Long-term Assessment of Floodplain Reconnection as a Stream Restoration Approach for Managing Nitrogen in Groundwater and Surface Water


 Stream restoration is a popular approach for managing nitrogen in degraded, flashy urban streams. Here, we investigated the long-term effects of geomorphic stream restoration on riparian and in-stream N transport and transformation in an urban stream in the Chesapeake Bay watershed. We examined relationships between hydrology, chemistry, and biology using a Before/After-Control/Impact (BACI) study design to determine how flashiness and N concentrations and flux changed after the restoration. We examined two independent surface water and groundwater data sets collected from 2002–2012 at our study sites in the Minebank Run watershed, modeled N flux, and compared our data to similar long-term data from the Baltimore Ecosystem Study LTER (BES) that served as reference sites. Restoration was completed during 2004 and 2005. Afterward, the monthly flashiness index, based on mean monthly discharge, decreased over time from 2002 and 2008. Groundwater nitrate (NO3−) concentrations trended slightly downward over time after the restoration at the restored site while dissolved organic carbon (DOC) concentrations trended upward whereas no trends were observed at the control site. Comparisons of NO3− concentrations with Cl− concentrations and specific conductance in both groundwater and surface water suggested that N reductions over time at the restored sites were not due to dilution. Similar patterns at BES sites suggested that declining NO3− was a function of restoration and watershed management, not larger regional factors such as decreased atmospheric inputs. DOC and NO3− were negatively related before and after restoration suggesting C limitation of N transformation. Long-term trends in surface water NO3− based on USGS data showed downward trends after restoration at both the restored and control sites while specific conductance showed no trend, suggesting that load reductions were not responsible for NO3− patterns. Modeled NO3− flux decreased post restoration in both the short and long-terms. Groundwater NO3− concentrations varied among stream features suggesting that some engineered features may be functionally better at creating optimal conditions for N removal. However, some engineered features eroded and failed post restoration thereby reducing efficacy of the restoration to reduce flashiness and NO3− flux. N management via stream restoration will be most effective where flashiness can be reduced, and DOC made available for denitrifiers. Stream restoration may be an important component of holistic watershed management including stormwater management and nutrient source control.


Introduction
Urban streams transport nitrogen (N) to downstream waters because channel degradation from ashy runoff, incision, and oodplain disconnection impair N transformation and uptake (Paul and Meyer 2001). Stream restoration designed to repair and reconnect stream channels, is an increasingly popular approach for managing N in urban streams. Such restoration attempts to improve hydrologic conditions favorable for N transformation and denitri cation by reducing ashiness, increasing residence times, and adding organic carbon for denitri ers (Mayer et al. 2010b: Duan et al. 2019). A recent synthesis suggested that there is potential for reducing N through stream restoration (Newcomer Johnson et al. (https://baltimoreecosystemstudy.org/). We targeted stream reaches for intensive study where channel geomorphology was reengineered to reconnect hydrology. We use both empirical and modeling approaches to corroborate results and examined our site both before and after restoration. Our results are intended to inform future stream restoration efforts designed to manage N in urban ecosystems.

Study design
We investigated how stream restoration in uenced N ux and concentration in groundwater and surface Here, we sought to place our research in the context of assessing the restoration of ecological processes and the identi cation of factors limiting those processes (Palmer 2009). Our speci c objectives were to examine relationships among DOC, NO 3¯, and Cl¯ in the surface water and groundwater pre-and post-restoration by employing a Before/After-Control/Impact (BACI) study design (Underwood 1992; Thompson et al. 2018). Our study used both long-term monitoring of a stream before and after restoration with intensive groundwater and surface water characterization, utilizing two independent data sets to produce empirical and modeling results. We used multiple data sets, including our own empirical data, USGS stream gage data, and published data from the BES LTER to develop corroborative nitrogen ux models spanning pre-and post-restoration study periods. We expected that nitrogen processing would be controlled by geomorphology, hydrology, and carbon supply.

Study area
Minebank Run is a 2 nd order urban stream located within an 8.47 km 2 watershed within Baltimore County, Maryland, USA in the eastern section of the Piedmont physiographic province (39 o 24'43"N and 76 o 33'12"W; Figure 1). Minebank Run ows in a northeasterly direction at approximately a 1% grade for 5.2 km, where it enters Gunpowder Falls, eventually draining into Chesapeake Bay (Doheny et al. 2006). Land use in the Minebank Run watershed is over 80% urban/suburban (Doheny et al. 2006). Between 1960-70s, rapid urbanization led to severe channel degradation that was addressed by installing concrete umes in the channel, which, by the 1990s, had been eroded out of place (Sortman 2004). The high proportion of impervious surface in the watershed, including the Interstate-695 Beltway, in a region of signi cant topographic relief, combined to produce ashy hydrology, eroded banks, incised channel bed, and general geomorphic instability. Hydrographs of storm events at Minebank Run (Doheny et al. 2006) are typical of urban streams in areas of high impervious surface (Paul and Meyer 2001).
Urban development around Minebank Run predates stormwater management regulations implemented in mid-1980's, and thus, uncontrolled runoff entering the stream was a signi cant water quality problem. Portions of the channel were encased in concrete, thereby increasing the ashiness of storm ows.
Sewer lines and storm drains were eroded and exposed. Riparian buffers were cleared for residential and commercial development.
Before restoration, the study reach was characterized by channel incision of 2-3 m, revealing bedrock in some places and causing lateral movement of the stream that impacted property and sewer infrastructure. Pre-restoration channel width ranged from 0.15-15.5 m, and depth ranged from 0.04-0.9 m, yielding channel cross-sectional areas ranging from 0.007 -13.908 m 2 (Doheny et al. 2006). Mean bank height along the study reach was 0.77 + 0. Throughout the 5.2 km length of Minebank Run, various stream restoration techniques were used based on the condition of the channel and surrounding land use and infrastructure. For example, the stream channel was redesigned to move the thalweg away from an exposed sewer line to protect against further erosion and channel meandering (Figure 2). At two points, the channel was redesigned to reduce bank erosion by creating oxbow wetlands (Harrison et al. 2012a, Harrison et al. 2012bHarrison et al. 2014), which effectively straightened the channel but allowed for greater overbank ow and stormwater retention. Prior to restoration, about 40-70% of the reach was ri es with runs and pools making up the remainder whereas, after restoration, the proportions of ri es, runs, and pools were more equitable at about 30-40% each (Doheny et al. 2012).
The restoration was intended to reconnect the stream channel with the oodplain by mimicking natural valley and oodplain morphology. For example, the project included step-pool and pool-ri e features as well as a stable meander pattern and cross-section. Natural channel design methods (Rosgen 1996, Rosgen 2011) were applied to control ow and erosion by a) raising the stream bed by lling the channel with gravel and cobble, b) removing concrete liners, c) reconstructing point bars, ri es and meander features, d) creating step-pool structure, e) armoring banks, f) creating oxbow wetlands, and g) revegetating the riparian zone (Sortman 2004;Duerksen and Snyder 2005). The restoration corresponded to typologies A, C, and I as described in Newcomer-Johnson et al. (2016).
Although the approach to stream restoration at Minebank Run was primarily intended to address channel erosion and protect sewer infrastructure, we hypothesized that restoration could also affect the hydrology and biogeochemistry of the system (Mayer et al. 2003; Figure 3). Speci cally, we speculated that the physical manipulations designed to accommodate the change in stream discharge rates would also have the potential to change surface and groundwater hydrology (Bukaveckas 2007, Tague et al. 2008) and reduce the hydrologic drought common in urban streams . We speculated that the approach of reshaping the banks and raising the stream bed to eliminate bank incision also might allow carbon-rich riparian soils to become saturated and/or remain wetter, resulting in biogeochemical conditions favorable for nutrient transformations (Kaye et al. 2006). We expected ow control structures installed in the stream channel to reduce erosion also may trap organic matter long enough to create enriched anoxic zones conducive for denitri cation to occur . We also expected revegetating the riparian zone would provide litter inputs and root biomass to supply carbon to denitri ers (Gift et al. 2010 Groundwater was collected from the piezometers using low-ow pumping methods (Puls and Barcelona 1996) with a peristaltic pump through a ow cell and multi-meter instrument (Hach Co., Loveland, CO, USA) Surface water was collected via peristaltic pump for consistency with groundwater sampling. Field measurements for all samples included dissolved oxygen (DO; mg/L), pH, temperature (Cº), oxidation reduction potential (ORP; mV), and speci c conductance (mS/cm). Samples for lab analysis were stored on ice and acidi ed to pH 2 and/or ltered with 0.45-micron lters, depending on analysis.
Piezometers were installed along transects aligned perpendicular to stream ow in groups of 3 (one group in the channel, and one group each on either bank) at 61, 122, and 183 cm below the surface to capture longitudinal and lateral ow (as described in Striz and Mayer 2008;Kaushal et al. 2008a). Transects crossed the stream at geomorphic and restoration features of interest including: cutbanks, gravel bars, terrace, riprap, and oxbows ( Figure 2). A total of 18 piezometers and 2 surface water stations were sampled at control IV and 33 piezometers and 3 surface water stations were sampled at restored CVP. At control IV, piezometers were arranged in 2 transects located 38 m apart across 2 consecutive meander bends. At restored CVP, piezometers were located downstream of USGS stream gage 0158397967 ( Figure 2) and arranged in 3 transects (71 m and 49 m apart). After restoration, some piezometers were replaced at the approximate original pre-restoration locations, where possible, or in comparable locations along the new post-restoration channel. Restoration involved redesigning the channel to ll heavily incised reaches that threatened damage to sewer infrastructure and, in the process, two bends in the channel were cut off to create oxbow wetlands ( Figure 2) that were the focus of previous studies (Harrison et al. 2012a).
Bi-weekly surface water chemistry Independent of the 19 groundwater EPA sampling episodes described above, a second set of surface water samples for NO3¯ and speci c conductance was collected by USGS (USGS National Field Manual; https://www.usgs.gov/mission-areas/water-resources/science/national-eld-manual-collection-waterquality-data-nfm?qt-science_center_objects=0#qt-science_center_objects) approximately every 2 weeks at

Reference sites: eight BES LTER streams with different land uses
We also compared Minebank Run to eight nearby reference streams included in the Baltimore BES LTER network of urban study streams (https://baltimoreecosystemstudy.org/). The BES LTER streams include urban and suburban degraded streams, a stream in an agricultural area, and a forested reference stream.
We used surface water chemistry data collected approximately weekly from 2002-2008 at 9 BES LTER stream study sites ( Figure 4). The Gwynns Falls urban streams at BES LTER were selected because they are near Minebank Run ( Figure 4) and are subject to similar urban in uences on hydrology and ashy ows and managed with stormwater BMPs, stream restoration, and/or sewer repairs. McDonogh, the agricultural stream, is more heavily in uenced by fertilizers than from urban runoff. Baisman Run and Pond Branch are in the Gunpowder Falls watershed, separate from Gwynn's Falls, which is in the Patapsco River watershed. However, both watersheds, and Minebank Run, are in the Piedmont physiographic region. Baisman Run is in an area without sewer infrastructure but with septic systems while Pond Branch is a forested reference stream. All streams were assumed to be potentially in uenced by atmospheric inputs of nitrogen (Lovett et al. 2000, Eshleman et al. 2013).

Laboratory chemical analyses
Chemical analyses followed methodology described in APHA (1998), USEPA (1983), and USGS (USGS National Water Quality Laboratory: https://www.usgs.gov/labs/nwql). Dissolved organic carbon (DOC) was measured directly on a Tekmar-Dohrmann instrument (Teledyne Technologies Inc., Los Angeles, CA, USA) via the UV-persulfate digestion method. Nitrite (NO 2¯) and nitrate (NO 3¯) were measured on un ltered samples using a Lachat Flow Injection Analyzer (Hach Co., Loveland, CO, USA). Because nitrite was negligible in our samples, we refer to combined nitrite and nitrate as nitrate (NO 3¯) . Cl¯ was measured using capillary electrophoresis with indirect UV detection (Waters Corp., Milford, MA, USA). We calculated the ashiness index, a metric to assess the variability in mean daily discharge over a given . Additionally, we applied the same methodology to precipitation data to calculate a "precipitation Flashiness Index," based on mean daily rainfall data.  Figure S1).

Statistical analyses
We used ANOVA analysis to test for differences in chemistry (nitrate, dissolved organic carbon and chloride concentrations) between groundwater and surface water and between pre-restoration and postrestoration periods at restored CVP and control Intervale. Tukey's post-hoc test was performed to compare means during the pre-restoration and post-restoration periods, respectively. We used regression analysis to test for relationships in nitrate and chloride over time. Data were analyzed using Systat 13.0 and SigmaPlot 14.0 software (https://systatsoftware.com).

Results
Precipitation trends and hydrologic response to streamoodplain reconnection

Seasonal groundwater and surface water chemistry
Groundwater vs surface water chemistry Pre-and post-restoration groundwater and surface water NO 3 concentrations did not differ (p ≥ 0.2; Table 1; Figure 9) at either the restored CVP reach or the control IV site (Table 1). In other words, groundwater resembled surface water at both reaches, although surface water NO 3 at control IV (1.12 + 0.06 mg/L) trended weakly higher (p = 0.09) than groundwater at control IV (0.97 + 0.04) after restoration, with mean concentration becoming similar to NO 3 concentrations at the restored CVP reach (1.08 + 0.07 mg/L; Table 1).
Prior to restoration there was no trend at restored CVP in groundwater NO 3 concentration (R 2 = 0.003, p = 0.6; Table 3; Figure 9). However, groundwater NO 3 trended weakly downward over time after the restoration at the restored CVP reach (R 2 = 0.017, p = 0.03; Table 3; Figure 9). No trend was observed in surface water NO 3 concentration either before or after the restoration at the restored CVP reach, suggesting that NO 3 transformations occurred only after restoration in NO 3 stored in groundwater. No trends were observed at the control IV reach, for groundwater or surface water NO 3 during the postrestoration period (p = 0.6; Table 3; Figure 9). However, the signi cant positive trend in surface water NO 3 during the pre-restoration period (p = 0.005; Table 3) may be a function of small sample size.
DOC was always signi cantly higher in surface water than in groundwater both before and after restoration at the control IV and the restored CVP reaches (p < 0.001; Table 1; Figure 9), suggesting that DOC was transported in surface water without being stored in groundwater or because DOC was consumed while in the subsurface. Groundwater and surface water Clconcentrations were similar at the restored CVP reach both before and after restoration (p > 0.13; Table 1; Figure 9). However, at the control IV reach, surface water Clwas double that in the groundwater both before and after restoration (p < 0.001; Table 1), suggesting that local runoff events in the headwaters in uenced surface water salt chemistry. Chloride concentrations at the downstream restored CVP reach ( Figure 9) were chronically elevated compared to the upstream control IV reach due to effects of a major freeway that received heavy deicer salt inputs. However, groundwater was a reservoir for salt loads (Cooper et al. 2014), leading to similar groundwater and surface water Clconcentrations downstream ( Figure 9).
Prior to restoration there was no trend at restored CVP in groundwater DOC concentration (p = 0.5; Table  3; Figure 9). However, groundwater DOC trended upward over time after the restoration at the restored CVP reach (R 2 = 0.022, p = 0.01; Table 3) in opposite trend to groundwater NO 3 concentration (p < 0.03; Table 3). No such trends were evident in surface water DOC at restored CVP (p > 0.3; Table 3). No trend was observed in groundwater DOC at the control IV site during the pre-restoration period (p = 0.9; Table 3; Figure 9), but a weak positive trend was observed in the post-restoration period (R 2 = 0.019, p = 0.07; Table 3). Surface water DOC at control IV showed a decreasing trend during the pre-restoration period (R 2 = 0.586, p = 0.03; Table 3), again perhaps owing to small sample size. No trend in surface water DOC was evident in the post-restoration period at control IV (p = 0.13; Table 3).

Groundwater vs surface water chemistry
Groundwater NO 3 concentration was signi cantly lower after restoration at the restored CVP reach (p = 0.01; Table 2) as was surface water NO 3 concentration at restored CVP (p < 0.001; Table 2; Figure 9).
However, groundwater NO 3 concentration at the control IV reach did not differ after restoration (p = 0.2, Table 2) and surface water NO 3 concentration at control IV was higher on average post-restoration than pre-restoration (p = 0.08; Table 2; Figure 9).
Neither surface water nor groundwater DOC differed after restoration at the restored CVP reach (p > 0.6; Table 2; Figure 9). Likewise, surface water DOC at the control IV reach did not differ after the downstream restoration (p = 0.4; Table 2; Figure 9). However, groundwater DOC was much lower after the restoration at the control IV reach for reasons we are not able to ascertain (p < 0.001; Table 2: Figure 9). This difference did not seem to propagate downstream to the restored reach suggesting that there were more local sources of organic matter that were transported to the stream.
Chloride is chronically higher at the downstream CVP site than the IV site because CVP is downstream of the I-695 Beltway, a signi cant source of road salts (Cooper et al. 2014). After restoration, however, the relationship between groundwater and surface water Clis more closely matched, suggesting stronger groundwater and surface water interaction. The IV site is characterized by ashier surface water Clconcentrations, perhaps due to local headwater sources and/or less riparian buffer, while ground water Clat IV is generally low and less variable over time.
Groundwater Clconcentrations at the restored CVP reach did not differ after the restoration (p = 0.4; Table 2; Figure 9). However, surface water Clconcentrations were lower after restoration (p = 0.02; Table   2; Figure 9), suggesting lower inputs from road salts along the I-695 beltway during this period. Overall, groundwater Clconcentration was lower during the post-restoration period at the control IV reach (p = 0.01; Table 2; Figure 9) owing to high concentrations prior to the restoration, suggesting less storage or increased ushing in the subsurface during the post-restoration time period. Surface water Cltrends did show increasing rate after the restoration (see next section). However, surface water Cl in groundwater did not differ at control IV after the restoration at CVP (p = 0.3; Table 2; Figure 9).
Chloride trends in groundwater at the restored CVP reach shifted from negative (p < 0.001) to positive (p = 0.03) after the restoration, suggesting an increasing rate of road salt application and/or increased storage of salts in the subsurface (Table 3). Surface water Clat restored CVP showed a negative trend before the restoration (p = 0.06; Table 3) but no such trend after the restoration (p = 0.6; Table 3). Only during the post-restoration period was there a positive trend in Clconcentration at the IV reach and only for groundwater (p < 0.001; Table 3). No other Cltrends were observed at control IV (p > 0.3; Table 3).

Relationship between NO 3 and DOC
Groundwater Chemistry response to channel geomorphology NO 3 -, DOC, and Clconcentrations in groundwater differed among stream features at restored CVP reach before and after the restoration (p < 0.05; Table 7). Before restoration, NO 3 was highest in cutbanks (2.6 + 0.3 mg/L; Table 7) followed by concentrations below the stream channel and in terrace features (1.53 + 0.06 mg/L and 1.51 + 0.08 mg/L, respectively; Table 7). DOC was highest in the subsurface of the stream channel (1.18 + 0.08 mg/L; Table 7), likely as a function of transport and groundwater-surface water mixing. Clwas highest in groundwater of gravel bars associated with meander features and below the stream channel (163.5 + 11.3 mg/L and 135.0 + 8.3 mg/L, respectively; Table 7) and lowest in cutbank features (59.8 + 6.3 mg/L; Table 7).
After restoration, oxbows and rip rap structures became new features of the system. NO 3 was highest in oxbow features (2.52 + 0.25 mg/L; Table 7), suggesting higher retention of N at this oodplain reconnection feature (Harrison et al. 2012a). Cutbank features were mostly eliminated after the restoration, however, NO 3 was relatively low for the few samples collected (0.91 + 0.09 mg/L; Table 7). NO 3 in the stream bed and at terrace features designed to connect the oodplain to the stream channel were similar (Table 7), suggesting again that the stream and terraced features were hydrologically connected. DOC remained highest (1.23 + 0.06 mg/L; Table 7) below the stream channel perhaps owing to higher transport of incoming organic matter mixing in the subsurface.
Chloride was highest at gravel bars and in the stream channel (212.3 + 13.9 and 128.5 + 3.3 mg/L, respectively; Table 7), suggesting enhanced Clstorage in these features. Chloride concentration was similar among stream channel, cutbank, oxbows, riprap, and terrace features (Table 7).
Restoration effects on biweekly surface water chemical concentrations and uxes Surface water NO 3 concentration in the intensive surface water USGS surveys showed increasing trends prior to restoration and during construction at restored CVP (p < 0.008; Table 4; Figure 11). However, after restoration, NO 3 trends declined steadily (p < 0.001; Table 4; Figure 11). Seasonal cycles were evident, with higher NO 3 observed in winter with maximum concentrations > 4 mg/L before restoration and > 2 mg/L even after restoration (Table 4)  . Like Cl -, speci c conductance was chronically higher at the downstream restored CVP site than upstream at control IV ( Figure 12). Speci c conductance was relatively variable, exhibiting peaks and outliers ( Figure 12). Speci c conductance increased over time at restored CVP and control IV prior to restoration (p < 0.05; Table 4). While post-restoration speci c conductance trends were not signi cant at either restored CVP or control IV (p > 0.11; Table 4), the overall trends appear to be increasing, suggesting that Minebank Run water chemistry was in uenced by ion inputs from road salts and that stormwater runoff is a factor dictating stream chemistry throughout this system ( Figure 12).

Surface water NO 3 trends at BES LTER reference streams
The BES LTER streams represent reference sites to our study sites at Minebank Run because they are urban streams in similarly degraded watersheds that have been variously impacted by urbanization but managed for water quality improvements via stream restoration, stormwater management, and sewer repair. Like Minebank Run, declining trends in NO 3 were observed at two BES LTER streams, Gwynns Falls at Carroll Park (urban; p = 0.01) and at Glyndon (suburban headwaters; p < 0.001; Table 5; Figure  18). Nitrate increased at the Gwynns Falls Gwynnbrook Ave site (suburban; p = 0.04; Table 5, Figure 18). There was no trend at Baisman Run (p = 0.07; Table 5; Figure 18) which is in a suburb with septic instead of piped sewer infrastructure. There was no NO 3 trend at forested reference site Pond Branch (p = 0.7) which should be affected by atmospheric and natural soil inputs alone. There was a strong increasing trend in NO 3 at McDonogh (p < 0.001; Table 5; Figure 18), the agricultural site most affected by manure fertilizers inputs and where there were no signi cant stream restoration efforts.
Like Minebank Run, increasing trends in Clwere observed among some of the 8 BES LTER streams examined (  Figure 19). Dead Run and Glyndon showed no trends (p > 0.2; Analysis #2; Table 6; Figure 19). Increases at the urban and suburban sites suggest increasing runoff from road salts. However, increasing Clat Pond Branch (p < 0.001; Table 6; Figure 19) was unexpected because we assumed this reference stream would not be affected by saline runoff. The increasing Cltrend at McDonogh, the agricultural site, may have been affected by fertilizers but unlikely from road runoff because of low road density (p < 0.001; Table 6; Figure 19). Surface water trends among BES LTER reference streams show similar trends in N over the same time period compared to Minebank Run, suggesting that stream restoration, stormwater BMPs and/or sewer repairs have also reduced N in certain watersheds over longer time scales. Declining trends in bioreactive NO 3 corresponding to increasing trends in conservative Cland/or speci c conductivity suggest that the decreases in NO 3 are a function of biological uptake or denitri cation (Mayer et al. 2010b). Below, we discuss the impacts of stream restoration on N and potential factors in uencing N retention and transport pre-and post-restoration.

Stream-oodplain reconnection (and other management activities) in uences long-term N transport and retention
Our study demonstrated that oodplain reconnection was an effective restoration approach for reducing N concentration and ux in an urban stream. Based on the BACI design, both groundwater and surface water NO 3 decreased at restored CVP after the restoration despite no change in groundwater NO 3 concentration at the control site and an increase in NO 3 in surface water upstream at IV. groundwater-surface water interaction and the initial reconnection of the oodplain to the channel.
However, with numerous structural failures appearing along the restored reach (see Figure S5a and b), long-term e cacy is in question.

Evidence that altered stream morphology enhances N transformation by increasing retention times in stream features and oodplains
The similarity of groundwater and surface water NO 3 and Clat both the restored CVP reach and control IV reach suggests mixing of groundwater and surface water. NO 3 and Clpatterns among stream restoration features varied before and after restoration suggesting that some features were more retentive of N. The oxbows created by the restoration had higher NO 3 concentrations perhaps because they were designed to retain stormwater runoff (Harrison et al. 2014). However, these features promoted high rates of denitri cation (Harrison et al. 2011), demonstrating that such wetlands have the potential to reduce Increased hydrologic connectivity at the groundwater-surface water interface can provide DO, N, and organic matter to microbes in subsurface sediments, and may foster "hot spots" for nitrogen removal via denitri cation due to low redox conditions (Hedin et al. 1998 Previous research at Minebank Run showed that denitri cation enzyme activity (DEA) and microbial biomass C were both higher in hyporheic sediments (in and near stream piezometers) than in deep oodplain sediments suggesting that the hyporheic zone is responding to and processing C and NO 3 from upstream and/or riparian sources (Mayer et al. 2010b). These results also suggest that restoration that increases C ow to these sediments could increase denitri cation capacity of the stream ecosystem.
At Minebank Run and other streams in the Baltimore area, denitri cation potential was highest in organic debris dams and other features high in organic matter . None of the stream features in Minebank Run were regions of high organic matter accumulation probably because ashy stream ows frequently wash debris from the channel and downstream. Also, pools in both the restored and unrestored reaches of Minebank Run had lower denitri cation enzyme activity than pools from other streams in the Baltimore area ). Strong positive relationships at Minebank Run between root biomass and soil organic matter, and between soil organic matter and denitri cation potential, suggests that deep rooted vegetation may be particularly important for maintaining an active denitri cation zone in restored riparian zones (Gift et al. 2010). However, restoration that improves hydrologic connectivity in the hyporheic zone and oodplain so that organic matter reaches subsurface zones where there is low DO and adequate NO 3 for anaerobic activity is key to enhancing denitri cation.
Reductions in NO 3 ux and ow normalized concentrations post-restoration Despite geomorphic failure of restoration features, the decline of ow normalized NO 3 concentrations over time indicate that the stream was still able to continue removing NO 3 through biogeochemical processing. Also, the lower slope of the NO 3 concentration-to-discharge relationship during postrestoration compared to pre-restoration, indicates that, even at higher ows, NO 3 concentrations were lower after the restoration, which con rms that the restoration improved biogeochemical NO 3 uptake activity across a range of ows.

Conclusion
Our studies support the idea that in-stream processes and hydrologic connectivity between the stream channel and subsurface zones may in uence N processing in urban streams (Craig et al. 2008, Kaushal et al. 2008a). Restoration activities focused on increasing hydrologic connectivity in riparian zones may enhance denitri cation rates by increasing soil organic carbon availability and altering hydrologic owpaths (e.g. Fennessy and Cronk 1997;Boulton 2007;Mayer et al. 2007). Because riparian soils, geomorphology, hydrologic owpaths, and geology all play roles in explaining variations in denitri cation rates (e.g. Alexander et al. 2000, Stanley and Doyle 2002, Groffman and Crawford 2003, Gücker and Boechat 2004, Wollheim et al. 2005, all of these factors should be considered and further evaluated in the e cacy of restoration designs aimed at increasing both denitri cation rates and mass removal of nitrate-N in riparian zones.
Restoration practices that improve connectivity between the stream and the riparian zone can increase NO 3 -removal. Yet, more work is necessary to better quantify the effectiveness of stream restoration practices under various applications and conditions and over time ( Employing this type of restoration solely for managing N is prohibitively expensive and likely will not address the impacts of future N loads and sources. While stream restoration is not the primary solution to N management in Chesapeake Bay, stream restoration has numerous cumulative, potential bene ts that may justify the costs of such efforts, including, sediment and erosion control, protection of property, increased property values, sh and wildlife habitat and migration corridors, green space, stream temperature control, improved ecosystem metabolism and maintenance of riparian zones. Longevity or e cacy of restoration projects, especially under repeated storm effects or increasing urbanization, or from the effects of salinization and chemical cocktails (Cooper et al. 2014, Kaushal et al. 2020), is currently unknown. Storms after the restoration may have exceeded the channel design discharge for which the restoration at Minebank Run was engineered and/or the bankfull dimensions may have been too di cult to accurately identify (Sortman 2004). Soil instability and poor vegetation reestablishment also may lead to erosion during overbank ows (Sortman 2004). Natural channel design (NCD) restoration approaches integrate uvial processes of "self-formed and self-maintained natural rivers" (Rosgen 2011 Restoration designs are heterogeneous efforts consisting of various components including bank reshaping, bank stabilization, channel reconstruction, riparian re-vegetation, etc. Therefore, study designs that distinguish the in uence of individual restoration components will help to identify those techniques that contribute most to nutrient uptake and other ecosystem functions of concern. Also, because in situ N uptake and transformation is notoriously di cult to measure , studies that measure denitri cation or surrogates of denitri cation (e.g. oxidation-reduction potential, DO) at watershed scales and over time will be most useful in quantifying restoration effects. Finally, long-term monitoring will better elucidate short and long-term patterns of nutrient dynamics (Goodale et  mg/L) excluded from NO 3 analyses. *One outlier (Cl -> 400 mg/L) excluded from Clanalyses.  ) concentrations (mg/L) during pre-versus post-restoration periods at restored CVP and control Intervale sites in groundwater and surface water. Table shows sample   ), dissolved organic carbon (DOC), and chloride (Cl -) concentration (mg/L) linear regression trends over time (2003)(2004)(2005)(2006)(2007)(2008) in groundwater and surface water at Minebank Run*. Statistical comparisons (R 2 , regression coe cient, p-value) represent a comparison of pre-vs post restoration at each site for groundwater and surface water, respectively. *Six outliers (NO 3 ->6 mg/L) excluded from NO 3 analyses. *One outlier (Cl -> 400 mg/L) excluded from Clanalyses. ) concentrations (mg/L) and speci c conductance (µS/cm) over time (2003)(2004)(2005)(2006)(2007)(2008) in surface water collected during bi-weekly surveys at Minebank Run. *One outlier (speci c conductance > 5000 µS/cm) excluded from analyses. bes.700.600; le: BES-stream-chemistry-data-for-WWW-feb-2018---core-sites-only.cvs Table 7. Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Aerial photos of a portion of the downstream study reach of Minebank Run showing the reconstructed stream before and after the restoration (aerial photos by K. Jewell). Note the bend in the stream channel (future oxbow) that became the oxbow after the restoration.

Figure 3
Conceptual gure of stream restoration to improve oodplain reconnection. Eliminating incision and reconnecting the oodplain unites saturated soils with organic matter from plants and roots and allows   USGS bi-weekly Minebank Run surface water speci c conductance (µS/cm) at restored CVP and control Intervale. Surface water in samples collected bi-weekly at restored CVP reach. Speci c conductance is generally higher at restored CVP than at the upstream control Intervale because of the in uence of the I-695 beltway and associated inputs of road salts. Trends overall appear to be increasing with occasional extreme peaks from storm runoff suggesting that road salts and/or other ions are in uencing water chemistry at Minebank Run.