Nitrogen cycling in large temperate floodplain rivers of contrasting nutrient regimes and management

Hydraulic connection between channels and floodplains (“connectivity”) is a fundamental determinant of ecosystem function in large floodplain rivers. Factors controlling material processing in these rivers depend not only on the degree of connectivity but also on the sediment conditions, nutrient loads, and source. Nutrient cycling in the nutrient‐rich upper Mississippi River (MISS) is relatively well studied, whereas that of less eutrophic tributaries is not (e.g., St Croix River; SACN). We examined components of nitrogen cycling in 2 floodplain rivers of contrasting nutrient enrichment and catchment land use to test the hypothesis that N‐cycling rates will be greater in the MISS with elevated nutrient loads and productivity in contrast to the relatively nutrient‐poor SACN. Nitrate (NO3−‐N) concentrations were greatest in flowing habitats in the MISS and often undetectable in isolated backwaters except where groundwater inputs occurred. In the SACN, NO3−‐N concentrations were greatest in the flowing backwater where groundwater inputs were high. Ambient nitrification in the MISS was twice that in the SACN and tended to be lowest in the main channel. Denitrification was 3× greater in the MISS than that in the SACN, N‐limited in both rivers. Community production/respiration was >1 in the MISS and likely provisioned labile C to fuel microbial metabolism and dissimilatory NO3−‐N reduction, whereas the heterotrophic (production/respiration < 1) nature of the SACN likely limited microbial metabolism and NO3−‐N dissimilation. It appears that N‐cycling in the SACN was driven by groundwater, whereas that in the MISS was supported mainly by water column N‐sources.

The geomorphology of large floodplain rivers is similar globally (Lewin, 1978), often consisting of isolated backwaters (IBW), flowing backwater (FBW) channels, and main channels (MC; Lewis Jr., Hamilton, Lasi, Rodriguez, & Saunders III, 2000). In FBW, hydraulic retention time is intermediate between IBW and MC. The rapidly flowing MC typically contains carbon-poor sediments and short hydraulic retention time, limiting certain microbial nitrogen uptake and removal, whereas oxygenated water inhibits denitrification (DEN) and release of sediment-bound phosphorus. In temperate floodplain rivers, IBW tend to exhibit lower water column nitrate and greater phosphorous concentrations during base flow periods than do more connected backwaters (Heiler, Hein, Scheimer, & Bornette, 1995). During extended periods of disconnection, water column nitrate concentrations often decline below detection due to uptake and removal processes and lack of surface water replenishment. Backwaters with groundwater inputs will exhibit less severe nitrate depletion. In contrast, dissolved phosphorus concentrations may increase, as phosphorus-rich sediments become anoxic due to the large microbial oxygen demand and release of sediment-bound phosphorus.
This general pattern of N-cycling is well described for enriched temperate zone floodplain rivers (e.g., Danube: Welti, Bondar-Kunze, Tritthart, Pinay, & Hein, 2011;Seine: Sebilo et al., 2006;and Upper Mississippi River [UMR]: Richardson, Strauss, Monroe, Bartsch, & Soballe, 2004;Strauss, Richardson et al., 2004). Yet the role of ecosystem productivity and process drivers (e.g., nutrient loading), which may have strong influence on N-cycling processes, have not been compared in rivers of contrasting nutrient loading and productivity. The  (Table 1). Dominance of agriculture and greater human population in the MISS catchment results in greater movement of phosphorus and nitrogen from land to river in the MISS, relative to the SACN (Kroening & Andrews, 1997). The St. Croix River is a nationally unique ecosystem with a relatively pristine environment, high biological diversity, exceptional water quality, and progressive management history. More than 30 years since federal designation, the St. Croix River Basin is still predominantly forested (Larson et al., 2002) and the river corridor above St. Croix Falls is representative of pre-European settlement conditions (Minnesota-Wisconsin Boundary Area Commission, 2002). In contrast, the study reach of the MISS corridor begins in rural environments, passes through two large metropolitan cities (Minneapolis and St. Paul), and then returns to a rural setting at its downstream end.
Despite substantial water quality degradation, the river valley contains diverse flora and fauna, and much of the river still has some connection to the floodplain, maintaining some historic river ecosystem function (Lafrancois, Magdalene, Johnson, Vandermelen, & Engstrom, 2013).
Our understanding of habitat-specific N-cycling in less enriched temperate zone floodplain rivers (e.g., SACN) is poorly tested. The striking contrast between the MISS and SACN in nutrient and sediment loading provides an opportunity to evaluate important questions relevant to management regarding the role of natural river ecosystem functioning as a determinant of nutrient retention and cycling (especially nitrogen) by floodplain rivers. In particular, Lake St. Croix, a natural lake within the SACN, is undergoing rapid increases in nitrate concentrations and is categorized as mesotrophic (Vander Meulen & Elias, 2008). The goals of our research were to (a) characterize nutrient conditions in MC, FBW, and IBW of the SACN and MISS over a range of river discharge and (b) investigate biogeochemical processes affecting N-cycling in these habitats.

| Site description
Our study focused on a set of MC, FBW, and IBW complexes (n = 2 sites per habitat) in each river ( Figure 1). Sediment composition was variable but primarily sand in MC sites, a mixture of sand, silt, clay in FBW, and a fine clay and silt in IBW sites. Water depth ranged from 0.5 to 1 m, depending upon discharge (range, SACN: 29.7 to 414.5 m 3 /s; MISS: 73.6 to 1,359.3 m 3 /s).

| Nitrogen biogeochemistry and sediment nutrient assays
Biogeochemical assays and sediment nutrient measurements were conducted in the SACN and MISS to determine spatial and temporal variation in rates of nitrogen cycling. In May and July 2008 and August 2009, at each site, we collected duplicate sediment cores (7.6-cm diameter, 5-cm depth), which were homogenized, subsampled, and analysed for nitrification, DEN, ammonium (porewater and exchangeable NH 4 + -N), and porewater nitrate-nitrite (NO 3 − -N), total carbon (TC), and total nitrogen (TN). Sediment porewater was collected by centrifugation for NH 4 + -N and NO 3 − -N and analysed following Richardson et al. (2004). Unionized porewater ammonia concentrations (NH 3 -N) were calculated based on in situ pH and temperature measurements, taken during sampling, following Emerson, Russo, Lund, and Thurston (1975). Sediment bulk density, water content (a surrogate of organic matter content), and porosity were determined following Robertson, Coleman, Bledso, and Sollins (1999

| Productivity and metabolism
Water quality sondes were deployed in each river at MC, FBW, and IBW sites for continuous measurements of DO and temperature for 1 week prior to water sampling (during July-August 2007 and May-June 2009).
Open-channel, single-point ecosystem metabolism was estimated using diurnal changes in DO (Owens & Crumpton, 1995). Photosynthetically active radiation was measured at the same locations with an LI-COR model LI-250 metre (Lincoln, Nebraska, USA) and underwater quantum sensor to estimate euphotic zone depth in accordance with Wetzel (2001). Gross and net production and respiration were modelled using the single-station method (Bales & Nardi, 2007).

| Statistical analyses
Our goal was to identify large differences in patterns of N-cycling rates between rivers and among habitats of the two rivers and to identify drivers of these rates and patterns. Analysis of variance (JMP®, 2016; Table S3) was used to test for differences in magnitude of response variables among rivers, habitats, interaction of habitats × rivers, and sample periods (dates). An information-theoretic approach was used to identify the best models that predicted DEA rates based on a suite of predictor variables. Separate analyses were conducted for each river that resulted in a set of predictor variables that had the minimum Akaike Information Criteria (AIC) for observed DEA rates. We present the best models based on Burnham and Anderson (2002) where competing models with Δ i values <2 are the best Kullback-Liebler model. where AIC min is the minimum AIC value for all models and AIC i is the ith competing model. Similarities in the suite of predictors were compared to conceptual models in an attempt to describe processes occurring under the different nutrient regimes in each river. We used model averaging, which tends to shrink estimates of weaker terms, yielding better predictions. Models were averaged with respect to AICc weight following: We averaged models with 1 to 8 terms using a cut-off AIC i weight quantile of 0.95 and present predictor coefficients with confidence intervals that do not include zero (JMP®, 2016).

| Porewater dissolved inorganic nitrogen
Average concentrations of sediment porewater NH 3 -N (Figure 3a

| Sediment C and N
The total C (% of total mass) in the top 5 cm of sediments was~5× greater in the MISS than that in the SACN (Figure 4). Average % C in the MISS was greatest in FBW, whereas in the SACN, the IBW sediments had the greatest % total C. Sediment total % N in the SACN IBW was 3× greater than in other SACN habitats. Sediment % total N in MISS were similar across all habitats. The great disparity in sediment total C between the two rivers resulted in a 2× greater C:N (atoms) in the MISS relative to the SACN.

| Potential nitrification
Average rates of potential nitrification were 6× greater in the MISS than In the MISS, the IBW was most strongly N-and C-limited, whereas the MC was the least. Addition of C-alone had little effect on DEN in any habitat. The IBW generally exhibited the strongest N limitation on all dates (Figure 6a,b,c), whereas FBW and MC showed less distinct limitation. In the SACN, the IBW also exhibited the strongest N limitation, particularly in summer when water temperatures were highest and river discharge was lowest.
DEA was significantly greater in the MISS (p < .0001) and across habitats (p < .0001) in both rivers (  Our results show that the MISS is "supercharged" with fuel needed to drive nitrogen biogeochemisty at high rates, compared to the SACN system, which has a leaner supply of nitrogen and carbon to drive the biogeochemical engine. In the MISS, DEA was generally much greater than reported in the UMR Pool 8   Note. Displayed are the best equivocal models (AICc < 2 different).
SW NH 4 + -N = surface water ammonium; SW NO 3 − -N = surface water nitrate; PW NO 3 − -N = sediment porewater nitrate; PW NH 4 + -N = sediment porewater ammonium; porosity = sediment porosity; water content = sediment % water content; C:N ratio = sediment C:N atom ratio; pot. (potential) nitrification = ambient nitrification; AICc = Akaike Information Criterion. studies have shown that nitrogen limitation is a general phenomenon related to delivery of nitrate (e.g., Richardson et al., 2004;Welti et al., 2012). Additionally, the consortia of denitrifying bacteria vary in their response to fluctuating porewater oxygen concentrations, exhibiting reduced DEA in habitats with variable oxygen concentration (Tatariw, Chapman, Sponseller, Mortazavi, & Edmonds, 2013), as likely found in the MC of both the MISS and SACN.
Further, regression models suggest that nitrate is being supplied differently to denitrifiers in the two rivers. In the MISS, water column sources of both nitrate and ammonia likely control DEN (Figure 7). In contrast, DEA in the SACN is relatively low and likely dependent on groundwater sources of DIN. Surface waters are low in DIN, and river discharge likely supplies little DIN to denitrifiers. In the SACN, greater concentrations of porewater nitrate and greater sediment water content suggest that groundwater, supplemented with nitrification, provides nitrate in excess of metabolic demands of the sediment microbial community (Welti et al., 2012). The relatively low sediment carbon content and low rates of primary production in the SACN likely further limit microbial metabolism and N-cycling (Stelzer, Scott, & Bartsch, 2014). In these two rivers of contrasting productivity, it is highly likely that Ncycling rates in the MISS are being supported by carbon produced in the water column and captured in sediments (Tatariw et al., 2013).
In the MISS, elevated sediment ammonium and nitrification, but low concentrations of sediment nitrate, suggests that DEN was rapid enough to remove nitrate at rates greater than nitrification. Further, in the MISS, a positive relationship between nitrification and DEA suggests a tight coupling between these processes and potential nitrate limitation. Sediment ammonium concentrations in the MISS are likely linked to high sediment carbon and nitrogen concentrations via mineralization and oxygen limitation of nitrifiers to process this ammonium.
In the MISS, percent sediment total C and N and C:N ratios were generally two to five times greater than those measured in the SACN sediments, particularly in the flowing channels. These elevated C and N concentrations in the sediments suggest that high rates of organic matter accumulation occur in the MISS. Dodds et al. (2013) found the lower Mississippi River (nr. Baton Rouge, Louisiana, USA) to be almost completely heterotrophic except for short periods in winter. Although Houser, Bartsch, Richardson, Rogala, and Sullivan (2015) found the MISS (nr. La Crosse, Wisconsin, USA) MC to be autotrophic during the growing season, associated with elevated chlorophyll a concentrations and a drawdown of soluble phosphorus. The elevated rates of primary production (e.g., P/R ratio 1.1 to 1.6) in the MISS channels suggest that the source of this sediment organic matter is likely autochthonously produced phytoplankton. Entrainment of phytoplankton into river sediments can provide a high-quality source (Canuel, 2001) of both carbon and nitrogen to drive N-cycling processes (Fischer et al., 2005). Movement of particulate C into benthic sediments is common in sandy sediments of large rivers and likely supports heterotrophic biogeochemical processes (Fischer, Sukhodolov, Wilczek, & Engelhardt, 2003). The high concentrations of C and N may also be derived from bacterial secondary productivity in sediments. Fischer et al. (2005) measured bacterial carbon production in channel sediments of the Elbe River in Germany (0.95 g C m 3 /h), which was three to five times greater than that in nearshore areas and N-cycling was a significant portion of the carbon budget.
In contrast to the MISS, origins of C and N in the SACN are likely to be substantially different and derived less from autochthonous but more from allochthonous sources. Reduced nutrient concentrations of the SACN limit water column productivity. Further, channels of the SACN course through wooded riparian zones where the forest canopy and shading often reach mid-channel. Water column metabolism in the SACN was typically more heterotrophic (P/R ranging 0.1 to 0.9) than that in the MISS. Sediment C and N in the SACN reflect the reduced Water content is a surrogate for sediment total carbon. Grey arrows indicate variable relationship with DEA rates water column productivity, such that C (%) and N (%) were at least half that of the sediments in the MISS. The implications are that labile C available to fuel microbial activity in the SACN is much reduced relative to those of the MISS; further, sediment ammonium concentration are less than half than those in the MISS, whereas sediment nitrate is 30 times that in the MISS sediments. We suspect that local groundwater inputs were the source of nitrate because nitrification rates were relatively low and likely not capable of producing such large concentrations of nitrate, particularly in the presence of relatively high DEA.
Differences between these two rivers in nutrient loading and productivity are likely related to differences in catchment land use and sewage outfall volume (Kloiber, 2004). In particular, greater human population densities, greater agricultural land use, and less forest result in greater nutrient loading in the MISS than in the SACN. These patterns have existed for at least a century (Lafrancois, Magdalene, & Johnson, 2009) but with indications of declining phosphorus and increasing nitrogen load in the SACN and MISS rivers. Management of agriculturally derived sediments may be contributing to declining P loads, whereas increasing nitrate loads likely originate from a number of sources including atmospheric, agricultural, and urban (Kloiber, 2004;Kroening & Andrews, 1997).

| CONCLUSIONS
Two large floodplain rivers in the UMR catchment of differing nutrient load and productivity showed similar spatial patterns of nitrogen biogeochemical functioning. DEN rate tended to be highest in backwater lakes and side channels where sediment C and N accumulate. However, rates of nitrogen cycling and carbon and nitrogen accumulation in sediments were much greater in the MISS where loads of N and P and productivity are much greater than in the SACN. We suggest that the fuel needed to support high rates of N-cycling is paradoxically provided by these elevated nutrient inputs from surrounding agricultural and urban landscapes. Current anthropogenic nutrient inputs likely far overwhelm the capacity of these rivers to naturally remove the inputs before being transported downstream.