VZJ sign up for etocs
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 26 May 2006
Published in Vadose Zone J 5:720-730 (2006)
DOI: 10.2136/vzj2005.0095
© 2006 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
Related Collections
Right arrow Dryland Soils
Right arrow Field-Scale Studies
Right arrow Ground Penetrating Radar, GPR

ORIGINAL RESEARCH

Estimating the Fine Soil Fraction of Desert Pavements Using Ground Penetrating Radar

Darren G. Meadowsa,*, Michael H. Younga and Eric V. McDonaldb

a Desert Research Institute, Nevada System of Higher Education, 755 E. Flamingo Rd., Las Vegas, NV 89119
b Desert Research Institute, Nevada System of Higher Education, 2215 Raggio Parkway, Reno, NV 89512
* Corresponding author (Darren{at}dri.edu)

Received 29 July 2005.



   ABSTRACT
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
A reconnaissance tool that can estimate the clay content and saturated hydraulic conductivity (Ks) of well-developed desert pavements has immediate applications for researchers and practitioners who work in arid environments. We examined the use of surface-based ground penetrating radar (GPR) to rapidly approximate these properties on six, 100-m-long transects for two different aged (~100 000 vs. 4000 yr) desert pavement surfaces. We determined early-time amplitudes from GPR transects, ground electrical conductivity, hydraulic properties, clay content, water content, and soil salinity at regular intervals along each transect. Both surfaces were low in water content and salinity; however, the older pavement contained substantial amounts of silt and clay in the surficial soil horizon. Using multivariate linear regression, which included GPR amplitude and a nominal measure of soil structure ascertained by visual field inspection, we show significant correlations between measured and predicted values of both silt plus clay content (r = 0.84, P < 0.0001) and Ks (r = 0.73, P < 0.0001) for the older surface. No significant correlations were found on the younger surface. This is probably due to the low concentrations of clays in the young soil. Including the metric of soil structure improved the predictive capabilities on the older surface. The GPR method provides higher spatial resolution than electromagnetic measurements. The results suggest that, at this location, the approach is not influenced by heterogeneities in lower soil horizons. The method can be used for reconnaissance surveys as a means of estimating the clay content, and in some cases Ks, of surficial soils on certain well-developed desert pavements.

Abbreviations: EC, electrical conductivity • EM, electromagnetic • GPR, ground penetrating radar • TDR, time domain reflectometry


   INTRODUCTION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
THE EXPLOITATION of resources in arid regions of the southwestern USA is an inevitable consequence of population growth in these areas. Overuse through either development or intensive land use (i.e., grazing and ranching) could have ecological impacts. Thus, as the population of the arid southwestern USA continues to grow, decisions regarding the management of the surrounding deserts and their ecosystems are becoming increasingly important; however, the dynamics among soil morphology, hydrology, and ecology are not fully understood, which impedes the development of long-term management strategies. These dynamics often affect the amount of precipitation that infiltrates into soil where plants can utilize it, or the runoff that can cause erosion and flooding.

Although management of desert lands is indeed complex and multifaceted, this research focused on understanding and predicting the amount of water infiltrating into soil and how this could change through natural soil development or anthropogenic soil disturbance. The ability to rapidly characterize the permeability of desert surfaces also has direct implications for regional flood control models and contaminant transport studies; however, standard point measurements of hydraulic properties are time consuming and the heterogeneity of desert soil surfaces is often significant. A means of rapidly estimating the complex variations in surface morphology in desert environments and the associated infiltration capacities would enhance the predictive capabilities of scientists, planners, and engineers.

Desert pavements are landscape-scale features common in arid regions worldwide; Evenari et al. (1985) estimated that 50% of desert surfaces in North America are stone pavements. Pavements are comprised of a surficial layer of closely packed gravel (typically one stone thick) partially embedded in an underlying vesicular A (Av) soil horizon. The Av horizon varies in thickness from <1 cm to >10 cm and is generally correlated with the age of the surface (McDonald, 1994). The Av horizon is comprised of interconnected prismatic soil peds, each with a platy internal structure as well as abundant air vesicles.

In desert pavement environments, pedogenesis and Av horizon development is primarily driven by deposition of eolian silt and clay (McFadden et al., 1987; McDonald, 1994). Soils develop, the surface aggrades, and structure increases over many millennia as more fines are deposited and incorporated into the soil profile. Thus, clay content, Av thickness, and structure generally increase as a function of age.

The highly structured, fine-grained nature of the Av horizon can affect the permeability of soil, and thus have important ecological consequences in water-limited desert environments. As surface age increases, Av thickness generally increases with a concomitant decrease in permeability (McDonald, 1994; Young et al., 2004). This relationship has direct implications for biodiversity (Shafer et al., 2004) and plant community structure (McAuliffe, 1994). Shafer et al. (2004) showed that soil surfaces with ages of approximately 5000 to 10000 yr exhibited the highest biodiversity, and that the diversity declined as the soils continued to age. Although the reasons for this are not completely understood and are still being investigated, there appears to be an optimal characteristic of the surface soil that balances water-holding capacity with infiltration capacity.

Methods to infer infiltration capacity of the Av horizon of desert pavements are also needed in arid regions for large-scale predictions of surface flooding. Flash floods are particularly common in desert environments where rainfall patterns, rates, and intensity are highly variable and where complex tectonic and geomorphologic forces juxtapose surfaces of widely varying infiltration capacity. Foody et al. (2004) used a hydrologic model based on Landsat images to derive land cover distribution and field soil measurements to predict regions prone to flash flooding. Their model used a single value (0.07 cm/h) for the infiltration capacity for desert pavement; however, infiltration capacity of desert pavements varies widely with surface age (McDonald, 1994; Young et al., 2004). Thus, a means to improve large-scale predictions may be to collect more field data on a variety of different-aged pavements and incorporate that heterogeneity into the model.

A variety of methods exist for estimating permeability in the field. The most popular are the tension disk infiltrometer (Ankeny et al., 1991; Jarvis and Messing, 1995; Simunek and van Genuchten, 1996; Evett et al., 1999; and others) and the ring infiltrometer (Youngs, 1987; Reynolds and Elrick, 1990; Elrick et al., 1995). Most infiltration methods, though, are point measurements. Therefore, data must be collected from a sufficient number of points to fully capture the variability in the soil properties. Depending on the size of the study area, this can easily become prohibitive.

Ground penetrating radar is a noninvasive and relatively quick means of imaging near-surface layering and has been used for soil studies since the 1980s. Early GPR studies focused on reflection profiling to map shallow stratigraphy (Davis and Annan, 1989; Beres and Haeni, 1991; Smith and Jol, 1992). More recently, cross-borehole tomographic techniques have been used to create three-dimensional radar velocity distributions to investigate soil moisture dynamism (Eppstein and Dougherty, 1998; Alumbaugh et al., 2000; Binley et al., 2001). Radar attenuation tomography has also been used (Lane et al., 2000) to delineate contaminants in the vadose zone with a saline tracer to map fractures. Hubbard et al. (2001) combined velocity and attenuation tomography to study relationships between the GPR-derived parameters and hydraulic conductivity. In a vadose zone study, inversion of tomographic attenuation data yielded high-resolution images of clay distribution due to the high attenuation caused by the clay (Chang et al., 2004). Attenuation of the GPR signal is proportional to the EC (electrical conductivity) of the medium:

Formula 1[1]
where {alpha} is the attenuation (dB/m), µ0 and {varepsilon} are the magnetic permeability (H/m) and dielectric permittivity (F/m), respectively, {sigma} is the EC of the medium (S/m), and k is the dielectric constant.

The use of geoelectrical properties to infer hydraulic properties is not new. Much of this work focused on using resistivity studies to estimate aquifer properties (Kelly, 1977; Heigold et al., 1979; Kosinski and Kelly, 1981; Kelly and Frohlich, 1985); however, Mazac et al. (1985) showed both direct and inverse correlations between aquifer and geoelectrical properties. These relationships were largely dependent on the rock type and clay content of the aquifer. In general, positive correlations existed when reduced values of resistivity and permeability resulted from surface conduction along grains; negative correlations indicated a dependence on porosity and the conductivity of the pore water. Mazac et al. (1989) extended application of geoelectrical methods to estimate the saturated hydraulic conductivity of unsaturated soils. Our study differs in that we used GPR, generally a more mobile geophysical instrument, to derive these relationships. The scale of measurement is also substantially smaller using GPR, so the spatial resolution is greater. Our focus was also on unsaturated, near-surface soils rather than aquifers.

Because attenuation of the electromagnetic wave is increased due to surface conduction along clay particles, it should be possible to estimate the clay content of a surficial soil horizon by examining variations in the amplitude of the direct wave and short-time waves that travel in the surface material. In short, we attempted to exploit one of the greatest limitations of the technology by inferring near-surface soil properties based on the degree of signal attenuation. The objective of this study was to investigate the efficacy of using variations in signal amplitude from surface-based GPR and electromagnetic survey measurements as a means of predicting the fine soil fraction and hydraulic properties of the surficial soil horizon on two different aged desert pavements. Specifically, we addressed two hypotheses: (i) GPR can be used to estimate the clay plus silt content in desert pavement environments, and hence provide an indication of the soil saturated hydraulic conductivity, and (ii) the ability of GPR to estimate clay plus silt does not depend on soil age.


   MATERIALS AND METHODS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The field work for this study was conducted on two different-aged desert pavement surfaces in the Mojave National Preserve (34.16, –115.06; 33.41, –115.06; 33.41, –115.37; 34.16, –115.37) in southeastern California, ~120 km south-southwest of Las Vegas, NV. The field area is within an alluvial fan complex flanking the Providence Mountains. The surfaces are designated in the QfX convention, with Q designating Quaternary, f being fluvial, and X being a number related to the age of the surface, 1 for the oldest and 8 for the youngest active washes. The two surfaces surveyed in this study were mapped as a Qf3 and a Qf6, representing ages of approximately 100000 and 4000 yr, respectively (McDonald, 1994). These surfaces were chosen because they represent end members regarding pedologic and structural development of desert pavements. The parent material for both surfaces is composed of mixed plutonic rocks. McFadden et al. (1986) showed that the clay mineralogy of a nearby Av horizon of a 140000-yr-old desert pavement was ~27% kaolinite, smectite, and quartz and 18% mica.

On each surface, six 100-m-long parallel transects were marked with 20-m lateral separation, resulting in two, 100- by 100-m squares. Figure 1 shows the transects and the measurement locations. The transects were situated approximately parallel to the strike of the fan beginning downslope. The square on the Qf6 pavement was slightly abridged due to the natural shape of the surface. We conducted common offset GPR surveys on each transect on both surfaces. A NogginPlus 1000 GPR unit (Sensors & Software, Inc., Mississauga, Ontario, Canada) was used for GPR surveying at 1000 MHz frequency. An advantage of this unit is its mobility across rough terrain; the antennas are encased in a hard plastic shell mounted on swinging arms, which are attached to a cart on four wheels. Although a modular GPR would allow more precise velocity determination, such as the walkaway procedure, the rocky terrain and presence of desert shrubs necessitated a more compact, mobile, and rugged unit. Traces were collected every 2 cm along all transects using four stacks. Ground penetrating radar data were processed using a high-pass "dewow" filter on each trace to remove low-frequency signal that can be induced by the transmit signal and superimposed on the trace. This correction removes the low-frequency component while preserving the high-frequency reflections. Traces were also enveloped, transforming the positive and negative portions of the wavelet into a positive-only monopulse wavelet. Thus, the positive and negative wave components do not cancel each other during the amplitude analysis.


Figure 1
View larger version (15K):
[in this window]
[in a new window]
  		Fig. 1. Illustration of GPR (ground penetrating radar) transects, EM-38 (electromagnetic survey) measurements, TDR (time domain reflectometry), infiltrometer, and particle size sampling locations.

 
Initial transects were analyzed using depth of signal penetration maps. Depth of penetration is calculated on every trace by establishing the average noise level and then determining the time that the signal amplitude drops below that value. Penetration maps were produced using EKKOMapper V2.0 (Sensors & Software, Inc.); however, a more robust means of quantifying GPR response, as opposed to measuring only depth of penetration, is through the average early-time amplitude. The early-time amplitude is the average enveloped amplitude values from 0 to 2 ns. Using the depth of penetration as the primary parameter relies on the presence of subsurface reflectors. In these extremely heterogeneous environments, large cobbles are nearly ubiquitous throughout the subsurface and reflectors were thus expected to be present; however, the homogeneity of reflectors may differ, resulting in a nonuniform comparison. We therefore used the early-time amplitude so that only the near-surface material was sampled.

Transect maps depicting the average amplitude from 0 to 2 ns of each trace were produced using EKKO_Mapper V2.0 (Sensors & Software, Inc., Mississauga, Ontario, Canada). The time-zero point for the traces coincides with the arrival of the air wave. The choice of 2 ns allows for full arrival of the direct wave as well as very early reflections, which travel in the surface horizon. We used material velocities of 0.11 m/ns for the Qf6 and 0.13 m/ns for the Qf3 surfaces, both of which were determined by hyperbola matching. When the electromagnetic wave impinges on a point object of different electrical properties than the surrounding media (e.g., a cobble), a diffraction hyperbola is produced in the record. The hyperbola is a result of the hemispherical wavefront that travels through the ground. The apex of the hyperbola is produced when the GPR is directly over the point object (the shortest travel path). As the antennas move away from the object, the signal travels farther and thus appears later in time or deeper in the record. The slope of the hyperbola asymptotes, which is measurable as a width at a point below the apex, is related to the velocity of the material through which the wave travels. The asymptotes become steeper as the velocity of the material decreases. The velocity values we determined are typical based on the water content of the material at the time of the surveys (generally <0.05 cm3/cm3). The vertical wavelength was calculated to be 0.13 m on the Qf3 surface and 0.11 m for the Qf6.

Electromagnetic surveys were conducted with an EM-38 (Geonics Ltd., Mississauga, Ontario, Canada) on the same GPR transects at sampling intervals of 2 m. Measurements were performed in the horizontal dipole mode, which has a shallower depth of investigation than the vertical dipole (McNeill, 1980). The readings were taken with the instrument perpendicular to the transects. The instrument was calibrated twice per day according to manufacturer guidelines.

Time domain reflectometry (TDR) measurements were made at the same locations as the EM-38 readings. Measurements were taken with a three-wire waveguide, 10 cm long, inserted vertically into the ground. The TDR probe was connected to a cable tester (Model 1502B, Tektronix Corp., Beaverton, OR), which was wired to a datalogger (CR10X, Campbell Scientific, Logan, UT). Volumetric water contents were then determined from the dielectric constants using Topp's equation (Topp et al., 1980). All geophysical and TDR data for each transect were collected on the same days (5 and 6 Apr. 2005) to avoid temporal changes in water content.

In addition to the geophysical measurements, the van Genuchten (1980) hydraulic properties were determined at regular intervals along each transect using a tension disk infiltrometer (Soil Measurement Systems, Tucson, AZ). Nine infiltrometer experiments were conducted per transect with 12.5 m between locations (54 measurements total). Experiments were conducted and data were analyzed as described by Young et al. (2004) using HYDRUS-2D (Simunek et al., 1999). A soil sample was then collected from the surface for particle size analysis using the laser light scattering technique (Saturn DigiSizer Model 5200, Micromeritics Instruments, Norcross, GA). As an indication of soil salinity, the soluble salt content of the soil samples was also determined using a 1:1 soil/water extract (Dahnke and Whitney, 1988). All readings were adjusted to 25°C. We used the method of Zhang et al. (2005) to convert the 1:1 soil/water extracts to the equivalent saturated paste value.

Regression analyses were conducted to better understand the relationships between GPR response, water content, soluble salt content, hydraulic properties, EM measurements, and soil texture components, specifically clay content. These analyses were performed to determine if a rapid, noninvasive survey could be used to estimate clay content of the near-surface soil. Two analyses were conducted. The first analysis was a univariate regression, which used only GPR amplitude to predict the silt plus clay content; however, the results of Meadows et al. (2005) showed the importance of structure in these complex soils. Thus, to improve the level of correlation between GPR and soil physical properties, a second analysis added a metric of structural development as a second independent variable. By identifying changes in surface conditions, a rapid examination of the soil structural development could be done by a person relatively untrained in soil morphology. A nominal scale was therefore devised, in which the degree of pavement development was assigned a categorical variable: 1 = no pavement, 2 = minimal pavement, and 3 = well-developed pavement. Values were dummy coded and transformed into two dichotomous variables for inclusion in a multivariate linear regression analysis. Thus, for our case, predicted y values of silt plus clay were made based on the following:

Formula 2[2]
where c0, c1, c2, and c3 are regression coefficients; x1 is the GPR amplitude; and x2 and x3 are the categorical variables, where (x2,x3) = (0,0) designating no pavement, (x2,x3) = (1,0) designating minimal pavement, and (x2,x3) = (0,1) designating well-developed pavement.


   RESULTS AND DISCUSSION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Figure 2 shows raw radar data from a preliminary test transect traversing from the Qf3 into an active wash and finally onto the Qf6. The figure shows a distinct lack of reflectors underneath the Qf3 surface, and then a reappearance of reflectors as the GPR was moved toward the wash. Assuming that reflectors were continuous across these surfaces, one can infer that the clay-rich Av horizon on the Qf3 surface caused the "blanking" of the radar signal. Figure 3 shows the same transect as a color-coded map depicting the depth of penetration of the radar signal. It is clear that the depth of penetration for the two surfaces is different. It is also possible to distinguish different depths of penetration between the active wash and the Qf6. The diamonds on the horizontal axis indicate soil sample locations and the percentages represent corresponding clay content.


Figure 2
View larger version (86K):
[in this window]
[in a new window]
  		Fig. 2. Raw radar data beginning on the older Qf3 desert pavement, traversing a wash, and ending on the younger Qf6 pavement.

 

Figure 3
View larger version (34K):
[in this window]
[in a new window]
  		Fig. 3. Color-coded map of radar signal penetration depth on the older Qf3 surface, traversing a wash, and ending on the younger Qf6 surface. Red stars indicate where soil samples were taken for particle size analysis, and percentages are clay content.

 
To show that the blanking phenomenon is not restricted to a particular parent material but can be more generally expected, we also surveyed a nearby desert pavement surface formed on a different parent material. Figure 4 shows a map of the GPR penetration depth on a Qf3 surface at the Cima Volcanic Field, located approximately 60 km from the Providence Mountains field site. The parent material at the Cima site is a lava flow. The clay content at three locations is shown in addition to the results of two infiltrometer experiments. The transect began here in poorly developed pavement, moved onto a plant mound where the penetration depth increased substantially, and then onto a well-developed pavement section, where penetration depth decreased quickly. Large differences in Ks were also observed, with significant reductions in value where tests were conducted on the pavement surface. The results here and at the Providence Mountain site show that the penetration depth, and hence the ability of the GPR to detect subsurface reflectors or targets, is affected by the presence of pavement structure.


Figure 4
View larger version (31K):
[in this window]
[in a new window]
  		Fig. 4. Color-coded map of radar signal penetration depth on an older Qf3 pavement of different parent material. Red stars indicate where soil samples and infiltrometer measurements were made, and percentages are clay content.

 
Qf3 Observations
Table 1 is a correlation matrix containing all components of EC, GPR amplitude, and the pertinent hydraulic properties. Electrical conductivity is primarily a function of clay content, salinity, and water content (McNeill, 1992). Neither the soil salinity nor water content is significantly correlated to the bulk EC, as measured by the EM-38, or the GPR amplitude. Apparently, the low values of salinity (mean <1 dS/m) and water content (mean = 0.014 cm3/cm3) did not influence the GPR and EM instruments. This observation was also reported by Rhoades et al. (1976), who showed that EM measurements are controlled by surface conduction when water contents are <5% for sand and 12% for clay (Rhoades et al., 1976). At our site, nearly all water contents were well below 0.05 cm3/cm3. A significant negative correlation, however, did exist between the early-time GPR amplitude and clay content (r = –0.60, P < 0.0001; Table 1). The increase in clay content caused a rise in the EC of the surface material, dispersing more of the radar signal into near-surface material. Higher clay contents should also increase EC due to a resultant increase in water-holding capacity of the soil; however, Table 1 shows no significant correlation between clay and water content or between water content and EC.


View this table:
[in this window]
[in a new window]
  		Table 1. Correlation coefficients (r) among hydraulic properties, components of terrain electrical conductivity, and GPR (ground penetrating radar) early-time amplitude for the older Qf3 desert pavement.

 
The significant correlation between clay content and bulk ground EC (r = 0.54, P < 0.0001; Table 1), as measured by the EM-38, supports the overall observation that the clay content of the soil was affecting the electromagnetic response of the instruments; however, the correlation between the EM-38 measurements and clay content of the Av horizon is weaker than between the GPR amplitude and clay content. This can be explained by the different sampling volumes between the GPR (shallower penetration) and the EM-38 (deeper penetration). Although most of the EM-38 response was derived from the near surface (i.e., ~60% of response from the top 50 cm of soil), it was also being affected by the bulk EC in soil horizons deeper than 0.75 m (see McNeill, 1980). The GPR interrogates a lesser volume of soil, focusing on the Av horizon. This fact is important because the Av horizon heavily impacts the near-surface water balance.

To evaluate whether the regressions could be improved, we examined the use of clay plus silt as the independent variable, instead of clay content alone. The correlation coefficient (r) increased by 5 and 9% for the GPR and EM regressions, respectively (Fig. 5 ). This increased correlation is most likely because the laser diffraction method for measuring particle size analysis does not distinguish between charged (i.e., clay) and relatively uncharged (i.e., silt) particles. Therefore, some particles >0.002-mm diameter (the silt–clay cutoff using the USDA classification) may be misclassified as silt. By only including the clay-sized fraction (<0.002 mm) of the soil samples in the regressions, there is potential for excluding a portion of charged (clay) materials that could be affecting the electromagnetic responses. Adding back the silt fraction to the textural variable more fully captures the clay mineral content.


Figure 5
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 5. Silt plus clay content of (A) the older Qf3 soil samples vs. average GPR (ground penetrating radar) amplitudes from 0 to 2 ns, and (B) the younger Qf3 soil samples vs. EM (electromagnetic) response.

 
Although the regressions are all significant, considerable uncertainty still exists in the prediction of the fine soil fraction. Using the field-determined degree of pavement development described above, the regression equations were recalculated using the form of Eq. [2], and replotted as Fig. 6A , which is a graph of actual vs. predicted silt plus clay content. The correlation coefficient for this fit is 0.84, a considerable improvement for cases in which only soil texture was used; however, this goodness of fit was impacted by an outlier. This point represents a measurement taken near the edge of the Qf3 surface where the pavement would have probably been more recently disturbed by natural erosive processes. Figure 6B shows the same regression with the site classified as a light pavement instead of a well-developed pavement. The correlation coefficient increased to 0.88. By including the nominal measure of pavement development, the regressions were able to capture an additional 40% of the variability in the fine soil fraction than by using amplitude alone.


Figure 6
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 6. Predicted vs. measured values of silt plus clay content for (A) the older Qf3 surface and (B) the younger Qf3 surface using GPR (ground penetrating radar) amplitudes and a categorical variable describing the degree of pavement development, outlier adjusted. For both figures, x1 = GPR amplitude and x2 and x3 are binary variables related to pavement development. Dotted lines are the predictive interval and dashed lines are the confidence interval, both at 95%.

 
Although the regression results demonstrate that the GPR measurements and simple visual observations of the field soil can provide good estimates of the silt plus clay content, the ability to predict the fine soil fraction based on GPR data measured at a single point is limited. The dotted lines in Fig. 6 represent the predictive error (95%), and the dashed lines are the confidence interval (95%). Though the regression envelope is tightly positioned around the regression line, the prediction envelope is considerably larger, indicating that the ability to predict fine soil texture is rather poor. The predicted single value includes uncertainty not only from the estimates of the regression parameters, but also from the randomness in the value itself.

We next examined the relationship between silt plus clay content and Ks derived from Wooding's (1968) solution. As stated above, a higher early-time amplitude indicates lower clay content of the surficial horizon. Therefore, one would expect a positive correlation between Ks and GPR amplitude if we assume that a higher clay content results in a lower Ks. Figure 7 shows measured vs. predicted log Ks. Thus, on the Qf3 surface, the easily collected GPR measurements could be used to spatially guide more time-intensive point measurements, such as the tension disk infiltrometer.


Figure 7
View larger version (18K):
[in this window]
[in a new window]
  		Fig. 7. Predicted vs. measured values of ln Ks for the Qf3 surface using ground penetrating radar amplitude (x1) and two binary variables (x2, x3) describing pavement development. Dotted lines are the predictive interval and dashed lines are the confidence interval, both at 95%.

 
The support scales of the GPR, EM-38, and infiltrometer methods used in this study are not equal. The GPR has a lateral support scale of ~7 cm, the distance between the transmitter and receiver antennas, and a total depth of investigation of ~10 cm; however, the early-time data are dominated by the direct wave because it is much stronger than any reflected arrivals. Thus, most of the amplitude information is coming from the soil sampled by the direct wave, as calculated using:

Formula 3[3]
(van Overmeeren et al., 1997) where w is the thickness of the soil being sampled, or 4.8 cm in this case, {lambda} is the wavelength (~13 cm), and d is the transmitter–receiver distance (7 cm). The GPR thus investigates a relatively small soil volume, leading to higher spatial resolution. For the EM-38, with a distance of 1 m between the transmitter and receiver coils, the instrument response can be influenced by material at depths >75 cm when operated in horizontal dipole mode, which is the shallowest mode of investigation. The tension infiltrometer, with a 20-cm-diameter disk, has a similar support scale as the GPR. Field experience has shown that on the older Qf3 soils, the wetting front does extend below the surficial Av horizon.

Qf6 Observations and Comparison
Table 2 shows the correlation matrix for the Qf6 surface. An important observation is the general lack of correlation between field measurements and soil properties. Figure 8 shows the measured vs. predicted values of silt plus clay using GPR amplitude and the metric of pavement development on the Qf6. The lack of correlation among GPR, EM, and clay content for the Qf6 in comparison to the highly significant correlations for the Qf3 indicates that a threshold value may exist such that a certain amount of clay is required in the near surface for the GPR method to be useful in the manner described here. The average clay plus silt content from the Qf6 surface (31.8%) is about half that found on the Qf3 surface (58.2%), and the mean EM value for the Qf6 (9.96 mS/m) is one-third the value of the Qf3 (29.4 mS/m). The Qf6 also has a lower coefficient of variation (0.38) than the Qf3 (0.46). Thus, large differences in the textural composition and resulting electrical properties of the near-surface material were observed on these two different-aged desert pavements. These differences also manifest in varying degrees of attenuation. Table 3 shows the mean, standard deviation, coefficient of variation, range, and the results of a t-test of the amplitude values for both surfaces. The higher range and variability of GPR amplitude on the Qf3 could be partially explained by the variability in structure across the transects. On this surface, intact pavements are separated by regions of disrupted pavements, typically found near plant mounds. These mound surfaces are typified by higher bioturbation and soil mixing, which leads to more variability of soil texture. The Qf6 surfaces have not developed intact pavement surfaces per se, and thus the textural variability would be lower. The statistically significant differences in soil structure and texture lead to differences in the GPR amplitude, as from the t-test results in Table 3.


View this table:
[in this window]
[in a new window]
  		Table 2. Correlation coefficients (r) among hydraulic properties, components of terrain electrical conductivity, and GPR (ground penetrating radar) early-time amplitude for the younger Qf6 desert pavement.

 

Figure 8
View larger version (18K):
[in this window]
[in a new window]
  		Fig. 8. Predicted vs. measured values of silt plus clay for the younger Qf6 surface using ground penetrating radar amplitude (x1) and two binary variables (x2, x3) describing pavement development. Dotted lines are the predictive interval and dashed lines are the confidence interval, both at 95%.

 

View this table:
[in this window]
[in a new window]
  		Table 3. Comparison of the mean, standard deviation, coefficient of variation, range, and the results of a t test of early-time GPR (ground penetrating radar) amplitudes at locations on the older Qf3 and younger Qf6 surfaces where soil samples were collected. Ho is the null hypothesis: there is no difference in the means of the GPR amplitude between the two surfaces.

 
The overall low terrain conductivities and the lack of correlation between GPR amplitude and clay content on the Qf6 suggest that the amplitude attenuation is dominated by factors other than conduction losses. The Qf6 surface was observed to have less clay and more loose gravel compared with the cohesive, highly structured peds of the Qf3. This probably resulted in the signal strength being more influenced by typical scattering effects from material heterogeneity than conduction losses.


   CONCLUSIONS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
In this study, we addressed two hypotheses, that (i) the GPR method could be used to estimate the clay plus silt content of desert pavements, and hence provide an indication of the soil saturated hydraulic conductivity, and (ii) the efficacy of the method does not depend on the soil age. The GPR method was used in this study because it provides much higher spatial resolution than other commonly used methods (specifically the EM-38) and can be equipped with antenna designed to interrogate shallower soils. The results supported the first hypothesis, in that the GPR amplitude was effective at predicting the clay plus silt content of the desert soils; however, a considerable amount of uncertainty existed in the predictive capabilities of the method. We then improved the predictive capacity by including a visual observation of the degree of pavement development. Although 77% of the variability in silt plus clay content on older soils was accounted for, point estimation of texture obviously should be made with some caution.

The results of the research did not support the second hypothesis, and suggest that the GPR method is less effective in rapidly estimating soil hydraulic properties in younger desert soils with lower clay plus silt content. The field results on the younger, Qf6, soils showed essentially no correlation between GPR amplitude, texture, or salinity.

Although the GPR method as a quantitative measure of silt plus clay content—and soil hydraulic conductivity—may be limited, it can be used effectively as a reconnaissance tool in some environments. The GPR method worked effectively on the well-developed desert pavement that was sampled. In similar environments, this approach could probably help to identify clay-rich areas, potential zones of lower hydraulic conductivity, and reduced soil infiltration. Furthermore, because clays can reduce infiltration and recharge (Cook et al., 1992; Gee et al., 2005), this method could be used to rapidly locate areas that would more likely experience deep percolation. In this way, a limitation of GPR, namely high attenuation in conductive environments, was converted into a useful diagnostic tool.

It is important to note that some variability in these measurements could be a result of variations in antenna contact with the ground. Even though the fixed-width antenna has advantages over the variable width, the ruggedness of the desert terrain caused by large cobbles, surface undulations, and vegetation undoubtedly causes the antenna-to-ground contact to be inconsistent despite efforts to avoid such problems. This probably reduces the data quality. A corollary to this is that GPR amplitudes may better predict Ks on less rugged surfaces.

Our results show an apparent threshold value of soil development (as indicated by clay plus silt content) for this method to be successful. At the site used for this research, the minimum clay plus silt value sampled on the Qf3 and Qf6 surfaces were 34.4 and 4%, respectively. Thus, the clay content on the Qf6 surface may not be high enough to cause significant surface conduction. The low EM readings on the Qf6 surface support the limited effect of surface conduction on this surface. Future work will focus on defining the threshold value of electrical conductivity and the role of soil structure for the successful application of the GPR method in desert environments.


   ACKNOWLEDGMENTS
 
Funding for this work was provided by the Department of the Army, Army Research Office (DAAD19-03-1-0159). The content presented here does not necessarily reflect the position or the policy of the federal government, and no official endorsement should be inferred. Thanks to Todd Caldwell, John Goreham, Li Chen, and Jun Yin for field assistance.


   REFERENCES
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Meadows, D. G.
Right arrow Articles by McDonald, E. V.
Related Collections
Right arrow Dryland Soils
Right arrow Field-Scale Studies
Right arrow Ground Penetrating Radar, GPR

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome