SSSAJ Grow Your Career with SSSA
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 Google Scholar
Google Scholar
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
Related Collections
Right arrow Vadose Zone Processes and Chemical Transport
Right arrow Soil Models
Right arrow Preferential Flow
Soil Science Society of America Journal 66:686-695 (2002)
© 2002 Soil Science Society of America

DIVISION S-1—SOIL PHYSICS

Measured and Predicted Solute Leaching from Multiple Undisturbed Soil Columns

S. D. Logsdon*,a, K. E. Kellerb and T. B. Moormana

a NSTL, 2150 Pammel Dr., Ames, IA 50011-3120
b BASF, Ag. Prod. Center, 26 Davis Dr., P.O. Box 13528, Res. Triangle Park, NC 27709-3528

* Corresponding author (logsdon{at}nstl.gov)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Macro Model
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Preferential flow may cause rapid leaching of solutes. The preferential flow model MACRO is a flexible, transient-state preferential flow model. The objective of this study is to compare measured and predicted Br and tritium leaching for multiple soil columns. We collected 48 undisturbed columns (0.35 m long and 0.20 m in diameter), that represented three soils (Mollisols), two tillage treatments (no-till and chisel), and two rainfall application rates for the first rain. We applied Br to the soil 2 d before the first rain, which contained tritium. There were four rain events, each 1 wk apart with drainage and evaporation between rains. We determined effluent breakthrough curves. We determined input parameters from soil properties measured on three extra columns for each soil and tillage combination and from calibration. For the first rain event, measured Br loss was 175% larger than tritium loss. Predicted bromide leaching for the first rain was 2% smaller than tritium loss. Tritium retention could have reduced tritium leaching. Concentrations of tritium and bromide were usually overpredicted for the first flush and underpredicted for the end of the first rain event through the start of the third rain event. The root mean square error values were as much as four times as high as a treatment mean for bromide, and up to eight times as high for tritium. Lack of agreement between measured and predicted bromide and tritium leaching may suggest further important mechanisms should be incorporated into MACRO.

Abbreviations: AE, anion exclusion • CDE, convection-dispersion equation • d, half spacing between equivalent parallel fractures • Db, bulk density • Dv, dispersivity • h, pressure head • |hb|, the boundary tension between the macropore and micropore regions • K, hydrualic conductivity • Kb, K between the macropore and micropore boundaries • Ks, saturated K • n*, exponent of the K–{theta} relation for the macropore region • RMSE, root mean squared error • SA, surface area • {theta}, water content • {theta}sma, macropore volume fraction • {lambda}, micropore scale index


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Macro Model
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
MANY HAVE COMPARED predicted preferential solute leaching with measured field data (Vinten and Redman, 1990; Parrish et al., 1992; Villholth and Jensen, 1993; Jabro et al., 1994; Larsson and Jarvis, 1999; Close et al., 1999; Hendriks et al., 1999), but out of necessity the numbers of treatments and replicates were limited, often to only one. Schoen et al. (1999) compared results for a field lysimeter 1.2 m in diameter by 1.5 m deep. Jarvis et al. (1991) examined results for eight soil monoliths 0.3 m in diameter and 1 m high. Most of the laboratory studies of preferential flow were under steady-state conditions (Parker and Albrecht, 1987; Dyson and White, 1987; Ward et al., 1995; and Wallach and Steenhuis, 1998; Schwartz et al., 2000). Meyer-Windel et al. (1999) have shown that transient-state leaching results in more rapid leaching of solutes (in terms of outflow volume rather than time).

Transient-state soil column and lysimeter solute leaching data includes boundary conditions not found in the field, or not of critical importance for steady-state studies. Under transient rainfall conditions, intermittent evaporation can occur, which results in an upward gradient. In the field, water and solute can move back up into the study depth after leaching below the depth of interest. In contrast for columns and lysimeters, the water and solute that have exited the column cannot move back into the column. A steady-state downward flux would also negate the need for consideration of upward movement. Because of the boundary constraints, constant–flux or constant–potential lower boundary conditions are not suitable for transient-state column leaching data. The only preferential flow model we could find that included a lysimeter, free-drainage boundary condition was the MACRO model.

The objective of this study was to compare transient-state measured and predicted Br and tritium leaching from many undisturbed soil columns covering different soils, tillages, and rain intensities. The MACRO model was selected to predict solute leaching because it is a two-flow domain, transient-state solute transport model with a lysimeter, free-drainage boundary condition option.


   Macro Model
TOP
 ABSTRACT
 INTRODUCTION
 Macro Model
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The MACRO model has been described in detail elsewhere (Jarvis, 1991, 1994; Jarvis and Larsson, 2001 [The software version used was 4.1, cited is an updated version 4.3]). Assumptions in the soil hydrology component of the MACRO model are: (i) gravity drainage is assumed in the macropores, resulting in a linear relation between the pressure head (h) and the water content ({theta}); (ii) hydraulic conductivity (K) is a power-law function of saturation in macropores (kinematic wave approach (Germann, 1985); (iii) the macropore volume is influenced by shrinking and swelling; (iv) soil water retention in micropores is described by Brooks and Corey (1964); (v) hydraulic conductivity in the micropores is described by Mualem (1976).

Assumptions for soil water flow are: (i) water exchange between domains is approximately first order, neglecting gravity (Booltink et al., 1993) and assuming parallel fraction orientation (van Genuchten and Dalton, 1986); (ii) micropore flow is described by Richard's equation; (iii) gravity flow of water is assumed in macropores.

The upper boundary conditions are: (i) water flows into surface-open macropores if the rainfall rate is greater than the K at the boundary between the macropore and micropore domains (Kb), and {theta} > {theta}b (the water content at the boundary between the macropore and micropore domains); (ii) oversaturation in the surface-layer micropores is prevented; (iii) if the macropores at the surface become saturated, the excess water is shunted to runoff (i.e., no ponding is allowed); (iv) evaporation rate is the maximum of potential evaporation or the minimum rate water is supplied to the soil surface.

The lower boundary assumption used was a lysimeter with free drainage, which assumes: (i) if the bottom layer is saturated, water drains out of the lysimeter; (ii) if the bottom layer is unsaturated, no water movement occurs in either direction.

The solute transport assumptions are: (i) the convection-dispersion equation (CDE) is used with source and sink terms; (ii) the surface boundary conditions assumes instantaneous mixing of rain with water stored in a shallow depth (mixing layer); (iii) solute flows out of the bottom layer with dispersion set to zero; (iv) solute transfer between domains (Ue) includes diffusion and dispersion, still assumes parallel crack orientation, and accounts for incomplete contact between domains.

Some key macropore-region input parameters are Ks, saturated K; |hb|, the boundary tension between the macropore and micropore regions; Kb; {theta}sma, macropore volume fraction; n*, exponent of the K–{theta} relation for the macropore region; and d, half spacing between equivalent parallel fractures. Key micropore region parameters are {lambda}, micropore scale index; AE, anion exclusion; and Dv, dispersivity.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 Macro Model
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Column Leaching Study
We collected undisturbed columns of Clarion (fine-loamy, mixed superactive, mesic Typic Hapludolls), Storden (fine-loamy, mixed superactive, mesic Typic Eutrudepts), and Harps (fine-loamy, mixed superactive, mesic Typic Calciaquolls) soils. These are glacial till soils in the Des Moines lobe. For each soil there were eight columns each from no-till and chisel plots. The columns were 0.35 m long and 0.20 m in diameter, taken by driving thin-walled steel cylinders (0.4 m long) into the ground. To facilitate sampling, the cylinders had been coated with vegetable oil. If any compression occurred during sampling, the samples were discarded.

The treatments were three soils, two tillage management systems, and two water application rates (fast and slow for first application and vice-versa for the last application) for a total treatment number of twelve. There were four replicates for each of the treatments for a total of 48 columns.

We wetted up the 48 columns and allowed to drain by free drainage (positive pressure necessary for drainage). We dripped Br (0.42 g column-1) slowly on to the wetted columns, applied in 0.6 mm of water. We covered the columns for 1 d and then uncovered for 1 d. After equilibration with Br, we applied tritiated water (528 drops min-1 mL-1) using a dripper system made with syringe needles (Weed et al., 1998). Half of the columns received 30 mm of tritiated rain in 15 min. (rate of 120 mm h-1), and the other half received 30 mm of tritiated rain in 180 min (rate of 10 mm h-1). After the rain, we allowed the columns to drain for 7 d, uncovered. We applied three additional rainfalls as deionized water only (no tritium). The second and third rains were at the slower rate of 120 mm h-1 for all columns, each again followed by 7 d of drainage (uncovered). The fourth rain was at the rate of 120 mm h-1 for the columns originally receiving the slower rate, and at the rate of 10 mm h-1 for the columns originally receiving the faster rate. After drainage ceased for all columns after the fourth rain event, we ended the experiment. Leachate concentrations were determined colorimetrically for Br using a Lachet system (Lachet Instruments, Milwaukee, WI), and by liquid scintillation spectrometry for tritium. For each rainfall and drainage cycle, we measured the total amount of drainage and load of tracer lost as well as intermittent values. There was no fraction collector, so we manually collected outflow at approximately equal outflow volumes (equal to ~5% of applied volume).

Data Collection for MACRO Input Parameters
We had taken three extra columns for each soil and tillage treatment for determining pertinent soil properties. We determined K(h), using tension infiltrometers (Logsdon and Jaynes, 1993) at three heads (-30, -60, and -150 mm) and four depths (0, 0.05, 0.15, and 0.25 m) for each column. Afterward, we collected 74-mm diameter undisturbed soil cores at each depth. The length of the top sample was 38 mm, and the length of the bottom three samples was 76 mm (Fig. 1) . For each sample, we determined Ks by the falling head method (Klute and Dirksen, 1986), and bulk density after oven-drying. We collected disturbed soil at each depth for particle-size analysis by the hydrometer method (Gee and Bauder, 1986). Table 1 gives ranges for these measured soil properties.



View larger version (77K):
[in this window]
[in a new window]
  		Fig. 1. Diagram of column showing the horizontal measurement planes for hydraulic conductivity as a function of head, K(h). Each measurement plane was the top depth for an undisturbed sample taken for bulk density and Ks (saturated). The model was run with eight depths, two for each of the measured sections. Loose soil was taken for particle-size analysis.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Ranges of measured boundary hydraulic conductivities (Kb), mass fractions of sand and clay, and subcore measured saturated hydraulic conductivities (Ks) and bulk density the three extra columns, with Ks adjusted as used in the simulations.

 
Table 2 gives the range of input parameters used in this study. We directly or indirectly (Logsdon, 2002) determined the input parameters from the measured data describe above. In this study, we used 30 mm for |hb| since larger |hb| values sometimes resulted in a fitted n* > 6 (Logsdon, 2002), which is out of range for the MACRO program. The macropore {theta}(h) relationship was not always linear across a broader range of |h| (data not shown). For each soil–tillage–rainfall intensity combination, we determined the values for {lambda} and Dv from Br-leaching calibration (two to twelve trials each). We attempted to match the effluent data and the water balance. We used these final parameters for predicting tritium leaching.


View this table:
[in this window]
[in a new window]
  		Table 2. Ranges of model parameters ({theta}sma or macropore fraction, d or diffusion path length, AE or anion exclusion zone, {lambda} or microscopic pore factor, and n* or macropore exponent and Dv or dispersivity) used to predict bromide and tritium leaching.

 
We determined {theta}sma from the saturated K for the macropore domain (Ksma) from the empirical equation

[1]
assuming that {theta}sma was at least 0.01 m3 m-3. The MACRO model did not run well for smaller {theta}sma (for unknown reasons). Perhaps the slower simulation runs for small {theta}sma were because of greater instability associated with more predicted runoff. We calculated fracture length assuming parallel fractures and fracture widths based on mean |h|. For fracture width, we subdivided {theta}sma into two sections, 0- to 15-mm (fracture aperture between 1 and 4 mm) and 15- to 30-mm negative head (fracture aperture between 0.5 and 1 mm), calculating the mean fracture width for each negative head band from the capillary equation for fractures. Assuming all the fractures were parallel, we calculated d.

Anion exclusion is another important mechanism for soil with high charge (Bond et al., 1982; James and Rubin, 1986; Schoen et al., 1999). We utilized surface area measurements for similar soils and textures. We determined specific surface area by humidifying the soil (Newman, 1983; Laird, 1999; Quirk and Murray, 1999) over Mg(NO3)2 and correcting to a Ca-saturated bentonite standard (750 m2 g-1) for 68 similar soils with clay ranging from 14 to 52%. From this data we obtained a relationship between surface area (SA) and clay content for these superactive soils:

[2]

The AE as a function of SA varies with the amount of internal vs. external surface area and with {theta}. For these superactive clay fractions (high SA and charge), the internal SA is often a large fraction of total SA. Assuming two or three total water layers between internal surfaces (1 to 1.5 layers per surface), and four water layers per external surface (Prost et al., 1998) for Ca-saturated systems, the net result would be 1.5 to 2 water layers per surface. Therefore, conservatively, we calculated AE as

[3]
where AE is a fraction of total soil volume, SA is g m-2, Db is bulk density in Mg m-3, and 3 x 10-10 is water layer thickness in meters. Using data from the Soil Survey Center for similar soils, we obtained values for shrink-swell separately for each soil.

We took the value for tritium diffusion in free water from Bowman (1984). The value they gave for Br was for coelution with K. Scotter and Tillman (1991) have shown that high levels of Ca in the soil result in Br coeluting with Ca rather than K, and the diffusion value for CaBr2 was 35% less than for KBr. Two of the soils used have free CaCO3 within the surface horizons (Storden and Harps), and Clarion is dominated by Ca on the exchange sites even though no free carbonates occur in the surface horizons; therefore, we used the lower diffusion value for CaBr2 (1.21 nm s-1).

Running MACRO
We set up the model for eight depth increments within the column, the top two depth increments of 25 mm each, and the bottom six depth increments of 50 mm each; hence there were two depths for each of the measured parameters (Fig. 1). We determined initial water content at each depth iteratively to achieve the measured flow amounts. We set the bottom boundary condition as a lysimeter with free drainage. Although we examined effluent concentrations and loading (flux-averaged), the model does not convert to flux-averaged concentrations. This might not be as critical a point since much of the flow was occurring in the macropore domain. Temperature was set at 18°C.

We ran the model in the daily rain mode, so that the rain intensity could be set. To simulate the dripping of Br on to the column, we set the model for an irrigation 2 d before the first rain with high concentration of Br, to achieve the correct amount of Br to be added to the column. The irrigation was also included for the tritium runs, but with no solute present.

We used the lysimeter free-drainage boundary conditions because it was necessary to prevent prediction of upward water and solute movement because of evaporation. Unfortunately the lysimeter free-drainage boundary condition did not account for the positive pressure build-up (ponding) that was necessary for drainage to occur.

The simulated runoff component of the model cannot be turned off, but runoff did not occur in our measured column leaching study. Since the height of the cylinders was longer than the soil, water that did not enter the soil immediately, ponded until it infiltrated. The effect of ignoring ponded conditions in the simulations could have been significant. To prevent runoff in the simulations, we tried both increasing the rate of Ks up to the application rate, and slowing the application rate to the measured Ks values (some of which were less than the slowest rain intensity or slower than the corresponding Kb value). The compromise was to use a minimum Ks value that was twice (20 mm h-1) the slowest application rate, and to set the highest application rate to the minimum Ks value for that column. This reduced the difference between the two rain intensities, but did not eliminate the intensity effect.

We did not measure evaporation in the laboratory conditions but determined by water balance, and then input the mean values into the model (rather than calculated within the model). We assumed evaporation rate to be constant over the days that the columns were not covered (1 mm d-1), and zero on the covered day. Evaporation probably would not have been constant but would have been more than 1 mm d-1 initially after simulated rainfall, declining below 1 mm d-1 before the next rainfall; however, we only had average evaporation values.

We ran MACRO separately for each solute, rain intensity, and column of the measured parameters. Since there were three replicates for measurements of parameter inputs, two soils, two tillages, and two rain intensities, a complete set included 36 runs for each solute. Because we measured the model input parameters on different columns than the actual leaching studies, we compared means of actual (four replicates each) versus means of simulated (three replicates each) data. For each rain event, we determine the twelve measured mean minus predicted mean values. Significant differences between measured and predicted were indicated when the 95% confidence interval of the difference did not include zero. In addition, we displayed the mean predicted versus mean measured drainage, tritium load lost, and Br lost for all the rain events.

The measured data determined concentrations as a function of approximately equal effluent volumes, but the predicted data are output as cumulative leachate quantities (per surface area) as a function of selected equal time steps. To facilitate comparisons, the predicted runs were output at 1 h time steps. In addition, each run was also completed at 15 min time steps for 24 h after the first rain. For each treatment, the two predicted files with different time steps were merged. Excess data in the merged file was deleted to approximate outflow volumes of 2.5% of applied rainfall volume. This resulted in similar number of output steps for the predicted data (22 to 31 with mean of 27) as for the measured data (24 to 32 with a mean of 29). Then the predicted effluent data was converted to concentration (Br) or relative concentration (tritrium) at each step.

For each soil–tillage–intensity combination at selected volumes (interpolated at 5% of applied rainfall volume), we determined the deviation (v) of predicted solute concentration from the measured soil–tillage–intensity range. The reason we interpolated selected volumes was because the measured volumes were not exactly the same, either within a sample or across samples. For this comparison, the number of selected volumes was slightly less than half the actual number of volume intervals. If the predicted concentration fell within the measured range for the soil–tillage–intensity combination, v was zero. If the predicted concentration fell above the measured interval, v was the difference between the highest measured value and the predicted value. If the predicted concentration fell below the measured interval, v was the difference between the lowest measured value and the predicted value. For each soil–tillage combination (two rain intensities with three replicates each for six total per combination), the root mean squared error (RMSE) was determined from v at each selected time step as follows:

[4]
where n is sample number (six). This is a very generous RMSE since there was a range of values for each measured soil–tillage–intensity treatment at short volume leaching fractions.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 Macro Model
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Ranges of Data for Extra Columns
The Storden soil had higher maximum values for Ks, whereas Harps had lower maximum Ks values (Table 1). All soil–tillage treatment combinations had at least one column with high Kb values in the surface layer. The range of other input values (Table 2) are internally consistent. Because of the large Ksma for Storden soil, the {theta}sma values were also large, and the d values small. The higher clay content in Harps soils (Table 1) resulted in a larger calculated AE (Table 2). Although the Dv had only a small effect on leaching loss, larger values were often more consistent with the data because large Dv caused the predicted data to be smoother than did small Dv. Inherently, this was not appropriate, but perhaps the boundary conditions resulted in better fit for larger Dv. Experimentally, free drainage should not occur until positive head builds up at the base of a column, yet free drainage was predicted whenever the tension at the bottom boundary drops to zero (potential increases to zero). (See discussion by Schwartz et al., 1999.)

Overall, for these soils the expected range for n* was from 0 to 5.2; for {theta}sma was from 0.003 to 0.05; and for d was from 3 to 100 mm (Logsdon, 2002). In the current study, the measured range for Ks was from 20 (cutoff value to prevent runoff) to 475 mm h-1, for Kb in the surface 0 to 50 mm was from 1 to 16 mm h-1, and for Kb in the subsurface depths (50–350 mm) was from 0.5 to 7 mm h-1. The d calculated from (Ks - Kb) ranged from 8 to 60 (assuming {theta}sma was >=0.01 m3 m-3).

Comparison of Measured and Predicted Loads Lost
The model underpredicted drainage (Table 3) because of predicted runoff for the high rain intensities (Rains 1 and 4, half of samples). The predicted runoff occurred in spite of adjustment to Ks and intensities, setting both to 20 mm h-1 whenever measured Ks < 20 mm h-1. The reason for the significant differences between measured and predicted mean drainage was because of the uniformity of the data (Fig. 2) . The mean predicted and measured drainage values were actually quite close.


View this table:
[in this window]
[in a new window]
  		Table 3. For each rain event, comparison of the twelve measured means (four samples) and predicted means (three samples) for drainage, tritium leaching, and Br leaching.

 


View larger version (27K):
[in this window]
[in a new window]
  		Fig. 2. Predicted drainage vs. measured drainage for the twelve treatment combinations. Each predicted point was the mean of three replicates, and each measured point was the mean of four replicates.

 
The model overpredicted the tritium loads lost (Table 3) for the first two rain and drain cycles, and underpredicted for the last two rain and drain cycles. Overall at the end of the study, the measured and predicted tritium loads lost were not significantly different, demonstrating the importance of examining the intermediate times as well as the final. Cancellation of errors could imply the prediction was good, but in reality overall the prediction was not very good. For small mean measured loads, the model overpredicted tritium lost (Fig. 3) , but for large mean measured loads, the model underpredicted loads lost. The measured tritium loss spread among soils, tillage treatments, and rain intensities was greater than for mean measured drainage. Some of the treatment combinations were predicted quite well for all rains, i.e., Clarion conventional till, high intensity first rain (Fig. 3).



View larger version (30K):
[in this window]
[in a new window]
  		Fig. 3. Predicted tritium leached vs measured tritium leached for the twelve treatment combinations. Each predicted point was the mean of three replicates, and each measured point was the mean of four replicates.

 
The model underpredicted the Br loads lost for the first, second, and fourth rain and drain cycles. The lack of significant difference for Br load lost in the third rain was because of the extreme variability in the measured data. Overall, the model did not match measured Br loss, in spite of the fact, we attempted the sensitivity analysis (per sample) to match predicted with measured Br leaching. The model underpredicted Br leaching loss, especially for larger mean measured losses (Fig. 4) . Even so, a few treatment combinations were predicted fairly well for all rain events, i.e., Storden no-till, moderate intensity for the first rain.



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 4. Predicted Br leached vs. measured bromide leached for the twelve treatment combinations. Each predicted point was the mean of three replicates, and each measured point was the mean of four replicates.

 
The greater amounts of measured Br leaching than tritium leaching requires further discussion (Table 3). If all other factors were equal, the tritium applied in the rain water should leach more rapidly through the macropore system than the Br preapplied to the soil surface (Kluitenberg and Horton, 1990). Several factors that could increase Br leaching relative to tritium are AE, evaporation of tritium, slower matrix diffusion of Br, tritium retention, large mixing depths, and ponded conditions allowing better mixing of infiltrating water with Br on the soil. This simulation effort was not able to distinguish the importance of these various mechanisms, some of which were not included in the model. MACRO does not include a ponded surface boundary condition.

Anion exclusion results in the anion-solute front appearing ahead of the tritium solute front (Bond et al., 1982; James and Rubin, 1986; Schoen et al., 1999). As the infiltration rate increases, AE has less effect than preferential flow (Melamed et al., 1994). The MACRO model includes provisions for anion exclusion. Saxena et al. (1994) also observed less leaching for tritium than for 36Cl, which was attributed to tritium evaporation. The MACRO model includes tritium evaporation. We utilized the slower diffusion of CaBr2 in free water (Scotter and Tillman, 1991), loss of tritium by evaporation, and increasing the mixing depth, all of which increased Br leaching slightly. Others have suggested that tritium leaching is retarded compared with water, indicating possible sorption of tritium (Mansell et al., 1973; Wierenga et al., 1975; Van de Pol et al., 1977; Seyfried and Rao, 1987; Langner et al., 1999). The MACRO model does not include tritium retention.

Jabro et al. (1994) matched field Br leaching better with the SLIM model than with an early version of MACRO. Larsson and Jarvis (1999) observed good agreement for tile leaching of Br when the model was calibrated to the data. Even with the good agreement, the predicted Br leaching still lagged behind the measured Br leaching at longer times. In their case, the model overpredicted that water would bypass the solute, perhaps because evaporation moved solute to the edges of aggregates and peds, where the solutes were more likely to be picked up by the water in the macropore domain. This mechanism is discussed by Cote et al. (1999).

Breakthrough Curves
Ranges of measured Br breakthrough curves revealed much variability (Fig. 5) . Some of the Storden columns under low–intensity rainfall did not appear to have any preferential flow, but the rest of the columns did. Except in the Storden chisel-plow treatments, concentrations were high for the first, second, and early part of the third rain and drain event, then concentrations dropped. Similar variability was apparent in the measured tritium breakthrough curves (Fig. 6) , although the trends were somewhat different than for Br breakthrough curves. Preferential leaching of tritium was less evident than for Br, even though tritium was in the applied rain.



View larger version (55K):
[in this window]
[in a new window]
  		Fig. 5. Measured Br concentration as a function of outflow volume for each soil–tillage–intensity combination.

 


View larger version (51K):
[in this window]
[in a new window]
  		Fig. 6. Measured tritium reduced concentration as a function of outflow volume for each soil–tillage–intensity combination.

 
None of the predicted Br or tritium breakthrough curves perfectly followed the measured trends (Fig. 7 and 8) ; however, all captured some of the preferential flow for the first rain event, and sometimes for the second rain event. The intensity variation had little effect on predicted Br concentration except for the first pulse, in which the high intensity rain resulted in a higher concentration of Br. The predicted effluent concentrations increased slowly over time, but after the initial peaks, the concentrations generally remained lower than the measured concentrations until the third rain and drain event. The predicted patterns were quite similar to each other after the initial preferential flush (Fig. 7 and 8), even across the range of input values (Tables 1 and 2). The agreement between measured and predicted breakthrough patterns was not perfect, which resulted in some high RMSE values for the midrange of outflow volumes (Fig. 7 and 8), as much as four times the mean concentration for Br leaching, and as much as eight times for tritium leaching. The model was predicting bypass flow after the initial preferential flush, much more than actually occurred. Some of these RMSE values were quite high even though predicted values were compared with the range of measured values rather than the mean. The long-term fit was less well for tritium (Fig. 7) than for Br (Fig. 8) because a few parameters were calibrated for Br.



View larger version (34K):
[in this window]
[in a new window]
  		Fig. 7. Mean predicted Br concentration at specified volume outflows for the high and low first rain intensity treatments and RMSE for each soil–tillage combination. Outflow for each rain event was about 20%.

 


View larger version (33K):
[in this window]
[in a new window]
  		Fig. 8. Mean predicted Br concentration at specified volume outflows for the high and low first rain intensity treatments and RMSE for each soil–tillage combination. Outflow for each rain event was about 20%.

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 Macro Model
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Overall the model was able to indicate whether or not macropore flow and transport would occur, but intermediate-time micropore transport was underpredicted. Several factors might have contributed to our inability to perfectly match predicted and measured data include: (i) incorrect model equations (Beck, 1987; Beven, 1991); (ii) inaccurate input and output measurements (Beven, 1991); and (iii) indirect, back-calculated parameters may have conversion errors (Ferreira and Smith, 1988).

The inadequate match of predicted and measured concentration outflow data suggested that the mechanisms are not being captured completely or correctly. Several factors could contribute to these inadequacies, including inaccurate micropore equations (Milly, 1987), lack of ponding boundary condition, no provisions for tritium retention, need for more realistic geometries than parallel slab, nonuniformity of solute within the micropore domain, or hysteresis. Examining the mechanisms would be easier if the output could be given as a function of volume rather than as a function of time.

Measurement errors might also have contributed to poor agreement between measured and predicted effluent concentrations. Schwartz et al. (1999) discussed the laboratory-induced boundary conditions for unsaturated columns under steady-state outflow with vaccum drainage. They decided it would be helpful to consider the exit apparatus as a separate layer with mixing. The same could be true in our free drainage system with trapped air and intermittent outflow through tubing exit material, that depended on a positive head buildup at the base of the column before drainage can occur.

There is always uncertainty in measured input parameters (lack of steady-state, small subsample volume size). Sensitivity analyses (not shown) indicated that smaller values of {lambda} and a larger degree of shrink-swell tended to broaden the first peak somewhat (closer to the measured effluent concentration patterns), but still underestimated the long-term concentrations. No combinations of {lambda} and Dv produced effluent concentration patterns like the measured concentrations over the full range of the rain and drain cycles when combined with the other measured or selected parameters.

The vast and ever-changing model MACRO incorporates many components important for predicting solute leaching. Critical components such macropore flow, AE, various boundary conditions, plant growth, winter, etc. tremendously increase the flexibility of the model. This study failed to obtain perfect agreement between predicted and actual leaching from column data, indicating we still need greater understanding of the mechanisms of solute transport for these soils and boundary conditions.

Received for publication August 26, 2000.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Macro Model
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Vadose Zone JHome page
M. S. Akhtar, M. S. Akhtar, T. S. Steenhuis, B. K. Richards, and M. B. McBride
Chloride and Lithium Transport in Large Arrays of Undisturbed Silt Loam and Sandy Loam Soil Columns
Vadose Zone J., November 1, 2003; 2(4): 715 - 727.
[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 Google Scholar
Google Scholar
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Logsdon, S. D.
Right arrow Articles by Moorman, T. B.
Related Collections
Right arrow Vadose Zone Processes and Chemical Transport
Right arrow Soil Models
Right arrow Preferential Flow

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