A procedure for mapping the depth to the texture contrast horizon of duplex soils in south-western Australia using ground penetrating radar, GPS and kriging.
Successful precision agriculture requires accurate and detailed soil data (Cook and Bramley 1998; Shatar and McBratney 1999) to identify the spatial variation of potential constraints to plant growth such as salinity, acidity, nutrient deficiencies, and rooting depth.
In many parts of southern Australia and the south-west of Western Australia, duplex soils are the dominant soil type, occupying ~60% of the south-west agricultural area (Tennant et al. 1992). Duplex soils are characterised by the presence of prominent very sandy A 1 and A2 horizons and a uniform strong texture contrast between the A and B horizons (Hubble et al. 1983). Three soil orders, Sodosols, Chromosols, and Kurosols, exhibit a strong texture contrast between the A and B horizons, and a fourth order, Podsols, exhibits accumulations of organic matter and sesquioxides in the B horizon and can be considered as duplex soils under the Australian Soil Classification (Isbell 2002).
In a soil classification for Western Australia, Schoknecht (2002) defines duplex soils as 'soil with a texture or permeability contrast layer within the top 80 cm of the profile'. The depth criterion is included to distinguish the response of annual crops and pastures, which dominate farming systems in the south-west of WA, to the limitations imposed by shallow A horizons in duplex soils. This is primarily caused through the interaction of the Mediterranean climate with the physical properties of the duplex soil (Turner 1992). The coarse-textured A horizon usually has low plant-available water capacity, low pH buffering capacity, and poor nutritional status. The depth of the A horizon can vary significantly at small spatial scales, which contributes to the commonly seen spatial variation in yield (Passioura 1992). Also, physical and chemical characteristics of the B horizon, including poor structure, sodicity, and high salt content, can influence water storage root exploration and thus crop performance. Differences in soil moisture balance between years of different rainfall quantities or distribution can also cause large variations in yield for the same duplex soil, due to the coarse texture and consequent low moisture-holding capability of the A horizon and low hydraulic conductivity of the B horizon, in particular waterlogging, in high rainfall years (Passioura 1992).
The morphological characteristics of duplex soils also contribute to long-term insidious processes that degrade the landscape over time, and include increasing acidity in the A horizon caused in part by nitrate leaching, lateral throughflow resulting in waterlogging of low-lying areas, and increased deep flow contributing to rising watertables and, ultimately, increasing secondary salinity (Passioura 1992).
Thus, farming duplex soils presents many challenges to farmers and an ability to map the natural distribution of duplex soil profiles across large areas in a rapid, cost-effective manner would aid crop management by allowing farmers to better target management practices including gypsum and lime application, deep ripping, clay delving, fertiliser strategies, and seeding rates.
Electromagnetic induction (EMI) is most commonly used in agriculture in Australia where it is seen as a cost-effective means to rapidly map the spatial variability of soil parameters including salinity, soil water, and clay content. However, this technique can oversimplify the nature of the soil variation since [EC.sub.a] reflects a complex relationship between several soil parameters that are often site-specific and needs to be well understood before correct interpretations can be made (Corwin and Lesch 2005). Furthermore, mapping depth to clay or any other subsurface feature using EM1 is inherently difficult since the interaction between EMI signal response and soil depth is not linear, so that estimations of depth, although possible (McNeill 1980; Hendrickx et al. 2002), are rarely used. Gamma spectroscopy, another potential technique, is limited to defining soil characteristics within the top 0.30-0.45 m (Grasty 1976) because gamma emissions travelling through dry soil will attenuate by about half for each 0.10 m depth travelled, masking the distinctive gamma-ray emissions of both ferricrete gravels and non-ferruginous clayey B horizons (Verboom and Pate 2003).
Ground penetrating radar (GPR) has been studied for many years as a tool to directly measure and map the depth and constitution of soil profiles, without recourse to the intensive ground-truthing required by EMI (Mokma et al. 1990; Doolittle and Collins 1995; Inman et al. 2001; Freeland et al. 2002; Doolittle et al. 2007; Collins 2008). There has been some application of GPR to agriculture, particularly in USA where there has been considerable research for many years both in field application and in its integration with GIS systems. However, the technique has yet to gain wide acceptance as a routine soil management tool in the USA (Collins 2008), and there has been little research on its practical application to agriculture in Australia. One of the principle issues is soil suitability. GPR works best when the dielectric permittivity contrast between 2 adjoining soil layers is high and soil attenuation is low. Therefore, soils that are saline or sodic, calcareous soils, and soil with clay content >20% are considered unsuitable to GPR analysis. However, duplex soils, where a strong contrast between surficial sandy soil with a low water-holding capacity (low permittivity) and clayey subsoil with a higher water-holding capacity (high permittivity) exits, should be highly suited to this technique. What is lacking is a process converting soil horizon information to a 2- or 3-dimensional format, which would be of practical benefit to farmers since many commercial software applications are limited in their ability to generate the type of subsurface topographic information, rapidly, in a form or format that would be useful in the agriculture.
This study was designed to: develop and test a procedure for rapidly generating depth to B horizon maps for duplex soils in a format that is easily incorporated into a GIS format; quantify the accuracy of the technique by comparing the results with depth to B horizon measurements on core samples; examine the effect of GPR line spacing on map accuracy; and assess the potential of amplitude mapping as a method for differentiating broad categories of B horizon materials.
Four sites from 3 locations in the Esperance district, Western Australia, were selected for this study. They represent the range of duplex soil profiles common to the region. The site details are provided in Table 1. Two adjoining sites, AG 1 and AG2, have a predominantly fine sand A horizon over sandy to light clay B horizon, though the clay content decreases in the southern section of AG2. Site E1 is predominantly sand and gravelly sand over gravelly and mottled clays. Site ST 1 has a complex pattern of soils consisting of shallow sand, gravelly sand, sandy loam, or gravelly sandy loam in the A horizon and a variety orb horizon substrates consisting mostly of sodic clay containing various amounts of reticulated sesquioxide gravel or pedogenic carbonate.
Measurements and analysis
GPR surveys were conducted in October 2007 using a Sensors and Software [Noggin.sup.PLUS([TM])]GPR with a 250MHz antenna and datalogger on a hand-driven cart. The [Noggin.sup.PLUS] holds the GPR transmitter and receiver antenna in a cart configuration, making it easy to traverse the cultivated landscape. The 250MHz antenna provides a spatial resolution of ~0.25m and depth penetration of approximately 2 m. Transects were taken at 10-m line spacing for 3 sites and 30 m spacing for ST1. Radar data were collected along each transect at 0.25-m intervals. Details of the GPR transects for each study site are provided in Table 1. All GPR study sites and transects were geo-referenced using a global positioning system (GPS). GPS data were not collected simultaneously with the GPR but were interpolated between the start and ends points for each GPR transect and merged into a single dataset. The efficiency of data collection would be significantly improved by towed GPR by a vehicle concurrently with the GPS (Freeland et al. 2002).
Fifty soil cores to 1 m deep were extracted from the 4 sites, geo-referenced using GPS, divided into 0.1-m intervals in the field, and stored for later processing. The number of cores extracted from each site is documented in Table 1. Soil morphology of all cores was described according to McDonald et al. (1990). The depth to the B horizon was identified and compared to GPR readings from the same location. The error associated with A/B horizon identification in the core samples was [+ or -] 0.05 m. Samples representing typical morphological horizons were analysed by CSBP soil laboratories for selected chemical and physical characteristics including: panicle size, [EC.sub.1: 5], CEC, ESP, and CaCO3 content. The results are summarised in Table 2.
The depth to the structure of interest can be calculated based on the half of the 2-way travel time of the reflected EM wave multiplied by the velocity of the EM wave through the soil. The velocity through the A horizon at each study site was determined experimentally where the depth of impeding objects (B horizons and agricultural pipes) was known. Velocities were checked against core samples from the sites and were accurate to within <0.01 m/ns (<0.05 m).
The post processing of all GPR data was performed using pulseEKKO[TM] software and included correction for background noise (DEWOW), zero time static correction, and 3-point spatial average smoothing across traces. To ensure that the duplex horizons could be clearly identified, radargram signals were amplified using spreading and exponential compensation gain (SEC) for each site as required.
After processing, maps of depth to B horizon were generated in a 2-step process. First, the B horizon was manually 'picked' from the radar image, as a means of separating the data of interest from the radargram as a whole. The technique derives from a process known as coherency filtering, which has been used for many years in the interpretation of seismic data (Telford et al. 1976). This process was performed in Matlab (v7.6), by placing markers manually at intervals along the identified horizon and linearly interpolating between the markers at 0.25-m intervals to generate a file containing location, depth, and amplitude of reflection. The amplitude is the RMS value over I cycle of radar signal, which is ~5 ns in time. In the second step, interpolated GPR files and GPS data for each site were combined and depth and amplitude raster maps generated using a kriging program (Vesper 1.6, ACPA, Sydney University). Because of the intensity of the data collected, a local kriging model and variogram was employed in the kriging process (Whelan et al. 2001). Significant improvements in both the time and labour can be made by the automation of the picking process as well as integrating GPR, GPS, and mapping programs into a single package.
Results and discussion
Soil data collected from core samples and summarised in Table 2 illustrate that the contrast in clay concentration between the A and B horizons in soils at each site was >15%. The exception is site El where the contrast of clay concentration in some areas was no greater than 5-6%. However, the B horizons at this site contained 50-70% reticulated gravel, which included significant quantities of clay. All B horizon materials were sodic and the EC values indicate that the majority of the soils were non saline.
The GPR was operated at a single centre frequency with a narrow bandwidth (in this case 250 MHz) which was selected based on site knowledge and project requirements. Antenna selection is a tradeoff between spatial resolution and depth penetration; the higher the frequency, the better the resolution but the poorer the depth penetration. In conditions where the A horizon varies in depth between 0.2 and 1.5 m, a 250 MHz antenna will resolve objects with a minimum thickness of 0.25m to a depth of approximately 2m. This affects the ability of the 250MHz antenna to distinguish shallow features at <0.25 m depth, effectively putting a shallow depth limit on the GPR technique. Higher frequency antennas, while potentially providing finer detail of subsurface features, do not have the penetration required for this application. Salinity in soils will increase signal attenuation and reduce reflection quality by increasing the electrical conductivity of the soil. Sites for this study were non-saline to ensure that signal interference would be minimal; however, the impact of soil salinity of GPR data quality will require future investigation.
For all sites the 250 MHz antenna produced strong reflections at the B horizon boundary. Typical radargrams from each site are given in Fig. 1a-d. Strong continuous reflectors are evident in the 2 adjacent sites, AG1 and AG2 (Fig. 1a, b). Based on core analysis, the reflection at AGI represents a transition from fine sand (A) to sandy-loam or sandy-clay (B) and in AG2 from clay-sand (A) to sandy-loam (B) in some areas and sand (A) to clay-loam (B) in others. The reflector at site El provided data that were noisier than at the AG sites (Fig. 1d).
Unlike sites AG1 and AG2, cores from El (Fig. 1c) showed loamy sand and sandy loam A horizons over B horizons typical of lateritic soils, which texture as sandy-loam to sandy clay loam when finely ground but in their native state are cemented with sesquioxides and contain dense ironstone gravel layers (>60% v/v) at the upper margin of the B horizon. We assume that the contrast between the 2 layers was a result of water stored within cemented layers of the B horizon and the different mineralogy of the A and B horizon materials. The radargram in Fig. 1d from STI, a cross-section of about l km length, showed a large variation in depth to B-horizon ranging from areas with clay almost at the surface to a sandy A horizon extending down to 0.70-0.80 m.
A key objective of this study, to ascertain the accuracy of GPR for duplex soils, was achieved by comparing the results to core data from the same locations. The results, presented in Fig. 2, show a good relationship between actual (core) and GPR depth to B horizon, with a correlation coefficient (r) of 0.90. Based on the regression analysis the standard error for the GPR prediction of B horizon depth is 4-0.1 m. The likely sources of error in this relationship are due to variations in the velocity in the A horizon across the site, which are most commonly caused by changes in lithology or soil water content (SWC) with depth. The lithology of A horizon will be relatively uniform for duplex soils with changes in clay content down profile of only a few percent. In the absence of mechanical compaction there are only small changes in the bulk density of soils caused by overburden, since the soils are predominantly coarse-textured sandy soils with a relatively uniform particle size distribution. Soil water content influences velocity (v) via its relationship with the dielectric permittivity of soil (e), where:
v = 0.3/[[epsilon].sup.0,5] (l)
Topp et al. (1980) described the relationship empirically between dielectric permittivity and SWC ([theta]) expressed as weight per volume (w/v) for any given soil, as follows:
[epsilon] = 3.03 + 9.030 + 146[[theta].sup.2] - 76.7[[theta].sup.3] (2)
Therefore, it is possible to estimate how changes in SWC for a given site can affect velocity and ultimately calculated depth. The GPR trials were conducted in the drier summer months, and soil samples from AGI and AG2 analysed in the field using capacitance equipment put the range for SWC in a typical l-m profile at 0.01-0.04 w/v. Using Eqns 1 and 2 we determined the difference in calculated GPR depth for a 2 way travel time of 25ns. Based on EM velocities derived from a SWC of 0.01 and 0.04, the difference was 0.18m, which falls within the error [+ or -]0.1 m determined in Fig. 1. For practical purposes only, a small number of velocity estimations were made at any given location and the use of a Noggin GPR with a fixed antenna length negates the use of the common midpoint method (CMP) of determining the average velocity of soils. Importantly, this exercise and previous studies (Boll et al. 1996) suggest that velocity errors in duplex soils can be minimised by performing the surveys during periods when soil moisture is at its most uniform, which in the agricultural area of south-western Australia may be either post harvest (December) when SWC is at its lowest or in periods of high rainfall when the soils are nearing field capacity (mid winter).
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
The slope of 0.68 for the regression line implies that the GPR underpredicts depth to B horizon with respect to analysis of core samples, though as stated previously there is some error in the depth estimation of the core samples. Figure 2 also shows there are no radar depth to B horizon values <0.25 m irrespective of the core depth measured, thus directing the regression line through an intercept of-0.18 m rather than zero. This work confirms the limited capability of GPR equipped with a 250MHz antenna to detect texture contrasts any shallower that ~0.25 m.
Depth to B Horizon maps
GPR data were kriged using Vesper spatial prediction software to generate contour maps showing depth to B horizon for all sites. Information in this format is easily integrated into a GIS. The maps for AG1, AG2, and E1 are shown in Fig. 3a-c for 10-m line spacing. Figure 3d shows the contour map for ST1 at 30-m spacing only. The map for AGI (Fig. 3a) shows the depth to the B horizon to be uniform across most of the site at 0.44).5 m, dipping to >1 m in the northern comer. For AG2 (Fig. 3b), the B horizon was shallow across the middle of the site with values of 0.4-0.5m, increasing in depth at both the northern and southern ends of the site. Both E1 and ST1 show greater variation in depth to B horizon across the sites.
[FIGURE 3 OMITTED]
In order to determine how representative the kriged maps were of the actual depth to B horizon, 2 GPR transects were taken across the each site at right angles to the original GPR lines. GPR depth data were overlaid onto kriged depth based on georeference information, and the data were compared. The results in Table 3 for 10-m line spacing show a very close linear relationship between measured values of depth to B horizon for GPR tie lines and the kriged GPR data for AG1 and AG2. In both cases the slope is ~1, the intercept is 0, and correlation coefficients (r) were high (0.99 and 0.98 for AG1 and AG2, respectively). For ST1 where the site was mapped at only 30-m spacing the value of r was 0.91. For all maps the standard error for the GPR prediction of B horizon depth is [+ or -]0.1 m. Comparisons were not made at the E1 site because the GPS information for the tie lines were not recorded correctly.
We also evaluated how increasing line spacing from 10 to 20 and 30 m affected map accuracy. Thirty metres is the line spacing used by some commercial practitioners mapping with ground-based EMI and gamma radiometrics on broadacre properties (Quenten Knight, Precision Agronomics, pers. comm.). Lines of GPR data were removed from datasets and raster maps re-kriged. Data from the resulting maps were compared to tie lines as done previously for the 10-m line spacing. The results are presented in Table 3 and the corresponding maps in Figs 4-6. For AG 1 where for the majority of the surveyed site the increase in depth to the B horizon is gradual (across 50 m) and the loss in map quality is minimal. The effect on map quality for AG2 is small with a decrease in the slope of the regression line from 0.91 for 10 m to 0.79 for 30-m spacing and the correlation coefficient decreasing from 0.98 to 0.91.
[FIGURE 5 OMITTED]
This analysis shows that kriging of GPR data can generate quite accurate maps of depth to B horizon for duplex soils, but map quality will deteriorate if there are significant changes in depth to B horizon over distances less than the line spacing used. However, in an agricultural context where precision management is unlikely to be practicable at a 10-m scale, such errors may be considered acceptable when balanced against the cost of performing collection of GPR data at 10 m v. 30 m intervals.
Reflected amplitude mapping
Spatial maps can also be generated using a technique known as amplitude mapping where the relative strength of the reflected radar signal is mapped, highlighting zones of weak and strong reflection and identifying potential anomalies that may exist in the subsurface. The technique is commonly used in the detection of subsurface drainage systems (Allred et al. 2005) and in the mapping of archaeological sites (Conyers and Cameron 1998). Attempts have been made to correlate amplitude intensity to soil characteristics such as [EC.sub.a], although with limited success (Inman et al. 2001; Bradford 2007).
[FIGURE 6 OMITTED]
We investigated whether the intensity of the reflection at the A-B horizon boundary could be related to soil characteristics such as clay and gravel content. As the contrast in the dielectric constant between the 2 adjacent media increases, the intensity of the reflected radar pulse would increase; thus, the character of the B-horizon affects the amplitude of the reflected signal. If such information could be discerned from GPR data it would be useful in mapping subsoil properties.
[FIGURE 7 OMITTED]
The B-horizon materials recovered in cores were classified into 6 texture categories: 1-5%, 5-10%, 10-20%, 20-30%, 30-35%, >35% clay content. These were then compared to the amplitude of the corresponding GPR signal at the same point.
The results in Fig. 7 show that only for sites AG1 and E1 was there a trend of increasing amplitude with increasing clay content. For example, at AG2 there is significantly less clay at the southern end of the mapped site, but the amplitude range for the traverse was small, ranging from 16000 to 25 000 BV. A classification of core materials grouped as clay, loam, or gravel was also unsuccessful in predicting amplitude values. We conclude that there no simple relationship between amplitude and soil texture. This is most likely to be caused inter alia by variations in moisture, clay content, salt content, and mineralogy affecting signal attenuation.
Mapping the depth to the B horizon in duplex soil using ground penetrating radar has demonstrated advantages over the current method of soil surveying in agriculture using EMI, since it can provide a direct accurate measure of depth to B horizon without the amount of ground truthing normally required for EMI. Additional investigations of amplitude mapping are required to determine whether influence of B horizon composition on GPR response may provide additional information of value to farmers.
Future work should focus on the comparison of GPR with other soil survey techniques and crop yield data to gain a better understanding of the extent to which GPR can explain crop variability in duplex soils. With improvements in soil mapping using a more streamlined package for processing, 'picking' and kriging GPR could prove a highly effective tool for agriculture in Australia.
This work was funded by the Grains Research and Development Council of Australia and by South Coast Natural Resource Management Inc., with funding provided by the Australian and Western Australian Governments through the National Action Plan for Salinity and Water Quality and the Natural Heritage Trust II. The authors wish to acknowledge the technical assistance of Miss Kelly Kong from DAFWA, Esperance, and the farmers who kindly allowed us to conduct research on their properties, Mr and Mrs Agnew and the Stead family.
Manuscript received 27 October 2008, accepted 6 May 2009
Allred B J, Daniels JJ, Fausey NR, Chen C, Peters L Jr, Youn H (2005) Important considerations for locating buried agricultural drainage pipe using ground penetrating radar. Applied Engineering in Agriculture 21(1), 71-87.
Boll J, van Rijn RPG, Weiler KW, Ewen JA, Daliparthy J, Herbet S J, Steenhuis TS (1996) Using ground penetrating radar to detect layers in a sandy field soil. Geoderma 70, 117 132. doi: 10.1016/0016-7061(95) 00077-1
Bradford JH (2007) Frequency-dependent attenuation analysis of ground-penetrating radar data. Geophysics 72(3), J7-Jl6. doi: 10.1190/1.2710183
Collins ME (2008) A history of GPR application in agriculture. In 'Handbook of agricultural geophysics'. (Eds BJ Allred, JJ Daniels, MR Ehsani) p. 45. (CRC: Boca Raton, FL)
Conyers LB, Cameron CM (1998) Ground-penetrating radar techniques and three-dimensional computer mapping in the American southwest. Journal of Field Archaeology 25(4), 417-430. doi: 10.2307/530637
Cook SE, Bramley RGV (1998) Precision agriculture--opportunities, benefits and pitfalls of site-specific crop management in Australia. Australian Journal of Experimental Agriculture 38, 753-763. doi: 10.1071/EA97156
Corwin DL, Lesch SM (2005) Apparent soil electrical conductivity measurements in agriculture. Computers and Electronics in Agriculture 46, 11-43. doi: 10.1016/j.compag.2004.10.005
Doolittle JA, Collins ME (1995) Use of soil information to determine the application of ground penetrating radar. Journal of Applied Geophysics 33, 101-108. doi: 10.1016/0926-9851(94)00031-I
Doolittle JA, Minzenmayer FE, Waltman SW, Benham EC, Tuttle JW, Peaslee SD (2007) Ground penetrating radar soil suitability map of the conterminous United States. Geoderma 141, 416-421. doi: 10.1016/ j.geoderma.2007.05.015
Freeland RS, Yoder RE, Ammons JT, Leonard LL (2002) Integration of real-time global positioning with ground penetrating radar surveys. Applied Engineering in Agriculture 18(5), 647-650.
Grasty RL (1976) Applications of gamma radiation in remote sensing. In 'Remote sensing for environmental sciences'. (Ed. E Schanda) p. 257. (Springer Verlag: Berlin)
Hendrickx JMH, Botchers B, Corwin DL, Lesch SM, Higendorf AC, Schlue J (2002) Inversion of soil conductivity profiles from electromagnetic induction measurements: theory and experimental verification. Soil Science Society of America Journal 56(6), 1933-1941.
Hubble GD, Isbell RF, Northcote KH (1983) 'Soils: an Australian viewpoint.' Division of Soils, CSIRO. pp. 40-42. (CSIRO: Melbourne/Academic Press: London)
Inman DJ, Freeland RS, Yoder RE, Ammons JT, Leonard LL (2001) Evaluating GPR and EMI for morphological studies of loessial soils. Soil Science 166, 622-630. doi: 10.1097/00010694-200109000-00006
Isbell RF (2002) 'The Australian soil classification.' Revised edn (CSIRO Publishing: Collingwood, Vie.)
McDonald RC, lsbell RF, Spieght JG, Walker J, Hopkins MS (1990) 'Australian soil survey.' 2nd edn (Inkata Press: Melbourne, Vic.)
McNeill JD (1980) Electrical conductivity of soils and rocks. Technical Note TN-5. Geonics Ltd, Mississauga, ON, Canada.
Mokma DL, Schaetzl EP, Johnson EP, Doolittle JA (1990) Assessing Bt horizon character in sandy soils using ground penetrating radar: implication for soil survey. Soil Survey Horizons 31(1), l 18.
Passioura JB (1992) Overview of the processes limiting crop production on duplex soils. Australian Journal of Experimental Agriculture 32, 987 990. doi: 10.1071/EA9920987
Schoknecht N (2002) Soil groups of Western Australia. A simple guide to the main soils of Western Australia Edition 3 Land Resource Assessment and Monitoring Group, Department of Agriculture and Food, Western Australia. Resource Management Technical Report 246.
Shatar TM, McBratney AB (1999) Empirical modeling of relationships between sorghum yield and soil properties. Precision Agriculture 1, 249 276. doi: 10.1023/A:1009968907612
Telford WM, Geldart LP, Sheriff, Keys D (1976) 'Applied geophysics.' (Cambridge University Press: Cambridge, UK)
Tennant D, Scholz G, Dixon J, Purdie B (1992) Physical and chemical characteristics of duplex soils and their distribution in the south-west of Western Australia. Australian Journal of Experimental Agriculture 32, 827-843. doi: 10.1071/EA9920827
Topp GC, Davis JL, Annan AP (1980) Electromagnetic determination of soil water content: measurements in coaxial transmission lines. Water Resource Research 16, 574-582.
Turner NC (1992) Crop production on duplex soil: an introduction. Australian Journal of Experimental Agriculture 32, 797 800. doi: 10.1071/EA9920797
Verboom WH, Pate JS (2003) Relationships between cluster root-bearing taxa and laterite across landscapes in south west Western Australia: an approach using airborne radiometric and digital elevation models. Plant and Soil 248, 321 333. doi: 10.1023/A:1022358014629
Whelan BM, McBratney AB, Minasny B (2001) Vesper-Spatial prediction software for precision agriculture. In 'ECPA 2001. Proceedings of the 3rd European Conference on Precision Agriculture'. Montpellier, France, agro-Montpellier, pp. 139 144.
M. A. Simeoni (A,D), P. D. Galloway (B), A. J. O'Neil (C), and R. J. Gilkes (A)
(A) School of Earth and Environment, The University of Western Australia, WA, Australia.
(B) Department of Agriculture and Food, Western Australia, Esperance, WA, Australia.
(C) Down Under GeoSolutions, Subiaco, Western Australia, Australia.
(D) Corresponding author. Email: email@example.com
Table l. Information for the study sites at Esperance, Western Australia Site ID Location Soil and site description AGI Neridup 40 km Predominantly uniform fine north-east of sand overlying light clay at Esperance; between 0.35 m and 1.40 m S33[degrees] 34'30", E122[degrees] 04'35" AG2 Neridup 40 km Predominantly fine sand north-east of overlying light clay at Esperance; between 0.45 1.00 m, with 533[degrees] 34'51", occasional enclaves of a E 122[degrees] 04'36" B1 horizon of either gravelly sand or sandy loam at the interface between A and B horizons EI Gibson, 30 km Predominantly fine sand north of overlying a complex mosaic Esperance; of either medium clay, S33[degrees] 36'26", or ironstone gravels and E121[degrees] 4707" reticulite. Some areas had gravel at the surface, while the fine sand varied in depth from 0.10 to 1.00 m STI Lort River, 90 km Complex mosaic of shallow west of sandy and loamy A horizons Esperance; varying in depth from S33[degrees] 40'07", 0.10 to 0.50 m, and often E121[degrees] 11'15" containing ironstone gravels, overlying medium to heavy clay. Some areas displayed gilgai microrelief and in these areas the soil was a heavy, cracking clay, sometimes with a shallow loamy A horizon, not >0. l0 m deep Site ID Australian soil No. of Line classification GPR length lines AGI Eutrophic, mottled, 19 200 mesonatric brown Sodosol AG2 Major: eutrophic, mottled, 7 400 mesonatric brown Sodosol, Minor: ferric-sodic, hypocalcic, brown Chromosol EI Bleached-ferric, hypocalcic, 7 430 brown Chromosol and densic-placic, sesquic, aerie Podosol STI Complex of. ferric, subnatric, 4 1000 grey Sodosol; ferric subnatric brown Sodosol; hypocalcic, subnatric grey Sodosol; ferric, mottled, subnatric grey Sodosol; self-mulching brown Sodosol Site ID Line No. of Land spacing core history (m) samples AGI 10 11 Permanent lucerne pasture AG2 10 12 Permanent lucerne pasture EI 10 15 Crop and annual pasture rotations STI 30 37 Crop and annual pasture rotations Table 2. Physio-chemical data ranges for soil cores for 4 sites Site Cores Soil Sand Silt Clay ID analysed horizon (%) AG1 2 A 96.2-96.9 0 3.1-3.8 B 69.3-83.4 0-1.0 15.6-30.7 AG2 2 A 94.29 1.32 2.9-3.8 B 65.0-80.2 1.04 15.2-34.0 E1 3 A 91.50 1.0-19 2.9-7.7 B 78.1-87.5 1.0-1.9 8.7-20 ST1 5 A 71.1-90.2 2.07 4.9-25.9 B 22.6-73.3 1.99 18.1-67.7 Site Carbonate Gravel CEC Sat. ESP ID ([cmol.sub.c]/ conductivity (%) kg) (dS/m) AG1 -0.10 0 -0.10 0.3-1.0 IS 0.2 (^) 0 4.7 (^) 1.1-3.6 38.1 (^) AG2 -0.10 0 -0.10 0.2-0.6 IS 20.50 0 2.8-8.3 19-2.8 29.94 E1 <0.1-0.l 0 1.00 0.2-0.4 20.13 0.1-0.2 50-70 15.34 0.40 25-39.7 ST1 -0.10 0 IS (B) 0.5-1.1 IS (B) 0.2-9.2 0-80 15.8-35.0 0.4-2.2 29.2-42.2 (^) Single core sample. (B) Insufficient sample. Table 3. An assessment of raster map quality by comparing kriged depth to B horizon data at different line spacing (x) against measured values for GPR tie lines collected at the field sites (y) n.r., No results are available for site E1 because of errors in the GPS data. -, No transects, only measured at 30 m Site ID GPR line Regression Correlation spacing (m) equation coefficient (r) AG1 10 y = 0.99x + 0.011 0.99 20 y = 0.99x + 0.045 0.95 30 y = 0.96x + 0.016 0.96 AG2 10 y = 0.91x + 0.089 0.98 20 y = 0.83x + 0.015 0.97 30 y = 0.79x + 0.019 0.93 E1 10 n.r. n.r. 20 n.r. n.r. 30 n.r. n.r. ST1 10 - 20 - - 30 y = 0.95x + 0.049 0.91
|Printer friendly Cite/link Email Feedback|
|Author:||Simeoni, M.A.; Galloway, P.D.; O'Neil, A.J.; Gilkes, R.J.|
|Publication:||Australian Journal of Soil Research|
|Date:||Sep 1, 2009|
|Previous Article:||An expert system to predict intricate saline--sodic subsoil patterns in upland South Australia.|
|Next Article:||Sequential indicator simulation and indicator kriging estimation of 3-dimensional soil textures.|