This paper presents the methodology used for artificial flood
experiments conducted in a small artificial, trained (regulated) channel on
the Nučice experimental agricultural catchment (0.5 km
Excessive soil erosion from upland areas resulting in the transport of soil particles with bound organic matter, nutrients, microbes, or pollutants into the rivers and reservoirs is considered as a major environmental problem (Drummond et al., 2014; Lal et al., 2007; Pimentel et al., 1995; Stoate et al., 2001). The processes of soil particle mobilization and transportation within agriculturally used fields, including the transfer into streams and rivers, have been extensively studied (Báčová and Krása, 2016; Boardman, 2003; Lal, 1998; Neal and Anders, 2015).
The headwater streams and drainage channels in sediment source areas, typically small rural catchments with intensively cultivated soils, have considerable retention capacities for sediment and nutrients (Hession et al., 2003). The streams are narrow with a small flow profile, the baseflow is usually low, the stream bed contains fine-grained particles with a high concentration of nutrients, and extensive vegetation can therefore often be found there. Soil particles that enter the channel during an erosion event can easily get embedded in the channel bed alluvium and be remobilized during subsequent runoff events, and the nutrients tend to be retained (Withers and Jarvie, 2008). The source of the sediment and the processes related to the suspended sediment dynamics in the closing profiles are therefore of fundamental importance for an assessment of the sediment budget and the transport of dissolved or absorbed substances in the catchment (Walling, 2005). However, even physically based mathematical models of soil erosion assume that the sediment transported through water courses originates from a recent (or current) rainfall–runoff event. Similarly, traditional experiments and soil erosion monitoring usually rely on measurements of the sediment yield at the catchment outlet, assuming that the measured sediment yield originates on the hillslopes. If any retention in the channel is expected, no resuspension is then assumed, and this affects the total sediment budget. Minella et al. (2008) point out that the transport capacity of the channel may increase, and that the stream bed sediment is easily mobilized during runoff events with no eroded sediment from the catchment. Zumr et al. (2015) and also Musolff et al. (2015) show that a quick runoff response with no soil erosion on the fields is very commonly observed on cultivated catchments where subsurface runoff or tile drains are the dominant controls. The resuspension regime of the stream bed sediment and the connected nutrient transport depend on the characteristics of the stream, the hydrograph of the flood wave and the actual conditions of the channel (Peterson and Benning, 2013).
The sources of the suspended sediments recorded at the catchment outlet also vary due to seasonally varying vegetation (Hearne et al., 1994). The development of aquatic macrophytes limits the discharge capacity of the channels. Keesstra et al. (2012) evaluated the effect of temporary variable vegetation cover within the natural and semi-natural headwater channels and the stream riparian zone on water and sediment transport. On the basis of numerical modelling, they concluded that vegetation affects resuspension especially during high-flow conditions in streams that are not sediment supply limited. Similarly, Huisman et al. (2013) showed that the previously suspended sediment is mobilized during the later parts of the year. In spring the recently eroded sediment is quickly flushed downstream. Shore et al. (2015) showed that in the case of well-trained channels there is greater potential for fast sediment transportation downstream. However, this is not necessarily the rule in sparsely maintained and over-vegetated channels, where the sediment retention capacity is not negligible.
The key questions that we will address in this paper are as follows.
Can
well-trained and well-regulated stream channels act as a temporal sediment
trap and sediment source due to the resuspension of sediments deposited from
previous erosion events? How does the flood wave transformation regime and the suspended
solids remobilization regime change within one season as a consequence of
various instream vegetation and baseflow conditions? How does the resuspended sediment concentration and the mass
movement change in the event of repeated short flood waves?
To answer these questions, we initiated two sets of three small artificial floods into a typical drainage channel in the rural landscape of central Bohemia, Czech Republic. The experiments were performed recurrently in September and in March, when the channel vegetation, the baseflow, and the channel saturation differ most.
The experiments were performed in the stream which drains the Nučice
rural experimental catchment, Czech Republic (Fig. 1). The Nučice
catchment (49
Map of the Nučice catchment with the measurement sites.
The catchment (0.531 km
The Nučice catchment is drained by an artificially trained narrow stream, which has been piped in the uppermost part. The channel was modified into its current form in the 1950s, with the aim to decrease the groundwater level and to prevent inundation of the fields. The piped section is 530 m in length, and the open channel down to the outlet profile of the experimental catchment extends to 424 m. The straight, deep channel is in direct contact with the surrounding fields. The riparian vegetation is only sparse.
The channel has a trapezoid profile which is 0.6 m in width at the stream
bed, and the slope of the banks is 1 : 2. The stream bed and footslopes up
to 0.3 m are stabilized with concrete tiles. There are two culverts on the
stream. One is 56 m from the start of the open channel, and it is 0.8 m in
inner diameter and 10.2 m in length. The second culvert is 337 m from the
start, and is 0.6 m in diameter and 7.8 m in length. The average depth of
the channel is 1.5 m. The current situation of the channel represents very
well the situation in most small regulated drainage channels in the country:
there has been very little maintenance during the last ca. 30 years. Locally,
therefore, the stabilization is defective and the channel profile has been
covered by extensive weed vegetation with a predominance of stinging nettles
(
The typical flow conditions at the gauging station, as observed during the
period of monitoring (2011–2016), are as low as 0.1–0.2 L s
Stream longitudinal profile with monitoring sites and culverts on the Nučice experimental catchment.
The experiments were performed in the open section of the channel (Fig. 1).
The total monitored length was 424 m, which is the distance between the
injection profile (profile S) and the basin closing profile (profile C). The
point where we injected water is considered as the start (0 m), and the
basin outlet is considered as the end (424 m). The release profile was
placed directly at the beginning of open channel section. Water was pumped
into the stream over a period of 7 min simultaneously from a filled water
reservoir and from the on-board supply of a fire truck, using four fire
hoses. We used drinking water from a nearby reservoir. The total pumped water
volume for each wave was approximately 17 m
We established three monitoring stations along the watercourse (Figs. 1 and 2) to monitor the discharge and the electrical conductivity, and to collect samples for measuring the concentration of the suspended solids. After taking the initial water sample for an evaluation of the baseflow water properties, we started the water sampling in each experimental profile immediately after the flood wave arrived. Samples approximately 1 L in volume were taken every minute during the rise, the peak, and shortly after the decrease of the discharge. After that, the sampling interval was reduced to 2 to 5 min intervals to obtain approximately 30 samples for each wave and observation profile. The samples were analysed in the laboratory. The concentration of suspended solids, phosphorus, and nitrogen were measured. Each station was equipped with a pre-programmed camera on a tripod to confirm the exact time of arrival of the wave and to document the progress of the wave and the sampling.
The first 66 m long section between pumping station S and profile A was
meant to be used for wave dispersion and fluctuation stilling due to
non-homogeneities in the pumping process. Station A was situated at the upper
outlet of the culvert. The culvert ends with a free outflow, where the
discharge was measured both volumetrically and hydrometrically. Monitoring
profile B, in a distance of 224 m, was equipped with a rectangular weir, and
the discharge was estimated on the basis of the measured depth of the water
and the known rating curve. Profile C was located at the outlet of the
experimental basin, positioned 424 m from the beginning of open channel
section. The outlet is permanently equipped with an H flume with capacity of
up to 400 L s
Initial conditions before the experiments.
Stream vegetation conditions during the two experiments. Dense
instream vegetation in September 2012
The experiments were conducted in September 2012 and in March 2013. Within
each of the campaigns we carried out three wave experiments (W1 to W3 in
September, W4 to W6 in March). Subsequent waves were always initiated after
the discharge in the outlet profile (station C) had dropped close to the
initial baseflow. The second waves in the series (W2 and W5) were enriched
with NaCl as a tracer to compare the water velocity and the water celerity
during wave propagation. The tracer was dissolved in the water reservoir to
obtain a concentration of approximately 6 g L
The general conditions within the catchment and the stream prior to the experiments differed in September 2012 and in March 2013 (Table 1). In September, the stream baseflow was at its annual minimum, the soil water content was below its field capacity, and the instream vegetation was densely overgrown. In March 2013, the baseflow was at its annual maximum because of saturated soil from the snow melting, the instream vegetation was sparse, and the remaining plants were flattened on the stream bed and banks (Fig. 3).
For a quantitative evaluation of the impact of the vegetation on the transformation of the flood wave, we built a simple 1-D hydraulic model in HEC-RAS unsteady flow. The variable parameter between the September and March experiments was the stream roughness factor, which we attribute as a proxy of the actual character and density of the vegetation (Brookes, 1986). The approach is similar to the one of Nikora et al. (2008), who showed that the flow resistance is determined mainly by the general characteristics of the bulk instream vegetation, rather than by individual species. We assume that the vegetation resistance is mainly due to the stem blockage factor causing frictional energy losses, rather than by volume displacement effect, as the dense canopy occupies the cross section of the channel (Green, 2005).
The simulation of the hydraulic conditions in a channel system is based on a system of Saint Venant's equations. The evaluation of the energy gradient (friction slope) is based on Chézy's equation for velocity, where complex energy losses are represented by friction (Manning's formula). Unsteady flow is solved by the one-dimensional version of the finite difference method using the Preissman's implicit differential scheme.
The geometry of the channel was based on a set of 28 measured cross sections obtained by land surveying. To improve the precision of the simulations and to overcome numerical stability problems, intermediate profiles were added to the model by geometrical interpolation between the original cross sections. The average distance between cross sections in the final geometry set was 0.5 m. Moreover, each cross section was extended by a narrow Preismann bottom slot to deal with numerical stability issues in cases of very low discharges (propagation of the wave in an originally dry channel). The model was loaded on the upper end with the measured flow. The rating curve based on the Chezy equation for uniform flow was used as a downstream boundary condition.
The water exfiltration into the channel banks was simulated with the Richards
equation; van Genuchten's model for soil water retention curve was used. The
hydraulic characteristics and the saturated hydraulic conductivity of the
stream banks were assumed to be the same as the measured subsoil matrix
characteristics on the surrounding fields (residual water content
Hydrographs of flood propagation along the monitored stream in the Nučice catchment. The different dynamics in September (W1–W3) and March (W4–W6) are caused by the current state of the stream and the vegetation conditions.
Measured outflow rates and the concentration of suspended solids in
the Nučice catchment outlet during experiments conducted in September
2012
The hydrographs and sedigraphs of all six waves are shown in Figs. 4–6. The shape characteristics of the waves and the transformation are summarized in Table 2. The water and sediment balance are presented in Table 3.
The hydrographs of the subsequent experiments differed on all the monitoring
profiles. The velocity of the waves and the maximum flow rates increased
between the successive waves (Fig. 4). All the waves approached the A profile
less than 5 min after the start of the experiments, and the subsequent waves
reached the profile slightly earlier. The time difference between the
approaches of waves W1 and W3 is 41 s. Waves W1 and W2 reached a similar
peak discharge of approximately 30 L s
The time lags in the B profile already differed. Wave W1 arrived after almost
20 min. Waves W2 and W3 were faster, and appeared 15 min after pumping
began. The peak discharge also increased, with subsequent waves starting at
12.8 L s
Hydrograph characteristics. n/a: data not available.
The wave celerity along the stream was calculated according to the wave
arrival time, which we defined as the time of the first rise of the
hydrograph (Table 2). The average wave celerity for W1 was 0.20 m s
The hydrographs of waves W3–W6 are very similar to each other, and the time
lags differ by less than 1 min. The waves approached profile A after
3 min, profile B after 9 min, and profile C after 16 min. The peak
discharge values observed in the individual profiles were also similar, but
the last wave, W6, reached slightly higher values. The average peak discharge
of W4 to W6 in profile A was 32.5 L s
For each flood wave, the simulated flow hydrograph at the downstream end was compared with the measured discharge data (Fig. 6). The accuracy of the fit was evaluated by comparing two characteristic parameters – the time and the discharge at the wave peak. Manning's hydraulic roughness was used as the calibration parameter, separately for the September scenario and for the March scenario.
Comparison of the flood wave characteristics measured at the gauging profile and simulated by HEC-RAS.
The approach of the W4 wave front at stationing of 400 m. The times (hh:mm:ss) stand for the duration from the start of the experiment.
Water and sediment budget as measured at the gauging stations (profile C).
The total amount of sediment released during the September experimental
campaign was 41.6 kg, and during the March experimental campaign the amount
was 124.5 kg. Assuming regular initial distribution and uniform release of
the sediment, this represents 0.10 kg m
The maximum suspended solids concentration in profile A, with a value of
9 g L
The amount of carried sediment measured at site C decreased from 48.5 to
30.7 kg between waves W4 and W6. The peak concentration of suspended solids
in profile C reached close to 8 g L
The set-up of our experiment was based on a study made by Eder et al. (2014), who carried out two flushing experiments in a natural stream in the HOAL experimental catchment in Austria (Blöschl et al., 2016). The catchment is similar in size, climate, soils, and management to the Nučice catchment. The HOAL stream meanders through a forested belt. The monitored length is 590 m, with an average slope of 2.4 %. The stream cross section is irregular, and the channel width varies from 0.6 to 1.0 m. The longitudinal slope is relatively homogeneous over the whole monitored length, with the exception of the initial 90 m, which are significantly steeper. Our experimental section had a convex course, with the slope gradually increasing from 2.3 % in the first section of ca. 70 m to 3.3 % in the last section ca. 200 m in length (nearly half of the length of the total monitored course).
The HOAL experiments were carried out in August 2011 on 2 days separated
from each other by a gap of approximately 1 week. The volume of pumped
water was 17 m
We have to keep in mind that the artificial flood waves used in this experiment were relatively small in volume and of short duration. Based on the monitoring of the natural runoff events, we estimate that the minimum time needed for complete bedload sediment removal with comparable discharge is in the range of 10–24 h (Zumr et al., 2015). Although the amount of sediment transported by the waves decreased within each set of experiments, there was still enough sediment left in the channel that could be released if there were to be a larger wave. The clockwise hysteresis of the sediment concentration–discharge relation suggests that the sediment originates from nearby. Similar results were observed, for example, by Molder et al. (2015) and Seeger et al. (2004). The amount of resuspended sediment was significantly lower in summer conditions. We relate this to the particular conditions in the channel with dense erect vegetation and dry conditions, which led to storage of a considerable proportion of the water.
Significant changes in surface roughness, which also affect this process, may be documented by mathematical modelling of the movement of the wave through the experimental section using a 1-D hydraulic model.
These processes, though with reverse trends exhibiting a decrease in the amount of resuspended sediment over sections downstream, due to very different stream channel characteristics, have also been confirmed by the similar HOAL experiment, performed by Eder et al. (2014). The results clearly show the potential of even well-trained channels without visible signs of sediment accumulation to release sediment during flood events.
The experiments showed that a well-trained stream can act both as a trap and as a sediment source. However, the hydraulic characteristics of the flood wave and the physical and geometrical characteristics of the channel will be crucial for indicating whether deposition or remobilization will occur in a given section and during a given event.
All the sedigraphs show similar behaviour (Fig. 5). The sediment
concentration increases rapidly immediately after the arrival of the wave.
The highest sediment concentration is always directly measured at the wave
front, and does not necessarily correspond to the peak discharge. After
culmination of the wave, the concentration of the sediment also decreases.
The highest sediment concentration peak was observed when waves W1 and W4
were approaching, i.e. in the initial experiment of each campaign. We relate
this to the fine-textured sediment that had been deposited in the stream
during previous events. Our initial assumption was that most of the solid
particles move only a short distance, because of low water velocity and short
wave duration. Only the finest particles would be mobile enough to travel
longer distances. However, our experiment showed that while the discharges
decreased along experimental sections A to C, the sediment concentration
increased (see Tables 2, 3 and Fig. 4). This suggests that the transport
capacity of the stream had not been reached, even for lower discharges at the
outlet point, and at least fine-textured soil particles were resuspended and
transported over the whole observed section of the stream channel. To test
this assumption, we estimated the maximum clear water transport capacity
during the observed flow according to the simple transport capacity equation
proposed by Govers (1990). The transport capacity of the peak flow was
990 g L
The remobilized sediment mass was two to three times higher in March than in
September. We relate this to higher water velocity, as a result of which
heavier particles contribute to the recorded amount, due to the higher
transport capacity. It is not technically possible to measure the total
initial mass of the sediment in the stream, and we can only make an estimate
on the basis of previous runoff events that the conditions in September and
March were similar, and were close to a quasi-steady state for the stream. In
both cases, the last antecedent erosion event had taken place more than 2 months before the experiment, followed by at least one runoff event when no
soil erosion was recorded and the discharge was above 5 L s
The experiments confirmed our assumption that vegetation development is a crucial parameter that affects flood wave retention and propagation, as well as the sediment dynamics. Contrasting vegetation conditions are documented by Fig. 3 – fully erect well-developed dense vegetation in the September set of experiments (W1–W3) vs. no erect vegetation in the March set of experiments (W4–W6).
The general behaviour of the sediment transport during both sets of experiments (September conditions vs. March conditions) is the same, since the bedload sediment is available throughout the year. In both cases, it decreases event by event, but the sediment load increases along the sections. The general difference between the resuspension in fully-developed vegetation (September) vs. the March conditions is 2.7 times higher for the first event and 3.2 times higher for the second and third events, as regards total transported sediment. As regards sediment concentrations, the peak values were the same for the first events and ca. 50 % for the second and third events. Well-developed vegetation therefore significantly increased the trapping capacity of the stream channel (Keesstra et al., 2012).
The flood waves propagate differently in September and in March. While in September the successive waves speed up, in March the wave velocities are very similar for all experiments. When we compare the speed of flood propagation in a vegetated channel and in an empty channel, the March waves propagate twice as fast, and reach a 30 % higher peak discharge. The reason is twofold: (i) there is higher vegetation resistance in September, and (ii) there is higher baseflow and therefore a greater difference between the water velocity and the wave celerity in March. The volume of water recovered on site C is slightly larger during the March experiments. It should be noted that in reality the general summer and winter regime may vary because of variable rainfall patterns and catchment conditions (Buendia et al., 2016; Walling and Amos, 1999).
There is a significant water loss in the case of wave W1, which was released into an almost empty channel with dry stream banks (Table 1). Because of water exfiltration, interception on the vegetation, and filling of the streambed depressions along the 424 m long channel, the water loss reached 31 %. During all other experiments, including W4, the water loss was only 10–15 %.
A comparison of the simulated hydrographs and the measured hydrographs is
presented in Fig. 6. It can be stated that the numerical simulations mimic
the monitored hydrographs, and the effect of the vegetation seems to be a
correct assumption. In the simulation of the September experiments, very high
Manning roughness values (
Our paper has presented the methodology for an artificial flood experiment
conducted on an experimental agricultural catchment, and the results of the
experiment. Three successive flood waves, each with an approximate volume of
17 m
On the basis of our results, we concluded that even well-trained and straight channels trap sediment, which can be mobilized by subsequent small floods.
The resuspension regime depends on the current conditions of the stream and the instream vegetation, and therefore changes significantly in the course of a year. The sediment moves quickly in winter and early spring, but in the later part of the year the channel serves as a sediment trap and the resuspension is slower, if dense vegetation is present.
The resuspension regime and the sediment loads within the succeeding small flood waves do not change considerably. The artificial waves that we initiated do not have sufficient magnitude to flush the bedload sediment out from the entire channel.
All data is available upon request.
DZ managed the experiments, analysed data, and prepared the paper with contributions from all the other co-authors. TD and JD co-organized the experiments. PV did the HEC-RAS simulation. PR was responsible for all the laboratory analyses. PS and AE participated on the experiments and the data evaluation.
The authors declare that they have no conflict of interest.
We thank our colleagues Josef Krása, Václav David, Petr Koudelka, Luděk Strouhal, Vladimír Říha and Dan Fiala for their great help during the experiments. This research was prepared within the framework of Czech Science Foundation postdoctoral project GP13-20388P, Ministry of Agriculture projects NAZV QJ1230056 and QJ1530181, and ÖAD WTZ Mobility project no. CZ18/2016 – 7AMB16AT002. Edited by: Patricia Saco Reviewed by: two anonymous referees