Soil & Tillage Research 174 (2017) 58–69

Contents lists available at ScienceDirect

Soil & Tillage Research

journal homepage: www.elsevier.com/locate/still

Computational assessment of sediment balance and suspended sediment MARK

transport pathways in subsurface drained clayey soils
⁎
M. Turunena, , L. Warstaa, M. Paasonen-Kivekäsb, H. Koivusaloa
a
Aalto University School of Engineering, Department of Built Environment. P.O. Box 15200, FI-00076 Aalto, Finland
b
Sven Hallin Research Foundation sr. Simonkatu 12 A 11, FI-00100 Helsinki, Finland

A R T I C L E I N F O A B S T R A C T

Keywords: Efficient mitigation of environmental loads from agricultural fields to surface waters requires sufficient
Preferential flow knowledge of dominating erosion processes and sediment transport pathways. Computational models are often
Lateral transport applied in sediment load assessments but their structure inherently includes assumptions and uncertainties,
Erosion which can hinder their predictive and explanatory capabilities. In this study, a 3D dual-permeability model was
applied with field-scale data from a high-latitude site to investigate sediment balances and structural un-
certainties in sediment transport components. The two-year data encompassed hourly records of water flow and
sediment concentration composite samples from tillage layer runoff and drain discharge in two adjacent clayey
fields with different slopes (1% and 5%). Three model structures with different assumptions of sediment
transport pathways were built to test their performance against the data. The simulations demonstrated how
different model structures can reproduce the data with varying results on sediment balance. The varying results
revealed the importance of flow, erosion, and sediment transport observations from the fields to improve the
simulations. Structural uncertainty analysis revealed uncertainties which parameter sensitivity analysis could
not describe. Concentration data was shown to include more information about erosion and sediment transport
processes than solely the load data. The results suggest that a major part (48–69%) of the detached particles
remained in the field and that lateral subsurface transport contributed to load generation (10–21% of total loads)
especially in the steeper field. The results demonstrated that the majority (84–87%) of sediment loads occurred
via subsurface drain discharge and groundwater outflow with the slope gradient of 1–5%, which suggests that
load mitigation measures should also be directed to decrease loads via subsurface transport pathways. The
simulations demonstrated how transport processes were controlled not only by water flow but also by soil
structure.

1. Introduction
Detachment and mobilization of particulate soil material from land
areas lead to harmful consequences including loss of fertile topsoil in
arable areas and eutrophication in surface waters (e.g. Quinton et al.,
2010). Soil erosion rates are typically high in agricultural land areas
due to tillage operations and periodic lack of vegetative cover (García-
Ruiz et al., 2015) and therefore, the arable land areas are of particular
interest from the point of view of sediment load mitigation. Factors
such as local climatic conditions, terrain topography and land-use are
known to impact erosion rates (e.g. Vanmaercke et al., 2011), but the
processes and factors leading to erosion and sediment transport are
wide-ranging and complex (Vereecken et al., 2016). Understanding of
these processes and particle transport pathways is, however, a pre-
requisite for controlling particle detachment and sediment loads.
Characteristic features of the northern land areas in the Baltic Sea
catchment area include wet spring and autumn periods, short growing
seasons, and cold winters with seasonal snow cover (e.g. Vagstad et al.,
2004). Clayey soils are abundant in the cultivated areas, where sub-
surface drainage methods are commonly applied to convey excess water
from agricultural fields to surface waters (Panagos et al., 2012; Eriksson
et al., 1999; Puustinen et al., 1994). In these regions, the main en-
vironmental consequence of soil erosion is considered to be the trans-
port of sorbed nutrients on sediment particles to surface water bodies
(Ulén et al., 2012). Due to the changing climatic conditions, winter
snow depths are going to decrease (Räisänen, 2008) and the amount of
erosion is expected to increase (Puustinen et al., 2007). Sediment loads
are regularly monitored in field-, plot- and catchment scales in the re-
gion (Bechmann 2012; Rankinen et al., 2010; Puustinen et al., 2007;
Øygarden et al., 1997). It is widely known that part of the sediment

⁎
Corresponding author.
E-mail addresses: mika.turunen@aalto.fi (M. Turunen), lassi@warsta.net (L. Warsta), maija.paasonen@hallin.fi (M. Paasonen-Kivekäs), harri.koivusalo@aalto.fi (H. Koivusalo).

http://dx.doi.org/10.1016/j.still.2017.06.002
Received 10 October 2016; Received in revised form 4 May 2017; Accepted 5 June 2017
Available online 21 June 2017
0167-1987/ © 2017 Elsevier B.V. All rights reserved.
M. Turunen et al.
Fig. 1. (a) Location of the Gårdskulla Gård experimental field and (b) layout of the two monitored field sections.
load can occur via surface runoff, but a distinct feature of sediment
transport in high-latitude clayey soils is the large share of total sedi-
ment loss via subsurface drains (Warsta et al., 2013a; Bechmann 2012;
Turtola et al., 2007; Uusitalo et al., 2001; Øygarden et al., 1997; Culley
et al., 1983). Empirical studies have suggested that suspended sediment
can be transported to subsurface drains in both vertical and lateral
macropores in soil (Ulén et al., 2014; Nielsen et al., 2010; Øygarden
et al., 1997). However, as pointed out by Allaire et al. (2009), lateral
transport is far less studied compared to vertical transport, and con-
tribution of these different flow routes to the distribution of eroded
particles in different load pathways remains unknown. Furthermore, as
the amount of particle detachment is rarely monitored on site, it re-
mains unclear how sediment balances are formed and how much se-
diment is detached from the soil, deposited back on the field surface or
retained in the soil profile.
Computational erosion and sediment transport models are com-
monly applied to assess sediment processes and their controlling factors
(Rankinen et al., 2010; Tattari et al., 2001; Jarvis et al., 1999; Wicks
and Bathurst, 1996), and to produce predictions of sediment losses and
loads (Nearing et al., 2005; Bhuyan et al., 2002; Zhang et al., 1996).
Empirical models such as the Universal Soil Loss Equation can be useful
tools for assessing long-term patterns and trends, but spatially and
temporally distributed models are required when erosion and sediment
transport processes within a field or catchment are of interest (Bryan,
2000). Conceptual models have been reported to yield satisfactory re-
sults in terms of spatially lumped sediment load reproduction in clayey
areas across different scales (Rankinen et al., 2010; Lundekvam, 2007),
but process-based models have a potential to describe erosion processes
in more detail. Field-scale process understanding can be beneficial
when designing and evaluating erosion mitigation methods, because for
example catchment outlet data contains only aggregated information of
the processes occurring within the catchment (Wellen et al., 2014) and
plot-scale is often considered to provide biased information of the
erosion rates (García-Ruiz et al., 2015). However, only few model codes
(Warsta 2011; Jarvis et al., 1999) are suitable for describing spatially
variable sediment transport on field surface and in the soil profile, and
performance of such models has been rarely comprehensively evaluated
against long-term concentration and load data representing different
59
Soil & Tillage Research 174 (2017) 58–69
pathways.
Since the processes and factors which influence detachment and
transport of sediment in clayey soils are not well known, the distinctive
features of computational models are their assumptions related to
erosion and sediment transport processes. Furthermore, it is well es-
tablished that different models and parameterizations can lead to a si-
milar fit of calibrated variables against observations but to different
results of variables outside the calibration setting (Refsgaard et al.,
2006; Højberg and Refsgaard, 2005). Assumptions due to the in-
complete knowledge of the studied processes (e.g. choice of hydro-
geological model structure and processes included in the model) are
often called model structural uncertainties (e.g. Højberg and Refsgaard,
2005). The structural uncertainties have been rarely evaluated in ero-
sion modelling studies, although such an evaluation would be poten-
tially beneficial to reduce uncertainty of erosion and sediment transport
models.
The objectives of this study were to (1) produce computational se-
diment balances of clayey subsurface drained agricultural fields using
different model structures and (2) to evaluate impacts of the structures
and parameterizations on sediment transport pathways and model
performance against sediment load data. Three different model struc-
tures were generated with the 3D process-based FLUSH model. The
main difference between the structures was the description of subsur-
face sediment transport pathways in vertical and lateral directions. The
reference data included discharge and suspended sediment concentra-
tion observations, and data-based estimates of sediment loads from
tillage layer runoff and subsurface drain discharge (Äijö et al., 2014).
The two-year data period represented winters with intermittent or low
snow cover and generation of runoff and sediment loads through all
seasons.
2. Materials and methods
2.1. The experimental site and data
The Gårdskulla Gård experimental site is located in southern
Finland (60° 10′ 32′′ N, 24° 10′ 17′′ E) and consists of two monitored
and subsurface drained field sections (1 and 2 in Fig. 1). The study
M. Turunen et al.
period covered two years (2008–2009) and the amount of precipitation
was 838 mm in 2008 and 565 mm in 2009. The maximum, mean and
median precipitation intensities were 11.1 mm h−1, 0.7 mm h−1 and
0.4 mm h−1, respectively. During the winter period (Jan–Mar) 2008 the
amount of precipitation was 197 mm, but there was no permanent snow
cover on the ground due to the relatively high mean daily air tem-
perature (-0.1 °C) which constantly fluctuated above the freezing point.
In Jan–Mar 2009, the mean daily air temperature was lower (−3.3 °C),
but the precipitation was only 64 mm and the air temperature occa-
sionally exceeded the freezing point, which resulted in a thin snow
cover with the maximum SWE of 35 mm (Turunen et al., 2015a). For
reference, in the period 2008–2014 mean daily air temperature in
Jan–Mar was −4.0 °C. The maximum observed soil frost depth was
0.2 m in 2008 and 0.56 m in 2009.
The field sections have clayey soils and Sections 1 (5.7 ha) and 2
(4.7 ha) have average slopes of 1% and 5%, respectively. The subsur-
face tile drains were installed in the 1940s with the spacing of 16 m and
average depth of 1 m in both field sections. The soils are classified as
Vertic Luvic Stagnosols (FAO, 2007) and the average clay (particle
size < 0.002 mm) content of the topsoil (depth 0.0–0.2 m) in Section 1
(49%) was slightly higher than in Section 2 (41%). The particle size
distributions and organic carbon contents were determined with sieve
analysis and dry combustion, respectively, from three parallel samples
at four different depths in both of the studied sections. Organic carbon
content of the topsoil was 3% and 2% of dry weight in Sections 1 and 2,
respectively. Standard deviation of the particle size fractions varied
mostly between 0 and 4%, and was clearly the highest (13%) in the clay
content of the depth 0.8–1.0 m in Section 2 (Table 1). According to a
Mann-Whitney U test, the observed particle size fractions of soil layers
0–0.4 m and 0.4–1.0 m differed statistically (0.05 significance level) by
depth in most of the cases, with the exception of fine sand and silt
contents of Section 1 (p values 0.94–1.00), when the fractions were
compared separately in Sections 1 and 2. The difference between the
particle size fractions of Sections 1 and 2 (soil layers 0–0.4 m and
0.4–1.0 m) was significant only between the observed fine sand and
clay contents of the soil layers 0–0.4 m.
Measurements of soil hydraulic properties were conducted from
four parallel undisturbed soil samples (height 0.048 m and diameter
0.072 m) in two locations in both field sections. The measurements
included water retention curves (pressure head values 0 m, −0.1 m,
−0.3 m −1.0 and −10.0 m as a drying curve) and macroporosities at
the depths of 0.10-0.15 m, 0.30-0.35 m and 0.50–0.55 m. Macropores
were defined as pores which empty with the suction pressure of 0.1 m
(Jarvis, 2007). The average measured macroporosities were 1.26%,
0.74% and 1.33% in the soil layers of 0.10–0.15 m, 0.30–0.35 m and
0.50–0.55 m, respectively, in Section 1. In Section 2, the macro-
porosities were 2.53%, 0.48% and 0.39% in the soil layers of
0.10–0.15 m, 0.30–0.35 m and 0.50–0.55 m, respectively. The data, the
measurement methods, and the site details are presented in more detail
in Turunen et al. (2015a) and Äijö et al., 2014.
Table 1
Average content [%] of organic carbon, clay (≤0.002 mm), silt (0.002–0.02 mm), fine sand (0.02–0.2 mm), coarse sand (0.2–2 mm) and gravel (2–20 mm) in Sections 1 and 2 of the
Gårdskulla Gård experimental site. Standard deviation of the measurements in parentheses. (modified from Turunen et al., 2015a).
Depth [m] Organic carbon Clay
Section 1 0−0.2 3 (0) 49 (2)
0.2−0.4 2 (1) 51 (1)
0.4−0.8 1 (0) 54 (2)
Depth [m] Organic carbon Clay
Section 2 0–0.2 2 (0) 41 (2)
0.2–0.4 1 (1) 43 (4)
0.4–0.8 1 (0) 56 (1)
0.8–1.0 0 (0) 69 (13)
60
Soil & Tillage Research 174 (2017) 58–69
Hydrological variables and total suspended solid (TSS) concentra-
tions were intensively observed in the fields from 1 Jan 2008 to 31 Dec
2009. Discharge from the subsurface drain networks and tillage layer
runoff (TLR) were monitored with the time interval of 15 min with
vertical helix meters (Datawater WS, Maddalena, Povoletto, Italy). TLR
was collected with shallow drains (depth 0.4 m), which had coarse
gravel as trench backfill (Fig. 1b). Embankments beside the shallow
drains directed surface runoff to the shallow drains (see Turunen et al.,
2015a Fig. 2 for details). Thus, the TLR measurements represented both
surface runoff as well as horizontal seepage of water in the uppermost
0.4 m deep soil layer. During the spring snowmelt periods of 2009, an
unquantified share of surface runoff bypassed the measurement devices
because of the frozen topsoil above the shallow drains which may have
led to a biased estimate of the amount of TLR. However, Turunen et al.
(2015a) suggested that even though soil frost can occasionally impact
distribution of water balance components, frost had only a small effect
on the water balance distribution in the Gårdskulla Gård site during the
spring and winter 2009 when the simulated SWE was relatively low
(max. 35 mm). In 2008 the discharge measurement devices functioned
satisfactorily.
TSS concentrations in drain discharge and TLR were measured from
composite water samples. From both outflow pathways in both field
sections, an automated system collected a small subsample of ap-
proximately 1.5 × 10−4 m3 per each 50 m3 (Section 1) or 20 m3
(Section 2) of the water flowing through the system, i.e., the sampling
strategy produced more samples during time periods of intense outflow.
The water volume triggering the sampling (50 m3 or 20 m3) was higher
in Section 1, since more water outflow occurred in Section 1 than in the
steeper Section 2. The higher measured water outflow in the subsurface
drain network of Section 1 occurred due to the flat topography and
influx of water from the surrounding areas (Turunen et al., 2015a). The
subsamples formed composite samples in containers, from which the
samples were manually collected weekly or bi-weekly and transported
to the laboratory. The composite samples were analysed for TSS by
weighing the evaporation residue according to the standard SFS 3008:
1990 (SFS, 1990). The number of analysed composite samples was 56
(803 subsamples) and 40 (1073 subsamples) from drain discharge in
Sections 1 and 2, respectively. The amount of analysed composite
samples from TLR was 17 in Section 1 (16 subsamples) and 22 (72
subsamples) in Section 2. Sediment loads for each sampling interval via
each pathway were estimated as the product of the measured TSS
concentration and the measured discharge.
Winter wheat (Triticum aestivum) and spring barley (Hordeum vul-
gare) were cultivated on the fields using mineral fertilizers during the
studied period. The annual crops were harvested in the middle of
August (Table 2). The fields were tilled after harvest and before seeding
during the studied autumn and spring periods. The crop rotations,
farming practices and tillage operations are shown in more detail in
Table 2.
Silt Fine sand Coarse sand Gravel
31 (1) 13 (1) 7 (3) 0 (0)
34 (2) 13 (3) 2 (1) 0 (0)
31 (2) 13 (2) 2 (1) 0 (0)
Silt Fine sand Coarse sand Gravel
36 (1) 20 (1) 4 (0) 0 (0)
34 (1) 21 (3) 3 (2) 0 (0)
30 (0) 13 (0) 1 (0) 0 (0)
21 (1) 9 (4) 0 (0) 0 (0)
M. Turunen et al. Soil & Tillage Research 174 (2017) 58–69

Fig. 2. Conceptual description of the main erosion and sediment transport processes of the FLUSH model.

Table 2 domains are derived with implicit finite volume based methods
Agricultural land-use and farming and tillage operations in field Sections 1 and 2 of the (Warsta, 2011).
Gårdskulla Gård experimental site during the period of 2008–2009. Autumn mouldboard
The simulated domain can include a user-specified number of dif-
ploughing and disc harrowing depths were about 0.25 and 0.05 m, respectively.
ferent subsurface drain and open ditch networks (e.g. Turunen et al.,
Year 2008 2009 2015b, 2013). The model treats the drainage systems as cell internal
sinks, which remove water, solutes and sediment from the simulated
Section 1 Crop Winter wheat (Triticum Spring barley
domain. Flow of water to the drainage systems is calculated with
aestivum) (Hordeum vulgare)
Tillage Autumn ploughing Spring harrowing and
Darcy’s law. Potential evapotranspiration (PET) and crop root depth are
autumn disc harrowing provided as a pre-computed time series to the model. During each time
Farming Harvest 12 Aug Sowing 28 Apr step, PET is distributed to the soil profile as a function of root depth,
operations Harvest 21 Aug and actual evapotranspiration is calculated by reducing PET in sub-
Sowing 13 Sep (winter
optimal soil moisture conditions with the function of Feddes et al.
wheat)
Section 2 Crop Winter wheat (Triticum Winter wheat (Triticum (1978). Snow and frost processes are calculated with the energy balance
aestivum) aestivum) based scheme of Koivusalo et al. (2001). Currently, soil frost does not
Tillage Autumn ploughing Autumn ploughing affect the water flow model as explained in more detail in Warsta et al.
Farming Harvest 14 Aug Harvest 19 Aug
(2012) and Turunen et al. (2015a).
operations Sowing 29 Aug
(winter wheat)
The erosion and sediment transport computation scheme is com-
posed of the surface and subsurface transport models combined with
different processes related to particle detachment and transport of
suspended solids (Warsta et al., 2013a; Warsta, 2011). A conceptual
2.2. FLUSH model representation of the main erosion and sediment transport processes is
presented in Fig. 2. In the model, particle detachment is described as
FLUSH is a dynamic and process-based hydrological model, which hydraulic sheet erosion and raindrop splash erosion (Warsta et al.,
describes overland water flow in two-dimensions (2D) and subsurface 2014, 2013a). All sediment detachment processes occur in the surface
flow in 3D (Warsta et al., 2013b; Warsta, 2011). The diffuse wave ap- domain of the model, i.e., it is assumed that subsurface erosion is
proximation of the Saint-Venant equations is applied to simulate negligible. The current approach describes sediment transport solely
overland flow and overland transport is described with an advection with one particle size class. In the surface domain, the detached par-
model (Warsta, 2011). FLUSH implements a dual-permeability con- ticles can be advectively transported with overland flow and be re-
ceptual structure where the subsurface pore volume is divided into two moved from the domain by the open ditches, be deposited on the field
mobile pore systems (matrix-macropore or aggregate-inter aggregate surface or be infiltrated into the subsurface soil domain. In the sub-
systems). Water movement in both matrix and macropore systems is surface domain, the sediment can be transported with soil water flow
described with the Richards equation. Volumetric fraction of static soil and be removed from the soil domain by subsurface drains and seepage
macropores is given as a model input and dynamic macroporosity to open ditches. Transport of suspended sediment in the subsurface
(change in macroporosity due to soil shrinkage and swelling) is calcu- domain is constrained into the macropore system as the pore pathways
lated following van Dam et al. (2008). Saturated hydraulic conductivity in the soil matrix are considered too small for movement of sediment
of the macropore system is calculated as the product of macroporosity particles. Sediment load occurs in the model when detached particles
and macropore saturated hydraulic conductivity multiplier (Warsta are transported to subsurface drains or open ditches by overland flow or
et al., 2013b; Jarvis, 2008). Water exchange between the pore systems subsurface water flow.
is described as a function of pore water pressure and soil characteristics, Raindrop splash erosion re [kg m−2 h−1] is computed as a function
following Gerke and van Genuchten (1993). Water retention properties of rainfall intensity:
and unsaturated hydraulic conductivities are calculated with the
Mualem-van Genuchten schemes (van Genuchten, 1980). Subsurface re = kr Fw (1 − CC ) Mr (1)
transport in matrix and macropore systems is simulated with the ad-
vection-dispersion equation combined with a description of solute ex- where kr [h2 kg−1 m−2] is the raindrop splash soil erodibility coeffi-
change between the pore systems (Salo et al., 2015). Numerical solu- cient, Fw [–] is the overland water depth correction factor, Cc [–] is the
tions of flow and transport equations in both surface and subsurface proportion of soil covered by canopy cover and Mr [kg2 h−3] is the

61
M. Turunen et al.
momentum squared for rainfall. The approach is a simplification of the
raindrop splash erosion scheme in the SHE model (Wicks and Bathurst,
1996). Fw was calculated as follows:
Fw = ( Dr)
⎧ exp 1 − hw , h w > Dr
⎨ 1, h w ≤ Dr
⎩ (2)
where hw [m] is the overland water depth and Dr [m] is the median
raindrop diameter, which was calculated with Eq. (3):
Dr = ki I 0.182 (3)
−1
where ki [–] is an empirical constant, I [mm h ] is the rainfall in-
tensity. Mr in Eq. (1) is computed with the empirical approach of Wicks
and Bathurst (1996). Hydraulic erosion occurs during overland flow
events when shear stress of the overland flow exceeds the critical shear
stress value. Shear stress of overland flow is computed as a function of
field surface slope and water depth in the cell where overland flow
occurs (e.g. Warsta et al., 2013a; Taskinen and Bruen, 2007). Critical
shear stress value is calculated with the modified Shields method for
small particle sizes (Yalin, 1977). Soil erodibility coefficients can be
given as time series for both rain drop splash and hydraulic erosion.
The transport capacity of overland flow is computed with the
equation of Yalin (1963). When concentration of overland flow exceeds
the transport capacity, the excess sediment is deposited on the field
surface and removed from the simulated domain. Sediment deposition
on the field surface due to the settling of the particles is described with
Stoke’s law. Departing from the earlier model version of Warsta et al.
(2013a), in the current study a computational process was added to the
model to avoid long-term accumulation of sediment mass to the drain
depth in the simulations. The process removes sediment from the sub-
surface domain during those time periods when TSS concentration ex-
ceeds a threshold value, i.e. it was presumed that sediment does not
accumulate in the soil water domain due to the restriction set by the soil
pore sizes. This process describes the impacts of subsurface sediment
retention on the loads and concentrations.
3. Modelling strategy and setup
The main modelling strategy was to assess erosion and sediment
transport processes by forming three different model setups, hereby
called model structures, and to calibrate and validate each of the
structures against sediment load and concentration measurements. The
aim of the modelling strategy was to produce a computational estimate
of the possible range of the sediment load component quantities. The
three model structures (Models A–C) applied different assumptions of
the ability of water flow to transport suspended sediments in the pre-
ferential flowpaths of the soil domain. In model A, the sediment was
allowed to be transported in 3D in all soil layers of the subsurface
macropore domain. In model B, sediment was allowed to be transported
in 3D but only above the subsurface drain depth (1 m). In model C,
sediment was allowed to be transported in the subsurface domain only
vertically and only above the subsurface drain depth. Model C was
included in the studied setups to provide a reference for demonstrating
the role of lateral sediment transport. Conceptual presentation of the
three models are shown in Fig. 3. The water flow model was similar in
models A–C and allowed flow of water in 3D in all soil layers of the
subsurface domain.
The parameterization of the water flow model was adopted from
Turunen et al. (2015a) and Koivusalo et al. (2017), who simulated the
water balance of the Gårdskulla Gård fields throughout the years
2008–2014. The boundary conditions and soil properties were set si-
milar as in Koivusalo et al. (2017), i.e. in Section 1 the hilly forested
area (4.6 ha) upslope of the drained area (see Fig. 1b) was included in
the simulated area, and the extended area was parameterized similarly
as Section 1 (Koivusalo et al., 2017; Turunen et al., 2015a). The other
boundary conditions were no-flow bottom boundary condition at the
62
Soil & Tillage Research 174 (2017) 58–69
depth of 2.4 m, atmospheric boundary condition on the field surface,
no-flow boundaries at the horizontal field borders and fixed-head
boundary condition (based on observed stream water elevations) at the
Kirkkojoki stream (see Turunen et al., 2015a Fig. 2b for details). The
soil hydraulic properties were assumed to be anisotropic above the
subsurface drain depth, based on the study of Berisso et al. (2013). The
ratio of horizontal to vertical macroporosity was set to 0.02, 0.83 and
1.0 in the soil layers of 0–0.425, 0.425–1.05 and 1.05–2.4 m, respec-
tively (Turunen et al., 2015a). While the field sections had relatively
similar soil properties, there were differences in the observed water
retention curves and macroporosities above the drain depth, as shown
in Tables 3–4 of Turunen et al. (2015a). The lateral size of the com-
putational grid cells was set to 4 × 4 m2. The time step of the input data
was one hour and the computational time step was set to 56 s when
overland flow occurred on the field surface and 225 s during the other
time periods.
The erosion and sediment transport models were calibrated and
validated against the sediment load and concentration data on both
drain discharge and TLR. The models were calibrated against the data
from Jan–Dec 2008 and validated against the data from Jan–Dec 2009.
The simulated TSS concentrations were computed from the hourly si-
mulation results in the same way as the composite samples were col-
lected. The hourly simulated concentrations were aggregated to com-
posite concentrations for the same time periods as the empirical data,
and similarly to the observations the mean values were calculated based
on the hourly concentrations during those events when 50 m3 (Section
1) or 20 m3 (Section 2) of simulated water flow had discharged via the
subdrains or TLR.
Previous sensitivity analysis (Warsta et al., 2013a) suggested that
the key parameters of the applied erosion and sediment transport model
are those that affect erosion rates. In the current study, the intensity of
TLR was small, only a portion of it was overland flow and the events
occurred sporadically in the studied field sections (see Fig. 6c,d of
Turunen et al., 2015a). Therefore hydraulic erosion could only have
had an intermittent impact on sediment loads and it was hypothesized
that raindrop splash erosion was the primary cause of particle detach-
ment. Thus, the model calibration was conducted by manual trial-and-
error adjustment of the values of the empirical coefficients kr (Eq. (1))
and ki (Eq. (3)), which control raindrop splash erosion in the model. For
calibration, the value of kr was varied in the range of 1.2–452.6
s2 kg−1 m−2 and ki in the range of 1.2 × 10−3–2.5 × 10−3. The cali-
bration range was determined by adjusting the calibration parameters
to achieve the closest correspondence between the calibration and ob-
servation values. Since Sections 1 and 2 had relatively similar soil
properties, a similar parameterization was applied for both field sec-
tions. During the time periods with a vegetation cover on soil surface
(the time period between sowing and harvest for spring crops and 1
May to harvest for autumn crops), Cc (Eq. (1)) was set to 1 and during
the other periods Cc was set to 0. Overland flow soil erodibility coef-
ficient and critical shear stress value was 10−7 kg m−2 s−1 and
0.174 kg s−2 m−1, respectively (Warsta et al., 2013a).
The concentration threshold controlling the removal of accumulated
sediment mass in the subsurface macropore domain was set to the
maximum measured concentration value of 1 489 g m−3 in the sub-
surface drain discharge. The threshold value is within the range of the
maximum concentrations (1300–1800 g m−3) measured by Turtola
et al. (2007) and Warsta et al. (2014) in tile drained clayey agricultural
fields in southern Finland.
Other parameter values for the erosion and sediment transport
model were adopted from Warsta et al. (2013a) and all parameter va-
lues, except Cc, were set spatially uniform and temporally constant for
the whole simulated period. The initial TSS concentrations in the
overland flow domain and in the subsurface domain were set to zero.
Since the TSS concentrations in discharge waters from boreal forest
areas are typically low (e.g. average ≤2.6 g m−3 in Mattsson et al.,
2003), the erodibility of the hilly forested area adjacent to Section 1
M. Turunen et al.
(Fig. 1) was assumed to be negligible and the soil erodibility coeffi-
cients of the raindrop splash and hydraulic erosion in the area were set
to zero.
The model calibration resulted in parameter values of 2.15 × 10−3,
1.99 × 10−3 and 2.38 × 10−3 for ki in Model A, B and C, respectively.
Regarding kr, the calibration resulted in parameter values of
9.9 s2 kg−1 m−2 in Model A, 11.1 s2 kg−1 m−2 in Model B and 452.6
s2 kg−1 m−2 in Model C.
The goodness-of-fit of the model simulations was evaluated in terms
of the difference between the simulated and measured annual accu-
mulations (D) and the modified Nash–Sutcliffe coefficient (N–S), where
the squared values of the original Nash–Sutcliffe function were replaced
with absolute values, following the suggestions of Legates and McCabe
(1999).
Sensitivity of the model results to changes in the calibrated para-
meter values was addressed with a sensitivity analysis, where both of
the calibrated parameter values were increased and decreased by 20%
(Section 4.2). To address the sensitivity of the simulated sediment loads
and water outflow to changes in soil hydraulic parameters, an addi-
tional simulation scenario was conducted where Section 1 was simu-
lated with the soil parameters of Section 2 and vice versa (Section 4.2).
4. Results
4.1. Calibration and validation
The simulation results of Models A and B showed comparable fit
against the measurements in terms of the model evaluation criteria, and
the models were able to reproduce the distribution of the sediment load
to the drain discharge and TLR components (Fig. 4a–d, Table 3). In the
calibration period, the D value of cumulative load via drain discharge
was 21 and 40 kg ha−1 a−1 with Models A and B, respectively, in the
flat Section 1 (measured load 1418 kg ha−1 a−1). In the steeper Section
2, the D values of drain discharge load during calibration (A: −61 and
B: −181 kg ha−1 a−1) showed a higher difference between the simu-
lated and measured accumulations, though also the measured load was
higher (2416 kg ha−1 a−1). In both field sections, the D values were
only from −7 to 3% of the measured loads during calibration. The D
values of TLR load during calibration were comparable to those of drain
discharge load (Table 3).
Model C was not able to reproduce the magnitude of the observed
loads in either of the field sections (Fig. 4e–f, Table 3). In the calibra-
tion period, Model C underestimated the simulated load via drain dis-
charge (D −923 kg ha−1 a−1 or −38% of the measured load) in the
steep Section 2 (slope 5%) and overestimated load via TLR (D
238 kg ha−1 a−1 or 441% of the measured load) in Section 1 (slope
1%). The differences were high compared to the D values of Models A
and B during calibration (Table 3, Fig. 4), which pointed out the need to
63
Soil & Tillage Research 174 (2017) 58–69
Fig. 3. Conceptual representation of the three dif-
ferent applied model structures (Models A–C) of the
erosion and sediment transport model of FLUSH.
account for lateral sediment transport in the model structure. However,
Model C was able to reproduce the observed load via drain discharge
(D: −6 kg ha−1 a−1, N–S: 0.29) in Section 1 (Fig. 4e), which suggests
that regarding the load via drain discharge, the model structure C
functioned better in the flat field (slope 1%) than in the steeper Section
2 (slope 5%). This implies that lateral fluxes had a more pronounced
role in sediment transport in the steep field section compared to the flat
section due to the steeper hydraulic gradient and higher amount of
groundwater flow.
Regarding Models A and B, the D values of the sediment loads via
drain discharge during validation were comparable to those of cali-
bration (Table 3). In terms of the D values of the loads via drain dis-
charge, Model A performed more accurately during calibration, and
Model B during validation (Table 3). In both field sections and models,
the D values of loads via TLR during the validation period were com-
parable to those of drain discharge (Table 3). Majority of the sediment
load occurred via subsurface drains in both field sections and most of
the load occurred outside the growing season during the studied period
(Fig. 4a–d). The amount of precipitation was higher during calibration
(838 mm) than validation (565 mm), while the mean precipitation in-
tensities were similar (0.7 mm h−1 during both years), which explained
the lower amount of load in 2009 than 2008.
The positive N-S values (0.13–0.21) of the drain discharge load
dynamics in Models A–B during calibration (Table 3) suggested a cor-
respondence between the simulated and observed temporal variations
in load generation during the wet year of 2008. According to the drain
discharge N-S values in Section 2, during the validation year 2009 when
only minor sediment loads occurred and only during the autumn
period, the models A and B reproduced the observed load dynamics via
drain discharge more accurately (N-S values 0.63–0.68) compared to
the calibration period (N-S values 0.13–0.18). In Section 1, the simu-
lations with Models A and B resulted in negative N-S values for drain
discharge load dynamics during validation (Table 3). However, the
mean absolute error between the simulations and observations in Sec-
tion 1 was lower during validation (18–19 kg ha−1) than during cali-
bration (26 kg ha−1), which demonstrates that the negative N–S values
during validation were induced by the low deviation of the simulated
and measured loads. As demonstrated also by the negative N-S values of
TLR load dynamics (Table 3), the performance of the Models A–C was
generally better during high loads than low loads. Regarding Model C,
the N-S values were mostly in the same order of magnitude as for
Models A and B during calibration and validation (Table 3).
N-S values for TSS concentrations in drain discharge and TLR were
all negative, which was caused by differences between the simulated
and measured concentration dynamics and by the tendency of the
model to overestimate the highest concentrations and underestimate
the lowest concentrations. However, when the simulated concentra-
tions during the time periods which generated only a minor part of the
M. Turunen et al. Soil & Tillage Research 174 (2017) 58–69

Fig. 4. Measured and simulated accumulated annual sediment loads via drain discharge and tillage layer runoff during the calibration and validation periods of (a) Model A in Section 1,
(b) Model A in Section 2, (c) Model B in Section 1, (d) Model B in Section 2, (e) Model C in Section 1 and (f) Model C in Section 2 of the Gårdskulla Gård experimental site.

Table 3
Results of the model evaluation criteria during model calibration and validation period in Sections 1 (slope 1%) and 2 (slope 5%) with the Models A, B and C.

Accumulated load Accumulated load Load

Model D [kg ha−1 a−1] M [kg ha−1 a−1] N–Sm [–]

Calibration Drain discharge Section 1 A 21 1418 0.21

B 40 1418 0.21
C −6 1418 0.29
Section 2 A −61 2416 0.13
B −181 2416 0.18
C −923 2416 0.30
Tillage layer runoff Section 1 A −6 54 −0.26
B −5 54 −0.27
C 238 54 −6.51
Section 2 A −80 232 −0.42
B −82 232 −0.40
C −168 232 −0.47

Accumulated load Accumulated load Load

Model D [kg ha−1 a−1] M [kg ha−1 a−1] N–Sm [–]

Validation Drain discharge Section 1 A −25 382 −0.30

B −13 382 −0.34
C −119 382 −0.15
Section 2 A 100 469 0.63
B 76 469 0.68
C −153 469 0.79
Tillage layer runoff Section 1 A −28 29 −0.26
B −28 29 −0.26
C −27 29 −0.24
Section 2 A −48 74 0.15
B −49 74 0.14
C −51 74 0.12

D = difference between the simulated and measured annual accumulations.

M = measured annual accumulation.
N–S = the modified Nash-Sutcliffe coefficient.

accumulated loads (concentrations < 20 g m−3 and < 1% of the load) drain discharge were comparable during the studied period
were neglected, the mean measured (228 g m−3 in Section 1 and (2008–2009). This indicated a correspondence between the simulated
480 g m−3 in Section 2) and simulated (206–207 g m−3 in Section 1 and measured concentration levels during the time periods generating
and 506–521 g m−3 in Section 2 with Models A–B) concentrations in the majority of loads. Similarly, the mean measured (456 g m−3 in

64
M. Turunen et al.
Section 1 and 851 g m−3 in Section 2) and simulated (338–341 g m−3
in Section 1 and 802–810 g m−3 in Section 2 with models A–B) con-
centrations in TLR were in the same order of magnitude, even though
the simulations underestimated the observed values. Both the simulated
and observed TSS concentrations were higher in the steep Section 2
(slope 5%) than in the flat Section 1 (slope 1%). Furthermore, regarding
water outflow, the simulated and measured drain discharge (D
6–37 mm or 1–5% of measured discharge) and TLR (D −23 to –6 mm
or −49 to –26% of measured TLR) values were comparable during the
studied period. Due to the correspondence between the simulated and
measured values of both the concentration levels and the water outflow
components, the model was capable of reproducing the measured cu-
mulative loads. However, the negative N–S values of the TSS con-
centrations pointed out that the temporal variation in erosion and se-
diment transport processes could be further enhanced in the model.
Furthermore, despite the mismatch between the simulated and mea-
sured concentration dynamics, both the simulated and measured mean
concentrations in TLR were higher than the mean concentrations in
drain discharge. In this sense the measured and simulated concentra-
tion levels were qualitatively similar.
4.2. Sediment balances and sensitivity analyses
The simulated sediment load components with Models A–C are
shown in Fig. 5a–b and Table 4. Load via subsurface drains dominated
the loads in both field sections and with all model structures (Fig. 5).
However, results from Models A and B suggested that part of the load
65
Soil & Tillage Research 174 (2017) 58–69
Fig. 5. The simulated mean annual sediment load
components computed with Models A–C in (a)
Section 1 and (b) Section 2 of the Gårdskulla Gård
experimental site. The error bars denote the
minimum and maximum loads from the parametric
sensitivity analysis, where parameters values were
increased and decreased by 20%.
occurred via seepage to open ditches (8–9% in Section 1 and 8% in
Section 2) and groundwater outflow (10–15% in Section 1 and 16–21%
in Section 2). The groundwater outflow component occurs below the
other drainage systems (open ditches and subsurface drains) and re-
presents the flux to the adjacent stream. As shown in Fig. 6a–b, the
amount of groundwater outflow was higher in Section 2 (slope 5%)
than Section 1 (slope 1%), which induced the higher sediment loads via
groundwater outflow in the steeper field section. Evapotranspiration
was the largest water balance component in both field sections, and the
share of evapotranspiration from precipitation was higher in Section 1
(58%) than Section 2 (50%). As shown in Table 4, the highest difference
between the load components of the field sections was seen in drain
discharge (770–773 kg ha−1 a−1 in the flat Section 1 and
1201–1288 kg ha−1 a−1 in the steep Section 2) and groundwater out-
flow (107–161 kg ha−1 a−1 Section 1 and 274–433 kg ha−1 a−1 Sec-
tion 2). Most of the eroded sediment at the surface infiltrated to the
subsurface domain in the simulations (Table 4) and both seepage to
open ditches and groundwater outflow formed a higher load component
than the load via surface runoff in Section 1 and 2. Furthermore, ma-
jority ( > 80%) of the load via overland flow was collected by the
Kirkkojoki stream. Results from Model C are shown for reference in
Fig. 5. Note that the small amount of groundwater outflow also oc-
curred in the simulations with Model C (Table 4), which was explained
by vertical sediment transport in those computational grid columns
which contained the groundwater outflow sink terms representing the
Kirkkojoki stream.
With Models A and B, the amount of eroded sediment
Fig. 6. Simulated mean annual water out-
flow components (Model A) in (a) Section 1
and (b) Section 2 and simulated sediment
load components in (c) Section 1 and (d)
Section 2 with the measured soil hydraulic
properties swapped between the field sec-
tions in the Gårdskulla Gård experimental
site.
M. Turunen et al. Soil & Tillage Research 174 (2017) 58–69

Table 4
The simulated mean annual sediment balance components [kg ha−1 a−1] in Sections 1 and 2 with Models A–C in the Gårdskulla Gård experimental site.

Model A B C

Section 1 (slope 1%) Erosion 3104 3317 11312

Deposition (field surface) 560 561 1168
Infiltration to the soil 2486 2698 9957
Retention 1467 1729 9164
Overland flow 58 58 187
Seepage to open ditches 88 90 52
Subsurface drains 770 773 737
Groundwater outflow 161 107 5

Model A B C

Section 2 (slope 5%) Erosion 3870 4171 12527

Deposition (field surface) 410 433 695
Infiltration to the soil 3327 3604 11595
Retention 1439 1983 10715
Overland flow 134 134 238
Seepage to open ditches 167 147 74
Subsurface drains 1288 1201 785
Groundwater outflow 433 274 20

(3104–3317 kg ha−1 a−1 in Section 1 and 3870–4171 kg ha−1 a−1 in
Section 2) was higher than the sum of the different load components
(1028–1077 kg ha−1 a−1 in Section 1 and 1756–2022 kg ha−1 a−1 in
Section 2) (Table 4). This suggests that about 31–52% of the detached
particles were transported out of the studied fields. Deposition on the
field surface removed part of the sediment (10–18% of erosion) in the
simulations (Table 4), but sediment retention in the soil profile formed
the largest sink term and it removed 37–52% of the eroded sediment in
the simulations. A higher share of eroded sediment was deposited on
the field surface and retained in the soil in Section 1 (slope 1%) than in
Section 2 (slope 5%) (Table 4). Thus, a higher share of eroded material
formed load in the steep field section (42–52%) than in the flat section
(31–35%). With Model C, the amount of eroded sediment was about
3–4 times higher than with Models A–B, which resulted in even higher
amounts of sediment retention in the soil profile (Table 4). As the
horizontal sediment fluxes were excluded in Model C, eroded sediment
infiltrating between the drain lines was not transported to subdrains or
other sinks in the soil profile in the simulations, and thus only 9% of the
eroded material formed load. However, the simulated loads in drain
discharge in Section 2 and TLR in Section 1 with Model C were not
consistent with the data (Table 3) and thus also the sediment balance
simulation results of Model C may not be physically plausible.
More erosion occurred in the steep Section 2 (slope 5%) than in the
flat Section 1 (slope 1%) (Table 4). The difference in the amount of
erosion between the field sections (Table 4) was explained by the dif-
ferences in soil hydraulic properties between the topsoil layer of Section
2 and 1. Differences in macroporosities and soil water retention curves
promoted more rapid water infiltration to the subsurface domain in
Section 2 than 1, and thus in Section 1 the overland water had a higher
protective impact against the erosive force of the raindrops (see Eq.
(2)).
The results of the parameter sensitivity analysis are shown in Fig. 5
as minimum and maximum simulated values for each load component.
The results demonstrate that the relatively high changes (+20% and
−20%) in the parameter values had a lower impact on the groundwater
outflow component than the model structure selection (Model A vs
Model B). It is also noteworthy that the load via subsurface drains was
the most sensitive load component to the parameter deviations. Model
C was the least sensitive to the parameter deviations.
The simulation scenario where Section 1 was simulated with the soil
parameters of Section 2 and vice versa (with Model A) demonstrated
how the water flow model was less sensitive to the soil parameteriza-
tion (Fig. 6a–b) than the erosion and sediment transport model
(Fig. 6c–d). In the scenario, the share of sediment retention of the
eroded sediment was higher in Section 2 (46%) than Section 1 (40%),
which revealed that the retention process was controlled more by the
soil hydraulic properties than the terrain slope. Similarly, the share of
deposition on the field surface of the eroded sediment was higher in
Section 2 (23%) than Section 1 (7%) in the scenario. Additional tests
(not shown) pointed out that the model was the most sensitive to the
differences in the soil macroporosity between the field sections, which
revealed that even though water flow controlled sediment transport in
the field sections, also soil structure had a high impact on load gen-
eration. Regarding the load pathways, the highest sensitivity in terms of
absolute changes in results due to the deviations in soil hydraulic
properties was seen in load via drain discharge.
5. Discussion
As pointed out by the data and simulations of the current study as
well as by many previous studies (e.g. Warsta et al., 2013a; Bechmann,
2012; Turtola et al., 2007; Uusitalo et al., 2001; Øygarden et al., 1997),
subsurface transport processes can have a major role in the sediment
loads from clayey soils. The measured total loads via drain discharge
and TLR in the current study (411–2648 kg ha−1) fit well within the
range (100–5030 kg ha−1 a−1) presented in previous studies in sub-
surface drained clayey monitoring sites in the Nordic countries
(Bechmann, 2012; Turtola et al., 2007; Øygarden et al., 1997). The
current data and simulations showed that sediment load via drain dis-
charge formed the largest load component during the studied two-year
period, which included winter seasons with a minimal snow cover. The
climate change is projected to decrease the amount of snowfall and
SWE in Finland (Räisänen, 2008) and to increase sediment loads
(Puustinen et al., 2007). The results of our study give insight in how
sediment loads can be distributed to different components during years
with a thin snow cover, although based on our results and earlier stu-
dies (e.g. Ulén et al., 2012; Turtola et al., 2007) there can be high in-
terannual and spatial variation in the amount and distribution of the
loads. The quantification of detailed sediment balance presented in our
study is not possible to attain based on measurements and has been
therefore rarely produced. As pointed out by our results and Deasy et al.
(2009), sediment load mitigation should have more focus on subsurface
load pathways. For example, even though buffer strips may have a re-
latively good efficiency in retaining sediment loading from surface
runoff (e.g. Uusi-Kämppä and Jauhiainen, 2010), they can potentially
have only a minor impact on the total load when most of the load oc-
curs via subsurface pathways.
In the current study, sediment transport pathways were studied with

66
M. Turunen et al.
different model structures (Models A–C). The results suggest that lateral
transport of suspended sediments in clayey soils is essential for de-
scribing subsurface sediment transport. Previously Øygarden et al.,
1997, who empirically studied transport of sediment in a silty clay loam
soil also claimed that sediment can originate from the field surface and
then be transport laterally and vertically in preferential flowpaths of the
soil profile. In the current study, the connection was shown to be pro-
nounced in the steep field section (slope 5%) where the hydraulic
gradient is steeper and groundwater flow higher than in the flat section
(Turunen et al., 2015a). Even though the load via TLR was higher in the
steep Section 2 (slope 5%) than Section 1 (slope 1%), a higher differ-
ence between the load components of the field sections was seen in the
load via groundwater outflow due to the lateral preferential connection
(Table 4). In all models, the subsurface drain discharge was the main
pathway for the sediment load and the variability in the amount of
annual loads was attributed to the amount of annual precipitation. Even
though the load via drain discharge was higher in Section 2 (slope 5%)
than Section 1 (slope 1%), the share of load via drain discharge from
the sum of total simulated load was higher in Section 1 (72–75%) than
in Section 2 (64–68%) (Table 4). The difference was attributed to the
lower amount of groundwater outflow and TLR in the flat section
(Turunen et al., 2015a). The water balances of the field sections are
analysed more comprehensively in Turunen et al. (2015a). When lateral
subsurface sediment transport was triggered (Models A and B), the
groundwater outflow component became the second largest sediment
load pathway in both field sections (Table 4). When lateral transport
description was excluded (Model C), the model was not able to re-
produce the measured loads. Furthermore, the soil erodibility value kr
in Model C (452.6 s2 kg−1 m−2, Eq. (1)) was high compared to the
range of the values (1.2–73.5 s2 kg−1 m−2) presented in previous stu-
dies in different soils (Warsta et al., 2014, 2013b; Wicks and Bathurst,
1996), which suggests that the Model C may not be physically plausible.
In addition to continuous biopores (e.g. Nielsen et al., 2010), it can
be assumed that the interface between the plough layer and the un-
derlying soil profile can provide preferential flowpaths for movement of
water and transport of particles and colloids (e.g. Haria et al., 1994). As
discontinuous macropores can facilitate preferential flow (Alberti and
Cey, 2011; Nieber and Sidle, 2010), they may also enable preferential
transport of colloid-size particles.
Several studies have indicated that particle sieving and retention
can occur when particles are transported from field surface to subsur-
face drains (van den Bogaert et al., 2016; Burkhardt et al., 2008;
Turtola et al., 2007; Jarvis et al., 1999). Kjær et al. (2011) suspected
that solutes can be transported laterally via preferential flowpaths
while particles are likely retained in the soil. However, suspended solids
can include many particle size classes, and Ulén (2004), who studied
the characteristics of suspended solids in drain discharge in a clayey
soil, noticed that the majority of TSS in drain discharge were of colloid
size and not highly sensitive to filtering. Retention and sieving pro-
cesses have been rarely studied in the context of sediment balances, but
the simulations of the current study suggested that a large share of the
eroded sediment stayed in the field due to the retention and sieving
processes. A simulation scenario also pointed out that the retention
process was controlled more by the soil hydraulic properties than by the
terrain topography in the simulations. Further development of the
model assessment could benefit of more detailed inclusion of the re-
tention processes as well as computational schemes to simulate trans-
port of several particle size classes simultaneously (e.g. Beuselinck
et al., 2002). Also process description of sediment retention could be
improved and frost impacts on water flow and erodibility should be
included to the model to address the load generation processes in more
detail.
The transport pathways and processes in the soil profile are complex
and the implications of model assumptions related to soil para-
meterisation are not well known. For example, Jarvis et al. (1999)
successfully simulated TSS concentrations in drain discharge in clay soil
67
Soil & Tillage Research 174 (2017) 58–69
with a 1D model and Warsta et al. (2014) succeeded in reproducing
observed sediment loads and TSS concentrations in drain discharge
with a 3D model with an assumption of isotropic water flow and se-
diment transport in the soil profile. Our model applications with the
three different model configurations demonstrated that different model
structures and parameterizations can lead to a qualitatively similar fit
against the load and concentration data from drain discharge and TLR,
but to different estimate of the groundwater outflow load component.
This points out that more detailed data and process understanding
would be needed for more accurate quantification of all transport
pathways. Whereas Models A–B produced an estimate of the sediment
balances in the current study, detailed empirical studies for example of
lateral transport in different soil layers would be of benefit for setting
up more detailed model structures. The model applications and the data
also suggest that models which exclude descriptions of subsurface se-
diment transport easily exclude one of the dominant load pathways and
models which include subsurface sediment transport are likely subject
to non-uniqueness issues. These issues can further cause structural
uncertainty in models, and structural uncertainty can cause predictive
uncertainty in the simulation results, as pointed out e.g. by Refsgaard
et al. (2006) and Højberg and Refsgaard (2005). Moreover, our results
demonstrate that parametric uncertainty analysis may not describe the
uncertainties related to the model structure, and can thus disregard
uncertainties related to groundwater outflow component.
In contrast to the simulation of sediment loads, the modelling ap-
proach of the current study was not able to reproduce the observed
concentration dynamics, even though the simulated concentration le-
vels corresponded with the observed levels. Previously also Rankinen
et al. (2010), who simulated long-term TSS concentrations and loads in
Finnish catchments with the INCA-Sed model, were unable to ade-
quately simulate concentration dynamics in a catchment dominated by
clayey soils, even though concentration dynamics in the catchments
with other soil types were satisfactorily described by the model. Typi-
cally computational models describe erosion as a result of the erosive
forces of overland flow and raindrops (Merritt et al., 2003) and soil
erodibility is often set as a constant value in the simulations (e.g.
Rankinen et al., 2010; Jarritt and Lawrence, 2007; Lundekvam, 2007;
Tattari et al., 2001). However, empirical studies (e.g. Soinne et al.,
2016; Muukkonen et al., 2009; Pietola et al., 2005; Bryan, 2000) have
shown that erodibility in clayey soils is a dynamic property and that in
clayey soils processes such as slaking and dispersion also induce erosion
(Aura et al., 2006; Amézketa, 1999; Bissonnais and Singer, 1992). Thus,
we presume that the concentration dynamics were not adequately re-
produced by our model since soil erodibility was treated as a constant
value and processes such as slaking and dispersion were neglected. On
the other hand, the composite sampling strategy of the current study
also aggregates information of the concentration dynamics, and it is
suggested that continuous observations of the dynamics would be
beneficial when testing models. However, the fact that simulated and
observed loads corresponded more closely with each other than the
concentration dynamics also demonstrated that concentration data can
include more information of the erosion and sediment transport pro-
cesses than load data. Additional tests conducted in the current study
(not shown) to include sediment sieving function of Jarvis et al. (1999)
to the model did not result in an adequate fit of concentration dynamics
against the data, even though the function provided a tool to control
concentration dynamics.
The simulation results also demonstrated how the erosion and se-
diment transport model is more sensitive to changes in soil hydraulic
properties than the water flow model (Fig. 6). Thus, TSS transport
processes in the subsurface domain appear to function differently than
the water flow processes and it is not only the water flow which de-
termines the sediment losses from a field. Previously also Ulén et al.
(2014), who empirically studied transport of substances in clay soils,
noticed that transport of substances in the subsurface domain can be
highly dependent on the characteristics of the soil structure.
M. Turunen et al.
6. Conclusions
According to the field-scale data and the 3D simulations with the
dual-permeability model, subsurface load pathways can dominate the
annual sediment balance in Nordic clayey fields when the winter per-
iods have a minimal snow cover. The results further demonstrate how
sediment load can be distributed to different components in clayey
subdrained soils in high-latitude during years with an intermittent or a
low snow cover, although a more comprehensive assessment would
require observations of the erosion and sediment transport processes
within the fields and simulations throughout several years.
The three different model structures and parameterisations de-
monstrated how different assumptions embedded in the model setup
can result in similar fit against the measured loads from drain discharge
and TLR. Thus erosion models are likely subject to a non-uniqueness
problem when calibrated solely against these data types. The results
also demonstrate how simulations with different model structures can
reveal uncertainties which parameter sensitivity analysis may not en-
compass. According to the simulations, lateral transport of suspended
solids in soil preferential flow pathways likely has a role in the sediment
load processes in clayey subsurface drained fields, and a large share of
detached particles do not form load in clayey fields due to the settling
and retention processes. Correspondence between the simulated and
observed loads, but the mismatch between the simulated and measured
concentration dynamics demonstrated that concentration data can in-
clude more information on the sediment erosion and sediment transport
processes than load data. Furthermore, the role of dynamic soil erod-
ibilities and slaking and dispersion processes in clay soils should be
further studied and included in erosion and sediment transport models
to describe concentration dynamics more accurately.
The sensitivity analysis demonstrated how water balances are less
sensitive to changes in soil hydraulic properties compared to sediment
balances. Thus, it is not only the water flow which controls sediment
transport in clayey soils but also soil structure can have a clear impact
on sediment load generation.
Acknowledgements
The study was funded by the Finnish Drainage Foundation sr, Maa-
ja vesitekniikan tuki ry, the Ministry of Agriculture and Forestry, Aalto
University School of Engineering, Tekniikan edistämissäätiö, Academy
of Finland and Sven Hallin Research Foundation sr. We would like to
thank Jyrki Nurminen, Laura Alakukku, Helena Äijö, Merja Myllys and
Markku Puustinen from the TOSKA—project for their valuable com-
ments. We also acknowledge CSC-IT Center for Science Ltd. for the
allocation of computational resources.
References
Äijö, H., Myllys, M., Nurminen, J., Turunen, M., Warsta, L., Paasonen-Kivekäs, M.,
Korpelainen, E., Salo, H., Sikkilä, M., Alakukku, L., Puustinen, M. 2014. PVO2-hanke,
Salaojitustekniikat ja pellon vesitalouden optimointi Loppuraportti 2014.
Salaojituksen tutkimusyhdistys ry:n tiedote 31. Finnish Field Drainage Association,
Helsinki, Finland, pp. 105.
Øygarden, L., Kværner, J., Jenssen, P.D., 1997. Soil erosion via preferential flow to
drainage systems in clay soils. Geoderma 76, 65–86. http://dx.doi.org/10.1016/
S0016-7061(96)00099-7.
Alberti, D.R., Cey, E.E., 2011. Evaluation of macropore flow and transport using three-
dimensional simulation of tension infiltration experiments. Vadose Zone J. 10,
603–617. http://dx.doi.org/10.2136/vzj2010.0104.
Allaire, S.E., Roulier, S., Cessna, A.J., 2009. Quantifying preferential flow in soils: A re-
view of different techniques. J. Hydrol. 378, 179–204. http://dx.doi.org/10.1016/j.
jhydrol.2009.08.013.
Amézketa, E., 1999. Soil aggregate stability: a review. J. Sustainable Agric. 14, 83–151.
http://dx.doi.org/10.1300/J064v14n02_08.
Aura, E., Saarela, K., Räty, M. 2006. Savimaiden eroosio. MTT:n selvityksiä 118 (in
Finnish). Maa- ja elintarviketalouden tutkimuskeskus, Jokioinen, Finland, 32 pp.
Bechmann, M., 2012. Effect of tillage on sediment and phosphorus losses from a field and
a catchment in south eastern Norway. Acta Agric. Scand. B 62, 206–216. http://dx.
doi.org/10.1080/09064710.2012.715183.
Berisso, F.E., Schjønning, P., Keller, T., Lamandé, M., Simojoki, A., Iversen, B.V.,
68
Soil & Tillage Research 174 (2017) 58–69
Alakukku, L., Forkman, J., 2013. Gas transport and subsoil pore characteristics: an-
isotropy and long-term effects of compaction. Geoderma 195–196, 184–191. http://
dx.doi.org/10.1016/j.geoderma.2012.12.002.
Beuselinck, L., Hairsine, P.B., Govers, G., Poesen, J., 2002. Evaluating a single-class net
deposition equation in overland flow conditions. Water Resour. Res. 38, 15-1–15-10.
http://dx.doi.org/10.1029/2001WR000248.
Bhuyan, S.J., Kalita, P.K., Janssen, K.A., Barnes, P.L., 2002. Soil loss predictions with
three erosion simulation models. Environ. Model. Softw. 17, 135–144. http://dx.doi.
org/10.1016/S1364-8152(01)00046-9.
Bissonnais, Y.L., Singer, M.J., 1992. Crusting, runoff, and erosion response to soil water
content and successive rainfalls. Soil Sci. Soc. Am. J. 56, 1898–1903. http://dx.doi.
org/10.2136/sssaj1992.03615995005600060042x.
Bryan, R.B., 2000. Soil erodibility and processes of water erosion on hillslope.
Geomorphology 32, 385–415. http://dx.doi.org/10.1016/S0169-555X(99)00105-1.
Burkhardt, M., Kasteel, R., Vanderborght, J., Vereecken, H., 2008. Field study on colloid
transport using fluorescent microspheres. Eur. J. Soil Sci. 59, 82–93. http://dx.doi.
org/10.1111/j.1365-2389.2007.00989.x.
Culley, J.L.B., Bolton, E.F., Bernyk, V., 1983. Suspended solids and phosphorus loads from
a clay soil: I. Plot studies. J. Environ. Qual. 12, 493–498. http://dx.doi.org/10.2134/
jeq1983.00472425001200040011x.
Deasy, C., Brazier, R.E., Heathwaite, A.L., Hodgkinson, R., 2009. Pathways of runoff and
sediment transfer in small agricultural catchments. Hydrol. Process. 23, 1349–1358.
http://dx.doi.org/10.1002/hyp.7257.
Eriksson, J., Andersson, A., Andersson, R. 1999. Texture of agricultural topsoils in
Sweden. In: Report, vol. 4955. Swedish Environmental Protection Agency,
Stockholm, Sweden, 48 pp.
FAO, 2007. World reference base for soil resources 2006 first update. World Soil
Resources Reports No. 103. FAO, Rome.
Feddes, R.A., Kowalik, P.J., Zaradny, H., 1978. Simulation of Field Water Use and Crop
Yield. Pudoc Wageningen, The Netherlands (189 pp).
García-Ruiz, J.M., Beguería, S., Nadal-Romero, E., González-Hidalgo, J.C., Lana-Renault,
N., Sanjuán, Y., 2015. A meta-analysis of soil erosion rates across the world.
Geomorphology 239, 160–173. http://dx.doi.org/10.1016/j.geomorph.2015.03.008.
Gerke, H.H., van Genuchten, M.T., 1993. Evaluation of a first order water transfer term
for variably saturated dual-porosity flow models. Water Resour. Res. 29, 1225–1238.
http://dx.doi.org/10.1029/92WR02467.
Højberg, A.L., Refsgaard, J.C., 2005. Model uncertainty −parameter uncertainty versus
conceptual models. Water Sci. Technol. 52, 177–186.
Haria, A.H., Johnson, A.C., Bell, J.P., Batchelor, C.H., 1994. Water movement and iso-
proturon behaviour in a drained heavy clay soil: 1. Preferential flow processes. J.
Hydrol. 163, 203–216. http://dx.doi.org/10.1016/0022-1694(94)90140-6.
Jarritt, N.P., Lawrence, D.S.L., 2007. Fine sediment delivery and transfer in lowland
catchments: modelling suspended sediment concentrations in response to hydro-
logical forcing. Hydrol. Process. 21, 2729–2744. http://dx.doi.org/10.1002/hyp.
6402.
Jarvis, N.J., Villholth, K.G., Ulén, B., 1999. Modelling particle mobilization and leaching
in macroporous soil. Eur. J. Soil Sci. 50, 621–632. http://dx.doi.org/10.1046/j.1365-
2389.1999.00269.x.
Jarvis, N.J., 2007. A review of non-equilibrium water flow and solute transport in soil
macropores: principles, controlling factors and consequences for water quality. Eur.
J. Soil Sci. 58, 523–546. http://dx.doi.org/10.1111/j.1365-2389.2007.00915.x.
Jarvis, N.J., 2008. Near-saturated hydraulic properties of macroporous soils. Vadose Zone
J. 7, 1302–1310. http://dx.doi.org/10.2136/vzj2008.0065.
Kjær, J., Ernstsen, V., Jacobsen, O.H., Hansen, N., de Jonge, L.W., Olsen, P., 2011.
Transport modes and pathways of the strongly sorbing pesticides glyphosate and
pendimethalin through structured drained soils. Chemosphere 84, 471–479. http://
dx.doi.org/10.1016/j.chemosphere.2011.03.029.
Koivusalo, H., Heikinheimo, M., Karvonen, T., 2001. Test of a simple two-layer para-
meterisation to simulate the energy balance and temperature of a snow pack. Theor.
Appl. Climatol. 70, 65–79. http://dx.doi.org/10.1007/s007040170006.
Koivusalo, H., Turunen, M., Salo, H., Haahti, K., Nousiainen, R., Warsta, L., 2017.
Analysis of water balance and runoff generation in high latitude agricultural fields
during mild and cold winters. Hydrol. Res (Accepted).
Legates, D.R., McCabe, G.J., 1999. Evaluating the use of goodness of fit measures in
hydrologic and hydroclimatic model validation. Water Resour. Res. 35, 233–241.
http://dx.doi.org/10.1029/1998WR900018.
Lundekvam, H.E., 2007. Plot studies and modelling of hydrology and erosion in southeast
Norway. Catena 71, 200–209. http://dx.doi.org/10.1016/j.catena.2007.03.004.
Mattsson, T., Finér, L., Kortelainen, P., Sallantaus, T., 2003. Brook water quality and
background leaching from unmanaged forested catchments in Finland. Water Air Soil
Pollut. 147, 275–298.
Merritt, W.S., Letcher, R.A., Jakeman, A.J., 2003. A review of erosion and sediment
transport models. Environ. Model. Softw. 18, 761–799. http://dx.doi.org/10.1016/
S1364-8152(03)00078-1.
Muukkonen, P., Hartikainen, H., Alakukku, L., 2009. Effect of soil structure disturbance
on erosion and phosphorus losses from Finnish clay soil. Soil Tillage Res. 103, 84–91.
http://dx.doi.org/10.1016/j.still.2008.09.007.
Nearing, M.A., Jetten, V., Baffaut, C., Cerdan, O., Couturier, A., Hernandez, M., Le
Bissonnais, Y., Nichols, M.H., Nunes, J.P., Renschler, C.S., Souchère, V., 2005.
Modeling response of soil erosion and runoff to changes in precipitation and cover.
Catena 61, 131–154. http://dx.doi.org/10.1016/j.catena.2005.03.007.
Nieber, J.L., Sidle, R.C., 2010. How do disconnected macropores in sloping soils facilitate
preferential flow? Hydrol. Process. 24, 1582–1594. http://dx.doi.org/10.1002/hyp.
7633.
Nielsen, M.H., Styczen, M., Ernstsen, V., Petersen, C.T., Hansen, S., 2010. Field study of
preferential flow pathways in and between drain trenches. Vadose Zone J. 9,
M. Turunen et al.
1073–1079.
Panagos, P., Van Liedekerke, M., Jones, A., Montanarella, L., 2012. European Soil Data
Centre: response to European policy support and public data requirements. Land Use
Policy 29, 329–338. http://dx.doi.org/10.1016/j.landusepol.2011.07.003.
Pietola, L., Horn, R., Yli-Halla, M., 2005. Effects of trampling by cattle on the hydraulic
and mechanical properties of soil. Soil Tillage Res. 82, 99–108. http://dx.doi.org/10.
1016/j.still.2004.08.004.
Puustinen, M., Merilä, E., Palko, J., Seuna, P., 1994. Kuivatustila, viljelykäytäntö ja
vesistökuormitukseen vaikuttavat ominaisuudet Suomen pelloilla. In: With an
English abstract: (Drainage Level, Cultivation Practices and Factors Affecting Load on
Waterways in Finnish Farmland). Vesi- ja ymparistohallinnon julkaisuja −sarja A
198. National Board of Waters and Environment, Helsinki, Finland, pp. 319.
Puustinen, M., Tattari, S., Koskiaho, J., Linjama, J., 2007. Influence of seasonal and an-
nual hydrological variations on erosion and phosphorus transport from arable areas
in Finland. Soil Tillage Res. 93, 44–55. http://dx.doi.org/10.1016/j.still.2006.03.
011.
Quinton, J.N., Govers, G., Van Oost, K., Bardgett, R.D., 2010. The impact of agricultural
soil erosion on biogeochemical cycling. Nat. Geosci. 3, 311–314. http://dx.doi.org/
10.1038/ngeo838.
Räisänen, J., 2008. Warmer climate: less or more snow? Climate Dyn. 30, 307–319.
http://dx.doi.org/10.1007/s00382-007-0289-y.
Rankinen, K., Thouvenot-Korppoo, M., Lazar, A., Lawrence, D.S.L., Butterfield, D.,
Veijalainen, N., Huttunen, I., Lepistö, A., 2010. Application of catchment scale se-
diment delivery model INCA-sed to four small study catchments in Finland. Catena
83, 64–75. http://dx.doi.org/10.1016/j.catena.2010.07.005.
Refsgaard, J.C., van der Sluijs, J.P., Brown, J., van der Keur, P., 2006. A framework for
dealing with uncertainty due to model structure error. Adv. Water Resour. 29,
1586–1597. http://dx.doi.org/10.1016/j.advwatres.2005.11.013.
SFS, 1990. SFS 3008: Determination of Total Residue and Total Fixed Residue in Water,
Sludge and Sediment. Finnish Standards Association (SFS) (3 pp).
Salo, H., Warsta, L., Turunen, M., Paasonen-Kivekäs, M., Nurminen, J., Koivusalo, H.,
2015. Development and application of a solute transport model to describe field-scale
nitrogen processes during autumn rains. Acta Agric. Scand. B 65 (Suppl. 1), 30–43.
http://dx.doi.org/10.1080/09064710.2014.971861.
Soinne, H., Hyväluoma, J., Ketoja, E., Turtola, E., 2016. Relative importance of organic
carbon, land use and moisture conditions for the aggregate stability of post-glacial
clay soils. Soil Tillage Res. 158, 1–9. http://dx.doi.org/10.1016/j.still.2015.10.014.
Taskinen, A., Bruen, M., 2007. Incremental distributed modelling investigation in a small
agricultural catchment: 2. Erosion and phosphorus transport. Hydrol. Process. 21,
92–102. http://dx.doi.org/10.1002/hyp.6190.
Tattari, S., Bärlund, I., Rekolainen, S., Posch, M., Siimes, K., Tuhkanen, H.-R., Yli-Halla,
M., 2001. Modeling sediment yield and phosphorus transport in Finnish clayey soils.
Trans. ASAE 44, 297–307. http://dx.doi.org/10.13031/2013.4691.
Turtola, E., Alakukku, L., Uusitalo, R., 2007. Surface runoff, subsurface drainflow and soil
erosion as affected by tillage in a clayey Finnish soil. Agric. Food Sci. 16, 332–351.
Turunen, M., Warsta, L., Paasonen-Kivekäs, M., Nurminen, J., Myllys, M., Alakukku, L.,
Äijö, H., Puustinen, M., Koivusalo, H., 2013. Modeling water balance and effects of
different subsurface drainage methods on water outflow components in a clayey
agricultural field in boreal conditions. Agric.Water Manag. 121, 135–148. http://dx.
doi.org/10.1016/j.agwat.2013.01.012.
Turunen, M., Warsta, L., Paasonen-Kivekäs, M., Nurminen, J., Alakukku, L., Myllys, M.,
Koivusalo, H., 2015a. Effects of terrain slope on long-term and seasonal water bal-
ances in clayey, subsurface drained agricultural fields in high latitude conditions.
Agric. Water Manag. 150, 139–151. http://dx.doi.org/10.1016/j.agwat.2014.12.008.
Turunen, M., Warsta, L., Paasonen-Kivekäs, M., Nurminen, J., Koivusalo, H., 2015b.
Simulating water balance and evapotranspiration in a subsurface drained clayey
agricultural field in high-latitude conditions. Acta Agric. Scand. B 65, 44–57. http://
dx.doi.org/10.1080/09064710.2014.975834.
Ulén, B., Bechmann, M., Øygarden, L., Kyllmar, K., 2012. Soil erosion in Nordic coun-
tries—future challenges and research needs. Acta Agric. Scand. B 62, 176–184.
http://dx.doi.org/10.1080/09064710.2012.712862.
Ulén, B.M., Larsbo, M., Kreuger, J.K., Svanbäck, A., 2014. Spatial variation in herbicide
leaching from a marine clay soil via subsurface drains. Pest Manag. Sci. 70, 405–414.
http://dx.doi.org/10.1002/ps.3574.
Ulén, B., 2004. Size and settling velocities of phosphorus-containing particles in water
from agricultural drains. Water Air Soil Pollut. 157, 331–343. http://dx.doi.org/10.
1023/B:WATE.0000038906.18517.e2.
69
Soil & Tillage Research 174 (2017) 58–69
Uusi-Kämppä, J., Jauhiainen, L., 2010. Long-term monitoring of buffer zone efficiency
under different cultivation techniques in boreal conditions. Agric. Ecosyst. Environ.
137, 75–85. http://dx.doi.org/10.1016/j.agee.2010.01.002. (Special section
Harvested perennial grasslands: Ecological models for farming’s perennial future).
Uusitalo, R., Turtola, E., Kauppila, T., Lilja, T., 2001. Particulate phosphorus and sedi-
ment in surface runoff and drainflow from clayey soils. J. Environ. Q. 30, 589–595.
http://dx.doi.org/10.2134/jeq2001.302589x.
Vagstad, N., Stålnacke, P., Andersen, H.-E., Deelstra, J., Jansons, V., Kyllmar, K., Loigu,
E., Rekolainen, S., Tumas, R., 2004. Regional variations in diffuse nitrogen losses
from agriculture in the Nordic and Baltic regions. Hydrol. Earth Syst. Sci. 651–662
(Discussions 8).
Vanmaercke, M., Poesen, J., Verstraeten, G., de Vente, J., Ocakoglu, F., 2011. Sediment
yield in Europe: spatial patterns and scale dependency. Geomorphology 130,
142–161. http://dx.doi.org/10.1016/j.geomorph.2011.03.010.
Vereecken, H., Schnepf, A., Hopmans, J.W., Javaux, M., Roose, D., Vanderborght, J.,
Young, M.H., Amelung, W., Aitkenhead, M., Allison, S.D., Assouline, S., Baveye, P.,
Berli, M., Brüggemann, N., Finke, P., Flury, M., Gaiser, T., Govers, G., Ghezzehei, T.,
Hallett, P., Hendricks Franssen, H.J., Heppell, J., Horn, R., Huisman, J.A., Jacques,
D., Jonard, F., Kollet, S., Lafolie, F., Lamorski, K., Leitner, D., McBratney, A.,
Minasny, B., Montzka, C., Nowak, W., Pachepsky, Y., Padarian, J., Romano, N., Roth,
K., Rothfuss, Y., Rowe, E.C., Schwen, A., Šimůnek, J., Tiktak, A., Van Dam, J., van der
Zee, S.E.A.T.M., Vogel, H.J., Vrugt, J.A., Wöhling, T., Young, I.M., 2016. Modeling
soil processes: review key challenges, and new perspectives. Vadose Zone J. 15.
http://dx.doi.org/10.2136/vzj2015.09.0131.
Warsta, L., Turunen, M., Koivusalo, H., Paasonen-Kivekäs, M., Karvonen, T., Taskinen, A.,
2012. Modelling heat transport and freezing and thawing processes in a clayey,
subsurface drained agricultural field. In: Proceedings of the 11th ICID Int. Drainage
Workshop on Agricultural Drainage, Needs and Future Priorities. Cairo 23-27.9.2012,
Egypt. (pp.10).
Warsta, L., Taskinen, A., Koivusalo, H., Paasonen-Kivekäs, M., Karvonen, T., 2013a.
Modelling soil erosion in a clayey, subsurface-drained agricultural field with a three-
dimensional FLUSH model. J. Hydrol. 498, 132–143. http://dx.doi.org/10.1016/j.
jhydrol.2013.06.020.
Warsta, L., Karvonen, T., Koivusalo, H., Paasonen-Kivekäs, M., Taskinen, A., 2013b.
Simulation of water balance in a clayey, subsurface drained agricultural field with
three-dimensional FLUSH model. J. Hydrol. 476, 395–409. http://dx.doi.org/10.
1016/j.jhydrol.2012.10.053.
Warsta, L., Taskinen, A., Paasonen-Kivekäs, M., Karvonen, T., Koivusalo, H., 2014.
Spatially distributed simulation of water balance and sediment transport in an agri-
cultural field. Soil Tillage Res. 143, 26–37. http://dx.doi.org/10.1016/j.still.2014.05.
008.
Warsta, L., 2011. Modelling water flow and soil erosion in clayey, subsurface drained
agricultural fields. (Doctoral dissertation). Aalto University, School of Engineering,
Department of Civil and Environmental Engineering, Espoo, Finland (pp. 212).
http://lib.tkk.fi/Diss/2011/isbn9789526042893/isbn9789526042893.pdf.
Wellen, C., Arhonditsis, G.B., Long, T., Boyd, D., 2014. Quantifying the uncertainty of
nonpoint source attribution in distributed water quality models: a Bayesian assess-
ment of SWAT’s sediment export predictions. J. Hydrol. 519, 3353–3368. http://dx.
doi.org/10.1016/j.jhydrol.2014.10.007.
Wicks, J.M., Bathurst, J.C., 1996. SHESED: a physically based, distributed erosion and
sediment yield component for the SHE hydrological modelling system. J. Hydrol.
175, 213–238. http://dx.doi.org/10.1016/S0022-1694(96)80012-6.
Yalin, M.S., 1963. An expression for bed-load transportation. Proceedings of the ACAE. J.
Hydraulic Division 89, 221–250.
Yalin, M.S., 1977. Mechanics of Sediment Transport. Pergamon Press, Toronto (298 pp).
Zhang, X.C., Nearing, M.A., Risse, L.M., McGregor, K.C., 1996. Evaluation of WEPP runoff
and soil loss predictions using natural runoff plot data. Trans. ASAE 39, 855–863.
http://dx.doi.org/10.13031/2013.27570.
van Dam, J.C., Groenendijk, P., Hendriks, R.F., Kroes, J.G., 2008. Advances of modeling
water flow in variably saturated soils with SWAP. Vadose Zone J. 7, 640–653. http://
dx.doi.org/10.2136/vzj2007.0060.
van Genuchten, M.Th., 1980. A closed-form equation for predicting the hydraulic con-
ductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892–898.
van den Bogaert, R., Cornu, S., Michel, E., 2016. To which extent do rain interruption
periods affect colloid retention in macroporous soils? Geoderma 275, 40–47. http://
dx.doi.org/10.1016/j.geoderma.2016.04.010.