SSSAJ 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 Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (15)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.
Soil Science Society of America Journal 64:319-326 (2000)
© 2000 Soil Science Society of America

DIVISION S-6-SOIL & WATER MANAGEMENT & CONSERVATION

Soil Water Evaluation Using a Hydrologic Model and Calibrated Sensor Network

D.C. Hymera, M.S. Moranb and T.O. Keeferb

a NASA–GSFC Hydrologic Sciences Branch, Code 974, Greenbelt, MD 20771 USA
b USDA–SWRC, 2000 E. Allen Road, Tucson, AZ 85719 USA

dhymer{at}hydro4.gsfc.nasa.gov


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and methods
 Theory
 Results and discussion
 Discussion
 Conclusions
 REFERENCES
 
Studies show that it may be possible to combine satellite-derived soil water maps with soil–vegetation–atmosphere transfer (SVAT) models to obtain spatially distributed, temporally continuous information on vadose zone water contents. However, before this method can be instituted, it is essential to determine the ability of a SVAT model to simulate vadose zone soil water contents. A study was designed to evaluate the simultaneous heat and water (SHAW) model by comparing its soil water predictions with measured soil water contents collected by electrical resistance sensors (ERS) during the Monsoon '90 multidisciplinary field experiment. ERS collected hourly soil water measurements at 5-, 15-, and 30-cm depths in a shrub-dominated site [Larrea tridentada (Sessé & Moc. ex DC.) Coville] with large bare interspace areas. Data collected by the ERS were calibrated to time domain reflectometer (TDR) sensor measurements placed adjacent to the ERS using an in situ calibration technique. Results indicated that the SHAW model overestimated soil water at each depth by 0.02 m3 m-3 under bare soil and underestimated soil water at each depth under shrub cover by 0.02 m3 m-3. The ability of the model to simulate ERS water content values gives it the potential to be periodically updated with remotely sensed data to predict vadose zone soil water content over large areas at high temporal resolutions.

Abbreviations: CI, confidence interval • ERS, electrical resistance sensors • MBE, mean bias error • MPD, mean percentage difference • RMSE, root mean square error • SAR, synthetic aperture radar • SHAW, simultaneous heat and water [model] • SVAT, soil–vegetation–atmosphere transfer [model] • TDR, time domain reflectometer • {theta}v, volumetric water content


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and methods
 Theory
 Results and discussion
 Discussion
 Conclusions
 REFERENCES
 
SOIL WATER PLAYS A KEY ROLE in the transfer of energy and mass between land surfaces and the atmosphere, rivers, and aquifers (Blyth et al., 1993). In fact, the spatial and temporal distribution of soil water is a critical part of many disciplines including agriculture, forest ecology, hydroclimatology, civil engineering, water resources, and ecosystem modeling (Houser, 1996). Unfortunately, difficulties with soil water measurement techniques have made long-term regional data sets difficult to compile. Therefore, it is important to define an approach to monitor, characterize, and model soil water over a wide range of temporal and spatial scales (Islam and Engman, 1996).

Long-term soil water data sets are rare because many soil water sampling techniques (e.g., gravimetric, neutron probe, and time domain reflectometry) can be labor intensive and difficult to learn without appropriate training. Gravimetric measurements are simple and accurate, but are destructive and require at least 24 h of post-processing. Traditional time domain reflectometer (TDR) sensors yield accurate measurements with calibration, but are expensive and take spatially discrete measurements. Neutron probes are nondestructive and can sample over great depths, but are expensive and potentially hazardous without appropriate training. Fiberglass electrical resistance sensors (ERS) connected to automated dataloggers, however, can record temporally continuous data with little maintenance over long time periods. Like many soil water sensors, ERS require a calibration to convert a measured signal to volumetric water content. Seyfried (1993) documented two difficulties with calibrating these sensors: (i) They exhibit large variability among individual sensors; and (ii) They are sensitive to variations in soil properties and temperature. This same study, however, demonstrated that a field calibration of ERS to TDR sensors accounted for site-specific soil conditions and yielded reasonable estimates of soil water (±0.05 m3 m-3).

Hydrologic models can also be used to estimate soil water at various spatial and temporal resolutions. In particular, soil–vegetation–atmosphere transfer (SVAT) models are gaining attention as a means of better representing interactions between the soil and atmosphere. One of these SVAT schemes, the simultaneous heat and water (SHAW) model, is a detailed physical-process model capable of simulating the effects of a multispecies plant canopy on heat and water transfer at the soil–atmosphere interface (Flerchinger, 1987). Unfortunately, numerous studies have shown that SVAT models in general are subject to errors as a result of simplified model physics and complicated meteorological and site-characteristic inputs (Houser, 1996; Henderson–Sellers, 1996; Desborough et al., 1996; Lau et al., 1996). Consequently, SVAT models do not always produce reliable soil water estimates. Furthermore, these models usually require extensive meteorological and site-characteristic input parameters that can be difficult to acquire.

Because both in situ soil measurements and SVAT models are problematic, no large-scale soil water information exists at the spatial and temporal resolutions required to investigate how soil water can influence various land and atmospheric processes. Recent studies have proposed that images from synthetic aperture radar (SAR) sensors can be used to map spatially distributed soil water patterns within 5 cm of the surface (Engman and Chauhan, 1995; Sano, 1997). This method would provide a way to estimate soil water at large-scale resolutions not achievable with in situ measurements or modeling methods. Unfortunately, many studies require vadose zone soil water measurements rather than surface soil water measured by the SAR sensor. Furthermore, soil water information from orbiting SAR sensors would only be available at the time of the satellite overpass, which could be as infrequent as once a month.

By combining SAR-derived surface soil water maps with a SVAT scheme, it may be possible to obtain spatially distributed, temporally continuous information on vadose zone soil water measurements. That is, using techniques described by Sano (1997) and Moran et al. (1998), SAR images may be converted into surface soil water maps suitable for SVAT processing. Using these data along with required model inputs (meteorological, initialization, and site parameters), one may utilize the SVAT scheme to model soil water at depth, between and during satellite overpasses. At each subsequent satellite overpass, model parameters may be reinitialized according to the newly acquired SAR-derived soil water data, ultimately yielding a large-scale, temporally continuous soil water data set. In theory, this synthesis should result in less error than with remotely sensed data or the SVAT model used alone.

The first step in developing such a combined approach is to investigate the accuracy and precision of a SVAT model to estimate surface and vadose zone soil water over time. In this study, we calibrated a network of in situ ERS to create a long-term, temporally continuous soil water data set that could be used to study the SHAW model. Two objectives of this research were to (i) develop a 12-mo, hourly soil water data set at 5-, 15-, and 30-cm depths under three bare and three shrub-cover surfaces, and (ii) study the relationship between measured and predicted soil water contents.


   Materials and methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
Theory
 Results and discussion
 Discussion
 Conclusions
 REFERENCES
 
Site Description
Soil water data were collected from the Lucky Hills subwatershed in the Walnut Gulch Experimental Watershed located in the vicinity of Tombstone, AZ (Renard et al., 1993). The Lucky Hills subwatershed was instrumented in 1990 by the USDA–ARS as part of the interdisciplinary Monsoon '90 field campaign for continuous measurement of local energy conditions and surface energy balance (Kustas and Goodrich, 1994) (Fig. 1) . Standard measurements included soil water, soil temperature, soil heat flux, relative humidity, incoming solar radiation, net radiation, wind speed, and wind direction (Stannard et al., 1994).



View larger version (45K):
[in this window]
[in a new window]
  		Fig. 1 Map of the Walnut Gulch Experimental Watershed showing its geographical location in the state of Arizona

 
The Lucky Hills subwatershed is at a 1371-m elevation and has an average annual precipitation of 300 mm, 70% of which falls during the summer monsoon between July and September (Tiscareno–Lopez, 1991). This 1.46-ha area is dominated by creosote bush (Larrea tridentada) at 2- to 5-m spacing (26% cover) with surface rock percentages ranging from 0 to 46% (Kustas and Goodrich, 1994). Larger creosote shrubs are about 1-m tall and can be characterized by a spatially averaged leaf-area index value of 0.4 (Flerchinger et al., 1998). The dominant soil type in the subwatershed is the Lucky Hills series with a very gravelly sandy-loam texture (Table 1) . Rock content of the soil profile is 28% by volume between 0 and 5 cm and then decreases with depth.


View this table:
[in this window]
[in a new window]
  		Table 1 Description of physical characteristics of the Lucky Hills soil unit.{dagger}

 
Sensor Placement and Description
Eighteen pairs of ERS and TDR probes were installed horizontally into trench faces under three bare and three shrub-covered surfaces at 5-, 15-, and 30-cm depths according to procedures outlined by Bach (1991). Detailed descriptions of each trench location and its respective surface vegetation and profile characteristics are also presented. ERS used in this study were similar to those described by Colman and Hendrix (1949) and identical to those of Amer et al. (1994). Each 4- by 4-cm ERS contains several fiberglass layers wrapped around stainless-steel screens that are connected by wire leads to an automated datalogger. Original laboratory- and site-calibration procedures for ERS used in the experiment were described by Amer et al. (1994). Three-pronged TDR probes constructed by the USDA–ARS were placed adjacent to the ERS at each depth in the trenches (Bach, 1991). Dataloggers recorded hourly ERS values and ARS scientists collected TDR samples in the field at time intervals varying from daily to biweekly between August 1990 and July 1991. ERS readings were stored as a series of resistances ({Omega}) while TDR readings were calibrated to yield the volumetric water content ({theta}v) of the soil. ERS readings were averaged, respectively, using data collected from three trenches under bare soil and three trenches under creosote cover. In this study, it was assumed that TDR measurements represent the actual water content of the soil.

Calibration
Seyfried (1993) found a nonlinear relationship between TDR-measured {theta}v and ERS resistance values, as did Amer et al. (1994). Based on these analyses, a modified nonlinear calibration equation (Eq. [1]) was developed with the form

(1)
where Y is TDR measured {theta}v, a and b are parameters to be optimized, and X is ERS-measured resistance ({Omega}). This expression (Eq. [1]) was modified from the original calibration expression used by Amer et al. (1994) by eliminating an additive coefficient so that each parameter, when optimized, was significantly different from zero. This transformation was designed so that the model would be more sensitive to water content changes at lower resistance values. Individual in situ field calibrations were completed for all 18 ERS–TDR pairs by optimizing a and b parameters using nonlinear curve fitting techniques with a statistical software package. TDR and ERS measurements collected at identical times between August 1990 and July 1991 were used in each calibration. The average sample size used in each calibration was 78, but ranged from 46 to 103, due to instrumentation problems and measurement errors. Statistical tests were used to show that the a and b parameter values were significantly different from zero ( and ).

In this analysis, it was assumed that the variability in {theta}v measurements could be attributed to the calibration errors of the TDR and ERS. Bach (1991) described the installation and calibration procedures of the TDR probes used in this experiment. These calibrations did not account for temperature fluctuations that could affect TDR readings. Halbertsma et al. (1994) found that TDR readings are only slightly influenced by changing temperature and salinity conditions; and Dalton (1992) found that TDR readings are independent of temperature in soil textures finer than sand. Without considering these limitations, calibrations revealed that for the 95% confidence interval (95% CI), the root mean square error (RMSE) of the TDR measurements was 0.02 m3 m-3. Ultimately, the total error in each calibration was quantitatively defined as the square root of the sum of squares for the TDR and ERS RMSE values (Eq. [2]). Quantile plots indicate that residuals for the TDR and ERS showed no serious departure from normality.

(2)


   Theory
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Theory
Results and discussion
 Discussion
 Conclusions
 REFERENCES
 
The Simultaneous Heat and Water Model
The SHAW model is a detailed process model that simulates heat and water movement through a plant residue–soil system (Flerchinger, 1987). A vertical, one-dimensional profile extending from the vegetation canopy to a specified depth within the soil is represented by this model. At various points in this vertical profile, user-defined nodes represent various canopy and soil layers. For each of these nodes, interrelated water, water vapor, heat, and solute fluxes are calculated.

Daily or hourly weather conditions above the upper boundary and soil conditions at the lower boundary define heat and water fluxes into the system (Flerchinger, 1987). Specifically, air temperature, wind speed, relative humidity, and solar radiation define upper-boundary conditions while temperature and water contents at the bottom of the soil profile define lower-boundary conditions. Other inputs required by the model include general site parameters, textural and physical properties of the soil, and initial temperature and water-content values. Model outputs include a summary of water balance, in addition to temperature, water, and solute profiles.

The SHAW model uses a modified form of Richards equation to calculate {theta}v at each node, based on water-content values defined at the lower boundary (Flerchinger, 1987). At any frequency, user-defined {theta}v values may be used to update lower boundary conditions to improve simulation accuracy. In each case, linear interpolation is used to predict {theta}v values between lower boundary {theta}v intermediate values.

Model Testing
Two separate trials were used to simulate hourly {theta}v under bare and shrub surface cover at 5-, 15-, and 30-cm depths. In both cases, many site parameters were used that were identical to those used in previous simulations completed by Flerchinger et al. (1998). Specific textural and hydraulic soil parameters, however, were modified to match trench data measured directly at the location of the soil water sensors under bare soil and shrub cover (Table 2) . Surface and root plant parameters were used in shrub-cover simulations only and were defined by Flerchinger et al. (1998). Bare-soil simulations did not include root-distribution parameters because no discrete measurements could be included. Hourly meteorological data inputs were collected from meteorological-energy flux (METFLUX) towers located in the subwatershed (Kustas and Goodrich, 1994). Initial and final soil temperature readings for the simulation period were recorded by thermocouples installed adjacent to ERS and TDR probes in the soil at 5-, 15-, and 30-cm depths. Because soil temperature values have only a minor effect on modeled soil water content values, intermediate temperature values were not used in these simulations. Finally, {theta}v values taken by TDR probes at the same depths were used as initial and intermediate model inputs. Specifically, 13 intermediate lower-boundary (30 cm) {theta}v values at approximately 30-d intervals were used to update model simulations.


View this table:
[in this window]
[in a new window]
  		Table 2 Original{dagger} and revised{ddagger} soil textural and hydrologic parameters at the Lucky Hills soil unit

 
Regression analysis was used as a tool to evaluate the ability of the SHAW model to simulate the respective averages of calibrated, hourly ERS {theta}v values at 5-, 15-, and 30-cm depths under bare soil and shrub cover. Furthermore, statistical discrepancies between measured and simulated hourly {theta}v values were evaluated by calculating the mean bias error (MBE), mean percentage difference (MPD) and RMSE. Definitions for each are given in Table 3 . Finally, hourly measured and simulated {theta}v values were plotted as a time series to analyze how well the SHAW model could simulate water profile dynamics.


View this table:
[in this window]
[in a new window]
  		Table 3 Description and definition of model performance measures.{dagger}

 

   Results and discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Theory
 Results and discussion
Discussion
 Conclusions
 REFERENCES
 
Calibration Statistics
Optimized parameter (Eq.[1]) and coefficient of determination (R2) values for all 18 ERS–TDR probes are given in Table 4 . In all cases, t tests indicated that optimized a and b parameters were significantly different from zero, demonstrating that the calibration function effectively translated resistance values into {theta}v without yielding an average water content. Data presented in Fig. 2 show a calibration curve with matched TDR and ERS values . Clearly, resistance values approach zero as {theta}v increases. In some cases, the calibration functions showed considerable scatter, especially at higher resistance values (>1000 {Omega}). Each calibration function, however, accounted for this scatter at high resistances by yielding a constant {theta}v value above 1000 {Omega}.


View this table:
[in this window]
[in a new window]
  		Table 4 Optimized parameter (a and b) and coefficient of determination (R2) values for individual electrical resistance sensors

 


View larger version (13K):
[in this window]
[in a new window]
  		Fig. 2 Calibration curve for electrical resistance sensors . Points represent matched time domain reflectometer and electrical resistance sensor measurements used to derive the calibration curve represented by the solid line

 
Coefficient of determination (R2) values for individual sensor calibrations at all three depths ranged from 0.33 to 0.88 (mean ) (Table 4). Wide ranges of R2 values at each depth and in various trenches indicate that calibration variability was the result of individual sensors and not soil differences. These findings are consistent with the calibration results reported by Seyfried (1993). Calibration parameters were applied to each of the 18 individual ERS. In all cases, estimated {theta}v values below the calculated residual water content of the soil were adjusted to a threshold value of 3.4% (Woolhiser et al., 1990).

Calibration Error
Individual sensor calibration errors were estimated as the square root of the sum of the squares for the TDR and ERS RMSE values (95% CI) (Table 5) (Eq. [2]). The total average calibration error for all 18 sensors was ±0.029 m3 m-3 and ranged between 0.023 and 0.039 m3 m-3 for individual sensors. These findings are similar to those of Seyfried (1993), who found an average error of 0.05 m3 m-3 in a similar calibration experiment.


View this table:
[in this window]
[in a new window]
  		Table 5 Estimated calibration error for individual electrical resistance sensors (ERS){dagger}

 
Model Evaluation: Bare Soil
Data presented in Table 6 summarize the results of statistical analyses used to evaluate the performance of the SHAW model under bare soil. In this case, the MBE between measured and simulated {theta}v gradually decreased from 0.04 m3 m-3 at a 5-cm depth to 0.01 m3 m-3 at a 30-cm depth. This trend, however, was expected because it was consistent with the SHAW model structure. That is, the SHAW model calculates surface and intermediate {theta}v values based on lower-boundary measurements updated in the simulation. MPD and RMSE values reflected this same trend, indicating that the model did not simulate {theta}v well at shallower depths.


View this table:
[in this window]
[in a new window]
  		Table 6 Calculated mean bias error (MBE), mean percentage difference (MPD), root mean square error (RMSE), and coefficient of determination (R2) statistics for electrical resistance sensors and Simultaneous Heat and Water model water content values under bare soil and shrub cover

 
The scatterplot of measured and simulated {theta}v values at 5 cm under bare soil produced a bimodal distribution (Fig. 3) . The bimodal distribution in this plot can be characterized by temporal association. That is, the lower group of matched points falls below the 1:1 line because it shows a period in the driest part of the summer season where the SHAW model is underestimating {theta}v. Conversely, the highest cluster on the scatterplot indicates that the model consistently overestimates {theta}v in moist conditions. This trend is evident when the data are plotted on a time series (Fig. 4) .



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 3 Scatterplot of predicted Simultaneous Heat and Water model volumetric water content values vs. calibrated electrical resistance sensor volumetric water content measurements at 5- and 15-cm depths under bare soil

 


View larger version (27K):
[in this window]
[in a new window]
  		Fig. 4 Predicted Simultaneous Heat and Water model volumetric water content values and calibrated electrical resistance sensor volumetric water content measurements at 5- and 15-cm depths under bare soil plotted as an hourly time series (August 1990–July 1991)

 
Statistical analysis indicates that model simulations improved at 15 cm. MBE, MPD and RMSE values were approximately half as large as those at the 5-cm depth. The scatterplot (Fig. 3) and time series (Fig. 4) of the calibrated ERS and SHAW {theta}v values at 15 cm under bare soil reveal that the SHAW model overestimates {theta}v throughout the entire simulation period.

At the lower-boundary layer (30 cm), the SHAW model simulated {theta}v values well. However, {theta}v values at the lower boundary were calculated by using linear interpolation between intermediate {theta}v values used as model inputs, and therefore, are expected to be nearly identical to measured {theta}v values.

Model Evaluation: Shrub Cover
The results of statistical analyses used to evaluate the performance of the SHAW model under shrub cover are summarized in Table 6. At each depth, SHAW model {theta}v values underestimated calibrated ERS {theta}v values. MBE, MPD, and RMSE values were lower, on average, than those calculated in the bare-soil simulations but were distributed differently. In this case, MBE and MPD values at the 15-cm depth were higher than those at 5 cm. Similar to the results for bare-soil simulations, MBE and MPD values were lowest at the lower-boundary layer. A possible explanation for the differences between measured and simulated {theta}v at 5 and 15 cm may be related to root distribution parameters defined as model inputs. That is, root extraction of soil water at both depths may have been too high, ultimately causing the model to underestimate {theta}v. A second possible explanation for the discrepancy between simulated and measured {theta}v may be due to a positive bias in transpiration demand (Jones, 1983). Model estimates of transpiration may have been too high, causing the model profile to lose water, ultimately yielding low {theta}v values.

Data presented in Fig. 5 show the scatterplot between measured and simulated {theta}v values at 5 cm under shrub cover. In this case, at higher values, the SHAW model overestimated {theta}v, while at lower values, the SHAW model underestimated {theta}v. When these data are plotted as a time series, the high peaks in this graph show where SHAW overestimated {theta}v, while the driest periods (DDOY 440–530) indicate where SHAW underestimated {theta}v (Fig. 6) .



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 5 Scatterplot of predicted Simultaneous Heat and Water model volumetric water content values vs. calibrated electrical resistance sensor volumetric water content measurements at 5- and 15-cm depths under creosote cover

 


View larger version (27K):
[in this window]
[in a new window]
  		Fig. 6 Predicted Simultaneous Heat and Water model volumetric water content values and calibrated electrical resistance sensor volumetric water content measurements at 5- and 15-cm depths under creosote cover plotted as an hourly time series (August 1990–July 1991)

 
The scatterplot and time series of measured and simulated {theta}v values at the 15-cm depth under shrub cover are presented in Fig. 5 and 6, respectively. These data indicate that during inter-storm periods throughout the year, SHAW {theta}v estimates were substantially lower than measured {theta}v values. Once again, inaccurate root distribution, a positive bias in transpiration demand, and soil hydraulic properties may be responsible for these discrepant results.

At the lower-boundary layer (30 cm), the SHAW model simulated {theta}v values well. However, just as in the bare-soil simulations, {theta}v values at the lower boundary are calculated by using linear interpolation between intermediate {theta}v values used as model inputs and are expected to be nearly identical.

Model Uncertainty
Simulations under bare soil and shrub cover both indicate that parameter calibrations would improve model estimations of {theta}v. Therefore, a sensitivity analysis was performed to examine the response of simulated {theta}v with respect to 10% parameter adjustments. That is, by varying individual parameters while holding all others constant, the sensitivity of the SHAW model to individual parameters was determined. In particular, bulk density, saturated hydraulic conductivity, surface roughness, saturated water content, root distribution percentages, and porosity were analyzed. Data presented in Tables 7 and 8 show the change in {theta}v for the entire simulation, based upon parameter adjustments under bare soil and shrub cover, respectively.


View this table:
[in this window]
[in a new window]
  		Table 7 Model parameter sensitivity analysis for simulations completed under bare soil. {Delta} Input is the change in the input value expressed as a percentage of the actual value. The Change from actual value = | new value - original value |

 

View this table:
[in this window]
[in a new window]
  		Table 8 Model parameter sensitivity analysis for simulations completed under shrub cover. {Delta} Input is the change in the input value expressed as a percentage of the actual value. The Change from actual value = | new value - original value |

 
In both simulations, distinct peaks that overestimated measured {theta}v were apparent immediately after rainstorm events (Fig. 4 and 6). These peaks suggest that parameters identifying soil porosity, saturated conductivity, or saturated water content may have been too high. In each case, with sufficiently large precipitation, the model will saturate the soil layers to the saturated surface porosity, ultimately yielding peaks that overestimate {theta}v. Sensitivity analyses indicated that a 10% increase in the porosity index value could change {theta}v for the simulation by as much as 0.005 m3 m-3 (Table 8).

Shrub-cover simulations underestimated {theta}v values at 5- and 15-cm depths during inter-storm periods. A possible source of these discrepancies may be attributed to root-distribution parameters rather than the soil-hydraulic properties highlighted in Tables 7 and 8. If the root-distribution values were too high, water extraction by the plant roots would be too large, ultimately producing unrealistically low {theta}v values. Sensitivity analyses indicate, however, that a 10% adjustment will change {theta}v by only 0.001 m3 m-3. Therefore, only major adjustments to root-distribution parameters will significantly affect modeled {theta}v.

As one may surmise, changing a combination of hydraulic and physical parameters, rather than a single parameter, would be the best way to calibrate the SHAW model. It is clear from these analyses that peaks and drying rates would be influenced by a combination of many hydrologic parameters including porosity, saturated hydraulic conductivity, root distribution, and bulk density.


   Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Theory
 Results and discussion
 Discussion
Conclusions
 REFERENCES
 
In future studies, it will be critical to examine the physical structure of the SHAW model to determine if it could be linked with remotely sensed data. That is, one must determine if the model could be modified to use remotely sensed soil water values to update and predict {theta}v at various depths. Currently, the model structure is designed so that {theta}v values calculated at nodes in the user-defined soil profile are directly influenced by water content changes at the lower boundary. Therefore, major modifications must be completed so that the SHAW model will calculate the soil water profile based on {theta}v values defined at the surface by remotely sensed inputs.


   Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Theory
 Results and discussion
 Discussion
 Conclusions
REFERENCES
 
The dual objectives of this study were to produce a complete, 12-mo soil water data set at three depths, and to use these data to investigate the accuracy and precision of SHAW model surface and vadose zone soil water estimates.

Individual, in situ calibrations of ERS resistance values with TDR {theta}v measurements worked well. An hourly soil water data set for a 12-mo period was composed for six replicate sites at three different depths. This simple calibration procedure demonstrated the utility of the ERS for use in future soil water studies. That is, high-frequency ERS measurements, although inaccurate, were easily calibrated with accurate, low-frequency TDR measurements to yield a large, temporally continuous {theta}v data set.

In general, the SHAW model simulated annual and diurnal {theta}v patterns well (Fig. 4 and 6). However, under bare soil, {theta}v values were consistently overestimated while under shrub cover, {theta}v values were consistently underestimated. These simulation errors may be attributed to errors in model dynamics, model parameters, or error in sensor calibration. In any case, model calibration and modification will be required so that remotely sensed {theta}v values may be assimilated as primary inputs.

With appropriate modifications, the SHAW model may be used to model soil water at depth, between and during satellite overpasses. Ideally, SHAW-model parameters could be initialized by SAR-derived surface soil water estimates to yield a large-scale, temporally continuous soil water data set. Ultimately, this synthesis would provide an effective way to monitor, characterize, and model soil water over a wide range of temporal and spatial scales.


   ACKNOWLEDGMENTS
 
The authors acknowledge Gerald Flerchinger, Dave Toll, Paul Houser, Mike Jasinski, Leslie Bach, Saud Amer, Bill Kustas, Gary Richardson, Jim Toth, Soroosh Sorooshian, Chawn Harlow, Jim Shuttleworth, Dave Goodrich, staff at the Water Conservation Laboratory (Phoenix, AZ), and the Tombstone field office staff for their efforts. In particular, Ted Engman, Carl Unkrich, Ginger Paige, and Scott Miller deserve thanks for their guidance and support. The authors appreciate the comments and suggestions of the anonymous reviewers. This research was funded, in part, by NASA EOS (NASA NAGW–2425), the Landsat Science Team (NASA S–41396–F), and the Monsoon '90 project (IDP–88–086).

Received for publication August 5, 1998.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and methods
 Theory
 Results and discussion
 Discussion
 Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
V. Sheikh and E. E. van Loon
Comparing Performance and Parameterization of a One-Dimensional Unsaturated Zone Model across Scales
Vadose Zone J., August 23, 2007; 6(3): 638 - 650.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
S. A. Amidu and J. A. Dunbar
Geoelectric Studies of Seasonal Wetting and Drying of a Texas Vertisol
Vadose Zone J., August 1, 2007; 6(3): 511 - 523.
[Abstract] [Full Text] [PDF]

Home page
 Soil Sci.Home page
X. L. Xin, F. A. Xu, J. B. Zhang, and M. X. Xu
A New Resistance Sensor for Monitoring Soil Matric Potential
Soil Sci. Soc. Am. J., May 16, 2007; 71(3): 866 - 871.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
G. N. Flerchinger, M. S. Seyfried, and S. P. Hardegree
Using Soil Freezing Characteristics to Model Multi-Season Soil Water Dynamics
Vadose Zone J., October 3, 2006; 5(4): 1143 - 1153.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
W. Xiao, Q. Yu, G. N. Flerchinger, and Y. Zheng
Evaluation of SHAW Model in Simulating Energy Balance, Leaf Temperature, and Micrometeorological Variables within a Maize Canopy
Agron. J., May 3, 2006; 98(3): 722 - 729.
[Abstract] [Full Text] [PDF]

Home page
 Soil Sci.Home page
J. P. Walker and P. R. Houser
Evaluation of the OhmMapper Instrument for Soil Moisture Measurement
Soil Sci. Soc. Am. J., May 1, 2002; 66(3): 728 - 734.
[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 Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (15)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Hymer, D.C.
Right arrow Articles by Keefer, T.O.

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