Effect of preferential transport and coherent denitrification on leaching of nitrate to drainage

To protect the quality of the aquatic environment, it is imperative to be able to assess the leaching of nitrate through various hydrogeological settings. Numerical model concepts have been developed in order to describe this leaching and possible routes of nitrogen at field scale, often without being evaluated in regard to their ability to account for dominant preferential transport and coherent denitrification, which is the rule rather than the exception in soils. This study evaluates whether it is possible to describe 10-years of nitrate concentrations, measured in drainage from a tile-drained agricultural clay till field in 5 Denmark, by applying the soil-plant-atmosphere model DAISY, capable of accounting for preferential transport and denitrification. A DAISY model concept, including macropores capable of capturing the water and bromide balance of the field within this specific timeframe, was able to predict the water transport to drainage, dry matter and N-yield of the harvested crops, while it was unable, with the standard default denitrification model, to predict dynamics and quantity of N-loss to drainage. This was caused by a fast saturation of the plow layer, where nitrate seemed to be denitrified almost instantly, and no surplus nitrate 10 remained to be transported to the drainage. To circumvent this and describe the measured N-loss, modification to the water reduction function affecting denitrification was conducted. The denitrification had to be reduced by approximately 50% from a seasonal average of 75 kg Nha−1 to 35 kg Nha−1 while 48% to 80% of the total N-loss to drainage had to be preferentially transported from the plow layer. This study therefore reveals that, by not accounting for preferential transport and coherent denitrification, there is a high risk of underestimating leaching of nitrate to the aquatic environment. 15

,where θ [VV −1 ] is the volumetric water content in the soil, ψ [L] is the soil water pressure potential, K [LT −1 ] is the hydraulic conductivity of the soil and S is a sink term, which represents the loss to the drain, macropore or plant water uptake. In order to solve the Richards equation, in this study, the van Genuchten (vG) soil water retention model was used (Van Genuchten, 1980).
, where α, n and m are empirical shape parameters, and θ s and θ r are the saturated and the residual water content respectively of the given soil. In this study, the van Genuchten model is coupled with the Mualem hydraulic conductivity theory (Mualem, 1976) (vGM), where m is achieved as m = 1 − n −1 . The Mualem hydraulic conductivity is expressed as (3) 135 , where K sat is the hydraulic conductivity at saturation , S e is the effective saturation, which is calculated as, S e = θ−θr θs−θr and l is the shape form, which represents the pore connectivity.
The fast flow domain in DAISY is described by a macropore module designed by Mollerup (2010) and tested in technical reports, prepared for and published by the Danish Environmental Protection Agency (Hansen et al., 2010a(Hansen et al., , b, 2012b. The macropore is a vertically oriented feature in the DAISY model, characterised by physical properties such as length, diameter 140 (d) and density (ρ). The macropore flow is initiated when the matrix pressure exceeds a specific pressure potential called ψ init .
If this pressure potential is exceeded, the macropore domain activates and water starts to fill up the macropore. When the pressure potential drops below a level called ψ term , the macropore flow is terminated. In a specific case, when a macropore is filled with water, it can be transferred back to the soil matrix at a certain point. It is initiated when the pressure difference between the macropore and the matrix exceeds a minimum pressure barrier ψ barrier . All pressure parameters are common for 145 all macropore classes. Macropores can be drain ended and matrix ended macropore, where the drain ended macropore has no water build up, because when the water enters into the macropore, it is instantaneously transported to the drain. In contrast, in the matrix macropores, the water has the availability to build up and be transported back to the matrix later, as discussed above. More details of the mathematical and physical description of the macropore domain can be found in Mollerup (2010).
Soil N pools: Nitrogen (N) is present in two different fractions in the soil as organic or inorganic forms. These two fractions are 150 in a constant exchange with each other, through mineralisation and immobilisation. According to Jarvis et al. (1996), inorganic N represents approximately 5% of the total soil N, although it may change after fertilisation for a short period. Hence, organic N, which occurs in many forms, including proteins, urea, amino and nucleic acids and nucleotides is by far the largest N fraction of the soil.
[ Figure 3 is about here] 155 Although the microbial state of N may cover only 3-5%, most of the transformation processes (mineralisation, immobilisation, and denitrification) are mediated by the microbial community. In DAISY, soil N is divided into six different pools (Fig. 3), and some N is in an inert pool (not shown in Fig. 3). The pools are separated into two distinct groups, pools with slow turnover rate (denoted 1), and pools with higher turnover rate (denoted 2). Thereby the soil organic matter pool 1 (SOM1) mainly contains chemically stabilised compounds which are relatively resistant to biological degradation. The other organic matter 160 pool (SOM2) is physically stabilised and more labile, although temporarily resistant to biodegradation due to sorption to soil colloids. Added organic matter refers to manure, crop residue or green manure and is typically divided into AOM1 -cell wall material and AOM2 cell extractable substances. The main driver of the C/N turnover is the soil microbial biomass (SMB), which controls the turnover processes of the dissolved organic matter, even though it only represents a small quantity of the total organic matter (Hansen, 2002). Mineralisation-immobilisation turnover: Net N mineralisation or net N immobilisation is 165 determined by the microbial activity and the overall N balance. If the content of N in the assimilated organic substance is higher than that required by the biomass for growth, ammonium is excreted to the soil solution. On the other hand, if the content of N in the assimilated organic substance is lower than that required by the biomass for growth, ammonium or nitrate is assimilated from the soil solution and transformed into nitrogenous organic compounds (Hansen, 2002). The measure used in DAISY for the available organic substrate is the content of carbon in the organic matter. Hence, the simulation of net mineralisation of 170 N is based on the simulation of the turnover rate of soil organic carbon. The potential decomposition rate of organic carbon in various pools in the soil is described by first-order kinetics, but is affected by the abiotic factors (soil water content, soil temperature, pH (5 to 8), oxygen pressure) and availability of inorganic N. The potential N mineralisation rate is strongly related to the carbon turnover, as every sub-pool has a C:N ratio and the decomposition of carbon leads to mineralisation of N carbon according to this ratio. Hence, the potential background N mineralisation from dead native organic matter in the 175 soil is highly dependent on the distribution of the dead native soil organic matter between SOM1 and SOM2, which in turn is strongly related. By default, the C:N ratio for SMB1 and SMB2 is assumed to be 6 and 10, respectively, however, it can be specified differently if required. Denitrification: In the present model, denitrification is simulated using a rather simple index type model considering the decomposition of organic matter, volume of anaerobic microsites expressed simply in terms of soil water content, soil temperature, and the concentration of nitrate in soil solution. This is a typical way of using a simplified 180 model for denitrification (Eq. 4), according to Heinen (2006): ,where D a is the actual denitrification rate, α d represents the potential but may in different models have different formulation, f N is a dimensionless reduction function for nitrate content in soil or represents the nitrate content in the soil (depending on the exact formulation determined by α d ), f S is a dimensionless reduction function for water content in the soil, f T is a 185 6 https://doi.org/10.5194/hess-2019-666 Preprint. Discussion started: 16 January 2020 c Author(s) 2020. CC BY 4.0 License. dimensionless reduction function for temperature in the soil, and f pH is a dimensionless reduction function for soil pH. The α d parameter can be considered in two ways, depending on the model concept; either α d represents the potential denitrification rate D p (same units as D a ) or it represents a first-order denitrification coefficient (constant) k d . In both cases, α can be a constant parameter or can be related to carbon dynamics. In the DAISY model, the potential denitrification rate (in case of anoxic conditions and sufficient nitrate concentration in the soil solution) is expressed as a linear function of the CO 2 evolution 190 rate: ,where xi * d is the potential denitrification rate of the soil, xi CO2 is the CO 2 evolution rate simulated by the mineralisation -immobilisation -turnover model (MIT-model), and α * d is an empirical constant, which was taken from Lind (1980), who measured the relationship between easily decomposable organic matter and denitrification capacity. The actual denitrification 195 rate is determined either by the actual microbial activity at the anaerobic microsites, or the transport of nitrate to the anaerobic microsites represented by the left and right solution, respectively, in Eq. (6). In the case of ample supply of nitrate, the actual denitrification rate is determined by multiplying the potential denitrification rate by a modifier function. Hence, the actual denitrification can be simulated as:

200
,where xi d is the actual denitrification, K d is an empirical proportionality factor when denitrification is governed by the microbial activity at the anaerobic microsites and N n i the nitrate concentration in the soil. The maximum transport of nitrate to microsites can be assumed to be relative to the nitrate concentration in the soil (N ni = θ C ni , where C ni is the concentration in the soil solution, and θ is the soil water content). In Eq. (6), the modifier function f S is assumed to be a function of the soil water content. Many models use a power reduction function of the form (Grundmann and Rolston, 1987): ,where f S is the dimensionless power water reduction function in the range [0, 1], S is the dimensionless degree of water saturation or water-filled pore space; S is always in the range [0, 1], S m is close to full water saturation above which f S =1, S t is a threshold value for S below which f S =0, w is a curve shape parameter determining the steepness of the curve (Heinen, 2006). The temperature modifier function f T is an Arrhenius like function (Rodrigo et al., 1997) by correlating the exponential 210 rate of biological processes to the increasing temperature. According to Heinen (2006), based on a sensitivity test on Eq.(4), f S is the most sensitive within all modifiers. Therefore, this study is only focusing on the water saturation effect on denitrification in the aspect of calibration Eq.(4).
settings and three different horizons (A, B and C). The macropore settings included vertical macropore transport supplying water directly to 1) the drainage pipes, 2) to drainage pipes and the matrix 3) to drainage pipes and the matrix added with fractures, supplying water to the matrix in the saturated zone. The best-calibrated concept with drainage ending macropores (DM1, DM2) and matrix (MM1, MM2, MM3) ended macropores (Fig 4a) (Fig. 4b, 4c).
[ Figure 4 is about here]

Objectives included in the automated calibration procedure
The soil and hydrological parameters are adjusted in order to improve the bromide (Br -) transport as in the hydrological model by Nagy et al. (2019), which was not calibrated on Brtransport. A better performance of the model could particularly be 225 expected just after the application of 30 kg KBr ha −1 , corresponding to 20.14 kg Br ha −1 (cf. Fig. 9 in Nagy et al. (2019)).
Therefore new objectives have been included in the calibration procedure, such as the harvested dry matter yield -DM yield, function, which is eligible for automated calibration (Criss and Winston, 2008;Nagy et al., 2019).
,where N is the number of observations of a given objective within a year and ,where k is the number of calibration years. Due to DM yield and N yield having only one value per season, a seasonal aggregation is not possible. Thus, nMAE was calculated for all seasons and normalized by means of the observations. The same applied to the Br -(BRD1, BRC1, BRD2, BRC2) objectives due to the tracer experiment being held for one season in [2000][2001]. The multi-objective function calculated for the automated calibration is expressed as: For further evaluation purposes nRMSE[%] (normalized Root Mean Squared Error), normalized on the difference on the minimum-maximum deviation of the observation and KGE (Kling-Gupta efficiency measure were calculated. KGE is presented by Gupta et al. (2009)). Singh et al. (2005) and Hansson and Hokfelt (1975) suggested that if MAE or RMSE of the model is lower than half of the standard deviation (SD) of the measured data, the model may be considered as an adequate representation 250 of the measured data. On the other hand, KGE with a range of -∞ to 1, if KGE is above 0.5, the model can be considered as satisfactory.

Parameters
The hydrological parameters were taken from Nagy et al. (2019) as it is considered to be a reasonable baseline for the calibration (Table 1 and 2). All initial hydraulic parameters represented presented in Table 1, and 2 were given a ±5 % uncertainty 255 boundary range in order to see which parameter would be influential on the "new" objectives.
[ Table 1 is about here] For the fast flow domain, the macropore model of DAISY using the conceptualisation of Nagy et al. (2019) was applied (Table   2).
[ To be able to evaluate the model behaviour considering the "new" objectives conditions, one has to determine essential input parameters important for calibrating N loss by crop harvest, nitrate leaching and gaseous N 2 loss due to denitrification (Table   3). Only crop parameters influencing dry matter formation were selected, as the data available did not allow a calibration of the N uptake parameters.
[ Table 3 is about here]

265
The simple photosynthesis description in DAISY requires a value for maximum assimilation rate ( −1 ] at low light, and a temperature factor for assimilating production, referred to as a piece-wise linear function (PLF) (Hansen et al., 2012a;Vries, 1989). Also, to get more control over the crop production the conversion efficiency (growth respiration) et al., 2016;Vries, 1989) was taken into the calibration besides the Photosynthetic Active Radiation extinction coefficient (P ar ext ) and the temperature sum at emergence (T sum , Table 3). Additionally, the crop uptake reflection factor of Br -(CU RF BR , (Hansen, 2002)), the SOM fraction ratio of the plow layer (SOM ratio ) which describes the ratio SOM1:SOM2 has been added as a parameter, as well as parameters from the denitrification module for both fast and slow pools: the anaerobic denitrification constant (α * d ) and the empirical proportionality factor (K d ) from Eq.(5) and Eq.(6), respectively, and S t and w from Eq.(7) ( Table 4).

275
[ Table 4 is about here] The initial values of the selected crop and denitrification parameters were based partly on the values recommended for DAISY in the model library (https://DAISY.ku.dk/, Hansen et al. (2012a)) and on literature screening. Due to the lack of data on the crop growth, N uptake, and partitioning during the growth season, no extensive calibration of the crop models was possible.

Sensitivity analysis 280
The range of crop parameter values is biologically constrained by the diversity of crops and their cultivars. Given the lack of knowledge associated with the range of the variability that is genetic for most of the crop model parameters, uniform distribution for each parameter was assumed with ±20% uncertainty bound except P ar ext where the bounds were set ±50%.
All denitrification related parameters were given ±10% uncertainty bounds except for the water reduction function, the ranges for which were taken from Heinen (2006). The SOM fraction ratio varied within from 0.43 to 2.33, which indicate fractions 285 of SOM1:SOM2 as 0.3:0.7 to 0.7:0.3 and CU RF BR could vary from 0 to 1, where 1 means no crop uptake. Key parameters directly related to crop development, leaf photosynthesis, and net mineralisation of plow layer as one of the input source of denitrification, were tested for sensitivity by the Morris sensitivity screening (Campolongo et al., 2007;Morris, 1991;Nagy et al., 2019) in order to help in finding sensitive parameters that have the most influence on objectives presented in section 2.4. The results of the sensitivity screening were turned into the Morris distance( ) (Ciric et al., 2012;Jabloun, 2015), which 290 represents the Euclidean distance of the parameter from (0,0) on the µ * -σ coordinate system (Campolongo et al., 2007). The decision on sensitivity threshold was made by K-Means Clustering (Jain and Dubes, 1988). For each objective, the parameters were clustered into 3 groups (Low, Medium, High) by its .

Calibration methodology
The overall objective function was based similar to the mathematical formulation, which was used in Nagy et al.

Sensitivity screening results
Overall six soil matrix parameters per horizon (A, B, C), the SOM ratio for the A horizon plus two horizon depth parameters, two macropore parameters per macropore type (DM1, DM2, MM1, MM2, MM3), two soil water pressure parameters for macropores, four denitrification parameter per pool (slow/fast) and one Brparameter called soil-hydraulic parameters (SH) and eight parameters per crop as crop parameters (CP) were tested for sensitivity. All parameters were selected as a sensitive 305 parameter, which belongs to group High, in at least one of the objectives (Fig. 5). However, to put more emphasis on the ND and NC objectives, parameters which belonged to sensitivity group Medium of ND and NC were also selected as sensitive and added to the calibration parameters (Fig. 5). All selected parameters which were involved in sensitivity analysis against 12 objective functions were listed with their associated sensitivity group (Fig. 5). Parameters with black letters are the selected sensitive parameters and with gray letters the non-sensitive. Even though a large number of parameters have been evaluated, 310 only 16 CP and 21 SH parameters showed to be sensitive to the objective functions. All the sensitive CP parameters were, as expected, sensitive to N yield, since all crop parameters are related to crop growth and therefore directly affecting the crop N uptake.
[ Figure 5 is about here] SH parameters showed more diversity regarding sensitivity towards the objectives. This is in contrast to CP parameters, which 315 mainly were sensitive towards N yield and DM yield. There was no single parameter which showed to be sensitive for all objectives. Only one SH parameter, S t , affecting the denitrification reduction factor, had an effect on DM yield (Fig. 5). As also seen in Figure 5, no matrix and macropore SH parameter was sensitive in the objective function for DM yield. Therefore, it can be inferred that the crops were not affected by water stress. Furthermore, no macropore SH parameters were sensitive in the objective functions for DM yield and N yield. Even though the uncertainty bound was only ±5 % for the SH macropore 320 parameters, the N dynamics, and quantity objectives (ND, NC) showed no sensitivity on the preferential transport. This could mean that the N transport is not affected dramatically by the preferential transport change, although earlier studies showed macropore influences on water and solute movement (Larsson and Jarvis, 1999;Nagy et al., 2019). However, the denitrification parameter, S t , showed one of the highest impacts on NC objective, which represents the N quantity in the drainage water. This high sensitivity might be related to the fact that the denitrification is limiting the amount of NO 3 -N, which would be transported 325 by hydrological means.

Model calibration
Through the calibration procedure, it was found that the ND and NC objectives were not responsive to the initially selected crop and hydraulic parameters. Since the Brtransport in the drainage simulated by the baseline model showed agreement to the accumulated measured transport, with underprediction of initial breakthrough after spraying of KBr (Fig. 4). The simulated 330 to reality. If all these conditions are probable, NO 3 -N can be only limited by gaseous loss; therefore, the denitrification model parameters had to be involved in the sensitivity and calibration process.
[ Table 5 is about here] As mentioned above, the sensitivity and the calibration process were done by minimising the mean nMAE performance mea-335 sures. For a broader evaluation, KGE and nRMSE [%] are also presented in Table 5. The calibrated objective results show that significant improvements were achieved in all solute transport accounts (BRD1, BRC1, BRD2, BRC2, ND, NC), without compromising any of the water balance objectives (DD, DC, S25, S60). By observing the N yield and NC objectives, one can see that N transport ND was improved without creating nitrogen stress in the crop. Brtransport improved during the tracer experiment period, as the model was able to provide a reasonable fit for the initial breakthrough (BRD2, and BRC2, Table 5, Fig.   340 6). Table 6 shows the calibrated parameters and their initial value. There was no substantial change within the SH parameters.
The SOM ratio increased, the SOM1 became 0.63, and SOM2 became 0.37 from the initial 0.5.
[ Table 6 is about here] The most significant changes appeared for parameters directly related to the denitrification reduction function (S t,f ast , w f ast , S t,slow , w slow ; Table 6), which may indicate that the default reduction function in DAISY overestimated this type of N loss. The crop uptake reflection factor increased only slightly from 0 to 0.34 %. The FB dry matter yield decreased from 14.5 Mg DM ha −1 to 13.7 Mg DM ha −1 . However, the Bruptake in fodder beet increased from 11.6 kg Br ha −1 to 12.1 kg Br ha −1 , which resulted in a 2% (0.5 kg Br ha −1 ) higher Bruptake of the initially sprayed 20.1 kg Br ha −1 . The 0.5 kg Br ha −1 proportionately removed from all Brleaching routes and macropore leaching remained at the same magnitude. Therefore, it can 355 be inferred that the matrix and macropore interchange did not change significantly (Fig. 7).

Nitrogen transport and harvest
Considering the calibrated Brtransport, the overall leached quantity of Brdid not change substantially. This contrasted with N transport, which responded differently to the calibration. The original model captured the N dynamics, and in one instance the magnitude, of the cumulated transport during the WW2003-2004 season. The application of N fertilisers was not reflected 360 in simultaneous or subsequent increases of N flux from the field, as well as there being no significant additive effect from crop type identified.
[ Figure 8 is about here] Figure 8 shows that the denitrification in the original model significantly outweighed the N loss by drains (measured NO 3 -N) with one order of magnitude with an average seasonal loss of 75 kg N ha −1 (18 -151 kg N ha −1 ), due to denitrification from 365 slow and fast pools combined. Hence, it seems that denitrification limited the amount of N transported to the drainage. Nagy et al. (2019) discovered that most of the water build up was above the plow pan in the A horizon. This could lead to this rapid denitrification, according to Eq.(4) to Eq. (7) where one of the modifying components is the f S water factor. In DAISY, f S had been set as default in order to increase linearly from 0 to 1 as a function of S from 0.7 to 1.0 (Hansen, 2002) (Fig. 9). This linearity does not fit the real condition for denitrification, since the stagnation of the water above the plow pan, as it presented Eq.(7) have been changed, which resulted in a steeper reduction of the denitrification (Fig. 9). By allowing the separation of the reduction factor for both the fast and the slow pool, the calibration of the model showed that the denitrification from the fast pool was shrunk to the range approx. from 0.9 to 1.0 relative saturation, while the slow pool remained like the default version of DAISY.

375
[ Figure 9 is about here] The modified f S showed a high impact on the N leaching, as the reduced denitrification enabled the model to depict the amount of N transported to the drain. The main increase appeared in the directly connected macropore flow (DM1 flow, cf. Fig. 4.), which mainly drained water from the A horizon. Since the mineralisation process is faster from the fast pool, the available mineral N was not readily reduced to gaseous N, but instead transported through preferential pores to the drain. The average 380 seasonal denitrification was reduced to 34 kg N ha −1 (9 -62 kg N ha −1 ), a reduction of more than 55% percent , which instead was made available for crop uptake, flow to the drainage system or the groundwater. The harvested DM and N did not show significant differences before and after calibration (Fig. 10). Some crop N and DM yield were closer to the measured values after calibration, but although the surplus of N was increased by the lowered denitrification, this additional N was not taken up by the crops. Overall, the objectives of DM yield and N yield were improved with the modified crop parameters (cf. Table   385 6 and Fig. 10). However, none of the models showed a satisfactory match for N yield. For both the original and the calibrated model, the comparison between measured and simulated N yield showed differences of more than 10 kg N ha −1 and even 20 kg N ha −1 for maize in 2002 and winter wheat in 2007. As earlier stated, a calibration of the N uptake parameters in the crop models governing the N uptake during the season was not possible, as the data available did not contain samples of biomass and hence N uptake during the season. aligned or missing , so the measured values could not be reliable. Besides data from this calibration independent evaluation period, data series of N concentration sampled with suction cups at depths of 1 m and 2 m for the period from 2000 to 2010 were used to validate the N transport to both drainage and groundwater.

Nitrogen transport in drainage water and deep leaching to groundwater
The N flux was remarkably improved according to all performance measures for the period 2007-2008 (Fig. 11). For the 400 original modelling of the two cropping seasons, the average KGE was below zero, indicating that the observed mean N flux was better predicted than the simulated one. In contrast, the calibrated model gave satisfactory results with KGE of 0.56 and 0.5 for the objective ND and NC, respectively, which in the case of solute transport modelling is an excellent result (Hansson and Hokfelt, 1975;Singh et al., 2005). Although the comparison for the drainage season 2008-2009 was improved even after calibration, less N loss to the drain system was simulated. The difference in simulated and measured N uptake in FB could not 405 explain this difference.
[ Figure 11 is about here] The loss due to denitrification has changed similarly for the calibration period 2000-2007 from seasonal 75 to 35 kg N ha −1 , which is approximately a 50 % percent reduction in total. In season 2007-2008, after the WW2006-2007 crop, a satisfactory match between measured and simulated accumulated N losses to the drainage could exclusively be explained from the reduced 410 denitrification loss (Fig. 11). As mentioned above, soil N concentration in soil water sampled with suction cups was used for validation purposes. The soil water N concentration had similar behaviour before and after the calibration, except the dynamics were better matched after calibration (Fig. 12). During the WW2003-2004 crop drainage period, the calibrated model was able to depict the 25 mg N L −1 measured peak as well as the fluctuations from 2002 to 2004. However, both models underestimated the N concentration substantially in the period after WW followed by FB 2007-2008. Again, these differences could not be 415 explained by an overestimated high N uptake (cf. Fig. 10). Although the drainage input of N massively increased due to the change in denitrification, the simulated N concentration only improved nRMSE[%] 5%-point for the whole period, which is equal to 1 mg N L −1 . Visual and mathematical inspection of the N concentration show, of course, a decent improvement on the soil N dynamics, but with this pronounced change in denitrification, the expected rise could be expected to be higher, if the more available N was not transported further down to the lower soil matrix (Fig. 1a) as DM1 and DM2 macropores start either 420 from the surface or from the bottom of the plow layer (A horizon). However, both are conceptually described to transport water and nutrients to the drainage system, and water build up is not allowed in the model. Besides, three other MM type macropores are present in this conceptual system with the possibility to transport water and solute: from surface to the bottom of the plow pan (MM1), from the surface to below the drain level at 150 cm depth (MM2) and from the bottom of the plow layer to below the drain level at 150 cm depth (MM3).

425
[ Figure 12 is about here] The latter two have the capability to transport N from the surface/plow layer below the drain level, bypassing the entire B horizon. This limits the transport through the mentioned horizon by matrix flow. The conceptual description of deep macropores points out another possible rapid transport route for N from the plow layer towards the groundwater. By comparing the measured suction cups samples with the simulated concentrations at 2 m depth, a remarkable 30 %-point decrease can be observed 430 in the nRMSE[%] and 0.3 nMAE performance measure (Fig. 13). Rosenbom et al. (2009) concluded that deep fractures might result in deep leaching of agrochemicals and nutrient. However, these flow phenomena were outside the scope of the present study.
[ Figure 13 is about here]

435
This study aimed to evaluate, by using one-dimensional physically based root zone model DAISY (Hansen, 2002), the effect of preferential transport and denitrification on leaching of nitrate to drainage during a 10-years period, as measured for an agricultural clay till field included in the Danish Pesticide Leaching Assessment Programme (PLAP; Lindhardt et al. (2001); web address: http://pesticidvarsling.dk). The results reveal a dominant effect on the leaching of nitrate through this clay till field. A large amount of N (48% to 80% of the total N-loss to drainage) was preferentially transported via macropores to 440 drainage, regardless of the application method and concurrent occurrence of precipitation. The current standard denitrification water reduction factor, fs, needed modification with a reduction of approximately 50% in the denitrification of the field from a seasonal average of 75 kg N ha −1 to 35 kg N ha −1 . The crop model provided acceptable results, and further studies are needed to improve the simulation of N uptake in crops. Overall, this study delineates the importance of accounting for preferential transport and coherent denitrification in the assessment of the leaching risk of nitrate to the aquatic environment. pulse rate, catecholamines in blood and urine, plasma renin activity and urinary aldosterone under basal conditions and following exercise, Eur J Clin Pharmacol,9,[9][10][11][12][13][14][15][16][17][18][19]https Lindhardt, B., Abildtrup, C., Vosgerau, H., Olsen, P., Torp, S., Iversen, B. V., Jørgensen, J. O., Plauborg, F., Rasmussen, P., and Gravesen, P.: The danish pesticide leaching assessment programme: Site characterization and monitoring design, Report, Geological Survey of Denmark Default (Hansen, 2002) and old default (Hansen et al., 1990) (Hansen, 2002) and old default (Hansen et al., 1990) water reduction factor of denitrification of DAISY and the calibrated water reduction factor of denitrification for the slow and fast pools. is the residual water content, θ s is the saturated water content, α and n are the retention curve shape parameter, Depth is the depth of the denoted horizon, ψ init/term is the initiating -terminating pressure of the macropore flow, ψ barrier is the required pressure difference between the macropore and matrix, ρ is the density of the  (2019)). θr is the residual water content, θs is the saturated water content, α and n are the retention curve shape parameter, Ksat saturated hydraulic conductivity, l is the pore connectivity parameter in the Mualem equation.   (Nagy et al., 2019). Depth is the depth of the denoted horizon, ψ init/term is the initiating -terminating pressure of the macropore flow, ψ barrier is the required pressure difference between the macropore and matrix, ρ is the density of the denoted macropore type, and d is the diameter of the denoted macropore type.