| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a USDA-ARS Hydrology and Remote Sensing Laboratory, Animal Natural Resources Inst., Beltsville, MD 20705
b Univ. of Wisconsin, Madison, WI 53706
c USDA-NRCS, Newtown Square, PA 19073
d Advanced Geological Services, Malvern, PA 19355
* Corresponding author (tgish{at}hydrolab.arsusda.gov)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Abbreviations: DGPS, differential global positioning system • EM, electromagnetic induction • GPR, ground-penetrating radar • PVC, polyvinyl chloride
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Theories have been developed and tested to describe water infiltration and chemical transport through matrix pores in a homogenized soil profile (Rawls et al., 1993), yet most soils have distinct layers in the vadose zone. Lack of proper technologies to monitor water infiltration, percolation, and chemical transport in soil profiles with distinct layers has limited our ability to accurately determine subsurface water or chemical fluxes on a field scale. Recently, Kung et al. (2000a)(b) demonstrated that monitoring a flux was essential for improving our understanding of how complex flow patterns influence chemical transport. It follows that subsurface water and chemical fluxes cannot be accurately determined without first characterizing soil water movement at the field scale. If water quality research is to be more applicable to agricultural production, a better understanding of field-scale subsurface soil water dynamics is needed.
Accurate estimates of field-scale subsurface water movement and chemical fluxes depend on a detailed knowledge of the spatial distribution and autocorrelation of soil hydraulic properties and subsurface soil layering structures. Traditional methods to assess the spatial nature of soil hydraulic properties include collection and analysis of soil core and well log data (Sudicky, 1986; Ritzi et al., 1994). These methods are of limited benefit because only a fraction of the subsurface is sampled. It is virtually impossible to ascertain the spatial behavior of soil hydraulic properties using point data because the sampling density of soil core and well log data is usually considerably below the inherent spatial variability of soil hydraulic properties. As a result, uncertainty associated with interpolating soil core and well log data to areas where no samples were acquired can be significant.
Soil layer properties have significant impact on water movement and chemical transport because abrupt changes in texture or density across the boundary of two adjacent layers causes a discontinuity of soil pores. The mismatch of both pore entry value and soil hydraulic conductivity across this boundary can trigger funnel flow (Kung, 1990a, 1990b, 1993; Ju and Kung, 1993). Under this condition, a uniform matrix flow could be congregated and become preferential flow, especially when the subsurface restricting layer is inclined.
To determine total field-scale subsurface flux, matrix and preferential flow components must be accounted for. Matrix flow is associated with water movement through pores among primary soil particles and is controlled by the matric potential gradient. Conventional instruments such as soil water sensors, tensiometers, and soil cores can be readily used to quantify subsurface flux through matrix pores. Gravitational forces, on the other hand, govern water movement when preferential flow is the dominant process. Because of the dynamic nature of preferential flow, both spatially and temporally, no current technology for directly mapping funnel-type preferential flow paths exists.
Water percolation is influenced by soil layer properties, causing a distinct layering of soil water above the impeding layer (Hill and Parlange, 1972) which can be formed by a clay layer, dense till, or sand lens. Although no single instrument can nonintrusively map locations where funnel flow is initiated, both electromagnetic induction (EM) methods and GPR can detect changes in dielectric properties (Sheets and Hendickx, 1995; Sandberg and Slater, 1999). Dielectric properties of soil are strongly affected by water content with the dielectric constant of dry soil between 3 and 5 and that of water around 80. Recently, multifrequency EM sensors have been used to characterize subsurface conductivity distributions (Zhdanov et al., 1996). Unfortunately, the resolution of EM is too poor to discern the depth of soil layering. However, Rea and Knight (1998) found that GPR images contained information about the spatial continuity of coarse and fine grained beds in sedimentary deposits. Kung and Lu (1993) and Casper and Kung (1996) also showed that GPR can detect the size, inclination, and spatial pattern of subsurface layers. This suggested that subsurface convergent flow pathways could be identified and monitored. This would eventually allow accurate quantification of fluxes of water and solutes leaving the root zone, including matrix and preferential flow components.
In plot-scale experiments, Kung and Donohue (1990) used GPR to reveal textural discontinuities to install suction lysimeters along potential preferential flow pathways. They reported that solution samples collected from suction lysimeters located along predicted preferential flow pathways had greater solution volumes. This suggested that samplers installed near preferential flow pathways were in a wetter region and perhaps an area of greater water flow compared with lysimeters installed outside GPR-identified flow pathways. Moreover, suction lysimeter samples from preferential flow pathways had chemical concentrations that were almost 400% greater than samples obtained in areas where matrix flow was expected to be the dominant subsurface flow process.
The objective of this investigation was to develop techniques that provide a continuous, nondestructive image of field-scale subsurface stratigraphy which could be used to map the depth to and lateral extent of subsurface layers that control subsurface flow paths. Ground-penetrating radar data was collected, analyzed, and an evaluation performed on how to best supplement these data with other analyses in order to characterize subsurface flow pathways.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
|
Electromagnetic induction (EM-38) was used to estimate relative clay contents near the soil surface to estimate infiltration rate (Doolittle et al., 1994). The results showed that the EM-38 values ranged from 5 to 30. The EM data were divided into three equally sized populations, low (EM < 12), intermediate (17 
EM 12), and high (EM > 17) values.
The entire 7.5-ha site was then divided into different hydrologic groups based on surface topographic features and EM data. We assumed that those areas with EM > 17 and slope > 2% would have a relatively low infiltration capacity, while those areas with EM < 12 and slope < 1% would have relatively high infiltration capacities. All other areas were assumed to have medium infiltration capacities. Within each hydrogeologic region of potential infiltration, 7 or 8 plots (22 in total) were randomly selected (Fig. 1) . Boundaries for each of these 25- by 25-m (0.06-ha) plots were established with a submeter DGPS receiver. The plots were evaluated with detailed GPR transects (2 by 2 m), and subsequently instrumented with soil moisture capacitance probes.
|
0.8 m, and finally a loamy sand C horizon that continued until the core ended, 1 m. Soil physical data from the installation of soil moisture probes is limited, but auger samples were retrived as deep as 2.5 m. In general, these samples supported the GPR data, and indicated that a finer textured horizon (typically, a clay loam lens) was located 1.3 to 2 m below the soil surface.
Ground-Penetrating Radar Data
A subsurface interface radar system-2 (Geophysical Survey Systems, Inc., North Salem, NH) was used. Two GPR calibration sites were established, one at the top of the watershed, and the other near the bottom so that the range of soil subsurface profiles found at the experimental field site would be represented. Each calibration site consisted of a trench (1.5 m wide by 4.0 m long by 2.5 m deep) in which five metal plates (0.4-m diameter) were laterally inserted into the trench sidewalls spaced 0.75 m apart horizontally at depths of 0.50, 0.75, 1.00, 1.25, and 1.50 m (soil profile overlaying plates remained undisturbed). Ground-penetrating radar antennas with frequencies of 150, 300, and 500 MHz were dragged across the soil surface above the calibration plates on both sides of the trench. Return times of the GPR signals from the steel plates were used to calculate an integrated soil dielectric constant. Of the three antenna frequencies, the 150-MHz antenna was selected because it provided the best combination of resolution and depth penetration resulting in superior subsurface images.
Ground-penetrating radar data were acquired for the entire 7.5-ha site along parallel north-south transects 25-m apart (Fig. 1). Within selected 25- by 25-m plots, additional GPR data were collected by towing the 150-MHz antenna along north-south transects that were 2 m apart. GPR data were acquired in digital format so that a trace of subsurface reflections could be produced using RADAN software (Geophysical Survey Systems). Prior to data interpretation, GPR data were distance normalized (conformed to known surface distances) and processed through a low pass filter to accentuate reflections in the soil profile image. The GPR trace followed the shallowest contrasting dielectric discontinuity. Strong dielectric reflections were considered to be a manifestation of water holding capacity differences due to textural discontinuities such as a clay lens under a sandy soil. Generally, the clay lens (high dielectric) occurred below the C horizon, which frequently contained large gravel (low dielectric). Depth to the strongest reflection was as shallow as 0.9 m and as deep as 3.4 m, but the majority of the data gave the strongest reflection at depths between 1.3 and 2 m. A GPR image profile is shown in Fig. 2 . In this figure, the first continuous restricting layer lies just above the first continuous strong reflection, shown here as a dotted line located between 1- and 2-m depths.
|
GEO-EAS (EPA, Las Vegas, NV) and GS+ (Gamma Design Software, Plainwell, MI) geostatistical software packages were used to determine the spatial autocorrelation of the depth to the first continuous restricting layer. These programs were used to produce omnidirectional semivariograms from point data derived from digitized traces (i.e., depth to the first continuous restricting layer and its associated geographic coordinates). Semivariogram models, which were fit using a least squares approach, provided kriging parameters (i.e., nugget, range, and sill) for subsequent spatial interpolation. As a result, contour and three-dimensional surface maps of the depth to the first continuous restricting layer were produced.
Soil Moisture Sensors
Soil moisture sensors were installed to independently determine how subsurface restricting layers detected by GPR influence subsurface water flow pathways. Capacitance probes (EnviroSCAN, SENTEK Pty Ltd., South Australia), were used to measure volumetric water contents within a 10-cm radius from the sensor's center (Paltineanu and Starr, 1997). Recently, Paltineanu and Starr (2000) demonstrated the ability of this capacitance probe system to monitor spatial and temporal changes of moisture within the soil profile.
To install soil moisture capacitance probes, 22 plots (25 by 25 m) were evaluated geostatistically according to average depth from the soil surface to the first GPR-detected continuous restricting layer. Each soil moisture probe was configured with either 3, 6, or 7 sensors at varying depths depending on EM data and depth to the first continuous restricting layer at each plot. The three-sensor probes had sensors at depths of 0.1, 0.3, and 0.8 m. They were inserted in areas where estimated infiltration capacity was low (L series in Fig. 1), and depth to the first continuous restricting layer was generally <1.5 m from the soil surface. The seven-sensor probes had sensors at depths of 0.1, 0.3, 0.5, 0.8, 1.2, 1.5, and 1.8 m, and were inserted in regions where the estimated infiltration capacity was high (H series in Fig. 1) and the depth to the first continuous subsurface restricting layer was >1.5 m. The six-sensor probes had sensors at depths of 0.1, 0.3, 0.5, 1.2, 1.5, and 1.8 m, and were inserted whenever the depth to the first continuous restricting layer was between 0.9 and 1.5 m and was identified as having an intermediate infiltration capacity (M series in Fig. 1). Each sensor was calibrated before installation and programmed to record one volumetric water content reading every 10 min.
A 6-m high tripod drilling rig with vertical leveling capability was used to install the soil moisture probes. It held the polyvinyl chloride (PVC) access pipe with its steel liner while a soil auger with a tungsten tip was inserted through the liner and pipe to remove a slightly undersized core of soil. The PVC pipe with an attached inward-tapered metal cutting edge was continuously pushed into the soil as the auger removed soil until the desired depth was reached. The steel liner was then removed, the inside of the PVC pipe was cleaned, and the bottom of the PVC pipe was sealed with a compression rubber plug. The capacitance sensors were mounted on a plastic extrusion for placement at specific soil depths, inserted into the PVC pipe, and connected by cable to a data logger.
Scale Verification
Large (25-m spacings) and small (2-m spacings) sample grids were used to collect GPR data on a 0.5-ha subsection of the watershed (eight shaded 25- by 25-m plots in Fig. 1). These data were used to evaluate GPR's ability to identify subsurface flow pathways. This intermediate-sized area was chosen because it provided a manageable number (5) of soil moisture probes to evaluate and a reasonable spatial scale for an initial assessment of field-scale subsurface water flow pathways.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
|
An omnidirectional semivariogram of the depth to the first continuous restricting layer for a 25- by 25-m plot (Plot C03) is shown in Fig. 3 . The spherical model shows little nugget effect, has a range of 16 m, and a sill of 0.05 m2. These parameters were used in ordinary kriging to produce an interpolated contour map (Fig. 4a) and surface representation (Fig. 4b) of the restricting layer depth for Plot C03. Visual representation of the subsurface layer shows that even though a given 0.06-ha plot may have been classified at a particular infiltration capacity, GPR data was essential for locating the soil moisture monitoring sites (i.e., a function of the depth to the first continuous restricting layer). Although it is instructive to understand the spatial behavior of subsurface restricting layers at a relatively small scale (25 by 25 m), a larger scale of observation was required to properly assess the use of GPR to identify flow pathways at the watershed scale.
|
|
|
|
Data acquired by soil moisture capacitance probes from 7 May to 14 July 1999 are presented to confirm GPR-identified features that might influence subsurface water movement (Fig. 7) . This particular time frame was selected because it avoids confounding effects of surface freezing during the winter months, and the heaviest period of water extraction by the overlying corn canopy in mid- to late summer (i.e., from the start of tasseling through grain fill).
|
Figure 6 indicates that soil moisture probe BH4 is above an essentially flat restricting layer or clayey lens at a depth of 1.5 m. The deepest sensor activated on probe BH4 is 0.8 m below the soil surface. Volumetric water contents at 0.1 and 0.3 m were similar while volumetric water content at 0.8 m was always lower than that at 0.3 m (Fig. 7A, bottom). Similar to soil moisture dynamics at probe BH3, matrix flow at BH4 does not collect above the flat restricting layer, and, except for three brief surface dry-down periods, volumetric water contents decrease with depth.
If water contents between probes BH3 and BH4 are compared, it is possible to see evidence of matrix flow. While volumetric water contents at 0.1 m were similar for both probes, water contents for the sensors at depths of 0.3 m and 0.8 m were 50 and 45% greater, respectively, at probe BH4 than at probe BH3. This is indicative of matrix flow following topographic gradients (i.e., BH4 is downslope from BH3). Fortunately, soil water content differences were not due to surface runoff because this year (1999) was a severe drought year and no runoff occurred during the growing season. In addition, differences in soil moisture are not likely to be due to soil texture differences as the soil core taken closest to BH3 show a 24-cm Ap horizon with a sandy loam texture (73% sand, 20% silt, and 7% clay) while the soil core extracted within a few meters of the BH4 probe shows a 27-cm Ap horizon with the same texture (70% sand, 22% silt, and 8% clay).
Soil moisture probe BL4 is above the inside lip of a large concave restricting layer (Fig. 6). The deepest sensor activated on this probe is at 1.8 m; the restricting layer below the probe is at 2.3 m. Volumetric water contents at 1.2, 1.5, and 1.8 m deep show consistent increases in soil water content with depth (Fig. 7B). This soil moisture profile may be explained by the interpretation that water accumulates in the GPR-identified, bowl-shaped restricting layer beneath probe BL4. Differences become more apparent as the soil core taken near BL4 shows a sandy loam texture down to 1 m (the maximum depth of the soil core at that location). However, observations from the soil moisture probe installation suggests that between 1.75 to 2.0 m the subsoil becomes a sand.
There is evidence of matrix and preferential flow at soil moisture probe BM3. Figure 6 shows that the probe was installed where GPR data identified an essentially flat restricting layer located at a depth of 1.54 m. Matrix flow is suggested because soil water contents increase from sensor to sensor with depth (Fig. 7D). Volumetric water content data from the sensor located immediately above the restricting layer (1.5 m) show evidence of a funnel flow pulse which lasts for several weeks. This pulse was the result of lateral groundwater movement because there is no evidence of vertical groundwater movement which could feed the pulse (Fig. 7D). To our knowledge, this is the first time funnel flow along a restricting layer has been documented in real-time.
Figure 6 indicates that an essentially flat restricting layer intercepts soil moisture probe BM4 at a depth of 1.5 m. A sensor located at this depth consistently had volumetric water content measurements greater than sensors either above (1.2 m) or below it (1.8 m) (Fig. 7C). This is evidence of a fine-textured soil at a depth of 1.5 m with a greater water holding capacity than soils adjacent to it.
Subsurface Flow Network
Subsurface flow dynamics were captured with a network of soil moisture capacitance probes and provided corroboration of the use of GPR images to identify subsurface features that control subsurface water movement. Because soil moisture data demonstrated that local subsurface stratigraphies influenced soil water, subsurface flow pathways will also be influenced by local differences in subsurface stratigraphies. However, before potential subsurface flow pathways could be determined from the contour map of depth to the first continuous restricting layer, the effect that slope and aspect of the surface topography had on the spatial orientation of the subsurface restricting layer had to be accounted for. After the geostatistical analysis was completed, the surface and subsurface topographies were divided into 10- by 10-m cells. A first approximation of the subsurface flow pathways were constructed by subtracting the depth to the first continuous restricting layer from the surface elevation. The Arc/Info GIS hydrologic modeling tools FLOWDIRECTION and FLOWACCUMULATION were applied to a raster grid of the elevation-corrected subsurface topography to determine potential flow pathways (Fig. 8)
. The FLOWDIRECTION routine provides a grid of flow directions from one cell to its steepest downslope neighbor, while FLOWACCUMULATION determines the accumulated water from all cells that flow into each downslope cell.
|
Although no distinct water pulse is observed at BM4, water contents are consistent with water flowing along the clayey lens interface. The moisture sensor at 1.5 m is consistently greater than the ones at the other two depths, even though the 1.8-m sensor is located within the same clay loam texture (as determined from GPR and soil auger data). Additionally, GPR data at this location indicate a restricting layer at 1.48 m, and so the 1.5-m soil moisture sensor should be centered just below the clayey interface. As a result, rapid changes in water content should be dampened by the high water holding capacity of the clay lens since it would dominate the 10-cm area of influence monitored by the 1.5-m sensor. In addition, since the BM4 probe is located between two subsurface flow pathways, water plumes from the two GPR identified flow pathways could be moving through that region at different times, thereby dampening each other out. As a consequence, although no distinct soil water plume was observed, GPR and soil moisture data supported the possibility of water flow along the clayey lens interface.
Spatial Scale
To assess how the spatial autocorrelation of the depth to the first continuous restricting layer changed as the spatial scale of observation is changed, the semivariogram of GPR data collected on transects with 25-m spacings over the entire watershed is presented (Fig. 9)
. This semivariogram was best fit with a spherical model having a nugget of 0.02 m2, a range of 95 m, and a sill of 0.18 m2. With increasing scale of observation, from block or plot to watershed subsection to small watershed, consistent and predictable trends in semivariogram model parameters are seen. Nugget values increased, implying that it is more difficult to capture small scale variability; sill values increased, which reflects the increase in population variance; and range values increased as distances with which data exhibit spatial autocorrelation increased. This suggests that although 2- by 2-m transects were effective for evaluating subsurface soil water dynamics, large scale GPR data would be needed to understand the uncertainty (variance) of the depth to the first restricting layer on a watershed scale.
|
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
In summary, soil moisture data coupled with GPR-identified flow pathways suggest that: (i) a coupling of GPR data with real-time soil moisture monitoring may be an effective tool for evaluating and monitoring subsurface flow processes; (ii) the spatial location of the soil moisture monitoring system is critical to monitoring water movement; and (iii) real-time monitoring of water movement is critical if preferential flow pathways are to be accurately monitored.
This study demonstrates that georeferenced GPR data sets have great potential to locate soil layers which control subsurface water flow. Soil moisture sensors confirmed the existence of funnel flow processes, which indirectly confirmed the existence of restricting layers. These techniques may have the capacity to monitor and evaluate subsurface water pathways which are necessary to determine agrichemical fluxes beyond the root zone.
|
ACKNOWLEDGMENTS |
|---|
The authors would also like to thank the assistance and contribution from Andy Russ, Lynn McKee, Dan Shirley, Galen Hart, Peter Buss, Dale Hardin, Phil Nedeau, Peter Tucker, James Starr, and Rob Parry.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Received for publication January 9, 2001.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSNOTESRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. P. Oden, G. R. Olhoeft, D. L. Wright, and M. H. Powers Measuring the Electrical Properties of Soil Using a Calibrated Ground-Coupled GPR System Vadose Zone J., February 25, 2008; 7(1): 171 - 183. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Cockx, M. Van Meirvenne, and B. De Vos Using the EM38DD Soil Sensor to Delineate Clay Lenses in a Sandy Forest Soil Soil Sci. Soc. Am. J., June 29, 2007; 71(4): 1314 - 1322. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. J. Gish, C. L. Walthall, C. S. T. Daughtry, and K.-J. S. Kung Using Soil Moisture and Spatial Yield Patterns to Identify Subsurface Flow Pathways J. Environ. Qual., January 1, 2005; 34(1): 274 - 286. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Vadose Zone Journal | Journal of Plant Registrations | ||||
| Journal of Natural Resources and Life Sciences Education |
Journal of Environmental Quality |
||||
