The main mechanisms of denitrification and their factors are generally un-derstood and described by many authors (e.g. Boyer et al. 2006, Groffman et al. 2006, 2009a) but the combination of the four most important settings, waterlogging, nitrogen supply, bacterica community and energy source (Boyer et al. 2006) vary with time and space, depending on environmental settings and hence makes the modelling and measurement of denitrification a com-plex task (Boyer et al. 2006, Groffman et al. 2006). Indirect measurements of denitrification can only be carried out and results from mesocosms and study wetlands have to be upscaled to basins (Groffman et al. 2009a). Since deni-trification is a biogeochemical process, in which microorganisms are involved, measurements do not measure denitrification, but results from denitrification processes (see Groffman et al. (2006) for an overview of denitrification
mea-surements). Consequently retention models do not account for denitrification itself, but environmental settings, which affect the process of denitrification.
There are various more or less complex deterministic and empirical retention models for aquatic and terrestrial ecosystems available (see Boyer et al. (2006) for an overview). Because of data availability this work focusses on yearly and monthly empirical retention models.
Modelling of nutrient retention in rivers
To quantify the effect of floodplains on nutrient retention in a river system, the retention in the river has also to be known in comparison. There are several river retention models (Alexander et al. 2009, Behrendt & Opitz 2000, Boyer et al. 2006, Venohr 2006). Since the retention model developed by Behrendt &
Opitz (2000) has been established for retention calculation in German rivers and modules for N- and P-retention calculations are both available (see below), it is applied in this work. Additionally, this approach is also applied for calcu-lating nutrient retention in floodplains, which is presented in the next section.
The riverine retention module for P depends on the hydraulic load (HL) as the main factor. HL as the reciprocal of the water residence time can be concluded from the water surfaces of rivers and lakes and the corresponding discharge re-spectively specific runoff (Behrendt & Opitz 2000, Venohr 2006). It represents the contact area between sediment and water (Behrendt & Opitz 2000) which is incorporated in many other retention models (Boyer et al. 2006, Groffman et al. 2009a). N-retention is modelled as the sum of sedimentation and den-itrification since TN (total nitrogen) is considered in the model and with it dissolved inorganic nitrogen as the most important fraction of which nitrate forms the main compound. This approach was enhanced by Venohr (2006) to be modelled according to aTemperature andHLdependent approach (THL) on a yearly basis. Further extensions were carried out by Venohr et al. (2011), by adding global Radiation as the third input factor (THLR-approach) on a monthly basis to consider N-uptake and N-release by plants.
Modelling of nutrient retention in wetlands
Three different approaches can be found in literature to quantify nutrient re-tention in wetlands and floodplains.
Firstly, nutrient retention can be described as a linear or exponential rela-tionship, depending on either only nutrient load (Mander & Mauring 1994) or the combination of nutrient load and the wetland area (Byström 1998), and
additional parameters such as nutrient concentration (Arheimer & Wittgren 2002, Dortch & Gerald 1995), water temperature (Arheimer & Wittgren 2002) and/or residence time respectively hydraulic load (HL) (Dortch & Gerald 1995, Fisher & Acreman 2004). This approach is chosen when retention processes are described in concrete study wetlands where water surfaces are assumed to be constant over the studied period.
Secondly, based on the results of the first approach, retention proxies are combined with given constant floodplain extent to calculate nutrient reten-tion on the catchment scale (Kronvang et al. 2004, Schulz-Zunkel et al. 2012).
Thereby, a linear relationship is assumed between retention rate and wetland area. Schulz-Zunkel et al. (2012) considered the National Floodplain Inven-tory as a spatial basis for a first estimate of N- and P-retention in floodplains for German river systems only very recently. Here, floodplain characteristics (landuse, soil type) were applied to modify denitrification and sedimentation rates described in literature, which were then applied to upscale these val-ues for landscape scale calculations. Kronvang et al. (2004) applied land-use characteristics of the catchment as an indicator for modifying denitrification, whereas one constant sedimentation rate is assumed.
Thirdly, retention can be considered as the difference of emissions from the catchment into the river and loads transported in the river. This connection was found by Behrendt (1996, 1999), Behrendt & Opitz (2000) who initially had developed a nutrient (N and P) emission inventory and found discrep-ancies between calculated nutrient emissions and measured nutrient loads in several European rivers. Based on these results Behrendt & Opitz (2000) de-rived an empirical retention model which has already been described in the section above because it defines retention as the sum of removal processes in the river system, including all water surfaces (lakes, rivers, wetlands, inun-dated floodplains). Thus, retention processes in the river and in the floodplain cannot be distinguished. Nevertheless, this retention model is also applied to calculate the role of floodplains for nutrient retention as a measure, when dyke relocation activities are carried out. Venohr et al. (2011) assume that the retention in inundated floodplains can be expressed by the same algorithm as for retention calculations in the river itself because retention processes are comparable. Here the crucial parameter water surface area for calculating the hydraulic load is assumed as a constant derived from land-use data or esti-mates. But the calculated effects of dyke relocation on nutrient retention have not been validated.
Since inundation of floodplains is temporally and spatially variable, the in-undated floodplain extent as well as th incoming load have to be modelled as variables and not as constants which has not been done so far.