| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
a J.M. Ham, Department of Agronomy, Kansas State University, Manhattan, KS 66506
b Plant and Soil Science Department, Texas Tech University, Lubbock, TX 79409
c Rocky Mountain Research Station, U.S. Forest Service, 240 W. Prospect Rd., Fort Collins, CO 80526
d Department of Biological and Agricultural Engineering, Kansas State University, Manhattan, KS 66506
* Corresponding author (gjk{at}ksu.edu).
Contribution no. 03-330-J from the Kansas Agric. Exp. Stn., Manhattan, KS.
Received 19 March 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
) and change in water content (
) near the soil surface. The method has been evaluated in laboratory and greenhouse experiments, but not in a field setting. Our objective was to test the DPHP method under field conditions and for a range of soil properties. Twenty-five DPHP sensors and five monitoring stations were constructed and installed at five locations in northeastern Kansas to measure and at 3-h intervals for 3 mo. In addition, was estimated by coupling measurements with independent measurements of obtained by soil sampling at sensor installation. Additional soil samples were collected from each location during the monitoring period to provide independent measurements of . Regression of DPHP and independent measurements revealed slight bias but substantial offset error (about 0.1 m3 m-3) in the DPHP method. The offset error could not be fully attributed to bias in any single input parameter, but could have been caused by a combination of biased parameters. Estimates of from measurements also revealed slight bias, but offset error was considerably smaller. Use of a published empirical calibration for DPHP sensors almost completely eliminated this bias and further reduced the offset error to approximately 0.01 m3 m-3. Thus, the approach combined with use of the empirical calibration appears to have practical utility.
Abbreviations: DACS, data acquisition and control system • DPHP, dual-probe heat-pulse
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
-ray absorption, time domain reflectometry, and electrical resistance) may be prohibitively complex and expensive or require soil-specific calibration (Topp and Ferré, 2002). Methods such as neutron scattering and time domain reflectometry also may present difficulties near the soil surface (Hignett and Evett, 2002; Nielsen et al., 1995). Hence, there is a need for a simple method with the capability to provide continuous measurements of water content in the upper soil profile.
One promising method for measuring volumetric water content () is the DPHP method of Campbell et al. (1991). Dual-probe heat-pulse sensors consist of two parallel cylindrical probes typically constructed from 1.27-mm-o.d. stainless-steel tubing and separated by a distance of 6 mm (Tarara and Ham, 1997; Campbell et al., 2002; Basinger et al., 2003). One probe contains a heating element, and the other contains a temperature sensor (thermocouple or thermistor). A timed heat pulse produced by the heating element causes a temperature change, which is measured with the temperature probe. The magnitude of this change in temperature is a function of probe spacing, the amount of heat introduced, and the soil volumetric heat capacity (C). Thus, with a known heat input and probe spacing, this allows calculation of C. The basis of the DPHP method for measuring is the strong linear relationship between C and for soils with minimal shrink–swell potential. Because this relationship can be modeled relatively well for mineral soils, it permits estimation of from measurements of C. Sequential measurements of the heat input and resulting temperature rise further allow estimation of the change in soil water content () without knowledge of soil-specific properties (Bristow et al., 1993). The DPHP method provides advantages in that sensors are small, inexpensive to construct, and require no special calibration. In addition, DPHP sensors are capable of providing spatial resolution of a few millimeters, making them ideal for measurement near heterogeneities such as the soil surface (Tarara and Ham, 1997; Kluitenberg and Philip, 1999; Philip and Kluitenberg, 1999).
The DPHP method for measuring and has been evaluated in laboratory and greenhouse studies (Bristow et al., 1993; Noborio et al., 1996; Tarara and Ham, 1997; Bristow, 1998; Song et al., 1998, 1999; Campbell et al., 2002), but these studies were conducted with a limited range of soil types and/or water contents. Basinger et al. (2003) conducted a laboratory experiment to evaluate the DPHP method in soil materials with a wide range of physical properties and water contents. They found that with an empirical correction the DPHP method provided unbiased estimates of with RMSE = 0.022 m3 m-3 for water contents ranging from 0.02 to 0.59 m3 m-3. It was also shown that the DPHP method yields unbiased estimates of , with RMSE = 0.012 m3 m-3 for water content changes ranging from 0.0 to 0.33 m3 m-3. The accuracy and precision achieved in these experiments indicates the effectiveness of the DPHP method for determining and .
The DPHP method also has been used for routine and measurement in field investigations (Bremer et al., 1998; Ham and Knapp, 1998; Bremer and Ham, 1999; Bremer et al., 2001; Bremer and Ham, 2002). Unfortunately, it cannot be assumed that field measurement of and can be obtained with the same level of accuracy and precision as achieved in a laboratory setting. One important factor is the potential for probe deflection upon insertion into the soil (Campbell et al., 1991; Kluitenberg et al., 1993, 1995; Noborio et al., 1996; Tarara and Ham, 1997). Probe deflection can cause and to be either overestimated or underestimated depending on the direction of deflection. For typical values of input parameters, Basinger et al. (2003) showed that a 5% error in probe spacing causes an error of 0.024 m3 m-3 at = 0.00 m3 m-3 and an error of 0.084 m3 m-3 at = 0.60 m3 m-3. Good soil–probe contact also may be more difficult to achieve in the field than in the laboratory.
Another important factor that must be considered in field settings is spatial and temporal variation in soil temperature. The theory underlying the DPHP method assumes that isothermal conditions exist in the vicinity of the sensor and that temperature rise is caused only by the heater probe of the sensor. Consequently, variations in soil temperature with time and depth during a DPHP measurement may cause measurement error (Bristow et al., 1993). This error, caused by the diurnal cooling and warming of the soil, would vary throughout the day.
In their field experiment, Tarara and Ham (1997) showed that DPHP sensor and -attenuation measurements of agreed to within 0.05 m3 m-3; however, their experiment was limited in size and duration. No further field evaluations have been reported in the literature. Thus, there remains a need for field evaluation of the DPHP method. The objective of this study was to test the DPHP method in a field setting and for a range of soils. The DPHP sensors and monitoring stations were constructed and placed in the Black Vermillion Watershed in northeastern Kansas. The stations automatically monitored precipitation and soil water content for 3 mo during the summer of 1999. Special consideration was given to soil–probe contact and probe deflection at sensor installation and postseason excavation. Measurements of obtained by the DPHP method were compared with independent measurements obtained by soil sampling. Prediction of water content with the approach and an empirical correction were also explored.
|
THEORY |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
 | [1] |
As suggested by Campbell et al. (1991) and Bristow et al. (1993), the method can be adapted to estimate soil volumetric water content from C. The relationship between C and (m3 m-3) is
 | [2] |
b (kg m-3) is the soil bulk density, and cs (J kg-1 °C-1) is the specific heat of the soil solid (mineral and organic) constituents. Equations [1] and [2] can be combined to yield
 | [3] |
Using Eq. [3], the measured quantities q and Tm can be coupled with an estimate of the product bcs to obtain . Bristow et al. (1993) further showed that change in volumetric water content () can be determined if sequential measurements of q and Tm are made. The relationship is
 | [4] |
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
m-1, Nichrome 80 Alloy, Pelican Wire Co., Naples, FL) through the housing four times, resulting in two loops within the housing and a resistance of 820 m-1 of heater probe. Completed heater probes had a total resistance of approximately 33 . Temperature probes were made by positioning a welded thermocouple junction (Type T, 36-American wire gauge thermocouple wire; Omega Engineering, Inc., Stamford, CT) 13.5 mm from the unflared end of the probe housing. Heater and temperature probes were filled with a high thermally conductive epoxy (Omegabond 101, Omega Engineering, Inc.). The temperature and heater probes were connected to extension wire containing 22-American wire gauge stranded conductors (Pelican Wire Co.). Thermocouple lead wires were wrapped tightly around corresponding Type-T extension wires and secured with epoxy to form solderless connections. The completed heater and temperature probes were fixed in a milled PVC block that held the probes parallel with a spacing of 6 mm.
|
The five sensors were read sequentially every 3 h. A measurement sequence included measurement of initial probe temperature, application of an 8-s heat pulse and measurement of applied power (q), and measurement of the resultant temperature increase at 0.5-s intervals for 70 s to determine Tm. When power was applied to each sensor, current was determined by measuring the voltage drop across a four-legged 1- resistor (0.1% tolerance; Model VPR5, Vishay Resistors, Malvern, PA) that was wired in series with the heater. Measurements of q were obtained by sampling the current-sensing resistor at 0.5-s intervals, converting current to power (current squared x 820 m-1), and integrating the result. This yielded a heat input per unit length of q 
0.9 kJ m-1.
Sensors were calibrated using agar-stabilized water (6 g L-1) on the corresponding DACS that would be used in the field, as described by Campbell et al. (1991). Apparent probe spacing was calculated by setting C = 4.18 MJ m-3 °C-1 in Eq. [1] and solving for r.
Site Descriptions
The five monitoring stations were placed at sites within the Black Vermillion Watershed basin in northeast Kansas, selected to provide a range of soil physical properties and cropping systems (Table 1). Weather conditions during the monitoring period were typical for the region; mean cumulative rainfall for the five sites was 36.1 cm during the monitoring period.
|
Specific heat was estimated from (Basinger et al., 2003):
 | [5] |
m and o are the mass fraction of the mineral and organic constituents, respectively. Values of 0.73 and 1.9 kJ kg-1 K-1 from Kluitenberg (2002) were used for cm and co, respectively.
Water Content Monitoring
Monitoring stations were installed in May and June 1999. The five DPHP sensors for each monitoring station were evenly distributed around the perimeter of an approximately circular area with a radius of 7.6 m. Sensors were installed midway between the plant rows and inserted so that the heater and temperature probes were parallel to the soil surface with both at a depth of 10 cm.
At each sensor location, soil was first excavated to a depth of 8.5 cm so that a sample could be obtained for determination of bulk density and water content. The sample was obtained by pressing an aluminum cylinder (4.82-cm diam., 3-cm height) into the soil. After the sample was removed, a vertical face was prepared using hand tools. A coring tool with a 6 by 16 mm rectangular opening, fabricated from thin sheet metal, was pushed approximately 26 mm into the face and then removed to create a cavity approximately the size of the sensor body (Fig. 1). The sensor was aligned in the cavity and then pushed into the soil until the sensor body rested completely within the cavity and the probes were positioned in undisturbed soil. Excavated soil was then carefully replaced. At each monitoring station, the DACS was activated immediately after completing sensor installation.
Soil samples were collected at least four times at each monitoring station to obtain independent measurements of soil water content. To minimize soil disturbance near the sensors, samples were not collected immediately adjacent to sensor locations. Instead, 10 soil samples were collected from designated sampling zones located at approximately the same radial distance from the monitoring station as the sensors. A core sampler was used to collect a sample with an aluminum cylinder (4.82-cm diam., 3-cm height), centered at a depth of 10 cm. Each hole was backfilled with soil and marked to avoid repeat sampling; care was taken to minimize compaction of the surrounding soil. Soil samples collected during the field experiment were weighed, dried for 24 h in a convection oven at 105°C, and then reweighed to determine bulk density and gravimetric water content. Tests were conducted to ensure that water loss was negligible en route from the field sites to the laboratory before determination of gravimetric water contents.
Each DPHP sensor was carefully excavated using hand tools at the conclusion of the 3-mo monitoring period in August and September of 1999. Objects (roots, stones, residue) or cracks near the sensors were noted, and a visual inspection of soil–sensor contact was made. A visual inspection of probe deflection also was made before sensor removal. After the sensor had been removed, a soil sample was taken near the sensor to obtain final measurements of bulk density and water content. These samples were collected using the same procedures as those used during sensor installation.
Estimation of with the DPHP Method and Data Analysis
Estimates of for each DPHP sensor were obtained from Eq. [3] by using measured values of q, Tm, b, and r for each sensor along with the value of cs estimated for each site. The values of b used in these calculations were those obtained at each sensor location before sensor installation. These estimates of were compared with water contents obtained via soil sampling to evaluate the accuracy of the DPHP method. Comparisons also were performed with DPHP estimates that were corrected according to the procedure of Basinger et al. (2003) to remove bias. Based on results from a laboratory experiment, they proposed that bias could be removed by employing the empirical calibration
 | [6] |
' (m3 m-3) is the corrected (unbiased) DPHP estimate of soil volumetric water content.
Estimates of for each DPHP sensor were obtained from Eq. [4] by using measured values of q, Tm, and r for each sensor. These estimates of were subsequently used to predict water content with two different approaches. The first approach yielded the predicted water content , obtained from
 | [7] |
i is the water content obtained by soil sampling at sensor installation. The second approach yielded the predicted water content ', which incorporated the empirical correction of Basinger et al. (2003). This quantity was obtained from
 | [8] |
 | [9] |
Because gravimetric samples could not be collected at actual probe locations and because of the small-scale variability inherent in the DPHP method (Campbell et al., 2002), mean values for , ', , and ' obtained from five DPHP sensors at a given time and location were computed and used for subsequent comparisons to gravimetric estimates. Similarly, a mean value from 10 gravimetric samples collected at a given time and location was used in analysis. Regression analysis was conducted using standard linear regression. Standard hypothesis tests were used to test slope and intercept parameters. Values for RMSE are not reported here because regressions were conducted using mean values, which limits interpretation of precision.
Error Analysis
Basinger et al. (2003) conducted a first-order error analysis of Eq. [3] and derived the expression
 | [10] |
) caused by absolute errors in b, cs, q, Tm, or r. Table 2 shows values for the sensitivity coefficients in Eq. [10] for typical values of q, Tm, and r, and the values of b and cs for Site 4. Multiplying the sensitivity coefficient for each parameter by the expected error in that parameter provides an estimate of the expected error in .
|
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
i = 0.35 m3 m-3), but no major difficulties were encountered. The DACS and corresponding DPHP sensors were maintained continuously at each site for approximately 3 mo with only minor problems related to power loss and heater circuit relay failure. While all sensors showed evidence of sound soil–probe contact at excavation, five of the 25 sensors showed slight evidence of deflection. However, data from these five sensors did not indicate increased error and were therefore included in subsequent analysis.
Temporal variations in water content from five DPHP sensors for the 100-d monitoring period at Site 3 are shown in Fig. 2. As expected, an increase in is evident following each precipitation event with a decrease in from evapotranspiration and/or drainage typically occurring between precipitation events. Variability can be noted among the five sensors with an average CV of 16% for measurements recorded simultaneously. The variability in water contents recorded by the DPHP sensors at a particular time, although difficult to separate from measurement error, is not surprising given the small spatial scale of measurement. Similar variability was observed at each site, with average CVs of 10, 8, 27 and 26% for measurements recorded simultaneously at Sites 1, 2, 4 and 5, respectively.
|
is also evident in Fig. 2 and was observed at each of the five monitoring sites. The mean obtained from five DPHP sensors at Site 3 during a 5-d period with no precipitation is shown in Fig. 3. Diurnal variation in is evident with a daily range in on the order of 0.01 m3 m-3. However, measurements taken at the same time on successive days show mean differences in of <0.006 m3 m-3. This suggests excellent repeatability in DPHP estimates of despite diurnal variations. Diurnal cycling of has been commonly observed near the soil surface (Jackson, 1973; Reginato, 1975; Idso et al., 1975); however, minimum daily water contents typically occur at approximately 1600 h. This cycling is not consistent with that observed in Fig. 3, which shows that minimum water contents occurred between 0600 and 0800 h. This suggests that the diurnal cycling evident in our data is likely the result of measurement error.
|
(Fig. 3), several variables used in Eq. [3] to estimate are temperature dependent and should be considered potential sources of error. The quantity cs is dependent on cm, which is in turn temperature dependent (Kluitenberg, 2002). Using Site 4 as an example, with m = 0.98, o = 0.02, and daily temperatures ranging from 20 to 30°C, the expected error in cs can be estimated from Eq. [5] and the linear approximation for temperature dependence of cm provided by Kluitenberg (2002). Because Site 4 had the largest mineral fraction of the five sites used in the experiment, it represents a worst-case scenario for error resulting from temperature dependence of cs. For this example, cs would be 0.753 and 0.767 kJ kg-1 K-1 at 20 and 30°C, respectively, yielding an error of approximately 2% in cs. Multiplying cs (0.014 kJ kg-1 K-1) by the sensitivity coefficient for cs (Table 2) yields an estimated error of = -0.004 m3 m-3. Because the magnitude of from cs temperature dependence is small, and because the observed temperature cycle is not consistently in phase with diurnal variations in DPHP estimates of , this source of error was not considered further.
A second concern raised by diurnal soil temperature cycling was the potential for error in Tm measurements. In this setting, the assumption of isothermal conditions in the vicinity of the sensor may be violated, thus creating temperature dependence in measurement of Tm. By assuming a maximum change in background temperature of 1.7°C h-1 calculated from successive measurements of initial temperature (Fig. 3), the change in background temperature during a sensor measurement (
20 s from initial temperature to Tm) would be approximately 0.01°C. Equation [10] predicts that an error of Tm = 0.01°C will give errors of magnitude 0.002, 0.004, and 0.008 m3 m-3 at = 0.1, 0.3, and 0.5 m3 m-3, respectively. Thus, the expected error in noted above may be a plausible source for the daily variations in . However, because temperatures were recorded at 3-h intervals in this experiment, it was difficult to estimate the actual change in background temperature during sensor operation as a result of diurnal temperature cycling or to assess the assumption of isothermal conditions in the vicinity of the sensor.
Based on available data, it appears unlikely that the diurnal variation shown in Fig. 3 is real or a result of temperature dependence in cs. Alternative explanations for observed diurnal variation in are error in measurement of Tm or temperature dependence of the electronic components used in the DPHP sensors and the DACS. Evaluation of these sources of error would require refinement of the methods reported herein. With daily variations in on the order of 0.01 m3 m-3, further evaluation is needed. Yet, the validity of the method in the field setting may still be reasonable with repeatability on the order of 0.006 m3 m-3.
DPHP vs. Soil Sampling Estimates of
The accuracy of the DPHP sensors in estimating was assessed with several regression relationships. Interpretations regarding precision are not included because regression analyses were conducted with mean values. The DPHP estimates of were first compared with estimates of obtained by soil sampling for all sites and sampling times (Fig. 4). Each point represents the mean obtained from five DPHP sensors as compared with the mean obtained from 10 soil samples for a given site and sampling time. Mean by soil sampling ranged from 0.11 to 0.38 m3 m-3 while mean DPHP estimates ranged from 0.21 to 0.47 m3 m-3. The average standard deviation was consistently higher for DPHP estimates (SD = 0.076 m3 m-3) than for the estimates obtained by soil sampling (SD = 0.028 m3 m-3). This result is not surprising inasmuch as fewer samples were used to compute the SD for DPHP estimates and the volume of individual soil samples was 55 cm3 whereas the soil volume characterized by a DPHP sensor is <1 cm3. It is well established that variability tends to decrease with increasing sample size (van Es, 2002; Hopmans et al., 2002). The effect of sample volume on variability may be particularly pronounced in this case because water content typically varies nonlinearly with depth near the soil surface during redistribution in the presence of evaporation.
|
estimates obtained by soil sampling and those obtained with the DPHP method (r2 = 0.91). Hypothesis tests showed that the regression slope is not significantly different than one (P = 0.46) whereas the intercept is significantly different than zero (P < 0.0001). The nonzero intercept suggests an offset error of approximately 0.1 m3 m-3, indicating that the DPHP sensors consistently overestimated . This relatively large offset error deviates markedly from results obtained in laboratory evaluations of the DPHP method (Bristow et al., 1993, 2001; Noborio et al., 1996; Tarara and Ham, 1997; Song et al., 1998, 1999; Bristow, 1998; Campbell et al., 2002; Basinger et al., 2003) in which little or no offset error was reported. The contrast with the results of Basinger et al. (2003) is particularly noteworthy because the sensors they used were nearly identical to those used in this study, the only difference being that their sensors were constructed with thermistors instead of thermocouples.
The relatively large offset error (Fig. 4) appears to suggest systematic error in the DPHP method stemming from bias (consistent underestimation or overestimation) in a parameter with an effect that is independent of . Examination of Eq. [3] shows that these parameters would include b and cs; however, simple calculations with the sensitivity coefficients for b and cs (Table 2) show that a relatively large bias error (>30%) in either parameter would be required to produce an offset error of 0.1 m3 m-3. Thus, it is unlikely that bias in b or cs was the sole cause for the offset error. Alternatively, the offset error could have been caused by bias in the measurements of q, Tm, and r. Note, however, that sensitivity coefficients for q, Tm, and r (Table 2) exhibit strong dependence. That is, measurement bias in q, Tm, or r causes water content error that changes significantly as a function of water content. Therefore, measurement bias in any of these parameters could produce offset error only if the magnitude of the bias varied as a function of water content. Inasmuch as the measurements of q and r are independent of soil properties, it is unlikely that bias in these measurements would exhibit dependence.
It is conceivable, however, that bias in the measurement of Tm could vary as a function of water content. Basinger et al. (2003) discussed several phenomena that may cause -dependent bias in the measurement of Tm. These included vapor distillation effects, contact resistance between the soil and temperature probe and nonideality of the temperature measurement because of differences between the thermal properties of the soil and the temperature probe, all of which could vary with water content. It is also important to note that Tm varies with . Thus, if measurement bias in Tm varies with the magnitude of Tm, it also has the appearance of a measurement bias that is dependent.
The sensitivity coefficients for Tm can be used to illustrate the -dependent or Tm–dependent bias in Tm that would be required to produce an offset error of 0.1 m3 m-3. Bias in Tm of approximately 31, 19, and 14% would be required to produce an error of 0.1 m3 m-3 at Tm = 2.17, 1.34, and 0.97°C, respectively ( = 0.1, 0.3, and 0.5 m3 m-3, respectively) without bias in the regression relationship. While such a scenario is conceivable, the magnitude of the errors required to produce the observed offset error appear to be unreasonably large. Furthermore, the mechanisms proposed above as potential causes for -dependent bias in Tm are not restricted to the field setting. If significant, they would likely cause -dependent Tm bias in the laboratory as well. Yet the results from laboratory evaluations of the DPHP method show no evidence of a significant offset error. The arguments presented above lead us to conclude that the offset error of 0.1 m3 m-3 (Fig. 4) was not caused solely by biased measurements of b, cs, q, or r. Although it seems unlikely that the offset error can be explained solely by -dependent or Tm–dependent bias in the Tm measurements, the only alternative explanation is that the offset was caused by some combination of biased parameters.
Although not statistically significant, the slight bias evident in the regression relationship (Fig. 4) also merits discussion. The regression equation shows a slight bias toward greater overestimation of at lower water contents, which is consistent with previous work (Tarara and Ham, 1997; Song et al., 1998; Bristow et al., 2001). Basinger et al. (2003) performed a comprehensive laboratory evaluation of the DPHP method and found that bias in the method was consistent across a variety of soil materials with different physical properties. This led them to propose Eq. [6] as a universal empirical relationship for removing bias in DPHP estimates. Regression of ' values from Eq. [6] with the estimates of obtained by soil sampling for all sites and sampling times (figure not shown) yielded the equation Y = 1.036X + 0.063, with r2 = 0.91. Comparison of this equation with the regression equation from Fig. 4 shows that the empirical correction reduced both bias and offset errors. While this result is desirable and appears to support the empirical correction approach of Basinger et al. (2003), it leaves unresolved a relatively large offset error of 0.063 m3 m-3.
Prediction of from DPHP Measurements
Estimates of obtained with the DPHP method can be combined with i to provide another means to predict soil water content, , from Eq. [7]. To assess the accuracy of this approach, values of were regressed against the estimates of obtained by soil sampling for all sites and sampling times (Fig. 5). As before, each datum represents the mean value of obtained from five DPHP sensors and i at sensor installation as compared with the mean obtained from 10 gravimetric samples at a given time and location. The regression relationship reveals a decrease in the offset error with improved agreement (r2 = 0.93). Hypothesis tests indicated that the slope is not significantly different than one (P = 0.21) and the intercept is significantly different than zero (P < 0.0001). The significant reduction in offset error is not surprising inasmuch as this approach forces agreement with the value of i obtained by soil sampling, although this value is not included in the regression. The approach therefore serves as a "field calibration" that can be expected to eliminate offset error in DPHP estimates resulting from biased input parameters. It should be noted that this approach eliminates the quantity bcs in DPHP estimation of soil water content, but, as discussed above, these input parameters are not the likely source of error in calculation of . Somewhat surprising is the greater regression bias in the approach (regression slope of 0.929, Fig. 5) when compared with that observed in Fig. 4 (regression slope of 0.951). This result can be explained by the fact that individual values of are consistently lower than corresponding DPHP estimates of . Thus, the increase in bias is likely caused by forcing the regression relationship through lower water contents where bias is typically most severe (Tarara and Ham, 1997; Song et al., 1998; Basinger et al., 2003).
|
from DPHP measurements of was achieved by incorporating the empirical correction procedure of Basinger et al. (2003) to produce ' (Eq. [8], Fig. 6). The regression relationship shows limited bias over the available range in (0.114–0.376 m3 m-3) again with strong agreement (r2 = 0.93). With a slope approaching one and an intercept approaching zero, the ' approach appears to improve accuracy in estimation of soil volumetric water content with the DPHP method and may provide practical utility. Use of this approach, however, does not resolve the source of error observed in DPHP estimation of (Fig. 4), which warrants further evaluation.
|
|
SUMMARY AND CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
with the DPHP method was successful in that sensors proved robust and automated measurements were possible for an extended period (approximately 3 mo) with only minor difficulties. Though not quantitatively evaluated, probe deflection at installation was limited and did not obviously hinder results. Individual sensors showed a large degree of variability, but the variability may be a result of the fine-scale measurement provided by the sensors. Small diurnal variations in also were observed, but these variations were inconsistent with previous reports and may be the result of temperature dependence in measurement parameters or error in the sensors and/or the DACS. Further work is needed to test the assumption of isothermal conditions in the vicinity of the sensor and to determine the source of diurnal variation in witnessed here.
Results of regression analysis indicate overestimation and bias in the DPHP method, but improvement was possible with correction procedures. Success in the use of and ', both dependent on , suggests the potential value of the DPHP method for estimating changes in soil volumetric water content in a field setting. By correcting for initial water content, these approaches also offer a means to continuously monitor in the field. Although improving practical utility, adjustments in the regression relationship made with ', , and ' do not resolve the consistent bias and overestimation of observed in the DPHP method. Further work is needed to evaluate the source of overestimation in the DPHP method and to improve bias as approaches 0.1 m3 m-3.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONTHEORYMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARY AND CONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. Liu, B. Li, T. Ren, R. Horton, and B. C. Si Analytical Solution of Heat Pulse Method in a Parallelepiped Sample Space with Inclined Needles Soil Sci. Soc. Am. J., September 1, 2008; 72(5): 1208 - 1216. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. A. Robinson, C. S. Campbell, J. W. Hopmans, B. K. Hornbuckle, S. B. Jones, R. Knight, F. Ogden, J. Selker, and O. Wendroth Soil Moisture Measurement for Ecological and Hydrological Watershed-Scale Observatories: A Review Vadose Zone J., February 25, 2008; 7(1): 358 - 389. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. H. Young, G. S. Campbell, and J. Yin Correcting Dual-Probe Heat-Pulse Readings for Changes in Ambient Temperature Vadose Zone J., January 23, 2008; 7(1): 22 - 30. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. M. DeSutter, T. J. Sauer, T. B. Parkin, and J. L. Heitman A Subsurface, Closed-Loop System for Soil Carbon Dioxide and Its Application to the Gradient Efflux Approach Soil Sci. Soc. Am. J., January 11, 2008; 72(1): 126 - 134. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Knight, W. Jin, and G. J. Kluitenberg Sensitivity of the Dual-Probe Heat-Pulse Method to Spatial Variations in Heat Capacity and Water Content Vadose Zone J., October 8, 2007; 6(4): 746 - 758. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Liu, B. Li, T. Ren, and R. Horton Analytical Solution of the Heat Pulse Method in a Parallelepiped Sample Space Soil Sci. Soc. Am. J., August 27, 2007; 71(5): 1607 - 1619. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. L. Heitman, R. Horton, T. Ren, and T. E. Ochsner An Improved Approach for Measurement of Coupled Heat and Water Transfer in Soil Cells Soil Sci. Soc. Am. J., May 16, 2007; 71(3): 872 - 880. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. E. Ochsner, T. J. Sauer, and R. Horton Field Tests of the Soil Heat Flux Plate Method and Some Alternatives Agron. J., June 5, 2006; 98(4): 1005 - 1014. [Abstract] [Full Text] [PDF] |
