Chloride transport in a recently reclaimed Dutch polder

Journal of Hydrology 257 (2002) 59±77

www.elsevier.com/locate/jhydrol

Chloride transport in a recently reclaimed Dutch polder J.A. de Vos a,*, P.A.C. Raats b, R.A. Feddes c a

Alterra, Green World Research, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands b Wageningen University and Research Centre, Wageningen, The Netherlands c Department of Environmental Sciences, Wageningen University and Research Centre, Wageningen, The Netherlands Received 26 October 2000; revised 21 September 2001; accepted 5 October 2001

Abstract Knowledge about the origin of solutes is required to de®ne reasonable limits for surface water quality. The origin of the high chloride (Cl) concentrations in the drainage water of a recently reclaimed Dutch polder is analysed. Historical data on soil formation are used to understand the layering of the soil pro®le and its hydraulic properties. The salinisation process of the lake bottom in the period 1600±1932 explains the Cl pro®les in the soil. After reclamation in 1942, the average application rate of Cl from fertiliser was 118 kg ha 21 yr 21. Natural Cl deposition was 37 kg ha 21 yr 21 and average Cl crop uptake was 10 kg ha 21 yr 21. Upward diffusion from the subsoil (.120 cm depth) of about 220 kg ha 21 yr 21 is the major term in the Cl mass balance. Annual Cl drain discharge is estimated at 365 kg ha 21 yr 21. A ®eld monitoring study in the winter leaching period 1991±1992 showed that the Cl drain discharge was only 180 kg ha 21 yr 21, probably, due to the relatively small precipitation excess in that period. The dynamics of water ¯ow and solute transport in a two-dimensional ¯ow domain are well described by the HYDRUS-2D model based on the Richards and Laplace equations for movement of water and the convection±dispersion equation for solute transport. The large lateral ¯ow component towards the drain in the topsoil (0±120 cm depth) in the zone just above and below the phreatic surface explains the peaks in Cl and nitrate concentrations in the drainage water. The combination of historical data, ®eld experiments and a simulation model was fruitful in explaining the origin and dynamics of Cl concentrations in the drainage water. The high Cl concentrations in the drainage and surface water are natural background concentrations originating mainly from the subsoil of the reclaimed polder. q 2002 Elsevier Science B.V. All rights reserved. Keywords: Solute transport; Water ¯ow; Diffusion; Modelling; Field-scale

1. Introduction Leaching of solutes from agricultural soils is a serious environmental problem (Strebel et al., 1989). Nitrate (NO3), phosphate (PO4) and other solutes may leach and pollute both groundwater and surface water (EU, 1991; Fletcher, 1991). Knowledge about the

* Corresponding author. E-mail address: [email protected] (J.A. de Vos).

origin of solutes is required to de®ne reasonable limits for groundwater and surface water quality. In reclaimed polders salt concentrations in the surface water can originate from different sources. This paper analyses the origins of the high chloride (Cl) concentrations in the drainage water of a typical Dutch polder (Noordoostpolder), 50 years after reclamation of the polder. Fertiliser applications and upward transport from the subsoil are the likely sources of Cl. By combining historical and recent ®eld data and a two-dimensional simulation model,

0022-1694/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved. PII: S 0022-169 4(01)00552-2

60

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Fig. 1. The experimental farm Dr H.J. Lovinkhoeve is located at Marknesse (latitude 5853 0 E, longitude 52845 0 N) in the Noordoostpolder, The Netherlands.

water ¯ow and solute transport are described and quanti®ed. The Noordoostpolder (Fig. 1) was reclaimed in 1942 and is a typical Dutch polder with respect to soil formation, hydrological situation and agricultural use (Wiggers, 1955). The soil developed in sediments, which were deposited under water. In the period before 1300 ad, the sea level and groundwater level in the region rose and marshes with peat deposition developed. Due to the changing hydrological conditions, part of this peat was abraded and redeposited. Finally, a lake was formed, the Flevomeer. Until about 1600 ad, the in¯uence of the sea was not very strong and the water in the lake remained fresh to slightly brackish due to the in¯uence of fresh water from the river IJssel, which had considerable discharge in that period. Around 1600 ad, the in¯uence of the sea increased. The Zuyderzee, a lagoon, was formed and marine sedimentation took place. In 1932 a dam, the Afsluitdijk, closed off the Zuyderzee from the sea and the lake rapidly became fresh again. This lake is called the IJsselmeer. The Noordoostpolder was reclaimed from the IJsselmeer in 1942. The salinisation process of the lake bottom of the Noordoostpolder has been described by Volker and Van der Molen (1991), based on the observations during the period 1936±1938, as well as a more recent theoretical formulation of solute transport in soils and aquifers. Groundwater currents and diffusion were the

main processes determining the Cl pro®les in the lake bottom. The high Cl concentrations in the soil pro®le after reclamation were a result of the salinisation period from about 1600 to 1932. During this period, salt in the seawater of the Zuyderzee moved downwards into the soil to a depth of about 10 m. In the period 1932±1941, the salt moved upwards to the relatively fresh water of the IJsselmeer. After the digging of trenches and installation of tile drains, desalinisation of the 0±100 cm depth soil layer in the Noordoostpolder proceeded rapidly. Measurements made at the Lovinkhoeve experimental farm near Marknesse (Noordoostpolder) showed that the salt content in the 0±80 cm soil layer dropped rapidly (Van der Molen, 1958). This study is an extension of the work of Volker and Van der Molen (1991), using the latest experimental methods and simulation tools to describe the dynamics of the Cl transport processes in typical polder soils. The two possible sources of Cl in the drainage water are upward transport of Cl from the subsoil and input of Cl at the soil surface. Our objectives are to identify and quantify the Cl sources, to describe the solute transport processes, and to further understand the dynamics of Cl concentrations in the drainage water. Historical data of the salt balance of the Noordoostpolder (Van der Molen, 1958) and the archives of the experimental farm Lovinkhoeve are used to make a long-term Cl mass balance for the period 1942±1991. The Cl and NO3 contents in the soil pro®le along with Cl and NO3 concentrations in the drainage water and the description of the ®eld transport processes are evaluated for the leaching period 1991±1992. We apply the water ¯ow and solute transport model HYDRUS-2D (Simunek et al., 1996) for the combined saturated±unsaturated zones in a two-dimensional ¯ow domain to explain the dynamics of solute transport in this tile-drained silt loam soil. Hydrological ®eld measurements are used to evaluate the simulations. 2. Material and methods The soil pro®le of the Lovinkhoeve experimental farm is described in relation to the soil formation process. Measured soil hydraulic characteristics for the different soil layers are used in the HYDRUS-2D

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

61

1962), accelerated by soil tillage and crop growth the soil pro®le shown in Fig. 2 developed (Zuur, 1951). The soil has an AC-pro®le and is classi®ed as a Typic Fluvaquent (Soil Survey Staff, 1975; Kooistra et al., 1989) or as a Calcaric Fluvisol (FAO-UNESCO, 1974). De Vos et al. (2000) presented the experimental data on water retention and hydraulic conductivity characteristics, and the corresponding parameters of the analytical functions of Van Genuchten (1980) and Mualem (1976) for the different soil layers.

Fig. 2. Soil pro®le of the Lovinkhoeve in 1991. For each layer, the particle size distribution (clay, silt and sand content as volume percentage), the organic matter content as weight percentage, and the cation exchange capacity are indicated.

model. To determine whether upward or downward seepage occurred, hydraulic heads were measured in the ®eld. This provides the bottom boundary condition for water ¯ow in HYDRUS-2D. Meteorological measurements are used to de®ne the boundary conditions at the soil surface. Measured soil Cl and NO3 concentrations are the initial conditions for the simulation of solute transport. The general aspects of the HYDRUS-2D model are described brie¯y. Special aspects of using HYDRUS-2D for the drainage situation studied have been given by De Vos et al. (1999, 2000). Field measurements of drain discharge rates, Cl and NO3 concentrations in the drainage water, and hydrological conditions in the soil pro®le are used to estimate the Cl mass balance and to evaluate the modelling concept. 2.1. Lovinkhoeve experimental ®eld 2.1.1. Soil pro®le and soil hydraulic characteristics The soil pro®le of the Lovinkhoeve after reclamation in 1942 consisted of a loamy topsoil (0±30 cm), an intermediate clay layer (30±39 cm), a very thin ®ne sand layer (39±42 cm), a subsoil consisting of a heavy loam layer (42±46 cm), and a light loam layer (46±100 cm). Due to ripening processes (Smits et al.,

2.1.2. Drainage system and hydrological situation For the period 1943±1996, the average annual precipitation at the Lovinkhoeve was 745 mm. The annual actual evapotranspiration for arable land is about 445 mm. So, an annual precipitation excess of about 300 mm needs to be drained. Maps of the occurrence of upward and downward seepage in the Noordoostpolder were made by analysing groundwater levels, salt concentrations and the quality of ice in the ditches in winter (Van der Molen and Sieben, 1955). Based on these maps, little (,0.2 mm d 21) or no upward seepage is to be expected at the Lovinkhoeve. The design of the drainage system in the Noordoostpolder was based on the clay content in the 0± 25 cm and the 25±100 cm soil layers (Sieben, 1951; Zuur, 1952; Fokkens, 1966). A phreatic surface below 40 cm depth at a steady precipitation rate of 10 mm d 21 was the steady-state design criterion for the drainage system. The heavy loam soil overlying a light loam soil called for a drain spacing of 12 m (Sieben, 1951). The subsurface tile drains (7.5 cm outer diameter) were laid by hand in the drain trench (width ˆ 20 cm). 2.1.3. Hydrological and meteorological measurements in 1991±1992 To monitor groundwater levels, observation wells were installed in December 1991 to 200 cm depth. To monitor hydraulic heads, piezometers were installed at 125, 150, 175 and 200 cm depths. Measurements were conducted once in a week in the 1991±1992 period. At one location, the groundwater level was also measured automatically every 15 min. All instruments were installed midway between two adjacent drains.

62

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Fig. 3. Finite element grid in the ¯ow domain representing half the drain spacing. A detail is given of the grid close to the drain. The entire grid consists of 2563 triangles and 1352 nodes.

A special drainage water sampling device was developed to collect ¯ow-proportional samples of the drainage water to monitor solute concentrations (De Vos, 2001). Each sample represents 0.5 mm cumulative discharge from the experimental ®eld. An automated meteorological station was located at 75 m from the experimental plot. Global and net radiations, air temperature, relative humidity, precipitation, wind speed, wind direction, soil temperatures, and groundwater level were measured every 5 s, then averaged and stored as one datum for each 10 min interval.

2.1.4. NO3 and Cl concentrations in soil in 1991 At the beginning of the experiment on 17th December 1991, soil Cl and NO3 contents, and volumetric soil water content (100 cm 3 samples) were measured between 0 and 150 cm depth at 10 cm intervals. Twenty soil samples from each depth interval were taken at similar positions on both sides of the drain. 2.2. HYDRUS-2D model for water ¯ow and solute transport The HYDRUS-2D model simulates Darcian water ¯ow in a two-dimensional ¯ow domain in the

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

63

Fig. 4. Distribution of the different soil layers in the ¯ow domain, including a drain trench in the zone above and below the drain. The numbers of the soil layers at the right-hand side and of the drain trench (9) at the left correspond to the hydraulic characteristics mentioned in the text and in Table 1.

unsaturated±saturated ¯ow system. In addition, HYDRUS-2D simulates solute transport using the convection±dispersion approach. The ®nite-element method is used to solve the partial differential equations numerically. In the present study, HYDRUS-2D is applied for an autumn±winter±spring leaching period only, when there is no water and solute uptake by plant roots. Diffusion, dispersion and deposition are accounted for in the simulations of Cl transport. The diffusion coef®cient D ˆ 0:4 cm2 d21 is used for all soil layers and the in®ltrating rainwater has a Cl concentration of 5 mg l 21 (Ridder, 1978). This diffu-

sion coef®cient is similar to the one used in the subsoil. In the topsoil, this coef®cient is not very important, since the spreading of solutes is dominated by hydrodynamic dispersion. In the subsoil, the diffusion coef®cient is much more important because there is nearly no convection and solute transport is dominated by diffusion. This justi®es the use of only one diffusion coef®cient. The net effect of N transformations is simulated by a constant NO3 production term of 0.34 kg N ha 21 d 21, homogeneously distributed through the 0±25 cm topsoil. This production term re¯ects the net effect of mineralisation, immobilisation

Table 1 Hydraulic characteristics of the different soil layers (De Vos et al., 1999). The subsoil (layer 8, 120±200 cm) has a relatively low Ks. Parameters in the analytical expressions of Van Genuchten (1980) and Mualem (1976) for describing the hydraulic characteristics of the different layers of the Lovinkhoeve soil: u r is the residual volumetric water content, u s is the volumetric water content at saturation, Ks is the hydraulic conductivity at saturation, a , n and L are empirical parameters, u k is the volumetric water content at transition point in hydraulic conductivity characteristic close to saturation, and Kk is the hydraulic conductivity at transition point Zone

Layer

Depth (cm)

u r (±)

u s (±)

Ks (cm d 21)

a (cm 21)

n

L

u k (±)

Kk (cm d 21)

1

1

0±25

0.04040

0.43709

200

0.02190

1.30740

22.02754

0.41072

0.206

2

2 3

25±35 35±40

0.08165 0.08165

0.47303 0.47303

200 300

0.00525 0.00525

1.62422 1.62422

21.84898 21.84898

0.46923 0.46923

0.949 0.949

3

4 5 6 7 8

40±50 50±70 70±95 95±120 120±200

0.10060 0.10060 0.10060 0.10060 0.10060

0.51506 0.51506 0.51506 0.51506 0.51506

120 40 20 10 2

0.00207 0.00207 0.00207 0.00207 0.00207

1.91432 1.91432 1.91432 1.91432 1.91432

21.46461 21.46461 21.46461 21.46461 21.46461

0.51461 0.51461 0.51461 0.51461 0.51461

0.334 0.334 0.334 0.334 0.334

4

9

Trench

0.08165

0.47303

200

0.00525

1.62422

21.84898

0.46923

0.949

64

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

and denitri®cation. During the growing season, the mineralisation rate is about 1 kg N ha 21 d 21 in the 0±30 cm topsoil. However, during the winter leaching period, temperatures are lower end there is less readily decomposable organic material left. So, the mineralisation rate is lower end due to immobilisation and denitri®cation, the production amounts to 39 kg N ha 21 throughout the winter leaching period, which corresponds to the estimate based on the ®eld N balance (De Vos et al., 2000).

Fig. 5. Hydraulic conductivity characteristics for three soil layers: topsoil (1): 0±25 cm; intermediate layer (3): 35±40 cm; and the subsoil (7): 95±120 cm depth. The lines show the description of the hydraulic conductivity characteristics in the HYDRUS-2D model, the symbols indicate laboratory measurements at the different depths (see Table 1 and Fig. 4).

Fig. 6. Initial Cl and NO3 concentrations in the soil water at different depths on 17th December 1991 …t ˆ 0† used as starting point for the simulations.

2.2.1. Flow domain and ®nite element grid We simulated water ¯ow and solute transport in a ¯ow domain of 6 m wide (representing half the drain spacing), and 2 m deep (Fig. 3), respectively. The drain was located at 97.5 cm depth, and was described as a semi-circular hole with the real physical dimensions (inner diameter ˆ 5 cm; outer diameter ˆ 7.5 cm). A ®nite element grid was created, the triangular elements being forced to be strati®ed according to the layered soil pro®le (Fig. 3). The distribution of the triangles was generated using an automatic grid generator (Simunek et al., 1996). The density of the grid was then evaluated, using the ®rst 40 days of leaching period 1991±1992 (0 , t , 40 d) as a test period. The grid density was judged suf®cient, when the solution always converged and the water balance error was below 1%. This percentage represents the errors in summed changes in volumetric soil water content in the entire ¯ow domain relative to the water ¯uxes across the boundaries of the ¯ow domain. We used a trial and error method to ®nd the grid represented in Fig. 3. Close to the soil surface and the drain, high grid densities were needed because the gradients in the hydraulic head are considerable and time variable. In the subsoil, especially, further away from the drain, relatively low grid densities could be used. 2.2.2. Soil hydraulic properties in HYDRUS-2D To reduce numerical problems at nodal points close to saturation in the HYDRUS-2D model, a volumetric water content u k with corresponding pressure head, hk, is de®ned, above which the unsaturated hydraulic conductivity varies linearly between Kk and Ks. The equation of Mualem (1976) for unsaturated hydraulic conductivity is used for water contents u , u k. A threshold hk ˆ 220 cm was chosen for the linear

0±20 20±50 50±80 80±120 0±120

Depth interval (cm)

970 1335 1760 2425

970 2000 2640 4850 10 460

60 120 245 610

Cl concentration (mg l 21)

Cl concentration (mg l 21)

Cl content (kg ha 21)

1947

1942

Year

60 180 365 1210 1815

Cl content (kg ha 21)

100 150 200 400

Cl concentration (mg l 21)

1991

60 180 240 800 1280

Cl content (kg ha 21)

Table 2 Cl concentrations in the soil water and Cl contents in the soil pro®le for different soil layers for different years. For the 0±20, 20±50, and 50±80 cm depth intervals, the data of 1942 and 1947 are taken from Smits et al. (1962). For the deeper layers extrapolation is used to estimate Cl concentrations in the 80±120 cm depth interval. An average volumetric water content u ˆ 0:5 is assumed to estimate the Cl contents in the soil layers. The data for 1991 were measured at the Lovinkhoeve farm, and ®eld water contents were used to calculate the Cl contents in the soil layers

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77 65

66

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

part of the hydraulic conductivity characteristics (De Vos et al., 1999). Lower hk values were tried, but resulted in numerical stability problems. In order to keep the number of model parameters to a minimum, one value of hk for all soil layers (Fig. 4) was chosen. The only difference between the hydraulic conductivity parameters of layers 4, 5, 6, 7 and 8 is the value of Ks (Table 1) and the corresponding steepness of K between h ˆ hk and h ˆ 0: Fig. 5 shows the hydraulic conductivity characteristics for three different soil layers. The consequence of the linear increase of K close to the saturation is that the zone with a thickness of uhk u ˆ 20 cm above the phreatic surface will contribute substantially to the horizontal water ¯ow and will behave almost like a saturated part of the soil. The choice of hk has an effect on the simulated depth of the phreatic surface. The simulated drain discharge rate is determined by both the distribution of K in the soil pro®le and the hydraulic head gradients, and is thus less sensitive to the exact position of the phreatic surface. In the comparison of measured versus simulated data, the drain discharge rate should therefore be weighted more than the groundwater level. 2.2.3. Calibration The use of the ®eld water retention curve for the topsoil (0±25 cm) (Table 1), together with the hourly precipitation and daily evaporation data resulted in simulated groundwater levels and drain discharge rates in close agreement with the measurements (De Vos et al., 2000). The dynamics of NO3 concentration in the drainage water appeared to be dominated by convective transport, which determines the peaks in NO3 concentrations. The longitudinal dispersivity, al ˆ 10 cm; and transverse dispersivity, at ˆ 1 cm; were obtained after calibration (De Vos et al., 2000). The order of magnitude of these dispersivities corresponds to the dispersivities of 5±20 cm given by Van Hoorn (1981) for ®eld-scale solute transport. 2.2.4. Initial and boundary conditions The initial …t ˆ 0† groundwater level midway between the drains was taken as representative of the depth of the initial phreatic surface. A constant hydraulic head, which corresponds with this phreatic surface, was assumed when calculating initial pressure heads. As an initial condition, the measured Cl and NO3 concentration pro®les on 17th December

1991 …t ˆ 0† were used, assuming a uniform distribution of Cl and NO3 as a function of the distance from the drain (Fig. 6). The simulation period ended on 8th April 1992 …t ˆ 114 d†: At the soil surface, precipitation and potential soil evaporation were given as top boundary ¯ux conditions. In the precipitation water, a Cl concentration of 5 mg l 21 and a NO3 concentration of 20 mg l 21 were used to simulate realistic deposition at the soil surface. The N deposition in the study period was 34 kg ha 21, which is normal for The Netherlands where N deposition can vary between 15 and 50 kg ha 21 yr 21 between different regions. At the bottom of the ¯ow domain (200 cm depth), a constant ¯ux boundary condition was prescribed, creating the ¯exibility to either use a no-¯ow condition (impermeable boundary) or a small upward or downward ¯ow. Because of the symmetry of the ¯ow problem, the left and right sides of the ¯ow domain were impermeable (Fig. 4). The inner wall of the drain was a seepage face, implying that the drain was practically always empty. 3. Cl mass balance 1942±1991 A Cl mass balance (kg ha 21) has been made for the period 1942±1991: M1942 2 M1991 1 Mfert 1 Mman 1 Mdep ˆ Mdisch 1 Muptake ;

…1†

where M1942 is the Cl in the soil pro®le in 1942, M1991 the Cl in the soil pro®le in 1991, Mfert the Cl applied as potassium chloride (KCl) fertiliser between 1942 and 1991, Mman the Cl applied as manure between 1942 and 1991, Mdep the Cl deposited in the precipitation water between 1942 and 1991, Mdisch the Cl discharged between 1942 and 1991, and Muptake is the Cl taken up by the crop between 1942 and 1991. Table 2 shows the total Cl concentrations and contents in the soil pro®le down to 120 cm depth in 1942, 1947 and 1991. We used an average volumetric water content u ˆ 0:5 for the calculations of the total Cl amounts in 1942 (M1942) and 1947, while for 1991 (M1991), the measured u values were used. To estimate the amount of Cl applied to the experimental ®eld, we studied the 1942±1991 archives of the Lovinkhoeve farm. KCl was applied as potassium

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

67

Fig. 7. (a) Measured precipitation rate, (b) groundwater level in the middle between the drains, (c) drain discharge rate, and (d) Cl and NO3 concentration in the drainage water, during the leaching period from 17th December 1991 …t ˆ 0† to 8th April 1992 …t ˆ 114 d†:

(K) mineral fertiliser (Mfert). Normally, KCl was applied before a potato or sugar beet crop. Before 1964, no KCl fertiliser was applied because the natural soil fertility with respect to K was adequate. Over the entire period 1964±1977, a total Cl amount of 120 kg ha 21 was applied. Later, the amounts of KCl fertiliser were drastically increased, resulting for the period 1977±1991 in a total Cl amount of 1450 kg ha 21. From 1975 to 1991, mushroom

compost was applied as organic manure, which resulted in a total Cl application Mman of 440 kg ha 21. The average Cl concentration in the precipitation water in the Noordoostpolder is 5 mg l 21 (Ridder, 1978). Taking the average annual 745 mm precipitation at the Lovinkhoeve results in a Cl deposition Mdep of 37 kg ha 21 yr 21. The crop uptake of Cl was relatively small. For the four year crop rotation, potato±winter wheat±sugar

68

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Fig. 8. Hydraulic heads measured midway between two adjacent drains at three depths at three distances from a ditch, during the leaching period from 17th December 1991 …t ˆ 0† to 8th April 1992 …t ˆ 114 d†:

beet±spring barley, taking average Cl contents in the crops according to Houba and Uittenbogaard (1994), the average amount of Cl removed by the harvested crop parts Muptake was 10 kg ha 21 yr 21. Summarising, over the period 1942±1991, a Cl amount of Mdisch ˆ M1942 2 M1991 1 Mfert 1 Mman 1 Mdep 2 Muptake ˆ 10 460 2 1280 1 1570 1 440 1 1810 2 490 ˆ 12 510 kg ha21 was discharged by the subsurface drain.

The annual precipitation excess of 300 mm over 49 years corresponds to an equivalent water column of 14.7 m. Using u ˆ 0:5; this results in a 29.4 m displacement of the soil water. For a one-dimensional piston ¯ow, this would mean that the water in the soil pro®le had been refreshed 24.5 times. The travel time density distribution approach (Raats, 1981) with a characteristic residence time taq ˆ 2 yr; in which the solute concentration of the drainage water behaves in a

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

69

similar way to the discharge of a well-mixed vessel, would also result in a nearly complete depletion of the Cl originally present in the 0±120 cm soil layer. However, the Cl contents in this soil layer in 1991 were still high, especially, at greater depths (.100 cm, Fig. 6). The period 1981±1991 was evaluated more closely to identify the origin of the high Cl contents in the soil pro®le measured in 1991. In the period 1981±1991, an average Cl amount of 90 kg ha 21 yr 21 was applied as KCl fertiliser, 28 kg ha 21 yr 21 was applied as mushroom compost, the natural deposition was 37 and 10 kg ha 21 yr 21 was removed by the crop. This results in a total Cl input of 145 kg ha 21 yr 21 at the soil surface. When all this Cl is dissolved in the 300 mm precipitation excess, it corresponds to a Cl concentration of 48 mg l 21 in the soil water. However, the initial Cl concentrations in the soil pro®le were at least a factor of 2 higher (Fig. 6), and the average concentration in the drainage water in 1991±1992 was a factor of 3 higher (Fig. 7). The high Cl concentrations in the topsoil can be explained by the relatively dry conditions before the sampling dates. However, the high Cl concentrations in the drainage water and subsoil (80±150 cm depth), and the imbalance between Cl input at the soil surface and drain discharge suggest another source of Cl, which increased its concentration in the soil pro®le.

extra Cl is entering the soil pro®le at 120 cm depth by convective transport because the hydraulic head gradients show no upward ¯ow.

3.1. Upward seepage and capillary rise

3.3. Diffusion of Cl from the subsoil

The initial Cl data (Fig. 6) show an increasing Cl concentration with depth. One has to know the application rate and water and solute extraction by plant roots and boundary conditions at the bottom of the soil pro®le to interpret the form of such a pro®le. During the growing season, Cl concentrations will become higher during passage through the root zone due to evaporation of water at the soil surface and selective uptake of solutes by the plant roots. Hydraulic heads at 150 cm depth are lower than the hydraulic head at 125 cm depth, indicating downward ¯ow and thus downward transport of Cl (Fig. 8). Capillary rise usually occurs in summer periods. The upward water ¯ux density can potentially reach 3 mm d 21 at depths between 50 and 100 cm (De Vos et al., 1994). This means a redistribution of water and Cl, resulting in an upward movement of Cl in the 0± 120 cm soil layer. However, on an annual basis, no

The Cl present in the soil after reclamation in 1942 in the 10 m deep soil is a result of diffusion of salt into the soil during the period from about 1600 to 1932, when the seawater above the lake bottom was saline (Van der Molen, 1958). The Cl content in the 0±10 m depth soil layer amounted to about 75 000 kg ha 21. After installation of the drains, the hydrological situation changed and most of the Cl was leached to the drains from the top (100 cm) of the soil pro®le. The Cl concentrations at greater depths (.100 cm) remained high (Fig. 6). The diffusion process now acts in the opposite direction: salt is transported upwards. We assumed that the impermeable layer starts at 120 cm depth and that the water ¯ow occurs only in the 0± 120 cm depth soil layer. The diffusive Cl ¯ux from the subsoil must be estimated at 120 cm depth to enable the Cl mass balance for the 0±120 cm soil layer to be estimated. Volker and Van der Molen (1991)

3.2. Subsurface in®ltration The ditch-water levels at the site were sometimes raised above drain depth to provide the neighbouring fruit growers with water for sprinklers used to prevent frost damage in the orchards. Under these conditions, ditch-water can subirrigate the soil pro®le through the subsurface drains. The water in the ditches originates mainly from the IJsselmeer and partly from the larger canals in the Noordoostpolder. The average Cl concentration in the IJsselmeer water was 150 mg l 21 (Dienst Getijdewateren en Dienst Binnenwateren, 1990). However, for the Cl concentrations to reach 300±600 mg l 21 at 100±150 cm depth in the soil pro®le (Fig. 6), originating from the subirrigation of ditch-water, the Cl concentrations in the ditchwater should have been above 600 mg l 21 because a certain degree of dilution due to dispersion after entering the soil may be expected. A Cl concentration of the ditch-water above 600 mg l 21 is unrealistically high compared to the Cl concentration of 150 mg l 21 in the IJsselmeer water. So, subirrigation did not play a substantial role and cannot explain the gap in the Cl mass balance for the 0±120 cm depth layer.

70

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Fig. 9. (a) Measured and simulated groundwater levels midway between the drains, (b) drain discharge rates, (c) Cl concentrations in the drainage water, and (d) NO3 concentrations in the drainage water, during the leaching period from 17th December 1991 to 8th April 1992 (0 , t , 114 d). The layer with a low permeability is assumed to start at 120 cm depth (Table 1) and there is assumed to be no downward seepage at 200 cm depth.

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

discussed the downward diffusion of salt in the subsoil of the Noordoostpolder. From the Cl pro®les in the soil, they found a diffusion coef®cient D ˆ 0:4 cm2 d21 for Cl in the pores of the sediments. Using the measured Cl concentration gradient at the Lovinkhoeve at 125±145 cm depth of 15 mg l 21 cm 21 (Fig. 6) and applying Fick's law, results in an upward diffusive Cl ¯ux of 220 kg ha 21 yr 21. So, upward Cl diffusion is a major term in the Cl mass balance. The characteristic length of the diffusion process is ldif ˆ 12 cm for t ˆ 1 yr; and ldif ˆ 85 cm for t ˆ 49 yr: This length ldif indicates the penetration height of the upward Cl diffusion. In the zone where water ¯ow plays an important role, the redistribution of Cl will be dominated by the convective transport because the annual water displacement of about 60 cm is greater than the characteristic diffusion length ldif. If the Cl concentrations in the subsoil (.120 cm depth) are not in¯uenced by processes, such as lateral groundwater ¯ow, the upward Cl diffusion will continue for many years and the Cl concentration in the drainage water originating from this source will only drop slowly due to depletion of the subsoil. 4. Water ¯ow and solute transport to a subsurface drain 4.1. Field measurements 4.1.1. Water balance 1991±1992 The soil water balance of the experimental ®eld was calculated for a well-de®ned volume of soil (Vs) for a unit surface area over the total measurement period: DW ˆ Qin 2 Qout ;

…2†

where DW is the increase in water storage in Vs (mm), Qin the cumulative amount of water entering Vs (mm), and Qout is the cumulative amount of water leaving Vs (mm). A depth of 120 cm was chosen as the basis for calculating Vs. Assuming that water (mm) can leave or enter Vs only via the top or bottom surface and via the drain, then Qin ˆ P 1 Qup ;

…3†

Qout ˆ E 1 Qdown 1 Qdrain ;

…4†

71

where P is the precipitation, E the soil evaporation, Qup the upward ¯ux from the subsoil, Qdown the downward ¯ux to the subsoil, and Qdrain is the drain discharge. Over the total period (0 , t , 114 d), the precipitation P was 213 mm, and the drain discharge Qdrain was 120 mm. Evaporation from the bare soil was estimated at E ˆ 70 mm (De Vos, 1997). There was an increase of DW ˆ 19 mm in the water content of the soil pro®le. From the integral water balance, it follows that Qdown 2 Qup ˆ 2DW 1 P 2 E 2 Qdrain ˆ 219 1 213 2 70 2 120 ˆ 4 mm water disappeared as net downward seepage. During the leaching period, the measured H gradients indicated downward ¯ow (Fig. 8). However, during relatively dry periods, the groundwater level never fell far below the drain depth (Fig. 7(b)). This indicates a small hydraulic conductivity in the subsoil, which is consistent with the measured hydraulic conductivities in the subsoil (Table 1). 4.1.2. Precipitation, drain discharge, groundwater levels, and Cl and NO3 concentrations in the drainage water 1991±1992 The drain discharge measurement period reported here ran from 17th December 1991 …t ˆ 0† to 8th April 1992 …t ˆ 114 d†; corresponding to the period of the ®rst post-summer drainage to the time of spring barley sowing. During the measurement period, the soil surface was bare. Fig. 7 shows a remarkably consistent inverse relationship between dissolved Cl concentration in the drainage water and drain discharge rate. Even peak drain discharge rates that appear as local maxima on larger peaks (e.g. 5 , t , 8 d in Fig. 7(d)) corresponded to local minima in the Cl concentration data. During the highest drain discharge around t ˆ 21 d; the Cl concentration fell from 180 to 90 mg l 21. The ¯uctuations of NO3 concentration in the drainage water (Fig. 7(d)) corresponded directly with those of the groundwater levels (Fig. 7(b)) and drain discharge rates (Fig. 7(c)). The ¯uctuations in NO3 concentrations were opposite to the ¯uctuations in Cl concentrations. Fig. 6 shows the initial Cl and NO3 concentrations in the soil pro®le. The lowest Cl concentrations occurred in the 0±50 cm soil depth interval. For depths deeper than 100 cm, the Cl concentrations

72

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Table 3 Hydraulic characteristics of the different soil layers for the simulation with a nearly impermeable subsoil (layer 8, 110±200 cm). The Ks of the subsoil allows a small downward seepage of 0.04 mm d 21. An explanation of the parameters is given in Table 1 Zone

Layer

Depth (cm)

u r (±)

u s (±)

Ks (cm d 21)

a (cm 21)

n

L

u k (±)

Kk (cm d 21)

1

1

0±25

0.04040

0.43709

200

0.02190

1.30740

2 2.02754

0.41072

0.206

2

2 3

25±35 35±40

0.08165 0.08165

0.47303 0.47303

200 300

0.00525 0.00525

1.62422 1.62422

2 1.84898 2 1.84898

0.46923 0.46923

0.949 0.949

3

4 5 6 7 8

40±50 50±70 70±95 95±110 110±200

0.10060 0.10060 0.10060 0.10060 0.10060

0.51506 0.51506 0.51506 0.51506 0.51506

120 40 20 10 0.005

0.00207 0.00207 0.00207 0.00207 0.00207

1.91432 1.91432 1.91432 1.91432 1.91432

2 1.46461 2 1.46461 2 1.46461 2 1.46461 2 1.46461

0.51461 0.51461 0.51461 0.51461 0.51461

0.334 0.334 0.334 0.334 0.005

4

9

Trench

0.08165

0.47303

200

0.00525

1.62422

2 1.84898

0.46923

0.949

were high. The trend in NO3 concentrations with depth was opposite to the trend in Cl concentrations. 4.2. Simulations The simulated Cl discharge appeared to be very sensitive to the initial Cl concentration distribution in the soil pro®le. Cl concentrations at the depth of the drain (Fig. 6) are initially high and cause high simulated Cl concentrations in the drainage water. It will be shown that even a small contribution of convective transport from the zone at depths .110 cm will give a large contribution to Cl discharge through the drain because the concentrations are very high in this zone. To illustrate this effect, we performed two simulation runs. The ®rst simulation used a small Ks in the subsoil with the impermeable layer at 200 cm depth. In the second simulation, we took as bottom boundary condition, a relatively small downward seepage of 4 mm during the 114-day period, with the impermeable layer at 110 cm depth. 4.2.1. Impermeable layer at 200 cm depth and a small Ks in the subsoil In the initial simulations, the bottom of the soil pro®le was assumed to be impermeable. No water ¯ow or solute transport was possible at 200 cm depth, and the hydraulic conductivity at saturation Ks was relatively small for depths deeper than 120 cm (Table 1), but water ¯ow and solute transport were still possible in this zone. Simulated ground-

water levels are generally at lower depths than measured, except at times of high precipitation rates, i.e. around t ˆ 7 d and t ˆ 21 d (Fig. 9(a)). Simulated and measured drain discharge rates in the course of time correspond well (Fig. 9(b)). Simulated peak discharge rates are sometimes higher than measured discharge rates. However, in dry periods, i.e. 40 , t , 80 d, the simulated drain discharge starts later than measured, which is related to the deeper simulated groundwater level and the corresponding larger water storage capacity. Measured cumulative drain discharge was 138 mm and simulated cumulative drain discharge was 136 mm. The simulated Cl concentrations in the drainage water are greater than the measured Cl concentrations by a factor of 1.75 (Fig. 9(c)). However, the relative changes in Cl concentrations are simulated very well. The minimum in Cl concentrations, correspond with high drain discharge rates, the maximum in Cl concentrations correspond with low drain discharge rates. Measured cumulative Cl discharge was 180 kg ha 21 and simulated Cl discharge was 316 kg ha 21. The changes in NO3 concentrations in the drainage water in the course of time are well simulated (Fig. 9(d)). Measured cumulative N discharge was 11 and simulated cumulative N discharge was 12 kg ha 21. 4.2.2. Impermeable layer at 110 cm depth and 4 mm downward seepage We repeated the simulations with a small hydraulic conductivity starting at 110 cm depth (Table 3). We

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

73

Fig. 10. (a) Measured and simulated groundwater levels midway between the drains, (b) drain discharge rates, (c) Cl concentrations in the drainage water, and (d) NO3 concentrations in the drainage water, during the leaching period from 17th December 1991 to 8th April 1992 (0 , t , 114 d). The layer with a very low hydraulic conductivity is assumed to start at 110 cm depth (Table 3) and a 4 mm downward seepage is assumed at 200 cm depth.

74

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

Fig. 11. Distribution of the water ¯ow for steady in®ltration rates of (a) 15 mm d 21 and (b) 1 mm d 21 for a situation with an impermeable boundary at 200 cm depth. The lengths of the arrows indicate the magnitude of the ¯ow relative to the maximum. At the same arrow length, the absolute water ¯ow is 20 times greater in the (a) top ®gure than in the (b) bottom ®gure. The thick curves show the depth of the phreatic surface. The hydraulic properties of Table 1 were used for this example.

also took as bottom boundary condition a relatively small downward seepage of 4 mm during the 114-day period. Hence, only a small water ¯ow and convective solute transport component were possible in this zone. The simulated Cl concentration in the drainage water was then closer to the measured data, while the dynamics in Cl concentrations remained realistic (Fig. 10(c)). The cumulative simulated Cl discharge amounted to 226 kg ha 21, which is closer to the measured 180 kg ha 21 than the initial simulations. The steep vertical gradient in Cl concentrations explains the sensitivity of the simulations to the position of the impermeable layer. By reducing the convective transport in this zone with initial Cl concentrations of 600 mg l 21, a substantial reduction in Cl transport was obtained. The effect of changing the hydraulic properties of the subsoil on the simulated groundwater level±drain discharge relationship was small. The more super®cial soil layers are more signi®cant for water ¯ow over the whole soil pro®le, which can be seen in the simulated and measured

groundwater level and drain discharge (Fig. 10(a) and (b)). The simulations of NO3 transport were also affected by changes in the hydraulic properties of the subsoil and the 4 mm downward seepage. NO3 concentrations in the drainage water are overestimated for the entire simulation period (Fig. 10(d)). The ¯uctuations in NO3 concentrations in the drainage water in the course of time are accurately re¯ected. Measured cumulative N discharge was 11 kg ha 21 and simulated cumulative N discharge, 15 kg ha 21. 4.3. Discussion Under discharge conditions with a groundwater level close to drain depth, as in the period 30 , t , 90 d, the origin of the discharged water was mainly from depths below the drain. The Cl concentrations in the soil water were relatively high in that zone (Fig. 6). When the groundwater level rose for a short period to a shallow depth (0±50 cm), as in the periods 20 , t , 25 d and

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

90 , t , 100 d, a substantial part of the discharged water originated from the zone where Cl concentrations were lower. The discharged water is always a mixture of water from different origins, but the sharp changes in Cl concentrations indicate an abrupt shift of the stream tubes and a contribution to drainage and leaching from correspondingly different zones in the soil. This explanation is supported by the NO3 data, which show opposite trends to Cl, both in soil pro®le distribution and drainage water concentration. These trends, which are inferred from ®eld data only, are con®rmed by the simulations. The simulations clearly demonstrate the importance of water ¯ow on convective transport in a twodimensional ¯ow domain. The vertical distribution of the soil hydraulic characteristics, e.g. hydraulic conductivity at saturation, had a pronounced effect on the ¯ow paths to the drain under different atmospheric conditions (Fig. 11). Under dry conditions with a relatively deep phreatic surface, drainage water mainly originated from the zone close to the drain depth. Under extremely wet conditions, the major part of the drainage water originated from the topsoil. Due to the relatively high hydraulic conductivity in this zone, rapid lateral transport occurred towards the drain. The opposite behaviour of Cl and NO3 concentrations in the drainage water is explained by these changing ¯ow conditions and the inverse distribution of Cl and NO3 in the soil pro®le. However, the quantitative simulation results are very sensitive to the initial solute concentration distribution and the hydraulic properties of the different soil layers, especially, the depth of the impermeable layer. In the simulations, the diffusion from the subsoil is accounted for, and serves as a Cl source for the topsoil. Since the simulations we were dealing with rather wet situations …u . 0:3†; we did not consider the impact of the poor ®t in the dry part of the hydraulic conductivity characteristics (Fig. 5). Instead, we focused on the wet parts of the characteristics close to saturation because the impact on solute transport will be much larger. However, a complete sensitivity and uncertainty analysis were beyond the scope of this study. Therefore, we presented the simulation results as a means to help us understand the transport processes, rather than attempt to give quantitative results. Sharp peaks in Cl or NO3 concentrations in drainage water are often explained by preferential ¯ow

75

mechanisms (Nieber and Misra, 1993). The high values of Ks for the topsoil suggest the occurrence of macro-pores, worm holes, old root channels and small cracks. Evidence of the presence and functioning of these larger pores was found during the Ks measurements and a later (1994) bromide tracer experiment (De Vos, 1997). However, our study shows that a two-dimensional convection±dispersion approach explains the main features of the peaks in solute concentration. De Vos et al. (2000) showed that for this soil an annual net N mineralisation of 120 kg N ha 21 yr 21 and a denitri®cation of about 20 kg N ha 21 yr 21 is to be expected. In the simulations, nitrogen transformations are not accounted for in the description of NO3 leaching. The relatively high simulated NO3 concentrations in the drainage water (Fig. 10(d)) will become lower when denitri®cation plays an important role, which is reasonable to assume during the wet, anaerobic conditions in the leaching period. However, to interpret the N dynamics properly, all N transformations have to be taken into account, which was beyond the scope of this study. Therefore, in our study, the data on NO3 leaching were only used to show the effect of dynamics of water ¯ow in a two-dimensional ¯ow domain and to support the results found for Cl. Due to uncertainties in transport parameters, we consider the simulation results to be an explanation of the dynamics of the transport processes, rather than being precise quantitative results. The simulation model HYDRUS-2D appeared to be a powerful tool for analysing the complex transport problem. In principle, we could predict trends in Clconcentration in the future using our simulation model. However, this depends also on the boundary conditions in the subsoil. We are not sure of the Cl pro®les at depths below 150 cm depth and this can be important in relation to the depletion of the subsoil. The salinisation process took a few hundred years, the reverse process will have a similar time-scale. It would take too much computer time to calculate future decreasing trends in Cl concentrations using the HYDRUS-2D model. Due to a lack of information about the current status of Cl concentrations in the deeper subsoil, we were

76

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77

not able to produce realistic quantitative predictions of Cl in the drainage water. 5. Conclusions The long-term 1942±1991 Cl mass balance showed that upward diffusion from the subsoil was the main source of high Cl concentrations in the drainage water, rather than the application of fertiliser. Historical data of Van der Molen (1958) indicated that the large amounts of Cl are still located in the deeper soil layers. Based on the diffusion coef®cient of D ˆ 0:4 cm2 d21 (Volker and Van der Molen, 1991), the characteristic length scale of annual diffusive transport is ldif ˆ 12 cm: The upward diffusion from the subsoil (.120 cm depth) of about 220 kg ha 21 yr 21 is a major term in the Cl mass balance, compared to the average fertiliser application of 118 kg ha 21 yr 21, natural deposition of 37 kg ha 21 yr 21, and crop uptake of 10 kg ha 21 yr 21. Annual Cl drain discharge is estimated at 365 kg ha 21 yr 21. In the winter leaching period 1991±1992, the Cl drain discharge was only 180 kg ha 21 yr 21, due to the relatively small precipitation excess in that period. High Cl concentrations in the drainage water in 1991 originate from the subsoil (.120 cm depth). A small downward seepage of 0.04 mm d 21, which corresponds to an annual vertical displacement of 30 mm, was measured during the monitoring period in 1991±1992. So, the subsoil diffusion is more important than convection. The dynamics of water ¯ow and solute transport are well described by the HYDRUS-2D model. The convective transport component in the topsoil dominated the occurrence of the peaks in solute concentrations in the drainage water. The large lateral ¯ow component towards the drain just above and below the phreatic surface explains the minima in Cl and the peaks in NO3 concentrations in the drainage water. This opposite behaviour can be understood from the opposite initial distributions of Cl and NO3 in the soil pro®le. The simulated Cl concentrations over the leaching period 1991±1992 were closer to the measured concentrations. The NO3 concentrations in the drainage water are overestimated, which can be related to our ignorance of N transformations in the soil pro®le. In this paper, NO3 data are only presented to support the conclusion that solute concentrations in

the drainage water are originating from different zones in the soil pro®le. Since the Cl concentrations in the subsoil are very high and the NO3 concentrations are very low, both Cl and NO3 transport simulations are sensitive to the location of the impermeable layer. The combination of historical data, ®eld experiments and a simulation model explain the origin of Cl. The high Cl concentrations in the drainage water are natural background concentrations for the reclaimed polder soil. Our research approach can be used for similar problems in which the source of the load on surface waters is not yet known.

References De Vos, J.A., 1997. Water Flow and Nutrient Transport in a Layered Silt Loam Soil. PhD Thesis. Wageningen Agricultural University, The Netherlands. De Vos, J.A., 2001. Monitoring nitrate leaching from submerged drains. J. Environ. Qual. 30, 1092±1096. De Vos, J.A., Raats, P.A.C., Vos, E.C., 1994. Macroscopic soil physical processes considered within an agronomical and a soil biological context. Agric. Ecosyst. Environ. 51, 43±73. De Vos, J.A., Simunek, J., Raats, P.A.C., Feddes, R.A., 1999. Identi®cation of the hydraulic characteristics of a layered silt loam soil. In: Van Genuchten, M.Th., Leij, F. (Eds.). Characterisation and Measurement of the Hydraulic Properties of Unsaturated Porous Media, Part 1. University of California, Riverside, USA. pp. 783±798. De Vos, J.A., Hesterberg, D., Raats, P.A.C., 2000. Nitrate leaching in a tile-drained silt loam soil. Soil Sci. Soc. Am. J. 64 (2), 517± 526. Dienst Getijdewateren en Dienst Binnenwateren, 1990. Jaarboek van afvoeren, waterstanden, golven en waterkwaliteit 1987. 's Gravenhage (in Dutch). EU, 1991. Richtlijn van de Raad van 12 December 1991 Inzake de Bescherming van Wateren Tegen Verontreininging Door Nitraten uit Agrarische Bronnen. , Richtlijn Nummer 91/676/EG. European Union, Brussels in Dutch. FAO-UNESCO, 1974. Soil Map of the World, vol. I. FAO, Rome. Fletcher, D.A., 1991. A national perspective. In: Follett, R.F., Keeney, D.R., Cruse, R.M. (Eds.). Managing Nitrogen for Groundwater Quality and Farm Pro®tability. Soil Sci. Soc. Am., Madison, USA. pp. 9±17. Fokkens, B., 1966. De Landbouwwaterhuishouding in De Noordoostpolder, Mede in Vergelijking Met Die in Andere Zeekleipolders. , Van Zee tot Land, nr. 44. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Houba, V.J.G., Uittenbogaard, J., 1994. Chemical Composition of Various Plant Species. International Plant-Analytical Exchange (IPE), Department of Soil Science and Plant Nutrition, Wageningen Agricultural University, The Netherlands.

J.A. de Vos et al. / Journal of Hydrology 257 (2002) 59±77 Kooistra, M.J., Lebbink, G., Brussaard, L., 1989. The Dutch programme on soil ecology farming systems. 2. Geogenesis, agricultural history, ®eld site characteristics and present farming systems at the Lovinkhoeve experimental farm. Agric. Ecosyst. Environ. 27, 361±387. Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resour. Res. 12, 513±522. Nieber, J.L., Misra, D., 1993. Modelling ¯ow and transport in heterogeneous dual-porosity drained soils. 15th International Congress of ICID, The Hague, Workshop on Subsurface Drainage Simulation Models. ICID-CIID, Cemagref, pp. 174-184. Raats, P.A.C., 1981. Residence times of water and solutes within and below the root zone. Agric. Water Mgmt 4, 63±82. Ridder, T.B., 1978. Over de Chemie van de Neerslag; Vergelijking Meetresultaten. , Scienti®c Report W.R. 78-4.. KNMI, De Bilt, The Netherlands in Dutch. Sieben, W.H., 1951. Drainage of Silt Loam Soils in the Noordoostpolder. , Van Zee tot Land, nr. 3. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Simunek, J., Sejna, M., Van Genuchten, M.Th., 1996. The HYDRUS-2D Software Package for Simulating Water Flow and Solute Transport in Two-dimensional Variably Saturated Media. , Version 1.0, IGWMC-TPS-53. International GroundWater Modeling Center, Colorado School of Mines, Golden, CO 167pp. Smits, H., Zuur, A.J., Schreven, D.A., Bosma, W.A., 1962. De Fysische, Chemische en Microbiologische Rijping der Gronden in de IJsselmeerpolders. , Van Zee tot Land, nr. 32. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Soil Survey Staff, 1975. Soil Taxonomy. A Basic System of Soil

77

Classi®cation for Making and Interpreting Soil Surveys. , US Department of Agriculture, Agricultural Handbook 436. US Governmental Printing Of®ce, Washington, DC. Strebel, O., Duynisveld, W.H.M., Bottcher, J., 1989. Nitrate pollution of groundwater in Western Europe. Agric. Ecosyst. Environ. 26, 189±214. Van der Molen, W.H., 1958. Over de Zouthuishouding in de Noordoostpolder. , Van Zee tot Land, nr. 23. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Van der Molen, W.H., Sieben, W.H., 1955. Over de Landbouwkundige Betekenis en de Kartering van de Kwel in de Noordoostpolder. , Van Zee tot Land, nr. 12. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Van Genuchten, M.Th., 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892±898. Van Hoorn, J.W., 1981. Salt movement in unsaturated soil pro®les. In: Verruijt, A., Barends, F.B.J. (Eds.). Flow and Transport in Porous Media. Proceedings on the Euromech. 143, September 2±4, DelftBalkema, Rotterdam/The Netherlands, pp. 227±232. Volker, A., Van der Molen, W.H., 1991. The in¯uence of groundwater currents on diffusion processes in a lake bottom: an old report reviewed. J. Hydrol. 126, 159±169. Wiggers, A.J., 1955. Het Ontstaan van de Noordoostpolder. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Zuur, A.J., 1951. Ontstaan en Aard van Bodem van de Noordoostpolder. , Van Zee tot Land, nr. 1. Tjeenk Willink NV, Zwolle, The Netherlands in Dutch. Zuur, A.J., 1952. Drainage and reclamation of lakes and of the Zuiderzee. Soil Sci., 75±89.