Fluvial sediment budget of a modern, restrained river: The lower reach of the Rhine in Germany

Catena 122 (2014) 91–102

Contents lists available at ScienceDirect

Catena journal homepage: www.elsevier.com/locate/catena

Fluvial sediment budget of a modern, restrained river: The lower reach of the Rhine in Germany Roy M. Frings a,⁎, Ricarda Döring a, Christian Beckhausen a, Holger Schüttrumpf a, Stefan Vollmer b a b

Institute of Hydraulic Engineering and Water Resources Management, RWTH Aachen University, Aachen, D 52056, Germany Department of Groundwater, Geology and River Morphology, Federal Institute of Hydrology, Koblenz, D 56068, Germany

a r t i c l e

i n f o

Article history: Received 3 December 2013 Received in revised form 17 May 2014 Accepted 18 June 2014 Available online xxxx Keywords: Rhine River Sediment budget Rating curve Bed erosion Annual sediment flux Human impact

a b s t r a c t The Rhine River is a restrained river which is intensely used for navigation. Its river bed is subject to humaninduced erosion and sedimentation processes. For river management, information on the amount, type, source, transport mode and fate of the sediments moving through the Rhine is indispensable. The objective of this study was to quantify the downstream fluxes of clay, silt, sand, gravel and cobbles through the Rhine between 1991 and 2010 and to identify the sources and sinks of these sediments. This was done by analysing a unique dataset containing thousands of sediment transport measurements and by evaluating the sediment budget. The river bed of the Rhine was found to be subject to a net bed degradation of 3 mm/a between 1991 and 2010. Bed degradation has been induced by 18th–20th century river training works and nowadays is concentrated in areas with Tertiary sands close to the bed surface, in areas with mining-induced subsidence and in the gravel–sand transition zone. Sediment transport was found to be dominated by suspended clay and silt. Morphologically relevant, however, are only the sand, gravel and cobble fractions. Despite the armoured, gravely river bed, sand is the main morphological agent. Sediment loads change in the downstream direction: sand and fine gravel loads increase due to erosion of the bed, whereas coarse gravel and cobble loads decrease due to a reduced sediment mobility caused by the downstream decreasing bed slope. Approximately one third of the sand and gravel load comes from upstream (Rhenish Massif), one third is supplied by bed degradation and one third is supplied artificially by humans for bed stabilisation purposes or as substitute for natural bedload. Slightly more than one half of the sediment was transported downstream into the North Sea Basin (Rhine Delta), a small amount was lost by abrasion, and the remainder must have been deposited in groyne fields, on floodplains or in ports. The transfer of sand, gravel and cobbles from the hinterland towards the Rhine delta equalled 0.66 Mt/a ± 26%. Despite the long history of human impact, this rate does not differ significantly from the Holocene rate of sediment transfer to the Rhine delta. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The Rhine River is the most intensely used inland waterway in Europe, but due to a long history of human impact, its river bed is not stable morphologically: long reaches of the river are subject to erosion or sedimentation (Frings et al., 2009, 2014), causing problems for navigation, infrastructure, ecology, drinking water supply and flood safety (e.g. Gölz, 1994). Commonly, erosion and sedimentation processes are investigated with echo soundings, although echo soundings fail to provide answers to questions such as: “How much sediment are moving downstream?”, “Which grain size fractions are in motion?”, “How are sediments being transported (as bed load or suspended load)?”, “Where are the sediments transported by the river coming from?”, and ⁎ Corresponding author. Tel.: +49 241 80 25265; fax: +49 241 80 25750. E-mail addresses: [email protected] (R.M. Frings), [email protected] (R. Döring), [email protected] (H. Schüttrumpf), [email protected] (S. Vollmer).

http://dx.doi.org/10.1016/j.catena.2014.06.007 0341-8162/© 2014 Elsevier B.V. All rights reserved.

“Where are the eroded sediments going to?”. Answers to these questions are indispensable for a proper understanding of morphological river behaviour and for making realistic and trustworthy predictions of future erosion and sedimentation rates. Furthermore, the answers are needed to calibrate numerical models, to optimize dredging strategies, and as a starting point for studies on climate change and human impact on river systems. The objective of this study was to provide an answer to the aforementioned questions by (1) quantifying the downstream fluxes of clay, silt, sand, gravel and cobbles through the Rhine, and (2) identifying the within-channel and upstream sources and sinks of these sediments. This was done by analysing a unique dataset containing thousands of sediment transport measurements, followed by a computation of the sediment budget of the study area. To support these computations, we analysed grain size data and bed-level data. We focus on the time period between 1991 and 2010 and deal with the 226 km long river reach just upstream of the Rhine delta, situated in the Lower Rhine Embayment (Fig. 1, Rhine-km 640–866). A morphological analysis for the northern

92

R.M. Frings et al. / Catena 122 (2014) 91–102

Fig. 1. Geology, A) chronostratigraphy and tectonics of the Rhine basin (Frings et al, 2014), B) lithology of the river bed in the Lower Rhine Embayment (based on studies in the 1970s and 1980s, after Gölz, 1992).

Upper Rhine Graben, Mainz Basin and Rhenish Massif (Fig. 1, Rhine-km 338–640) was already provided by Frings et al. (2014). In the Discussion section, the results of the present study are interpreted and compared to Quaternary-geologic sediment budgets in order to evaluate whether the long history of human impact on the Rhine has led to a reduction of land-coast sediment transfer.

2. The river Rhine The Rhine originates in the Swiss Alps and flows through Switzerland, Germany and The Netherlands towards the North Sea (Fig. 1). Its drainage basin covers 185,000 km2. This study focuses on the area between the village of Koenigswinter at the edge of the Rhenish

R.M. Frings et al. / Catena 122 (2014) 91–102

Massif (Rhine km 640, 44 m a.s.l.) and the village of Millingen a/d Rijn (Rhine km 866, 3 m a.s.l.) at the German–Dutch border (Fig. 1). The discharge regime of the study area is rain-dominated, with maximum discharges in the winter period. The mean discharge (gauge Rees) between 1991 and 2010 was 2311 m3/s, whereas the maximum discharge ever recorded was 12,200 m3/s (DGJ, 1926). The river width varies between 230 and 300 m. Within the study area no major tributaries discharge into the Rhine. Mean flow velocities vary between 0.9 and 1.6 m/s (Fig. 2). During floods, flow velocities decrease downstream, partly due to a decrease of bed gradient towards the Rhine delta (from 20 to 11 cm/km), partly due to a downstream increase in river and floodplain width. Geologically, the study area belongs to the Lower Rhine Embayment (Fig. 1), a subsidence zone belonging to the European Rift system. The river bed consists of a sand–gravel mixture of Quaternary age, locally reaching thicknesses of over 100 m (Rothe, 2000). In the central and northern part of the study area the Quaternary cover is locally absent (Fig. 1B), exposing the Tertiary deposits underneath to the flow. Tertiary deposits predominantly consist of fine marine sands with varying clay and silt content. Locally sandstone, clay and peat layers occur. Human impact on fluvial morphodynamics in the Rhine Basin started during the Iron Age with deforestation of valley slopes and increased alluvial deposition rates (Erkens et al., 2006). In the early Middle Ages, inhabitants built small engineering works for flood protection, land reclamation and irrigation (e.g., Middelkoop, 1997; Tümmers, 1999). In the centuries thereafter, the river was gradually completely embanked (Schmidt, 2000). The first large-scale engineering works in the channel itself were carried out in the 18th century: in order to reduce flood risks, the Rhine was straightened (by cutting off meander bends and connecting islands to the banks) and narrowed (by building groynes and bank revetments) (Tümmers, 1999). Because of the increasing importance of the Rhine as an international shipping route in the 19th century, the river was further narrowed and straightened (Jasmund, 1901). Embankments were regularly reinforced and heightened (Schmidt, 2000). In the 20th century numerous smaller engineering works and flood protection measures were carried out. Furthermore, in many tributaries dams were built to generate energy. Since the 1920s, coal has been mined underneath the Rhine, which led to rapid subsidence of the river bed (Rommel, 2005; Wenka, 2009). In the period 1991–2010 subsidence was limited to the river reach between Rhinekm 791.5–809. The subsidence funnels were continually filled with mining waste by river managers (about 13.6 Mt since 1976). The Rhine is the most intensely used inland waterway in Europe, connecting the port of Rotterdam to the industrial areas in the hinterland. The annual transport of goods over the river amounts to 300 ± 25 million tonnes (J.P. Weber, CCR-ZKR, pers. comm.). In order to allow year-round navigability, continuous maintenance is required: newly formed sediment deposits that hinder navigation are dredged (Fig. 3A) and re-allocated to the river elsewhere (Fig. 3A). Furthermore, since 2000 AD allochthonous sediments have been artificially supplied to river stretches with a sediment deficit. This concerns relatively fine

93

Fig. 3. Amounts of sediment dredged from the river bed or artificially supplied to the river between 1991 and 2010: A) dredging and re-allocation (supply) of dredged sediments, B) artificial supply of fine gravel (typically 4–32 mm) as substitute for natural bed-load, C) artificial supply of coarse gravel and cobbles (8–150 mm) for bed stabilisation purposes (volumes exclude the sediment used for filling the subsidence funnels).

material (typically 4–32 mm) that is meant to be a substitute for natural bed-load and serves to stop the general bed degradation trend (Fig. 3B). In order to stabilise the bed in areas prone to local scouring, coarser allochthonous sediments (8–150 mm) are supplied the river bed too (Fig. 3C). 3. Methods 3.1. Approach In order to quantify the downstream fluxes of clay, silt, sand, gravel and cobbles through the Rhine and to identify the sources and sinks of these sediments, we first made a reconstruction of the geomorphological development of the Rhine in the period 1991 to 2010 by analysing bed-level data (Section 3.2). Then, we analysed a unique dataset containing thousands of sediment transport measurements (Section 3.3). The data that were obtained in this way were combined with information on tributary sediment supply, floodplain deposition, abrasion and anthropogenic sediment fluxes to compute the sediment budget of the study area (Section 3.4). To support these computations, we analysed grain size data (Section 3.5). 3.2. Bed level analysis

Fig. 2. Longitudinal velocity profile of the lower Rhine Embayment (SOBEK model computation, period 1993–2010).

The German Federal Waterways and Shipping Administration regularly carries out bed-level measurements in the entire river. For

94

R.M. Frings et al. / Catena 122 (2014) 91–102

reconstructing the long-term morphological development in our study area, we analysed the surveys of 1934, 1960, 1975, 1985, 1991, 1995, 2000, 2002, 2004, 2006, 2008 and 2010. During these surveys, most of which took several months, detailed cross-sections were sounded at 100 m distances with single-beam echo sounders. Afterwards, the measured water depths of each cross-section measurement were projected onto a profile perpendicular to the river axis. Then, the mean water depth was calculated and converted into the mean bed level (in m a.s.l.) by subtraction of the measured water depth from the known water level. This was done for the morphologically active part of each cross-section (i.e. that part of the channel that is situated between the groynes). 3.3. Sediment transport analysis Using sampler techniques, the Federal Waterways and Shipping Administration has carried out systematic measurements of the sediment transport in the Rhine since 1974. Until 2001, a bed-load sampler with a 1.0 mm mesh was used. After this date the sampler construction was adapted and a 1.4 mm mesh (upstream of km 800) or 0.5 mm mesh (downstream of km 800) was used. Suspended load is measured by taking a water sample of 50 l using a pump sampler. The sample is passed over 0.063-mm sieve in order to collect the suspended sand. Subsequently, another 3–5 l water sample is passed over the 0.063-mm sieve and a 0.006-mm filter in order to collect the suspended fines (silt, flocculated clay minerals, and organic matter). The filter is sent to the laboratory for determination of the solid matter content. A detailed description of measuring techniques for bed load and suspended load (including both suspended bed material load and wash load) is provided by Frings et al. (2014). In this study we analysed 19 bed-load and 19 suspended-load sampling sites in the lower reach of the Rhine in Germany, with altogether 322 suspended-load measurements and 820 bedload measurements made between 1991 and 2010. Each measurement consists of several samples across the river width. Data were exported from the transport database (SedDB) on 4 February 2011. For each of the 19 study sites, the transport data were plotted against flow discharge and a power function (a so-called rating curve) was fitted. The functions had the following form: T = aQb, with T sediment transport (kg/s), Q discharge (m3/s), and a, b coefficients. The coefficients were determined by linear least-squares regression after log-transforming T and Q. This was done separately for suspended clay/silt, suspended sand, sand transported as bed-load, fine gravel, coarse gravel (including cobbles) and bed-load as a whole (Table 1). Because rating curves are sensitive to outliers in sediment transport we deleted data points that deviated more than 2.5 standard deviations from the rating curve and recalculated the rating curves. Only rating curves that were statistically significant (F-test, 0.05 level) and representative for the entire discharge range (i.e. including high-flow measurements) were used. The rating curves were combined with time-series of daily discharges in order to estimate the average annual sediment load at each of the study sites. To correct for the systematic underestimation of the annual loads that is inherent to the use of power-type rating curves, we applied the Duan (1983) bias correction method. More details about the procedure for calculating annual loads are given by Frings et al (2014). In this way, we were able to establish accurate estimates of the annual transport rates for about 50% of the measuring sites.

For the other sites, annual loads could not be determined in this way due to either a scarcity of suspended-load measurements during high flow or a strong scattering in the bed-load data. Transport estimates for these sites were obtained as follows. Firstly, we assumed the few suspended load measurements at discharges greater than 5000 m3/s to be valid for all measuring sites, thus implicitly assuming that suspended loads during floods do not change in the downstream direction. This assumption may be incorrect and certainly reduces the resolving power of the analysis, but renders rating curves significant and representative and thus allows for a first approximation of annual loads. Secondly, we approximated the annual bed load fluxes at locations with insignificant or unrepresentative rating curves by Test, with: Test = k (c T1000–2000 + d T2000–4000). T1000–2000 and T2000–4000 represent the mean transport rate in the period 1991–2010 in the discharge classes 1000–2000 m3/s and 2000–4000 m3/s respectively. c and d are the coefficients indicating the proportion of time the respective discharge class prevails (0.38 and 0.52 respectively). The term between brackets represents the sediment flux in 90% of the year and can be determined accurately due to the high number of bed-load measurements. k is a coefficient to upscale the result to the entire year. It was determined by relating the annual loads to the term between brackets (Fig. 4) for those measuring sites with significant and representative rating curves. The value of k was determined for each grain size fraction separately. Because the k-values were all very similar and statistically indifferent at the 95% level, we used the average value of k for all grain size fractions. 3.4. Sediment budget analysis After quantifying the annual sediment fluxes through the Rhine, we determined their sources and sinks by evaluating the sediment budget of the Rhine for the period 1991–2010, focusing on the sand and gravel sediments. The sediment budget (Fig. 5) is the balance (I − O = ΔS) between the mass of sediment entering the study area (I), the mass of sediment leaving the study area (O) and the change in sediment mass stored in the study area (ΔS). The sediment budget for the lower reach of the Rhine in Germany is expressed as follows (all units in tonnes/a): I u þ It þ Ia −Od −Odr −Ofgp −Oa ¼ ΔS with Iu sediment input from upstream, It sediment input by tributaries, Ia artificial sediment supply, Od sediment output to the downstream area (Rhine delta), Odr dredging, Ofgp sediment loss due to sedimentation on floodplains, in groyne fields or in ports, Oa abrasion and ΔS net bed level change. Implicit to the budget equation is the assumption that bank erosion is effectively prohibited by bank protection works and that tectonic bed level changes are negligible. Obviously, the second assumption is violated in the mining subsidence area. However, because subsidence effects are very local and funnels are continually filled with

Table 1 Sediment fluxes analysed in this study. ID

Mode

Sediment type

Grain size (mm)

F1 F2 F3 F4 F5 F6

Suspension Suspension Bed load Bed load Bed load Bed load

Silt, flocculated clay minerals Sand Sand Fine gravel Coarse gravel, cobbles Sand, gravel, cobbles

0.006–0.063 0.063–2 0.063–2 2–16 16–125 0.063–125

Fig. 4. Estimation of the contribution of discharges below 1000 m3/s and discharges over 4000 m3/s to the annual sediment load.

R.M. Frings et al. / Catena 122 (2014) 91–102

95

the average bed level change in the morphologically active part of the channel (m/a) (Section 3.1), Wn the width of the morphologically active part of the channel (i.e. between the groynes) (m), L the length of the study reach (m), ρs the mineral density (2.60 tonnes/m3) and p porosity (Fig. 6). Because of the downstream variation in Wn, ρs, p and Δz / Δt, we divided the 226 km long study area into subreaches of 100 m long and calculated ΔS for each of these subreaches. Subsequently, ΔS values were accumulated over the entire study area. 3.5. Grain size analysis

Fig. 5. Sediment budget terms for the Rhine (Frings et al, 2014), with Iu sediment input from upstream, It sediment input by tributaries, Ia artificial sediment supply, Od sediment output to the downstream area, Odr dredging, Ofgp sediment loss due to sedimentation on floodplains, in groyne fields or in ports, Oa abrasion and ΔS net bed level change.

The German Federal Waterways and Shipping Administration has carried out thousands of bed grain size measurements in the Rhine since 1968. The common sampling procedure consists of lowering a caisson to the river bed and photographing the bed, followed by manual sampling of the bed sediments at five positions across the river width in cross-sections situated 500 or 1000 m apart. All samples are visually inspected and sieved. For this study, we analysed two grain size datasets (Fig. 7). The first covers the period 1981–1983 and contains about 1100 grain-size distributions from the bed surface (0–10 cm depth) measured with a circular mesh. The second covers the period 1992–2010 and contains about 5600 grain-size distributions of the bed surface (0–10 cm depth) and the subsurface (10–50 cm depth) measured with a quadratic mesh. Because the sieve set has changed several times since the beginning of the measurements, we log-linearly interpolated the GSDs onto an imaginary sieve set with 1-phi classes in order to make the datasets homogeneous. Before, the circular mesh sizes of 1981–1983 were according to DVWK (2000) guidelines multiplied by 0.8 to allow comparison with the quadratic mesh of 1992–2010.

mining waste (Section 2), we decided to disregard both the subsidence and the supply of mining waste. Following the approach of Frings et al. (2014), we quantified each of the budget terms independently, except for Ofgp which was used as closing term. The sediment input from upstream (Iu) was set equal to the mean annual flux of sand and gravel in the downstream part of the Rhenish Massif between 1991 and 2010. To determine this flux, we analysed the transport data of the measuring site at Rhine km 620.8 according to the procedure described in Section 3.2. The sediment input by tributaries (It) was assumed negligible, because tributaries in the Lower Rhine Embayment do not supply significant amounts of water to the Rhine (Section 2). The sediment input due to artificial supply (Ia) and the sediment output due to dredging (Odr) were taken from Fig. 3. Volumetric data were converted into mass units by multiplication with ρs(1 − p), where ρs represents the mineral density (2.60 tonnes/m3; Frings et al., 2011) and p the porosity of the sediments (−). For the porosity a spatially variable value was used (see Fig. 6), calculated using the semi-empirical porosity predictor for the Rhine, developed by Frings et al. (2011). The output to the downstream area (Od) was set equal to the average annual flux of sand and gravel measured at the downstream measuring site of Griethausen (km 857.5) (Section 3.2). The sediment output due to abrasion (Oa) represents the production of silt and mud (wash-load) during the abrasion process (cf. Frings et al., 2014). According to Gölz et al. (1995) the maximum mass loss of natural Rhine gravel due to abrasion equals 11% per 50 km, or 0.02% per hm. In order to calculate the abrasion rates in mass units, we multiplied this value with the average gravel transport rate (Section 3.2). Finally, the change in sediment mass stored in the study area (ΔS) was set equal to Wn Lρs(1 − p) Δz / Δt, with Δz / Δt

Between 1934 and 1991, the upper part of our study area had a relatively stable river bed, whereas the middle and lower part featured significant bed degradation of 1.0 m (or 1.7 cm/a). Between 1991 and 2010 the study area was subject to net average bed degradation of 3 mm/a (Fig. 8). Bed degradation was concentrated in four areas: Rhine-km 640–698, 732–769, 789–819 and 842–865, with in between areas of aggradation (Fig. 8). Suspended load transport showed a strong correlation with discharge, whereas bed-load transport did much less: determination coefficients for rating functions averaged 0.87 for suspended load and 0.30 for bed load. The ratio of bed-load to suspended-load transport decreased with discharge. On average, suspended load transport rates were a factor of 15 larger than bed-load transport rates. Predominantly clay and silt were transported (Table 2). These fractions are considered

Fig. 6. Sediment porosity estimated with the porosity predictor of Frings et al. (2011).

Fig. 7. Grain size measurements of bed grain size analysed in this study.

4. Results 4.1. Data and trends

96

R.M. Frings et al. / Catena 122 (2014) 91–102

Fig. 9. The average composition of the sand and gravel load at Rhine km 857.5 (period 1991–2010).

Table 2) equalled 2.22 Mt/a for suspended clay and silt, 0.48 Mt/a for suspended sand, 0.08 Mt/a for bed-load sand, 0.09 Mt/a for fine gravel (2–16 mm) and 0.02 Mt/a for coarse gravel and cobbles (16–125 mm). The budget analysis for sand and gravel (Fig. 11, Table 3) shows the natural sediment supply to the study area to be limited. Approximately one third of the total sediment flux came from upstream (Rhenish Massif), one third was supplied by bed degradation and one-third was supplied artificially by humans for bed stabilisation purposes or as substitute for natural bed-load (Fig. 11). The sum of all these fluxes is 1.26 Mt/a. Slightly more than half of this sediment was transported downstream into the Rhine Delta, whereas a small amount was lost by abrasion. It was assumed that the remainder (0.54 Mt/a or about 2300 t/a/km) was deposited either in groyne fields, on floodplains or in ports. A spatial differentiation of the sediment budget (Fig. 12) shows that the relative importance of the budget terms varies in a downstream direction. The river bed of the study area consists of a sand–gravel mixture (size range 0.063–63 mm) with about 9% coarser constituents (63– 125 mm). In the period 1992–2010, the geometric mean size of the bed surface decreased from 16 mm at Rhine km 640 to about 10 mm at Rhine-km 820 (Fig. 13A), and 2–3 mm at Rhine-km 866. The strong downstream fining between km 820 and 866 represents the gravel– sand transition zone of the Rhine at the onset of the delta. Within this

Fig. 8. Bed level change: A) variation over time; B) variation in downstream direction.

to be wash load. Within the morphologically relevant part of the sediment load, sand transport dominated over gravel transport (Fig. 9). The average annual transport rate at the upstream end of the study area (Table 2) equalled 1.73 Mt/a for suspended clay and silt, 0.32 Mt/a for suspended sand, 0.04 Mt/a for bed-load sand, 0.04 Mt/a for fine gravel (2–16 mm) and 0.02 Mt/a for coarse gravel and cobbles (16–125 mm). Sediment transport rates varied in the downstream direction (Fig. 10). After remaining more or less constant until Rhine km 810, the transport of clay, silt and sand strongly increased (Fig. 10A, B, and C). Especially the strong increase of sand transported as bed-load is remarkable. Gravel transport rates remained constant to about Rhine km 730, increased until Rhine km 765 and reached a local minimum around Rhine km 810. Further downstream, the transport of fine gravel increased again, whereas the transport of coarse gravel (including cobbles) remained low (Fig. 10D and E). The average annual transport rate near the downstream end of the study area (km 857.5;

Table 2 Rating functions and annual loads for the river Rhine for 6 grain size fractions. The rating functions had the following form: T = aQb, with T sediment transport (kg/s), Q discharge (m3/s), and a, b coefficients. The grain-size fractions (F1–F6) are defined in Table 1. Km

Rating curve coefficient (a, ×10−6) F1

645.8 681.3 703.6 720.0 729.3 749.5 756.0 762.0 768.0 782.0 794.6 800.0 808.5 813.0 818.2 825.0 838.4 845.0 857.5 866.0

F2

14.5 1.25 21.2 0.57 22.5 8.94 31.3 0.86 55.5 0.62 25.7 0.15 38.8 0.56 34.0 0.49 62.4 0.88 58.0 1.45 33.4 0.62 1454 24.2 120 3.13 77.6 0.97 376 7.16 72.7 0.42 327 4.10 118 1.84 460 26.8

F3

F4

Rating curve coefficient (b) F5

41.3 7.50 18.3

2.40 97.7 10.6 19.7 85 1203 20.5 230

47.7 0.01 0.63

88.9 0.17 0.12

28.59 1245 459 67.1

82.1 0.89

F6 31.9 50.3 251 125

F1

1.94 1.89 1.88 1.82 1.75 1.85 1.80 3.93 44.5 1.82 11.2 0.87 1.74 0.05 1.50 1.76 1.82 1.38 260 1.68 1.73 1086 1.53 1.72 1.56 0.11 29.3 1.68 1.52

F4

F5

Bias correction factor (−)

F2

F3

F6

2.04 2.13 1.80 2.08 2.11 2.28 2.11 2.13 2.06 2.00 2.10 1.68 1.92 2.06 1.82 2.15 1.89 1.99 1.68

1.64 1.05 1.43 1.21 1.49 1.42 1.42 1.40 1.21 0.86 1.20 1.33 1.42 1.05 1.29

1.21 1.30 1.66 1.49 2.23 2.11 1.49 1.99 1.76 2.15 2.05 1.88

1.35

1.16

0.94 0.81 0.98 1.29 1.32 1.88 1.85 1.53

F3

F4

Annual load (×0.1 Mt/a)

F1

F2

F5

1.14 1.09 1.12 1.07 1.07 1.06 1.06 1.07 1.07 1.09 1.07 1.06 1.07 1.06 1.07 1.07 1.08 1.10 1.09

1.08 1.88 1.58 1.07 1.78 1.50 1.22 1.14 1.89 1.36 1.34 1.12 1.84 1.45 1.42 1.12 1.15 1.11 1.09 1.83 1.24 1.33 1.12 2.04 1.24 1.23 1.13 1.83 1.28 1.71 1.11 1.11 1.12 1.39 1.09 1.11 1.42 1.36 1.12 1.48 1.13 1.43 1.14 1.36 1.36 1.71 1.15

F6

F1

F2

F3

F4

F5

F6

1.61 1.30 1.25 1.43

17.3 16.9 17.4 14.8 15.0 14.4 15.0 15.3 16.8 18.4 16.6 22.5 19.7 19.3 19.5 17.4 20.8 20.5 22.2

3.22 3.12 3.78 3.30 3.02 2.97 2.84 2.90 3.27 3.31 3.15 4.41 3.91 3.65 4.01 3.30 4.04 4.04 4.77

0.35 0.25 0.22 0.30 0.22 0.49 0.50 0.29 0.37 0.34 0.37

0.44 0.50 0.42 0.53 0.59 0.94 1.00 0.79 1.09 1.02 1.10

0.15 0.44 0.35 0.33 0.31 0.47 0.45 0.62 0.46 0.30 0.36

0.93 1.20 1.04 1.15 1.22 1.95 1.97 1.75 1.93 1.53 1.83

0.32 0.25 0.44 0.43 0.56 0.79 0.78 1.16

0.47 0.44 0.79 0.68 0.82 0.97 0.90 0.89

0.14 0.12 0.24 0.23 0.19 0.11 0.16 0.08

0.90 0.80 1.67 1.43 1.60 1.79 1.84 2.14

1.27 1.20 1.32

1.36

1.25

R.M. Frings et al. / Catena 122 (2014) 91–102

zone, the gravel content decreases rapidly downstream, whereas the sand content increases from below 25% to over 50%. Between the periods 1981–1983 and 1992–2010, the river bed became markedly

97

coarser, especially near the downstream border of the study area (Fig. 13B). The subsurface sediments were generally somewhat finer than the surface sediments (Fig. 13C), indicating the presence of an armour layer. In most of the study area, the bed-load grain size equaled the subsurface bed grain size (Fig. 13D).

4.2. Uncertainties A systematic error of up to 5 cm can locally be present in the bed level data that were used (Frings et al., 2014). The bed level change between 1991 and 2010 (Fig. 8), however, at most locations was greater than 5 cm (or 0.25 cm/a), which means that the observed bed level changes are significant. Due to measurement errors and uncertainties in the fitting of rating curves, the uncertainty of annual loads is quite high. Using the standard error of the estimate for each rating curve, we calculated the 95% confidence interval of the annual loads (Fig. 10). According to these confidence intervals, differences in annual load between two successive measuring sites are statistically not significant at the 95% confidence level. The large-scale longitudinal trends in annual loads, on the other hand, are statistically significant. Note that the confidence intervals do not include the effects of any systematic errors. The risk of a systematic underestimation of the transport due to the loss of sand particles through the 1.4 mm meshed bed-load sampler is substantial, though (Frings et al., 2014). The risk was reduced by applying a 0.5 mm mesh at places with high sand content (downstream of km 800), but this made the data series a little inconsistent. It is certain, therefore, that the bed-load transport of sand in this study (e.g. Figs. 9 and 10) was underestimated at some places. The same is true for the suspended load transport of clay and silt. The use of a filter with a 0.006 mm mesh allows fine silt and clay to pass through the filter, causing an underestimation of the real load. This underestimation is limited though (probably less than 20%), because only about 20–40% of the suspended load is finer than 0.006 mm (B. Brandstetter, Federal Institute of Hydrology, Pers. Comm.), whereas most of this fine material is present in flocculated state (flocs are not destroyed prior to the measurements) and therefore most likely to be trapped on the filter. With respect to the longitudinal trends in suspended load, the assumption had to be made that the few suspended load measurements at discharges greater than 5000 m3/s are valid for all measuring sites (Section 3.3). This implies that suspended loads during floods are supposed not to change in the downstream direction, ignoring the likelihood that some of the suspended sediments are deposited on the floodplains and introducing a small error in the fluxes of suspended material as stated in Section 3.3. However, without this assumption, it would have been impossible to obtain significant and representative annual load estimates at all. The uncertainty of sediment budget calculations is related to the assumptions that were made prior to the budget analysis (Section 3.3) and to the estimation of the individual terms of the budget. Artificial supply rates (Ia) are well documented (because of their high economic value) and the related uncertainty therefore is small. Its maximum error is estimated here at 20%. The uncertainties of the upstream supply (Iu) and the sediment output to the downstream area (Od) are equal to the uncertainty of the calculated annual sand and gravel load: their maximum stochastic error (at the 95% confidence level) equals about 40% (cf. Fig. 10). The maximum error of the sediment loss by abrasion (Oa) was estimated to be 50%. Sediment losses due to sedimentation on floodplains, in groyne fields and in ports (Ofgp) have never been Fig. 10. Longitudinal variation in sediment transport (period 1991–2010): A) clay and silt (b0.063 mm), B) sand (suspended) (0.063–2 mm), C) sand (bed-load) (0.063–2 mm), D) fine gravel (2–16 mm), E) coarse gravel (16–125 mm), F) bed-load (0.063–125 mm). The shaded area indicates the 95% confidence interval. Method 1: based on rating curves; Method 2: based on average transport rates in discharge class 1000–4000 m3/s (see Section 3.2).

98

R.M. Frings et al. / Catena 122 (2014) 91–102

Fig. 11. Sediment budget for gravel and sand for the Rhine reach between km 640–865 (period 1991–2010). 100% = 1.26 Mt/a * estimated.

quantified and therefore are rather uncertain. We assume a maximum error of 75%, based on the consideration that the estimate is indeed uncertain, but that its order of magnitude must be correct (see Section 5.1). The uncertainty in the last term of the budget, the change of sediment mass stored in the study area (ΔS), is relatively small, because it is unlikely that a systematic measuring error occurred throughout the study area. The maximum error of ΔS was assumed to be 20%. 5. Discussion 5.1. Verification of sediment fluxes We verified the sediment fluxes calculated in this study by comparing them to flux data from other datasets that were determined independently. The suspended load transport was verified against suspended load data from the SchwebDB dataset of the Federal Institute of Hydrology, containing daily values of the suspended load transport, based on

Table 3 Summary of the source terms, sink terms and storage terms of the sediment budget for gravel and sand, including estimated uncertainty ranges. For symbol definition, see text. Sources (Mt/a) Iu Ia

0.40 ± 0.16 0.42 ± 0.08

Sinks (Mt/a)

Storage change (Mt/a)

Od Ofgp Oa

ΔS

0.66 ± 0.26 0.54 ± 0.40 0.05 ± 0.03

0.43 ± 0.09

daily surface water samples in the middle of the river. The match between both transport estimates is good (Fig. 14) with the SchwebDB (2.44 Mt/a) differing less than 10% from the results of this study. The suspended sand data from this study are consistent with suspended sand data published by Frings and Kleinhans (2008) (Fig. 15): the relation between discharge and suspended sand load is the same for both datasets. The data by Frings and Kleinhans (2008) stem from acoustic measurements in the main Rhine branch (i.e. Waal) of the Rhine delta. Because this branch only discharges 67% of the Rhine's discharge, transport values were multiplied by 1.5 before they were plotted in Fig. 15. In order to validate the calculated bed load transport rates, we compared the sampler-based bed-load estimates from this study to bedload estimates obtained by dune tracking (Frings and Kleinhans, 2008), which were converted into mass units by assuming a porosity of 0.30 and a mineral density of 2600 kg/m3. Again the match is good (Fig. 15). The comparison also shows the presence of bed-load transport rates measured during floods to be necessary for producing reliable rating curves (Fig. 15A). In this study, the annual bed load at the downstream border of the study area was found to be 0.18 Mt/a (Table 2). If one includes the high flow dune tracking data (Fig. 15), the annual load becomes 0.25 Mt/a, which is still within the error band of Fig. 10F. The sediment budget as a whole was verified by comparing the annual sediment loads calculated from the sand and gravel budget to those calculated from the transport measurements (Fig. 12B). Both estimates correlate surprisingly well. This is not only the case at the beginning and end of the study area (where the match was forced), but also in the river stretches in between. Only one discrepancy exists, viz. around km 680. An explanation for the discrepancy has not been

R.M. Frings et al. / Catena 122 (2014) 91–102

99

Fig. 12. The components of the sediment budget (A) and a comparison between the transport rates calculated from the sediment budget to those calculated from the transport measurements (B).

found, but the discrepancy remains within the uncertainty band provided in Fig. 10. The sediment (sand and gravel) input to the Lower Rhine Embayment (period 1991–2010) calculated in this study was compared to the sediment output from the Rhenish Massif (period 1986–2006) as determined by Frings et al. (2014). The latter value was 35% higher (Fig. 16). The discrepancy does not arise due to the minor differences in analysis methods but is caused by the fact that the measured transport rates from the period 1986–1990 were distinctly higher than those from the period 2007–2010. Finally, the plausibility of the magnitude of the closing entry of the sediment budget (i.e. the sedimentation on floodplains, in groyne fields and in ports) was evaluated. Floodplain sedimentation includes deposition of coarse sediments (mainly sand) on the natural levees and deposition of fine sediments (silt, clay) further away from the river. For the sediment budget, only the levee deposition of sand is relevant. Measurements of levee deposition in our study area do not exist, but measurements in the Rhine delta show levee deposition rates to attain values up to 649 m3/km annually (Ten Brinke et al., 2001, Table 1, based on Ten Brinke et al., 1998). With a mineral density (ρs) of 2.60 tonnes/m3 and a porosity (p) of 0.40 (typical for sand deposits), this corresponds to an annual sand loss of 1010 t/a/km. Information on the deposition of sand in ports and groyne fields does not exist. It is to be expected that the combined deposition on floodplains, in ports and in groyne fields in our study area is of the order of 103 t/a/km. The deposition rate that was determined by the budget analysis (2.3 · 103 t/a/ km) is relatively high, but has the correct order of magnitude.

Fig. 13. Grain size: A) downstream variation of bed grain size; B) temporal variation of bed grain size, C) difference in bed grain size between surface and subsurface sediments, D) comparison of bed grain size to bed-load grain size. In panel A, the Dutch grain size data of Ten Brinke et al.(1997), corrected by Frings et al. (2009), are shown in addition to the German data. Panel A shows individual grain size measurements, Panels B, C and D show cross-section averaged grain size values.

5.2. Interpretation of morphodynamics The bed degradation trend that was observed in our study area (Fig. 8) has also been observed in other parts of the Rhine catchment (e.g. Frings

Fig. 14. Comparison of the average annual suspended load (clay, silt, sand) calculated in this study to those from the SchwebDB database (period 1991–2010).

100

R.M. Frings et al. / Catena 122 (2014) 91–102

Fig. 15. Comparison of sediment transport rates analysed in this study to those published by Frings and Kleinhans (2008): A. Bed-load, B. Suspended sand.

et al., 2009, 2014). The degradation has been induced by the human interventions (especially the narrowing of the river) of the 18th, 19th and 20th centuries (Section 2). These increased the bed shear stress (e.g. Frings et al., 2009) and sediment transport capacity, to which the river reacted by bed degradation. Bed degradation rates have decreased over time (Fig. 8A), but degradation is expected to continue into the future, because it takes several centuries before the Rhine attains a new equilibrium after disturbances (De Vries, 1975). Present-day degradation rates (Fig. 8B) would have been much higher without the countermeasures that were taken (Section 2). Especially the sediment management is important: between 1991 and 2010 8.4 million tons of allochthonous sediment (mainly gravel) were supplied to the river (Fig. 3), corresponding to a sediment layer of 8 cm (or 4 mm/a) thickness, evenly spread over the 226 km long study area. These numbers do not include the 13.6 million tons of mining waste disposed between km 791.5 and 809 in this period (Section 2). The artificially supplied sediments not only compensated bed erosion, but also coarsened the river bed (Fig. 13B) thereby protecting the underlying sediments from being eroded. Bed degradation rates vary spatially (Fig. 8A). For a large part of the study area, degradation rates are higher in the more sandy stretches (Fig. 17), suggesting that it is mainly sand that is eroded from the river bed. This has also been observed in other parts of the Rhine (see Frings et al., 2014). In order to release sand from the subsurface, the

Fig. 16. Comparison of annual sand and gravel loads from the northern Upper Rhine Graben and Rhenish Massif (Frings et al, 2014) to those calculated in this study for the Lower Rhine Embayment.

Fig. 17. Average bed degradation between 1991 and 2010 in relation to the average sand content of the subsurface (data from between 1992 and 2010): A. km 640–780, B. 780–860.

armour layer must be disrupted. This can occur due to e.g. passing ships, dredging activities and migrating bed forms. The erosion of sand accelerates the process of bed coarsening (Fig. 13B) thereby reducing future erosion rates. The downstream variation of sand and gravel loads (Fig. 10) downstream of Rhine-km 730 can be explained as follows. The increase in sediment loads between Rhine-km 730 and 765 (Fig. 10) must be caused by the erosion of sand and gravel from the river bed (Fig. 8B). The high erosion rates (cf. Figs. 8B and 12A), probably are partly due to the presence of highly-erodible sandy Tertiary deposits at shallow depths below the bed surface (Fig. 1B) and partly due to recent engineering works. The sudden drop in sediment loads between km 794 and 810 (Fig. 10C, D, E, and F), probably is induced by mining subsidence, which was highest in this area (Section 2). Subsidence results in bed level lowering (Fig. 8A) and accommodation space and therefore causes sedimentation and a downstream decrease of sediment load. This process will have been reduced, but not stopped, by the regular filling of the subsidence funnels by mining waste (Section 2). The sedimentation in the subsidence area must have led to a sediment deficit in the area directly downstream. Thus, the sudden increase of the annual loads in the area directly downstream of the subsidence area (km 810–820; Fig. 10C, D, E, and F) can be explained by erosion of sediments from the river bed (Fig. 8A). The presence of sandy Tertiary deposits might contribute to the high erosion rates (Fig. 1B). To reduce bed degradation, large amounts of allochthonous sediments are supplied to the river in this area (Fig. 3). The downstream decrease of coarse gravel loads from km 820 onwards is caused by the decreasing flow velocities (Fig. 2) in this area (Section 2). Km 820 marks the onset of the gravel–sand transition in the Rhine (Frings, 2011), in which an abrupt change from a gravel-bed channel to a sand-bed channel occurs (Fig. 13A). In the gravel–sand transition, the sand content of the river bed rapidly increases downstream (cf. Fig. 13), which facilitates bed erosion (Fig. 8A). The erosion of sand from the bed is reflected by a strong increase in sand transport rates downstream of km 845 (Fig. 10B and C).

R.M. Frings et al. / Catena 122 (2014) 91–102

Suspended silt and clay loads remain more or less constant throughout most of the study area. From km 800 onwards, however, a marked downstream increase in these loads occurs (Fig. 10). Given the combination of bed degradation in this area with the presence of Tertiary deposits close to the bed surface, a part of the increase in suspended load may be caused by erosion of the silt and clayey part of the Tertiary sediments. Furthermore, there are indications that parts of the 13.6 Mt of mining waste that were used to fill the subsidence funnels in the mining area between km 791.5–809 (Section 2) directly disintegrated after coming in contact with water, thus producing an increase in suspended load concentrations. 5.3. Human impact on land–coast sediment fluxes The transfer of sediment by rivers from the hinterland to coastal zones and oceans is a major pathway for material transfer on Earth (Walling, 2006). Although soil erosion rates are increasing globally due to land clearance for agriculture, the sediment transfer to the coastal zone often decreases due to the construction of dams (Syvitski, 2003), leading to coastal erosion. Almost all studies of land–coast sediment fluxes focus on silt and clay sized particles, which are better preserved in the geological record and more easily measured in today's rivers than coarser particles. As a result, little is known on the transfer of sand and gravel from the hinterland to river deltas and the sea, and the human impact thereon. Our dataset presents an excellent opportunity to evaluate the human impact on the transfer of sand and gravel from the hinterland to the Rhine delta. First, we estimated the transfer of sand and gravel to the Rhine delta during the Holocene using Quaternary-geologic data. Because the Rhine delta is known to have been a near-complete sediment trap for Rhine sediments during the Holocene (Beets and Van der Spek, 2000), the Holocene transfer of bed material towards the delta must equal the total Holocene accumulation of sand and gravel in the Rhine delta. Based on geological cross-sections, Erkens et al. (2006) estimated the total Holocene accumulation of sand and gravel to be 4.68 km3 ± 20%. After multiplication with the mineral density and solid fraction of the sediments (2600 kg/m3 and 66% respectively; Frings et al., 2011) and division by the duration of deposition (9000 years; Gouw and Erkens, 2007), the average transfer of sand and gravel from the hinterland to the Rhine delta is found. It must have been of the order of 0.89 Mt/a ± 20%. According to our study (Table 2, Fig. 10), the present-day transfer of sand and gravel from the hinterland to the Rhine delta equals 0.66 Mt/a ± 26%. In his sediment budget analysis, Ten Brinke (2005) also provided an estimate of the present-day sand and gravel transfer from the hinterland to the Rhine delta. His estimate was somewhat higher (0.85 Mt/a ± 74%). Given the high uncertainty ranges, all estimates must be considered statistically indifferent, leading to the conclusion that the transfer of bed-material (sand, gravel) from the hinterland to the Rhine delta did not change significantly over time. This is a remarkable conclusion, because the Rhine has been subject to strong human impact for more than 1000 years (see Section 2). Many of the river training works that have been carried out (e.g. embankments, meander cut-offs, river narrowing, bank protection) are known to have increased the bed shear stress of the river (Frings et al., 2009, 2014), something which may be expected to lead to increased sediment transport rates and an increased transfer of sediment to the sea. The fact that this does not occur in the Rhine can be explained by the increase in the critical bed-shear stress for incipient motion that occurred simultaneously. The increase in the critical bed-shear stress is caused by the increase of bed grain size that occurred as a result of preferential erosion of sandy sediments (e.g. Frings et al., 2009) and the artificial supply of gravel material (e.g. Fig. 3). In the natural (Holocene) situation, the bed material of the lower reach of the Rhine in Germany consisted for more than 90% of sand (e.g. Erkens et al., 2011). Today, however, the bed-material predominantly consists of gravel with general less than

101

25% sand (Section 4.1). This shows that the increased bed shear stress was counteracted by an increased critical shear stress, leaving the overall transport capacity apparently about the same. 6. Conclusions The river bed of the lower course of the Rhine in Germany is subject to bed degradation. Degradation rates averaged 3 mm/a in the period 1991–2010. Bed degradation has been induced by 18th–20th century river training works, and nowadays is concentrated in areas with Tertiary sands close to the bed surface, mining-induced subsidence and in the gravel–sand transition zone. Sediment transport in the lower reach of the Rhine in Germany is dominated by suspended clay and silt. Morphologically relevant, however, are only the sand, gravel and cobble fractions. Despite the armoured, gravelly river bed, sand is the main morphological agent. Sediment loads change in the downstream direction: sand and fine gravel loads increase due to erosion of the bed, whereas coarse gravel and cobble loads decrease due to a reduced sediment mobility caused by the downstream decreasing bed slope and flow velocities. Approximately one third of the sand and gravel load comes from upstream (Rhenish Massif), one third is supplied by bed degradation and one third is supplied artificially by humans for bed stabilisation purposes or as a substitute for natural bed-load. Slightly more than one half of this sediment was transported downstream into the North Sea Basin (Rhine Delta), a small amount was lost by abrasion, and the remainder must have been deposited in groyne fields, on floodplains or in ports. The transfer of sand, gravel and cobbles from the hinterland towards the Rhine delta equalled 0.66 Mt/a ± 26% in the period 1991–2010. This rate does not differ significantly from the pre-human Holocene rate of sediment transfer to the Rhine delta. Although human activities caused an increase in bed shear stress, which increases the mobility of the bed sediments, human activities also caused a coarsening of the river bed and hence an increase of the critical bed shear stress for incipient motion, which decreased again the mobility of the bed sediments. Acknowledgements This study benefited from the help of Nicole Gehres, Gudrun Hillebrand, Klaudia Krötz, Wilfried Otto, Markus Promny, Jürgen Schmegg, Sönke Schriever (Federal Institute of Hydrology), Franziska Ribbe, Imke Evers, Arthur Bär (RWTH Aachen University) and two anonymous reviewers. The effort of the measuring crews during the field campaigns is particularly acknowledged. The support of Dietmar Abel (WSA Duisburg-Rhein) and his remarkable knowledge about morphodynamics, engineering and maintenance works in the Rhine was invaluable for this study. Funding was provided by the Federal Institute of Hydrology. References Beets, D.J., van der Spek, A.J.F., 2000. The Holocene evolution of the barrier and backbarrier basins of Belgium and The Netherlands as a function of late Weichselian morphology, relative sea-level rise and sediment supply. Geol. Mijnb. (Neth. J. Geosci.) 79 (1), 3–16. DGJ, 1926. Deutsches Gewässerkundliches Jahrbuch 1926, Rheingebiet Teil I–III. Bundesanstalt für Gewässerkunde, Koblenz. Duan, N., 1983. Smearing estimate: a nonparametric retransformation method. J. Am. Stat. Assoc. 78, 605–610. DVWK, 2000. Geschiebemessungen, DVWK-Regeln zur Wasserwirtschaft 127. Deutsches Verband für Wasserwirtschaft und Kulturbau, Bonn. De Vries, M., 1975. A morphological time-scale for rivers. XVIth IAHR Congress São Paulo. IAHR publications, p. 147. Erkens, G., Cohen, K.M., Gouw, M.J.P., Middelkoop, H., Hoek, W.Z., 2006. Holocene sediment budgets of the Rhine Delta (The Netherlands): a record of changing sediment delivery. In: Rowan, J.S., Duck, R.W., Werritty, A. (Eds.), Sediment Dynamics and the Hydromorphology of Fluvial Systems. IAHS Publication 306, International Association of Hydrological Sciences, Wallingford, UK, pp. 406–415. Erkens, G., Hoffmann, T., Gerlach, R., Klostermann, J., 2011. Complex fluvial response to Late-glacial and Holocene allogenic forcing in the Lower Rhine Valley (Germany). Quat. Sci. Rev. 30 (5–6), 611–627.

102

R.M. Frings et al. / Catena 122 (2014) 91–102

Frings, R.M., 2011. Sedimentary characteristics of the gravel–sand transition in the river Rhine. J. Sediment. Res. 81, 52–63. http://dx.doi.org/10.2110/jsr.2011.2. Frings, R.M., Kleinhans, M.G., 2008. Complex variations in sediment transport at three large river bifurcations during discharge waves in the river Rhine. Sedimentology 55 (5), 1145–1171. http://dx.doi.org/10.1111/j.1365-3091.2007.00940.x. Frings, R.M., Berbee, B.M., Erkens, G., Kleinhans, M.G., Gouw, M.J.P., 2009. Human-induced changes in bed shear stress and bed grain size in the river Waal (The Netherlands) during the past 900 years. Earth Surf. Process. Landf. 34 (4), 503–514. Frings, R.M., Schüttrumpf, H., Vollmer, S., 2011. Verification of porosity predictors for fluvial sand‐gravel deposits. Water Resour. Res. 47. http://dx.doi.org/10.1029/ 2010WR009690 (W07525). Frings, R.M., Gehres, N., Promny, M., Middelkoop, H., Schüttrumpf, H., Vollmer, S., 204, 2014, 573-587. Today's sediment budget of the Rhine River channel, focusing on the Upper Rhine Graben and Rhenish Massif. Geomorphology. http://dx.doi.org/10. 1016/j.geomorph.2013.08.035. Gölz, E., 1992. Recent Morphologic Development of the Rhine River. 5th International Symposium on River Sedimentation, Karlsruhe. Gölz, E., 1994. Bed degradation — nature, causes, countermeasures. Water Sci. Technol. 3, 325–333. Gölz, E., Schröter, M., Mikos, M., 1995. Fluvial abrasion of broken quartzite used as a substitute for natural bed-load. In: Varma, C.V.J., Rao, A.R.G. (Eds.), Management of Sediment: Philosophy, Aims and Techniques. Sixth International Symposium on River Sedimentation, New Delhi, India, pp. 387–395. Gouw, M.J.P., Erkens, G., 2007. Architecture of the Holocene Rhine–Meuse Delta (The Netherlands) — a result of changing external controls. Neth. J. Geosci. (Geol. Mijnb.) 86 (1), 23–54. Jasmund, R., 1901. Die Arbeiten der Rheinstrom-Bauverwaltung 1851 bis 1900 — Denkschrift. Mittler, Berlin. Middelkoop, H., 1997. Embanked floodplains in The NetherlandsPublished PhD thesis , Utrecht University, Netherlands Geographical Studies 224Royal Dutch Geographical Society, Utrecht.

Rommel, J., 2005. Historische Entwicklung des Niederrheins und seiner Vorländer als Folge des dort seit 1934 betriebenen Bergbaus – Voruntersuchungsbericht zum Abschnitt Rhein-km 775 bis 814 im Zeitraum bis 1975. Bundesanstalt für Wasserbau, Karlsruhe. Rothe, P., 2000. Erdgeschichte – Spurensuche im Gestein. Wissenschaftliche Buchgesellschaft, Darmstadt. Schmidt, M., 2000. Hochwasser und Hochwasserschutzs in Deutschland vor 1850 – Eine Auswertung alter Quellen und Karten. Oldenbourg, München. Syvitski, J.P.M., 2003. Supply and flux of sediment along hydrological pathways: research for the 21st century. Glob. Planet. Chang. 39, 1–11. Ten Brinke, W.B.M., 1997. De bodemsamenstelling van Waal en IJsselin de jaren 1966, 1976, 1984 en 1995. Netherlands Institute for Inland Water Management and Waste Water Treatment, Arnhem. Ten Brinke, W., 2005. The Dutch Rhine, a restrained river. Veen Magazines, Diemen. Ten Brinke, W.B.M., Schoor, M.M., Sorber, A.M., Berendsen, H.J.A., 1998. Overbank deposits in relation to transport volumes during large-magnitude floods in the Dutch sandbed Rhine river system. Earth Surf. Process. Landf. 23, 809–824. Ten Brinke, W.B.M., Bolwidt, L.J., Snippen, E., Van Hal, L.W.J., 2001. Sedimentbalans Rijntakken 2000. Een actualisatie van de sedimentbalans voor slib, zand en grind van de Rijntakken in het beheersgebied van de Directie Oost-Nederland. RIZA rapportRijksinstituut voor Integraal Zoetwaterbeheer en Afvalwatermanagement, Arnhem. Tümmers, H.J., 1999. Der Rhein - Ein europäischer Fluss und seine Geschichte. 2. überarbeitete und aktualisierte Auflage. Beck, München. Walling, D.E., 2006. Human impact on land–ocean sediment transfer by the world's rivers. Geomorphology 79 (3–4), 192–216. Wenka, T., 2009. Untersuchungen zur langfristigen Lagestabilität der zum bergsenkungsbedingten Sohlausgleich eingebrachten Waschberge im Rheinstrom. AbschlussberichtBundesanstalt für Wasserbau, Karlsruhe.