1 Introduction

The deposition processes in river channels are commonly affected by multiple forms of human intervention, such as local hydraulic measures, reduced sediment transport caused by upstream reservoirs and water diversions (Hohensinner et al., 2022). Diversion dams/weirs have been built on large rivers, as well as on many streams and smaller rivers, around the world, deflecting water for various reasons, such as to supply water to a hydropower plant, for irrigation and flood protection purposes or simply due to demands on the available water (Kuriqi et al., 2021). Designed for hydroelectric power generation, irrigation and/or navigation, the dams on large European rivers such as the Rhine, the Danube, the Durance and the Rhône divert most of a river’s water discharge into a bypass channel running parallel to the original river course, and the diverted water is returned to the main river course at some downstream point. In this way, the original channel is bypassed for several kilometres of flow length (Arnaud et al., 2015). At present, other mega-water diversion projects are being designed in China, India and South Africa. Such inter-basin water diversion projects greatly affect the environment and ecology of both water sources / rivers and the receiving basins (Yu et al., 2018).

Common to most diversion structures is the reduction of the flow in the sources/original rivers and also the reduction of the stream power and the related sediment-carrying capacity (Batalla et al., 2004; Brandt, 2000; Grams & Schmidt, 2002; Klingeman et al., 1998; Pal, 2016; Sammut & Erskine, 1995; Vauclin et al., 2019), resulting in changes in the bio-physical diversity and ecological values. For these reasons, it is challenging to investigate the complex adjustment responses of bypassed river channels to changing discharge regimes in a holistic way; understanding the morphological and sedimentary bases of their ecosystems is highly desirable. The biophysical analysis of four bypassed reaches of the French Upper Rhone performed by Klingeman et al. (1998) demonstrated the complex response of reaches to dewatering. Geomorphic and vegetation ecological responses to changes in the flow regime were also detected below diversion structures in nine steep, coarse-grained streams in the Colorado Basin above Glenwood Springs (Ryan, 1997). Baker et al. (2011) investigated the effects of variable levels of flow diversion on fine-sediment deposition, hydraulic conditions and geomorphic alteration in 13 mountain streams and concluded that there is a limit to the sediment-transport capacity of downstream channels. The channel narrowing and great reduction of aquatic biota habitats of the Durance River after a diversion were pointed out by Warner (2006). A comparison of channel landforms with and without flow diversions for eight hydropower schemes was performed along the meandering middle and lower reaches of the Aragón River (Spain) by Ibisate et al. (2013), who used several research methods, including field measurements, topographical measurements and hydrological analysis. Their results indicated that changes in the channel morphology and landforms occurred after land cover changes and became even more noticeable after a dam was built upstream from the channel. Caskey et al. (2014) stated that diversions can have a discernible effect on a channel’s geomorphic parameters and suggested that much of the shift in the riparian vegetation community’s composition occurs between 0.25 and 1 m above the channel. Vauclin et al. (2019) analysed 15 sediment cores from the Rhône floodplain, adjacent to bypassed reaches, and found a common pattern in the grain size sequences. The bottoms of the cores were characterised by coarse and well-sorted deposits. Then, the sediment abruptly became finer, poorly classified and uncommonly homogenous, which indicated changes in the hydrological regime of the reach after bypassing.

Commonly, the biophysical behaviour of bypassed channels is characterised, on the one hand, by a relatively steady and lower discharge compared to the original channel and, on the other hand, by the fact that the discharge rapidly increases due to the release of larger amounts of water during large flood events, with discharges exceeding the power plant’s or head-race canal’s capacity. A particular case of bypassed channel adjustment occurs when groyne fields have been designed and maintained before dewatering (Seignemartin et al., 2023). According to Sear et al. (2003) and Yossef and Vriend (2004), groynes are constructed in lowland rivers at an angle to the flow direction in order to deflect the flowing water away from critical zones. Studies of active groyne fields are focused mainly on sediment balance and behaviour or flow dynamics (e.g., Henning & Hentschel, 2013; Przedwojski, 1995; Savić et al., 2013; Sukhodolov et al., 2002; Tritthart et al., 2009; Yossef & Vriend, 2010). When a channel is bypassed, the groyne fields can no longer perform their original function, and due to the fact that they trap mainly suspended loads during large flood events, they are gradually changed into a new kind of landform (Przedwojski., 1995; Seignemartin et al., 2023). Sandy sediments with different accumulation properties are nowadays causing various problems in the river channel and surrounding floodplains considering the faster and more dynamic transport, but also the riparian succession (Habersack et al., 2013).

Therefore, vertical accretion causes the development of complex landform we addressed as groyne-induced benches (GIBs), which are progressively covered by different stages of vegetation, supporting sediment trapping (e.g., Belčáková et al., 2019; Corenblit et al., 2009; Gurnell et al., 2016; Solari et al., 2016) and influencing flood conveyance in bypassed channels (e.g., Carson, 2006; Yang et al., 2019). Groyne-induced benches are characterised by complete deposition (Sukhodolov et al., 2002), and they are considered the successors of the original groyne fields. Thus, the set of benches represents a new morphological pattern in bypassed channels. Only the works of Arnaud et al. (2015) and Räpple (2018) attempted to assess the behaviour of bypassed channels in a more complex, geomorphic-sedimentary and ecological way; these authors also made recommendations for the management of bypassed channels. Räpple (2018) pointed to the vertical accretion from 0.2 cm·year−1 to 10.8 cm·year−1 after diversion of the main flow in the Rhône River. Increased accretion of riverbank alluvium after bypassing was recorded in the Lower Rhône (Provansal et al., 2010) and Lehotsky et al. (2013) highlighted sedimentation rate of 13–16 cm per flood event detected form the stratigraphic analysis of the Danube bank upstream from Čuňovo dam near Bratislava between 1992 and 2011.

Our study is focused on the detailed investigation of the spatio-temporal variability of the vertical accretion thickness and volume caused by the response of the Danube River reach to bypassing. Five groyne-induced benches (GIB1 to GIB5) attached to the left (Slovak) bank of the bypassed Danube River channel (‘The Old Danube’) developed between rkm 1840 and 1824 after the diversion of the water in 1992; they represent our study area. We made the following hypotheses: (i) reach scale controls and the geometric characteristics of channel and groyne-induced benches can be considered important phenomena in accreted sediment thickness spatial variability; (ii) we hypothesize that the thickness of accreted sediments of individual benches is on average invariant both (spatially) and over time; (iii) the spatial variability of the vertical accretion thickness to the 100-year recurrence interval (RI) flood event is the same as the spatial variability of the sediment thickness accreted over 24 years to several flood events; and (iv) the land vegetation cover alters the vertical accretion thickness. In addition to the above, the other objectives of the article are the assessment of the spatiotemporal variability of the vertical accretion volume and the assessment of the channel’s conveyance before and after bypassing, as well as the presentation of some recommendations for bypassed channel management.

2 Study Area

2.1 General Situation

With a total length of 2,850 km and a drainage basin of 801,463 km2, the Danube River is the second largest fluvial system in Europe. One of the basic problems concerning the entire Danube River Basin is the modification of the longitudinal and lateral sediment continuity and the related regime. Especially over the last few decades, the sediment regime of the Danube River has been drastically changed. Between 1950 and 1980, 69 reservoirs, with an overall storage volume of about 7300 million m3, were constructed in the Danube River Basin. In total, 78 barriers exist along the Danube main river channel, leaving only five free-flowing sections. Moreover, the bed load deficit in the Danube has been strongly affected by the decline of the former bed load input from the main tributaries in the basin, where more than 700 large hydropower plants/weirs have been constructed. Therefore, the bed load input and output from the upper reaches of the Danube dramatically declined after 1960 (Habersack et al., 2013, 2016).

The study reach (Fig. 1a) lies in the sub-section type ‘the inland delta’ of the section type 4 (‘Lower Alpine foothills Danube’) of the Danube River classification (Sommerhäuser et al., 2003). In the past, it was characterised by an anastomosing planform with dominant substrates of medium to coarse gravels. The first projects that impacted the anastomosing pattern were carried out during the eighteenth century: dykes were locally constructed along the main, central meandering channel (Pišút, 2002). For navigation purposes, at the end of the nineteenth century and in the early years of the twentieth century, the channel, reinforced with rip-rap, was transformed into a 300-m-wide, single-thread, slightly sinuous waterway. The increased demands on water transport after World War II caused the construction of groyne fields for maintaining a suitable water depth at the fairway. However, the longitudinal and lateral hydrological-sedimentary connectivity of the anastomosing system was still maintained by ‘open banks (dykes)’ at confluence/diffluence nodes where water naturally ran from the main stem into side channels and vice versa. Such conditions, involving channel maintenance projects (gravel mining, groyne field maintenance), a mean annual discharge of 2025 m3·s−1 and a relatively well-functioning anastomosing system with hydrological and sedimentological connectivity, existed up until 1992 (Illyová & Matečný, 2014).

Figure 1
figure 1

Study area and Danube River reach near Gabčíkovo Waterworks: a study area location; b the regional position of the bypassed Danube with marked location of the prospected landforms; c preserved original part of the anastomosing river channel; d evolution of the groyne-induced bench GIB3

Full size image

2.2 Bypassed Danube Channel

After 1992, when the Gabčíkovo Waterworks was launched into service (Fig. 1b), this anastomosing system was transformed into a system consisting of the Čunovo diversion weir (1851.7 rkm), the 29-km-long head race canal ending by the Gabčíkovo weir (1821 rkm), the 10-km-long trail-race canal and the original Danube channel, which is a 40-km-long bypassed channel (‘The Old Danube’). The Čunovo diversion weir upstream from this channel modified its flow regime by decreasing the mean annual discharge from 2000 m3·s−1 to about 400 m3·s−1 (during the vegetation period, depending on the hydrological conditions, it should fluctuate from 400 to 600 m3·s−1; in the non-vegetation period, the discharge should not be less than 250 m3·s−1), lowering the water level by about 3–4 m, narrowing the channel to 150–200 m (Lehotský et al., 2010) and causing a loss of hydrological/sedimentary connectivity with the side, floodplain channels (Fig. 1c). Thus, the bypassed channel itself is confined by margins of anthropogenic origin – constructed dykes – that constrain its lateral adjustment.

As for the inter-dyke landform pattern, the Old Danube, after dewatering and narrowing, is characterised by 15 alternate groyne-induced benches developed at localities where groyne fields were constructed and maintained before 1992 (Fig. 1d). The averaged longitudinal friction slope of the study reaches, approximated from both the bed slope (Topoľská & Kľúčovská, 1995) is 0.003 ‰. Its longitudinal profile with regard to erosion/accumulation processes is characterised as quasi-stable (Holubová & Capeková, 2006). As for hydrological characteristics, from 1992 to2013 were recorded 11 large flood events in the bypassed channel at the Dobrohošť gauging station in the years: 1997–2013 7(Table 1) Three of these floods (two in 2002 and one in 2013) are classified as 100-year RI events. According to the Handling Instructions for Water Structure System Gabčíkovo-Nagymaros, such events occur only when the discharge at the upstream Bratislava-Devín gauging station reaches 5700 m3·s−1.

Table 1 Highest water levels on Dobrohošť gauging station on Danube in 1992–2013
Full size table

2.3 Settings of Groyne-Induced Benches and Their Geomorphic Units

In order to assess the spatial variability of the vertical accretion thickness, five groyne-induced benches (GIB) were analysed in this study (marked from GIB1 to GIB5 and position located on the Fig. 1b). Groyne-induced benches have different basic morphometric parameters and positions along the Danube, and the geometric properties and settings of the GIBs are presented in Table 2. Two benches (GIB1 and GIB4) are located at straight channel positions, while the other three (GIB2, GIB3 and GIB5) are located on bends of the river channel.

Table 2 Number of drilling probes and cross-sections at groyne-induced benches 1–5
Full size table

3 Materials and Methods

3.1 Field Mapping of Sediment Depth

The allostratigraphic approach was applied for the analysis of 548 drilling probes (Table 3) throughout five GIBs during field campaigns in 2014 to determine the vertical accretion thickness. All probes were spatially localised with the Global Positioning System device Trimble model 3, which was equipped with transmitters using the Slovak real-time positioning service (SKPOS) and the coordinate reference system S-JTSK / Křovak East North (EPSG: 5514). The maximum horizontal error for the points surveyed in the forested area was 50 cm, and it was 3 cm for open land. The probes were organised into cross-sections perpendicular to the original dyke of the main channel; this allowed us to identify the depth of the unconformity between the gravel layer of the former channel bottom and the total vertical accretion, as well as the depth of the unconformity between the uppermost sandy layer, which represents the vertical accretion caused by the last large flood (June 2013; Fig. 2a) and is easily distinguishable, due to its distinct colour (10 YR 6 / 2–3 to 10 YR 7/2–3, Munsell Soil Color Charts, 2000), and the sandy layer below that represents vertical accretion before 2013 and after 1992.

Table 3 Values of vertical accretion thickness for groyne-induced benches (GIBs 1–5) and their geomorphic units (BCH – backchannel, SP – supra-platform, T – tail, H – head), and their heights above the mean water level (300 cm) in the active channel
Full size table
Figure 2
figure 2

Photographs taken during field research in 2014 on groyne-induced bench 1 demonstrating the different geomorphic units: a sedimentary accumulation over gravel bed with visible layer from the flood 2013; b growing accumulation on the toptop part geomorphic unit; c supra-platform; d backchannel scour; e erosion of the part of GIB1 after flood in 2013

Full size image

3.2 Geomorphic Unit Mapping

The morphology of the river was mapped based on topography of the GIB. The elevation surfaces of the GIBs were identified from a height above river channel model, which was a modified version of the height above the nearest drainage (HAND) model presented by Nobre et al. (2011). This modified topographic model detrended channel slope and represent the topography of the benches based to their relative position above the water level during imaging. The model is calculated from the LiDAR point cloud with resolution 50 cm. It consists of two sets of procedures. First, we identify the longitudinal water-level elevation from the contact edge of the water and the GIBs from the LiDAR DEM. The second step consist of GIBs’ topography recalculation based on the differences between the normal elevation of the GIBs and the elevation of this closest water level edge elevation detected based on the Euclidean distance.

Finally, this relative model was classified based on the analysis of the frequency curve to six classes: (i) 0–2.30 m, (ii) 2.301–2.90 m, (iii) 2.901–3.30 m, (iv) 3.301–4.30 m, (v) 4.301–5.0 m and (vi) > 5.0 m. The discrimination of the elevation surfaces of the GIBs, along with the positions of groynes in each individual GIB, allowed us to identify and vectorise the following four geomorphic units in each GIB in the sense of Bluck (1971) and Reid et al. (2019): (i) the top (H), which is the only unit comprising the original groynes; (ii) supra-platform (SP), (iii) tail (T) and (iv) backchannel (BCH) units (Fig. 2b–e).

3.3 Processing of LiDAR Data

The topography of the groyne-induced benches (GIBs) was determined from a LiDAR point cloud dataset. Airborne LiDAR data were obtained with a Leica ALS70 – HP scanner by the National Forest Centre (Slovakia) in December 2014 during low vegetation cover. The density of the point cloud was 3 points/m2 in multiple, partially overlapping fly strips. Although the vast majority of the points were reflected directly from the surface, further filtering was required. The LiDAR dataset from 2017 were supplied by the Geodesy, Cartography and Cadastre Authority of the Slovak Republic in the form of a LAZ point cloud. This data was captured in December 2017, reaching a vertical accuracy of 0.05 m and an average density for the last reflective points of 17 points/m2. Both LiDAR point clouds was processed usingthe LAStool software. After merging (lasmerge) the scanned strips, lasclip was used to clip the points to the study area around groyne-induced benches, and the lasground module (terrain type: Forest or Hills and granularity value: Fine) was used for the classification of ground points. Finally, LAS dataset to raster was applied for the interpolation of the DEM using the natural neighbour triangulation method with 1 m resolution (Fig. 3).

Figure 3
figure 3

Digital elevation models of groyne-induced benches (1–5) and their geomorphic units (H – headtop, SP – supra-platform, T – tail, BCH – backchannel) based on LiDAR data

Full size image

3.4 Identification of Vertical Accretion Thickness

In order to determine the spatial variability of accreted sediments, the depth of sandy deposition at sampling points was analysed at three spatial scales: scales of individual benches comparing GIB1 to GIB5, scale of geomorphic units, present on each bench: toptop, supra-platform, tail and backchannel, and scale of individual benches relative to their geomorphic units. The set of values of the accretion thickness, for benches and geomorphic units, has been expressed using median values.

The temporal sediment thickness variability was assessed during three accretion periods (Fig. 3), namely, from 1992 to 2012 (20 years); period effected by floods events in 2013 and post flood accretion between 2014 and 2017 (3 years). Vertical accretion was assessed based on the detection of four surfaces on every bench. First is gravel bed bottom emerged after dewatering in 1992. Second surface is sedimentation discontinuity of sandy layer from 2013-flood. Last two surfaces are ground classified point cloud from 2014 and 2017. Vertical accretion depth was for preFlood and Flood period identified form drilling probe at the sampling points. Accretion depth for postFlood period was detected from differences of elevation models between 2017 and 2014 at the same sampling points. In this case, only differences largest than 7 cm was considered as effect of deposition and smaller changes was omitted and attributed to inaccuracy as an effect of uncertainty (δu) in the comparing of elevation changes between these topographic surfaces. Minimum detection level of 7 cm was calculated from vertical errors of LiDAR datasets.

The statistically significant difference between median value sets of the deposition thickness in benches as well as geomorphic units (Fig. 4) was examined using the Kruskal–Wallis H test. If the p-value is less than the significance level 0.05, we conclude that there are significant differences between the sets of median values of sediment depths, i.e., sets of five benches and sets of four geomorphic units. Moreover, a Shapiro–Wilk test was performed to calculate normality and data was processed in RStudio software (package PerformanceAnalytics).

Figure 4
figure 4

Workflow for identification of three main accretion period on groyne-induced benches (GIBs) of Old Danube channel (a); process of detection four main surfaces from field drilling probes (interpolated surface from sampling points) and LiDAR datasets topography model of ground (b); and final identification of accretion period before and after flood events in 2013 (c)

Full size image

3.5 Identification of Vertical Accretion Volume

Accretion of sediment deposition volume was calculated for the period from 1992 to 2012, period of the flood event that occurred in 2013 and period from 2014 to 2017. First two volume estimation represent two unconformities, i.e., the surface of the sediment layers deposited during the flood in 2013 and the surface of the sediment layers deposited before the 2013 flood. This volume was estimated from interpolation of field drilling probe at sampling points. Third volume estimation for period from 2014 to 2017 was calculated based on DEM of differences (DoD) from LiDAR topography models.

3.5.1 Interpolation of Sediment Unconformity

The volume estimation has been calculated based on values of the total sediment thickness and the thickness of the 2013 flood sediment recorded at the sampling points. These values were interpolated using a regularised spline with tension (RST; GRASS GIS version 7.2.1, v.surf.rst) with the combination of two parameters (smooth and tension). A configurable Python script, executable in GRASS GIS (in a Python editor), was designed to run a sequence of modules to interpolate the sediment thickness, deviations and cross-validation for each of the measured attributes. This script combines the modules in GRASS GIS for importing the vector layer (v.in.ogr), setting the region and mask (g.region, r.mask), interpolating sediment depths, calculating deviations and cross-validations (v.surf.rst) and exporting the raster (r.out.gdal) in csv format, in our case (function incorporated into the script).

Every attribute of each individual groyne / induced bench was interpolated with all possible combinations of values from the smooth and tension ranges:

  • Smooth range: bottom boundary equal to 0.1, top boundary equal to 2, step equal to 0.1;

  • Tension range: bottom boundary equal to 20, top boundary equal to 80, step equal to 2.5.

The selection of the most suitable parameters was managed in two phases/iterations. The script iterated through all vector layers (first iteration) and calculated basic statistical parameters (maximum, minimum, mean absolute error [MAE], median absolute deviation [MAD], standard deviation [SD], root-mean-square error [RMSE], interquartile range [IQR]) using the deviations and cross-validations of the sediment thicknesses for period preFlood and Flood2013.

The chosen models (approximately 100 models) were visualised in ArcGIS (v. 10.3.1), considering the best statistical accuracy (second iteration); the most relevant models were selected (Supplementary Material 1). Their RST outputs were used as inputs to the r.volume GRASS module that calculated the total sediment volume and the volume of sediment deposited by the flood in 2013. Thus, the creation of three DEMs allowed us to calculate accreted sediment volumes for period 1992 – 2012 and flood event in 2013.

3.5.2 Calculation of Volumetric Changes from DoD


The differences between ground elevation model from LiDAR 2014 and 2017 was used for detection sediment volume deposited between 2014 and 2017. For calculation we subtracted elevation values from topography model in the raster calculator in ArcGIS software. Only ground point clod was used for generation bench surface topography. Pixel resolution for the final model was 1 m and for uncertainty analyses we evaluate Level of Detection (minLoD) for entire DoD form verticals errors of LiDAR scan (Brasington et al., 2003; Wheaton et al., 2009):

$$\delta u = \sqrt {\left( {\delta z_{2014}^{2} + \delta z_{2017}^{2} } \right)}$$

where propagated uncertainty (δu) in the topography model comparing elevation changes (z) for old (2014) and new (2017) topographic surface.

In order to calculate the LoD, the standard deviation error SDE was combined for individual LiDAR datasets. Ultimately, the propagated uncertainty for the DoD calculation was 0.069 m, and the minLoD detection threshold was set to 7 cm. Therefore, all areas below this threshold were excluded from the DoD analyses.

3.6 Effect of Vegetation to Deposition and Conveyance Capacity of Groyne-Induced Benches

In order to assess the influence of the vegetation cover on the thickness of the accretion caused by the large flood in 2013, four types of vegetation cover were defined: (i) herbs (H), (ii) scarce shrubs < 3 m with a herb layer (SS), (iii) dense shrubs < 3 m (DS) and (iv) adult trees > 3m without shrubs and herb layers. Vegetation cover was surveyed according to general physiognomy (Corenblit et al., 2009) at all sampling points during field work in 2014 and also vectorised using a 2014 orthophotomap in ArcGIS (v. 10.3.1).


The appraisal of the flood conveyance capacity of the individual GIBs was performed for four-time horizons: (i) in 1996, that represent state after water diversion; (ii) in 2003, after the 2002 flood and at the first stage of vegetation succession; (iii) in 2011, before the 2013 flood; and (iv) in 2014, after the 2013 flood. The cross-section bankfull discharge and the mean value of the roughness zones, expressed by the Manning coefficient (n), were calculated using Winxspro software and following Arcement and Schneider (1989). On the cross-section, land cover categories were identified and ordinal values indicating the floodplain roughness were assigned to land cover types as follows: channel, bare surface (very low roughness), herbs (low roughness), scarce shrubs (medium roughness), dense shrubs (high roughness), adult trees (high roughness) and cleared forest (medium roughness). For each roughness category, there were associated Manning’s coefficient (n) values: channel (0.03), very low roughness (0.03), low roughness (0.035), medium roughness (0.07) and high roughness (0.12). The flood conveyance capacity was calculated according to the widely used Manning equation:

$$Q = \frac{{AR^{2/3} }}{n} \cdot S_{0}^{1/2} ,$$
(1)

where Q is the flow conveyance, and A and R are the cross-sectional area and hydraulic radius, respectively; they both depend on the flow depth. S0 is the river channel bed slope, and n is the Manning’s roughness coefficient. The bankfull discharge has been calculated for each cross-section separately.

4 Results

4.1 Spatio-Temporal Variability of Accretion Thickness

The accretion thickness during the study period on individual benches (Fig. 5a) express the similar trend and overall deposition from 88.6 cm on GIB3 to 107.2 on GIB4. Prevailing deposition in first preFlood period from 1992 to 2012 approximately in a ratio of 2:1:1 with comparison to Flood-2013 and postFlood deposition (2014–2017). The response of sediment accretion to the large flood event in 2013 is somewhat similar to the trend of the previous period of accretion. On GIB4 and GIB5, 20 cm of deposited sediments were detected. However, on GIB3, lower values around zero were recorded; the maximum thickness of accretion is 20 cm. Lower values indicate the sparse spreading of the 2013 flood sediment layer across the bench. The Kruskal–Wallis test indicates that GIB1 significantly differs from GIB2, GIB4 and GIB5.

Figure 5
figure 5

Vertical accretion thickness for the 1992–2012, 2013, 2014–2017 a whole study period for groyne-induced benches 1–5 (a) and their geomorphic units (H – top, SP – supra-platform, T – tail, BCH – backchannel) (b)

Full size image

The vertical accretion calculated for geomorphic units in whole study period Fig. 5b) is highest on the toptop geomorphic unit with median of 122 cm,, followed by the tail (91 cm) and the supra-platform (86 cm) units. As far as the thickness of accretion values recorded for geomorphic units for the flood event in 2013, the highest values belong to H (18 cm), followed by BCH (12 cm). The lowest values are calculated for T (6 cm), which is followed by SP (8 cm). The vertical accretion for pre and post flood period on geomorphic units across all benches is the same as the long-term trend.

4.2 Accretion Thickness Rate

The average accretion rate for individual benches during whole study period range from 3.8 cm.year−1 for GIB3 to 5.3 cm.year−1 on GIB4. Overall accretion rate is 4.2 cm.year−1 for study region of the Danube Old channel and temporally change from 3.1 cm.year−1 in first period, with flood-2013deposition 13.9 cm.year−1 and 8.1 cm.year−1 in the last time horizon. Flood event in 2013 reach highest intensity of deposition on the bench top (22.4 cm.year−1) and backchannel (14.9 cm.year−1) and lower accretion rate in the central part with rate 10 cm.year−1 on supra-platform and 9.8 cm.year−1 on tail (Fig. 6).

Figure 6
figure 6

Accretion rate for the 1992–2012, 2013, 2014–2017 a whole study period for groyne-induced benches 1–5 (a) and their geomorphic units (H – top, SP – supra-platform, T – tail, BCH – backchannel) (b)

Full size image

4.3 Spatial Distribution of Vertical Accretion on Individual Benches

To understand the variability of the spatial accretion thickness on individual benches, we used trendline-joining median values in the downstream direction represented by geomorphic units arranged from upstream to downstream. X-axis shows the succession ‘H-SP-T-BCH’, reflecting their longitudinal variation. Thus, their shapes allowed us to assess the commonalities/dissimilarities in the spatial variability of the accretion thickness on individual benches for different periods, as well as to determine the variability of individual bench deposition and assess the evolutionary trend of the accretion. Further characteristics of the GIB1-5 considering their area and groyne fields are in Table 4.

Table 4 Detailed characteristics of groyne-induced benches: ratio of bench area to geomorphic unit area, classification of benches using a criterion value of this ratio of more than 20%, number of groynes and groyne fields, groyne lengths and angles of groynes to the flow line (GIB – groyne-induced bench, BCH – backchannel, SP – supra-platform, T – tail, H – head)
Full size table

The total accretion on GIB1 (Fig. 7d) exhibits the highest thickness of deposits in H and SP and so it is called the toptop-supra-platform bench; it is characterised by the parabolic shape of its trendline and sediment deposition on the front edge of the GIB and backchannel parts. A similar spatial trend (parabolic trendline of toptop-supra-platform) is observed for GIB5; however, the difference between GIB1 and GIB5 consists in the larger values of the H (highest accretion in the study period: 144 cm) and T geomorphic units in GIB5, which has approximately the same values for the SP and BCH units. GIB2 can be considered a tail-toptop (T-H) bench, with the highest values of vertical accretion in the tail and top geomorphic units (140 cm and 104 cm, respectively). The shape of the trendline has inflexion points. The benches GIB3 and GIB4 are top-supra-platform (H-SP) benches, with the highest values of vertical accretion in the H and SP geomorphic units (GIB4: 150 cm and 105 cm, respectively; GIB3: 99 cmand 86 cm). The vertical accretion decreases in the downstream direction and differs in the amount of sediment deposition. Overall, the linear trendlines of GIB3 and GIB4 indicate a different method of accretion in comparison to the parabolic (GIB1 and GIB5) and two-headed curve (GIB2) trendlines.

Figure 7
figure 7

Vertical accretion thickness during the 1992–2012 (a), 2013 (b), 2014–2017 (c) a whole study period (d) for the groyne-induced benches and their geomorphic units, with a trendline of the accretion for each of the groyne-induced benches 1–5 (H – top, SP – supra-platform, T – tail, BCH – backchannel)

Full size image

4.4 Spatio-Temporal Variability of Vertical Accretion Volume

A total of 1,146,589 m3 of accreted sediments was deposited on all benches. Almost a half of this total amount specifically calculated to 669,137 m3, was accreted during first period (1992–2012). During flood event in 2013 was deposited 209,752 m3 and 267,700 m3 in the last study period. Overall, an accretion rate of 47,775 m3·year−1 could be calculated for the period of 1992–2017.

It is logically found in the largest supra-platform geomorphic units; conversely, the lowest volume is recorded in the smallest tail units. On the supra-platform units, overall, 511,578 m3 of sediments was accreted and of which 93,152 m3 during flood in 2013. In total 100,271 m3 was accreted on tail units and, 12,762 m3 represented the 2013 flood and 59,405 m3 was accreted during the period from 1992 to 2013. However, generally the relationship between the areas and volumes of accretion is evident. The highest total volume of sediments, as well as the highest volume deposited during the period of the 2013 flood accretion volume was deposited on GIB5 with overall volume 292,746 m3 and on GIB4 of 208,461 m3. The lowest volume of sediments was deposited on GIB1, and it was 163,495 m3, and the second-smallest volume was exhibited by GIB3 (167,629 m3). In terms of the area, GIB3 is the smallest of the GIB landforms.

4.5 Vegetation Cover and 2013 Flood Accretion Thickness

The sediment accretion thickness caused by the 2013 flood in relation to the four groups of vegetation cover was assessed for individual benches and for each geomorphic unit. Generally, the lowest values (6 cm) of the accretion thickness were exhibited by adult trees; conversely, the highest (16.5 cm) values were exhibited by dense shrubs, followed by scarce shrubs (13 cm) and herbs (10 cm). The maximum thickness of accretion under adult trees was recorded at the top geomorphic unit (26 cm), and the minimum thickness was recorded at the SP and T units (Fig. 8a). An increased thickness value (7 cm) is also exhibited by the BCH unit. Accretion under dense shrubs dominates in the SP (19 cm) and BCH (22 cm) units, and accretion under scarce shrubs dominates in the T unit.

Figure 8
figure 8

Effect of vegetation cover on the vertical accretion caused by the 100-year flood event in 2013 within groyne-induced benches 1–5 and different geomorphic units (H – top, SP – supra-platform, T – tail, BCH – backchannel)

Full size image

It should also be noted that the second highest values of accretion in the top unit were measured for the herb vegetation group. As far as benches (Fig. 8b), despite the fact that the deposition under adult trees is in general the lowest, at GIB1, the values of such accretion are higher than the values for all other vegetation groups (40 cm). At other benches, the highest values of accretion are indicated under dense shrubs, and the lowest under herbs and scarce shrubs. GIB3 is an exception because of the forest clearing in 2012, which thus led to negligible accretion during the 2013 flood.

4.6 Changes in Flood Conveyance Capacity

Based on DEMs compiled for the GIBs’ original gravel-bed surfaces, the surfaces of the sediment discontinuity represented by sediment layers before the deposition caused by the 2013 flood and also contemporary surfaces (2014) were used to apply divided channel approaches to the cross-sectional channel, which was divided into more than one flow zone. Thus, we could estimate the conveyance capacity for five GIB cross-sections (Fig. 9). Comparing the 1996 and 2011 cross-section conveyances, all GIBs exhibit an evident drop. On the one hand, bare gravel surfaces and the negligible succession of vegetation, following water diversion in 1992, conditioned a relatively high capacity of conveyance in 1996. On the other hand, ten large flood events accompanied by sediment accretion and the development of shrubs and trees supported the evident decrease in the conveyance in 2011 (Fig. 10). However, if the conveyances in 2011 and 2014 are compared, three different modes of behaviour can be observed. The GIB1 conveyance also decreases in 2014, but the GIB4 conveyance seems to be about the same as the GIB4 conveyance in 2011.

Figure 9
figure 9

Channel conveyance capacity calculated for cross-sections of groyne-induced benches 1–5 over time; it exhibits an overall decreasing trend after channel bypassing. Changes in vegetation cover on the GIB3 and GIB5 cross-sections due to their striping during the 2013 flood event, as well as shrub and forest clearing, caused the conveyance capacity values to increase

Full size image
Figure 10
figure 10

Changes in the ‘bare surface’, ‘herbs’, ‘scarce shrubs’, ‘dense shrubs’, ‘adult trees’ and ‘cleared forest stands’ land cover categories over time calculated for all five groyne-induced benches. A decrease in the area covered by dense shrubs is accompanied by an increase in the area covered by adult trees starting in 2000 and a decrease in the area covered by adult trees after harvesting. After 2015, decreases in the areas covered by cleared forest and scarce shrubs are accompanied by an increase in the area covered by dense shrubs

Full size image

5 Discussion

5.1 Channel Modification and Sedimentation on Groyne Fields After Water Diversion

Human manipulations of river systems often change and arrest evolutionary trajectories (James, 2017). Therefore, the construction of artificial levees and groynes in the past, damming, water diversion and regulated discharge (since 1992), as structural changes to the Old Danube, induced its arrested trajectory, causing changes in the channel geometry and the lateral and longitudinal hydrological and sedimentological disconnectivity (Gautier et al., 2009). The Čunovo weir upstream represents a barrier (sensu Brierley et al., 2006) where sediments are trapped.

As far as the channel geometry, the Old Danube channel’s sinuosity after diversion slightly increased from 1.03 to 1.05, and the erosion of the artificial levee is occurring. This is contrary to the findings of Caskey et al. (2014), who assumed that reduced sinuosity in diverted streams may reflect lower local bank erosion and the preservation of meander bends. Horáčková et al. (2017) found that the dewatering and narrowing of the Old Danube channel caused 57.66% of the original channel to be transformed into groyne-induced benches covered by terrestrial vegetation; the settings of these benches play an important role in the control of the spatial variability of vertical accretion and sediment connectivity.

The benches’ settings and channel connectivity with side arms seem to be very important factors influencing the spatial variability of sediment deposition in the GIBs. Tthe lowest thickness of sediments are observed in upstream position at the straight section of the channel (width/depth ratio = 35.6) near Čuňovo dam. There we can detect absence of receiving additional suspended load input from side arms or from upstream (above dam) in-channel bars or benches.

Position of the groyne fields can significantly modified sedimentation pattern (GIB3). and can be explained by the proximity of the very well-developed alternate bench upstream (Fig. 11). It is characterised by (i) a bench wavelength of ca. 7 times the channel width, which is approximately the same as that suggested by Lewin (1976); (ii) a relatively low radius of curvature (2298 m); (iii) the confluence of the main channel with a side arm on the opposite side of the channel and proximity of well-developed system of side channel on the left side floodplain; and (iv) a width-depth ratio of 24.1. Thus, all these circumstances condition the higher input and deposition of the suspended load and the slight migration of the bench downstream, which are manifested in both the maximal portion of the tail area (27.5%, Table 4) of GIB3 and the sediment accumulation from 1992–2012, where sediment thickness was 56 cm.

Figure 11
figure 11

Scheme of the settings of groyne-induced benches and stages of Danube channel development: a groyne-induced bench 1 at the straight part of the channel; b groyne-induced bench 2, which has no side-channel connection, causing the lowering of the flow velocity and increasing sediment accumulation on the largest groyne-induced bench; c the bypassed channel and side-arm connection, with an alternate bench upstream that likely influences the sediment input of the back-channel geomorphic unit of GIB3; d groyne-induced bench 4 at the straight part of the channel; e hydrological and sedimentation connectivity of the bypassed channel with the side arm during large flood events on groyne-induced bench 5. f The main stages of groyne-induced bench development: A the first stage, which is characterized by free sedimentary transport/input before groyne field construction and channel bypassing; B groyne field construction for the adjustment of fairway and dyke construction along the former main channel; C vertical accretion on groyne fields after channel bypassing, the lowering of the water level and the development of groyne-induced benches; D vegetation succession on the new surfaces of groyne-induced benches, causing an increase in channel roughness and lowering the conveyance capacity

Full size image

The position of GIB4 at the straight section of the channel (width-depth ratio = 38) can cause the input of more sediment from upstream during large floods, as well as from the side-arm system; this is manifested in the high sediment thickness of the top geomorphic unit. The position of GIB4 on the bend with the lowest radius of curvature 1677 m Table 2), the smallest width of the channel and the maximum channel depth (10 m, width-depth ratio = 13) caused this development (the highest sediment thickness values of the backwater geomorphic unit).

GIB2 represents the groyne-induced bench characterised by the largest number of groynes (10) organised into two groups at a distance of ca. 1100 m. These circumstances caused GIB2 to have the greatest length (2889 m), the largest proportional BCH area (26.93%) and the lowest proportional SP area (19.35%). This is the only bench with the highest sediment thickness values of the tail geomorphic unit (GU) for all time spans.

5.2 Flood and Vegetation Impact on Sedimentation

The spatial variation of the sediment thickness is variable for individual GIB; however, the evolution of this thickness during individual time spans is the same. Thus, if the spatial variation of deposition during the flood event in 2013 is similar to the total accretion and the accretion from 1992 to 2012, major flood events (>10-year floods) exclusively dictate the general mode of the accretion thickness for individual benches and their GUs. It is worth mentioning that the uniqueness of the 2013 flood event with 100-year reoccurrence interval (Slovak Hydrometeorological Institute, 2013) is demonstrated by the fact that its values of the vertical accretion thickness are high on the BCH geomorphic units, despite the fact that these units show the lowest total accretion thicknesses. Spatial sedimentary pattern differs from lower-magnitude flood events to large flood events. Also, we expected that sedimentation in the last period 2014–2017 is result of the higher water level at Dobrohošť gauging station in beginning of September 2017. There, 1st flood alert was exceeded, but higher water level in the Old Danube channel was result of a rise in the water level on the Danube from precipitation associated with a cold front over the Alps, and increasing of discharge in Old Danube channel was effect of the manipulation of the VDG to lower the level in the inlet channel. Result of this manipulation was flushing of the sediments from Čuňovo dam.

The analysis of the vertical accretion thickness for the 2013 flood event, the period from 1992 to 2012 and period 2014–2017 does not decrease with the heigher topography of benches, and the thickness of accretion on benches and their GUs varied, likely due to the variability of the vegetation cover: shrubs are considered to trap sediment the most effectively (e.g., Corenblit et al., 2009; Arnaud et al., 2015). Vegetation-induced landform accretion results in changes in the composition of the vegetation communities, while the reduced frequency and magnitude of flooding resulting from landform accretion, combined with successional changes in the vegetation, alters the vegetation’s contribution to sedimentation (Bendix & Hupp, 2000; Corenblit et al., 2007).

The rate of the vertical accretion of the studied groyne-induced benches ranged from 3.5 cm·year−1 to 5.8 cm·year−1 or 13.6 cm·year−1 for specific flood event. For comparison, Räpple (2018) estimated average annual sedimentation rates that ranged from 0.2 cm·year−1 to 10.8 cm·year−1 on the post-diversion surfaces of groyne fields in the Rhône River. Provansal et al. (2010) found that due to an increase in the frequency of major floods and changes in the annual rate of vertical accretion in the Lower Rhône, the accretion of alluvial banks changed from 4 cm for the period from 1950 to 1972 to 6.5 cm from 1972 to 1993, finally reaching 23 cm from 1993 to 2003. Dufour et al. (2007) estimated that the annual sedimentation rates in the embanked reach along the River Drôme varied from 0.2 cm·year−1 to 2.8 cm·year−1, compared to rates of 0.2–10.1 cm·year−1 in the unconstrained reach. In terms of the rate of vertical accretion on benches during an individual flood event, Lehotský et al. (2010) estimated a rate of 13–16 cm per flood event based on the stratigraphic analysis of a groyne-induced bench and discharge records in the Old Danube near the Bratislava from 1992 to 2011 located above Čuňovo dam. Grams and Schmidt (2002) found that the 1999 post-diversion flood deposition on the intermediate bench on the Green River typically occurred in discrete patches, and individual deposits varied from traces to deposits 50 cm in depth. The 2013 flood vertical accretion thickness in the Old Danube also varies from 1–2 cm up to 39 cm on the top geomorphic unit of GIB5. Wood and Ziegler (2008) found that the sediment thickness of a floodplain after a 100-year flood event on the Ping River (northern Thailand) was 3.3 cm on average (sediment thicker than 8 cm was found only in 16% of the area), which is smaller than the thicknesses we found. In Canada, Smith and Pérez-Arlucea (2008) measured sediment thicknesses of 0–70 cm per large flood event in 2005 on the Saskatchewan River, which has an in-channel levee flooding period of 7 weeks. According to this study, all new flood sediment was created by channel scour. There are also the compensating effects of erosion and deposition, which occurred throughout the flood event; these effects are difficult to separate because erosion occurring during the peak of the flood discharge can be masked by the deposition that occurs with the falling limb of the flood (Croke et al., 2013).

5.3 Future Development and Management

Accretion thickness rates are related to the flood pulse of Danube cannel. Therefore, only more extreme/pulse events can deliver sediments downstream into the Old Danube channel as well as into the right (Hungarian) side-arm system and thus stimulate more intensive ‘side-arm–main channel’ interactions. According to the Handling Instructions for Water Structure System Gabčíkovo-Nagymaros, such events occur only when the discharge at the gauging station Bratislava-Devín reaches 5700 m3·s−1. However, the Čunovo reservoir is not always the main source of the suspended sediment downstream. Based on the comparison of the suspended sediment load of the August 2002 flood between the Bratislava (14,600 kg·s−1) and the Medveďov (3371 kg·s−1; downstream from the confluence of the Old Danube with the trail-race canal) gauging stations, most of the suspended sediments were trapped by the Čunovo weir, as well as by the Old Danube benches and the side-arm system (Borodajkevyčová, 2015). On the contrary, the suspended sediment load of the June 2013 flood was 4594 kg·s−1 at Bratislava, which is lower than that at Medveďov (5944 kg·s−1). This fact indicates that some amount of the suspended sediments deposited in the Čunovo weir have been washed downstream to the Old Danube channel and into the side-arm system. Habersack et al. (2016) also pointed out that the large flood event in 2013 on the Danube caused the remobilisation of sediments and flushed over 5 million m3 of fine material from the Aschach hydropower reservoir (located upstream in Austria); this was followed by sedimentation in inundated areas downstream. Therefore, the sources of the sediments on the Old Danube benches likely vary with individual flood events.

Based on analyses of Lehotský et al., (2010, 2013) it can be assumed that the sediment grain size varies from around 1.8 Phi during 50–100-yr RI floods to around 3.3 Phi during 5–50-yr RI floods. However, not only the magnitude of floods but also the source of the suspended load, the frequency of flood events and local hydraulic conditions influence the sediment particle size and its spatial variability (Fig. 12). Regarding bypassed channel management, the Old Danube and the adjacent side-arm system belong to a Dunajské Luhy protected area, so the development and occurrence of groyne-induced benches in the Old Danube channel have caused the establishment of new biotopes and an increase in the protected.. Therefore, on the one hand, there is nature protection; on the other hand, as mentioned above, non-organised forest clearing is practised throughout benches because some volume of the discharge of large floods is usually diverted into the Old Danube channel, which operates as an important floodway. Hence, forest clearing is considered a flood protection measure for the town of Bratislava upstream. We suggest that forest clearing should be used to increase the flood conveyance capacity, but it should not be performed throughout the whole area of the groyne-induced benches; instead, it should only be performed in stripes along benches and possibly along backchannel geomorphic units. Backchannels represent ‘cut-offs’ along benches and are characterised by a relatively low position above the channel water level; they are high-velocity floodways, which is also proved by the low thickness of accreted sediments. They offer the most suitable space for forest clearing and flood water conveyance. Due to the fact that banks (artificial levees) opposite groyne-induced benches are subjected to undercutting, due attention, encompassing serious investigation and monitoring, should be given to this issue. Recently, ideas concerning re-establishing the hydrological connectivity between almost the whole length of the bypassed channel and the adjacent side-arm system have emerged, and several variants of solutions for this issue have been proposed. However, final conclusions and agreement between the Slovak and Hungarian sides have not yet been reached. To the future we propose further research of several LiDAR data updates to record upcoming large flood events on Danube to compare the sedimentary impact with our research, as from 2013 no such an event occurred. With availability of such a high-accuracy data, the research results could be more precise.

Figure 12
figure 12

Scheme showing the adjustment of the groyne fields (GFs) to channel bypassing, resulting in the development of groyne-induced benches (GIBs). Components influencing sediment input and output are marked as ( +) or ( – ), respectively

Full size image

6 Conclusion

Petts (1984) had already pointed out some time ago that the consideration of human activity within the environment, using three orders of impact, provides a basic framework for the appreciation and evaluation of long-term problems. Consequently, upon dam construction, major changes in the flood magnitude and frequency and the quantity and calibre of sediment loads (first-order impacts) will induce the readjustment of the channel morphology and ecology (second-order impacts) and the readjustment of the macrophyte and macro-invertebrate populations (third-order impacts).

This article deals with the first-order impacts on the original channel, which contained a system of groyne fields, and its adjustment when upstream weir construction caused the discharge and water level to decrease and then caused channel narrowing and the development of alternate groyne-induced benches. The availability of a remote sensing data set for several time periods and 548 drilling probes of a 16.3-km-long reach of the bypassed Danube allowed the investigation of the morphology and vertical accretion dynamics of five groyne-induced benches.

Overall, during the period 1992–2014 on the studied GIBs was deposited 1,146,589 of fine-grained sediments and 209,752 m3 of sediments during 2013 flood events with a reoccurrence interval of 100-years. An accretion rate of 47,775 m3·year−1 can be estimated for the period from 1992 to 2017. The sediment accretion was influenced mostly by larger flood events: the 100-year flood contributed almost 20% of the total volume of sediments above the gravel bed of the benches. The median of the vertical accretion thickness does not decrease with height above the mean water level, and the thickness of accretion on benches and their GUs varied, likely due to the variability of the vegetation cover, which can cause variable hydraulic conditions; shrubs are considered to trap sediment the most effectively.

The article, inter alia, also provides a methodological template for the detailed spatio-temporal investigation of geomorphic units terrestrialization of bypassed river channel on alluvial margin obtained via GIS diachronic analyses and LiDAR DEMs. Such kind of research also allows to define maintenance measures, identify the areas where such measures are appropriate, and provide operational data such as spatial variability of sediment thickness, sediment deposit volumes, groyne-induced benches roughness, and flood conveyance capacity, respectively. In the context of profoundly modified “Anthropocene rivers”, it is essential to understand their eco-hydromorphologic trajectories in order to effectively target ecosystem dysfunctions and lost functionalities with the objective of proposing functional rehabilitation measures adapted to the river’s hydrosedimentary reality (Seignemartin et al., 2023).Danube River reach around the original anastomosing channel, also so-called inland Delta, has a potential to National Park establishment in Slovakia. Currently it belongs spatially to the Dunajské Luhy protected area, so the morphological and sedimentological development and spatial variability of groyne-induced benches initiated the development of new habitats. Protection of the new landforms with riparian vegetation habitats could be potentially part of the future discussions with considering different management approaches weighing the flood risk and habitat protection sustainably. Continuing forest clearing on these in-channel landforms should be reconsidered to support the newly formed riparian habitats where possible and stabilise the sedimentary transport.

Currently it: In view of the fact that during large flood events, a large amount of discharge and thus suspended sediment is usually diverted by the Čuňovo weir into the Old Danube, the management of this channel needs to consider not only the geomorphic-sedimentary character and adjustment of the Old Danube channel, but also upstream sedimentary dynamics and forest clearing, which should be performed selectively.