Project no. GOCE-CT-2003-505540
Project acronym: Euro-limpacs
Project full name: Integrated Project to evaluate the Impacts of Global Change on
European Freshwater Ecosystems
Instrument type: Integrated Project
Priority name: Sustainable Development
Deliverable No. 264
Calibration of INCA-N to the Tovdal catchment, southernmost
Norway
Due date of deliverable: 31 July 2007
Actual submission date: 10 January 2008
Start date of project: 1 February 2004
Duration: 5 Years
Organisation name of lead contractor for this deliverable: 10 NIVA
Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006)
PU
PP
RE
CO
Dissemination Level (tick appropriate box)
Public
Restricted to other programme participants (including the Commission Services)
Restricted to a group specified by the consortium (including the Commission Services)
Confidential, only for members of the consortium (including the Commission Services)
1
X
Deliverable 264. Workpackage 6. Task 2.
Calibration of INCA-N to the Tovdal catchment, southernmost Norway
Øyvind Kaste1, Mirjam Bloemerts2, Heleen A. de Wit3, and Richard F. Wright3
1
Norwegian Institute for Water Research, Southern Branch, Televeien 3, N-4879 Grimstad, Norway
Environmental Systems Analysis, Wageningen University, 6700 AA Wageningen, The Netherlands
3
Norwegian Institute for Water Research, Gaustadalléen 21, N-0349 Oslo, Norway
2
________________________________________________________________________
This report provides a short description of the calibration of INCA-N to Tovdal, which is
selected a Norwegian core catchment for studies and modelling within Workpackages 4
and 6 in Eurolimpacs. This will contribute to a planned joint WP6 publication on INCAN applications and simulation of climate change effects on N runoff across Europe.
________________________________________________________________________
INTRODUCTION
Human activities such as fossil fuel combustion, fertiliser application, cultivation of
nitrogen N fixing crops, and discharge of domestic and industrial effluents have caused N
enrichment of many terrestrial and aquatic ecosystems. This has increased the importance
of N in acidification of upland lakes and rivers and resulted in eutrophication of many
sensitive estuaries and coastal areas. N concentrations and output fluxes in river systems
reflect the integration of several diffuse or point sources within the catchment and various
terrestrial and aquatic N retention processes. To deal with this large complexity in river
basin management, integrated and spatially distributed catchment models are useful tools.
The process-oriented INCA model attempts to integrate these factors by linking
hydrology and N inputs from atmospheric deposition, agriculture and populated areas
with the microbial processes controlling N behaviour in soils and river reaches
(Whitehead et al. 1998; Wade et al., 2002). By computing a mass balance for all sources
and sinks of N in up to six land cover types within a catchment, the model assesses the
contribution of multiple sources to catchment N pools and river nitrate (NO3) and
ammonium (NH4+) concentrations.
The main objectives of this report are to provide a short description of the calibration of
INCA-N to Tovdal, which is selected a Norwegian core catchment for studies and
modelling within Workpackages 4 and 6 in Eurolimpacs. This will contribute to a
planned joint WP6 publication on INCA-N applications and simulation of climate change
effects on N runoff across Europe.
2
MATERIAL AND METHODS
Site description
The Tovdal River, southernmost Norway, runs north-to-south from its headwaters in the
uplands above 1000 m above sealevel to the Topdalsfjord at the coast at Kristiansand
(Figure 1). The catchment area is 1863 km2 and underlain predominantly by Precambrian
granitic and gneissic bedrock, with thin and patchy moraine of the same lithology. The
higher lying areas of the catchment are characterised by alpine, heathland and peaty soils.
Much of the lower parts are forested with pine, spruce and birch. There is very little
farming, industry or habitation in the catchment, and it has not been developed for
hydropower significantly. The Tovdal River basin is thus dominated by “semi-natural”
unproductive uplands (Table 1).
Figure 1. Map of the Tovdal River basin showing the 5 sub-regions used for modelling.
3
The Tovdal River basin was divided into 5 sub-basins for the modelling. These are the
major tributaries and reaches of the main river with suitable water chemistry data (Figure
1). The surface area, average height and the dominant land-use categories for each reach
are given in Table 1.
The Tovdal catchment contains nearly 300 lakes, the majority of which are highly
acidified and lost their native fish populations (mainly brown trout) in the period 19401980. The river itself is acidified and the salmon population disappeared in the 1960’s
following decades of steady decline. In 1980 the river was included as one of about 20
rivers in SFT’s monitoring programme (SFT 2006). Beginning in late 1995 the river has
been limed to raise the pH such that water quality is adequate for salmon recruitment.
The annual precipitation in the area is 1200-1500 mm/yr and the 30-year mean annual
temperature at the meteorological station Herefoss, located in the central part of the
Tovdal catchment, is 5.6 °C. The river has a mean water discharge of 65 m3/s at
Flaksvatn (1777 km2; bottom reach 4). Atmospheric N deposition is the main N input to
the catchment, and wet and dry deposition of NO3 and NH4 constitute about 15 kg
N/ha/yr on average (Aas et al., 2006).
Table 1. Catchment characteristics of the Tovdal catchments. Upper panel: Surface area and average
altitude in the five reaches. Lower panel: Percentages of land-use cover. Data from Statistics Norway
(SSB) and the Norwegian Water Resources and Energy Directorate (NVE).
Reach 1
Reach 2
Reach 3
Reach 4
Reach 5
Reach 1-5
Reach 1
Reach 2
Reach 3
Reach 4
Reach 5
Reach 1-5
Surface area
(km2)
371.5
279.9
959.8
217.1
34.7
1863.0
Average altitude
(m)
595
234
314
161
70
Arable
Forest Productive
Forest Unproductive
Mountainous
vegetation
Peat land
Lake
surface
0
0
0
5
40
9
22
87
64
84
51
62
33
5
18
2
0
12
32
0
6
1
0
7
3
3
5
4
1
3
10
5
7
4
8
7
4
For calibration of the INCA-N model to Tovdal, two small long-term monitored
catchments were selected to represent the major land cover types: Birkenes for the
forested areas and Storgama for the unproductive, mountainous areas.
Birkenes (0.41 km2) is located in the lower parts of the Tovdal catchment. Of this area,
90% is forest dominated by Picea abies L., 7% is peat land and the remaining 3% is
mountainous vegetation. Altitude in the catchment ranges from 200 to 300 m.a.s.l. and
bedrock is mainly granite and biotite (SFT, 2006). Annual precipitation is 1400 mm/yr
(SFT, 2006). The nearest weather station is Herefoss, where the annual mean annual
temperature corresponds to 5.6°C. Due to higher altitude, the mean annual temperature at
Birkenes is approximately 1°C lower. Atmospheric N deposition is currently around 15
kg N/ha/yr (Aas et al., 2006).
Storgama (0.6 km2) is located close to the Tovdal catchment and covers an altitude range
of 580-690 m.a.s.l. The dominant land-use in Storgama is mountainous vegetation (59%),
typical for the upper parts of the Tovdal catchment. The remaining area is made up by
scattered forest (11%), peat land (22%) and lakes (8%). Bedrock in the area is mainly
granite. Mean annual temperature during 1970-2003 was 3.2°C (calculated with altitude
correction from weather station 3723 Tveitsund). Annual precipitation was 960 mm/yr
during the same period (data from Tveitsund, no altitude correction).
Data sources
Climate
Table 1 gives an overview of the meteorological stations used for modelling in Tovdal,
Birkenes and Storgama. Temperature data are corrected for altitude with a lapse rate of
-0.6°C for each 100 meter increase in elevation.
Table 2. Meteorological stations used to obtain temperature and precipitation data for the modelling.
The temperature column includes a temperature correction factor based on altitude and a lapse rate
of -0.6 oC per 100 m elevation.
Catchment
Birkenes
Storgama
Tovdal Reach 1
Tovdal Reach 2
Tovdal Reach 3
Tovdal Reach 4
Tovdal Reach 5
Temperature station
Herefoss – 0.99°C
Tveitsund – 2.26°C
Herefoss – 3.1°C
Herefoss – 0.9°C
Herefoss – 1.4°C
Herefoss – 0.5°C
Kjevik – 0.3°C
Precipitation station
Rislå/Senumstad
Tveitsund
Dovland
Mykland
Dovland
Rislå/Senumstad
Kjevik
Discharge, flow routing and lakes
The Norwegian Water Resources and Energy Directorate (NVE) operates two gauging
stations on the Tovdal River, one at Austenå in upper Tovdal (sub-basin 1) and one at
Flaksvatn (sub-basin 4) (Figure 1). Daily mean discharge for these two stations was used
5
for calibration. The Tovdal River comprises of 102 REGINE sub-catchments. The
REGINE database operated by NVE includes the 1961-1990 mean specific discharge for
each sub-unit. A database of all the major lakes in Norway is also held by NVE; 24 of
these are located in the Tovdal basin.
Water chemistry
Routine monitoring of water chemistry began in 1980 with monthly sampling at Boen
(outlet of reach 4), near the the Tovdal River outlet, and the data are reported annually
(SFT 2006). Additional stations on the river and its major tributaries were initiated in
1995-96 in conjunction with the planning and operation of the river basin liming. These
include north inflow to Herefossfjord (outlet of sub-basin 3), and outflow of Tveitvatn
(outlet of sub-basin 1); these data are also reported annually (DN 2005). All water
samples are analyzed by NIVA for concentrations of various chemical constituents, of
which only NO3 is used for the INCA-N modelling. NO3 is analysed by ionchromatography using a Dionex DX 320 duo.
Precipitation chemistry
Monitoring of air and precipitation chemistry at two stations in or near the Tovdal River
basin is conducted by NILU (Norwegian Institute for Air Research) as part of the
Norwegian monitoring programme for long-range transported air pollutants (Aas et al.
2006). Samples are collected daily at Birkenes (used for sub-basins 2, 3, 4 and 5) and
weekly at Treungen (used for sub-basin 1) and analyzed by accredited methods at NILU.
Used here are values for bulk deposition of inorganic N (NO3 + NH4).
Land-cover and point sources of nutrients
Land cover and land use data came from Statistics Norway (SSB). N supplied from
inhabitants in urban areas and individual dwellings came from SSB database (by
municipality), portioned to REGINE unit based on number of dwellings on the digital
maps of the Norwegian Mapping Authority (Statens kartverk).
Model description
The process-based and semi-distributed Integrated Nitrogen in Catchments model
(INCA-N) integrates hydrology, basin and river N processes, and simulates daily NO3
and NH4 concentrations as time series at key sites, as profiles down the river system, or
as statistical distributions (Whitehead et al. 1998; Wade et al., 2002). The term semidistributed is used, as it is not intended to model the catchment land surface in a detailed
manner. River, soil water and ground water NO3 and NH4 concentrations and fluxes are
produced as daily time series.
Three components are included: the hydrological model, the catchment N process model,
and the river N process model. Sources of N include atmospheric deposition, the
terrestrial environment and direct discharges. Hydrological processes in soil are simulated
in the hydrological sub-model. The mass balance equations for NO3 and NH4 in both the
soil and groundwater zones are solved simultaneously with the flow equations. The key N
6
processes modelled in the soil water zone are nitrification, denitrification, mineralization,
immobilisation, N fixation and plant uptake of mineral N. Rate coefficients of N
processes are temperature and moisture dependent.
Model Calibration
The INCA-N model requires input of daily time series of air temperature (AT), actual
precipitation (P), soil moisture deficit (SMD), and hydrologically effective rainfall (HER;
the fraction of P that contributes directly to runoff). In addition, the model requires
information about sub-catchment structure (number, size, reach length), physical
properties of the selected sub-catchments and inputs of N from atmospheric deposition,
fertiliser application, and effluent discharges. A full description of the required model
parameters is given by Wade et al. (2002).
The daily change in SMD was calculated from an estimated evapotranspiration rate (ET)
minus P. Evapotranspiration was expressed as a function of AT, and according to longterm water balances for Norwegian catchments (Otnes and Ræstad, 1978) annual ET
amounts to roughly 0.15 mm per oC per day in this region. The volumes and dynamics of
simulated vs. observed flow indicate that this ET factor was appropriate. The time series
of HER was calculated as:
HER = ( P + M ) – ET – ∆S
(1)
where P is liquid precipitation, M is snowmelt water, ET is evapotranspiration and ∆S is
soil water storage. In periods with saturated soils (SMD=0), ∆S will be zero. Water input
from melting snow was estimated by a separate snow accumulation and snow melt model
(Vehviläinen, 1992; Rekolainen and Posch, 1993).
When calibrating the model, procedures recommended by Wade et al. (2002) were
applied. After including the appropriate initial values, INCA was set up to simulate the
actual hydrology both in terms of dynamics and absolute flow before any parameters
controlling N storage, transformations or transport were adjusted. Secondly, the
parameters controlling land phase and in-stream N transformation rates were adjusted
such that annual process loads were within the ranges reported in the literature and a
reasonable match between simulated and observed streamwater NO3 concentrations was
obtained.
In the relatively large and complex Tovdal watershed, the model was first calibrated for
the small, homogenous catchments Birkenes (0.41 km2) and Storgama (0.6 km2), see
previous sections. When scaling up to the entire Tovdal catchment (1863 km2), the main
river was first divided into five reaches (Figure 1). Hydrological time series (SMD, HER,
AT, P) were then assigned to each of the individual reaches, and hydrological parameters
such as storage volumes and velocity/flow relationships were calibrated. Further, N
process parameters from the small catchments were applied to the corresponding land
cover classes in the main catchment.
7
The initial model runs was done for two periods; one calibration period 1996-2000 and
one control period 2000-2003.
RESULTS AND DISCUSSION
Calibration results
Model performance was evaluated by means of the Nash-Sutcliffe efficiency criterion
(E). E-values for both the calibration period 1996-2000 and the control period 2000-2003
are listed in Table 3.
Hydrology
The Birkenes catchment is characterised by short retention time and a low contribution of
ground water to the water flow. Hence, the water discharge data showed remarkable high
peaks rapidly followed by low water flows. Two other complicating factors are the small
catchment area and the local precipitation patterns, often making measured precipitation
data from the nearest met-station little representative for the actual catchment area. Due
to the challenges mentioned above the best hydrological calibration at Birkenes gave a
Nash-Sutcliffe efficiency criterion (E-value) of only 0.29.
Storgama shares some similarities with Birkenes as it is a small catchment with relatively
short retention time of water in the water. However, for Storgama, the main challenge
was to tune the temperature data (measured at the meteorological station Tveitsund) to
reflect freezing-thawing cycles occurring in the catchment during winter (see temperature
correction factor in Table 2). After correction, the relation between precipitation amounts
and measured discharge during winter shows relatively good agreement. The NashSutcliffe efficiency criterion (E-value) for the hydrological calibration at Storgama
corresponded to 0.51.
The larger Tovdal river basin has much longer retention time, which dampens the effects
of precipitation episodes on the water discharge. Judged by the E-values, the quality of
the hydrological simulation is better for Tovdal than for Birkenes and Storgama. An Evalue of 0.61 was found for reach 1 and the results for reach 4 were even better with a
value of 0.74.
In-stream nitrate concentrations
E-values for the simulation of in-stream NO3 concentrations in Birkenes and Storgama
were 0.49 and 0.67, respectively. This must be considered good, especially as the
hydrological calibration at Birkenes was problematic. In the Tovdal sub-catchments the
E-values were 0.38 in reach 1 and 0.53 in reach 4. The exceptionally low E-value for
reach 3 is caused by long retention time in the relatively large Lake Herefossfjord. This
result was to be expected as the INCA-N model structure doses not fully account for lag
times in water chemistry caused by large lakes.
8
Control period simulation
The years 2000-2003 were used as a control for the calibration. For the calibration period
the observations are used to adjust the parameters in order to improve the simulation. For
the control period the observations are used to check the model simulations without
adjusting any parameters. The E-values were used again as efficiency criterion and are
summarized in Table 3. The E-values largely show that the model simulations for the
hydrology in the control period were successful. For most catchments, the E-value was
even higher than for the calibration period. By contrast, the simulations of NO3
concentrations during the control period generally gave lower E-values than the
calibration period. The best results were obtained at Birkenes and Storgama, with Evalues of 0.33 and 0.57, respectively. In the Tovdal catchment, the results for reach 4
were regarded acceptable although a short period of the control run caused instability in
the model (Table 3). Observed and simulated NO3 concentrations for the whole period
1996-2003 are showed in Figure 2.
Table 3. The Nash-Sutcliffe efficiency criterion for the calibration period (1996-2000) and the control
period (2000-2003) for both the hydrology and the nitrate simulations by INCA-N.
Hydrology
Birkenes
Storgama
Tovdal
Reach 1
Reach 4
Nitrate
Birkenes
Storgama
Tovdal
Reach 1
Reach 3 1
Reach 4
1)
2)
E for
Calibration
period
0.29
0.51
E for
Control Period
0.61
0.74
0.66
0.75
0.49
0.67
0.33
0.57
0.38
0.04
0.53
0.02
-0.18
X2
0.22
0.60
For reach 3 the measurements were taken at the outlet of a large lake and therefore gave no adequate
representation of the NO3 concentrations in the river.
INCA-N experienced an error in this simulation run due to instability of the model.
9
NO3 concentration (mg N/l)
0.4
0.3
0.2
0.1
0
Jan-96
Jan-97
Jan-98
Jan-99
Simulated NO3
Jan-00
Jan-01
Jan-02
Observed NO3
Figure 2. INCA-N simulated and observed nitrate concentrations in Tovdal, Reach 4 for the years
1996-2003.
Overall model performance and simulation of N loads
Nitrate seasonal pattern
NO3 concentrations in the main river follow a clear seasonal pattern with the highest
concentrations during winter and the lowest during summer. The low summer values are
mainly due to vegetation uptake of NO3 during the growing season. Figure 3 shows one
year of observed and simulated NO3 concentrations at Boen (reach 4). Monthly
observations, as displayed for Boen, are not capable of covering the complete seasonal
variations, but observed and simulated concentrations were in good agreement.
10
NO3 concentration (mg N/l)
0.4
0.3
0.2
0.1
0
Jan Feb Mar Apr May Jun
Simulated NO3
Jul
Aug Sep Oct Nov Dec
Observed NO3
Figure 3. One year (1999) of observed nitrate concentrations measured at Boen (reach 4) and nitrate
concentrations simulated by INCA-N in reach 4.
600
2500
500
2000
400
1500
300
1000
200
500
100
0
1995
Annual Precipitation (mm)
NO3-N (tonnes per year)
Annual nitrate load
The annual NO3 load is calculated by multiplying daily discharge and daily NO3
concentrations and summing up the daily values for the whole year. These values were
compared with the estimated annual export at Boen (at the outlet of reach 4), which are
based on monthly measurements and daily water discharge. Comparing these
measurements with the model simulations at reach 4 shows a similar pattern between the
two series (Figure 4). Plotting of observed vs. simulated annual NO3 loads in an XY-plot
gave a R2 of 0.96. The highest NO3 loads were underestimated by the model, and the
performance of the model in the control period was poorer than in the calibration period.
0
1997
Measured NO3-N
1999
2001
Modelled NO3-N
2003
Annual precipitation
Figure 4. Calculated annual nitrate loads based on measured and INCA-N simulated data(19962003).
11
Nitrate per land-use type
NO3 leaching differs among land-use types and the differences give an indication of the
relative contribution of each land-use type to the total nitrogen load in the catchment.
However, NO3 leaching is not equal to the total NO3 export in the catchment. After NO3
has leached from the soil to the water, in-stream processes like nitrification and
denitrification alters the N loads delivered to downstream areas. Considering the fact that
forests cover between 51% and 92% of each reach, the relative contribution to NO3
leaching is not high, varying between 7% and 70%. Arable land is only present in reach 4
and 5 with percentages of 5 and 40, respectively. Nevertheless, the contribution to total
NO3 leaching of this land-use type is up to 37% and 85%, respectively. The land-use
specific nitrogen leaching per hectare is given in Table 4 while Figure 5 shows the overall
contribution per land-use type.
Table 4. Nitrogen leaching from the different land-use types; weighted average of the five reaches in
the period 1996-2003.
Forest – Productive
Forest - Unproductive
Peat Land
Mountainous vegetation
Arable land
Lake surfaces
Nitrogen leaching
(kg N/ha/yr)
1.3
2.5
2.0
1.9
159
8.3
100%
80%
60%
40%
20%
0%
Reach 1
Forest - Productive
Peat Land
Arable Land
Reach 2
Reach 3
Reach 4
Reach 5
forest - Unproductive
Open mountainous vegetation
Lake Surface
Figure 5. Leaching of nitrate from the different land-use types in the five reaches of the Tovdal
catchment in the period 1996-2003.
12
Acknowledgements
This work was conducted as part of the Norwegian CLUE project (NORKLIMA
programme of the Research Council of Norway) and the EU Eurolimpacs project (the
Commission of European Communities GOCE-CT-2003-505540) with additional support
from the Norwegian Institute for Water Research. Major parts of the analyses are from
the diploma thesis of Mirjam Bloemerts, Wageningen University for Life Sciences,
whose work was supported in part by a fellowship from Wageningen University.
References
Aas, W., S. Solberg, T. Berg, S. Manø, and K. E. Yttri. 2006. Monitoring of Long-Range
Transported Air and Precipitation -- Atmospheric deposition, 2005. Report
955/2006, State Pollution Control Authority, Oslo, Norway.
DN. 2005. Kalking av vann og vassdrag. Effektkontroll av større prosjekter 2004. Notat
2005-2. Direktoratet for naturforvaltning, Trondheim, Norway.
Otnes, J. and Ræstad, E.: 1978, Hydrology in practice [in Norwegian]. Engineering
Publishers (Ingeniørforlaget), Oslo, Norway, 314 pp.
Rekolainen, S. and Posch, M.: 1993, ‘Adapting the CREAMS Model for Finnish
Conditions’, Nordic Hydrol. 24, 309-322.
SFT 2006. The Norwegian monitoring programme for long-range transported air
pollutants. Annual report - Effects 2005. TA-2205/2006, The Norwegian
Pollution Control Authority (SFT), Oslo.
Vehviläinen, B., 1992. Snow cover models in operational watershed forecasting.
Doctoral Thesis, National Board of Waters and the Environment, Helsinki,
Finland, 112 pp.
Wade, A.J., Durand, P., Beaujouan, V., Wessel, W.W., Raat, K.J., Whitehead, P.G.,
Butterfield, D., Rankinen, K., and Lepistö, A., 2002. A nitrogen model for
European catchments: INCA, new model structure and equations. Hydrol. Earth
System Sci., 6, 559-582.
Whitehead, P.G., Wilson, E.J., and Butterfield, D., 1998. A semi-distributed Integrated
Nitrogen model for multiple source assessment in Catchments (INCA): Part I –
model structure and process equations. Sci. Tot. Environ., 210/211, 547-558.
13