Você está na página 1de 21

Journal of Hydrology 234 (2000) 228–248

www.elsevier.com/locate/jhydrol

Fitting and interpretation of sediment rating curves


N.E.M. Asselman 1
Department of Physical Geography, Marine and Atmospheric Research, The Netherlands Centre for Geo-Ecological Research (ICG),
Utrecht University, P.O. Box 80.115, 3508 TC Utrecht, The Netherlands
Received 28 September 1999; revised 28 April 2000; accepted 1 May 2000

Abstract
A large part of the sediment in lowland rivers is transported as wash load. As wash load is a non-capacity load it often is
modelled using empirical relations, such as the rating curve technique. Sediment rating curves in the form of a power function
are derived for several locations along the river Rhine and its main tributaries, using different fitting procedures. Inaccuracies in
estimated sediment loads are analysed, and spatial differences in the shapes of the fitted rating curves are related to watershed
characteristics.
Rating curves obtained by least squares regression on logarithmic transformed data underestimate long-term sediment
transport rates by 10–50%. Better estimates are obtained when nonlinear least squares regression is applied. The steepness
of the fitted rating curves decreases along the main channel of the Rhine in a downstream direction. Contrary to what is
generally believed this is not related to changes in the sediment transport regime. A better indication of the sediment transport
regime is obtained when the slope/intercept pairs of the fitted rating curves are plotted in a graph. All locations that plot on the
same line have a common discharge-concentration value and appear to be characterised by a similar sediment transport regime.
At locations of which the slope/intercept pairs plot on higher lines, a larger part of the annual sediment load is transported
during high discharge. 䉷 2000 Elsevier Science B.V. All rights reserved.
Keywords: Rhine River; Sediment transport; Suspension material; Modelling; Regression analysis

1. Introduction clay, i.e. wash load. It can thus be concluded that wash
load plays an important role in the sediment transport
In many lowland rivers a major part of the sediment in the river Rhine.
is transported in suspension. In the German lowland As the finest fraction of the suspended sediment
rivers suspended sediment makes up about 85% of the load often is a non-capacity load it cannot be predicted
total solid sediment load (Hinrich, 1974). In 1984, using stream power related sediment transport
93% of the total sediment load of the river Rhine models. Instead, empirical relations such as sediment
near the German–Dutch border was transported in rating curves often are applied. A sediment rating
suspension (Bundesanstalt für Gewässerkunde, curve describes the average relation between
1987). About 85% of this load consisted of silt and discharge and suspended sediment concentration for
a certain location. The most commonly used sediment
rating curve is a power function (e.g. Walling, 1974,
E-mail address: nathalie.asselman@wldelft.nl
1978):
(N.E.M. Asselman).
1
Present address: WL/DELFT HYDRAULICS, P.O. Box 177,
2600 MH Delft, The Netherlands. C ˆ aQb …1†
0022-1694/00/$ - see front matter 䉷 2000 Elsevier Science B.V. All rights reserved.
PII: S0022-169 4(00)00253-5
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 229

with C is suspended sediment concentration (mg/l), Q tributaries. The study has three objectives. The first
is water discharge (m 3/s), and a and b are regression objective is to determine spatial differences in the
coefficients. Eq. (1) covers both the effect of increased relation between discharge and suspended sediment
stream power at higher discharge and the extent to transport for different locations in the Rhine drainage
which new sources of sediment become available in basin and to relate these differences to differences in
weather conditions that cause high discharge. Despite the sediment transport regime. Secondly, inaccuracies
its general use several problems are recognised that in estimated sediment loads will be analysed and
regard the accuracy of the fitted curve as well as the related to the type of rating curve and the statistical
physical meaning of its regression coefficients. methodology applied to fit the rating curves. Finally,
Inaccuracies in predicted instantaneous suspended the study aims at determining the physical meaning of
sediment concentrations are related to the statistical the regression coefficients. Questions that need to be
method used to fit the sediment-rating curve and to the answered with regard to these objectives include
scatter about the regression line. Statistical inaccura- ‘Which rating curve produces the most accurate esti-
cies related to the fitting procedure are discussed by mates of the suspended sediment load at different
Ferguson (1986, 1987), Jansson (1985), Singh and locations in the Rhine drainage basin?’, ‘How does
Durgunoglu (1989), and Cohn et al. (1992). They the shape of the fitted curves differ between different
concluded that the sediment load of a river is likely locations?’, and ‘Do sediment rating curves provide
to be underestimated when concentrations are esti- information on the sediment transport regime of the
mated from water discharge using least squares river and if so, what kind of information can be
regression of log-transformed variables. Scatter derived from them?’
about the regression line is, among other things,
caused by variations in sediment supply due to, for
instance, seasonal effects, antecedent conditions in the 2. The Rhine drainage basin
river basin, and differences in sediment availability at
The Rhine is one of the largest rivers in north
the beginning or the ending of a flood. This is not
west Europe. Its drainage basin is situated between
accounted for by the rating curve.
the Alps and the North Sea (Fig. 1). The drainage
As a sediment rating curve can be considered a
area upstream of the German–Dutch border is
‘black box’ type of model, the coefficients a and b
about 165 000 km 2 (CHR/KHR, 1976). The total
in Eq. (1) have no physical meaning. Nevertheless,
length of the river is about 1320 km. The Rhine
some physical interpretation is often ascribed to
basin is characterised by a temperate climate.
them. Peters-Kümmerly (1973) and Morgan (1995)
Mean annual precipitation varies between 600 mm
state that the a-coefficient represents an index of
in the lower downstream parts and 2500 mm in the
erosion severity. High a-values indicate intensively
Alps (CHR/KHR, 1976). Average annual discharge
weathered materials, which can easily be transported.
increases from about 370 m 3/s at the outlet of the
According to Peters-Kümmerly (1973), the b-coeffi-
Bodensee, to about 2300 m 3/s near Rees at the
cient represents the erosive power of the river, with
German–Dutch border. The average annual
large values being indicative for rivers where a small
suspended sediment load of the river Rhine at
increase in discharge results in a strong increase in
Rees is 3.14 × 10 6 ton.
erosive power of the river. Others state that the b-
coefficient indicates the extent to which new sediment
sources become available when discharge increases. 3. Discharge and suspended sediment data
Several authors compare the values of the b-coeffi-
cient obtained for different rivers to discuss differ- Daily discharge and suspended sediment concen-
ences in sediment transport characteristics in the tration measurements have been carried out by the
different basins (Peters-Kümmerly, 1973; Walling, Bundesanstalt für Gewässerkunde (BfG) in Germany
1974; Sarma, 1986; Morgan, 1995; Kern, 1997). at several locations along the river Rhine and its main
In this study sediment rating curves are derived for tributaries. Suspended sediment concentrations are
different locations along the river Rhine and its main determined from samples collected about 1 m
230 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Fig. 1. Measurement locations in the Rhine drainage basin.

below the water surface. Details on the measurement 4. Method


methods are provided by Deutscher Verband für
Wasserwirtschaft und Kulturbau (1986). Data were 4.1. Fitting of sediment rating curves
provided by the BfG for the measurement locations
indicated in Fig. 1. The data records vary in length, In most studies, a power function is used, in which
but at most locations measurements have been the regression coefficients a and b are obtained by
carried out since the early or mid seventies (Table ordinary least squares regression on logarithms of
1). Average water discharge and average annual sedi- concentration and discharge data (e.g. Walling,
ment loads are also given in Table 1. Sediment loads 1974, 1977a,b, 1978; Church and Gilbert, 1975;
are computed by multiplication of daily river Mossa, 1989):
discharge values and daily suspended sediment
concentrations. log C ˆ log a ⫹ b log Q ⫹ log 1 …2†
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 231

Table 1
Measurement periods, drainage area, annual average discharge and suspended sediment transport of various locations along the river Rhine and
its tributaries

Location Period Area (km 2) Discharge (m 3/s) Sediment load (10 6 t/year)

Rheinfelden (Rhine) 1977–1990 34 550 1100 1.12


Maxau (Rhine) 1975–1990 50 200 1300 1.16
Worms/Nierstein (Rhine) 1984–1990 68 827 1450 1.40
Kaub/Bacharach (Rhine) 1971–1990 103 730 1688 2.16
Andernach/Weissenthurm (Rhine) 1975–1990 139 800 2219 3.48
Rees/Emmerich (Rhine) 1975–1991 159 300 2386 3.14
Rockenau (Neckar) 1972–1990 12 676 142 0.47
Schweinfurt/Viereth (Main) 1973–1989 12 715 108 0.10
Kleinheubach (Main) 1987–1990 21 505 191 0.34
Kalkofen (Lahn) 1971–1990 5305 46 0.08
Hauconcourt (Mosel) 1975–1980 9400 131 0.24
Cochem/Brodenbach (Mosel) 1982–1990 27 088 384 0.88

which is then back-transformed to obtain: method is inappropriate since the assumption of


constant variance or scatter of the dependent variable,
C ˆ aQb 1 …3† homoscedasticity, often is not met. The degree of
scatter in a graph of sediment concentrations plotted
in which 1 is a lognormally distributed error. Statis-
against discharge usually increases with discharge.
tical considerations show that the sediment load of a
Also, the estimate of the variance of the normal errors
river is likely to be underestimated when concentra-
will often be biased because there can be no negative
tions are estimated from water discharge, using least
concentration values.
squares regression of log-transformed variables
As both fitting procedures have their own statistical
(Ferguson, 1986, 1987; Jansson, 1985; Singh and
shortcomings, they are both applied and evaluated on
Durgunoglu, 1989; Cohn et al., 1992). To correct for
the measurement locations along the main channel of
this underestimation, or bias, several forms of bias
the Rhine. First, a regression was carried out on the
correction factors have been suggested (Ferguson,
logarithmic transformed data Eq. (2). Then, two
1986, 1987; Jansson, 1985). Since the degree of
power functions were fitted through the data, using
underestimation is proportional to the degree of scat-
nonlinear regression. The first power function is
ter about the regression, Ferguson (1986) proposed
given in Eq. (5):
the following bias correction factor (CF):
C ˆ aQ b ⫹ d …5†
CF ˆ exp…2:651S2 † …4†
where d is a normally distributed error. The second
where S 2 is the mean square error of the log-trans- power function consists of a power function with an
formed regression (in log-10 units). It is important additive constant term (p):
to notice that application of this correction factor
only results in unbiased estimates of suspended C ˆ p ⫹ aQb ⫹ d …6†
sediment concentrations when the residuals of C are Finally, four sets of rating relationships were
log-normally distributed and C is a power function of constructed, using data sets subdivided after season
Q. This assumption often is not met. and changes in discharge (i.e. summer and winter
An alternative approach is to assume a power func- months, and rising or falling limbs of the hydrograph).
tion with additive error, which consists of random,
normally distributed white noise, with zero mean 4.2. Selection of the calibration period
and variance s 2. In this case unbiased estimates of a
and b can be obtained using a nonlinear least squares A sediment rating curve can only be regarded as
regression procedure. However, in statistical terms the representative for a certain location under the present
232 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

range of environmental and climate conditions, when measurement periods provides an indication of the
the relationship between discharge and suspended robustness of sediment rating curves under slightly
sediment concentration is consistent over the entire different discharge regimes.
measurement period. Homogeneity of the data was
checked using a double-mass curve approach (Nordin
and Sabol, 1973). For most stations except for 5. Results
Rheinfelden and Basel, the double-mass curve is an
almost straight line, indicating that the relation 5.1. Fitted rating curves
between sediment and water discharge is consistent
for the period of record. The bend in the double-mass Sediment rating curves were fitted for locations
curve of Rheinfelden was hardly visible. Also, no along the river Rhine and its tributaries. The regres-
inconsistency occurred during the first 8 years of the sion coefficients of the fitted curves for locations
measurement period. Therefore the data were along the main channel are given in Table 2. Although
assumed applicable. The data of Basel were not correlation coefficients for some curves are low, they
used for further analyses. are all significant at the 0.05 level, because the
As sediment rating curves should be based on a number of data points is large. Several of the regres-
wide range of discharges including a sufficient sions given in Table 2 have negative p-values Eq. (6).
number of measurements carried out at high This may suggest that when discharges are extremely
discharge, it was decided to fit the rating curves on low negative concentration estimates can be produced.
a relatively wet period. A relatively wet period with However, during the selected periods no flows of less
discharge exceeding the long-term average discharge than the critical discharge levels occurred.
was selected after examination of the cumulative Large differences in regression coefficients are
monthly mean discharge anomalies, as described by found for different fitting procedures as well as for
Lozowski et al. (1989). At all gauging stations a simi- the different calibration periods. To facilitate compar-
lar trend was observed, with low discharges in 1976, ison, some fitted curves obtained for locations along
and during the period 1989–1991. Relatively high the main channel of the Rhine are shown in Fig. 2. The
discharges occurred during the periods 1979–1983 rating curves based on least squares regression of the
and 1986–1988. The period 1979–1983 was the long- log-transformed data seem to underestimate concen-
est wet period. Also, during this period no inconsis- tration values at high discharge (curve 1). The degree
tencies were found in the data. Therefore it was of underestimation decreases when the bias correction
decided to develop a complete set of sediment rating factor is used (curve 2), or when the rating curve is
curves for the period January 1979 until December obtained by fitting a power function based on
1983. All fitting procedures and bias correction nonlinear least squares regression (curve 3). However,
factors described previously Eqs. (1), (2), (4), and at gauging stations where suspended sediment
(5) were applied on the 9 sub samples of the data concentrations are relatively high at low discharge
set. The 9 sub samples are (1) all data, (2) summer and increase only slightly with increasing discharge,
and (3) winter months, (4) rising discharge and (5) none of the first three methods fit the data well. This is
falling discharge, (6) rising discharge during summer, best shown by the data obtained from the gauging
(7) falling discharge during summer, (8) rising station near Rees. This shortcoming can be overcome
discharge during winter, and (9) falling discharge by fitting a rating curve in the form of a power func-
during winter. As 4 rating curves were fitted on each tion, based on nonlinear least squares regression, with
data set, this resulted in a total of 36 curves per addition of a constant term (curve 4). At all locations,
location. Because data records for some of the main the rating curves obtained by fitting a power function
tributaries start after 1979, sediment rating curves (curves 3 and 4) visually result in the best fit.
were also fitted for a 5 year period including the Fig. 3 shows sediment rating curves for locations
second wet period (1986–1990), and for all available along the river Rhine as well as its main tributaries,
data, i.e. the period 1975–1990 for most gauging fitted by means of nonlinear least squares regression
stations. Comparison of the curves fitted on different (curve 4). The rating curves are based on all data
Table 2
Regression coefficients of rating curves fitted for gauging stations along the river Rhine (callibration period: 1979-1983, CF ˆ correction factor, N ˆ number of data points, fall/
rise ˆ falling/rising discharge)

Location Data Log-transformed data Nonlinear Nonlinear N


log C ˆ log a ⫹ b × log Q C ˆ a × Qb C ˆ p ⫹ a × Qb

10 a b CF r2 a b r2 p a b r2

Rheinfelden All 3.35 × 10 ⫺03 1.21 1.18 0.34 7.70 × 10 ⫺04 1.44 0.27 13.7 8.00 × 10 ⫺14 4.43 0.35 1764
Winter 2.42 × 10 ⫺03 1.25 1.21 0.36 1.22 × 10 ⫺06 2.34 0.36 10.7 2.12 × 10 ⫺11 3.73 0.40 849
Summer 6.44 × 10 ⫺03 1.12 1.16 0.24 1.31 × 10 ⫺04 1.67 0.24 17.1 2.10 × 10 ⫺17 5.47 0.30 915
Fall 6.60 × 10 ⫺03 1.10 1.17 0.30 7.40 × 10 ⫺04 1.43 0.22 14.1 1.10 × 10 ⫺15 5.70 0.32 1473
Rise 1.17 × 10 ⫺03 1.38 1.20 0.36 4.26 × 10 ⫺06 2.16 0.34 16.4 1.70 × 10 ⫺11 3.73 0.36 291

N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248


Winter-rise 1.71 × 10 ⫺03 1.33 1.23 0.37 1.10 × 10 ⫺05 2.05 0.33 15.7 4.70 × 10 ⫺10 3.32 0.34 140
Winter-fall 5.52 × 10 ⫺03 1.13 1.20 0.30 1.71 × 10 ⫺07 2.60 0.33 11.9 3.92 × 10 ⫺18 5.78 0.48 709
Summer-rise 3.08 × 10 ⫺04 1.55 1.17 0.37 2.81 × 10 ⫺06 2.21 0.37 19.3 5.44 × 10 ⫺16 5.06 0.39 151
Summer-fall 1.92 × 10 ⫺02 0.96 1.15 0.19 3.97 × 10 ⫺03 1.20 0.15 16.8 1.06 × 10 ⫺16 5.24 0.19 764
All 75-90 8.84 × 10 ⫺03 1.10 1.28 0.28 1.41 × 10 ⫺03 1.40 0.18 14.8 3.38 × 10 ⫺08 2.75 0.19 5047
Maxau All 8.09 × 10 ⫺03 1.07 1.13 0.37 2.92 × 10 ⫺03 1.22 0.37 7.9 3.90 × 10 ⫺05 1.75 0.38 1826
Winter 5.08 × 10 ⫺03 1.13 1.14 0.43 3.87 × 10 ⫺04 1.49 0.49 12.3 1.50 × 10 ⫺09 3.03 0.53 911
Summer 3.33 × 10 ⫺02 0.88 1.11 0.47 2.40 × 10 ⫺02 0.94 0.23 ⫺7.4 1.38 × 10 ⫺01 0.74 0.23 915
Fall 1.32 × 10 ⫺02 1.00 1.13 0.32 7.29 × 10 ⫺03 1.09 0.32 4.4 9.10 × 10 ⫺04 1.35 0.32 1455
Rise 5.13 × 10 ⫺03 1.14 1.12 0.40 1.23 × 10 ⫺03 1.35 0.40 16.6 1.70 × 10 ⫺08 2.70 0.42 371
Winter-rise 2.83 × 10 ⫺03 1.23 1.14 0.45 4.49 × 10 ⫺04 1.49 0.43 22.3 4.40 × 10 ⫺17 5.16 0.48 189
Winter-fall 1.06 × 10 ⫺02 1.02 1.13 0.36 8.50 × 10 ⫺04 1.38 0.48 11.4 2.86 × 10 ⫺09 2.95 0.54 722
Summer-rise 1.52 × 10 ⫺02 1.00 1.09 0.32 6.15 × 10 ⫺04 1.13 0.37 14.7 2.00 × 10 ⫺06 2.07 0.38 182
Summer-fall 5.37 × 10 ⫺02 0.81 1.12 0.18 4.91 × 10 ⫺02 0.84 0.17 ⫺161.7 7.97 × 10 ⫹01 0.12 0.18 733
All 75-90 1.48 × 10 ⫺02 1.01 1.15 0.36 2.83 × 10 ⫺03 1.25 0.41 11.1 4.35 × 10 ⫺06 2.05 0.43 6148

Kaub/Bacharach All 8.30 × 10 ⫺03 1.06 1.11 0.44 3.70 × 10 ⫺03 1.19 0.36 3.8 1.25 × 10 ⫺03 1.31 0.36 1826
Winter 2.05 × 10 ⫺03 1.24 1.15 0.52 1.32 × 10 ⫺03 1.32 0.39 ⫺1.5 2.00 × 10 ⫺03 1.27 0.39 911
Summer 1.57 × 10 ⫺01 0.68 1.06 0.28 3.15 × 10 ⫺02 0.90 0.31 16.9 3.10 × 10 ⫺05 1.69 0.33 915
Fall 4.54 × 10 ⫺03 1.17 1.09 0.46 9.80 × 10 ⫺03 1.09 0.35 10.9 1.20 × 10 ⫺04 1.55 0.33 1416
Rise 2.40 × 10 ⫺02 0.91 1.13 0.38 1.30 × 10 ⫺02 1.01 0.32 ⫺44.2 7.11 × 10 ⫺01 0.63 0.36 410
Winter-rise 2.79 × 10 ⫺03 1.24 1.17 0.47 1.67 × 10 ⫺02 1.04 0.31 ⫺132.5 1.44 × 10 ⫹01 0.33 0.33 238
Winter-fall 5.68 × 10 ⫺03 1.09 1.12 0.49 8.98 × 10 ⫺04 1.35 0.44 9.8 4.47 × 10 ⫺06 1.97 0.68 673
Summer-rise 1.71 × 10 ⫺02 0.99 1.07 0.46 5.81 × 10 ⫺03 1.14 0.55 6.8 1.31 × 10 ⫺03 1.31 0.55 171
Summer-fall 5.77 × 10 ⫺01 0.50 1.05 0.18 5.00 × 10 ⫺01 0.53 0.14 10.1 3.81 × 10 ⫺02 0.81 0.14 744
All 71-90 5.35 × 10 ⫺02 0.85 1.16 0.30 1.89 × 10 ⫺02 1.01 0.25 7.8 2.60 × 10 ⫺03 1.23 0.25 7305
Andernach Weissent All 1.61 × 10 ⫺02 0.99 1.09 0.49 1.72 × 10 ⫺03 1.29 0.44 6.6 4.24 × 10 ⫺04 1.44 0.44 1826
Winter 4.32 × 10 ⫺03 1.15 1.12 0.59 6.46 × 10 ⫺04 1.41 0.44 1.7 4.58 × 10 ⫺04 1.45 0.44 911
Summer 1.87 × 10 ⫺01 0.67 1.07 0.28 9.22 × 10 ⫺03 1.07 0.50 22.1 8.54 × 10 ⫺06 1.94 0.54 915
Fall 5.34 × 10 ⫺02 0.83 1.09 0.41 8.34 × 10 ⫺03 1.07 0.41 11.4 3.60 × 10 ⫺04 1.42 0.42 1376
Rise 3.52 × 10 ⫺03 1.20 1.13 0.56 2.38 × 10 ⫺03 1.27 0.45 ⫺20.3 2.10 × 10 ⫺02 1.03 0.45 450

233
234
Table 2 (continued)

N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248


Location Data Log-transformed data Nonlinear Nonlinear N
log C ˆ log a ⫹ b × log Q C ˆ a × Qb C ˆ p ⫹ a × Qb

10 a b CF r2 a b r2 p a b r2

Winter-rise 1.21 × 10 ⫺03 1.34 0.14 0.61 1.71 × 10 ⫺03 1.31 0.42 ⫺29.1 3.02 × 10 ⫺02 1.01 0.43 260
Winter-fall 1.52 × 10 ⫺02 0.98 1.10 0.55 2.58 × 10 ⫺03 1.22 0.44 5 6.87 × 10 ⫺04 1.37 0.44 651
Summer-rise 4.64 × 10 ⫺02 0.86 1.09 0.64 4.21 × 10 ⫺03 1.17 0.63 21.6 2.70 × 10 ⫺05 1.72 0.64 190
Summer-fall 4.59 × 10 ⫺01 0.55 1.06 0.20 4.36 × 10 ⫺02 0.87 0.36 21.7 2.10 × 10 ⫺05 1.71 0.41 725
All 75-90 3.08 × 10 ⫺02 0.91 1.13 0.40 2.99 × 10 ⫺03 1.22 0.40 9.6 2.89 × 10 ⫺04 1.48 0.64 5844

Rees All 3.36 × 10 ⫹00 0.32 1.08 0.11 7.93 × 10 ⫺01 0.51 0.16 34.3 4.00 × 10 ⫺06 1.86 0.20 1826
Winter 1.64 × 10 ⫹00 0.53 1.09 0.28 1.89 × 10 ⫺01 0.69 0.27 21.2 1.22 × 10 ⫺03 1.24 0.27 911
Summer 6.90 × 10 ⫹01 ⫺0.07 1.05 0.01 1.66 × 10 ⫹01 0.12 0.10 41.5 2.90 × 10 ⫺13 3.61 0.12 915
Fall 1.39 × 10 ⫹01 0.13 1.06 0.02 8.51 × 10 ⫹00 0.20 0.04 37.4 5.20 × 10 ⫺08 2.25 0.06 1392
Rise 2.84 × 10 ⫺01 0.64 1.10 0.31 1.08 × 10 ⫺01 0.78 0.31 12.5 1.82 × 10 ⫺02 0.96 0.31 434
Winter-rise 1.26 × 10 ⫺01 0.75 1.10 0.38 1.17 × 10 ⫺01 0.77 0.31 ⫺18 5.85 × 10 ⫺01 0.61 0.32 270
Winter-fall 2.15 × 10 ⫹00 0.36 1.07 0.19 9.70 × 10 ⫺01 0.47 0.18 26.4 1.80 × 10 ⫺04 1.40 0.20 641
Summer-rise 2.36 × 10 ⫹00 0.37 1.08 0.13 1.25 × 10 ⫺01 0.74 0.27 35.2 1.00 × 10 ⫺06 1.99 0.32 164
Summer-fall 3.06 × 10 ⫹02 ⫺0.26 1.04 0.09 2.48 × 10 ⫹02 ⫺0.23 0.06 35.7 5.15 × 10 ⫹06 ⫺1.79 0.07 751
All 75-91 2.62 × 10 ⫹00 0.33 1.12 0.09 6.85 × 10 ⫺01 0.52 0.15 29.3 1.96 × 10 ⫺06 1.93 0.44 6148
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 235

Fig. 2. Sediment rating curves established for gauging stations along the river Rhine, using different fitting procedures.

without subdivision after season or discharge stage. nonlinear least squares regression yields the best esti-
The steepest rating curves are observed in the tribu- mates of suspended sediment concentrations (curves 3
taries of the river Rhine. Along the main channel of and 4), the distribution of the concentration data indi-
the river Rhine the steepest rating curve is found near cate that linear regression on logarithmic transformed
Rheinfelden. The steepness decreases in a down- data might produce better results. It can easily be seen
stream direction. The flattest rating curve is observed in the scatter plots of Fig. 2 that the concentration data
near Rees. is not normally distributed and that the variance is not
constant with discharge. Also, analyses of the resi-
5.2. Model efficiency of the fitted rating curves duals of these models indicate that they are not
normally distributed either. Best ‘unbiased’ estimates
Although the scatter plots in Fig. 2 suggest that of a and b would thus be obtained by least squares
236 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Fig. 3. Sediment rating curves for locations along the Rhine and its main tributaries fitted using nonlinear least squares regression.

regression on logarithmic transformed data in combi- residual variance given by


nation with the correction factor Eq. (4). However,
X
N
when the data distribution is studied in more detail F02 ˆ  2
…Ct ⫺ C† …8†
with the Kolmogorov–Smirnov test (Blalock, 1981), tˆ1
it appears that the data is not log-normally distributed
either. Thus, as statistical considerations indicate that X
N
both fitting procedures are inappropriate for these data F02 ˆ …Ct ⫺ C 0t †2 …9†
sets, it can be suggested that the rating curve with the tˆ1

highest model efficiency should be used for further in which Ct and C 0 t are measured and computed
computations. concentrations at time t, and C is the average
The efficiency of the different rating curve models measured concentration. The model efficiency R 2
in estimating time series of suspended sediment loads should not be confused with the correlation
was evaluated using the model efficiency criterion coefficient, which indicates the degree of inter-
(R 2) as defined by Nash and Sutcliffe (1970). R 2 relation between two or more variables or between
is a criterion that determines the efficiency of a estimated and measured values of a single
model in comparison with the average value. It is variable. Instead, it shows whether the applied
computed as: model provides better estimates than application
of the average value. Values of R 2 range between
F02 ⫺ F 2 minus infinity and plus 1, where plus 1 indicates
R2 ˆ …7†
F02 perfect agreement between measured and computed
sediment transport rates. Negative values indicate
in which F02 is the initial variance, and F 2 is the that the average measured concentration value is a
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 237

Table 3
Model efficiency R 2 (Nash and Sutcliffe, 1970) of different sediment rating curves in estimating suspended sediment transport rates (data
measured during the period 1979–1983 were used for model calibration)

Rees Andernach Kaub Maxau Rheinfelden

Log, all data 0.50 0.56 0.60 0.57 0.32


Log × CF, all data 0.52 0.59 0.63 0.60 0.36
Power, all data 0.57 0.64 0.63 0.61 0.41
Power ⫹ C, all data 0.58 0.64 0.63 0.62 0.51
Power ⫹ C, sum/win 0.59 0.65 0.65 0.69 0.53
Power ⫹ C, fall/rise 0.65 0.68 0.68 0.65 0.45
Power ⫹ C, sum/win ⫹ fall/rise 0.66 0.69 0.71 0.72 0.57

better estimate of instantaneous suspended sediment regression model. Hence, differences in the residual
concentrations than the estimates obtained by the sum of squares predominate over the penalty for the
model. As R 2 depends on the initial variance in number of model parameters. Consequently, the best
measured sediment concentrations, which may be rating curve model is a power function with additional
quite different for different time periods or different constant term, based on nonlinear least squares regres-
gauging stations, values of R 2 can only directly be sion on concentration and discharge data, subdivided
compared for different model results of sediment after season and discharge stage.
transport time series, obtained for the same gauging
station, and for the same validation period. 5.3. Model bias and accuracy
Evaluation of the R 2-factors (Table 3) shows that
for most gauging stations the rating curves based on Validation of the fitted rating curves also indicates
least squares linear regression of the logarithmic that rating curves fitted using nonlinear least squares
transformed data have the lowest efficiency. The effi- regression yield the most accurate estimates of long-
ciency increases when the correction factor is applied. term average suspended sediment loads. The model
The best results are obtained when sediment-rating validation results of 4 locations along the river Rhine
curves in the form of a power function with additional and 1 location along a major tributary are summarised
constant term are used. The efficiency also increases in Table 4. When no statistical bias occurs, sediment
when the data is subdivided after season and discharge loads computed for the calibration period should
stage. This, however, is to be expected as addition of equal measured sediment loads derived by multiplica-
parameters to a model usually improves the fit. A tion of daily discharge and suspended sediment
criterion for selecting reasonable models in regression concentration values. As variations in discharge and
analysis that embraces a penalty for increased suspended sediment concentrations during a single
complexity is Akaike’s Information Criterion (AIC). day are small, it is expected that errors in calculated
It is estimated as (Webster and McBratney, 1989): measured loads are small. When the rating curves
    fitted for Rees are applied to compute sediment
2p
A ˆ n ln ⫹ n ⫹ 2 ⫹ n ln R ⫹ 2m …10† loads over the period used for model calibration it
n
appears that the rating curve based on logarithmic
in which n is the number of observations, m the transformed data underestimates the long-term sedi-
number of model parameters, and R the residual ment load up to 12% (Table 4). When a power func-
sum of squares. Differences in model performance tion based on nonlinear least squares regression is
depend on differences in the residual sum of squares applied an error of less than 1% is attained. At loca-
and the number of model parameters. The number of tions that are characterised by steeper rating curves
model parameters only has a significant effect on the the differences in bias are even larger. At Kalkofen for
AIC when regression is carried out on a limited example, rating curves fitted on logarithmic trans-
number of observations. In this study, however, formed data tend to underestimate sediment loads by
more than 1800 observations were used to fit the 56%, whereas the rating curves based on nonlinear
238 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

least squares regression overestimate sediment loads Table 4


by 0.2% or less (Table 4). When sediment-rating Differences (%) between estimated and measured sediment loads
for different types of sediment rating curve (types of rating curve:
curves are used to compute sediment transport during (1) logarithmic; (2) logarithmic with correction factor; (3)
single years-larger errors may occur. At Andernach nonlinear; and (4) nonlinear with constant. Rating curves are cali-
for instance (not shown in Table 4) errors in annual brated and validated using the same data series (1979–1983))
sediment loads computed with a rating curve based on
Data Types of rating curve
nonlinear least squares regression, varied from ⫹33%
in 1983 to ⫺39% in 1976, whereas at Kaub errors 1 2 3 4
ranged from ⫺61% in 1971 to ⫹23% in 1983.
Rees (Rhine)
All ⫺12.1 ⫺5.2 ⫺0.4 4.1
Subdivision after season ⫺11.5 ⫺5.0 ⫺0.7 0.6
6. Discussion
Subdivision after stage ⫺9.0 ⫺1.8 0.5 1.0
Subdivision after season ⫺8.7 ⫺2.1 0.1 0.1
6.1. Interpretation of sediment rating curves and stage
Andernach (Rhine)
As discussed previously, the values of the regres-
All ⫺18.5 ⫺11.4 ⫺0.1 0.2
sion coefficients of sediment rating curves are Subdivision after season ⫺18.3 ⫺9.7 ⫺0.0 2.6
assumed to depend on the erosion severity, or the Subdivision after stage ⫺14.3 ⫺5.0 1.4 1.2
availability of sediment in a certain area, the power Subdivision after season ⫺14.1 ⫺5.2 ⫺2.0 0.8
of the river to erode and transport the available mate- and stage
rial, and on the extent to which new sediment sources Kaub (Rhine)
become available in weather conditions that cause All ⫺13.6 ⫺4.2 0.5 ⫺0.0
high discharge. According to Peters-Kümmerly Subdivision after season ⫺13.8 ⫺4.5 ⫺0.1 0.2
Subdivision after stage ⫺4.1 5.8 10.5 3.7
(1973) and Morgan (1995), the erodibility of the
Subdivision after season ⫺9.0 0.7 2.7 2.4
soils is represented by the a-coefficient Eq. (1). High and stage
values of the a-coefficient occur in areas characterised
Maxau (Rhine)
by intensively weathered materials, which can easily
All ⫺13.1 ⫺7.1 ⫺0.1 ⫺0.5
be eroded and transported. The b-coefficient repre- Subdivision after season ⫺13.8 ⫺3.2 ⫺0.8 ⫺0.1
sents the erosive power of the river, with large values Subdivision after stage ⫺11.2 ⫺0.3 0.9 1.6
being indicative for rivers with a strong increase in Subdivision after season ⫺11.9 ⫺1.3 0.1 ⫺0.8
erosive power and in sediment transport capacity and stage
when discharge increases. According to Walling Kalkofen (Lahn)
(1974) b-values are also affected by the grain size All ⫺56.0 ⫺40.0 ⫺0.0 0.2
distribution of the material available for transport,
i.e. in streams characterised by sand sized sediments
the power of the stream to transport sediment will be concentrations, indicating that either the power of
more important than in streams that mainly transport the river to erode material during high discharge peri-
silt and clay. This will result in high b-values. ods is high, or that important sediment sources
However, as the a- and b-coefficients of sediment become available when the water level rises. Flat
rating curves are inversely correlated (Rannie, 1978; rating curves should be characteristic for river
Thomas, 1988) it seems more appropriate to use the sections with intensively weathered materials or
steepness of the rating curve, which is a combination loose sedimentary deposits, which can be transported
of the a- and b-values, as a measure of soil erodibility at almost all discharges.
and erosivity of the river. Steep rating curves, i.e. low When this line of reasoning is accepted for the river
a- and high b-values, should thus be characteristic for Rhine, the following interpretation can be assigned to
river sections with little sediment transport taking the rating curves shown in Fig. 3. The steepest rating
place at low discharge. An increase in discharge curves are found in the tributaries of the river Rhine.
results in a large increment of suspended sediment This suggests a limited amount of fine sediment,
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 239

Table 5
Importance of high and low discharges in transporting suspended sediment at several locations along the Rhine and its main tributaries

Qavg (m 3/s) a Qavg% (% time) b Qs,avg (% Qs) c Q10% (m 3/s) d Qs,10% (%Qs) e

Rockenau (Neckar) 142 69 9 264 77


Kalkofen (Lahn) 46 71 10 98 77
Hauconcourt (Mosel) 131 67 12 297 66
Cochem (Mosel) 384 68 10 830 70
Schweinfurt (Main) 108 68 24 204 53
Kleinheubach (Main) 191 70 20 371 69
Rheinfelden (Rhine) 1100 57 23 1708 35
Maxau (Rhine) 1300 58 26 2020 36
Worms (Rhine) 1450 60 26 2300 37
Kaub (Rhine) 1688 60 27 2630 38
Andernach (Rhine) 2219 61 23 3569 46
Rees (Rhine) 2386 61 34 3896 34
a
Qavg, annual average discharge.
b
Qavg%, percentage of time during which discharges are lower than Qavg.
c
Qs,avg, percentage of the total annual sediment load transported at discharges below Qavg.
d
Q10%, discharge that is exceeded 10% of the time, i.e. about 36 days per year.
e
Qs,10%, percentage of the total annual sediment load transported at discharges higher than Q10%.

which can be picked up from the bed at low discharge. shown in Table 5 the sediment transport regime varies
Once a certain discharge threshold is exceeded, sedi- little between different locations along the main chan-
ment supply to the river increases, and sediment can nel of the river Rhine. At all locations along the main
be picked up from the riverbed, resulting in a rapid stream, about 25–30% of the annual sediment load is
increase in suspended sediment concentrations. The transported at discharge levels below the long-term
presence of weirs in most tributaries will also result average discharge. Little over 35% of the annual sedi-
in steep rating curves. During low discharge, much ment load is transported at very high discharges,
suspended sediment will settle behind the weirs. which occur only 10% of the time (i.e. about 36
When discharge increases this temporary deposited days per year). In the tributaries low discharges are
sediment will be flushed out, resulting in a large less important in transporting suspended sediment,
increase in suspended sediment concentrations. while high discharges are more important. In the
Along the river Rhine, the steepness of the rating Main and the Mosel, about 10–20% of the annual
curves decreases in a downstream direction. The flat- load is transported at discharges below the annual
test rating curve is found near Rees, indicating that average discharge, whereas more than 60% of the
near the German–Dutch border large quantities of fine annual load is transported at high discharges which
material are available for transport at low discharge. occur only 10% of the time. At Rockenau (Neckar)
This argument leads to the hypothesis that steeper and Kalkofen (Lahn) about 77% of the annual load is
rating curves are indicative of rivers, or river sections, transported during these extreme and infrequent
where most of the sediment transport takes place at discharges. High discharge can thus be considered
high discharge. To test this hypothesis, the sediment more important in the tributaries than in the Rhine
transport regime of the river Rhine and its tributaries itself.
was studied. To describe the regime, Qs,avg and Qs,10% Although the steepness of the rating curves
were computed. Qs,avg is defined as the percentage of decreases along the Rhine in a downstream direction,
the total annual sediment load that is transported at no differences in the sediment transport regime are
discharges below the average discharge (Qavg). Qs,10% observed. This contradicts the previous interpretation
represents the percentage of the total annual sediment of the sediment rating curves. It can thus be concluded
load that is transported at very high discharges that are that differences in steepness of sediment rating curves
exceeded 10% of the time, i.e. 36 days per year. As do not necessarily indicate differences in the sediment
240 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Fig. 4. Correlation between slope/intercept values of sediment rating curves fitted using nonlinear least squares regression.

transport regime. Instead, the difference in steepness with a rating curve fitted through a different part or
of the rating curve observed in Fig. 3 may be related to subset of the data (see also Table 2).
scale. For instance, an increase in discharge of It is striking that all coefficients obtained for loca-
100 m 3/s can be regarded as a relatively larger tions along the main channel plot as a single line,
increase near Rheinfelden than near Rees, as the aver- whereas the coefficients obtained for the tributaries
age discharge is much higher at the latter location. As plot above this line. Three parallel lines can be
a result, a stronger increase in suspended sediment drawn. The lower line coincides with the coefficient
concentrations will be observed near Rheinfelden values obtained for the main channel. The upper line
than near Rees. High discharges can however, still coincides with sample locations in the river Neckar at
be of equal importance in transporting suspended Rockenau, and in the river Lahn at Kalkofen. In
sediment at both locations. The question that remains between these lines a third line can be drawn for
is whether another method exists to determine the measurement locations in the Mosel at Hauconcourt
importance of high discharges in transporting and Cochem, and in the Main near Schweinfurt and
suspended sediment using the shape of sediment Kleinheubach.
rating curves. A negative correlation between the regression coef-
ficients has been observed more often (e.g. Rannie,
1978; Thomas, 1988). Thomas (1988) states that this
6.2. Correlation between the regression coefficients is due to the fact that a linear regression line fits
through the average values of both variables, here
To study the effect of the sediment transport regime log-discharge (Q) and log-suspended sediment
on the shape of the sediment rating curve, the regres- concentration (C). Data sets that have similar means
sion coefficients a and b Eq. (1), obtained by nonlinear of log-Q and log-C will result in rating curves that
least squares regression, were plotted in a graph have this one point in common. The slope of the fitted
(Fig. 4). Per location up to 10 combinations of a- curves can then be expressed as a linear function of
and b-values are plotted, each of which corresponds the intercepts with parameters that depend on the
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 241

co-ordinates of the common point. In slope/intercept i.e. relief, drainage area, and drainage density. The
plots these points will result in a straight line. position of the rating curve coefficients on this line,
Thomas (1988) sub-sampled a data record of however, varies in time and therefore must mainly be
discharge and suspended sediment concentrations at related to watershed characteristics that also vary in
a single gauging station and fitted rating curves time, such as average or maximum discharge, and
through the simulated data sets. However, as all rating sediment availability. However, no such relation-
curves are based on the same period of record it can be ships could be identified in this dataset. For instance,
expected that the simulated data sets will have similar drainage area increases along the Rhine in a down-
discharge and suspended sediment concentration stream direction, while relief and channel slope
values. Hence, the rating curves fitted through the decrease, still all measurement locations plot on the
data sets will have one point in common and the same line.
slope/intercept parameters will plot on a straight line. As the sample locations along the main channel of
For the river Rhine different data sets were used, the river Rhine are characterised by a similar sediment
measured at different locations. Still the slope inter- transport regime, it can be argued that the common
cept values plot on a single straight line, indicating point, and hence the position of the line plotted in
that the rating curves have one common discharge- Fig. 4, provides an indication of the sediment trans-
concentration value, which can be regarded as a port regime. In other words, all locations that plot on
fulcrum around which the rating curves rotate. To the same line have a similar sediment transport
study this common point in more detail, sediment- regime, which can be characterised by the communal
rating curves were developed per calendar year. The point. The locations that plot on the upper left side of
regression coefficients obtained by least squares this line have a steeper rating curve than the locations
regression on the logarithmic transformed variables, that plot on the lower right side. This probably is
were plotted in a graph similar to the graph in Fig. 4, related to the size of the river, or the distribution of
and a logarithmic relation was fitted. For the 16 rating the samples taken at high and low discharge. Loca-
curves established for Andernach the following tions that plot on different lines not only are charac-
relation was obtained (Fig. 5A): terised by steeper or flatter rating curves, but also by a
different sediment transport regime. It can thus be
b ˆ 0:4668 ⫺ 02963 log a …r2 ˆ 0:997† …11† assumed that the lower line in Fig. 4 represents river
sections where a large part of the annual sediment
When no scatter around this regression line occurs, all load is transported at relatively low discharge. Lines
possible combinations of a- and b-values result in that have a higher position in this graph, such as the
rating curves that plot through a communal point, line plotted through the coefficients obtained for
indicated by …Q 01 ; C 01 † in Fig. 5B. However, as in Rockenau and Kalkofen, would then be representative
reality some scatter is present, true-rating curves for river sections where high discharges are more
will plot through a communal area rather than through important in transporting suspended sediment.
a communal point. Measurement locations that plot on The hypothesis that information on the sediment
a different line in the graph in Fig. 4 will have a transport regime of a river section can be obtained
different point in common. For Rockenau, the idea- by from slope/intercept plots is based on a limited
lised relation established between the a- and b-coeffi- amount of data and therefore remains debatable.
cients result in rating curves that plot through the However, when the results presented in Table 5 are
common point …Q 02 ; C 02 † (Fig. 5B). used to verify the hypothesis, no contradictions or
The value of the common discharge/concentration inconsistencies are found. In the main channel about
point is related to the position of the line in Fig. 4. Its 35% of the total annual sediment load is transported
value is different for the different tributaries, i.e. it during infrequent high flows. In the Mosel and the
varies spatially, but it appears to be relatively constant Main this percentage increases to more than 60%,
in time. Hence, it can be suggested that the common and in the Neckar and the Lahn about 77% of the
point, or the position of the line in Fig. 4, is a function annual load is transported during these extreme and
of watershed characteristics that do not vary in time, infrequent discharges.
242 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Fig. 5. Sediment rating curves fitted for Andernach and Rockenau (A) slope/intercept relations, (B) fulcrum of the fitted rating curves as derived
from the slope/intercept relations.

6.3. Comparison with the literature sediment concentrations were observed, all rating
coefficients plot on a single line (Fig. 6A).
To assess the validity of the above hypothesis, Schulte (1995) derived sediment rating curves for
rating curve coefficient values published in the litera- the river Elsenz, a tributary of the river Neckar, using
ture were plotted in slope/intercept plots (Fig. 6). Fenn different intervals of measurements carried out during
et al. (1985) established sediment rating curves for a a flood in March 1988. The resulting a- and b-coeffi-
proglacial stream of the glacier de Tsidjiore Nouve in cient values plot on a single line (Fig. 6A). The posi-
Switzerland, using hourly instantaneous stream tion of the line above the lines obtained in the present
discharge and suspended sediment concentration study implies that during this particular flood a very
data. The data were subdivided using meteorological large part of the sediment (more than 80%) must have
and discharge data. Although differences in suspended been transported at very high discharge.
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 243

Fig. 6. Correlation between slope/intercept values of sediment rating curves obtained from the literature. Slope/intercept relations established
for the rivers Rhine and Neckar are shown for comparison.

Kern (1997) fitted rating curves using data of 3 Rockenau (Fig. 6A). This implies that high discharges
measurement locations within a 11 km river section are more important in transporting suspended sedi-
of the river Neckar, about 65 km upstream of ment at the locations in the river Neckar studied by
Rockenau, and 1 location at the confluence of the Kern than at Rockenau. As the latter gauging station is
river Enz with the river Neckar. The a- and b-values located downstream of the confluence of the Neckar
derived from the rating curves plot on a single line with the rivers Kocher and Jagst (Fig. 1), a decreasing
slightly above the line obtained for the river Neckar at importance of very high discharge can be expected.
244 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Walling (1977b) established sediment rating curves transport regime remained the same even though the
for three small rivers in England, i.e. the rivers total load decreased.
Creedy, Exe, and Dart. Data were grouped according Although none of the examples cited above proves
to season and discharge stage to obtain several rating validity of the hypothesis that the relationship
relationships. The a- and b-values of these relation- between the regression coefficients provides infor-
ships are plotted in Fig. 6B. All values plot well above mation on the importance of extreme discharge stages
the lines found for the river Rhine and its tributaries. in transporting suspended sediment, they do not
This suggests that in these rivers most sediment is contradict it either.
transported at extreme and infrequent discharges and
that the Qs,10% value as defined in Table 5 exceeds 6.4. Implications and applicability
80%. According to the suspended sediment load dura-
tion curves published by Webb and Walling (1984) The advantage of the sediment rating curve techni-
Qs,10% for the river Creedy is in the order of 95%. For que is that once a rating curve has been developed it
the river Dart Qs,10% is even higher. can be applied to past streamflow data to reconstruct
Finally, rating curves developed by Kesel (1989) long-term sediment transport records or to fill gaps in
for the Mississippi River at New Orleans were used existing sediment transport records. However, one of
to assess the validity of the hypothesis. As the rating the major shortcomings of the application of rating
curves are based on annual concentration and curves in the extrapolation of sediment transport
discharge data the results can not directly be time series is that the requisite assumption of statio-
compared with the results of other studies. Kesel narity is often questionable. To judge the validity of
(1989) identified four groupings in the data: a prehis- the stationarity assumption, we need to know more
toric period (1851–1952), a pre-dam period (1930– about the factors that determine the shape of the sedi-
1952) and two post-dam periods (1953–1962 and ment-rating curve. In other words, we can only judge
1963–1982). For each period he derived a rating rela- the stationarity of one or both of the rating coefficients
tionship. According to Kesel (1989) a 43% decrease in if we have more insight into the physical meaning of
sediment load occurred between the prehistoric and these coefficients. For instance, can we expect one or
the pre-dam period, which was mainly caused by both of the rating coefficients to remain constant when
changes in land use practice, whereas river discharge river discharge increases due to a change in climate,
remained almost constant. The regression coefficients or when flood discharge decreases due to the construc-
of both periods plot close to each other (Fig. 6C). tion of dams? If not, can we predict in which way the
Thus, although the sediment load decreased, the sedi- coefficients will change? In the previous discussion on
ment transport regime probably did not change much. the interpretation of sediment rating curves it was
During the first dam period (1953–1962) the annual shown that a logarithmic relation exists between the
sediment load declined with 51%. The position of the rating coefficients which seems to be related to the
regression coefficients in the plot is significantly sediment transport regime of the river section under
lower, which suggests a change in the sediment trans- investigation. For larger rivers, such as the river
port regime, with a decreasing importance of years of Rhine, variations in annual discharge or sediment
high discharge. This can be expected as reservoirs and load do not seem to have a major effect on the sedi-
lakes tend to trap large amounts of sediment, which ment transport regime as the coefficient values
will not be flushed out during high discharge events, obtained for periods of low and high discharge plot
decreasing the importance of high discharge in trans- on a single line. This implies that gaps in sediment
porting suspended sediment. During the next period transport records can be filled using the relation
(1963–1982) the sediment load decreased by another between the rating coefficients, the daily discharge
27%. A line can be drawn through the coefficients of record, and an estimate of the total sediment load.
the first and the second post dam periods. As the line Also, as variations in the sediment transport regime
runs more or less parallel to the lines observed for along the river Rhine are small this method could be
other rivers (see the line of the river Rhine for applied to estimate sediment concentrations at loca-
comparison) it can be assumed that the sediment tions where only water discharge has been measured.
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 245

Data of the gauging station Andernach were used to in which Qs is the total sediment load (kg/5 year), Qi
illustrate how the logarithmic relation between the the discharge of day i (m 3/s), Ci the estimated sedi-
rating coefficients can be applied to extrapolate the ment concentration of day i (mg/l), a the rating curve
sediment transport record. The measurement record coefficient determined by iteration and 86.4 the
at Andernach covers a period of 16 years, 1975– conversion factor (86400 s/day × 0.001 g/kg)
1990. When no sediment concentrations would be Sediment transport rates estimated using a sedi-
available for the period since 1986 the following ment-rating curve always differ from measured sedi-
procedure can be used to obtain an estimate of daily ment loads. These errors are shown by ‘method I’ in
sediment transport at Andernach since 1986. Fig. 7. The sediment-rating curve was fitted on
First, sediment rating curves are established for all discharge and suspended sediment data measured
years between 1975 and 1985. The regression coeffi- between 1986 and 1990. This sediment-rating curve
cients are plotted in a graph similar to Fig. 5A and the produces reasonable estimates of long-term total sedi-
slope/intercept relation is determined: ment loads. For the period 1986–1990, the estimated
total sediment load is 6% less than the measured sedi-
b ˆ 0:4701 ⫺ 0:2948 log a …r2 ˆ 0:998† …12†
ment load. For the different discharge intervals esti-
As data from the measurement period 1986–1990 are mation errors are larger and range from ⫺40% to
omitted, the relation is slightly different from the about 120% (Fig. 7).
relation given in Eq. (11). When sediment transport at the different discharge
Second, the total sediment load during the period intervals is calculated with the slope/intercept relation
1986–1990 is estimated. Four methods can be determined at Andernach in combination with esti-
applied: (1) The annual sediment load equals the aver- mated long-term total sediment loads, similar errors
age annual sediment load measured between 1975 and are obtained. The long-term total sediment load can be
1985. (2) The annual sediment load equals the sum of estimated using any of the four methods described
the loads measured at Kaub (Rhine), Cochem (Mosel) previously. An example is given by method II
and Kalkofen (Lahn). (3) The sediment load is a func- (Fig. 7). Depending on the method applied to estimate
tion of the load measured near Rees. During the period the long-term sediment load, errors in estimated total
1975–1985 the load at Andernach was 1.07 times the sediment load vary from ⫺5% to ⫹4%. Estimation
load measured near Rees, hence the annual sediment errors for the different discharge intervals are larger
load can be estimated as 107% of the sediment load but in the same order of magnitude as those obtained
measured at Rees. (4) The sediment load is a function with the sediment rating curves fitted on measured
of the annual discharge. The relationship between data of the period 1986–1990 (method I, Fig. 7).
annual discharge and annual sediment load can be Method III (Fig. 7) presents the errors that result
determined from the 1975–1985 data. when the slope/intercept relation determined at other
Third, the daily discharge frequency distribution locations along the river Rhine is applied. Since the
over the period 1986–1990 is determined. Combina- errors fluctuate around zero it can be concluded that
tion of this daily discharge frequency distribution with the steepness of the rating curve, which is determined
the slope/intercept relation given in Eq. (12) and the by this slope/intercept relation, is correct (method III,
estimated annual sediment load provides the opportu- Fig. 7). When an inappropriate slope/intercept relation
nity to deduce a sediment rating curve and to estimate is applied the errors exhibit a clear trend. This effect
sediment transport rates as a function of discharge becomes apparent when the slope/intercept relation
during the period 1986–1990. This is done iteratively obtained for Rockenau (Fig. 5) is combined with
using: discharge data from Andernach (Fig. 7, method IV).
(n ) Because the total sediment load is set equal to the
X
Qs ˆ Qi Ci 86:4 …13† measured sediment load, errors are only the result of
iˆ1 inaccuracies in the shape of the sediment-rating curve.
This method (IV) overestimates sediment loads at low
with
discharge and underestimates sediment loads at high
Ci ˆ aQ…0:4701⫺0:2948
i
log a†
…14† discharge. The resulting trend in the errors indicates
246 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Fig. 7. Differences between measured and computed sediment transport. Sediment loads are computed as a function of discharge, using four
methods. Method I: a sediment rating curve based on sediment concentrations measured at Andernach between 1986 and 1990. Method II:
combination of the slope/intercept relation determined at Andernach, and the long-term average sediment load measured between 1975 and
1985. Method III: combination of the slope/intercept relation determined at other locations along the Rhine, and the measured sediment load.
Method IV: combination of the slope/intercept relation determined at Rockenau (river Neckar) and the measured sediment load.

that the shape of the applied rating curve is wrong, and locations along the river Rhine and its main tribu-
that only the slope/intercept relations of gauging taries. Rating curves obtained by least squares regres-
stations that plot on the same line in Fig. 4 should sion on logarithmic transformed data tend to
be used to fill gaps in the sediment transport record. underestimate sediment transport rates by about
Finally, the method can be applied to estimate sedi- 10% to more than 50%. The degree of underestima-
ment transport at locations where only water tion decreases when a correction factor is applied.
discharge is measured. As all gauging stations along Better estimates are obtained when rating curves are
the main channel of the river Rhine are characterised fitted in the form of a power function, based on
by a similar sediment transport regime, it can be nonlinear least squares regression. For most locations,
assumed that a location in between two measurement computed long-term average sediment transport rates
locations will have a similar sediment transport differ less than 20% from measured sediment loads.
regime. Hence, it can be expected that the slope/inter- The steepness of the fitted rating curves decreases
cept relation established for the river Rhine is valid for along the main channel of the Rhine in a downstream
all other locations between Rheinfelden and Rees, so direction. The steepest rating curves are found for the
that sediment transport rates can be estimated from tributaries. It is generally believed that the steepness
discharge measurements only. of the rating curve is related to the availability of
suspended sediment in a certain area, in combination
with the erosive power of the river to transport this
7. Conclusions material. Steep rating curves are expected to be char-
acteristic for rivers where most sediment is trans-
Sediment rating curves were fitted for different ported at high discharge. However, although the
N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248 247

steepness of the rating curve decreases along the river nutrient loads entering Chesapeake Bay. Water Resources
Rhine in a downstream direction, no changes in sedi- Research 28, 2353–2363.
Deutscher Verband für Wasserwirtschaft und Kulturbau (DVWK),
ment transport regime occur. Thus, rating curve steep- 1986. Regeln zur Wasserwirtschaft, Schwebstoffmessungen.
ness does not provide information on the sediment Regeln zur Wasserwirtschaft, 125.
transport regime. Fenn, C.R., Gurnell, A.M., Beecroft, I.R., 1985. An evaluation of
A better indication of the sediment transport regime the use of suspended sediment rating curves for the prediction of
of a river section is obtained when the slope/intercept suspended sediment concentration in a proglacial stream.
Geografiska Annaler 67A, 71–82.
pairs of the fitted rating curves are plotted in a graph.
Ferguson, R.I., 1986. River loads underestimated by rating curves.
All locations that plot on the same line have a Water Resources Research 22, 74–76.
common discharge-concentration point and appear Ferguson, R.I., 1987. Accuracy and precision of methods for esti-
to be characterised by a similar sediment transport mating river loads. Earth Surface Processes and Landforms 12,
regime. Locations that plot on higher lines, i.e. that 95–104.
have higher slope values at a given intercept value, von Hinrich, H., 1974. Schwebstoffgehalt, Gebietsniederschlag,
Abfluß und Schwebstofffracht der Ems bei Rheine und Versen
have a different discharge-concentration point and are in den Jahren 1965 bis 1971. Deutsche Gewässerkundliche
characterised by a sediment transport regime where a Mitteilungen 18, 85–95.
larger part of the annual sediment load is transported Jansson, M., 1985. A comparison of detransformed logarithmic
during high discharge. regressions and power function regressions. Geografiska
For the river Rhine all slope/interval pairs plot on a Annaler 67A, 61–70.
Kern, U., 1997. Transport von Schweb- und Schadstoffen in
single line. This suggests that for any location along
staugeregelten Fließgewässern am Beispiel des Neckars.
this river a rating curve can be established using the Mitteilungen Institut für Wasserbau, Universität Stuttgart, Stutt-
relationship between the slope/interval values deter- gart, Germany.
mined at other locations along the Rhine, in combina- Kesel, R.H., 1989. The role of the Mississippi River in wetland loss
tion with water discharge data, and an estimate of the in Southeastern Louisiana, USA. Environmental Geology and
Water Sciences 13, 183–193.
long-term sediment load.
Lozowski, E.P., Charlton, R.B., Nguyen, C.D., Wilson, J.D., 1989.
The use of cumulative monthly mean temperature anomalies in
the analysis of local interannual climate variability. Journal of
Acknowledgements Climate 2, 1059–1068.
Morgan, R.P.C., 1995. Soil erosion and conservation, 2nd ed.
Longman, London.
The author thanks the Bundesanstalt für Gewasser- Mossa, J., 1989. Hysteresis and nonlinearity of discharge-sediment
kunde (Germany) for providing the discharge and relationships in the Atchafalaya and lower Mississippi. Sedi-
suspended sediment data. ment and the environment (Baltimore symposium, May 1989).
IAHS publication, Wallingford, 184, 105–112.
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through
conceptual models. Part I. A discussion of principles. Journal
References of Hydrology 10, 282–290.
Nordin, C.F., Sabol, G.V., 1973. Estimating average sediment yield
Blalock, J.R., 1981. Social statistics, McGraw-Hill, Singapore. from annual streamflow and sediment records. International
Bundesanstalt für Gewässerkunde, 1987. Schwebstofftransport in association for hydraulic research, international symposium on
den Bundeswasserstrab en (Übersicht). Bundesanstalt für river mechanics. Bangkok, Thailand, pp. 93–104.
Gewässerkunde, Koblenz. Peters-Kümmerly, B.E., 1973. Untersuchungen über Zusammenset-
CHR/KHR, 1976. Das Rheingebiet: Hydrologischer Monographie. zung und Transport von Schwebstoffen in einigen Schweizer
Internationale Kommission für die Hydrologie des Rheinge- Flüssen. Geographica Helvetica 28, 137–151.
bietes (CHR/KHR), ’s Gravenhagen. Rannie, W.F., 1978. An approach to the prediction of suspended
Church, M., Gilbert, R., 1975. Proglacial fluvial and lacustrine sediment rating curves. In: Davidson-Arnott, R., Nickling, W.
environments. In: A.V. Jopling and B.C. McDonald (Eds.), (Eds.), Research in fluvial systems. Geoabstracts, Norwich,
Glaciofluvial and glaciolacustrine sedimentation. Society of pp. 149–167.
economic paleontologists and mineralogists. Special Sarma, J.N., 1986. Sediment transport in the Burhi Dihing River,
Publication, 23:40–100. India. In: Hadley, R.F. (Ed.). Drainage basin sediment delivery,
Cohn, T.A., Caulder, D.L., Gilroy, E.J., Zynjuk, L.D., Summers, IAHS publication 159, pp. 199–215.
R.M., 1992. The validity of a simple statistical model for esti- Schulte, A., 1995. Hochwasserabfluß, Sedimenttransport und Gerin-
mating fluvial constituent loads: an empirical study involving nebettgestaltung an der Elsenz im Kraichgau. Heidelberger
248 N.E.M. Asselman / Journal of Hydrology 234 (2000) 228–248

Geographische Arbeiten 98. Universität Heidelberg, waters. IAHS publication, Wallingford, 122, 34–48 (proceed-
Heidelberg, Germany. ings of the Paris symposium, July 1977).
Singh, K.P., Durgunoglu, A., 1989. Developing accurate and reli- Walling, D.E., 1977b. Assessing the accuracy of suspended sedi-
able stream sediment yields. Sediment and the environment, ment rating curves for a small basin. Water Resources Research
IAHS publication, Wallingford, 184 (Proceedings of the Balti- 13, 531–538.
more symposium, May 1989: pp. 193–199). Walling, D.E., 1978. Suspended sediment and solute response char-
Thomas, R.B., 1988. Monitoring baseline suspended sediment in acteristics of the river Exe, Devon, England. In: Davidson-
forested basins: the effects of sampling on suspended sediment Arnott R., Nickling W. (Eds.), Research in fluvial systems.
rating curves. Hydrological sciences Journal 33, 499–514. Geoabstracts, Norwich, pp. 169–197.
Walling, D.E., 1974. Suspended sediment and solute yields from a Webb, B.W., Walling, D.E., 1984. Magnitude and frequency char-
small catchment prior to urbanization. In: Gregory, K.J., Walling, acteristics of suspended sediment transport in Devon rivers. In:
D.E. (Eds.). Fluvial processes in instrumented watersheds, Insti- Burt T.P., Walling D.E. (Eds.), Catchment experiments in
tute of British geographers special publication, 6, pp. 169–192. fluvial geomorphology. Geobooks, Norwich, pp. 399–415.
Walling, D.E., 1977a. Limitations of the rating curve technique for Webster, R., McBratney, A.B., 1989. On the Akaike information
estimating suspended sediment loads, with particular reference criterion for choosing models for variograms of soil properties.
to British rivers. Erosion and solid matter transport in inland Journal of Soil Science 40, 493–496.

Você também pode gostar