alluvial soils of Indo Gangetic plains. Soil compaction is described as decreasing porosity or increasing dry bulk density (BD) as a result of firm-pack soil particles (McKyes, 1985). Highly compacted subsoil layers (hardpans) restrict root penetration growth and plant growth, and limits water and air transport, decrease soil moisture availability that results in a decline in crop productivity, particularly during drought periods. Compaction of soils causes nutrient stress and retards seed germination and may accelerate erosion and runoff. Soil compaction is mostly affected by soil water content, BD, soil texture and soil porosity (Okursoy, 1992). The effect of each factor changes mainly with the soil texture. In a sandy loamy soil, an increase in dry BD and a decrease in volumetric moisture content cause high penetration resistance (PR) values. Meanwhile, in clayey soil, similar PR values depend on mainly dry BD (Alexandrou and Earl, 1998). Farmers apply uniform subsoiling either on an annual or a biannual basis to mechanically disrupt the hardpan layers such that roots can penetrate into deeper soil layers to access the soil moisture and nutrient reserves for optimal crop growth. With the current subsoiling practice, the tillage implements are set at uniform depth for the entire field. Uniform tillage practices may cause unnecessary fuel consumption as the hardpans exhibit spatial variability. Application of site -

130

Pragati Pramanik & Pramila Aggarwal

specific tillage management that addresses spatial variability of soil compaction may reduce tillage energy and fuel costs and also create the desired soil conditions for crop growth. Soil variability in the field is generally defined with classic statistical methods and is assumed to have a random variability. Soil variability occurs as a result of effect and interaction of various ongoing processes in soil profile (Parkin, 1993). Soil characteristics generally show spatial dependence (Webster, 1985). Samples close to each other have similar properties than that of samples further from each other. The classical statistic is not sufficient to analyze the spatially dependent variables and assumes that measured data are independent (Vieira et al., 1983). Several research studies have been published on spatial variability of penetration resistance (Utset and Cid, 2001; Kilic et al., 2004). The main purpose of researchers is to map the quantitative and qualitative aspects of soil compaction (Silva et al., 1989; Bach et al., 2000; Couto and Mello, 2000; Souza et al., 2001). Geostatistics is a branch of applied statistics that quantifies the spatial dependence and spatial structure of a measured property and, in turn, uses the spatial structure to predict values of the property at unsampled locations. Geo statistics, which together with classical statistics constitutes an extra ordinary important tool of precision farming. The knowledge of spatial variability of soil compaction parameters can be used to identify the zones with soil compaction problems and suggest the appropriate soil management options accordingly. Now a days due to the availability of advanced and more sophisticated geostatistical softwares (ArcGIS with geostatistical module, GS+ etc), soil data taken at few locations with defined spatial coordinates can be put to GIS environment and details of spatial structure of the parameters can be studied easily, which can be used to generate prediction maps of parameters even with low sampling intensity (Bakhsh et al., 2000; Sarangi et al., 2005). Spatial variability analysis of soil compaction and application of site specific tillage management has not progressed as the site specific application of fertilizers and chemicals due to lack of appropriate technology or procedures to characterize soil compaction. Effective implementation of Variable Rate Technology (VRT) depends on an accurate characterization of the spatial variability of the factors used to determine desired application rates. Research is needed to develop methods to precisely characterize and quantify the relative strength and depth of hardpans and determine their spatial pattern for precision (site-specific) tillage management to improve sustainability of crop production. The objective of the experiment was to carry out two-dimensional spatial variability analysis of soil penetration resistance at field capacity soil water content, bulk density, soil water content, clay and sand contents of representative sandy loam and loam soils of IARI experimental farm and preparation of kriged map for delineating compact zones.

MATERIALS AND METHODS

The study area is situated at 28.38 N, 77.20 E and is at an altitude of 228.7 m above the mean sea level in the IARI Research Farm, New Delhi, India. The climate of the area is semiarid and sub tropical with hot summer and cool winters. The mean monthly maximum and minimum temperature during the year ranges from 21.3 C to 40.5 C and 7.3 C to 28.7 C, respectively. The annual rainfall is 708.6 mm of which on an average 597 mm (84 %) is received during the month from June to September. Land use was a maize wheat rotation. Data on soil penetration resistance (PR) and other soil physical parameters such as bulk density (BD), soil water content (SWC), clay and sand content of two texturally different soils of sandy loam (Main Block VIII; size- 300 m x 200 m; soil type- Sandy loam; class- Typic Ustocrepts (Soil Survey Staff,1994); - referred to as Site I) and loam soils (Main Block XI; size- 300m x 200m; soil type- loam; classTypic Ustocrepts; - referred to as Site II) were collected from 0-15 cm and 15-30 cm soil depth. Each block was divided into three strips- A, B, C each of 100 m x 200 m. From each strip 3 points were selected at an interval of 33.3 m. Soil samples were collected from three locations in each strip, 50 m apart. In all, 27 bench marked

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

131

sampling points were selected in a block on a rectangular grid at 33.3 m x 50 m interval. Samples were collected from 1 m x 1 m plot. Before collection of samples, the area thus marked was bunded for impounding of water and allowed to drain for 48 hours to reach to field capacity level. The sampling was done twice. The first sampling was done in the month of October, 2005 before plowing for winter sowing and the second sampling in the month of November, 2005, immediately after plowing and sowing of winter crop. This enables creating two states of compaction. During rainy season maize was grown in the field by ridge and furrow methods and during winter season wheat was grown by conventional method. In loamy field, where wheat was sown under no tillage condition, soil physical parameters after plowing could not be collected. Data on Soil penetration resistance (PR), Soil bulk density (BD), and Soil water content (SWC) and Soil texture was obtained by following procedures. Soil penetration resistance for each sampling point up to 30 cm soil depth was measured by using a Rimik cone penetrometer (model no.CP20) (automatic penetrometer with data logger) at field capacity (2nd day after watering a bunded plot of 1m x 1m size). The maximum value of PR, which this instrument could measure, was 5000 kPa; maximum depth of penetration was 600 mm, which was measured by an ultrasonic transducer. The diameter of the base of cone was 12 mm, slant height of cone was 24 mm and angle of cone was 30. The capacity of load cell was 60 kg. Penetrometer had a data logger system attached to it for automatic recording of data in the field. Retrieval of data on PC using CP20V2 software gave the recorded PR data at an interval 15 or 20 mm soil depth in graphical as well as tabular form (detail can be obtained from hardware manual of Rimik cone penetrometer CP20). Bulk densities (BD) of 0-15 cm and 15-30 cm soil layer both before and after plowing were measured at each bench marked point using a core sampler (Blake and Hartge, 1986). Soil water content (SWC) was measured simultaneously along with penetration resistance using gravimetric method for both depths before and after plowing. Sand, silt and clay percentages of 0-15 cm soil layer at each site of the field were determined by Bouyoucos hydrometer method (Bouyoucos, 1962) from same BD samples and soil textural classes were determined by USDA textural triangle. ArcGIS 8.3 was used for carrying out the geostatistical analysis. Geostatistical analyst module of ARC GIS 8.3 had been used for carrying out spatial variation studies of all observed soil physical parameters namely PR, SWC and BD of all 54 data points (27 points in Sandy loam field and 27 in Loam field) located on a grid at an interval of 33.3 m x 50 m interval covering a total of 12 hectare farm area. Geo statistical analyst module of this software was used to construct semivariograms and spatial structure analysis for variables. The X, Y coordinates representing the location of sampling point along with Z coordinate represented by magnitude of BD, PR and SWC values for all 54 sites of each soil layer, were entered in Excel spread sheet format and saved as csv. extension. The spreadsheet format was then imported to the table of content (TOC) in ArcGIS and the measured point locations of the rectangular plots were generated as point feature classes. These point arrays representing the data values were further subjected to exploratory data analysis and kriging to predict maps of BD, PR, SWC, sand and clay over the entire area. Exploratory Spatial Data Analysis (ESDA) allows us to examination of the data in different ways such as exploring the distribution of the data, looking for global and local outliers, looking for global trends, examining spatial autocorrelation, and understanding the co-variation. The ESDA tools are Histogram, Voronoi Map, Normal QQ Plot, Trend Analysis, Semivariogram/ Covariance Cloud, General QQ Plot, and Cross covariance Cloud. Exploratory Spatial Data Analysis Geostatistical Analyst requires that the data is normally distributed and there are no outliers present. When the data are skewed (the distribution is lopsided), histogram tool is used for transformation of data to make it normally distributed. The transformations should be done using linear, box-Cox or logarithmic techniques. This tool allows

132

Pragati Pramanik & Pramila Aggarwal

exploring the effects of different transformations on the distribution of the dataset. If the data are transformed before creating a surface using geostatistics, the predictions will be transformed back to the original scale for the interpolated surface. In addition to this, the histogram tool can detect global outliers, which if present should be removed before proceeding for spatial structure studies. Voroni mapping of data should be carried out for examining the presence of local outliers .But local outliers, if present, need not to be rejected if they represent abnormal variation in parameter value rather than the measurement error. The Trend analysis tool can help in identifying global trends in the input dataset. Trend if present should be removed from the data set before using kriging. The semivariogram/covariance cloud can be used to examine the local characteristics of spatial autocorrelation within a dataset and look for outliers. The semivariogram/covariance cloud shows the empirical semivariogram (half of the difference squared) and covariance for all pairs of locations within a dataset and plots them as a function of the distance between the two locations. (h) =1/2 N(h) [z(xi)-z(xi+h)]2 (1)

Where z (xi) and z (xi + h) are measured sample values at xi and xi+h points. N is the number of pairs separated with distances h (lag space). Each red dot shown in the figures Figure 1 of semivariance cloud actually is an empirical semivariogram between a pair of locations in the dataset. The values in the semivariogram cloud are put into bins based on the direction and distance between a pair of locations. These bin values are then averaged and smoothed to produce a surface of the semivariogram. It is possible to explore the spatial autocorrelation that is present in one's data by examining the different pairs of sample locations. If data are spatially dependent, pairs of points that are close together (on the far left of the x-axis) should have less difference (be low on the y-axis). As points move farther away from each other (moving right on the xaxis), in general, the difference squared should be greater (moving up on the y-axis). Often there is a certain distance beyond which the squared difference levels out. Pairs of locations beyond this distance are considered to be uncorrelated. The directional influence seen in semivariograms also called as anisotropy can be explored and if detected should be taken into account while developing semivariance models. Spatial Structure Analysis and Kriging Using Geostatistical Analysis Wizard The first step of this analysis is to choose the appropriate geostatistical method of interpolation which could be ordinary kriging, simple kriging, universal kriging or cokriging .The decision for a particular interpolation technique can be taken from the observations made in ESDA. If no trend in data is observed oridinary kriging can be chosen. If trend is present universal kriging could be used. If one wants to prepare the surface of the parameter whose value exceeds a particular threshold value then indicator kriging could be used. Since all methods of kriging require that data is normally distributed, hence the data if not normal should be made normal by using some transformations like log or Box-Cox and trend if present; its order should be entered before proceeding for further analysis. Again using the informations obtained from ESDA, different available anisotropic or isotropic semivariance models can be fitted to empirical semivariance cloud data (scatter plot). Three most commonly used semivariogram models are circular, spherical and exponential and has been described below: The spherical model is given by: (h,)= s[1.5h/ r-0.5 (h3/ r3) = s Where s is sill value, h is the lag and r is the range for 0 h r for rh (2) (3)

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

133

The exponential model is given by: (h,) = s[1-exp(-3h/ r)] The circular model is= (h,)=2 s/[h/ r(1-h2/ r2)0.5+arcsin(h/ r)] = s for 0 h r for rh (5) (6) for all h. (4)

The above mentioned models were tried with different combinations of lag size (h in meters) and n (no of lags). The h, n combinations were chosen in such a way that product of h and n were less than the distance between the extreme sampling points. Neighborhood shape is kept spherical if the empirical semivariogram of data shows isotropy and elliptical if it shows anisotropy. Again in order to avoid bias in a particular direction the ellipse was divided in four sectors and a minimum of 2 or maximum of 5 sampled data points were selected in each sector. All the models were cross validated by plotting the predicted values against observed value and fitting a line through the scatter plot. The closer is this line to 1: 1 line, the better is the fit. For all layers, the best fitted model along with appropriate h and n combination was the one which gave lowest RMSE and also its average standard error (AVSE) nearest to root mean square prediction error or alternatively root mean square standardized error (RMSS) nearest to one. Finally, the values at the unvisited location were predicted by multiplying the weights of the sampled data points by their values and then adding them together. In this way the prediction map of a parameter was prepared. For presenting the prediction map as filled contours, the maximum, minimum and contour interval should be specified. By inspecting the prediction map, one can delineate the area having the value of PR and BD parameters greater than their critical limits.

RESULTS AND DISCUSSIONS

As explained earlier in material and methods section, ESDA was carried out to examine the normality, spatial correlation, trend and outliers in data using various ESDA tools provided by geostatistical analyst module of ArcGIS 8.3. The results of ESDA have been summarized in Table 1. ESDA thus revealed that data of all parameters i.e. PR, BD and SWC were nearly normal with little skewness. There were no global outlier present, no trend present in data. Data of all parameters exhibited spatial correlation and their empirical semivariogram showed anisotropy. Spatial Structure Analysis and Interpolation By making use of the information obtained in the ESDA that the data sets of all above parameters were nearly normal and no trend was found, oridinary kriging was chosen as the appropriate model for interpolation. Data on PR, BD and SWC data for 0-15 and 15-30 cm soil layers before and after ploughing were used to prepare empirical semivariograms. Since data of all parameters showed anisotropy, among the various available anisotropic semivariance models, three most commonly used semivariogram models i.e. Circular, Spherical and Exponential were chosen and were tried with different combinations of lag size (h in meters) and n (no of lags). The h, n combinations were chosen in such a way that product of h and n were less than the distance between the extreme sampling points. Selection of sampling distance intervals is important in ensuring the quality of spatial variability analysis and interpolation of points for unsampled locations using geostatistical techniques (Donald and Ole, 2003). A sampling interval distance less than a range, a distance over which pairs of observations exhibit spatial dependence, was considered appropriate in grid sampling. In case of sampling before plouging, the maximum distance between extreme data points with coordinates 16.5,50 and 283.3,360 was nearly 400 m; hence h x n combinations tried were 10 x 18, 18 x 10, 12 x 15 and 15 x 12. In case of sampling after plouging the maximum distance between extreme data points with coordinates 16.5,50 and 283.3, 150 was nearly 300 m

134

Pragati Pramanik & Pramila Aggarwal

hence combinations tried were l0 x l5, 15 x l0 and 12 x 12. Since data set of most of the parameters showed anisotropy, the shape of neighborhood was chosen to be elliptical. Figure 2 showed semivariance model of PR of 0-15 cm soil layer before ploughing. Model was cross validated to compare the results of measured and predicted values. Since the blue line (fitted line through the predicted points) was nearer to 1: 1 line the chosen model could be rated as fairly good model. For all layers, the best fitted model along with appropriate h and n combination was the one which gave lowest root mean square error (RMSE) and also its average standard error (AVSE) nearest to root mean square prediction error or alternatively root mean square standardized error (RMSS) nearest to one. The best-fit models and model parameters, such as nugget variance, sill variance, and the range of influence, were presented in Table. The table 2 showed that before ploughing, the best suited semivariance model for PR, BD and SWC were spherical with 18 x I 0 combination, exponential with 12 x 15 combination and circular with 10 x 18 combination, respectively, for surface layer. Similarly for 15-30 cm soil layer the best suited models for PR, BD and SWC were exponential with l0 x 18 combinations, circular with l0 x 18 combinations and exponential with l5 x 12 combinations, respectively. After ploughing (Table 3), the best suited semivariance model for PR, BD and SWC were circular with 12 x 12 combination, exponential with 15 x l0 combination and exponential with 15 x 10 combination, respectively, for surface layer. Similarly for 15-30 cm soil layer the best suited models for PR, BD and SWC were spherical with 10 x 15 combination, circular with 15 x 10 combination and circular with l0 x 15 combination, respectively. The spatial structure of BD both before and after ploughing for best suited models is given in table 4. Results revealed that spatial structure of BD for both layers before ploughing was nearly isotropic with spatial range of 177m , nugget variance nearly 0- 0.002 and partial sill value of 0.003- 0.0047Mgm-3. But after ploughing the semivariogram exhibited anisotropy and although major range reduced slightly (from 177m before ploughing to 148 m after ploughing), the reduction in minor range was appreciable (from 177m before ploughing to 36m after ploughing). Smaller range value indicates that after ploughing reduced the distance over which pairs of BD values remain spatially dependant. The nugget variances for before and after ploughing were very low, which indicated strong spatial dependence in soil BD values. The results thus suggested that ploughing increased shorter-range variations and hence the range of spatial dependence reduced. The % nugget ratio (nugget variance x100/total sill variance value) was low (0-30%) before ploughing and medium (30-70 %) after ploughing; it could be concluded that BD parameter of both layers showed strong spatial dependence before ploughing, but after ploughing it became moderate. The low nugget variance/total variance ratios and small range values for these properties suggested that the distributions were patchy (Kilic et al., 2004). Spatially dependent variables may be controlled by intrinsic variations in soil characteristics such as texture and mineralogy (Rao and Wagenet, 1985; Cambardella and Karlen, 1999), and extrinsic variations such as tillage (Cambardella and Karlen, 1999). Spatial structure of semivariograms of PR and SWC before ploughing (Table 5 and 6) also showed their major range (l76-178 m) similar to that for BD but their minor ranges were relatively less as unlike BD, they exhibited anisotropy even before ploughing. After ploughing the range of minor axis reduced to below 60 m for both these parameters, which again showed the increased short range variations due to ploughing. In general for PR, the nugget ratio was low which showed strong spatial dependence of the parameter for both depths before and after ploughing. Similarly, for SWC nugget ratio was medium to low, which showed medium to strong spatial dependence of the SWC parameter for both depths. Prediction maps were prepared by using the best fitted semivariance model for all observed soil physical parameters. The map for BD was drawn as filled contours with 6 equal classes (Mgm-3) 1.2-1.3, 13-1.4, 1.4-1.5, 1.5-1.6, 1.6-1.7 and 1.7- 1.8. The prediction maps of BD for surface layer before ploughing showed the presence of 1.2-1.3, 13-1.4, 1.4-1.5 contour classes (Figure 3a) whereas the prediction maps of BD for sub surface 15-30 cm layer showed the 1.5-1.6,

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

135

1.6-1.7 and 1.7-1.8 contour classes. The map thus clearly showed the presence of plow pan area with BD > 1.6 Mg/m3 in subsurface in the entire area of study. Similarly a bulk density level of 1.60-1.65 Mg m-3 for sandy loam and > 1.50 Mg m-3 for loam can be considered as the critical bulk density beyond which crop yield declines drastically. The BD map drawn for after ploughing (Figure 3b) situation for sandy loam field also revealed the presence of plough pan in the subsurface but its magnitude and area was reduced substantially by ploughing. Kriged map of bulk density showed that the predicted soil hardpan depth seems to vary across the field. These results were similar to the ones reported by prediction maps of BD (Utset and Cid, 2001; Kilic et al., 2004; Fluton et al., 1996). The prediction map of PR (measured at field capacity water content) was prepared for 9 contour classes (kPa) namely 1200-1400, 1400-1600,1600-1800,1800-2000,2000-2200,2200-2400,2400- 2600,2600-2800 and 2800-3000 (Figure 4a ) . The areas with PR classes having value >2000kPa were treated as compacted areas. A penetrometer measurement of 2.0 MPa is generally regarded as sufficient to hinder the growth and developments of crops (Taylor et al., 1966; Kay and Anger, 2002). Results revealed that before ploughing the 0-15 cm soil layer showed contours with values <2000 kPa but subsurface had contours with values ranging between 2400 to 3000 kPa. After ploughing the range of PR was shifted to 1800-2600 kPa (Figure 4b) but still the major portion of the sandy loam field had PR >2000 kPa, which again indicated the presence of permanent plow pan in some portion of the subsurface. The maps for PR of the field indicate that the values exceeded the critical root limiting cone index value of 2 MPa (Taylor and Gardner, 1963) in most parts of the field with the values being higher before ploughing. The Prediction map of SWC (field capacity water content) was prepared for 6 contour classes (%) namely 1213.5, 13.5-15,15-16.5,16.5-18,18-19.5 and 19.5-21 (Figure 5a) .The prediction maps of SWC (at field capacity) showed that for surface layer the SWC ranged between 16.5-19.5 % whereas the range was 13.5-18% for 15-30 cm soil layer. However, the range of surface layer was reduced (13.5-18%) after ploughing (Figure 5b). SWC2MPa is an indicator of critical level of soil water content at which soil mechanical impedance become sufficiently high to reduce the root growth by more than 50% in most of the crops (Aggarwal et al., 2006). The results presented here indicated that soil compaction depended on extrinsic variations in soil characteristics such as intensive soil tillage and intrinsic variations such as BD, clay content, sand content and water content. Nevertheless, these results could be related to the overflowing size and deposition regularity during soil formation (Kilic et al., 2004). Generally the spatial distribution pattern of the soil compaction within the field seems similar as predicted by BD or PR values. It is well known that there is a close agreement between soil BD and PR. As seen in Fig. 5a and 5b and Fig. 6a and 6b, the mean PR was higher in 15-30 cm depth before ploughing, and so was the BD. However, while PR slightly decreased after ploughing in both depths, BD went down similarly. Kriged maps of PR values indicate that most part of the field requires deep tillage. The depths of tillage need to vary according to the predicted soil compaction depths. This indicates that applications of depth-specific tillage IARI farm may improve the sustainability of crop management.

CONCLUSIONS
In brief, it can be concluded that geostatistical analyst of Arc GIS 8.3 is a very useful tool for carrying out geostatistical /spatial variability analysis of soil properties. Such analysis is required for developing appropriate semivariograms of the parameters which are used for preparing the surface maps of the parameters using kriging as interpolation technique. These maps are then used for deciding the site specific inputs requirement which is the main objective of precision farming. For site specific tillage recommendation, BD or PR (at field capacity water content) measured at few locations could be used for generating their prediction maps through kriging which could be used for delineating compaction zones so that necessary soil tillage techniques could be recommended in the compacted areas for a

136

Pragati Pramanik & Pramila Aggarwal

suitable seedbed, plant grow and producing greater yields and to economize the use of inputs. Continuous moldboard plowing at the same depth resulted in the formation of a plow pan below the tillage depth. This plow pan can be alleviated by varying the depth of tillage over time or by special tillage operations. Tillage practices and traffic management can be changed to avoid re-compaction the soil. In recent years, emphasis has shifted to developing more dynamic, integrated forms of site-specific management. Variable Rate Technology (VRT) for input application in precision farming based on spatial variability information of related soil/plant parameter can be made more effective if the temporal variability is also looked into. For this purpose time should be added as fourth dimension along with the already existing three dimensions (two for spatial coordinates and one for the magnitude of parameter) in the existing geostatistical softwares / analysis methods.

REFERENCES
1. Aggarwal, P., Choudhary, K.K., Singh, A.K., & Chakraborty, D. 2006. Variation in soil strength and rooting characteristics of wheat in relation to soil management. Geoderma, 136, 353363. 2. Alexandrou, A., & Earl, R. 1998. The relationships among the pre-compaction stress, volumetric water content and initial dry bulk density of soil. Journal of Agricultural Engineering and Research, 71, 7580. 3. 4. B.D. Kay & D.A. Anger. 2002. Soil structure. In: Warrick, A.W. (Ed.), Soil Physics Companion, pp. 249296. Bach, A.V., Silva, V.R., Secco, D., Reichert, J.M., & Reinert. D.J. 2000. Variabilidade espacial da resistencia mecanica do solo a` penetracao. In: Marrocos, P.C.L., Arau jo, Q.R., Sodre, G.A. (Eds.), Proceedings of the 13th Conference of RBMCSA (Reuniao Brasileira de Manejo e Conservacao do Solo e da A gua). Ilheus, Brazil, Soc. Bras. Cienc. Solo, 611 August/July (CD-ROM). 5. Bouyoucos, G.J. 1962. Hydrometer method improved for making particle size analysis of soils. Agronomy Journal, 54, 464-465. 6. Couto, E.G., & D. Mello. 2000. O uso da krigagem para identificar a resistencia a` penetracao num Latossolo Vermelho Ferrico sob diferentes coberturas e sistemas de preparo do solo. In: Marrocos, P.C.L., Arau jo, Q.R., Sodre, G.A. (Eds.), Proceedings of the 13th Conference of RBMCSA (Reuniao Brasileira de) 7. Fluton, J.P., Wells, L.G., Shearer, S.A., & Barnhisel. R.I. 1996. Spatial variation of soil physical properties: A precursor to precision tillage. ASAE paper no. 961002. International meeting, Phoenix, Arizona. 14-18 July, 1996. ASAE, St. Joseph, MI. 8. G. R.Blake & K.H. Hartage. 1986. Bulk density In: Klute,A. ed., Methods of soil analysis, Part I. Physical and Minerological Methods. Agronomy Monograph no. 9 (2nd ed.), pp. 363-375. 9. Kilic, K., Ozgoz, E., & Akbas. F., 2004. Assessment of spatial variability in penetration resistance as related to some physical properties of two fluvents in Turkey. Soil and Tillage Research, 76,1-11. 10. McKyes. 1985. Soil Cutting and Tillage. Elsevier, Amsterdam. 11. Okursoy, R. 1992. Soil compaction model. In: Proceedings of the 14th National Congress of the Agriculture Mechanization, October 1416, pp. 564573. 12. Parkin, T.B. 1993. Spatial variability of microbial process in soil: a review. Journal of Environmental Quality, 22, 409417.

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

137

13. R.N. Donald & W. Ole. 2003. Spatial and temporal statistics. Sampling field soils and their vegetation. Catena Verlag GMBH, 35447 Reiskirchen, Germany. 14. Rao, P.S.C., & Wagenet, R.J. 1985. Spatial variability of pesticides in field soils: methods for data analysis and consequences. Weed Sci. 33, 1824. Cambardella, C.A., Karlen, D.L., 1999. Spatial analysis of soil fertility parameters. Precision Agriculture, 1, 514. 15. Sarangi, A., Cox, C.A., & Madramootoo, C.A. 2005. Geostatistical methods for prediction of spatial variability of rainfall in a mountainous region. Transacsion in American Society of Agrcultural Engineering, 48, 943-954. 16. Silva, A.P., Libardi, P.L., & Vieira, S.R. 1989. Variabilidade especial da resistencia a` penetracao de umLatossoloVermelho-Escuro ao longo de uma transecao. Rev. Bras. Cienc. Solo 13, 15. 17. Soil Survey Staff, 1994. Key to Soil Taxonomy, 6th eds. ACS USDA, Washington, DC. 18. Souza, Z.M., Silva, M.L.S., Guimaraes, G.L., Campos, D.T.S., Carvalho, M.P., and Pereira, G.T. 2001.

Variabilidade espacial de atributos fsicos em um latossolo vermelho distrofico sob semeadura direta em Selviria (MS). Rev. Bras. Cienc. Solo. 25, 699 707. 19. Taylor, H. M.,& Gardner, H. R. 1963. Penetration of cotton seedling taproots as influenced by bulk density, moisture content and strength of soil. Soil Science, 96,153-545. 20. Taylor, H.M., Roberson, G.M., and Parker, Jr. J.J. 1966. Soil strength root penetration relations for medium to coarse textured soil materials. Soil Science, 102, 1822. 21. Utset, A., and G. Cid. 2001. Soil penetration resistance spatial variability in a Ferrasol at several soil moisture conditions. Soil and Tillage Research, 61,193-202. 22. Vieira, S.R., Hatfield, J.L., Nielsen, D.R., & Biggar J.W. 1983. Geostatistical theory and application to variability of some agronomical properties. Hilgardia, 51 (3), 149. 23. Webster, R. 1985. Quantitative spatial analysis of soil in the field. Advances in Soil Science, 3, 170.

APPENDICES
Table 1: Exploratory Spatial Data Analysis for All Kriged Parameters Type of Analysis No transform Nature of histogram Box-Cox Log transform Global outliers Local outliers seen through Voroni mapping Global trend examined by Trend analysis tool Nature of semivariance tool Not present present Not present Anisotropic nature Nature of Data No transformation

138

Pragati Pramanik & Pramila Aggarwal

Table 2: Estimated Error Parameters of Different Fitted Semivariance Models with Variable Lag Size (H in Meters) and Number of Lags (N) for before Ploughing h=10, n=18 Parameters Cir. Sph. Exp. Cir. Sph. Exp. Cir. Sph. Exp. Cir. Sph. Exp. PR (0-15cm) RMSE 178.1 178.3 178.6 177.9 177.1 178.3 177.7 177.2 177.7 177.6 177.6 179.2 h=18, n=10 h=12, n=15 h=15, n=12

BD (0-15cm) RMSE 0.036 0.035 0.035 0.036 0.036 0.035 0.036 0.035 0.035 0.036 0.035 0.035

SWC (0-15 cm) RMSE 1.426 1.431 1.424 1.421 1.426 1.426 1.422 1.431 1.43 1.422 1.428 1.425

PR (15-30 cm) RMSE 195.4 196.8 193.1 195.1 196.3 199.4 195.5 197.3 193.8 195.3 196.5 196.4

BD (15-30 cm) RMSE 0.069 0.07 0.073 0.07 0.069 0.072 0.069 0.07 0.073 0.069 0.069 0.072

SWC (15-30 cm) RMSE 1.301 1.303 1.299 1.3 1.3 1.297 1.301 1.302 1.298 1.3 1.3 1.294

Table 3: Estimated Error Parameters of Different Fitted Semivariance Models with Variable Lag Size (H in Meters) and Number of Lags (N) for after Ploughing h=10, n=15 Parameters Cir. Sph. Exp. Cir. Sph. Exp. Cir. Sph. Exp. h=15, n=10 h=12, n=12

PR (0-15 cm) RMSE 317.8 318 318.5 312.8 326.9 320.1 309.6 319.1 319.6

BD (0-15 cm) RMSE 0.054 0.054 0.055 0.054 0.054 0.053 0.055 0.056 0.055

SWC (0-15cm) RMSE 1.519 1.529 1.503 1.508 1.514 1.499 1.534 1.529 1.505

PR (15-30cm) RMSE 302 301.4 305.5 301.7 302.4 305.5 301.5 303.1 308.2

BD (15-30cm) RMSE 0.038 0.038 0.038 0.038 0.039 0.039 0.039 0.039 0.0386

SWC (15-30cm) RMSE 1.055 1.084 1.067 1.061 1.073 1.056 1.06 1.084 1.068

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

139

Table 4: Spatial Structure of Kriged Models of BD (Mgm-3) Before Ploughing Parameters 0-15 cm Best fitted model Lag size & no. of lag Major range(m) Minor range (m) Direction of major range (degree) Partial sill Nugget variance Anisotropic factor RMSE AVSE RMSS Nugget ratio Spatial class Exponential 12,15 177.34 177.34 8.1 0.0031 0 1 0.0346 0.0399 0.88 >70 Weak 15-30 cm Circular 10,18 177.34 177.34 8.4 0.0047 0.0026 1 0.0687 0.067 1.036 35.8 Moderate 0-15 cm Exponential 15,10 148.87 36.23 301.4 0.0016 0.0014 4.1088 0.0533 0.058 0.92 47.3 Moderate 15-30 cm Circular 15,10 148.87 36.23 0.8 0.0009 0.0004 4.1088 0.038 0.035 1.072 34.4 Moderate After Ploughing

Table 5: Spatial Structure of Kriged Models of PR (kPa) Before Ploughing Parameters 0-15 cm Best fitted model Lag size & no. of lag Major range(m) Minor range (m) Direction of major range (degree) Partial sill Nugget variance Anisotropic factor RMSE AVSE RMSS Nugget ratio Spatial class Spherical 18,10 178.64 164.33 270.2 19.98 0.1648 1.087 177.1 162.1 1.089 81.8 Weak 15-30 cm Exponential 10,18 177.34 92.185 356.4 44.26 0.6468 1.9237 193.1 196.7 0.988 1.4 Strong 0-15 cm Circular 12,12 138.13 63.93 13 21.93 0.791 2.16 309.6 325.5 0.954 3.4 Strong 15-30 cm Spherical 10,15 148.49 44.2 44.2 75.26 0 3.54 301.4 254 1.03 >70 Weak After Ploughing

140

Pragati Pramanik & Pramila Aggarwal

Table 6: Spatial Structure of Kriged Models of SWC Parameters Best fitted model Lag size & no. of lag Major range(m) Minor range (m) Direction of major range (degree) Partial sill Nugget variance Anisotropic factor RMSE AVSE RMSS Nugget ratio Spatial class Before Ploughing 0-15 cm Circular 18,10 178.64 173.3 290.4 0.8421 1.3157 1.03 1.421 1.321 1.073 60.9 Moderate 15-30 cm Exponential 15,12 176.53 94.811 0.5 2.2146 0.0421 1.861 1.294 1.26 1.036 1.8 Strong After Ploughing 0-15 cm Exponential 10,15 148.49 42.613 9.8 2.8255 0 3.1484 1.503 1.61 0.931 >70 Strong 15-30 cm Circular 10,15 148.49 60.922 59.4 0.6592 0.5183 2.437 1.055 0.995 1.037 44 Moderate

Figure 1: Examination of Trend by Trend Analysis Tool

Figure 2: Cross Validation of Predicted and Observed Data of PR of 0-15 cm of Soil before Ploughing

Delineation of Compact Zones through Spatial Variability Analysis of Soil Physical Properties

141

Figure 3: a. Kriged Map of BD before Ploughing

Figure 3: b. Kriged Map of BD after Ploughing

Figure 4: a. Kriged Map of PR before Ploughing

142

Pragati Pramanik & Pramila Aggarwal

Figure 4: b. Kriged Map of PR after Ploughing

Figure 5: a. Kriged Map of SWC before Ploughing

Figure 5: b. Kriged Map of SWC after Ploughing