Modelling the inorganic nitrogen behaviour in a small Mediterranean forested catchment, Fuirosos (Catalonia)

Abstract

Abstract. The aim of this work was to couple a nitrogen (N) sub-model to already existent hydrological lumped (LU4-N) and semi-distributed (LU4-R-N and SD4-R-N) conceptual models, to improve our understanding of the factors and processes controlling nitrogen cycling and losses in Mediterranean catchments. The N model adopted provides a simplified conceptualization of the soil nitrogen cycle considering mineralization, nitrification, immobilization, denitrification, plant uptake, and ammonium adsorption/desorption. It also includes nitrification and denitrification in the shallow perched aquifer. We included a soil moisture threshold for all the considered soil biological processes. The results suggested that all the nitrogen processes were highly influenced by the rain episodes and that soil microbial processes occurred in pulses stimulated by soil moisture increasing after rain. Our simulation highlighted the riparian zone as a possible source of nitrate, especially after the summer drought period, but it can also act as an important sink of nitrate due to denitrification, in particular during the wettest period of the year. The riparian zone was a key element to simulate the catchment nitrate behaviour. The lumped LU4-N model (which does not include the riparian zone) could not be validated, while both the semi-distributed LU4-R-N and SD4-R-N model (which include the riparian zone) gave satisfactory results for the calibration process and acceptable results for the temporal validation process.

Progressive development of a hydrologic and inorganic nitrogen conceptual model to improve the understanding of small Mediterranean catchments behaviour

“…For example, the percolation capacity to the deeper aquifer, K pp became the third most influent parameter and the shallow aquifer residence time, T 3 , became the fourth most influential parameter. The results show that streamwater nitrate and ammonium concentrations are especially sensitive to parameters related to the shallow aquifer, in particular the nitrification constant (K nitr-aquif ), which is in agreement with the conclusion obtained by Medici et al (2010) , RRMSE(NH 4 ) and RRMSE(Q), illustrating the degree to which the three objective functions were minimized simultaneously with the lumped LU4-N model. The red points represent the best parameter sets corresponding respectively to the smallest value obtained for RRMSE(Q) and RRMSE(NO 3 ).…”

“…Figure 2 shows the E tot (Q) index plotted against the four most flowsignificant parameters. This figure allows the identification of the 'optimum' parameter value to be visualised, which is far away from the expert calibration result (Medici et al, 2008 and2010). Figure 2a also shows that the best Monte Carlo behavioural parameter set (depicted in Fig.…”

“…Random sampling of parameters values from uniform distributions was done, so that each parameter value had an equal chance of being chosen as part of a Monte-Carlo procedure whereby 100,000 input parameters sets were sampled independently from the user-defined ranges specified in Table 1. The ranges were specified using a priori knowledge and previous calibration and testing (Medici et al, 2008 and2010). Due to a lack of data with which to specify the initial soil available water and the soil, groundwater and streamwater nitrate and ammonium concentrations, then these initial conditions were also included in the GSA.…”

“…The LU4-N model is a lumped model that describes the Fuirosos catchment as homogeneous and represents the simplest conceptualization adopted; it is characterized by 27 parameters that require calibration (of which 8 are for the rainfall- N) is that the two semi-distributed models simulate water movement and inorganic nitrogen dynamics in the riparian zone, which was introduced to represent successfully the catchment drying-up period and the non-linear hydrological behaviour during the wetting-up period (Medici et al 2008 and2010). The riparian zone was also identified as a possible source of nitrate that entered the streamwater, especially after the summer drought period and as an important nitrate sink (due to denitrification) during winter/spring, the wettest period of the year (Butturini et al, 2002;Medici et al, 2010).…”

“…All model calibrations and subsequent predictions are subject to uncertainty (Seibert, 2003) show where knowledge gaps are most severe and which of the model parameters or aspects of the model structure most strongly affect prediction uncertainty (Wagener et al, 2003 and. In this paper, a general sensitivity analysis (GSA) using Monte Carlo simulations (Hornberger and Spear, 1980) and the Generalized Likelihood Uncertainty Estimation (GLUE) methodology (Beven and Binley, 1992) was done to three catchment-scale nitrogen models of varying complexity when applied to a small Mediterranean forested catchment, the Fuirosos, located in the northeast of Spain (Medici et al, 2008 andMedici et al, 2010). Specifically the aim of the work is to determine if additional model complexity gives a better capability to model the hydrology and nitrogen dynamics of the Fuirosos catchment.…”