Agriculture, Ecosystems and Environment, 1 1 ( 1 9 8 4 ) 1 0 5 - - 1 1 5 105 Elsevier Science Publ ishers B.V., A m s t e r d a m - - Pr in ted in The Ne the r l ands

SIMULATION OF NITRATE MOVEMENT IN UNDISTURBED SOIL COLUMNS

B.O. M O C H O G E '

Institute of Soil Science and Forest Nutrition, University of Goettingen (W. Germany)

(Accep t ed for pub l i ca t i on 2 J a n u a r y 1984)

ABSTRACT

Mochoge, B.O., 1984 . S imula t ion of n i t r a t e m o v e m e n t in u n d i s t u r b e d soil co lumns . Agric. Ecosystems Environ., 11: 105 - -115 .

In o rder to s imula te t he behav iour of n i t ra te m o v e m e n t in soils, two loessal soils were used wh ich d i f fered in the i r phys ica l and chemica l fer t i l i ty . U n d i s t u r b e d soil co lumns of 30 cm in l eng th and 15 cm d i ame te r were sampled . An u n s a t u r a t e d s teady-s ta te wate r f low in the co lumn s was es tab l i shed and m a i n t a i n e d by a Darc ian f low of 5 cm s o l u t i o n / day. The n i t r a t e fer t i l izer ( tagged) was appl ied in so lu t ion f o r m to t he t o p of co lumns and in one lot. The s t u d y t o o k place at 4 and 23°C in t he l abora to ry .

The resul t s show t h a t t he re is good ag reemen t b e t w e e n c o m p u t e d and measu red curves. (This was achieved by the curve-f i t t ing a p p r o a c h . ) This d e m o n s t r a t e s t h a t mea- sured curves can easily be descr ibed by s imple c o m p u t e r models . However , t o be able to merge the two curves, one has to t ake in to cons ide ra t i on the physical and chemica l charac te r i s t ics of t he soil. These fac tors have great in f luence o n t he t r a n s p o r t of n i t r a t e and o t h e r ions in soils.

I N T R O D U C T I O N

Intensive research during the past decade has resulted in a multi tude of models which simulate the behaviour of nitrogen in soil--water--plant sy~ tems. The diversity of these models results from incomplete understanding of the major transformation processes of nitrogen in soils and their inter- relationships. This makes it even more difficult to integrate the available information (Bartholomew and Clark, 1965) into a form that can be used to develop accurate relationships which are needed for simulation and pre- diction purposes.

Tanji and Gupta (1978), in their recent review of nitrogen simulation models, showed that these models range from totally empirical to those that are mechanistic in nature. The empirical models are designed on the basis of

' P r e s e n t address : D e p a r t m e n t of Soil Science, Univers i ty o f Nairobi , P.O. Box 30197 , Nairobi , Kenya.

0 1 6 7 - 8 8 0 9 / 8 4 / $ 0 3 . 0 0 © 1984 Elsevier Science Publ ishers B.V.

1 0 6

experience, observation and the use of regression equations which correlate input with output parameters, whereas mechanistic models are based on well-established physical, chemical and biological laws that describe in- dividual processes. Among many models developed to explain the tranw formation and transport of nitrogen are those of Beek and Frissel {1973), Mehran and Tanji (1974), Hagin et al. (1976), van Veen (1977), Tanji and Gupta (1978), van Veen et al. (1981) and Tillotson and Wagenet (1982).

The purpose of this paper is to present a simple simulation model which describes the behaviour of nitrate during its passage in two loess-derived soils. The model verifies its validity by comparing the computed with the experimental results. Hence, the model is intended to serve as a subsystem of N-models.

THEORETICAL APPROACH OF THE MODEL

A mechanistic model is presented to describe simultaneously the transport and transformation of nitrate during its movement in unsaturated steady- state of water flow through the soil columns. However, processes such as partial displacement of soil-water channelling (Quisenberry and Phillips, 1976), the presence of mobile and immobile water (Addiscott, 1977) and the incorporation of nitrate in the organic fraction are not included in the model.

The convection--dispersion equation normally used to describe the move- ment of non-reacting solute through the soil (eqn. 8-110, Kirkham and Powers, 1972) is also used here.

5c ~2c ~c - D ~ - V o - - ( 1 )

5t 5x 2 ~x

where c = concentration (mg cm -3) D = dispersion coefficient (cm:/day) Vo = pore-water velocity (Darcian flux/vol, water content (0)) x = distance (cm) t = time (days)

Where there is a source or a sink of the solute during its move- ment through the soil, eqn. (1) should be changed to eqn. (2).

5c 52c 5c --St = D 5 x 2 - V ° --Sx + Q ( 2 )

where Q = source or sink (rag cm -3 ) In the case of solutes which interact with the soil matrix, eqn. (2) should

be changed to eqn. (3).

pSs 5c 52c 5c + ~ = D - V o + Q

Of t 8 t ; x 2 -~x

107

where c = concentra t ion of chemical in soil solution (mg cm -3) 0 = volumetric water content (cm 3 cm -3) p = bulk density (g cm -3) s = sorbed concentra t ion (mg/g soil)

In this model, the adsorbed concentra t ion (s) and the actual concentra t ion (c) in the soil solution were assumed to be at equilibrium and linear (Lindstrom et al., 1967). A Freundlich equation was therefore used to de- scribe this state (eqn. 4).

s = Kc N (4)

where K and N are constants. Differentiating eqn. (4) with respect to time gives eqn. (5).

6s KNcN_I 6c - - = - - ( 5 ) 6t 6t

Substituting eqn. (5) in eqn. (3) allows the transport equation to be written in terms of one dependent variable (eqn. 6).

~c I182c ~c } - D - - - V o - - + Q ( 6 )

6t Rf dx 2 6x

with Rf = 1 + pKNcN-1/O (7)

where Rf = retardation factor (expression of the movement of the solute relative to the movement of water).

For linear adsorption, the retardation factor (eqn. 7) reduces to

Rf = 1 + pK/O (8)

Nitrogen transformations

The sink term Q in eqn. (6) is related to nitrogen transformation in soils. The microbiological nitrogen transformation considered in this model is denitrification of NO~ to gaseous form. In addition, ion-exchange NO~ is also considered. These processes are summarised below.

- - kD (NO3)e ~ (NO;)s

(N2 + N~O)g

(e = exchangeable; s = solution; g = gaseous; k D = Freundlich distribution coefficient; k, = first-order rate coefficient). The ion-exchange process is considered to be instantaneous, whereas the other process is of first-order kinetic type. The rate coefficient associated with this first.order reaction is k, for NO~ denitrification per day.

108

The kinetic rate coefficients for nitrogen transformations are frequently assumed to be constant (Mehran and Tanji, 1974; Misra et al., 1974), al- though their magnitude depends upon several soil environmental factors. McLaren (1971) suggested that these rate coefficients are dependent upon the size of the microbial population responsible for the transformation. The population and activity of any group of microbes is determined, in part, by the energy source available at any given depth in the soil profile. Based upon this suggestion, Rao et al. (1976b) assumed that the magnitude of k~ de- creased exponentially with depth, in a similar fashion to the organic matter content distribution in the soil profile.

In the present model, it was assumed that k l was a function of depth. A first-order rate process was therefore used (eqn. 9).

5c - ~ = k l c (9)

8t

First, order rate processes have been used by Mehran and Tanji (1974) and Hagin and Amberger (1974) to describe denitrification in soils. Hagin and Amberger (1974) included the effect of pH, temperature, oxygen and organ- ic carbon content in their models.

A continuous system modelling program (CSMP) was developed to solve eqns. (6), (8) and (9) numerically for the following conditions:

t = 0 , ( 1 0 )

t/> O, (11)

0 = const., c = 0, s = 0

O = const., qo = R

D 6 c x = 0, 0 < t < tl, c - Co (12)

V o 6 x

X ~ O ~ ,

where

D6c t > t l , c V o S x - 0 (13)

6c = O , q x = R (14)

~t

t = time (days) tl = time during which the solute is applied with the water (days) Co = concentration of irrigation water with NO3 (mg cm -3 ) qo = flux of water entering the soil surface (cm/day) q = flux of water in the soil (cm/day) R = acons tan t

109

EXPERIMENTAL

Soils

Two soils, one from arable land and the other from forest management but both originating from loess, were used for this study. The soil from arable land (Orthic Luvisols) was sampled from the research farm of the University of Goettingen. This soil is characterised by its high silt con- tent {83%), low organic matter (1.1%) and by total absence of CaCO3. The soil is neutral (pH 7) and has a high calcium content, which occupies almost 87% of the total C.E.C.

The soil from forest management (Dystric Cambisols), is slightly podzol- ized and is occupied by beech trees. The area lies on the rolling mountains of W. Germany and has been under International Biological Programme (IBP) studies. The soil is characterised by its low pH (3) and low bulk density (1.03 g cm -3 ), which is about 30% less than the bulk density from arable soil. Chemically, the soil has a high aluminium content, about 76% of the total C.E.C., and very low calcium. The soil is, however, well aggregated and has higher physical stability than the arable soil.

Methods

For this study, undisturbed soil columns 30 cm long and 15 cm wide were used. (For experimental set, up see Mochoge, 1981.) The columns were treated with about 130.0 mg N (corresponding to 80 kg N ha -1) of lSN tagged Ca(NO3)~. This amount was applied to the surface of each column in solution form and in one lot. The application followed after establishment of an unsaturated steady-state water flow condition in the columns. Thereafter, a daily application of 5 mm (about 81.5 m l ) o f a prepared solution (Mochoge and Beese, 1983) to each column was maintained and lasted 60 days. An effluent from the bot tom of each column was collected after every two days and analysed for NO3. The mean concentration values of NO~ analysed from 5 replicates for the 60 days plotted against the cor- responding volumes of effluent collected led to the construction of the break-through curves (BTC's). The determination of lSN was done by an atomic emission spectrometer (Straton NOI-5, 5N analyser).

Procedure

The parameters used in the simulation were those obtained from the experiment with respect to the two soils. These are shown in Table I.

Parameters such as bulk densities (p), volumetric water content (0) and water pore velocities (V o) were assumed to be uniform in the whole soil column, so average values were used for computation.

Other parameters not measured quantitatively, such as dispersion co-

110

T A B L E I

Soil parameters used for c o m p u t a t i o n

Orthic luvisols Dystr ic cambiso ls

4°C 23°C 4°C 23°C

p ( g c m - ~ ) 1.37 1.39 0 (cm ~ cm - a ) 0.298 0.299 qo (cm/day) 0.5 0.5 Vo (cm/day) 1.678 1.672 Co ( M g N c m - 3 ) 0.8 0.8

1.01 1.03 0.435 0.435 0.5 0.5 1.150 1.149 0.8 0.8

efficients (D), adsorption coefficient (KADS) in case of acid soil, and kinet- ic rates (k ~ ) were determined by a curve-fitting method. In order to obtain D values, the model curves were fitted to the measured ones at 4°C treatments. This approach has been used previously and found to be adequate (Beese and van der Ploeg, 1979). Otherwise, no satisfactory method of calculating D quantitatively from soil physical properties, especially of field soils, has yet been proposed (Tinker, 1980). This is because dispersion is physically com- plex (Rao et al., 1976a).

R E S U L T S A N D D I S C U S S I O N

Because of the fact that the dispersion coefficient (D) of eqn. (1) was not determined independently, the BTC's at 4°C treatments were used to find

0.24

0.20

o (J ,~ 0.16

0.1;

8 O.OE

~: 00~

p i. AGRICULTURAL SOIL (4'C)

~t. 0 : 0 2 9 8 i ! I Vo= 1.678 c m / d a y

i I . . . . . . . . . . . d . . . . . c a l c u l a t e d c u r v e s :

! ~ - - - D = 2.6 cm2/day ! , ! j . . . . D : 0.8 cm2/day ! I I !

~,o i

I, o°l ~i Jio toi !1o i ~ I ; ° t

| p •

// ,. £ J ' . . 5 o ~ _ - % ' Lo- ' 6o DAYS

1 2 3 P V .

C, .08

o =

06

8

0;

l i o: I i o"

2'0

Fig. 1. D e t e r m i n a t i o n o f D-values by the f i t t ing me th o d .

.: 0 187 !

: !

!: '~ i !~ t o -

,7, ;:ti i i : o t : o I , i o ~ i:.%

• ::oO \ ' ~ ! % I j . .:{,

I I :o l i °

~ o

FOREST SOIL ( 4 " C )

S = 0./*35 Vo= 1 150 c m / d a y aDo m e a s u r e d c u r v e

c a l c u l a t e d c u r v e s :

- - - D = 3 6 0 cruZ/day . . . . D = 2.0 cruZ/day ...... O : 0.45 cmZlday

~Oo %~..o o

I RV6' 0 D A Y S 40 2

111

the D-values of these soils. The nitrogen losses at 4°C were low, so this temperature was used to fit the calculated BTC's to the measured ones. To do this, D in eqn. (1) was changed in a trial and error procedure until the deviations of the curves were small. Some results of the procedure are shown in Fig. 1. The curves show that using low D-values, it is possible to find BTC's which fit the very first part of the measured BTC's, but have very high relative concentrations before they break, and using high D-values, the cal- culated BTC's appear much earlier than the measured ones and exhibit pronounced tails. These values do not agree with the experimental results. Therefore, the best fit was obtained by using D-values of 1.1 and 1.7 cm3/ day for the agricultural and forest soils, respectively. These D-values were used in all calculations reported here.

However, the use of D-values alone was not enough to allow the calculated BTC's to be fit ted to the measured BTC's. In the case of agricultural soil, a sink-term had to be introduced to account for denitrification losses, as shown in eqn. (2). All further calculations for the agricultural soil are there- fore based on eqn. (2). In the case of forest soil, the introduction of the sink term did not give the required results until a retardation factor was included in the model. NO~ ions in acid soils are adsorbed on positive charges formed by A1 and Fe hydroxyl species which retard its movement (Rao, 1974; Balasubramanian, 1974). Equation (3) was therefore used for all calculations for the forest soil.

Using the D-values mentioned above in a trial an error procedure, kl was changed until the calculated curves nearly fitted, or fell within, the 95% confidence limits of the measured BTC's. Denitrification losses in agri- cultural soil were approximately the same in both temperature treatments. The same pattern of losses in the columns was therefore used. The losses here were assumed to increase down the profile. This is due to the decrease of 02 down the soil profile (van Veen, 1977). 02 deficit gives way to facul- tative microorganisms which use NO~ as a hydrogen acceptor. In the agri- cultural soil, the sink term was found to rate from 0.0001 at the top surface to 0.01 at the bot tom of the 30-cm soil column/day. Similar rates have also been reported by Davidson et al. (1978). The calculated curves (Fig. 2) show a close fit to measured ones and lie within, or close to, 95% confidence limits. It therefore seems that when a sink term is used, a simple convection--dispersion equation, as shown in eqn. (2), can be used for com- putat ion with neutral soils.

The sink-rate constants with the forest soil decreased from top to bot tom of the soil columns. That is from 0.0001 to almost 0.0 at 4°C, and from 0.01 to 0.0055/day at 23°C. This behaviour is contrary to agricultural soil, where rate constants increase with depth. The reason for this difference is the magnitude and distribution of energy sources to microbes. While in the agricultural soil the organic matter was more or less uniformly distributed over the whole plough layer, in the forest soil, the organic matter con- centration was higher in the upper parts and decreased steeply with depth.

112

This therefore had a great influence on denitrification in the upper parts of the soil columns. Consequently, this uneven distribution of energy source might be superimposed on the effect of the 02 supply. Nevertheless, since the upper part of this soil is well aggregated, microsites are likely to exist

AGRICULTURAL SOIL

.20 .20

i

0 (.i

.16

z 0

E .12 z .I rJ z 0 /J

<

n...0.'-

i 0 120

(23"C)

(4"C) e • 0.298

%= 1.678 cm/day oooomeasured curve - - - c a l c u l a t e d curve

Confidence l imit ~/~//~ 95 %

D = 1.1 cm2/doy

2 40 I~V. 3

~ . t E

.12 .=,

~ .08

~ .04

60 DAYS '0

e = 0.299 / /

cm, ,

" E

g

20 " ~ 40 60 DAYS 1 2 RV.3

Fig. 2. Measured and calculated NO 3 BTC's for agricultural soils.

FOREST SOIL

( 4 " C ) e • 0 .~s .1 v o f 1.150 cm/day

oeo measured curve - - c a l c u l a t e d curve

M ~ Confidence [imit ~O° ~ ~ g5 % .08

O z

. . . , . o ,0 KADS - 0.09 z .0E

z

z w ' ~ ~ .04

z

,,-I, "' e,- .02 ~ .02

0 S~ • 0 I 4'0 i p VS0 DAYS 0

2() 1 2 . . 1

Fig. 3. Measured and calculated NO 3 BTC's for forest soil.

(23"C)

e ffi 0.435

lay

2U ~U DAYS 2 RV.

113

which lack oxygen and therefore give way to denitrification. In the case of the retardation factor, an adsorption coefficient was deter-

mined by the fitting method. It was found to be 0.09 for both temperatures and was assumed to be linear and to obey the Freundlich equation. Further, this coefficient was assumed to be constant over the whole soil column (30 cm). The slight shift of the curve and the peak to the right therefore reflects the low adsorption of NO~ ions in this soil (Fig. 3).

However, at 23°C (Fig. 3), the calculated curve does not fit well to the measured one, nor does it lie within the confidence limits as intended. This discrepancy is attributed to NO~ incorporated into organic matter. This process was not included in the model. While in the other treatments, the NO~ incorporated was low (4.6%), in this t reatment (Fig. 3, 23°C), in- corporation was high (15%) and therefore had an impact on the fitting of the curves (Mochoge, 1981).

CONCLUSION

These results clearly show that there is a fair agreement between the measured and calculated curves. The study therefore demonstrates that the measured curves can easily be described by simple computer models, using the CSMP III language. Further, the study shows that the soil structure has great influence on the movement of NO3 in soils. This can be seen from the dispersion coefficients of the two soils. The high D-values in forest soils is a reflection of good aggregation of the soil, while tillage has destroyed such aggregation in agricultural soft, resulting in low D-values.

The widening and the tailing of BTC's with the forest soils is due to high D-values and anion adsorption.

The way of determining the rate constants to describe the losses of ni- trogen through denitrification has to be regarded as a first approach. Our ability to describe the process of denitrification quantitatively as a function of soil and environmental conditions is still very poor. The parameter fit, as it has been done here, can help to explain more about the factors influencing the process and also help to locate the zones where the process takes place in the soil.

ACKNOWLEDGEMENT

I am grateful to Dr. F. Beese, Senior Research Fellow, Institute of Soil Science and Forest Nutrition, University of Goettingen, for introducing me to simulation models and for his assistance in the present work.

REFERENCES

Addiscott, T.M., 1977. A simple computer model for leaching in structured soils. J. Soil Sci., 28: 554--563.

1 1 4

Balasubramanian, V.T., 1974. Adsorption, denitrification and movement of applied ammonium and nitrate in Hawaiian soils. Ph.D. Thesis, University of Hawaii, 73-72, 675 University Microfilms International, Ann Arbor, MI, London.

Bartholomew, V.W. and Clark, F.E. (Editors), 1965. Soil Nitrogen Agronomy Monograph No. 10. Am. Soc. Agron., Madison, WI, 615 pp.

Beek, J. and Frissel, M.J., 1973. Simulation of nitrogen behaviour in soils, Pudoc, Wageningen.

Beese, F. and van der Ploeg, R.R., 1979. Simulation des Anionen -- Transports in unge- stSrten Bodens~iulen Unter Stationiiren Fliessbedingungen. Z. Pflanzenernaehr. Bodenkd., 142: 69--85.

Davidson, J.M., Graetz, D.A., Rao, P.S.C. and Selin, H.M., 1978. Simulation of nitrogen movement, transformation and uptake in plant root zone. Rep. U.S. Environmental Protection Agency.

Hagin, J. and Amberger, A., 1974. Contribution of fertilizers and manures to the N- and P-load of waters. A computer simulation. Final Report to the DFG (W. Germany) from Technion Israel.

Hagin, J., Amberger, A., Kruh, G. and Segall, E., 1976. Comparison of nitrate leaching results in a lysimeter experiment to those predicted by a simulation model and esti- mates of influences of varying parameters on simulation results. Z. Pflanzenernaehr. Bodenkd., 4: 457--464.

Kirkham, D. and Powers, W.L., 1972. Advanced soil physics. Wiley-Interscience, New York, pp. 413--415.

Lindstrom, F.T., Haque, R., Freed, V.H. and Boersman, L., 1967. Theory on the move- ment of some herbicides in soils. Environ. Sci. Technoh, 1: 561--565.

McLaren, A.D., 1971. Kinetics of nitrification in soils: Growth of nitrifiers. Soil Sci. Soc. Am. Proc., 35: 91--95.

Mehran, J. and Tanji, K.K., 1974. Computer modeling of nitrogen transformations in soils. J. Environ. Quali., 3: 391--395.

Misra, C., Nielsen, D. and Biggar, J.W., 1974. Nitrogen transformation in soil during leaching: I, II, III. Soil Sci. Soc. Am. Proc., 38: 289--304.

Mochoge, B.O., 1981. The behaviour of nitrogen fertilizers in neutral and acid loess soils. Ph.D. Thesis, University of Goettingen, Goettinger Bodenkd. Berichte, p. 69.

Mochoge, B. and Beese, F., 1983. The behaviour of nitrogen fertilizers in neutral and acid loess soils. I. Transport and transformations of nitrogen. Z. Pflanzenernaehr. Bodenkd., 146: 89--100.

Quisenberry, V.L. and Phillips, R.E., 1976. Percolation of surface-applied water in the field. Soil Sci. Soc. Am. J., 40: 484--489.

Rao, P., 1974. Pore geometry effects on solute dispersion in aggregated soils and evalua- tion of a predictive model. Ph.D. Thesis, University of Hawaii, 75-17, 123 Xerox University Microfilms, Ann Arbor, MI.

Rao, P.S.C., Green, R.E., Ahuja, L.R. and Davidson, J.M., 1976a. Evaluation of a capil- lary bundle model for describing solute dispersion in aggregated soils. Soil Sci. Soc. Am. J., 40: 815--820.

Rao, P.S.C., Sclim, H.M., Davidson, J.M. and Graetz, D.A., 1976b. Simulation of trans- formations, ion-exchange and transport of selected nitrogen species in soils. Soil Crop Sci. Soc. Fla. Proc., 35: 161--164.

Tanji, K.K. and Gupta, S.K., 1978. Computer simulation modeling for nitrogen in ir- rigated crop lands. In: D.R. Nielsen and J. McDonald (Editors), Nitrogen and the Environment. Academic Press, New York, pp. 79--131.

Tillotson, W.R. and Wagenet, R.J., 1982. Simulation of fertilizer nitrogen under cropped situations. Soil Sci., 133: 133--143.

Tinker, P.B. (Editor), 1980. Soils and Agriculture. Critical Reports on Applied Chemistry, Vol. 2. Blackwell Scientific, Oxford, London, pp. 35--70.

115

Van Veen, J.A., 1977. The behaviour of nitrogen in soil. A computer simulation model. Doctoral Dissertation, The Free University of Amsterdam.

Van Veen, J.A., McGill, W.B., Hunt, H.W., Frissel, M.J. and Cole, C.V., 1981. Simulation models of the terrestrial nitrogen cycle. In: F.E. Clark and T. Rosswall (Editors), Terrestrial Nitrogen Cycles. Ecology Bulletin 33.