Effects of urban stream burial on organic matter dynamics and reach scale nitrate retention

Nitrogen (N) retention in streams is an important ecosystem service that may be affected by the widespread burial of streams in stormwater pipes in urban watersheds. We predicted that stream burial suppresses the capacity of streams to retain nitrate (NO3−) by eliminating primary production, reducing respiration rates and organic matter availability, and increasing specific discharge. We tested these predictions by measuring whole-stream NO3− removal rates using 15NO3− isotope tracer releases in paired buried and open reaches in three streams in Cincinnati, Ohio (USA) during four seasons. Nitrate uptake lengths were 29 times greater in buried than open reaches, indicating that buried reaches were less effective at retaining NO3− than open reaches. Burial suppressed NO3− retention through a combination of hydrological and biological processes. The channel shape of two of the buried reaches increased specific discharge which enhanced NO3− transport from the channel, highlighting the relationship between urban infrastructure and ecosystem function. Uptake lengths in the buried reaches were further lengthened by low stream biological NO3− demand, as indicated by NO3− uptake velocities 17-fold lower than that of the open reaches. We also observed differences in the periphyton enzyme activity between reaches, indicating that the effects of burial cascade from the microbial to the ecosystem scale. Our results suggest that stream restoration practices involving “daylighting” buried streams have the potential to increase N retention. Further work is needed to elucidate the impacts of stream burial on ecosystem functions at the larger stream network scale.

Abstract Nitrogen (N) retention in streams is an important ecosystem service that may be affected by the widespread burial of streams in stormwater pipes in urban watersheds. We predicted that stream burial suppresses the capacity of streams to retain nitrate (NO 3 -) by eliminating primary production, reducing respiration rates and organic matter availability, and increasing specific discharge. We tested these predictions by measuring whole-stream NO 3 removal rates using 15 NO 3 isotope tracer releases in paired buried and open reaches in three streams in Cincinnati, Ohio (USA) during four seasons. Nitrate uptake lengths were 29 times greater in buried than open reaches, indicating that buried reaches were less effective at retaining NO 3 than open reaches. Burial suppressed NO 3 retention through a combination of hydrological and biological processes. The channel shape of two of the buried reaches increased specific discharge which enhanced NO 3 transport from the channel, highlighting the relationship between urban infrastructure and ecosystem function. Uptake lengths in the buried reaches were further lengthened by low stream biological NO 3 demand, as indicated by NO 3 uptake velocities 17-fold lower than that of the open reaches. We also observed differences in the periphyton enzyme activity between reaches, indicating that the effects of burial cascade from the microbial to the ecosystem scale. Our results suggest that stream Electronic supplementary material The online version of this article (doi:10.1007/s10533-014-9971-4) contains supplementary material, which is available to authorized users.

Introduction
Streams draining urbanized basins exhibit a suite of ecological degradations including flashy hydrology, elevated concentrations of nutrients and contaminants, and reduced biotic richness (Walsh et al. 2005). These degradations largely result from increased impervious surfaces in the basin and have been the focus of most urban stream ecology research (Wenger et al. 2009). Despite the focus on impervious surfaces, perhaps the most extreme expression of urbanization on stream ecosystems is their burial in drainage pipes to facilitate above-ground construction or stormwater management (Kaushal and Belt 2012). While stream burial has occurred since the Roman empire (Hopkins 2007), ecologists have only recognized stream burial as a common consequence of urbanization in recent decades (Leopold 1968), and only in the last few years has the extent of stream burial been quantified (Elmore and Kaushal 2008). Stream channels are frequently routed into pipes and buried during the process of urbanization. Over time, the total length of buried stream channels can far exceed that of open stream channels. For example, it is estimated that over half the streams in Baltimore, Maryland (USA) are buried with the probability of stream burial reaching 98 % in the most densely populated areas of the city (Elmore and Kaushal 2008). Similarly, O'Driscoll et al. (2010) reported that stream density in urban catchments was approximately 40 % lower than surrounding rural catchments due to urban stream burial. Although similar data are lacking from other cities, stream burial is likely ubiquitous throughout large urban centers (Conradin and Buchli 2005;Nam-choon 2005;Wild et al. 2011).
Stream burial fundamentally changes the structure and function of stream ecosystems through several mechanisms including the inhibition of photosynthesis, reduced organic matter inputs, and altered hydrology. Despite the prevalence and obvious ecosystem impact of stream burial, little research has been conducted on the topic, and the existing literature focuses on the effects of short road crossings on macroinvertebrates and fish. For example, Meyer et al. (2005) found that stream burial decreased macroinvertebrate and taxonomic richness in a buried stream in Georgia (USA). Others have reported that stream burial can hinder the in-stream migration of macroinvertebrates and fish (Dedecker et al. 2006;Foster and Keller 2011;Poplar-Jeffers et al. 2009) as well as the longitudinal dispersal of emerging adult insects (Blakely et al. 2006). To our knowledge, however, there are no published reports of the effect of stream burial on other important ecosystem services provided by streams, such as nutrient removal.
Streams draining developed catchments carry elevated nitrogen (N) loads (Howarth et al. 1996;Kaushal et al. 2008a), often in the form of nitrate (NO 3 -), which stimulates algal blooms in N-limited aquatic ecosystems, leading to a cascade of biogeochemical processes that ultimately results in oxygen-depleted waters and eutrophication. Estuarine and coastal zones have been particularly affected by eutrophication where dissolved oxygen routinely falls below levels required to support sensitive aquatic organisms at more than 400 sites across the globe (Díaz and Rosenberg 2008). Management strategies to enhance the capacity of streams and rivers to remove NO 3 from the water column via biological activity may reduce N delivery to coastal waters (Bukaveckas 2007;Kaushal et al. 2008b). Heterotrophic and autotrophic organisms in streams can assimilate water column NO 3 into biomass, which may be subject to long term storage in stream beds and floodplains (O'Brien et al. 2012). Some heterotrophic organisms can permanently remove NO 3 through denitrification, a dissimilatory metabolic process where NO 3 is converted to dinitrogen (N 2 ) or nitrous oxide (N 2 O) gas. We speculated that stream burial may severely reduce the efficacy of these NO 3 removal mechanisms by eliminating N assimilation associated with photosynthesis, reducing the quantity and quality of organic matter needed to support denitrification and heterotrophic assimilation, and by enhancing hydrologic export from the channel, thus augmenting downstream NO 3 transport. Stream burial may therefore exacerbate elevated N loading from urban basins to coastal waters.
The biological demand for NO 3 in temperate zone streams often follows a predictable seasonal cycle (Lutz et al. 2012). High light levels and rising water temperatures during the spring can lead to benthic algal blooms and high assimilative NO 3 demand, which decreases during the summer when the forest canopy shades streams and alga populations senesce. A secondary peak in NO 3 demand often occurs during the fall when allochthonous organic matter inputs stimulate heterotrophic metabolism whereas low water temperature and short days tend to suppress winter-time NO 3 demand. By contrast, light and organic matter availability in buried streams is low throughout the year, which likely dampens seasonal patterns in biological NO 3 demand. As a consequence, the effect of burial on biological NO 3 demand may be most pronounced during the spring and fall when the difference in light and organic matter between open and buried streams is most extreme.
Understanding the degree to which ecosystem N retention is impaired is a first step towards setting targets for watershed restoration objectives. In order to determine the effect of burial on NO 3 removal in streams, we used whole stream 15 NO 3 tracer additions to measure NO 3 uptake in a paired comparison of open and buried reaches in urban streams located in Cincinnati, OH (USA) and Baltimore, MD (USA). Here we report the results of the work in Cincinnati whereas Pennino et al. (2014) report the results from Baltimore in this special issue of Biogeochemistry. We measured NO 3 uptake seasonally (e.g. summer, fall, winter, spring) over the course of 1 year. To identify the mechanisms controlling NO 3 uptake, we measured a broad suite of functional and structural stream characteristics. At the patch scale (i.e. 225 cm 2 ) we estimated bacterial abundance and assessed dissolved organic carbon (DOC) quality using measurements of periphyton extracellular enzyme activity (EEA) rates (Sinsabaugh and Shah 2011). At the ecosystem scale (stream reach) we measured metabolism and hydraulic characteristics. Water chemistry was measured at the top and bottom of the reaches and organic matter standing stocks were estimated by scaling habitat weighted means to the reach.

Study sites and experimental design
We selected three streams draining urban watersheds within and near the city of Cincinnati, Ohio (USA) (Figs. 1, 2; Table 1). Each stream contained buried and open reaches separated by 30-100 m of open stream channel. The buried reach at Eastgate Creek (EAS) was 740 m long, constructed from interconnected segments of ovoid corrugated metal pipe ranging from 1.5 to 1.8 m in height, and approximately 0.5 m wide at the base. The pipe was relatively clean of debris and sediment. The open reach at EAS had an average wetted width of 2.1 m, was downstream of the buried reach, deeply incised, and lined by a vegetated riparian zone. Sediment in the open reach at EAS was predominantly cobbles, boulders, and exposed bedrock, though restricted portions of the stream bed were composed of pebbles and sand. The buried reach at Amberly (AMB) was 430 m long, 2.5 m in height, 4.5 m wide at the base, constructed from segments of corrugated metal pipe, and filled with sand and pebbles to a depth of ca. 0.15 m. The open reach at AMB had an average wetted width of 3.9 m, was downstream of the buried reach, flowed between parking lots, and the width of the vegetated riparian zone was less than 5 m in many places. The buried reach at Este (EST) was downstream of the open reach, 290 m long, 2.5 m in height, 1.0 m wide at the base, constructed from concrete, and relatively free of sediment. The open reach at EST flowed through a vegetated riparian zone and had a deeply entrenched channel with a wetted width of 2.1 m composed of highly mobile sandy sediments.
We measured NO 3 uptake rates in the buried and open reaches of all three streams during summer and fall 2011, and winter and spring 2012. This experimental design provided four paired measurements of NO 3 uptake in buried and open reaches of each stream. Nitrate uptake rate measurements were made during the daylight hours. Water chemistry samples were collected from the top and bottom of each experimental reach immediately prior to the 15 N-NO 3 tracer release and again immediately before the release was terminated. Stream hydrologic parameters, photosynthetically active radiation (PAR), and ecosystem metabolism were measured concurrently with the 15 N-NO 3 tracer release. Samples for benthic organic matter and chlorophyll standing stocks, algal abundance, ecoenzyme activity, and DNA extraction were collected within 48 h of the 15 N-NO 3 tracer release.
Hydrology, metabolism, light, and water chemistry Average stream width was calculated from a minimum of 30 measurements of wetted width distributed evenly along the study reaches. Water velocity was calculated as time to maximum slope of the rhodamine breakthrough curve recorded at the downstream end of each experimental reach using a data sonde and rhodamine sensor (YSI, Yellow Springs, OH, USA). Discharge (Q) was calculated as: We estimated reach scale rates of gross primary production (GPP) and ecosystem respiration (ER) using temperature and dissolved oxygen measurements made with data sondes (YSI, Yellow Springs, OH, USA) at 5 min intervals for a minimum of 36 h at the top and bottom of each experimental reach. Rates were calculated using the two-station method following Marzolf et al. (1994) with the corrections discussed in Young and Huryn (1998), and the onestation calculations following Roberts et al. (2007). The two-station method was used for the open reach whenever possible (i.e. data from up and downstream sondes were available). If the downstream end of the buried reach exhibited no diel dissolved oxygen trend, then dissolved oxygen dynamics at that site were considered to be decoupled from the Fig. 1 The three study streams located near Cincinnati, Ohio (USA). The shaded areas represent the drainage area relative to the most downstream sampling location upstream open reach and the one-station method was used. We measured air-water gas exchange rates using sulfur hexafluoride (SF 6 ) tracer injections conducted concurrently with the 15 NO 3 tracer injections (Marzolf et al. 1994). Photosynthetically active radiation (PAR) was measured 1 m above the water surface in the middle of each reach at 5 min intervals using a cosine corrected photosynthetic irradiance sensor (Odyssey, New Zealand) (Shaffer and Beaulieu 2012).
All water samples were filtered (0.45 lm) in the field, stored on ice during transport to the laboratory, and either acidifed or frozen, depending on the analyte, prior to analysis. Nitrate ? nitrite, hereafter referred to as NO 3 -, dissolved reactive phosphorus (DRP), bromide (Br -), and ammonium (NH 4 ? ) were measured using standard colorimetric methods (APHA 2005) and flow injection analysis (Lachat Instruments, Loveland, CO USA). Dissolved organic carbon (DOC) was measured using a total organic carbon (C) analyzer with high-temperature Pt-catalyzed combustion and NDIR detection (Shimadzu TOC-VCPH, Columbia, MD, USA). See below for 15 N-NO 3 processing details. Fig. 2 Pictures of buried streams included in this study. Este (EST) (a, b) flowed through a concrete tunnel with a narrow wetted width at baseflow. Amberly (AMB) (c) flowed through a wide corrugated metal pipe which was filled with sediment to a depth of ca 15 cm. Eastgate (EAS) (d) flowed through a corrugated metal pipe that was relatively free of sediment and debris Benthic organic matter We measured coarse ([1 mm size fraction), fine (\1 mm size fraction), and attached (e.g. periphyton) benthic organic matter standing stocks by collecting material from a known area of the stream bottom at 10-20 locations randomly distributed among habitat units (e.g. pool, riffle, run) according to their relative abundance in a stratified-random sampling design. An open-ended plastic cylinder (0.052 m 2 cross sectional area) was placed up to a maximum depth of 5 cm in the sediment and coarse benthic organic matter (CBOM) was removed by hand. A hand trowel was then used to agitate the sediments, and a subsample (*200 mL) of the resultant slurry was collected and poured over a 1 mm mesh sieve. The material retained on the mesh was composited with the coarse material and the filtrate was collected for fine benthic organic matter (FBOM). Periphyton was collected by scraping a known area of rock surface area (0.006-0.04 m 2 ) with a toothbrush or stiff-bristled brush operated by a portable drill. The dislodged periphyton was washed into a sample bottle with stream water. Periphyton standing stocks were assumed to be zero on large patches of sand. Dry mass and ash-free dry mass, determined from mass loss after ignition at 500°C, were measured on all samples. Reach scale organic matter standing stocks were calculated as the sum of the mean value for each habitat unit scaled to the reach. Chlorophyll a was measured on subsamples of the periphyton samples using the trichromatic method (APHA 2005) following extraction with hot ethanol (Sartory and Grobbelaar 1984).
To assess the effect of stream burial on periphyton, while controlling for differences in substrate (e.g. metal pipe vs. natural cobble) and colonization time, we deployed 0.15 m 9 0.15 m unglazed clay tiles in each reach and allowed a minimum of 6 weeks for colonization by periphyton. Within 48 h of each tracer release, the tiles were collected and the attached periphyton was dislodged with a toothbrush and razor blade and rinsed into a collection bottle with site water and stored on ice during transport to the laboratory. Three samples were analyzed for algal abundance using a Palmer-Maloney counting cell (Charles et al. 2002), four to ten samples were analyzed for total bacterial counts using quantitative PCR (see below), and five to ten samples were analyzed for EEA (see below).

DNA extraction and Quantitative PCR (qPCR) assays
We extracted DNA from four to ten tiles per reach within 6 h of sample collection. For each sample, biomass was concentrated from a 6 mL subsample via microcentrifugation (3,0009g for 15 min). Biofilm pellets were resuspended in 500 lL of lysate buffer and transferred to tubes prior to the bead-beating step. DNA extractions were performed using Mo Bio Power Lyzer/Power Soil kits according to the manufacturer's protocol (MO BIO Laboratories, Carlsbad, CA, USA). DNA extracts were stored at -20°C until further processing.
Total bacterial densities were estimated using a eubacterial universal 16S rRNA gene qPCR assay developed by Suzuki et al. (2000) with the BACT1369F (5 0 -CGGTGAATACGTTCYCGG-3 0 ) and PROK1492R (5 0 -GGWTACCTTGTTACGACTT-3 0 ) primers and TM1389F hydrolysis probe (5 0 -6FAM-CTTGTACA CACCGCCCGTC-TAMRA-3 0 ). The qPCR assay was performed in 25 lL reaction volumes containing 19 TaqMan universal PCR master mix with AmpErase uracil-N-glycosylase (Applied Biosystems, Foster City, CA, USA), 0.2 lg lL -1 bovine serum albumin, the corresponding primers at 1 lM and the TaqMan reporter probe at 0.5 lM. Ten-and 50-fold dilutions of each DNA extract were used to test for PCR inhibition (Ryu et al. 2011;Toledo-Hernandez et al. 2013). The amplification protocol involved an initial incubation at 50°C for 2 min to activate uracil-N-glycosylase, a 10 min incubation at 95°C, followed by 40 cycles of 95°C for 15 s and 56°C for 1 min, and a 10 min incubation at 72°C. The qPCR assay was performed using a 7900 HT Fast Real-Time Sequence Detector (Applied Biosystems, Foster City, CA, USA) in duplicate MicroAmp Optical 96-well reaction plates with MicroAmp Optical Caps (Applied Biosystems, Foster City, CA, USA). PCR data were analyzed using the instrument's Sequence Detector software (version 2.2.2, Applied Biosystems). qPCR standard curves were generated in duplicate for each 96 well plate by plotting threshold cycle (CT) values against the number of target copies corresponding to serially diluted plasmid standards (Integrated DNA Technologies, Coralville, IA, USA). The target copy numbers (T) were estimated by the following equation where D (g lL -1 ) is plasmid DNA concentration, and PL (bp) is plasmid length in base pairs. The standard curve was generated from five 10-fold plasmid dilutions. Two no-template controls per PCR plate were used to check for cross-contamination.

Extracellular enzymes activities (EEA)
Periphyton collected from five and ten tiles deployed in the open and buried reaches, respectively, was analyzed for extracellular enzyme activities (EEA). Extracellular enzymes are produced by microbial assemblages to aid in the degradation of organic matter and the acquisition of limiting nutrients and can serve as an index of resource availability (Sinsabaugh and Foreman 2001). Labile C acquisition was measured as the activity of b-D-glucosidase. Two oxidases (polyphenol oxidase and peroxidase) were analyzed to assess recalcitrant C acquisition and the ratio of recalcitrant C acquisition (phenol oxidase) to total C acquisition (b-D-glucosidase ? phenol oxidase) represents an index of the quality of the dissolved organic C pool (LCI; Sinsabaugh and Shah 2011). All EEA assays used the microplate protocols originally developed by Sinsabaugh and colleagues (Sinsabaugh et al. 1997;Sinsabaugh and Foreman 2001) and modified according to Hill et al. (2010). Each microplate array included quadruplicate assays for each enzyme and each reference standard. All substrate and reference solutions were prepared in sterile deionized water. Quenching, the decrease of fluorescent emissions caused by the interactions of enzyme substrates with non-reactant chemicals in the assays, was estimated by comparing the fluorescence of the supernatant of standard solutions mixed with sample to that of the standard solution mixed with buffer. Substrate and sample controls (each mixed with buffer) were assayed in quadruplicate on the same microplate. Fluorescence was measured using a fluorometer (Model FLX800T, BioTek Instruments, Winooski, VT, USA) with an excitation wavelength of 350 nm and an emission wavelength of 450 nm.

NO 3 tracer injections, sampling and analysis
We measured NO 3 uptake rates in each reach by conducting whole stream 15 NO 3 tracer additions. We used a reciprocating piston pump (Fluid Metering Inc, Syosset, NY, USA) to meter a solution of 99.99 % 15 N potassium nitrate (K 15 NO 3 -), sodium bromide (NaBr, conservative tracer), and rhodamine (conservative tracer) into each reach. The target tracer enrichments in the stream water were ?5,000 % 15 NO 3 -, ?500 lg L -1 Br -, and ?15 lg L -1 rhodamine. The tracer additions were maintained until the tracer attained steady state (i.e. plateau) for 2-4 h as determined from in situ rhodamine concentration measurements made at the downstream end of each experimental reach.
We collected filtered water samples (0.45 lm pore size) for 15 NO 3 -, NO 3 concentration, and Brat the upstream and downstream ends of the reach before (background samples) and immediately prior to terminating the tracer releases (plateau samples). Background and plateau samples were collected in triplicate and quintuplicate, respectively. Rhodamine was measured continuously at the upstream and downstream ends of the reaches using a data sonde (YSI, Yellow Springs, OH, USA). We located the upstream sampling station at the furthest upstream point where the tracers were well mixed across the stream width. We assumed the site was well mixed if the rhodamine concentration differed\5 % across the width of the stream channel during the tracer release. The downstream sampling site was located at the end of the pipe for the buried reaches, resulting in a 20-209 min water travel time (mean = 81 min), and for the open reaches the downstream station was located a distance equivalent to 60 min of water travel time from the upstream station.
Filtered 15 NO 3 samples were stored on ice during transport to the laboratory and frozen at -20°C until analysis. Tracer 15 NO 3 was measured at Colorado Plateau Laboratory (NAU, Flagstaff, AZ, USA) using the denitrifier method (Coplen et al. 2012;Sigman et al. 2001). Briefly, the method utilizes P. aureofaciens, a denitrifier lacking the enzyme to reduce N 2 O to N 2 , to convert NO 3 and nitrite (NO 2 -) to N 2 O which is subsequently analyzed for N isotopic composition using an isotope ratio mass spectrometer (Thermo-Finnigan Delta Plus We calculated NO 3 uptake rates at the whole stream scale using the nutrient spiraling framework (Newbold et al. 1981;Stream Solute Workshop 1990). Nitrate uptake was determined from the downstream decrease in 15 NO 3 flux according to the first order model: where 15 NO 3 is the tracer 15 NO 3 flux at the upstream and downstream ends of the reach, k is the first order rate constant for NO 3 uptake, and (upstream -downstream) is the length of the measurement reach (m). We calculated NO 3 uptake length (S w ), the average distance traveled by a NO 3 ion prior to removal from the water column, as the inverse of k. We calculated NO 3 uptake velocity (m f , mm min -1 ), a metric which accounts for the influence of water depth and velocity on S w , as: where Q is discharge and w is stream width. For reaches where NO 3 uptake was below detection (i.e. the 15 NO 3 tracer flux did not decrease across the length of the experimental reach) we calculated the maximum possible k value the stream could have supported without resulting in a measurable decrease in tracer 15 NO 3 flux (e.g. method detection limit). The method detection limit was calculated using simulation modeling based on the observed variability in the replicate NO 3 concentration, 15 NO 3 -, and Brmeasurements. Following the approach of Hanafi et al. (2007), we defined a distribution of possible values for each analyte during the background and plateau sampling using the mean and standard deviations of the 3-5 replicate measurements. We then randomly picked values from the defined distributions for each analyte and calculated k as described above. This process was repeated 10,000 times and the mean and 95 % confidence interval was calculated from these 10,000 k estimates. We then decreased the mean 15 NO 3 tracer flux value for the downstream station during the plateau by increments of 0.1 % until the 95 % confidence interval no longer included 0. We defined the minimum detection limit as the smallest k value with a 95 % confidence interval that did not include 0.

Statistical analysis
We used paired t-tests to test for differences in NO 3 uptake rates (e.g. S w and v f ) and ER between open and buried reaches. Paired t-tests block for spatial (e.g. stream to stream) and temporal (e.g. season) variability that may otherwise obscure the effect of reach. We substituted the minimum detection limit for the true NO 3 uptake rate in the 3 buried and 1 open reach where the 15 NO 3 tracer flux did not decline across the reach. The true uptake rate lies somewhere between 0 and the minimum detection limit, therefore this approach will overestimate the uptake rates. Since three of the four reaches where the NO 3 uptake was below detection were in buried reaches, this statistical analysis will underestimate the effect of stream burial on NO 3 uptake. Stream burial could influence NO 3 uptake via hydrologic and/or biologic mechanisms. If the effect is purely hydrologic, S w should scale linearly with specific discharge (Q s ), defined as the ratio of discharge and width Stream Solute Workshop 1990). Any deviation from this relationship indicates that burial has affected the biological demand for NO 3 -. To assess the relative importance of these two mechanisms we plotted the ratio of S w in the buried and open reaches of each stream versus the ratio of Q s in the buried and open reaches of each stream. Deviation from the 1:1 line indicates that burial has affected the biological demand for NO 3 -. We calculated the ratio of NO 3 v f in the open and buried reaches of each stream and season as an indicator of the magnitude of the burial effect on biological NO 3 demand. To determine the factors controlling the burial effect, we used simple linear regression to relate this metric to the length of the buried reach, water travel time through the buried reach, and the ratio of ER in open to buried reaches.
We used general linear models to test for an effect of reach (e.g. buried or open), season, and a reach X season interaction on fine, coarse, and attached benthic organic matter standing stocks. The 10-20 replicate samples of each organic matter type collected from each reach-season combination were used in the model. The reach scale organic matter standing stock values (i.e. one value per reach) were used to explain variation in ecosystem process rates (see below).
When ecosystem process rates (e.g. NO 3 uptake or ER) were found to differ between open and buried reaches, we used general linear models to determine which variables were responsible for the differences. We used a backward selection procedure starting with a model containing all candidate independent variables and their interaction with reach (e.g. buried or open). A significant interaction would indicate that a specific variable is controlling the process, but is doing so differently in the buried and open reaches. We first removed all non-significant two-way interactions from the model. If no interactions were retained, we also removed reach as a main effect because it yielded no additional insight into the factors responsible for the stream burial effect. We repeated this procedure using data from only the buried reaches to determine which factors control variation in uptake rates among buried streams. We used one-way analysis of variance, with season as the main effect, to determine if the temporal component of the experimental design (i.e. seasonal sampling) generated variability in NO 3 cycling rates and other stream characteristics.
When necessary, we applied a natural log transformation to the dependent variable to ensure model residuals conformed to the assumptions of normality and homogeneity. The data set was composed of 24 observations of 32 independent variables, therefore not all independent variables could be included as candidates in the statistical models without overfitting. We reduced the number of candidate variables by choosing representative variables from groups of related measures. We further reduced the pool of candidate variables by using expert opinion to remove intercorrelated variables. No two variables in final candidate set had Pearson correlation coefficients greater than 0.5. We conducted all statistical analyses in R (R Development Core Team 2011).

Physicochemical characteristics
Water temperature ranged from 3.0°C in winter to 23.8°C during the summer (Table 2). Stream discharge (Q) ranged from 0.9 to 80 L s -1 during the study, was highest during the winter, lowest during the summer, and greater in AMB than EST or EAS. Water velocities in the buried reaches (range: 1.9-14.2 m min -1 ) exceeded those in the open reaches (range: 0.8-4.0 m min -1 ) of the same stream by a factor of 3.8, on average (p \ 0.001, Table 2). Stream width was greater in the open reaches than in the buried reaches at EST and EAS (p \ 0.001), though not in AMB. Specific discharge (Q s = Q/w), an index for the potential hydrologic control of S w , was 2.0 and 3.0 times greater in the buried than open reaches at EAS and EST, respectively. At AMB, however, Q s in the buried reaches was only 0.76 of that in the open reach.
Photosynthetically active radiation (PAR) in the open reaches was lowest during the summer (mean = 2.0 mol photon m -2 day -1 ) and equally high during the fall and winter (mean = 3.6 mol photon m -2 d -1 ; Table 2). No PAR penetrated into the middle of the buried reaches. Due to instrument malfunctions, PAR data are not available for the open reaches during the spring season.
Stream nutrient concentrations were moderately high throughout the study (Table 2). Nitrate ranged from 164 to 822 lg N L -1 (mean = 476 lg N L -1 ) and did not differ by stream or reach (p [ 0.07). Ammonium averaged 9 lg N L -1 and composed less than 2 % of the dissolved inorganic N pool among all reaches. Dissolved reactive phosphorus was lowest in EAS (mean = 25 lg P L -1 ) and averaged 65 lg P L -1 in EST and AMB. DOC averaged 3.0 mg L -1 , ranged from 1.7 to 4.5 mg L -1 , and did not differ by reach (p = 0.54).
Benthic organic matter, bacteria abundance, and extracellular enzyme activity At the per sample scale, benthic organic matter and chlorophyll a standing stocks were greater in the open than buried reaches (p \ 0.001; Fig. 3a-d). CBOM standing stocks varied seasonally in both open and buried reaches, with the highest values occurring during the fall (p \ 0.001). Both periphyton and chlorophyll a standing stocks were greatest during the spring and winter in the open reaches (p \ 0.001), but did not vary seasonally in the buried reaches (p [ 0.05). Periphyton was more abundant in the buried reaches lined with concrete (EST) and metal (EAS) than the reach that had filled in with sediment (p \ 0.001). Reach-scale standing stocks are presented in Table 2 and follow the patterns described above, with the exception of periphyton at EST. At  (Table 2).
Algal cell density on tiles was three orders of magnitude greater in the open (mean = 2.1 9 10 6 ind cm -2 ) than buried (mean = 1.0 x 10 3 ind cm -2 ) reaches (p \ 0.001; Fig. 4a). The number of 16S rRNA gene copies per cm 2 of colonized tile surface ranged from 6.5 x 10 3 to 2.6 x 10 7 and was related to stream and season (p \ 0.001), but did not differ by reach (p = 0.59; Fig. 4b). The LCI calculated from the EEA levels associated with periphyton collected from the tiles ranged from 0.43 to 0.99 (mean = 0.84) and was 20 % greater in the buried than open reaches (p = 0.002; Fig. 4c).

Stream metabolism
Dissolved oxygen (DO), expressed as percent saturation, exhibited strong diurnal variation in the open reaches that was greatly attenuated, or completely absent, in the buried reaches. GPP ranged from 0.05 to 5.25 g O 2 m -2 day -1 (mean = 1.28) in the open reaches and the highest values were observed during winter or spring ( Table 2). The buried reaches were isolated from PAR and supported no GPP. ER was lower in the buried (mean = -0.86 g O 2 m -2 day -1 ) than open (mean = -1.28 g O 2 m -2 day -1 ) reaches (p = 0.02). Log transformed ER in the open and buried reaches was positively correlated with total organic matter standing stocks (AFDM m -2 ) (p = 0.01, r 2 = 0.24). Net ecosystem production in the open reaches was negative, with the exception of AMB during the spring (NEP = 0.4 g O 2 m -2 h -1 ; Table 2).
We detected a decline in 15 NO 3 tracer flux across the reach in 19 of 24 tracer experiments (Table S1 in Supplementary material). Three buried reach and one open reach tracer injection did not yield a decrease in 15 NO 3 flux and we calculated the method detection limit for these measurements. One measurement was not executed properly (i.e. poor mixing at the upstream station) and yielded no information on NO 3 uptake rates.
Nitrate uptake length (S w ) was greater in the buried reach of every stream during all seasons, with the possible exception of the summer measurement in EST where NO 3 uptake in the open reach was below detection (Fig. 5a-d; Table S1 in Supplementary material). The detection limit for k in this reach is Fig. 3 a Coarse benthic organic matter (CBOM) standing stocks, b fine benthic organic matter standing stocks, c benthic chlorophyll a, and d periphyton standing stocks in the buried and open reaches during each sampling season. Error bars are standard errors of the mean 2.04 9 10 -3 m -1 , indicating that the true value for the uptake length was C490 m (e.g. 1/k), whereas S w in the buried reach was 2,016 m. Nitrate uptake lengths in the buried reaches exceeded those in the open reaches of the same stream by a factor of 29 (p \ 0.001), on average. Nitrate uptake length shortened with increasing ER (p = 0.002, r 2 = 0.33) and was not significantly related to season (p = 0.53).
In all cases the effect of burial on S w was disproportionate to the effect on Q s , indicating that burial suppressed biological N demand (Fig. 6). This was also evident in the NO 3 m f measurements which were greater in the open reaches of all streams during all seasons, with the possible exception of the summer measurement in EST where NO 3 uptake in the open reach was below detection (Fig. 5e; Table S1 in Supplementary material). The true value of m f in this reach was between 0 and 0.46 mm min -1 , whereas m f in the buried reach was 0.32 mm min -1 . Nitrate uptake velocities in the open reaches (range: 0.05-1.98 mm min -1 ) exceeded those in the buried reaches (range: 0.003-0.74 mm min -1 ) of the same stream by a factor of 17.1 (p \ 0.001). Uptake velocity was not significantly related to season (p = 0.79). The only factor retained in the full model was ER (positive correlation, p = 0.046, r 2 = 0.14), but in pair-wise comparison, m f was weakly predicted by algal abundance (positive correlation, p = 0.04, r 2 = 0.15) and was marginally correlated with reachscale GPP (positive correlation, p = 0.052). Nitrate uptake velocity was not related to the standing stock of any benthic organic matter compartment.
Nitrate uptake velocity in the buried reaches varied considerably (range: 0.003-0.74 mm min -1 ), but was unrelated to stream, season, or any other measured variable (e.g. water chemistry, temperature, organic matter, etc.). Similarly, the difference in NO 3 m f between pairs of open and buried reaches was highly variable, but unrelated to stream, season, buried reach length, or water travel time through the buried reach.

Discussion
Our data clearly indicate that stream burial reduces NO 3 retention and fundamentally alters the flow of energy through streams. We found that NO 3 uptake length (S w ) was 29 times greater in buried than open reaches, reflecting differences in hydrology and biological N demand between reaches. Biological N demand in buried reaches was lower than that of open reaches due to reduced organic matter quantity and quality, as well as the absence of assimilatory NO 3 demand by autotrophs. Stream burial also increased specific Effect of stream burial on NO 3 uptake: hydrologic mechanisms We found that stream burial increased NO 3 -S w , indicating that burial substantially reduces the N-removal capacity of streams. This effect could have occurred through biotic or abiotic mechanisms, or a combination of both. An abiotic factor that affects S w is stream water NO 3 concentration. As NO 3 concentration increases, the pool of NO 3 molecules that biological activity could remove increases, making it less likely that any one molecule will be removed per unit time, causing S w to lengthen, even under constant biological N demand. However, NO 3 concentration can also influence biologic N demand through Michaelis-Menten kinetics  which predicts that NO 3 concentration can increase more rapidly than biological N demand, resulting in lengthening S w with increasing NO 3 concentration. We intentionally minimized the effect of NO 3 concentration in our analysis by comparing uptake length in buried and open reaches that were nested within streams and separated by only 30-100 m, which ensured that NO 3 concentrations did not differ substantially between reaches. Specific discharge (Q s ), calculated as discharge divided by width, is another abiotic factor that has a direct effect on S w . Specific discharge is reported in units of area per time and expresses the benthic surface area over which stream water passes per unit time. Large Q s values indicate that water column nutrients are rapidly transported over large expanses of stream sediment with little opportunity for biological uptake.
The effect of stream burial on Q s varied across streams and was related to pipe design. The buried reach at AMB flowed through an oversized corrugated metal pipe (i.e. 4.5 m wide at base) with a Q s about 20 % lower than the open reach. In EST and EAS, however, the buried reaches were constructed of narrow pipes that restricted the stream width to less than 1 m and the Q s values ranged from 130 to 420 % of that in the open reaches. If the biological demand for NO 3 did not differ between open and buried reaches, then the effect of stream burial on uptake length should scale linearly with the effect of burial on Q s . It follows that if the effect of stream burial on uptake length was strictly due to hydrology, all observations should fall along the 1:1 line of a plot of the effect of burial on S w versus the effect of burial on Q s , where the 'effect size' is expressed as the ratio of the parameter value in the buried and open reaches (Fig. 6). All observations fell above the 1:1 line, indicating that differences in hydrology were not sufficient to account for differences in S w between open and buried reaches in these streams. The unexplained effect, represented by the residuals between the observations and the 1:1 line, reflects reduced biological demand for NO 3 in the buried reaches.
The relative importance of biological NO 3 demand and Q s in determining the effect of stream burial on S w differed by stream. In AMB, stream burial enhanced hydrologic NO 3 retention, as indicated by the lower Q s , yet S w was longer in the buried reach, a clear indication that burial increased S w through purely biological mechanisms. On the other extreme, the data points from EST fell near the 1:1 line, indicating that the effect was predominantly hydrological. EAS represents an intermediate case where both hydrologic and biological effects were important, but the biological effect in EAS outweighed that of the hydrologic effect by a factor of 39 (median = 10).

Effect of stream burial on NO 3
uptake: biological mechanisms An alternative measure of NO 3 uptake that accounts for Q s is uptake velocity (v f ) which is expressed in units of distance per time and can be thought of as the depth of the water column from which NO 3 can be removed via biological activity per unit time (Stream Solute Workshop 1990). Uptake velocity is independent of hydrology and represents an index of biological NO 3 demand within the reach. Stream burial reduced v f by a factor of 17, on average, representing the net effect of burial on biological NO 3 removal processes. Our study included numerous indices of autotrophic and heterotrophic removal processes (e.g. GPP, ER, algal abundance) that explained variation in v f nearly as well. However, all these indices were affected by stream burial and as a result were correlated, complicating our efforts to resolve the relative importance of these variables in controlling v f . Nitrate uptake could be directly and indirectly affected by the absence of light in buried streams. Without sunlight, the reach is incapable of supporting GPP and is completely dependent on energy subsidies from outside the local channel to support metabolic activities. The absence of autochthonous production in buried streams can cascade throughout the ecosystem, ultimately affecting organic matter quality, quantity, and the metabolic rates of heterotrophic organisms. In addition to the indirect effects of GPP on NO 3 uptake, GPP directly affects NO 3 uptake through the assimilative demand of primary producers, which can be the dominant N uptake mechanism in streams (Hall and Tank 2003;Hall et al. 2009). Although algae in the buried reaches were not contributing to NO 3 uptake via autotrophic processes, it is possible that they contributed to NO 3 retention through heterotrophic mechanisms. Some algal taxa have been documented to survive prolonged darkness through facultative heterotrophy, increasing the oxidation rates and the variety of organic substrates utilized in the dark (Tuchman et al. 2006). The contribution of algal respiration to carbon cycling in the buried reaches is unlikely to be large, however, given their extremely low abundance (Figs. 3c, 4a).
Although the effect of burial on the abundance of autotrophs was expected, the effect on total periphyton biomass was not as easily predicted. Periphyton consists of a mixed community of autotrophs and heterotrophs which may interact antagonistically or mutualistically (Findlay et al. 1993). Elimination of autotrophs could benefit heterotophs if the two groups were competing for space or nutrients. If heterotrophs are relying on algae as a source of labile carbon, however, the loss of autotrophs could result in reduced heterotrophic biomass. While burial clearly reduced the abundance of the autotrophic community, periphyton abundance (g AFDM m -2 ) was not affected to the same extent (Figs. 3c,d,4a) and bacterial density was unaffected (Fig. 4b). These data indicate that the trophic link between bacteria and algae was not strong in these streams and that bacteria density is relatively resistant to stream burial. Therefore, the net effect of stream burial on periphyton was to produce a less dense assemblage dominated by heterotrophs. The shift in the composition of the periphyton assemblage is supported by the ratio of periphyton to chlorophyll a, an index of the relative abundance of heterotrophs in the periphyton assemblage, which was two orders of magnitude greater in the buried (mean = 68) than open reaches (mean = 0.7). Future studies need to take into consideration the identity, diversity, and functional (denitrifying) potential of the bacterial fraction to further understand how the bacterial network is impacted by stream burial.
Nitrate uptake velocity correlated positively with ER, which was greater in the open reaches. ER can be an important assimilative mechanism for water column NO 3 removal in streams (Fellows et al. 2006;Hoellein et al. 2007;Newbold et al. 2006) and low ER in the buried reaches likely contributed to their reduced biological NO 3 demand. In addition to functioning as a direct NO 3 sink, ER can indirectly promote NO 3 removal by creating the anoxic conditions necessary for denitrification, a type of anaerobic respiration in which NO 3 is reduced to N 2 or N 2 O gas and permanently removed from the ecosystem ). An important energy source for aerobic and anaerobic respiration is particulate organic carbon (Hedin 1990). Particulate organic carbon standing stocks were substantially greater in open reaches, partially explaining differences in ER between reaches. Low organic matter standing stocks in the buried reaches likely resulted from a combination of reduced allochthonous and autochthonous inputs and limited retention. The buried reaches had greatly simplified geomorphology with few in-stream devices to physically retain organic matter for subsequent microbial processing (Hoover et al. 2006).
Dissolved organic carbon is another important energy source for microbial respiration (Wiegner et al. 2005). While we found no differences in DOC concentration between reaches, the microbial LCI indicated a strong reduction in DOC quality in the buried reaches. This is most likely explained by the absence of photosynthesis and photodegradation in the buried reaches, which can be important sources of labile DOC in aquatic ecosystems (Bertilsson and Jones 2003;Moran and Covert 2003). Reduced DOC quality may have further limited heterotrophic metabolic processes, including NO 3 uptake, in the buried reaches. Carbon limitation and low heterotrophic metabolic rates suggest that buried streams are functionally analogous to cave streams where C limitation greatly depresses biological N demand (Simon and Benfield 2002).
We measured the abundance of the bacterial component of the heterotrophic community, but our method did not quantify fungal abundance. Fungi can be important consumers in streams and the relative abundance of bacterial and fungal biomass is largely a function of particle size and composition, where fungi tend to dominate coarse particulate organic matter while bacteria dominate inorganic substrates (Findlay et al. 2002). The buried reaches in this study had little coarse particulate organic matter (Fig. 3a) and predominantly inorganic substrates, which suggests that they had lower fungal biomass than the open reaches. Differences in fungal biomass between open and buried reaches, determined by substrate characteristics at the mm to cm scale, may contribute to differences in ER and NO 3 uptake at the reach scale. Although our experimental design captured seasonal patterns in organic matter standing stocks and stream metabolism (Fig. 3a-d; Table 2), there were no statistically significant seasonal patterns in NO 3 -S w , v f , or the magnitude of the burial effect on these uptake parameters. This may be due to counteracting seasonal drivers. For example, the stimulation of algal NO 3 uptake by high light levels in the spring may be balanced by the inhibitory effect of cool water temperatures (Piña-Ochoa and À lvarez-Cobelas 2006). Similarly, the effects of elevated water temperature during the summer may be balanced by low light. These offsetting effects were likely compounded by a relatively small sample size within each season (i.e. 3 streams per season). Nevertheless, the seasonal component of the study demonstrates that stream burial consistently increases NO 3 spiraling length and reduces v f , regardless of water temperature, stream discharge, organic matter availability, or any of the other controlling factors that vary seasonally.
Nitrate uptake velocity in the buried reaches are among the lowest reported in the literature for daylighted streams and two of the buried-reach m f values are an order of magnitude lower than the minimum value reported from a survey of 72 streams using similar measurement methods (Mulholland et al. 2008). Although the buried-reach m f values spanned a wide range (0.003-0.74 mm min -1 ), the variability was not attributed to season, stream, or any of the other variables measured in this study. Similarly, the magnitude of the burial effect on NO 3 v f spanned a broad range among streams and seasons, but was unrelated to the length of the buried reach, water travel time through the buried reach, or differences in ER between open and buried reaches. Future studies should employ experimental designs better suited to explaining variation in NO 3 v f among buried reaches, possibly by examining a larger number and greater variety of buried reaches. This information could be used to inform management actions designed to minimize the effect of stream burial on NO 3 removal.

Stream restoration implications
Stream burial reduces the NO 3 removal capacity of streams, but this capacity may be restored through 'stream daylighting' (Pinkham 2000), a stream management practice where buried streams are removed from underground pipes and placed in new stream channels on the land surface. Numerous streams have been daylighted around the world, ranging from small headwaters to non-wadeable rivers (Buchholz and Younos 2007;Conradin and Buchli 2005;Pinkham 2000). The motivation for these projects is typically economic, including the need to eliminate the costly maintenance of aging infrastructure, or to create an aesthetic centerpiece for a revitalized commercial district. Although many of these projects claim a water quality benefit (Wild et al. 2011), none have collected the data to demonstrate it. Demonstrating the efficacy of stream daylighting to restoring a stream's NO 3 removal capacity remains an important research need that our dataset has begun to address.
While daylighting may restore the NO 3 removal capacity of buried reaches, it should be recognized that the NO 3 removal capacity of most open reaches in urban areas is also impaired. The mean m f value for the open reaches (0.50 mm min -1 ) in this study was a factor of 3 lower than the mean m f reported from a much larger study of NO 3 uptake rates in streams using similar isotope tracer methods ). Urban streams are subject to a suite of stressors that are likely to inhibit stream ecosystem functioning including flashy flows, emerging contaminants, and degraded riparian zones (Rosi-Marshall & Royer 2012;Walsh et al. 2005). It should be recognized, therefore, that stream daylighting is unlikely to restore the NO 3 removal capacity of streams to pre-development levels, rather daylighting may increase NO 3 removal capacity to a level equivalent to that of adjacent open reaches. The best results will likely be realized when the daylighted channels include elements specifically designed to promote NO 3 retention such as a reduction of stream bank incision to reconnect floodplains or physical mechanisms to reduce stream velocity and retain organic matter Kaushal et al. 2008b;Mayer et al. 2010;Mayer et al. 2013;Passeport et al. 2013;Sivirichi et al. 2011).
Stream daylighting may be impractical in some space-limited urban environments. A possible alternative may be 'skylighting' where large grates, or other screened openings, are installed directly above the buried stream channel. Even if these devices are only a few meters in length, the patches of light would likely create localized areas of primary production that could stimulate biological NO 3 removal. A related concept is to minimize the length of continuous stream burial whenever possible. Even short patches of the stream channel exposed to sunlight could stimulate NO 3 retention for some distance downstream. Nitrate uptake length is directly affected by Q s , which is largely dictated by the design of the underground stream channel. Wide pipes may reduce Q s , enhancing NO 3 retention, while narrow pipes tend to increase Q s and promote hydrologic NO 3 export. Wider pipes with a lower slope may also promote the retention of alluvium which will increase bed roughness and transient storage and further enhance organic matter and nutrient retention. A potential management intervention to promote NO 3 retention in buried streams is to adopt a policy of promoting wider pipes. This should be tempered, however, with the recognition that extremely shallow buried reaches can be a dispersal barrier for aquatic organisms (Jungwirth et al. 1998), though this may not be a large concern in concentrated urban environments where the biological communities are already impaired. Another consideration is that disruption to the stream during pipe construction, which may be proportional to the size of the pipe, should be minimized according to best management practices.
An alternate management action is to adopt channel designs that increase water travel time during periods of low flow, but do not impede water at high flows when flood risks are greater. One such design may include a small meandering channel within the pipe that effectively increased the length of the flow path at low flows, but at high flows the water would over flow the channel walls within the pipe and fill the entire pipe diameter. Another approach may be to increase bed roughness and/or install structures in pipes that slow water velocity and retain sediment, thereby creating a shallow benthic zone that may function to support microbial activity.
Urban stream managers can also control the bed material in buried streams, which may have implications for organic matter standing stocks and NO 3 uptake. The buried streams with exposed metal (Fig. 2d) or cement stream bottoms (Fig. 2a, b) supported larger periphyton standing stocks than the buried stream that had filled in with a shallow layer of pebbles and sand ( Fig. 2c; Table 2), likely because the more stable substrate provided better growing conditions (Cattaneo et al. 1997). This effect was somewhat offset, however, by the greater particle trapping ability of the sand and pebbles which retained substantially more FBOM than the metal or cement stream bottoms (Table 2). Whether FBOM or periphyton provides a stronger NO 3 sink is not well known, though measurements in forested streams suggest FBOM supports higher NO 3 uptake rates (Hoellein et al. 2009). Despite the differences in organic matter composition between buried streams, NO 3 v f did not differ between these reaches. Furthermore, our experimental design lacked replication within the three buried stream types (i.e. pebble, metal, concrete) making it difficult to associate any differences in NO 3 v f with streambed composition. Our study did demonstrate, however, that regardless of streambed composition, stream burial suppresses NO 3 uptake. Future work should investigate whether this effect can be minimized by designing buried streams that support periphyton, trap particulate organic matter, and support hyporheic flow paths (Lawrence et al. 2013).

Effect of stream burial on NO 3
transport at the watershed scale Our data demonstrate that burial substantially reduces NO 3 retention in streams. What may be of more interest to watershed managers, however, is the cumulative impact of stream burial on NO 3 transport at the watershed scale. A related question is how much additional NO 3 retention may be realized by daylighting currently buried reaches. In the next phase of this research, we will design a watershed model to explore these questions as a first step toward setting targets for watershed restoration. We anticipate the approach will be particularly useful for investigating the relationship between the spatial distribution of buried stream reaches and watershed scale NO 3 retention along stream networks.

Summary and Conclusions
We found that stream burial increased NO 3 uptake length through a combination of mechanisms. Burial eliminated primary production in the buried reaches which reduced biological NO 3 demand, autochthonous organic matter inputs, and DOC quality. Burial also reduced allochthonous particulate organic matter availability by isolating the channel from the riparian zone and reducing in-stream organic matter retention. Low organic matter availability and quality translated to depressed ER rates and reduced biological NO 3 demand. In some instances, stream burial increased Q s , which further reduced the capacity of the stream channel to retain NO 3 -. The effect of stream burial on Q s was not uniform across sites, however, and oversized pipes can reduce Q s and enhance hydrologic NO 3 retention. Designing buried reaches with low Q s values is one management action that may minimize the effect of stream burial on NO 3 uptake. Further research will be necessary to elucidate the factors that control variation in N uptake among buried streams and to quantify the mechanisms of N transformation (e.g. denitrification).