Text S1
2012JG001985
The INCA model: structure, input data and calibration procedures
1. Model structure.
The Integrated Catchment Model of Nitrogen (INCA-N, v1.11.10) is a process-based model that
integrates hydrology and catchment/river N processes and simulates daily discharge and NO3 and
NH4 concentrations [Whitehead et al., 1998; Wade et al., 2002]. The model structure accounts for
four ecosystem compartments: plant, soil, groundwater and stream. Sources of N include
atmospheric deposition, N fixation and anthropogenic inputs (fertilizer application and direct
discharges). Internal cycling of N includes biological uptake, mineralization and nitrification.
Outputs of N consider denitrification and hydrological export. The stream compartment operates
as a multi-reach system where equations are solved to maintain a mass balance along the river.
For the present study, we used the INCA-N version 1.11 that is available at
http://www.reading.ac.uk/INCA/.
2. Input data and model calibration.
The INCA model requires daily measurements of precipitation, air temperature, soil moisture
deficit (SMD), and hydrologically effective rainfall (HER). To derive input data for the Santa Fe
(subhumid) and Fuirosos (semiarid) catchments during the period 2004–2006, we used climatic
data provided by the Montnegre-Corredor and the Montseny Natural Protected Areas from two
meteorological stations located nearby our study sites (the Can Lleonart station for the subhumid
catchment; the Dos Rius station for the semiarid catchment). We calculated SMD and HER as
described by Bernal et al. [2004]. In the model, the HER drives the water flow and N flux
through the soil compartment. For the semiarid catchment, where unsaturated flow paths often
occur due to low water availability [Medici et al., 2008], we adjusted the HER initially calculated
with standard methods to get the best fit between observed and simulated stream discharge. We
used field measurements to estimate the a and b parameters from the velocity/discharge
relationship V=aQb from which the model derives the residence time of water.
Land uses, catchment area and reach length for each study site were derived from a digital
land use/land cover and a river network LandSat maps 1:50000 (BT-50M v.3.0) provided by the
Catalan Government (http://www.aca-web.gencat.cat) and the Centre for Ecological Research
and Forestry Applications (CREAF) (http://www.creaf.uab.es/mcsc/esp/descriptiu.htm). For
atmospheric N inputs, we used the long-term average of N deposition in the study area that is
18.6 kg N/ha/year (60% and 40% as dry as wet N deposition, respectively) [Àvila et al., 2010].
We followed the calibration procedure proposed by Butterfield et al., [2006]. First, we
calibrated the hydrological component of the INCA model until we got a good fit between
simulated and observed values. Then, we adjusted the terrestrial component, that is the soil N
processes (nitrification, denitrification, mineralization) and those parameters related with tree
physiology (growth period, uptake rates). During this setting-up process, we kept the in-stream
model parameters inactive (i.e, in-stream denitrification, in-stream nitrification, stream length).
To fit the plant/soil N parameters, we used either, field or published data, if available. Otherwise,
we used “soft-data” (based on the experimental knowledge of N processes) provided by
Butterfield et al. [2006] following recommendations from Rankinen et al. [2006]. Parameters
were adjusted to the peaks of the NO3 and NH4 chemographs to obtain: (i) the best match
between simulated and measured stream inorganic N concentrations, and (ii) simulated N annual
rates and N fluxes similar to those reported in the literature [Bernal et al., 2004; Butterfield et al.,
2006]. Due to the inherent difficulty of measuring hydrological and soil processes in the field at
the same scale as conceptualized in model structures, all the calibrated parameters have limited
physical meaning [Oreskes et al., 1994]. Futhermore, several sets of parameters may give a good
fit (or even, a better fit) between simulated and measured values, a problem known as equifinality
(the impossibility to find an optimal parameter set in hydrological modelling) [Beven, 2006].
Despite these unavoidable limitations when dealing with complex distributed models, the
application of the INCA model in our study was useful because, once set-up and calibrated, it
allowed for running in-stream scenarios to test the functioning of this model compartment.
For each sub-catchment, we estimated the goodness of fit for both, discharge and stream
water dissolved inorganic N concentration (DIN-N = NO3−N + NH4−N) with the coefficient of
determination (r2) between simulated and measured data. Moreover, we quantified differences
between water outputs and N loads estimated from observed and simulated data.
3. Model simulations.
The INCA model was able to successfully recreate the variability of stream discharge for both,
the subhumid (r2 = 0.88, p < 0.01, n = 50) and the semiarid (r2 = 0.95, p < 0.01, n = 46)
catchments. However, the model tended to overestimate stream discharge, in particular for the
subhumid catchment (Table A1). Regarding DIN, the model matched the magnitude as well as
the seasonality of stream water concentrations, though the explained variability was lower than
for stream discharge (r2 = 0.61, n = 50 for the subhumid catchment; r2 = 0.46, n = 38 for the
semiarid catchments; p < 0.01 in both cases). Simulated and empirically estimated annual stream
N fluxes differed by < 40% (Table S1).
Table S1. Annual discharge and catchment DIN export measured empirically and simulated with
the INCA model for the subhumid and semiarid catchments during two consecutive water yearsa.
Catchment
Subhumid
Water Year
2004-2005
2005-2006
Water Flux (mm)
Observed
Simulated
94.1
102.8 (+9.2 %)
229.4
381.6 (+66.3 %)
Semiarid
2004-2005
2005-2006
24.8
76.9
21.6 (-12.9 %)
48.8 (-36.5 %)
2
DIN (kg N/km /year)
Observed
Simulated
9.5
10 (+5.2 %)
39.4
33.5 (-14.9 %)
12.6
31.4
11.1 (-11.9 %)
43.5 (+38.5 %)
a
The relative difference between observed and simulated values is shown in parenthesis. The plus
sign (+) indicates that simulated values were above empirical values. The minus sign (–) indicates
the opposite.
References
Àvila, A., R. Molowny-Horas , B.S. Gimeno, and J. Peñuelas (2010), Analysis of decadal time
series in wet N concentrations at five rural sites in NE Spain, Water Air Soil Pollut., 207(1–4),
123–138, doi:10.1007/s11270-009-0124-7.
Bernal, S., A. Butturini, J.L. Riera, E. Vázquez, and F. Sabater (2004), Calibration of the INCA
model in a Mediterranean forested catchment: the effect of hydrological inter-annual
variability in an intermittent stream, Hydrol. Earth Syst. Sci., 8(4), 729–741.
Beven, K (2006), A manifesto for the equifinality thesis, J. Hydrol., 320:18–36.
Butterfield, D., A.J. Wade, and P.G. (2006), INCA-N v 1.9 user guide, Univ. of Reading, UK, pp
106.
Medici, C., A. Butturini, S. Bernal, E. Vázquez, F. Sabater, J.I. Vélez, and F. Francés (2008),
Modelling the non-linear hydrological behaviour of small Mediterranean forested catchment,
Hydrol. Process., 22, 3814–3828.
Oreskes, N., K. Shrader-Frechette, and K. Belitz K (1994), Verification, validation and
confirmation of numerical models in earth sciences, Science, 263, 641–646.
Rankinen, K., T. Karvonen, and D. Butterfield (2006), An application of the GLUE methodology
for estimating the parameters of the INCA-N model, Sci. Total Environ., 365, 123–139.
Wade, A.J., P. Durand, V. Beaujouan, W.W. Wessel, K.J. Raat, P.G. Whitehead; D. Butterfield,
K. Rankinen, and A. Lepisto (2002), A nitrogen model for European catchments: INCA, new
model structure and equations, Hydrol. Earth Syst. Sci. 6(3), 559–582.
Whitehead, P.G., E.J. Wilson, and D. Butterfield (1998), A semidistributed Nitrogen Model for
Multiple Source Assessments in Catchments (INCA): Part 1 – Model Structure and Process
Equations, Sci. Total Environ., 210/211, 547–558.