VZJ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

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 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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Serbin, G.
Right arrow Articles by Or, D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Serbin, G.
Right arrow Articles by Or, D.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Serbin, G.
Right arrow Articles by Or, D.
Related Collections
Right arrow Water Content
Published in Vadose Zone Journal 2:500-510 (2003)
© 2003 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA

SPECIAL SECTION - ADVANCES IN MEASUREMENT AND MONITORING METHODS

Near-Surface Soil Water Content Measurements Using Horn Antenna Radar

Methodology and Overview

Guy Serbin*,a and Dani Orb

a Dep. of Plants, Soils, and Biometeorology, Utah State Univ., Logan, UT 84322-4820, USA
b Dep. of Civil and Environmental Engineering, Univ. of Connecticut, 261 Glenbrook Road, Unit 2037, Storrs, CT 06269-2037
* Corresponding author (gserbin{at}mendel.usu.edu).

Received 1 April 2003.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Ground-penetrating radar (GPR) with a suspended 1-GHz horn antenna was deployed over bare and vegetated soil surfaces using surface reflection (SR) magnitudes and propagation times (PT) to calculate bulk soil dielectric permittivity and soil water contents. Concurrent radar measurements over sand, Millville silt loam (coarse-silty, carbonatic, mesic Typic Haploxerolls), and sand–bentonite surfaces showed rapid drainage from sand and slower drainage from higher-surface-area textured soils. Soil texture and temperature affected diurnal variations in measured water content (occurrence of minima and maxima) for both SR and 2-cm time-domain reflectometry (TDR) water content data. Measurements over wheat canopy showed that while SR values were strongly altered by canopy biomass, PT measurements remained unaffected. Wheat canopy influence on SR gradually intensified during the growth season until the canopy was removed and SR-based measurements rejoined with PT data. Hornantenna radar measurements over natural surfaces offer a promise for remote mapping of soil texture and truthing of radar data collected from air- and spaceborne platforms, and they may be used in the field for water content and vegetation biomass measurements.

Abbreviations: EC, electrical conductivity • GPR, ground-penetrating radar • LAI, leaf area index • PT, propagation time • SAR, synthetic aperture radar • SR, surface reflection • TDR, time-domain reflectometry


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
TIME- AND FREQUENCY-DOMAIN electromagnetic methods are commonly used for soil water content measurement in laboratory and field applications. These often employ in situ (TDR) and remote sensing (radar) methods. Common to these methods is the determination of bulk soil dielectric permittivity ({epsilon}b), from which the volumetric water content ({Theta}v) is inferred. Topp et al. (1980) introduced the TDR method for measurement of volumetric water content in porous media using signal propagation time (tp) along a transmission line (probe) with a known travel path of length L, capitalizing on the strong dependence of the travel time upon {epsilon}b (von Hippel, 1954; Birchak et al., 1974; Wang et al., 1978; Topp et al., 1980, 1988; Dasberg and Dalton, 1985; Dobson et al., 1985; Or and Wraith, 1999). Frequency-domain methods, such as most narrowband air- and spaceborne radar remote sensing methods, determine {epsilon}b by measuring the SR magnitude in conjunction with surface geometric and biophysical properties (Dubois et al., 1995; Ulaby et al., 1996; Blumberg and Freilikher, 2001; Blumberg et al., 2002). The field portability of TDR and frequency sensitivity around 1 GHz [similar to many L-band radar sensors (Heimovaara, 1994; Friel and Or, 1999; Or and Rasmussen, 1999)] has made it a popular tool for ground validation of radar remote sensing (Schmullius and Furrer, 1992; Dubois et al., 1995; Blumberg and Freilikher, 2001; Blumberg et al., 2002).

Radar remote sensing allows for high resolution (up to 1 m) mapping of soil water content and vegetation biomass from both air- and spaceborne synthetic aperture radar (SAR) and scatterometer systems (Dobson and Ulaby, 1986a; 1986b; Mihaly and Bozsoki, 1988; Oh et al., 1994; Dubois et al., 1995; Ulaby et al., 1996; Le Hegarat-Mascle et al., 2002; Njoku et al., 2002). Furthermore, these methods allow for data acquisition at any time of day or night, irrespective of weather conditions. These measurement methods, however, are sensitive to surface and subsurface geometric features, dielectric properties, and vegetation cover, and average information from a profile to a single value (Dobson and Ulaby, 1986b; Ulaby et al., 1996). Moreover, measurement intervals for most radar remote sensing platforms range from hours to days, which precludes detailed observation of highly dynamic near-surface hydrological processes.

Time-domain GPR technology has been used for measurements of soil water content from a variety of configurations, including aboveground air-coupled (suspended antenna) and ground-coupled (antennas are placed in direct contact with ground) measurements (Chanzy et al., 1996; van Overmeeren et al., 1997; Weiler et al., 1998; Huisman et al., 2001; Redman et al., 2002; Serbin and Or, 2003, unpublished data), and cross-borehole (Binley et al., 2001; Alumbaugh et al., 2002; Rucker and Ferré, 2003) configurations, with results showing good correlation with TDR and gravimetric measurements (van Overmeeren et al., 1997; Weiler et al., 1998; Huisman et al., 2001; Serbin and Or, 2003, unpublished data). Chanzy et al. (1996) and Redman et al. (2002) showed that aboveground air-coupled GPR could be used for measurement of surface soil water content. Redman et al. (2002) showed that this technique could be used for measuring near-surface water contents of transects by using surface reflectivity measurements and had an accuracy of 2% in water content measurement. Ground-coupled and cross-borehole GPR measurements are acquired via bistatic antenna configurations (i.e., transmit and receive antennas), whereas air-coupled measurements can utilize both bistatic and monostatic (the same antenna acts as both to transmit and receive signals) configurations. Bistatic antenna configurations allow the operator to alter the configuration geometry, but can suffer from crosstalk effects, whereby the signal is transmitted directly from one antenna to another (Chanzy et al., 1996). Monostatic configurations have the distinct advantage of no crosstalk since the same antenna transmits and receives.

Serbin and Or (2003, unpublished data) showed that diurnal measurement of near-surface water content dynamics using a GPR with a 1-GHz horn antenna suspended over the surface provided a detailed picture of evaporation and drainage processes based on accurate water content measurement in a frequency range similar to both TDR and L-band SAR systems. The radar setup used in the study reported herein is a commercially available remote platform with continuous temporal measurement capabilities that could easily be validated by a variety of means. This capability circumvents some of the obvious limitations of space- and airborne SAR and scatterometer systems that either suffer from poor temporal resolution (35 d) or require expensive flight time. Although scatterometers provide similar measurement capabilities in the frequency domain, these systems must be custom manufactured and calibrated.

Since all aboveground air-coupled GPR measurements of soil water content currently in the literature involved bistatic dipole or bowtie antenna configurations (Chanzy et al., 1996; Redman et al., 2002) which suffer from crosstalk, and since these systems use an unfocused radiation pattern that can easily be affected by aboveground objects (Chanzy et al., 1996), we deemed it necessary to develop a methodology for GPR measurement of soil water content dynamics with a horn antenna, which provides a ground footprint of known area. Furthermore, a time-domain radar system would allow for the study of discrete scattering elements in a vegetation canopy, something that is not possible with frequency-domain analyses.

Here, we provide an overview of a methodology for using measurements from GPR with a horn antenna to study (i) concurrent drying and wetting patterns from both bare and vegetated soil surfaces, (ii) comparability of radar with TDR and gravimetric measurements, (iii) canopy biophysical parameters via radar, and (iv) effects of vegetation canopy on radar measurements of soil water content. For brevity, we focus on key aspects of the methodology and on main results, and refer interested readers to the supplemental reference for additional details.


   THEORETICAL CONSIDERATIONS AND METHODOLOGY
TOP
 ABSTRACT
 INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Instrumentation and Measurement System
A Penetradar IRIS-L GPR unit (Penetradar Corp., Niagara Falls, NY) equipped with a monostatic horn antenna was used to acquire measurements in both the field and greenhouse. The GPR unit consists of a radar control unit which connects to the antenna via cable. The antenna, which is typically used for high-speed profiling of roadways and bridges, is encased in polyurethane and polyethylene for protection from the elements, and utilizes a monocycle signal with a center frequency of 1.025 ± 0.025 GHz (with frequency components extending from below 50 MHz and up to 2.5 GHz), a pulse width of 1 ns and an accuracy of 1% in both signal amplitude and time. After travel through the system, the pulse may be described as a Ricker wavelet (Ricker, 1953; Sneddon et al., 2002) with a width of about 2 ns.

The GPR unit acquires waveforms consisting of 1600 data points, each denoting a 0.025-ns interval for a total waveform length of 40 ns. The waveform can be divided into three regions: antenna, air, and surface, as depicted in Fig. 1A. Antenna reflections are due to the antenna and enclosure impedance mismatches, followed by signal travel in air (Ulaby, 1999). Surface reflection and subsurface reflections occur at dielectric discontinuities between the air and soil and within the soil, respectively. The borders between these regions tend to be diffuse as the wavelet is about 2 ns wide, with physical boundaries occurring at the center peak of the wavelet. Wavelet overlaps mask the exact location of the peak at the end of the antenna reflection; however, the following approximation (ns) was found to be in excellent agreement with physical distance measurements:

[1]
where t(n) denotes the time along the waveform of reflection number n as depicted in Fig. 1A, with (1) being the second to last antenna minimum, (2) the last antenna maximum, and (3) the last antenna minimum. The diffuse nature of these boundaries means that the last antenna (3) and first soil surface (4) reflections fall into the air region. Furthermore, the near-field antenna region (which is proportional to frequency) extends beyond the antenna itself up to a distance of about 67 cm at 2.5 GHz (26.7 cm at 1 GHz) from the antenna, as denoted by (NF) in Fig. 1A. The near field is comprised of two zones, namely reactive and radiative (Balanis, 1982). Within the reactive zone, which is most proximal to the antenna, reflections can influence or react with antenna impedance characteristics, and for the 30AGC antenna, this extends to 0.16 m at 2.5 GHz (0.1 m at 1 GHz). Beyond that exists the radiative near-field zone, in which the electrical field is not assumed planar (Balanis, 1982). To correct for near-field effects, a scaled free-space waveform is subtracted from the measured data. The SR is approximately 2 ns wide, with significant reflections occurring between (4) and (6), where the actual interface occurs at (5). A subsurface reflection may be seen at (7). The asymmetric nature of the Ricker wavelets between the leading and trailing sides that comprise the surface and subsurface reflections are most likely because of frequency-dependent differences in dielectric properties over the bandwidth of the GPR device (Heimovaara, 1994; Friel and Or, 1999; Serbin, 2001).



View larger version (39K):
[in this window]
[in a new window]
  		Fig. 1. A ground-penetrating radar (GPR) waveform of a wheat canopy overlying a wet soil. (A) Physical boundaries are denoted by dash-dot lines, diffuse boundaries of air–soil interface are shown by dashed lines, and the border of the near-field antenna range (NF) by dash-dot-dot lines. Antenna reflections of interest are (1–3), soil surface (4–6), and subsurface at (7). (B) Reflection magnitudes and distances. Vant and Vsurf denote antenna and surface voltages, respectively; tair and tp denote propagation times between the antenna and soil surface and within soil between surface and subsurface reflections, respectively. Canopy reflections (CR) are shown as small aboveground reflections in air region.

 
In some cases, we were able to obtain two types of radar water content measurements: one based on SR, and the other from signal PT, based on features shown in Fig. 1B and using Eq. [5] for SR, and Eq. [6] for PT (see below). The SR measurements were simpler to implement and similar to other standard remote sensing radar applications, but required voltage amplitude corrections relative to a calibration measurement. In contrast, PT measurements required prior knowledge of medium thickness or depth to a strong reflective interface beneath the soil layer of interest, but required no amplitude correction. The SR and PT measurements provided two different types of information: SR focused on the surface skin of the soil, whereas PT integrated dielectric properties (and water content) along the signal travel path as marked by two reflections (similar to analyses used in standard TDR methods). Measurements by Serbin and Or (2003, unpublished data) showed that SR reflections agreed with gravimetric water contents from the top 1 cm of the soil but not with measurements from a 2-cm-deep horizontal TDR probe or 1- to 5-cm depth gravimetric measurements (TDR and 1- to 5-cm gravimetric measurements did agree with each other), confirming that the SR responded to water content at the soil surface skin. Both measurement methods determined bulk soil dielectric permittivity, {epsilon}b, which was used to estimate soil water content {Theta}v (m3 m-3) via the Topp et al. (1980) relationships:

[2]

These relationships do not account for changes in soil temperature. Furthermore, they assume that soil texture has no effect upon soil dielectric properties.

Surface Reflection Measurements
Surface reflection measurements utilize the voltage magnitude of the SR, Vsurf, relative to that of a perfectly reflective (i.e., metal, or flat plate) surface, Vfmp, at the same distance from the antenna to determine the surface's reflection coefficient, {Gamma}(t), and corrected for temperature-induced changes in antenna impedance (which affect the magnitude of the measured SR) and distance discrepancies between the antenna and surface (which also cause differences in measured voltage due to air spreading losses):

[3]
where ACF denotes the antenna impedance correction factor and is the ratio of the flat plate calibration antenna voltage to that of the voltage of the waveform of interest. Flat plate measurements were acquired at the end of an experiment by covering the soil surface with thick aluminum foil whose area well exceeded that of the ground radiation footprint and then acquiring a series of waveforms. The Vant, Vsurf, and Vfmp (V) are determined from the voltage magnitudes of points (4–6) as seen in Fig. 1A and B:

[4a]

[4b]
where V(n) is the voltage of location n. The correction factor and choice of antenna reflections were determined from analysis of 2 d of measurements over a metal surface.

For measurements collected within the near-field range (<53 cm from the base of the antenna used in this study), a scaled free space correction was also required for surface reflectivity measurements, whereby a waveform that was collected with the antenna pointed skyward was subtracted out of the sampled waveform. The reflection coefficient may be used to determine {epsilon}b at normal incidence via (Ulaby, 1999):

[5]
and the volumetric water content as a function of {epsilon}b may then be determined via the Topp et al. (1980) relationships (Eq. [2]).

Propagation Time Measurements
Radar PT measurement of soil water content is possible when a subsurface reflection exists at a known depth, such as an interface between two different soil types (i.e., sand overlying a silt loam), between a soil and a metal (silt-loam overlying a layer of aluminum foil) or between a sand and the water table (Smith et al., 1992), all of which would be denoted by a strong dielectric discontinuity. If the distance between two reflections is known, then the two-way propagation time between them, tp, as illustrated in Fig. 1B, can be used to calculate {epsilon}b via (Topp et al., 1980):

[6]
where c is the speed of light in a vacuum (3 x 108 m s-1) and L is the thickness of the medium in meters. This technique requires that the location of wavelet maxima for each interface, that is, (5) and (7) as seen in Fig. 1B, be located, and the two-way time between them determined.

Confounding Issues Affecting GPR Measurements
In addition to the effect of water content, near-surface GPR measurements of soils are affected by additional factors, including (i) the geometry of interfaces, (ii) soil texture, (iii) salinity, (iv) temperature, and (v) antenna radiation patterns and height above surface.

Surface and interface geometric considerations which affect radar measurements at normal incidence include surface roughness, vegetation canopy height, and architecture characteristics. Increased surface roughness will decrease the reflected signal magnitude (and thus SR response) by scattering the incident radiation away from the antenna (Blumberg and Freilikher, 2001). Vegetation canopy will enhance scatter of incident and reflected radar signal away from the antenna, effectively attenuating the measured surface reflectivity, but can increase backscatter to the antenna, depending on antenna-surface geometry, canopy architecture, and transmit–receive polarization. This requires modeling to separate the canopy from the soil surface contribution, and in case of dense canopies and higher frequencies, can entirely eliminate any soil surface contribution (Lang and Sidhu, 1983; Dobson and Ulaby, 1986b; Karam et al., 1995; Ulaby et al., 1990, 1996).

Soil dielectric and conductive properties that affect radar measurements include water content, texture, salinity, and temperature. Soil texture is a key factor controlling partitioning of soil water into free and bound water, with bound water being rotationally hindered by surface forces and thus less visible to the radar (Hasted, 1973; Dobson et al., 1985; Or and Wraith, 1999). Thus, coarse and low surface area soils such as sands contain little bound water and exhibit low values of water content at field capacity; high surface area soils will have high bound water contents and field capacity. Furthermore, with increasing surface area, soils generally contain more salts, hence higher electrical conductivities (ECs) that affect GPR measurements in two ways. First, EC affects the magnitude of reflections at interfaces, that is, larger reflectivity with larger EC of the soil. Second, EC attenuates the radar signal within the medium and limits its penetration depth (von Hippel, 1954; Ulaby, 1999), such that in very saline soils PT measurements could be problematic.

Soil temperature has been found to control partitioning of bound to free water in soils, particularly in higher surface area soils where a large proportion of the soil water is bound (Bockris et al., 1966; Or and Wraith, 1999; Jones and Or, 2002). Increases in soil temperature serve to desorb bound water layers and decrease the viscosity of free water, which (i) cause an increase in bulk dielectric permittivity due to a greater volume of free water, and (ii) increase EC due to increased ion mobility.

Antenna radiation patterns also affect SR and PT measurements via air-spreading losses which are caused by the decrease in signal strength with distance from an interface due to the noncollimated nature of the antenna radiation pattern (Ulaby, 1999). These losses can limit the ability of GPR systems to do subsurface sensing, particularly when other loss mechanisms such as EC become significant.


   EXPERIMENTAL SETUP
TOP
 ABSTRACT
 INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Bare Soil Measurements
Radar measurements were acquired from bare soils at Utah State University's Greenville Farm in North Logan, UT, and utilized a moving platform, as seen in Fig. 2A and B. The soils were placed in a large box (which was placed upon the local soil, Millville silt-loam, and oriented in the east-west direction lengthwise) that was about 10 m long by 2 m wide by 0.4 m high and partitioned into three sections with three different soils. Each box section was divided into two subsections, where the bottom of the eastern subsection was lined with thick aluminum foil (punctured for free drainage) which served as an electrically terminating subsurface interface. The moving platform, which supported the antenna at a height of about 45 cm above the soils, consisted of a motorized wooden cart that moved east and then west at 10-min intervals. Because of the close proximity of the antenna to the soil, free-space corrections were used with this data set.



View larger version (77K):
[in this window]
[in a new window]
  		Fig. 2. (A) Representative diagram of moving platform measurements, including ground truthing setup. (B) Photograph of setup at Greenville Farm.

 
Thermocouples were placed at depths of 2, 5, 10, 15, and 25 cm in the soil, and TDR probes placed horizontally at 2 cm and vertically at 0- to 15-cm and 15- to 30-cm intervals, excepting one section containing a sand–bentonite mixture whereby the 0- to 15-cm probe was replaced with a horizontal probe at the 7.5-cm depth. The TDR measurements were acquired with WinTDR software (Soil Physics Laboratory, Utah State University, Logan, UT), which controlled a Tektronix (Beaverton, OR) 1502B TDR and four Campbell Scientific (Logan, UT) SDMX50 multiplexers. Both thermocouple and TDR data were acquired every 10 min concurrent to GPR measurements. The 0- to 15-cm and 15- to 30-cm TDR water content values were then averaged for the whole profile.

Soils were irrigated by ponding for 24 h to induce homogenous water content distribution and to smooth out the surface by aggregate slaking. The time scale used is elapsed from the midnight before the end of ponding (18 Aug. 2002 at 0000 h), with measurements acquired every 10 min. Soil types reported in this study were for unterminated coarse sand, terminated Millville silt loam, and an unterminated 80% sand–20% bentonite (mass basis) artificial mixture (their physical properties are given in Table 1), where terminated and unterminated denotes whether the subsurface was lined with aluminum foil. Not all of the soils and termination schemes sampled in this paper are shown, but will be reported on in a future manuscript. The sand–bentonite mixture was placed in a layer above the midsection of the sand box, as shown in Fig. 2A and B. Soil thicknesses were determined by excavation at the end of the experiment and by measurement of heights at three locations.


View this table:
[in this window]
[in a new window]
  		Table 1. Soil properties used in ground-penetrating radar measurements.

 
Vegetated Surface Measurements
The radar unit was deployed in a greenhouse to measure soil water drying patterns during the growing season of the wheat canopy. The dwarf wheat cultivar planted was USU 9-2-2 (developed at the USU Crop Physiology Lab) with maximal canopy height of 0.4 m to facilitate antenna placement at 1 m above the soil surface to ensure a uniform canopy footprint.

Wheat was planted in a 1.4 m2 square planter on a 2.5-cm2 square grid. The planter was filled with a peat–perlite artificial soil mixture (bulk density of about 140 kg m-3) to a depth of 0.14 m. The bottom of the planter was terminated with aluminum foil to reflect radiation and mark the location of the bottom for PT measurements. After planting, the soil surface settled to a total thickness (including gravel) of about 12 cm. A representative diagram and photograph of the setup may be seen in Fig. 3A and B.




View larger version (145K):
[in this window]
[in a new window]
  		Fig. 3. (A) Representative diagram and (B) photograph of greenhouse wheat canopy measurements. GPR, ground-penetrating radar; PT, propagation time; SR, surface reflection.

 
The wheat was planted on 9 Mar. 2002 and GPR canopy measurements were acquired every 30 min starting on 15 Mar. 2002. For the first few weeks, water was applied by spraying to prevent damage to wheat plants, followed by flood irrigation for the remainder of the season. Canopy height was monitored every few days using a tape measure. On 7 May 2002 the canopy was removed and biophysical parameters such as number of tillers (shoots, stalks), tiller lengths and tiller section lengths, number of leaves, and leaf area index (LAI) were measured. The bare soil was irrigated again the next day at 2130 h and the drying curve was measured until 21 May 2002 at 1500 h. Soil water content measurements were acquired after canopy removal via gravimetric sampling with a trowel and drying at 110°C for 24 h and then converted to volumetric water contents. Because of the organic nature of the soil and its low bulk density, the Topp et al. (1980) relationships were not used to relate bulk dielectric permittivity to soil water content. da Silva et al. (1998) derived a relationship for a peat-perlite mixture:

[7]


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
SUMMARY AND CONCLUSIONS
 REFERENCES
 
Drying of Bare Soil
Bare soil measurements from sand (Fig. 4A), Millville silt loam (Fig. 4B), and the sand–bentonite mixture (Fig. 4C) are shown with 2-cm soil temperatures (Fig. 4D). Radar and TDR measurements were acquired from onset of irrigation through the following 10 d while soils were gradually drying.



View larger version (50K):
[in this window]
[in a new window]
  		Fig. 4. Ground-penetrating radar and time-domain reflectometry (TDR) measurements of drying from (A) sand, (B) Millville silt loam, and (C) the sand–bentonite mixture. (D) 2-cm soil temperatures for sand (S), Millville (M), and sand–bentonite (S–B). Solid black and gray lines in B and C denote surface reflection (SR) and 2-cm TDR minimum value baselines, respectively. PT, propagation time.

 
Radar measurements of the sand (Fig. 4A) showed quick drainage for SR measured water content values and slower drainage for PT measured water content values. Most of the reduction in SR occurred within the first 12 h after irrigation because of drainage. Propagation time measurements showed more gradual decrease in water content and attained a steady value after 4 d, probably because of the slow drainage through the underlying and wet Millville silt loam soil. Shallow (2 cm) TDR data followed SR patterns, whereas 0- to 30-cm measurements were more similar to PT results. The TDR data were relatively noisy because of lack of clear first peak for TDR waveforms from sand, which can be exacerbated by variations in initial peak location due to temperature changes affecting the propagation velocity along the coaxial cable (Robinson et al., 2003) and could result in water content estimation errors of up to 0.02 m3 m-3.

Millville silt loam data (Fig. 4B) showed slower drying than sand, with PT data gradually decreasing, and agreed with 0- to 30-cm TDR data, albeit that PT values were about 0.03 m3 m-3 less than TDR towards Day 10. The SR and 2-cm TDR data showed close agreement and evidence of diurnal thermodielectric effects relative to minimum value baselines (denoted by black and gray solid dotted lines, respectively).

The sand–bentonite mixture (Fig. 4C) showed a gradual decrease in water contents. For the first 3 d, water content values measured by SR and 2-cm TDR were highest and 7.5-cm TDR were the lowest. The maximal values of SR were probably attributable to the relatively high EC that enhanced surface reflectivity. The PT values initially showed agreement with SR, but as drying progressed PT values exceeded those of SR by about 0.06 m3 m-3. After 6 d, TDR data showed agreement with each other but were about 0.03 m3 m-3 over PT values. The difference between PT and TDR values at late times was probably because of the lower water content in the top 2 cm of the soil. It is possible that a sharp water content gradient at the top 1 or 2 cm of the soil resulted in a 0.09 m3 m-3 difference in water contents between SR and TDR values. The PT values were also shown to be noisier than SR measurements because of the small thickness of the sand–bentonite layer and system jitter that would cause fluctuations in peak location.

Data from both the Millville silt loam and sand–bentonite mixture showed noticeable midday thermodielectric effects which were most pronounced between Days 2 and 5 of drying, and decreased thereafter. These effects were seen in the SR and shallow (2-cm depth) TDR measurements. Air temperature was minimal around 0730 h (just after sunrise over the mountains), whereas the 2-cm soil temperatures lagged up to 1.5 h and varied with soil type. The timing of sand SR maxima (which were very small in magnitude) occurred around the time of the 2-cm minimum temperature (Fig. 4A). Millville silt loam SR maxima, as seen in Fig. 4B, were about 0.07 m3 m-3 above the extrapolated SR minimum values between Days 2 and 4. The TDR values also showed similar behavior to SR measurements, albeit that 2-cm TDR maxima occurred later in the morning, and the differences between extrapolated water content values were generally less than SR differences. Sand–bentonite also showed evidence of thermodielectric effects. The SR differences of 0.06 m3 m-3 above extrapolated values occurred on Day 2. Measured SR minima for sand–bentonite coincided with minimal 2-cm temperatures, whereas maximal values occurred up to 5 h after 2-cm soil temperature minima. This is best shown in Fig. 5A and B for the sand bentonite mixture from Days 6 to 10. In these figures, minima agreed for respective depths and water content values, namely that 2-cm temperature minima coincided with SR minima, and 2-cm TDR minima occurred slightly thereafter. The SR maxima occurred between late morning and midday when the soil was heating, and resulted in bound water desorption, increased EC, and evaporation, the first two of which served to increase reflectivity and the last which had the opposite effect. The 2-cm TDR maxima lagged by a few hours behind SR values because of the time lag in the temperature wave reaching the subsurface, but clear changes were seen.



View larger version (43K):
[in this window]
[in a new window]
  		Fig. 5. (A) Ground-penetrating radar and time-domain reflectometry (TDR) measurements of diurnal thermodielectric effects on the sand–bentonite mixture. (B) Data logger and 2-cm soil temperatures. Red up- and blue down-pointing arrows denote 2-cm temperature maxima and minima, respectively. mbl, minimum baseline; PT, propagation time; SR, surface reflection.

 
Low surface area soils such as sand showed maximal SR values before the 2-cm soil temperature had achieved a minimum when soil heating was still low, whereas higher surface area soils such as the Millville silt loam and the sand–bentonite mixture showed maximal SR values when the top few cm of the soil were warming up despite the overall drying trend. Simply put, even though the soil was warming (and thus water is evaporating), bound water was being liberated to free water and we see a net increase in measured SR and 2-cm TDR.

It should be noted that because of the cart movement direction and measurement triggering method, measurements were often collected 1 cm apart laterally for opposite acquisition directions, which caused some noise due to microscale variability with discrepancies between directions being sometimes on the order of 0.01 m3 m-3. This was partly attributable to the height at which measurements were acquired (z = 0.45 m) and the electric field density of the horn antenna, which concentrated most of the energy at the center of the illuminated area (0.13 m2 total). Furthermore, measurements from this setup were based upon a single waveform measurement, such that no averaging was performed on acquired data. The total measurement RMSE between SR water contents and the extrapolated baselines were 0.020 and 0.015 m3 m-3 for the Millville and sand–bentonite mixtures, respectively. For 2-cm TDR data, the total variations were 0.014 and 0.007 m3 m-3, respectively.

Greenhouse Measurements
Radar measurements of vegetation canopy and the underlying soil can be seen in Fig. 6. Figure 6A compares SR and PT for initial stages of the growing wheat with incomplete canopy cover between 20 and 33 d after planting. Figure 6B depicts measurements for a mature canopy between 50 and 59 d after planting, and Fig. 6C shows measurements over bare soil surface after harvest and canopy removal. Differences in irrigation patterns can clearly be seen in Fig. 6A. Prior to 25 d after planting, the soil was gently sprayed to prevent the plants from being damaged. After this initial period, the surface was flood irrigated for a few minutes at each irrigation time to produce spatially homogenous water contents. Because of gentle spraying and the lack of a full canopy (canopy height hc was below 22 cm), SR measurements were frequently greater than PT values except during drying. However, once flood irrigation commenced, this pattern reversed itself and PT exceeded SR. Additionally, as canopy height (and biomass) increased, so did the discrepancies between the two measurements. As canopy height hc increased, the overall range between measured maximal and minimal water contents decreased for the SR. Figure 6B shows that with a fully developed canopy, SR values were much lower, often by almost 0.2 m3 m-3 than PT values, and the two only approached one another or crossed after several days of drying under a transpiring canopy.



View larger version (53K):
[in this window]
[in a new window]
  		Fig. 6. Surface reflection (SR) and propagation time (PT) measured soil water contents with canopy development and bare soil. (A) Period from 20-35 d after planting; (B) 50-59 d; (C) drying curve after canopy removal. hc, canopy height.

 
The decrease in SR correlated with an increase in canopy biomass, which scattered away radiation and induced canopy reflections, as seen in Fig. 7. It should be noted that the location of the highest canopy reflection was typically about 8 cm below the top of the canopy, reflecting the vertical averaging by the measurement (canopy height is strongly influenced by the tallest plants in the canopy and not the plant average height or vertical density distribution). A canopy reflection became evident approximately a month after planting; in the following week, a second reflection beneath the first canopy reflection became apparent. The top reflection correlated with flag leaves and wheat seed heads (specifically with the average tiller length at the end of the experiment), and the bottom layer correlated with the second leaf layer beneath it, with the intensity of these reflections increasing with time. Vegetation canopy reflections were not visible prior to one month because of the size and magnitude of the surface refection (on the order of several volts), which would have masked canopy reflections (on the order of tenths of a volt). Canopy LAI, a measure of canopy leaf area per ground area, was measured at the end of the experiment and was found to be 4.5 m2 leaves m-2 ground directly beneath the antenna, and 6.9 m2 m-2 for the entire box. It is assumed that the leaf area of the plot increased in a similar manner to that of canopy height. The LAI, however, does not give any information as to how the leaves are oriented spatially within the canopy, but is a commonly used parameter in vegetation scattering models.



View larger version (33K):
[in this window]
[in a new window]
  		Fig. 7. Wheat canopy reflection development from a low canopy to maturity, with distance from the soil surface. hc, canopy height.

 
The PT measurements tended to be higher than SR and were not affected by canopy cover because of their dependency on travel time between surface and subsurface reflection locations and not upon reflection amplitude. Removal of the canopy (Fig. 6C) resulted in an immediate and large increase in SR values, with SR values exceeding or equaling PT values. The SR values appeared to sharply decrease between Day 64 to Day 65 and then level off from Day 65 and onward, with PT decreasing at a constant rate (the sharp decrease is also echoed in the PT as well, though not as greatly). Comparison of SR and PT with daily gravimetrically determined volumetric water content measurements obtained after canopy removal (Fig. 8) showed that at higher water contents both SR and PT appeared to underestimate water content, with the RMSE values being 0.05 m3 m-3 for SR and 0.08 for PT. The SR appeared to show the closest agreement with gravimetric with r2 = 0.89, with PT showing lower agreement with r2 = 0.86, particularly when the soil was wet or very dry. Potential explanations for this would either be because of the dryness and low bulk density of the top surface layer, which would cause a region of very low permittivity similar to air (observed waveforms of the dried-out organic soil several months later showed no noticeable SR), and thus caused the radar to sample at a greater depth. This would essentially lower the effective distance of the travel path L, and thus bias PT measurements such that the measured propagation time or distance would be less than it should be and would underestimate water content. Errors in measurement of L could have also biased PT—any error on the order of a few millimeters could bias measurements by a few percentage points or more, particularly on smaller scales. It should be noted, however, that PT measurements did show less noise for this data set than for the sand–bentonite set, albeit that the travel thicknesses were similar because of averaging of 30 waveforms per sample in greenhouse data prior to analysis.



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 8. Bare soil surface reflection (SR) and propagation time (PT) vs. gravimetrically measured volumetric water contents.

 
Potential Applications of Small-Scale Surface Radar Measurements
Mapping of Soil Texture
The differential responses of soils because of thermodielectric properties allow for the mapping of soil texture, particularly between times of day when the variability between soils are the greatest, namely at around dawn and late morning or noontime, and ideally under clear summer conditions (Serbin et al., 2001, p. 69), in conjunction with a laser profilometer or surface scanner (Blumberg et al., 2002) to eliminate the effects of surface roughness and microtopography. Soil texture mapping could be performed via the use of a grid that would be sampled twice at identical locations, assuming that this could be done in a rapid manner and that roughness conditions are known or can be accounted for, possibly with concurrent laser roughness measurements (Blumberg et al., 2000; Blumberg et al., 2002). The magnitude of the response between SR values at sunrise and late morning would be indicative of soil type. While these effects may not always be large, the sensitivity of the radar allows us to monitor very small changes in reflectivity. This would be especially useful for fields where the soil texture was suspected to vary greatly, for example, where a sandy soil would lie next to or interfinger with a clayey soil.

Field Mapping of Soil Water Content and Crop Biophysical Parameters
Measurements from both bare and vegetated soils show that a horn antenna could be deployed in the field for a number of purposes, as suggested by results in this paper. Should soil texture be known, SR and PT (assuming that a reflective subsurface dielectric discontinuity exists at a known depth) responses could be used to measure soil water contents within the profile, with soil temperature data being used to correct for thermodielectric biases with the soil.

Vegetation canopy characteristics could also be measured, albeit that an antenna with a higher center frequency (narrower pulse width) would allow for greater resolution, particularly with lower canopies whose canopy layer reflections are likely to be swallowed up by a large SR, and as these higher frequencies are more sensitive to vegetation canopy characteristics and are less transmissive to microwave radiation (Ulaby et al., 1996). The number, locations, and intensities of canopy reflections could be used for correction of measured SR values, but these first require application and possibly development of vegetation canopy models to determine the effect of such vegetation on SR. This would allow for independent evaluation of the canopy impact on SR that serves as the backbone of many SAR platforms. The apparent reduction in inferred SR water content could be quantified and incorporated in inversion algorithms. Additional experiments for well-defined antenna footprint and canopy geometry, and surface properties can be used for definitive testing of canopy-related radiation transfer models.

Near-Surface Validation of Microwave Remote Sensing
Ground validation of radar and passive microwave remote sensing can often be a tedious point sampling process. Such truthing often employs acquisition of gravimetric and TDR (or other dielectrically based measurement methods, such as portable dielectric or capacitance probes like those used by Dean et al., 1987; Brisco et al., 1992; Dubois et al., 1995; and Nadler and Lapid, 1996) samples, as well as vegetation canopy parameters. This can be both expensive with respect to manpower and equipment and very time-consuming for limited coverage. Furthermore, most microwave applications in the L-band and higher frequencies are typically sensitive to only the top 5 cm or less of the soil or even less, depending upon water content and salinity, with some measurements being sensitive to shallower depth, particularly for C-band measurements. While the 0- to 5-cm depth can be easily monitored via TDR, soil water contents of the top 1 cm are very difficult to sample except via gravimetric measurements. Furthermore, the top 1 cm of soil will have lower water contents than the soil layers beneath it, particularly as drying progresses (Capehart and Carlson, 1997; Hillel, 1998), even though soil dielectric properties could limit the effective penetration depth even at longer wavelengths to the top layer of the soil (von Hippel, 1954; Elachi, 1988). Since the radar unit and horn antenna can be easily mounted on a vehicle, profiling of soil dielectric or reflectivity, vegetation canopy, and near-subsurface feature parameters could be acquired for large areas in a relatively short time, particularly with L-band satellites that utilize similar frequencies, and the antenna could be oriented according to the incidence angle and polarizations of the sensors being used. If possible, higher frequency antennas could be developed to allow for better studies of vegetation canopy and truthing of 5.3 GHz radar sensors such as ERS, ENVISAT, and RADARSAT.


   SUMMARY AND CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
REFERENCES
 
Radar with an aboveground horn-based antenna was shown to provide valuable information on dynamics of bare soil water contents via both SR and PT methods, with these data largely agreeing with TDR. Furthermore, this method successfully recorded concurrent differences in wetting and drying patterns between three soil types, namely a sand, a silt-loam, and a sand–bentonite mixture. Results showed that SR measurements were limited to surface soil skin (2 cm) and may not reflect critical hydrological processes at depths not exceeding a fraction of a meter. These measurements were greatly influenced by diurnal temperature oscillations which may require a correction to be implemented, or can be gainfully exploited for soil texture mapping. Additionally, SR measurements were found to be biased by the existence of vegetation canopy, and showed a consistent decrease in intensity with increase in canopy height and thus biomass; in contrast, PT measurements were unaffected. These canopy reflections could be accounted for in intensity and location, offering a promise for development of vegetation canopy corrective indices and future studies of biomass with GPR.

This technology and setup potentially allows for mapping of soil texture in heterogeneous fields, near-surface remote truthing of radar data from air- and spaceborne sensors, and the study of vegetation canopy cover.


   ACKNOWLEDGMENTS
 
We acknowledge Bill Mace for his assistance with fieldwork and construction. Ariel Serbin, Seth Humphries, and Louis Koberstein are acknowledged for their help in data processing. Special thanks to Sally Maxwell and Scott Jones for assistance in manuscript preparation. Funding for the GPR unit was provided in part by Phil Rasmussen, the United States-Israel Binational Agricultural Research and Development Fund (BARD), through project no. IS-2839-97, and the Utah Agricultural Experimental Station (UAES). The Rocky Mountain Space Grant Consortium is also acknowledged for providing a graduate research fellowship.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 THEORETICAL CONSIDERATIONS AND...
 EXPERIMENTAL SETUP
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
S. R. Evett and G. W. Parkin
Advances in Soil Water Content Sensing: The Continuing Maturation of Technology and Theory
Vadose Zone J., November 11, 2005; 4(4): 986 - 991.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
S. Lambot, J. Rhebergen, I. van den Bosch, E. C. Slob, and M. Vanclooster
Measuring the Soil Water Content Profile of a Sandy Soil with an Off-Ground Monostatic Ground Penetrating Radar
Vadose Zone J., November 1, 2004; 3(4): 1063 - 1071.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
S. Lambot, M. Antoine, I. van den Bosch, E. C. Slob, and M. Vanclooster
Electromagnetic Inversion of GPR Signals and Subsequent Hydrodynamic Inversion to Estimate Effective Vadose Zone Hydraulic Properties
Vadose Zone J., November 1, 2004; 3(4): 1072 - 1081.
[Abstract] [Full Text] [PDF]

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 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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Serbin, G.
Right arrow Articles by Or, D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Serbin, G.
Right arrow Articles by Or, D.
GeoRef
Right arrow GeoRef Citation
Agricola