Increases in nitrogen inputs from fertilizer application, the planting of legumes, nitrogen oxide emissions, and land use changes [Galloway et al., 2008] have led to nitrogen saturation [Aber et al., 1989], eutrophication [Diaz and Rosenberg, 2008] and groundwater contamination [Howden et al., 2011] in ecosystems. These increases in nitrogen inputs, and their effects, have led to a growing interest in processes that can remove nitrate and other forms of available nitrogen from ecosystems [Burgin and Hamilton, 2007; Lansdown et al., 2012] and have led to reach-scale [Bernhardt et al., 2002; Stelzer et al., 2011a], network-scale [Wollheim et al., 2008; Alexander et al., 2009], and continental-scale studies [Alexander et al., 2000; Mulholland et al., 2008; Hall et al., 2009] of nitrogen transport and removal in rivers. It is well known that processes in upland soils [Groffman and Tiedje, 1989; Seitzinger et al., 2006], groundwater [Groffman et al., 1996; Tesoriero and Puckett, 2011], riparian zones [Hedin et al., 1998; Hill et al., 2000; Duff et al., 2007], hyporheic zones (where groundwater and surface water mix) [Triska et al., 1989; Hill and Lymburner, 1998; Böhlke et al., 2009; Zarnetske et al., 2011] and in the surface water of streams and rivers [Mulholland et al., 2008; Heffernan et al., 2010] can transform, retain and remove available nitrogen. Much less is known about the role of deep sediments (which we operationally define as those sediments greater than 5 cm beneath the stream channel) in nitrogen processing. Most studies of nitrogen processing in streams do not include deep sediments, particularly those deeper than the hyporheic zone, which can be very shallow in gaining streams [Hill and Lymburner, 1998; Stelzer et al., 2011b]. Nitrogen processing studies at reach scales are usually designed to measure nitrogen uptake or transformation in stream channels [Mulholland et al., 2008; Heffernan et al., 2010] and those studies at smaller spatial scales typically emphasize shallow sediments. For example, most studies of denitrification in stream sediments only include denitrification measurements from surficial sediments (cores less than 5 cm deep) [Arango et al., 2007; Herrman et al., 2008] (also see review in Stelzer et al. [2011b]). In gaining streams groundwater often passes through substantial quantities of sediment before discharging to the stream. Previous studies have shown that available nitrogen is retained along upwelling flow paths in deep sediments [Duff et al., 2008; Puckett et al., 2008; Stelzer et al., 2011b]. However, most previous studies have not included process-oriented measurements in deep sediments (but seeStorey et al. , Fischer et al. , Inwood et al. , Stelzer et al. [2011b], and Lansdown et al. ) or have not included the fine-scale vertical profiles of available nitrogen necessary to infer where nitrogen retention occurs in deep sediments.
1.1. Conceptual Model for Nitrate Removal in Deep Sediments
 We showed in a previous study [Stelzer et al., 2011b] that nitrate removal occurred in deep sediments in Emmons Creek, a gaining stream in the Central Sand Ridge Ecoregion in Wisconsin. Based on this work and previous studies demonstrating nitrate loss along upwelling flow paths in stream sediments [Duff et al., 2008; Puckett et al., 2008; Krause et al., 2009] and redox-driven nitrogen transformations in riparian soils [Hedin et al., 1998] we developed a conceptual model for nitrate removal in deep sediments of groundwater-fed streams (Figure 1). The model illustrates nitrate-rich oxic groundwater upwelling through deep sediments before discharging to the surface water of a stream. Deposition and burial of POM (including fine and coarse particulate organic matter) in sediments [Metzler and Smock, 1990] creates a zone with favorable redox conditions (severe hypoxia or anoxia) for nitrate to serve as the terminal electron acceptor during microbial respiration, facilitating denitrification. POM and the presence of various metals [Puckett et al., 2008] serve as electron donors. As the groundwater encounters this zone, denitrification, and perhaps other nitrate removal processes, causes a decrease in nitrate concentration along the flow path. We think the depth and thickness of this zone will vary within and among streams and will be affected by vertical hydraulic gradient and sediment porosity [Triska et al., 1989]. This zone with favorable redox conditions for denitrification may occur below the depth at which surface water and groundwater mix (i.e., hyporheic zone), as shown in Figure 1, or it could encompass all or part of the hyporheic zone. This model has some similarities with previous conceptual models that have emphasized groundwater-surface water interactions in streams [Triska et al., 1989; Vervier et al., 1992; Krause et al., 2009], but is different than most previous conceptual models of hyporheic zone processes, which have focused on transport of surface water into the hyporheic zone and subsequent return to the channel [Jones and Holmes, 1996; Boulton et al., 1998; Mulholland and DeAngelis, 2000; Boulton et al., 2010]. The notion that changes in redox state can drive nitrogen transformation has also been applied in many previous studies of nitrogen loss in groundwater systems at the watershed scale [e.g., Böhlke and Denver, 1995].
Figure 1. Conceptual model of nitrate processing in stream sediments indicating deposition and burial of particulate organic matter (POM) and decline in nitrate and dissolved oxygen concentrations along upwelling flow paths in a zone favorable for denitrification.
Download figure to PowerPoint
 We developed three hypotheses from this conceptual model: (1) Nitrate concentration of groundwater will decline along upwelling flow paths before discharging to surface water, (2) denitrification potential in deep stream sediments will substantially contribute to total (depth-integrated) denitrification potential in sediments, and (3) dissolved oxygen concentration will decline along upwelling flow paths as groundwater passes through the zone of anoxia.
 We tested these hypotheses by developing vertical nitrate profiles to about 1 m depth, by measuring denitrification potential at sediment depths to 30 cm, and by comparing dissolved oxygen (as a proxy for redox state) between deep groundwater and pore water at 5 cm sediment depth at eight sites in streams and rivers in the Waupaca River Network in Central Wisconsin. Our primary goal was to determine if our conceptual model for nitrate removal in deep stream sediments was applicable throughout a river network. Our hypotheses and conceptual model were broadly supported. In particular, we found there was substantial denitrification occurring at depth and most streams showed a loss of nitrate as groundwater upwelled through sediments. Changes in dissolved oxygen between deep groundwater and pore water at 5 cm were consistent with redox-driven nitrate loss.