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


     

Published online 29 October 2007
Published in Soil Sci Soc Am J 71:1902-1908 (2007)
DOI: 10.2136/sssaj2007.0011
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow 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 Jiang, P.
Right arrow Articles by Sadler, E. J.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Jiang, P.
Right arrow Articles by Sadler, E. J.
Agricola
Right arrow Articles by Jiang, P.
Right arrow Articles by Sadler, E. J.
Related Collections
Right arrow Water Content
Right arrow Water Conservation

SOIL & WATER MANAGEMENT & CONSERVATION

Estimating Plant-Available Water Capacity for Claypan Landscapes Using Apparent Electrical Conductivity

Pingping Jianga,*, Stephen H. Andersonb, Newell R. Kitchenc, Kenneth A. Sudduthc and E. John Sadlerc

a Dep. of Environmental Sciences, 2323 Geology Bldg., Univ. of California, Riverside, CA 92521
b 302 ABNR Bldg., Dep. of Soil Environmental and Atmospheric Sciences, Univ. of Missouri, Columbia, MO 65211
c 269 Agricultural Engineering Bldg., USDA-ARS, Cropping Systems and Water Quality Research Unit, Columbia, MO 65211

* Corresponding author (pingping.jiang{at}ucr.edu).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Information on plant-available water (PAW) capacity (PAWc) variation within a field is useful for site-specific management. For claypan soils, established relationships between soil apparent electrical conductivity (ECa) and topsoil thickness suggested the hypothesis that profile PAWc could be estimated by assuming a two-layer soil composition, a silt loam topsoil layer and a silty clay sublayer, with known PAW fraction values for each layer. Objectives were (i) to investigate the direct relationships between ECa and the upper and lower limits of PAWc, and (ii) to test the previously stated hypothesis. Nineteen and 18 soil profile samples were taken from two Missouri claypan fields in October 2005. The lower limit of PAWc was determined at –1500 kPa soil water pressure. Samples were taken again from the same locations in March 2006 to determine the upper limit of PAWc. Calculations were on a 1.2-m basis. The direct relationship between ECa–1 and profile PAW (PAW1.2) was significant, with regression r2 values of 0.67 and 0.87 and RMSEs of 30 and 20 mm for Fields 1 and 2, respectively. The RMSEs for two-layer-estimated PAW1.2 were 14 and 16 mm for Fields 1 and 2, respectively, or 7.6 and 8.6% of the respective mean measured PAW1.2. With the two-layer approach, some underestimates of PAW1.2 resulted from underestimation of topsoil thickness, whereas overestimates were attributed to soil horizons being short of field capacity at sampling due to slow recharge. The resulting field-scale PAWc information is useful in site-specific decision making for soil and water management.

Abbreviations: C, clay • ECa, bulk soil apparent electrical conductivity • LL1.2, lower limit of plant-available water for a 1.2-m soil profile • PAW, plant-available water • PAWc, plant-available water capacity • PAW1.2, plant-available water for a 1.2-m soil profile • SIC, silty clay • SICL, silty clay loam • SIL, silt loam • UL1.2, upper limit of plant-available water for a 1.2-m soil profile


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The ability of soil to store and supply water to plants is one of its fundamental properties related to crop production. Knowledge about plant-available water (PAW) capacity (PAWc) is useful for many soil management practices as well as for crop yield modeling applications. Quantitative determination of PAWc, however, is not an easy task. Determination of PAW involves determining the two limits (i.e., field capacity and permanent wilting point), which can be either monitored from field measurements (Ritchie, 1981) or approximated under laboratory conditions (Jamison and Kroth, 1958). The former requires permanent installation of soil moisture devices and repeated monitoring, while the latter involves destructive sampling and water extraction. Either way, the time-consuming nature prohibits extensive assessment of the spatial variability of this soil property for a given field or watershed. Further difficulties include the limited value of soil survey information (e.g., texture and bulk density) for estimating PAW due to potentially large errors and bias in the estimation (Fortin and Moon, 1999).

A key component of site-specific management is quantification of the spatial variability of soil properties that affect crop yields (Atherton et al., 1999). A map of PAWc would help advance management decisions such as adjusting fertilizer input and optimizing water management options. This information could also be incorporated in management zone delineation or in crop models. To meet this need, alternative approaches have been proposed (Timlin et al., 2001b; Morgan et al., 2003). Timlin et al. (2001b) used a simple water budget model to simulate yield, and then applied a procedure to match the simulation to observed yield. During the matching procedure, the amount of PAW was varied until the closest match between predicted and observed yield was found. Then available water was estimated at the closest match. By similar principles, Morgan et al. (2003) devised an inverse yield model to create a "look-up" table where corn yields were simulated at a range of PAW levels. Using this correspondence, a map of PAW could be inversely generated based on yield maps. These approaches take advantage of readily available yield data made possible through yield-mapping technologies and assume that observed yield is only affected by PAW. Through inherent biophysical relationships between crop yield and water balance, the yield information can be transferred into PAW information.

Apparent profile soil electrical conductivity (ECa) has become an important tool in site-specific management practices because it relates to a wide range of soil chemical and physical properties that affect crop yield (McNeill, 1992; Lund et al., 2001; Kitchen et al., 2003; Sudduth et al., 2005). Applications of mapped ECa have included characterizing soil spatial variability (Corwin and Lesch, 2005) and delineating management zones (Kitchen et al., 2005; Jaynes et al., 2005). The direct regression relationships between ECa and PAW, however, have been examined by only a few (Morgan et al., 2000; Wong et al., 2006), even though high and consistent correlations between ECa and soil water content on nonsaline soils have been reported by many (Kachanoski et al., 1988, 1990; Sheets and Hendrickx, 1995; Khakural et al., 1998; Sudduth et al., 2001; Reedy and Scanlon, 2003).

For claypan soil landscapes in the U.S. Midwest, the contrasting electrical properties of the claypan and the overlying topsoil have lent ECa a unique utility of estimating and mapping spatially variable topsoil thickness above the claypan horizon (Doolittle et al., 1994; Sudduth et al., 2003; Sudduth and Kitchen, 2006). Topsoil thickness has been found to highly correlate with crop yield, especially in dry growing seasons (Gantzer and McCarty, 1987; Kitchen et al., 1999), because it serves as a crucial reservoir for PAW and nutrients and provides a suitable rooting environment for plants (Timlin et al., 2001a). Compared with the topsoil, the claypan horizon has a substantially lower PAWc due to high clay content (usually >50%), low organic matter, and poorly developed structure (Jamison and Kroth, 1958). Yet roots of annual crops, e.g., corn (Zea mays L.) and soybean [Glycine max (L.) Merr.], can penetrate through the claypan down to a depth of 1.35 m (Grecu et al., 1988; Myers et al., 2007). These characteristics of claypan soils led to the formulation of our study hypothesis: the maximum PAWc can be approximated with a hypothetical two-layer soil profile comprised of a topsoil layer (usually silt loam in texture) and a sublayer (silty clay or clay in texture) to the bottom of the rooting depth. The proposed procedure, if proven, would provide quick and inexpensive PAWc estimates at high spatial resolution because topsoil thickness can be estimated by ECa. The mathematical representation for this hypothesis is given as:

Formula 1[1]
where PAWc is the profile PAW capacity to the bottom of a presumed rooting depth; Ttopsoil and Tsubsoil are thicknesses of the topsoil layer and sublayer, respectively (Tsubsoil is obtained by subtracting Ttopsoil [estimated using ECa] from the rooting depth); and PAWSIL and PAWSIC are PAW fraction values for the two soil textures obtained from the USDA-NRCS soil survey (Young et al., 2001; NRCS staff, personal communication, 2006).

Thus, the specific objectives of this study were to: (i) investigate direct relationships between ECa and profile PAW and its upper and lower limits, and (ii) test how well the hypothesis expressed in Eq. [1] can approximate profile PAW at a field scale.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Study Sites
Study sites were two claypan soil fields, Field 1 (39°38'N, 92°20'W) and Field 2 (39°38'N, 92°25'W), located within 2 km of each other near Centralia in central Missouri. Field 1 was 28 ha and Field 2 13 ha in size. Elevation ranged from 262 to 266 m in Field 1 and from 256 to 266 m in Field 2. The primary soil series found in the study fields included Mexico (fine, smectitic, mesic Aeric Vertic Epiaqualfs), Adco (fine, smectitic, mesic Aeric Vertic Albaqualfs), both with 1 to 5% slope, and Leonard (fine, smectitic, mesic Vertic Epiaqualfs), with 2 to 14% slope. All these soil series are somewhat poorly or poorly (i.e., Leonard) drained (Soil Survey Staff, 2006). They are typical claypan soils characterized by an abrupt claypan horizon at varying depths, depending generally on slope and landscape position. The depth to claypan ranged from several centimeters in eroded areas to >1 m in depositional areas. The texture above the claypan ranged from the typical silt loam texture to an occasional silty clay loam texture. Both fields had been managed in a corn–soybean rotation with mulch tillage for about 20 yr. No-till was initiated in 2004 on Field 1 and in 1997 on Field 2. The mean annual temperature in the area is 12°C, and the mean annual precipitation is 96.9 cm (National Climate Data Center, 2002).

Sampling Procedures and Laboratory Analyses
Profile samples were taken at 19 locations in Field 1 and 18 locations in Field 2 in October 2005 using a hydraulic soil coring probe (38.1-mm diameter). The sampling sites were distributed throughout the fields such that major landscape features were represented. Soil properties and characteristics (e.g., topsoil thickness, horizon designation, and horizon texture) were already available at these sites as they had served as calibration sites for other research projects (e.g., Sudduth et al., 2003, 2005). The texture data indicated that there were four textural classes found at these sites: silt loam (SIL), silty clay loam (SICL), silty clay (SIC), and clay (C). Topsoil was considered as those soil horizons above the claypan whose texture was silt loam or, occasionally, silty clay loam. For sites where the surface texture was silty clay, topsoil thickness was considered zero (i.e., high-erosion areas). During the sampling for this investigation, profile horizons were reexamined guided by the original designation. Horizon lengths were recorded, and then soil profiles were separated by horizon and each horizon sample was collected and sealed in a plastic bag. These horizon samples were air dried for 2 wk before an air-dry weight was obtained. A subsample of about 50 g was oven dried to determine water content for the air-dry horizon samples. Thus, bulk density for each horizon was calculated using air-dry soil mass, water content of the air-dried subsample, and sample volume. Bulk density was used to convert gravimetric water content to volumetric water content. The remaining samples were broken, and small aggregates were used to determine water retention at –33 kPa. Further, sample material passed through a 2-mm sieve was used to determine water retention at –1500 kPa, which was used as the lower limit (LL) of PAWc. Water retention was determined using pressure chambers (Dane and Hopmans, 2002).

The same sites were resampled on 29 Mar. 2006, following wintertime profile recharge, to determine field capacity, which was used as the upper limit (UL) of PAWc. An 11-mm rainfall was recorded 2 d before the sampling. Sampling procedures followed those of the October sampling, using the same horizon designations and depths. There was a cumulative 19-cm deficit from normal precipitation during the recharge months (October–March; National Climate Data Center, 2002). To ensure the soil condition was as close to field capacity as possible before sampling, several test samples were taken approximately 2 wk before to compare with historical neutron probe moisture data that had been collected from some of the sampling sites at the beginning of June 1997 (seven sites in Field 1) and 1999 (five sites in Field 2) after profile recharge. We judged these neutron data to represent field capacity conditions, especially at deeper depths, because precipitation leading up to the measurement dates was 11 and 18 cm above normal for 1997 and 1999, respectively (from the previous October–May). The average water content measured by the neutron probe on a 1.2-m profile basis was 495 and 483 mm for Fields 1 and 2, respectively. A good comparison was obtained between the neutron-probe field capacity determination and the preliminary sampling in mid-March 2006.

For the actual samples, PAW was determined by the difference between the UL and LL values for each horizon. Profile upper limit (UL1.2) and lower limit (LL1.2) were obtained as a depth-weighted average of soil horizons to a 1.2-m depth. Profile PAW (PAW1.2) was then the difference between the UL1.2 and LL1.2.


   DATA ANALYSIS PROCEDURES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Validation of Estimation Equations for Topsoil Thickness
Our previous research determined regression relationships of soil ECa to topsoil thickness for the two fields (Sudduth et al., 2003; Sudduth and Kitchen, 2006). Soil ECa data used to develop these relationships were collected at different times of the year during multiple years using several types of commercial ECa sensors. The sensors included Geonics EM38 (Geonics Ltd, Mississauga, ON, Canada), Veris 3100 (Veris Technologies, Salina, KS), and DUALEM-2S (Dualem Inc., Milton, ON, Canada). The EM38 had a vertical dipole and a horizontal dipole with respective effective sensing depths of 1.5 and 0.75 m. The Veris 3100 used rolling coulter electrodes to directly sense both shallow (0.3-m effective sensing depth) and deep (1.0-m effective sensing depth) readings of ECa. The Dualem-2S sensor was designed with a single transmitter and two receivers, allowing simultaneous shallow and deep ECa readings, with respective effective sensing depths of 1.2 and 3.0 m. Additional details on the sensors and ECa data collection can be found in Sudduth et al. (2003) and Sudduth and Kitchen (2006). The regression equations of ECa vs. topsoil thickness were different due to sensor design, effective sensing depth, and variation in field conditions (e.g., moisture and temperature) when ECa data were acquired. Furthermore, the samples for the current study differed slightly in measured topsoil thickness from the original data used to develop the regression equations because of local variations in topsoil thickness and the subjectivity involved in determining the boundary of the claypan horizon using visual cues in the field. Therefore, a validation of the existing regression equations against the current measured topsoil thickness data was conducted to select the best relationship. All ECa data sets were kriged to a 5- by 5-m cell size with identical spatial extent. The ECa values from cells that contained sampling sites were used to develop the regression with measured topsoil thickness. A regression equation for each field was selected based on minimal bias between the measured and estimated topsoil thickness, standard error for the regression coefficient (β), and RMSE. The bias was tested by evaluating the hypothesis of β = 1 in the regression. The validation results showed that the DUALEM-2S sensor used in shallow mode performed the best for Field 1, and the DUALEM-2S sensor used in shallow or deep mode performed equally well for Field 2. For consistency and comparison purposes, we selected the DUALEM-2S sensor in shallow mode to estimate topsoil thickness for further analyses. The selected regression equations and the selection criteria are given in Table 1 .


View this table:
[in this window]
[in a new window]

  		Table 1. Mean measured topsoil thickness (TT), selected apparent bulk soil electrical conductivity (ECa) sensor, regression equation, and regression statistics for measured and ECa–estimated topsoil thickness.

 
Statistical Analyses
The mean distances between any two sampling sites were 363 and 244 m for Fields 1 and 2, respectively, and soil properties determined at these sampling sites were assumed spatially independent. Several statistical procedures were used in the data analyses. For each textural class, a two-sample t-test was performed between UL and water content at –33 kPa ({theta}–33), and then a one-sample t-test was used to test whether the measured PAW values were equal to the USDA-NRCS PAW values used in Eq. [1]. Both ECa and the reciprocal function, ECa–1 (Sudduth et al., 2003; Sudduth and Kitchen, 2006) were used to regress against measured UL1.2, LL1.2, and the difference between the two (i.e., PAW1.2). These ECa values were the same as those used to validate the topsoil thickness. Normal errors were assumed for these simple regression models, and therefore model residuals were tested for normality. For the two-layer profile approach, RMSE calculation and bias tests were performed for measured PAW1.2 vs. estimated PAW1.2 (Eq. [1]). Furthermore, the estimated PAW1.2 (Eq. [1]) and the measured PAW1.2 were examined against a 1:1 reference relationship. All statistical procedures were conducted using SAS software (SAS Institute, 2005), and the significance level for all statistical procedures was {alpha} = 0.05.


   RESULTS AND DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Texture Distributions and Plant-Available Water by Texture
In Field 1, the measured topsoil thickness ranged from 11 to 120 cm with an average of 34.8 cm (Table 1), and all topsoil horizons but one were SIL texture. In Field 2, the measured topsoil thickness ranged from 0 to 120 cm with an average of 40.1 cm. Seven out of the 18 sample profiles had SICL texture for the topsoil horizon, and one profile had no topsoil (i.e., SIC at the surface). The higher clay content in the surface horizons for Field 2 was an indication of more severe erosion having occurred in Field 2 than in Field 1 and a possible result of tillage mixing of the subsoil into the shallow surface horizon. Furthermore, 12 SIL horizons in Field 2 were found in the subsoil (below 40 cm), underlying SIC and SICL textures, while all SIL horizons but one in Field 1 were surface horizons. The SICL horizons were more dispersed across the depth of the profile in Field 2 than in Field 1.

Particle size distributions, UL, LL, calculated PAW, and {theta}–33 for each textural class are given in Table 2 . The two-sample t-tests indicated that the ULs for the SIL texture in both fields were significantly higher than the corresponding {theta}–33. The rain event that occurred 2 d before sampling with the somewhat poorly drained subsurface was a possible reason for this result. The UL of SICL, SIC, and C textures were all lower than the {theta}–33 in Field 1, but were all the same as the {theta}–33 in Field 2. The UL being lower than {theta}–33 in Field 1 was an indication that, on average at the time of sampling, subsurface soils had not been fully recharged during the fallow period, a result of below-normal precipitation. The fact that our UL compared well with the historical neutron probe data collected in Field 1, however, suggested that the observed UL values represented field conditions normally encountered for this type of soil. A closer agreement between the UL and {theta}–33 may have been found had the field been wet for a longer period of time to allow the subsurface to fully recharge. Between Field 1 and Field 2, PAW values for all texture classes except for the SIL were statistically the same. The PAW for the SIL was higher in Field 1 (0.250 m3 m–3) than in Field 2 (0. 219 m3 m–3), with a P value of 0.031 (data not shown). This difference stemmed from significantly higher LL for the SIL in Field 2 (0.150 m3 m–3) than in Field 1 (0.130 m3 m–3, P value = 0.028, data not shown). The higher LL value for the SIL in Field 2 was a result of the SIL horizons distributed deeper in the sample profiles. These deeper SIL horizons had lower organic matter and higher clay content than the SIL horizons found at shallower depths, hence a slightly higher LL.


View this table:
[in this window]
[in a new window]

  		Table 2. Particle size distributions, measured upper limit (UL), lower limit (LL), plant-available water (PAW), and water content at –33 kPa ({theta}–33) by textural class, and t-test results for the UL vs. {theta}–33 and for measured vs. NRCS PAW (numbers in parentheses, except for textural class, are standard deviations).

 
The PAW fraction values are also included in Table 2. These PAW fraction values were obtained by averaging the two values (a high value and a low value) given by Young et al. (2001) for a given texture. The NRCS PAW values matched well with the measured PAW for SIL in Field 2 and for SIC and C in both fields. The NRCS PAW value was lower than the measured PAW for the SIL in Field 1 and higher on average for the SICL in both fields (Table 2).

Relationships between Apparent Electrical Conductivity and the Upper and Lower Limits of Plant-Available Water Capacity
Simple regression models using ECa–1 yielded better results with all variables (UL1.2, LL1.2, and PAW1.2) than models using ECa. Thus, the results using ECa–1 are presented and discussed here. The mean and standard deviation for the UL1.2, LL1.2, and PAW1.2 expressed in millimeters of water, as well as for ECa–1, are given in Table 3 . The regression coefficients of ECa–1 were significant for LL1.2 for both fields. The r2 values were 0.66 and 0.75 for Fields 1 and 2, respectively (Fig. 1 ).


View this table:
[in this window]
[in a new window]

  		Table 3. Basic statistics for measured upper limit (UL1.2) and lower limit (LL1.2), and calculated plant-available water capacity of a 1.2-m soil profile (PAW1.2 = UL1.2 – LL1.2), along with apparent bulk soil electrical conductivity (ECa) and ECa–1 statistics.

 

Figure 1
View larger version (12K):
[in this window]
[in a new window]

  		Fig. 1. Plots of the reciprocal of bulk soil apparent electrical conductivity (ECa–1) vs. the upper limit (UL1.2) and lower limit (LL1.2) of plant-available water for a 1.2-m soil profile, along with regression equations fit to the data and r2 values. The ECa–1 values were obtained from the kriged 5- by 5-m cell containing each sampling site. * Significant at the 0.05 level. *** Significant at the 0.001 level.

 
Soil water content is one of the chief factors affecting ECa. Kachanoski et al. (1988) reported high correlations between volumetric water content measured over a 0.5-m soil depth and ECa measured over a series of soil depths ranging from 0.5 to 6 m. Good correlations remained between one-time measured ECa and water content measurements taken over time, provided that the spatial variability of water content was relatively temporally stable (Kachanoski et al., 1990), and that potential temporal correlation among water content measurements was small enough not to impact the estimation equation (Reedy and Scanlon, 2003). The sample LL1.2 ranged from about 160 mm (~0.13 m3 m–3) to 340 mm (~0.28 m3 m–3, Fig. 1), which was consistent with the water content ranges (<0.30 m3 m–3) where highly significant relationships were reported in the literature. Because the lower limit water content was obtained at a fixed soil water pressure, however, the variation in LL1.2 was mainly caused by soil texture and horizonation, rather than by field conditions such as structure and drainage.

From the relationships between ECa–1 and topsoil thickness and between ECa–1 and LL1.2, a relationship between topsoil thickness and LL1.2 could be expected. Correlation analysis showed significant correlation coefficients between topsoil thickness and LL1.2 (–0.92 and –0.93 for Fields 1 and 2, respectively; P < 0.0001). For a Mexico soil, the amount of water retained at –1500 kPa in an Ap horizon with a SIL texture (~0.12 m3 m–3) is normally only about one-half of the amount retained in a Bt horizon with a SIC texture (~0.24 m3 m–3; Chung, 1989). Thus, the thicker the topsoil, the more water can be released before the lower limit is reached. This result explained the significant relationship between ECa–1 and LL1.2 shown in Fig. 1.

There was a small significant increase in the UL1.2 with increasing ECa–1 for Field 1 (r2 = 0.24), but no relationship was found for Field 2 (Fig. 1). Kachanoski et al. (1988) showed that the curvilinear relationship between ECa and water content, both measured over a 0.5-m soil depth, leveled off at higher water content (>0.30 m3 m–3), and the slope of the fitted curve changed to negative (which would be positive with ECa–1) when water content increased above 0.36 m3 m–3. This finding is supported by our result that ECa–1 was insensitive to UL1.2, which ranged from about 400 mm (~0.33 m3 m–3) to 510 mm (~0.43 m3 m–3) across the two fields.

Having examined how UL1.2 and LL1.2 were related to ECa–1, the relationship between the PAW1.2 and ECa–1 could be readily examined (Fig. 2 ). The regression models in Fig. 2 yielded RMSE values of 30 and 20 mm for Fields 1 and 2, respectively. With the two fields combined, the r2 value was 0.76 and RMSE was 27 mm. These results indicated that soil ECa–1 can be directly used to estimate field-variable profile PAW with certain confidence intervals once a relationship between ECa and profile PAW to a chosen soil depth is calibrated.


Figure 2
View larger version (10K):
[in this window]
[in a new window]

  		Fig. 2. Plot of the reciprocal of bulk soil apparent electrical conductivity (ECa–1) vs. measured plant-available water for a 1.2-m soil profile (PAW1.2), along with regression equations fit to the data, r2 values, and RMSEs. The PAW1.2 was calculated as the difference between the profile upper limit (UL1.2) and lower limit (LL1.2). * Significant at the 0.05 level. *** Significant at the 0.001 level.

 
Estimating Plant-Available Water Capacity with a Two-Layer Soil Profile
As presented in Table 1, there was an average RMSE of 12.0 cm for measured vs. ECa–estimated topsoil thickness for the two fields. To give an insight into how these topsoil thickness errors contribute to estimating PAW1.2 with the two-layer approach (Eq. [1]), we applied Eq. [1] to both the measured topsoil thickness and the ECa–estimated topsoil thickness and obtained two PAW1.2 estimates. Then RMSE values were calculated for the measured PAW1.2 vs. each of the two PAW1.2 estimates. Using the ECa–estimated topsoil thickness, the RMSE values were 14 and 16 mm as shown in Table 4 , which were 7.6 and 8.6% of the mean measured PAW1.2 for Fields 1 and 2, respectively. Using the measured topsoil thickness, the respective error percentages were 7.0% (13 mm) and 6.4% (12 mm) of the mean measured PAW1.2 (data not shown). The increase in error by using ECa–estimated topsoil thickness (0.6 and 2.2% for Fields 1 and 2, respectively) was considered relatively minor, confirming our assumption in Eq. [1] that ECa could be used to estimate topsoil thickness.


View this table:
[in this window]
[in a new window]

  		Table 4. Regression statistics for measured (upper limit [UL1.2] – lower limit [LL1.2]) vs. two-layer-estimated (Eq. [1]) plant-available water for a 1.2-m soil profile (PAW1.2).

 
Figure 3 plots the regression relationship between the measured PAW1.2 and the two-layer-estimated PAW1.2, along with a 1:1 line. The regression parameters and test statistics are given in Table 4. The estimated PAW1.2 ranged from 146 to 249 mm for Field 1 and from 148 to 274 mm for Field 2, smaller ranges than those for the measured values (Table 4). The regression slopes significantly deviated from the 1:1 line. The estimation procedure tended to overestimate for lower PAW1.2 values and underestimate for higher values. The data point of Field 1 indicated by an arrow in Fig. 3 had the largest residual error of 102 mm because there was a 25-cm underestimation in topsoil thickness at this sampling site located in a depositional area of the field. Thus, the estimated PAW1.2 was greatly reduced. Deposited topsoil often has higher clay content than in situ topsoil, and this higher clay content detected by the EC sensor may be partially responsible for underestimating topsoil depth. The same reason also applied to the Field 2 data point indicated by an arrow, where there was a 24-cm underestimation in topsoil thickness. The overestimation at the lower end of the regression line, however, was not attributed to topsoil thickness errors because these errors did not correlate with the PAW1.2 residual errors (graph not shown). Instead, the overestimation of PAW1.2 occurred regardless of whether the residual errors for topsoil thickness were positive or negative. This trend was probably because NRCS PAW values smoothed the variation observed in individual horizons, especially for horizons that were potentially still less than field capacity at sampling. Our soil-sampling field notes confirmed that the greatest overestimation (data points circled at the lower end of the regression line in Fig. 3) occurred at the most eroded sites, where parts of the soil profile may not have reached field capacity.


Figure 3
View larger version (8K):
[in this window]
[in a new window]

  		Fig. 3. Plot of measured vs. estimated plant-available water for a 1.2-m soil profile (PAW1.2), along with the regression line for the combined data (solid) and a 1:1 reference line (dashed). The measured PAW1.2 is the difference between the profile upper limit (UL1.2) and lower limit (LL1.2), and the estimated PAW1.2 was calculated using Eq. [1]. The arrows and circle indicate the data points with the largest underestimating and overestimating residual errors, respectively.

 
Overall, the hypothetical two-layered soil body in conjunction with NRCS PAW values and ECa–estimated topsoil thickness yielded reasonable estimates for the PAWc over a 1.2-m profile. One key factor in the success of this simplified estimating procedure was that the SIC and SICL textures, dominant textures beneath the topsoil, had similar measured PAW values. Thus, the presence of SICL would not bias the estimation even though this texture was not included in the model (Eq. [1]). The procedure tended to overestimate PAW for soil profiles with higher clay content in one or more horizons (usually eroded areas). With reduced hydraulic conductivity near the soil surface, these profiles may take much more time to recharge to field capacity than what is normally assumed.


   CONCLUSIONS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Our ultimate objective was to quantitatively determine PAWc at a field scale using soil ECa information, which can be acquired relatively quickly and inexpensively at high spatial resolutions. Two approaches were examined in this study. The simple regression model showed a significant relationship between ECa and profile PAWc. The r2 values were 0.67 and 0.87 and the RMSE values were 30 and 20 mm for Fields 1 and 2, respectively. These results were derived from the significant relationship of ECa to the lower limit of the profile PAWc, which is highly correlated with topsoil thickness.

The second approach further simplified PAWc estimation by hypothesizing a two-layer soil profile comprised of a SIL topsoil layer and a SIC subsurface layer, whose boundary can be conveniently estimated by ECa. The RMSE between the measured and two-layer-estimated PAW1.2 was 16 mm for the two fields combined. The potential of this approach is that once a good calibration is established between topsoil thickness and ECa, the map of ECa can be translated into a PAWc map. In this case, the chief error source for this method came from sample sites that did not reach field capacity. The NRCS PAW values are given as an average PAW fraction value for a given texture class and do not take into account variability caused by field factors such as recharge and drainage conditions, landscape position, and organic matter content. This, in turn, presents a potential problem in applying this approach for a claypan soil landscape, because soils at certain locations in a claypan field may practically never reach field capacity throughout the whole soil profile even in normal and above-normal precipitation years, due to slow recharge. Another drawback of this approach, due to its deterministic nature, involves the difficulty in assessing estimation errors.

In all, for similar claypan soil types, both approaches can be used as quick and cost-efficient methods to quantify within field profile PAWc with reasonable accuracy. Being aware of their advantages and disadvantages, the resulting PAWc maps can be useful for site-specific decision making with regard to soil and water management.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

Received for publication January 6, 2007.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 DATA ANALYSIS PROCEDURES
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
P. Jiang, N. R. Kitchen, S. H. Anderson, E. J. Sadler, and K. A. Sudduth
Estimating Plant-Available Water Using the Simple Inverse Yield Model for Claypan Landscapes
Agron. J., May 7, 2008; 100(3): 830 - 836.
[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 Jiang, P.
Right arrow Articles by Sadler, E. J.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Jiang, P.
Right arrow Articles by Sadler, E. J.
Agricola
Right arrow Articles by Jiang, P.
Right arrow Articles by Sadler, E. J.
Related Collections
Right arrow Water Content
Right arrow Water Conservation

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