M. Venohr*, I. Donohue**, S. Fogelberg***, B. Arheimer***, K. Irvine** and H. Behrendt*
*Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany.
**Department of Zoology, Trinity College, University of Dublin, Ireland.
***Swedish Meteorological and Hydrological Institute, Norrköping, Sweden.
Abstract The mean annual transfer (loss and retention) of nitrogen in a river system was estimated using a
conceptual approach based on water surface area and runoff. Two different approaches for the calculation
of water surface area were applied to determine riverine nitrogen retention in four European catchments,
ranging between 860 –14,000 km2 in area, and differing considerably in the proportion and distribution of
surface waters, specific runoff and specific nutrient emissions. The transfer rate was estimated sequentially
as either the mean value for the total catchment, on a sub-catchment scale, or considering the distribution of
water surface area within a sub-catchment. For the latter measure, nitrogen retention in larger lakes was
calculated separately. Nitrogen emissions modelled with MONERIS and HBV-N were used to calculate
nitrogen river loads and compare those with observed loads. Inclusion of the proportion of water area within
a sub-catchment improved modelled results in catchment with large lakes in sub-catchments, but not where
there was a homogenous distribution of surface waters among sub-catchments.
Keywords emissions; hydraulic load; nitrogen; retention; water surface area
Introduction
Knowledge of the mechanisms, dynamics and spatial distribution of nutrient emissions
and transfer through catchments is critical for assessing risk to, and effective management
of, water quality. Transfer mechanisms and dynamics can and have been studied in field
experiments on a river stretch scale or on even smaller scales in laboratories. Resulting
transfer on a river system scale, however, can only be modelled. Modelling on a river
system scale can either be done by extrapolating results from experiments or physical
based micro-scale models (up-scaling) or by interpolating less sophisticated (conceptual)
model approaches (down-scaling). Model applications always face the conflict between
accuracy in the considered processes and increasing demand in input data. This problem
automatically defines a limit for applicable scales in both directions, up- and down-scaling. Down-scaling, as focused on in this study, was not explored to delineate a minimum
scale range for conceptual models, but to show in which range model results can be
improved by a limited increase of spatial information. Different hydro-morphological
conditions were considered to test sensitivity of these river systems on variation in spatial
information.
The retention and loss of nutrients in river systems can impact considerably on nutrient budgets in catchments and have significant detrimental consequences on downstream
water quality (Donohue et al., submitted). Rates of nutrient transfer vary considerably in
both time and space owing to variability in residence time and hydrological conditions
along river systems (Seitzinger et al., 2002; Grizetti et al. 2003; Billen and Garnier,
2000; Kronvang et al., 1999, Windolf et al., 1996).
The loss of nitrogen to the air from denitrification and the temporary or permanent
retention of nitrogenous compounds in sediments dominate nitrogen transfer within surface waters (Seitzinger et al., 2002; Alexander et al., 2000). Denitrification, generally the
Water Science & Technology Vol 51 No 3-4 pp 19–29 Q IWA Publishing 2005
Nitrogen retention in a river system and the effects
of river morphology and lakes
19
M. Venohr et al.
dominant mechanism of nitrogen transfer from rivers occurs mainly in the top few millimetres of the sediment surface (Behrendt and Opitz, 2000; Windolf et al., 1996). This
suggests that sediment surface area is important for modelling of nitrogen transfer, and in
this study was considered as an estimate of water surface area. Previous work (Behrendt
et al., 2002) has shown surface area of sediment to relate strongly to sedimentation and
denitrification, so distribution of water surface area will likely be of major importance for
nitrogen retention and loss in catchments. Modelling of N-loss and retention needs, therefore, to include spatial information on surface waters.
Accordingly, in some dynamic models water surface area is included in the rate
equations for nitrogen losses in lakes or wetlands instead of water volume (e.g. Arheimer
and Brandt, 1998; Arheimer and Wittgren, 2002). Also in this study, a distinction was
made between the main river course (directly located between monitoring stations) and
its tributaries on a sub-catchment scale. The accurate description of these characteristics
within river systems can be as important to the modelling of nutrient transfer in rivers as
reliable and sufficient input data.
This paper compares two conceptual approaches to estimate catchment water surface
area under differing hydro-morphological conditions. We investigate whether these surface water estimates together with measured runoff are capable of explaining the differences between nitrogen emissions (i.e. Gross load before retention) and resulting river
loads. In order to get an idea of the range of nitrogen emissions the two models HBV-N
and MONERIS were applied to three of the four study catchments (only MONERIS was
applied for Lough Mask yet).
Methods
Study sites
20
An empirical relationship describing nitrogen transfer as dependent on water surface area
(WSA) and runoff was used to estimate nitrogen transfer rates in four European study
catchments with differing drainage patterns (Fig. 1); in Warnow (Germany), Mask
(Ireland), Rönneå (Sweden) and Neckar (Germany).
The Mask catchment, located in the west of Ireland, has the highest WSA, and is the
smallest of the four catchments examined in this study (Table 1). It also has the highest
specific runoff, and contains the lowest population density. All surface waters in the
catchment flow through Lough Mask (surface area 82 km2; Zmax 58 m). The western part
of the catchment is mountainous with extensive peatlands, while primarily agricultural
grasslands overlie the eastern part.
The Rönneå catchment in the southwest of Sweden (Fig. 1) is divided into 7 subcatchments, ranging in area from 150 km2 to 550 km2, and discharges into Skälderviken
bay. This catchment is dominated by agriculture and forest and contains only a few small
lakes, mostly located in the upper parts of the system.
The catchment of the River Neckar, which originates in the hilly Schwarzwald in
southwestern Germany and discharges into the River Rhine, is the largest of the four
study catchments (Table 1). This catchment also contains the highest population density
and the highest specific nitrogen emissions of all catchments studied. Owing to a lack of
large lakes, the catchment has the lowest WSA in this study (Table 1).
The low-lying catchment of the River Warnow in the northeast of Germany is used
intensively for agriculture (Table 1). Gentle slopes, sandy soils and an annual precipitation of 600 mm a21 results in considerably lower specific runoff than the other
catchments.
Riverine nitrogen concentrations in the German and Swedish catchments could be
taken from governmental monitoring programmes. In the Irish catchments these
M. Venohr et al.
Figure 1 Location of the four study catchments and the distribution of the sub-catchments and lakes within
each catchment
measurements were collected through a dedicated monitoring programme. The available
datasets include dissolved inorganic (DIN) and total nitrogen (TN) for Lough Mask and
river Warnow, TN for Rönneå and DIN for the Neckar.
Emission calculations
The nutrient emission model MONERIS (Behrendt et al., 2002) was applied to all
four study catchments. The model considers six diffuse pathways (direct atmospheric
deposition, surface runoff, erosion, tile drainages, groundwater, urban areas) and
Table 1 Site characteristics and nitrogen load of the four study catchments
Study period
Catchment area
Catchment area range
No. of sub-catchments
Specific runoff
Specific TN load
Mean population density
Land-use data by CORINE
Urban area
Agricultural
Forest
Water surface area
1
Lough Mask
Rönneå
Neckar
Warnow
[l.s21.km22]
[kg N.ha21.yr21]
[inh.km22]
06.01 –05.02
859
1–118
33
49.0
10.8
22
93–97
1,897
153 –558
7
12.3
8.5
52
93 –97
13,957
2.4 –1,954
74
11.8
18.53
375
95–98
3,067
51 –210
26
4.3
5.6
57
[%]
[%]
[%]
[%]
0.26
43.3
2.9
12.4
3.2
32.2
46.5
3.0
11.9
53.4
37.9
0.1
2.8
70.8
22.1
3.6
[km2]
[km2]
Calculated with MONERIS; 2calculated with HBV-N; 3DIN load.
21
M. Venohr et al.
emissions from point sources. MONERIS was developed originally for medium-sized to
large catchments (. 500 km2; Behrendt et al., 2000; Behrendt et al., 2002), but has also
been shown to deliver reliable results in catchments down to 50 km2 (Behrendt et al.,
2003, Venohr et al., submitted). MONERIS utilises data from both GIS analyses and statistical reports. For the application of MONERIS to the catchments used in this study, no
modification of model approaches or constants took place.
The process-based, semi-distributed HBV-N model (Arheimer and Brandt, 1998) was
applied to the Rönneå catchment. The hydrological part (i.e. HBV-96, Lindström et al.,
1997) includes routines for accumulation and melt of snow, accounting for soil moisture,
and lake routing and runoff response. In the N routine, leakage concentrations are
assigned to the water percolating from the unsaturated zone of the soil to the response
reservoir of the hydrological HBV model. Nitrogen concentrations are applied to water
originating from areas with differing land-uses. Emissions from point sources, such as
rural households, industries, and wastewater treatment plants are included. Atmospheric
deposition is added to lake surfaces, while deposition on land is implicitly included in
soil leaching. The model simulates residence, transformation and transport of N in
groundwater, rivers and lakes and calculations are made step-wise for each of these compartments. A number of free parameters, which are calibrated against observed timeseries of water runoff and nitrogen concentrations (Petterson et al., 2001), are included in
the model. The equations used to account for the nitrogen turnover processes are based
on empirical relations between physical parameters and concentration dynamics. The
retention routine included in the HBV-N modelled has not been used for these calculations. The model application in Rönneå was based on the TRK database (Ejhed and
Brandt, 2003).
Nitrogen transfer approach
A general nutrient balance was used to describe the sum of all nitrogen retention and loss
processes: Load (L) = Emission (E) – Retention and Loss (R). By dividing both sides of
the equation by the river load, the ratio between the river load and total nutrient emissions can be described by a load-weighted retention coefficient RL. Behrendt and Opitz
(2000) used the following exponential equation to calculate RL:
L ¼ E
1
1
¼ E
1 þ RL
1 þ aHLb
ð1Þ
where L is the observed load (metric tonnes a21), E is the calculated nitrogen emission
(metric tonnes a21), HL is the hydraulic load (runoff/water surface area, m a21) and RL is
the load-weighted retention and loss coefficient. The parameters a and b were determined
for both DIN and TN in rivers (Behrendt et al., 2002), and for TN in lakes (unpublished
data from the EU-Project EUROHARP; Table 2).
Table 2 Parameters used for retention approaches
Parameter set
22
1 (TN rivers)
2 (DIN rivers)
3 (TN-lakes)
a
b
1.9
5.9
7.279
2 0.49
20.75
21
MONERIS approach to estimate WSA
In the current study, two methods were used to estimate WSA. The first (WSAMO, Eq. 2)
was developed for MONERIS (Behrendt et al., 2002) to estimate the total water surface
area connected to a river system on a sub-catchment scale from the 100 m CORINE land
cover grid:
ð2Þ
where WSAMO is the total water surface area (km2) connected to the river system used in
the model MONERIS (Behrendt et al., 2002) and calculated on basis of the water surface
area considered in the CORINE landcover map (WSAC) (km2), the sub-catchment area
(Acatch) (km2) and slope (%). Small rivers and lakes are overlooked at this resolution,
which necessitated the addition of a certain amount of surface waters to the water surface
area shown in the CORINE map as follows (Eq. 2). The parameters a and b are derived
by comparison of the water surface area in the CORINE map and detailed WSA information from administrative reports (Behrendt and Opitz, 2000).
M. Venohr et al.
WSAMO ¼ WSAC þ aAcatch slopec
River width approach to estimate WSA
A second approach, the river width approach, was developed for this study. This
approach differentiates between the water surface area of the main river (MR, river
stretch between two monitoring stations, Fig. 2) and that of all other river stretches
(ARS) Table 3 in a catchment. River width (RW) was calculated as dependent on the
total catchment area (Acatch), specific runoff (q) and the mean slope in the catchment
Figure 2 Overview on the different input data and methods used for the estimation of riverine nitrogen
retention
23
Table 3 Parameters used for the estimation of water surface area (WSAMO) and the mean river width for
the main river (MR) and the all river stretches (ARS)
M. Venohr et al.
WSAMO
MR
ARS
a
b
c
d
0.0052
0.350
0.152
1.087
0.468
0.102
0.360
1.018
2 0.028
2 0.027
2 0.250
(from a 1 km2 digital elevation model) (Eq. 3), while flow length was taken from
digital maps:
RW ¼ aAcatch b qc sloped
ð3Þ
The parameters a, b, c and d (Eq. 3) for the calibration of river width of the main river
were estimated using the measured width of several river stretches in the Lough Mask
catchment and river width information from the river Warnow. Calibration of parameters
for the estimation of mean width of all river stretches was done with data from the River
Warnow. These calculations allowed the direct estimation of water surface area of all
river stretches (WSAARS) and of the main river (WSAMR) at the sub-catchment scale.
Lake areas were taken from detailed digital maps and were added to the estimated river
surface area, separately for main river and tributaries on a sub-catchment scale.
Results and discussion
River width estimation
Mean difference between measured and calculated widths of the main river was 30.5%
(r 2 ¼ 0.98, n ¼ 11, p ¼ 0.001) for Lough Mask and 32.5% (r 2 ¼ 0.71, n ¼ 25,
p ¼ 0.001) for the Warnow. The mean width of all river stretches (ARS) calibrated with
data from river Warnow showed similar deviations (30.7%, r 2 ¼ 0.85, n ¼ 25,
p ¼ 0.001). Comparison of the total water surface area calculated by WSAMO and
WSARW showed high deviations related inversely to water surface areas (Table 4). At the
single sub-catchment scale, in particular in the smaller Mask and Neckar catchments,
considerable deviations up to several hundred percent were found. For water surface
areas smaller than 1 km2 these differences grew rapidly.
Nitrogen transfer calculations
Small differences in the mean calculated specific emissions by MONERIS and HBV-N
were found for Rönneå (HBV-N 15% less than MONERIS) and Neckar (HBV-N 7%
more than MONERIS) (Table 5). Specific emissions in the Warnow calculated by HBVN were 70% higher than those calculated by MONERIS. For the German rivers we determined deviations up to 450% among emissions calculated by MONERIS and HBV-N for
the sub-catchments. For Rönneå the highest deviation between the emissions by the two
models was 38%.
Table 4 Deviation between the water surface areas WSAMO and WSAARS for the total catchment and the
median of the deviations in the single sub-catchments
Deviation [%]
Total catchment
24
Lough Mask
Rönneå
Warnow
Neckar
0.3
7.8
21.0
31.6
Sub-catchments
68.2
23.8
23.6
54.9
Table 5 Mean modelled emissions, based on the models MONERIS and HBV-N and the deviation from
observed spec. loads at catchment outlet.
MONERIS
TN
TN
TN
DIN
Deviation [%]
Emissions [kg.ha21.yr21]
Deviation [%]
15.5
14.0
12.3
26.6
30.7 (0.42)2
39.0 (0.51)2
54.1 (0)2
30.51 (0.51)*
12.0
21.4
28.6
21.6 (0.27)2
73.7 (0.02)2
35.3 (0.45)1
1
Deviation between TN emissions and DIN river load.
( 2 not significant, þp ¼ 0.05, pp ¼ 0.1).
M. Venohr et al.
Lough Mask
Rönneå
Warnow
Neckar
HBV-N
Emissions [kg.ha21.yr21]
A large deviation among nitrogen loads measured at the catchment outlet and emissions estimated with MONERIS and HBV-N was found (Table 5), with substantial differences in the deviation among catchments. Neglecting possible model errors this deviation
expresses the retention as balance between calculated emissions and observed loads. The
regression between the emissions and the observed loads is very weak for all catchments
(Table 5). This suggests strongly that the retention and loss of nitrogen is of considerable
importance in our study catchments and also that it varies substantially between
catchments.
The four methods used (Fig. 2) to calculate the transfer rate were highly sensitive to
the calculated water surface area. In general, increased water surface area leads to an
increased retention rate, but no clear trend for over or under estimating the retention
among the four methods could be identified. The four methods also showed large variations in calculated retention and loss among sub-catchments (Table 6). The application
of method 3 and 4 led to considerable differences in accumulated retention calculated
with emissions by MONERIS and HBV-N (Table 6). These deviations in the retention
rates are caused by differences in the spatial distribution of the emissions and transfer
rates. In almost all cases, mean deviations from observed loads were reduced considerably by the application of the revised method 4 used to estimate riverine transfer, and
increased predictive power (Table 7). This improvement did not apply for the River
Neckar, which contains no big lakes. Here, the comparison between methods 1 and 4 led
to a reduction of the regression coefficient with, additionally, increasing deviations
between measured and calculated loads. Nevertheless, a consideration of sub-catchment
conditions (comparison between methods 1 and 3 and methods 2 and 4) led to a slight
improvement of the calculated loads (Table 7).
Table 6 Accumulated retention rate of the total-catchments calculated for TN and DIN using the four
methods outlined in Fig. 2
Mean calculated retention rate [%]
Methods
Lough Mask
Rönneå
Warnow
Neckar
TN-MONERIS
DIN-MONERIS
TN-MONERIS
TN-HBV-N
TN-MONERIS
TN-HBV-N
DIN-MONERIS
DIN-HBV-N
DIN-MONERIS
DIN-HBV-N
1
2
3
4
36.7
48.9
38.0
38.0
50.0
50.0
68.8
68.8
26.7
26.7
36.1
47.9
36.8
36.8
52.3
52.3
71.8
71.8
19.9
19.9
44.8
53.3
32.7
32.3
49.5
65.6
64.1
64.4
27.8
29.9
47.4
70.5
41.3
39.7
64.5
62.9
70.9
69.7
32.8
34.9
25
Table 7 Mean deviation (and regression coefficient r2) between measured and calculated TN and DIN loads
Mean deviation [%]
Methods
1
Lough Mask
M. Venohr et al.
Rönneå
Warnow
Neckar
TN-MONERIS
DIN-MONERIS
TN-MONERIS
TN-HBV-N
TN-MONERIS
TN-HBV-N
DIN-MONERIS
DIN-HBV-N
DIN-MONERIS
DIN-HBV-N
2
23.2 (0.65p)
126.1 (0.66p)
27.9 (0.89**)
22.5 (0.642)
32.0 (0.41)
101.6 (0.132)
36 (0.66***)
101.6 (0.53p)
12.8 (0.64**)
22.7 (0.49p)
24.0
112.8
29.3
21.8
29.5
89.9
32.3
78
18.7
29.1
(0.611)
(0.78**)
(0.89**)
(0.711)
(0.51p)
(0.252)
(0.71***)
(0.62**)
(0.55p)
(0.48p)
3
25.8
131.7
31.7
23.7
28.2
102.2
36.1
113.9
12.8
18.7
(0.61)
(0.64p)
(0.82p)
(0.62)
(0.71***)
(0.112)
(0.73***)
(0.43p)
(0.6**)
(0.51p)
4
22.8 (0.73p)
109.7 (0.76**)
25.6 (0.86p)
25.6 (0.71)
19.6 (0.77***)
64.5 (0.342)
21.6 (0.77***)
69.9 (0.61**)
20.5 (0.53p)
32.8 (0.362)
1
Emissions calculated with HBV-N.
( 2 not significant, þp ¼ 0.1, pp ¼ 0.05,
pp
p ¼ 0.01,
pp p
p ¼ 0.001).
The strongest reduction of the error in the calculated load was found for the Warnow
catchment (Table 7), which contains many lakes distributed throughout the catchment. In
the case of the Mask catchment, the deviation between measured and calculated load was
only slightly reduced, but the results from the application of Method 4 delivered the
strongest association with the observed load. In the Lough Mask catchment, the calculated DIN load showed extraordinary high deviations from the measured DIN load. Surface waters in the Lough Mask catchment contain a high concentration of organic
nitrogen in the form of humic substances, leading to relatively low proportions of DIN:
TN. The mean ratio of 2.9 between TN and DIN concentration in Lough Mask is almost
double that of the River Warnow (1.5). Even for Rönneå, which contains only a few
upstream lakes, using MONERIS emissions Method 4 delivered a stronger association
between measured and calculated load than the other methods (Table 7). Using HBV-N
emissions, on the other hand, the deviations but also the regression coefficient grew.
Differences in the emissions by MONERIS and HBV-N could not be identified to cause
this contradictory results.
Regardless to the reduction of the error in the calculated loads great differences in the
remaining deviations among the results calculated with MONERIS and HBV-N could be
found. In the Rönneå loads calculated with emissions by HBV-N generally showed lower
errors in the calculated loads than those calculated on basis of MONERIS emissions. For
the Neckar on the other hand the mean deviation of MONERIS-loads are much lower
than HBV-N-loads (Table 7). Some outliers cause the higher deviations of the HBV-Nloads and for most of the catchments the models show a similar model deviation. The
biggest differences in the modelled loads were calculated for the river Warnow catchment
(Table 7). Again, MONERIS-loads show lower deviations from the observed loads, but
here, MONERIS-loads always underestimated and HBV-N-loads always over estimated
the observed loads. The reasons for differences in model performance among MONERIS
and HBV-N could not be clarified in this study. Apparently the nutrient emissions should
be somewhere between the modelled emissions.
Discussion
26
Estimation of the water surface area of river systems was difficult to calibrate, owing to
limited information available on the width of different river stretches, in particular for
smaller tributaries in large catchments. Further, anthropogenic impacts on surface water
hydrology and morphology, as in the case of the River Warnow in this study, can lead to
considerable uncertainties in the estimation of water surface area. Increasing deviations
M. Venohr et al.
for WSA smaller than 1 km2 are likely caused by the spatial resolution (100 m) of the
CORINE land use map. Small lakes and rivers are not included at this scale, so that the
higher the number of small surface waters, the greater the underestimate of the total
water surface area. A second problem relates to the location of surface waters in the
CORINE map and topographical maps. For large lakes the shoreline can only be displayed roughly and often does not agree with the shoreline on the topographical map.
Owing to inaccuracies in the processing of digital maps, a general dislocation of lakes
and rivers between maps may also occur, which can cause problems if parts of a lake is
moved into a neighbouring catchment. In our study catchments we were not able to quantify the extent of these uncertainties.
The inclusion of information on sub-catchment WSA (Methods 3 and 4) did not
improve model results for the Neckar catchment. A homogenous distribution of the water
surface area, combined with low variation of specific runoff within the sub-catchments,
result in stable retention and loss rates throughout the river Neckar catchment. Behrendt
et al. (2002) reported increasing errors in the calculated nutrient emissions and the
measured river loads with decreasing catchment size. More than 25% of the sub-catchments in the Neckar catchment have a size less than 50 km2. Small catchment sizes
together with homogenous retention and loss rates could be a reason for the lack of
improvement in the calculated loads from method 1–4 in the Neckar catchment.
This study was based on the premise that increasingly detailed calculation of nutrient
transfer in the main river, tributaries and lakes in a sub-catchment reduces the error of
the calculated net load. Differentiating between the water surface area of the main river
and its tributaries did result in improved estimations of nutrient loads in relation to the
degree of heterogeneity of WSA in the sub-catchments. Where WSA was distributed
more homogeneously across the sub-catchments, improvement in model performance was
not evident. The spatial distribution of lakes is normally considered in catchment modelling when dividing the catchment into sub-catchments, so that the lakes are located at
sub-catchments outlets. However, this procedure may be quite subjective and it would be
interesting to further quantify how detailed the sub-catchment division should be for
reliable results, with respect to the modelling purpose.
Problems for model applications are faulty and inconsistent input data. Lacking input
data, however, is often a stumbling block for modelling at the catchment scale. This is
especially relevant for sophisticated, dynamic and physical based models with demanding
data requirements, which led to the application of a conceptual approach in this study.
The transfer approach is driven by the hydraulic load (runoff divided by water surface
area). High errors in runoff measurements can lead to increased uncertainties of model
output.
Discharge originating from a sub-catchment was calculated by subtraction of
measured discharge from monitoring stations of successively linked sub-catchments. The
improvement from detailed transfer calculations on a sub-catchment scale is, in this case,
simultaneously reduced by uncertainties in the runoff or water surface area calculation
for each sub-catchment.
Conclusions
Rates of nitrogen transfer in catchments were strongly dependent on the proportion and
distribution of surface waters within the sub-catchments. Consideration of the surface
areas of main rivers and tributaries separately for the retention and loss calculation of
nitrogen resulted in improvement of the calculated load (emissions minus retention) compared with the observed load. Uncertainties in the estimation of water surface area, in
addition to those in the emission and transfer calculations, are likely to originate from
27
M. Venohr et al.
insufficient input data, including the accuracy of digital maps and runoff measurements.
It could, nevertheless, be shown that retention and loss play an important part in the total
nitrogen household of a river system. High variability in transfer rates between sub-catchments and river systems could mostly be explained with a simple approach using water
surface area and runoff as driving forces. The detailed consideration of changing transfer
rates (method 4) completed the picture of the nitrogen budget of these river systems,
which could help to identify suitable programmes for water quality improvements.
Acknowledgements
We would like to thank all members of the EU-Projects EUROHARP (EVK1-CT-200100096) and BUFFER (EVK1-CT-1999-00019.), who helped to collate the data and to
conduct this study.
References
28
Alexander, R.B., Smith, R.A. and Schwarz, G.E. (2000). Effect of stream channel size on the delivery of
nitrogen to the Gulf of Mexico. Nature, 403, 758 –761.
Arheimer, B. (2003). Handling scales when estimating Swedish nitrogen contribution from various sources to
the Baltic Sea. Landschap, 20(2), 81 –90.
Arheimer, B. and Wittgren, H.B. (2002). Modelling Nitrogen Retention in Potential Wetlands at the
Catchment Scale. Ecological Engineering, 19(1), 63 –80.
Arheimer, B. and Brandt, M. (1998). Modelling Nitrogen Transport and Retention in the Catchments of
Southern Sweden. Ambio, 27, 471 – 480.
Behrendt, H., Bach, M., Kunkel, R., Opitz, D., Pagenkopf, W.G., Scholz, G. and Wendland, F. (2003).
Nutrient Emissions into River Basins of Germany on the Basis of a Harmonized Procedure, Federal
Environmental Agency (Umweltbundesamt), Research Report 82/03, 191.
Behrendt, H., Kornmilch, M., Opitz, D., Schmoll, O. and Scholz, G. (2002). Estimation of the nutrient inputs
into River systems – experiences from German rivers. Regional Environmental Changes, 3, 107 –117.
Behrendt, H. and Opitz, D. (2000). Retention of nutrients in river systems: Dependence on specific runoff and
hydraulic load. Hydrobiologia, 410, 111 –122.
Billen, G. and Garnier, J. (2000). Nitrogen transfer through the Seine drainage network: a budget based on
the application of the ‘Riverstrahler’ model. Hydrobiologia, 410, 139 – 150.
Brandt, M. and Ejhed, H. (2003). TRK Transport-Retention-Källfördelning, Belastning på havet, Swedish
Environmental Protection Agency, Sweden, Report No. 5247.
Donohue, I., Styles, D., Coxon, C. and Irvine, K. (Submitted). Importance of spatial and temporal patterns for
assessment of risk of diffuse nutrient emissions to surface waters. Journal of Hydrology.
Fogelberg, S., Arheimer, B., Venohr, M. and Behrendt, H. (acc.). Catchment modelling of nitrogen flow with
two different conceptual models, 23rd Nordic Hydrological Conference “ Fresh water Rescources
Management” and UNESCO IHP joint session “Climate change and hydrological processes”, Tallinn,
Estonia, 8th– 12th August 2004.
Grizetti, B., Bouraoui, F., Granlund, K., Rekolainen, S. and Bidoglio, G. (2003). Modelling diffuse emission
and retention of nutrients in the Vantaanjoki watershed (Finland) using the SWAT model. Ecological
Modelling, 169, 25 – 38.
Kronvang, B., Hoffmann, C.C., Svendsen, L.M., Windolf, J., Jensen, J.P. and Dørge, J. (1999). Retention of
nutrients in river basins. Aquatic Ecology, 33, 29– 40.
Lindström, G., Johansson, B., Persson, M., Gardelin, M. and Bergström, S. (1997). Development and test of
the distributed HBV-96 hydrological model. Journal of Hydrology, 201, 272 – 288.
Petterson, A., Arheimer, B. and Johansson, B. (2001). Nitrogen concentrations simulated with HBV-N: new
response function and calibration strategy. Nordic Hydrology, 32, 227 – 248.
Seitzinger, S.P., Styles, R.V., Boyer, E.W., Alexander, R.B., Billen, G., Howarth, R.W., Mayer, B. and van
Breemen, N. (2002). Nitrogen retention in rivers: model development and application to watersheds in the
northeastern U.S.A. Biogeochemistry, 57/58, 199 –237.
Venohr, M. & Behrendt, H. (2002). Modelling the dependency of riverine nitrogen retention on hydrological
conditions and temperature. Proceedings 6th Internat. Conf. On Diffuse Pollution Amsterdam, 30.09. –
04.10.2002, 573 –574.
Venohr, M., Behrendt, H. and Kluge, W. (Submitted). Effects of different input data and their spatial
resolution using a concepotual nutrient emission model: The river Stör case study, Hydrological
Processes.
Vollenweider, R.A. (1969). Möglichkeiten und Grenzen elementarer Modelle der Stoffbilanz von Seen. Arch.
Hydrobiol, 66, 1– 36.
Windolf, J., Jeppesen, E., Jensen, J.P. and Kristensen, P. (1996). Modelling of seasonal variation in nitrogen
retention: a four year mass balance study in 16 shallow lakes. Biogeochemistry, 33, 25 – 44.
M. Venohr et al.
29