Groundwater Nitrate Removal in Forested and Suburban Vegetated Riparian Wetland Soils: A Mesocosm Mass Balance Approach

Although riparian zones can remove nitrate (N03") from groundwater, the interaction between vegetation type, subsurface patches of elevated organic matter, and N03removal rates is still uncertain. The goal of my study was to examine the effects of vegetation type on groundwater N03removal rates in poorly drained riparian subsoils with four treatments, two "forested" and two "suburban" treatments. I also examined the variability in N03removal rates between riparian sites and the relationship between patches of elevated organic matter in the subsoil and groundwater N03removal rates. My study was based on a mesocosm approach to simulate N03dynamics in the shallow groundwater. These mesocosms were obtained from paired forested and suburban (mowed vegetation dominated by non-woody species) vegetated areas at two different sites. Intact horizontal mesocosms (15 cm diam., 40 cm long PVC cores) were extracted from fine to medium textured sands. The soils were derived from stratified glacial drift and were classified as Typic humaquepts or Umbric endoaquods. At each site, I obtained mesocosms from at least 35 cm below the dormant season water table depth. These mesocosms were placed in a controlled environmental chamber and continuously dosed with bromide and N03amended groundwater for 71 days. Mass balance calculations were completed to determine the rate of groundwater N03--N removal from each saturated mesocosm. There was no significant difference in groundwater N03removal rates between forested and suburban vegetated areas within sites. However, there were significant


INTRODUCTION
Nitrate (N03-) is the most commonly detected groundwater pollutant in wells in the United States . N0 3 -is a federally registered drinking water contaminant and at high concentrations can cause methemoglobinema (blue-baby syndrome) in infants. N0 3 -also has been linked to eutrophication of coastal waters, and watershed management of nitrogen (N) is a major coastal issue Howarth et al., 1996;.
As a form of non-point source pollution, N0 3 -originates at diffuse sources, including septic systems and agricultural fields Weiskel and Howes, 1992). Researchers studying coastal watersheds ranging from local scales ( 1000 ha) to regional scales (Mississippi Basin) have found that less than one-third of the estimated net anthropogenic N input to watersheds is discharged to coastal waters Howarth et al. , 1996). These studies suggest that we need additional research to identify and improve our understanding of watershed sinks of N.
Many studies have shown that riparian zones have the ability to remove N0 3 -from gro undwater . Riparian zones are functionally defined as areas where energy and matter are exchanged between uplands and surface waters . However, there is considerable uncertainty as to the relative importance of plant uptake and rnicrobially mediated processes of groundwater N0 3 -removal and if the type of vegetation cover influences this removal .
Within riparian zones, there is considerable spatial variability in groundwater N0 3removal. Generally, denitrification activity is unevenly distributed throughout the soil, with the greatest activity occurring in uppermost soil horizons . Numerous studies Groffman and Tiedje, 1989;Groffman et al. , 1992; have found that N0 3 -removal and denitrification activity are greater in poorly drained (PD) soils (wetland edge) than in moderately well drained soils (upland edge). PD soils are water-saturated for longer periods of time, promoting anaerobic conditions and the accumulation of organic matter, thus enhancing the potential for denitrification ).
An interesting aspect of riparian zone research is the discrepancy between groundwater N0 3 -removal in companion field and laboratory microcosm studies.
Microcosm laboratory studies using media from shallow aquifers have consistently found low denitrification rates (Pennock et al. , 1992;Groffman et al. , 1996). In contrast, high 0 3-removal rates in shallow riparian groundwater have been observed in a large number of field studies based on groundwater well networks . The disparity between these studies may be due to plant N0 3 -uptake, which would not be included in the laboratory microcosm studies. In addition, differences in sample volume between field and laboratory studies may be responsible for such discrepancies. Laboratory measured microbial activity may not adequately represent field-measured activity because field studies encompass a greater extent of the heterogeneity of natural soils (Smith et al. , 2 1996).  and Christensen et al. (1990a) postulate that the majority of denitrification in surface soils may occur in microsites or " hotspots" of organic matter that are often unevenly distributed through the soil. As sample volume increases, there is a greater chance of incorporating these hotspots of activity, and denitrification variability may decrease .
Recently, a laboratory mesocosm study (Gold et al ., in review; Jacinthe et al. , in review) with 12 kg of soil per mesocosm produced groundwater N0 3 -removal rates comparable to a riparian field study at the same site . Mesocosms (" medium-sized worlds") are intermediate between field and microcosm studies.
Mesocosms are samples large enough to incorporate the natural physical and biological heterogeneity of soil but still small enough to manipulate under controlled experimental conditions . In contrast to this mesocosm study, microcosms with less than 50 g of subsoil showed negligible removal rates at these sites (Groffman et al., 1996). In the mesocosm study, PD sites exhibited high groundwater N0 3 -removal and contained dark-stained patches of elevated organic matter. Moderately well drained mesocosms displayed no groundwater N0 3 -removal and did not contain any such patches.
Most recent riparian zone research has focused on forested areas Ambus and Lowrance, 199 1; Lowrance, I 992; Groffman et al. , 1996;Gold et al ., in review;Jacinthe et al ., in review). As a result of these studies, it is generally accepted that forested riparian zones have the ability to serve as N sinks. Fewer studies have examined the difference in N removal between forested and 3 non-forested riparian zones (Haycock and Pi nay, 1993;Osbourne and Kovacic, 1993 ;. Grassed riparian sites have often been found to have lower N removal rates than forested riparian sites, but in a few instances, substantial removal has been observed in grassed riparian sites. More studies are needed to determine whether there are substantial differences in groundwater N removal between forested and non-forested riparian subsoils. There is great interest by policy makers in using riparian zones to protect water quality. Restoration of vegetated riparian zones may play a critical role in the mitigation ofN loading from both agricultural and unsewered suburban development. I suggest that riparian zone restoration and maintenance programs may meet with greater success if the landowner has flexibility in the use and design of the riparian area (Desbonnet et al ., 1994). In developed suburban areas, riparian zones may obtain greater support ifthe public can use the riparian area for a variety of purposes, including nature walking, picnicking, and obtaining a scenic view of an adjacent waterbody (Desbonnet et al. , 1994).
Therefore, an obvious question arises : within a riparian zone, can a mix ("suburban" vegetation) of lawn, shrubs, and trees maintain the same extent of groundwater N removal as forested areas? The answer to this question will have great implications in the development of many riparian zone management plans.
The goal of my study was to examine the effects of vegetation type on groundwater N0 3 • removal in the subsoil of PD riparian zones . The carbon available for microbial activity from different vegetation types may differ both quantitatively and qualitatively (McKenney et al. , 1995). I hypothesize that greater amounts of groundwater N03will be removed from forested riparian areas than suburban vegetated riparian areas because forests may generate deeper roots that may create more patches of elevated organic matter in subsoils to facilitate denitrification in groundwater.
My study was based on a saturated mesocosm approach. These mesocosms were obtained from paired forested vegetated and suburban vegetated areas at two different riparian sites. Since the groundwater N0 3 -in these saturated mesocosms was not subjected to plant uptake, only microbially mediated removal processes were examined.

Objectives
The specific objectives of this study were: 1) to compare groundwater N03removal rates between forested and suburban vegetated riparian areas; 2) to explore variability in groundwater N0 3 -removal rates between sites of similar soils, drainage, and vegetation; 3) to investigate the relationship between subsoil patches of elevated organic matter and groundwater N0 3 -removal rates within the shallow groundwater of vegetated npanan zones.

Overview of Methods
I created saturated mesocosms to simulate shallow aquifer dynamics in forested and suburban riparian zones. The mesocosms consisted of large, undisturbed soil cores (15 cm diam. , 40 cm long Schedule 35 PVC pipe) obtained from the upper portion of the saturated zone. The mesocosms received a constant inflow of groundwater amended with Bromide (Br) and N0 3 --N at a 1: 1 ratio for 71 days. N loss within the saturated mesocosms was determined by a mass balance approach, based on the difference between the inflow and outflow mass of N0 3 --N over the entire study. At the conclusion of the study, I related the extent of patches of elevated organic matter within the mesocosms to observed N losses.

Site Characterization
Two study sites, "N orth Kingstown" and "Charlestown,'' were selected for this project. At each site, [ selected two different vegetated areas, "suburban" and " forested ." The four treatments in my study were: Suburban areas ab utted forested areas . Suburban areas were defined as areas that were at 6 least 5 m away from the adjacent forest and were dominated by mowed non-woody emergent vegetation, such as grasses and sedges. Within each site the paired suburban and forested vegetated riparian areas were similar in soil type and hydrology.
One criterion in site selection was that all sites be located in PD soils, since PD riparian soils have been shown to have a substantial capacity for N0 3 -removal . At each site, intensive field soil surveys were performed to delineate the extent of PD soils. I also sought sites with glacial outwash parent material of fine to medium sands at depths below the seasonal high water table. Finer textured soil media tend to have low hydraulic conductivities and often generate surface seeps that negate the importance of groundwater removal process investigations in these areas. Coarser soil media also would not have been suitable for this study because it would have made the core extraction procedure more difficult.
We dug three pits in the suburban area and three pits in the forested area at both the North Kingstown and Charlestown sites. One core was extracted from each pit. I conducted detailed soil morphological investigations at each pit (Appendix C) . I recorded thickness and depth of 0 , A, B, and C horizons in each pit. Within each of these horizons, I described the soil texture, soil color, and boundary divisions . Samples of matrix material in 0 , B, and C horizons and patches of elevated organic matter found below 50 cm from the surface were collected and stored at 4 °C to be used later for analyses of texture, percent moisture, percent carbon (C), and additional tests for a companion study on microbial processes.

7
North Kingstown Site

Setting
The first study site is located in North Kingstown, Rhode Island in the Pettaquamscutt watershed (41°32'N, 71°28') ( Fig. 1). The site drains to a first order tributary of Silver Spring Lake. The site is approximately 75 m from the tributary . The site consists of two abutting vegetated areas: a suburban backyard and a hardwood forest (Fig. 2). The soil at both areas consists of PD outwash material. Surface water from both the suburban and forested areas drains to a vernal pool located within the forested area.

Geology and Soils
The surf1cial geology of the North Kingstown site is composed of glaciofluvial deposits with 0-3% slopes. Soils are classified as Typic humaquepts . The pH of the C horizon was 4.6 and 6.0 at the forested area and the suburban area, respectively.

Vegetation
The mowed suburban area is dominated by sedges ( Carex sp. ), a variety of mosses, vio lets (Viola sp. ), dandelions (Taravacum officinaLe), and clover (Trifolium sp.   The suburban area is closest to the road (Fig. 4). From about 1960 to 1980, this site was used as a large garden. Horse manure was applied as the principal form of fertilizer. The garden was then abandoned in the early 1980' s. A portion of the site reverted back to forest and another portion was managed by mowing.

Geology and Soils
The surficial geology of this site is composed of glaciofluvial deposits with 0-3 % slopes. The soils in the suburban area are classified as Typic humaquepts, but the soils in the forested area are classified as Umbric endoaquods. The only difference between these two soils is that the latter has a 5 cm thick Bh layer, an illuviation of organic matter, General location is within the box. At the owners ' request, the exact location is not identified . immediately above the C horizon. The pH of the C horizon was 6.5 and 7.0 at the forested area and the suburban area, respectively.

Vegetation
The mowed suburban area is dominated by emergent vegetation, including Mowing occurs twice a year. Pits within the suburban area were dug at least 6 m from the forest edge (Fig. 4) . The pits were not located below the canopy of the adjacent forest.
Speckled alder (8-16 cm diam.) dominates the overstory of the Charlestown forested area. The overstory speckled alder trees were approximately 18 to 23 years old .
The understory includes red maple saplings, spice bush, skunk cabbage (Symplocarpus foetidus) , sensitive fern , vibernum, royal fern , cinnamon fern, sweet pepperbush, and blueberry (Vaccinium corymbosum) . Pits in this area were dug only l to l.5 m within the forest because the PD section of the forest was a fairly narrow strip. Pits were dug within I m of woody vegetation in this forested area (Fig. 4).

Groundwater Monitoring and Sampling
Water

Core Extraction
Three replicate, horizontal soil cores (15 cm diam. , 40 cm long Schedule 35 PVC pipe) were taken from each forested and suburban area (6 cores per site) . One pit was dug for each core extracted . We dug pits to the approximate dimensions of 90 cm by 120 cm by 90 cm deep. We randomly located three pits within each forested and suburban area. Each pit was at least 2 m away from the adjacent pit. Cores were extracted in July and August of 1996 at the low annual water table to obtain media at least 30 cm below the dormant season water table level of the area. When possible, cores were extracted on the same day as the pits were dug, but the pit was not exposed longer than 5 days before the core was extracted.
We extracted cores horizontally from a randomly selected side of the pit by pressing the core into the soil with an 8 ton hydraulic jack (Appendix B). At the orth Kingstown site, we extracted cores from a depth of 67.5 cm (midpoint of the core) from the surface. We extracted Charlestown cores from a depth of 57.5 cm (midpoint of the core) because an elevated water table prevented deeper extraction. Extracted cores were stored at 4 °C.
A slight bias may have been introduced into the forested cores when we selected which side of the pit to extract cores. Cores were not directed into the side of a pit immediately below a main tree trunk. However, pit selection was not biased by tree distribution; we used hacksaws to cut through large tree roots close to the surface.
Therefore, it was possible to extract cores relatively close to trees within forested areas.
We did not find impenetrable, thick tree roots at core extraction depths, 50-75 cm below the ground ' s surface.

Experimental Design
Once all 12 cores were extracted, I transported them to the Institute of Ecosystem Studies (IES) in Millbrook, NY in August of 1996 where the experiment was completed .
At IES , I capped both ends of the cores with flat PVC end-caps. Above a l cm PVC spacer in the interior of the cap, the soil within the core rested on a 1/8 in. circular plexiglass plate with a series of I /8 in. holes drilled into it. Before placement in the cap, l wrapped this permeable plexiglass plate in nylon to act as a coarse filter and stabilize the soil within the core. I screwed one l/ 16 in. polypropylene adapter into each cap within the spacer section. I then attached the caps to the cores with silicone. To allow the silicone to dry, the cores were stored at room temperature overnight.

20
Throughout dosing, the mesocosms sat vertically in a controlled climate chamber maintained at a constant temperature of 13 °C (Fig. 5) . This temperature represents the groundwater temperature observed in a riparian forest in November when the water table is expected to inundate the sampling location . An Ismatec multi-channel peristaltic pump (Cole Palmer, IL) continuously pumped dosed groundwater from four l 0 L carboys (wrapped in aluminum foil to prevent light penetration) into the bottom of each mesocosm through the adapter. Each carboy fed three mesocosms. Water exited the mesocosms through the adapter in the top cap. The water flowing out of the cap passed through a PVC enclosure with a silicone cell  inside to collect gases (for a companion study on microbial processes) and then into an acid-washed container. The pumping rate was set to 234 ml day°' . The mesocosms had an estimated pore volume of 2665 cm 3 (bulk density=l.65 g cm-3 ; porosity=0.38) and an expected retention time of 11.4 days . Every 2 to 3 days, outflow volumes were recorded and representative samples were stored in 20 ml polyethylene scintillation vials . The samples were stored at 4 °C until analysis.

Groundwater Dosing
All mesocosms received the groundwater from the same source of amended groundwater. Groundwater was collected from a shallow well located in the PD soil drainage class at Peckham Farm, University of Rhode Island, Kingston, Rhode Island (Gold et al. , in review) . I pumped groundwater out of the well with a submersible drainage pump into 20 gal storage containers. These containers were stored at 4 °C until transport to IES where they were stored in the controlled climate chamber. The average ambient concentrations of Br-and N0 3 --N in this groundwater were 0.2 mg L-1 (the detection level) and 0 .4 mg L-1 , respectively.
Deionized water was pumped through the mesocosms for seven days (days -6 to 0) until the mesocosms were saturated and the outflow rate equaled the inflow rate. For the next 9 day period (days 1-9), groundwater amended with 5.2 mg L-1 B( (as KBr) was pumped through the mesocosms. This B( only dosing period was included to observe the extent of mineralized nitrogen within the mesocosms. The use of B( also served to document if preferential flow was occurring in the mesocosms. We carefully monitored 8( concentrations to assure that high 8( concentrations did not appear in the first several days following the inception of B( dosing.
For the fo llowing 71 days (days 10-80), the mesocosms received groundwater amended with 5.2 mg L-1 8( and 5.2 mg C 1 N0 3 --N (as KN0 3 ) . Addition of B( was 24 continued to ensure that the dosing concentration remained stable throughout this period. Due to a pump malfunction, flow through the mesocosms was suspended for 6 days (days 54-60).
After the 71 days of B( and N0 3 --N dosing, unamended groundwater was pumped through all mesocosms for 36 days (days 81-116) . This stage was completed to flush the amendments from the mesocosms. Flushing was deemed complete when B( concentrations dropped to negligible levels. In total, the mesocosms were dosed with groundwater for a total of 116 days.  oc.

Characterization of Mesocosms
Once the groundwater dosing of the mesocosms was complete ( , 1986). Inflow groundwater samples were also prepared by potassium persulfate digestion and analyzed for total N on the Alpkem 300 RF A. Soil pit samples were analyzed for% C on a Carlo-Erba CNS analyzer, for percent moisture , and for texture (Troeh and Thompson, 1993). Matrix soil samples and plant roots taken directly from the mesocosms during final destructive sampling were also analyzed or for percent moisture. We obtained the dry mass of soil and roots by drying the samples at 104 °C and 60 °C, respectively, for 24 hours. A companion study is underway to examine gas outflow from the soil mesocosms as well as to determine immobilization and microbial transformation rates in patch and matrix material from the mesocosms. The total mass ofN03--N (MassINFLOWNo 3 -N) that entered each mesocosm over 116 days of groundwater dosing was then determined as the sum of the daily inflow masses.

Mass Balance Calculations
The mass ofN0 3 --N leaving each mesocosm per day was determined by: The total outflow mass ofN0 3 --N (MassOUTNo 3 -N) that left each mesocosm over the whole course of the experiment was then determined as the sum of the daily outflow masses.
The mass of groundwater N0 3 --N removed or retained in the mesocosms was : where MRNo 3 -N =the total mass of groundwater N0 3 --N removed or retained in the mesocosms over 116 days of groundwater dosing (mg) Outflow Br" and N0 3 --concentrations versus time were also plotted for each mesocosm to examine the trends in 0 3 -removal (Appendix E) .
In order to compare the mesocosm g roundwater

Data Analysis
Statistical analyses were performed on the N0 3 --N removal rate data to determine significant differences between treatments. The Mann-Whitney U Test  for nonparametric data was conducted to determine significant differences in N0 3 --N removal rates between forested and suburban areas at each site. The Mann-Whitney U Test was used because the sample sizes were small (only 3 replicates per treatment) . No significant differences were found between forested and suburban areas; therefore, forested and suburban data were pooled at each site. Another Mann-Whitney U Test was then performed to determine significant differences in N0 3 --N removal rates between North Kingstown and Charlestown sites. Similar Mann-Whitney U Tests were performed on NH/ -N generation rates and net DIN removal rates.
I used Spearman ' s Rank Correlation test  to explore the relationship between : l) N03--N removal rates and the mass of patch material C, 2) N0 3 --removal rates and the mass of matrix material C, 3) N0 3 --N removal rates and the mass of total C in the mesocosms, 4) N0 3 --N removal rates and the dry mass of roots in the mesocosms, and 5) the mass of match material C and the dry mass of roots in the mesocosms. Spearman's Rank Correlation test was used because the data were not related linearly. All statistical analyses were performed on Statistica for Windows (StatSoft, Inc., Tulsa, OK) .
Differences and correlations were considered statistically significant at the P<0.05 level.

Site Characteristics
Soil morphology differed between sites but was generally similar within sites (Table 1 and Appendix C). Mean depth to the C horizon differed significantly between sites, but not between vegetated areas at the same site. There was one distinct soil characteristic at the Charlestown forested area. This was the only site that contained a 5 cm thick Bh horizon, which is an illuvial accumulation of organic matter, just above the C horizon (Appendix C). Two out of three Charlestown forested mesocosms contained portions of this Bh horizon. Grain size distribution between sites and vegetated areas was very similar ( Table 1 ) . The effective diameter ( d 10 ) of grain size had a small range, from 0.065 to 0. 15 mm. Likewise the uniformity index varied very little, from 2.33 to 4.31 .
Background N0 3 --N concentrations in groundwater at both sites were low, ranging from 0.065 to 0.585 mg L-1 (Table 1). Water

Flow Characteristics
The pattern and rate of flow through the saturated mesocosms can be characterized by the breakthrough of the conservative tracer, B(. Bypass flow or macro pore flow did not occur in any of the mesocosms. No high B( concentration was observed in the outflow of any mesocosm during the first several days of dosing. Instead, there was a gradual increase in B( concentration over time ( Fig. 6 and Appendix E) .
After one pore volume of outflow, the concentration of a conservative tracer in a homogeneous media is expected to be half the inflow concentration (Hillel, 1980). The 0.5 relati ve concentration (outflow concentration/inflow concentration, i.e. C/Co) of B( indicating very little retardance of flow or preferential flow through the mesocosms (Li and Ghodrati, 1994). There were no observable differences in between sites. This suggests that any differences in N0 3 -removal rates observed between mesocosms was the result of internal N processing as opposed to retention ·time.

Nitrate in Blank Treatment
No groundwater N0 3 -removal was observed in the blank treatment. B( and N0 3 --N concentrations were not exactly at a 1: l ratio since the groundwater itself contained a small amount of background N0 3 --N, ranging from 0.3 to 0.4 mg L-1 (Fig. 7). Once the groundwater amendments were discontinued, the outflow N0 3 --N and B( concentrations gradually declined to negligible levels. N0 3 --N and B( concentrations did not completely flush from the blank. This lack of flushing may have occurred because the blank had a larger volume (3 L) and a corresponding longer retention time than the pore volume of the mesocosms (2 .7 L) . The absence ofN0 3 -removal in the blank treatment indicates that any N0 3 -removal observed in mesocosm was due to processes within the mesocosm soil.

Groundwater Nitrate Removal Rates
There were no significant differences in groundwater N03--N removal rates between forested and suburban areas at either site ( Fig. 8) Therefore, the removal rates from forested and suburban areas were pooled together at each site. The Charlestown site rates than the North Kingstown site (mean: 6.8 µg kg-' d-1 ; S.E .: 3.9).

Patterns in Outflow Bromide and Nitrate Concentration
After B( amendments of groundwater were initiated (day 1 ), outflow B( concentrations gradually increased and stabilized at the inflow concentration ( As expected, the rise in outflow N0 3 --N concentration lagged behind the rise in outflow B( concentration because N0 3 --N amendments did not commence until 9 days after B( amendments. Like the outflow B( concentrations, outflow N0 3 --N concentrations gradually rose after N0 3 --N amendments began; however, the outflow N03--N concentrations rose to varying levels in different mesocosms (Fig . 6 and Appendix E). In the absence of N0 3 -removal , the N0 3 --N concentration should mimic the B( concentration because they were added at a l : l rati o. 43 The difference between outflow groundwater B( and N0 3 --N concentrations indicates the extent of groundwater N0 3 -removal (Fig. 6). For example, because outflow N0 3 --N concentrations in Fig. 6a are relatively close to the outflow B( concentrations from days 30-95 , comparatively little N0 3 -was removed from mesocosm #4 . In contrast, outflow N0 3 --N concentrations in mesocosm #7 (Fig. 6b) remained at less than half the outflow B( concentration for most of the experiment, indicating greater groundwater N0 3 -removal. In general, outflow N0 3 --N concentrations were closer to outflow B( concentrations in the North Kingstown mesocosms than the in Charlestown mesocosms indicating less groundwater N0 3 -removal in the North Kingstown mesocosms. Over time, the outflow N0 3 --N concentrations gradually rose. This suggests that removal rates decreased over the course of the study. As expected, once B( and N0 3 --N amendments were discontinued (day 80), the outflow N0 3 --N concentrations gradually declined, approaching the detection level (0 .2 mg L-1 ) . The final outflow N0 3 --N concentration in five of the mesocosms was below this detection level. In the remaining seven mesocosms, the final N0 3 --N concentrations ranged from 0.3 to 0.6 mg L-1 .
Following the six days of pump stoppage (days 54-60), I observed a rapid decline and then rebound in outflow N0 3 --N concentration in several of the mesocosms . The Charlestown mesocosms experienced a greater decline in outflow N0 3 --concentrations at this point than the orth Kingstown mesocosms. In general , the hi gher the outflow N03--N concentration before the pump stoppage, the less the concentration dipped in response to this increase in groundwater retention time.

Ammonium Generation in Mesocosms
No NH/ -N was added to the groundwater inflow of the mesocosms. The inflow NH/-N concentration was consistently less than the detection level (0 . 1 mg L-1 ) . All the inflow total N was in the form ofN0 3 --N. NH/ -N was observed in the outflow groundwater even before N0 3 --N amendments commenced. This suggests that the NH/-N detected in the outflow groundwater had been generated internally within the mesocosms. However, the observed NH/-N concentrations in all mesocosms were generally low, never rising above 1.9 mg L-1 .
The outflow NH4 + -N concentration varied substantially in magnitude and pattern between vegetated areas ( Fig. 9 and Appendix F). Among the suburban vegetated mesocosms, the outflow~+ -N concentrations were generally either negligible throughout groundwater dosing or were initially negligible and then experienced a slight increase before the completion of groundwater dosing (Fig. 9a) . All the forested mesocosms had initial outflow NH/ -N concentrations greater than 0.9 mg L-1 . In general, the outflow~+ _N concentrations in these forested mesocosms gradually declined throughout groundwater dosing (Fig. 9b ). The pattern of outflow ~ + -concentrations did not correspond with the outflow N0 3 --N concentration patterns.
By day 60, outflow NH4 + -N concentrations in all mesocosms were less than 0. 9 mg L-1 . By the end of groundwater dosing (day 11 6) , NI-Li t-N concentrations in all mesocosms declined to or near the detection level. There was no apparent rise or fall in NH/ -N concentrations that can be attributed to the pump stoppage (days 54-60) .
The~+ _ generation rates were significantly higher in o rth Kings town  S.E. : 2.9) although this difference was not significant. These trends in NH/ -N generation rates contrast with the trends in N0 3 --N removal rates which only exhibited significant differences between sites and not between vegetated areas .

Net Dissolved Inorganic Nitrogen Removal Rates
Because NHi + -N was generated within the mesocosms, I determined the net dissolved inorganic nitrogen (DIN) removal rates. I did not observe any differences in overall trends between N0 3 --N removal rates and DIN removal rates. No significant differences in net DIN removal rates were found between the forested areas and the suburban areas at either site. After pooling forested and suburban vegetated removal rates at each site, I found that Charlestown net DIN removal rates (mean: 32.3 µg kg-1 d-1 ; S.E.: 3.8) were significantly higher than North Kingstown net DIN removal rates (mean: -0 .7 µg kg-1 d-1 ; S.E. : 3.4) . When taking into account NH/ -N generation, N removal rates at the Charlestown site remain substantial.

Root and Patch Characteristics
Dark-stained patches of elevated organic matter were found in all mesocosms (  within a site. However, the mass of patch material in Charlestown mesocosms (mean: 1495 .6 g; S.E.: 433.8) was significantly greater than the mass of patch materiaJ in North Kingstown mesocosms (mean : 2 72. 9 g; S. E. : 218. 5) . The appearance of individual patches of elevated organic mater did not vary between treatments. Some of the darkstained patches of elevated organic matter resembled root channels where the root had completely decomposed . Fine roots or fine root masses were sometimes found within these patches. However, most frequently these patches were amorphous features in the soil, ranging in size from 2 mm by 2 mm to aJmost covering the entire cross-section of a mesocosm' s 15 cm diameter face.
Occasionally in the Charlestown mesocosms, I observed streaking of organic matter that was difficult to separate from the matrix material as " patch." Also unique to the Charlestown forested area was the 5 cm thick Bh horizon that two of the three mesocosms intercepted . Aside from these observations, I did not observe any marked differences in patch material between sites or vegetated areas .
Patch material contained between 0. 63 and 3 . 51 % C while matrix material contained only between 0.16 and 0 .59% C. There was no significant difference in the mass of patch materiaJ C in mesocosms between forested and suburban areas at either site (Table 3) . After pooling the forested and suburban vegetated data, l determined there was significantly more mass of patch material C in Charlestown mesocosms (mean : 3 1.3 g; S.E. : 8 .8) than in North Kingstown mesocosms (mean : 3.0 g; S.E.: 2.4) . The mass of patch material C per mesocosm was significantl y correlated (rs= 0.81) with N03--N removal rates (Fig. lOa) . However, 0 3 --removal rates were not significantly [ Rs=o. 7· -5] correlated with the mass of matrix material C per mesocosm (Fig. lOb) _ The total mass of C within the mesocosms was also found to be significantly correlated (rs= 0.75) with N0 3 --N removal rates (Fig. I Oc ); however, this correlation was not independent of the mass of patch material C and matrix material C.
Roots were found in all mesocosms . I was unable to quantify the extent of root decomposition or mineralization in the mesocosms. The mass of dry roots contained within mesocosms was significantly correlated (r. = 0. 62) with the mass of patch material C per mesocosm (Fig. 11). However, unlike the patch material C, the dry mass of roots per mesocosm was not significantly correlated with N0 3 --N removal rates (Fig . 12).
55 Figure 11 . Relationship between mass of patch material C and dry mass of plant roots per mesocosm.  These results may have profound impacts on the role of riparian zone management for the control of N losses in coastal watersheds. However, it is important to recognize that forested riparian zones serve many other functions, including wildlife habitat, surface water sediment and pollutant removal, and stream habitat . Up to now, riparian zone management plans have considered forested vegetation to be a critical component of riparian zones in order for groundwater N0 3 -removal to occur (Desbonnet, 1994;. My study results cannot be extended to encompass large grassed areas. In suburban areas, vegetated areas are often comprised of a mix of mowed vegetation and woody vegetation. At both sites in my study, suburban soil pits were located only 5-6 m from trees (Figs . 2 and 4). In addition, at the North Kingstown site, L noted tree roots in the soil whi le digging the suburban vegetated pits. At the Charlestown suburban area, I noticed alder saplings that would grow to be 45 cm tall in the interval since the last mowing of the previous summer (September) and the first mowing of the following year (June) . Thus, the lack of intrasite differences in my study may be the result of similar below ground rooting patterns within the two vegetated areas.

Relationship Between Patch Material Carbon and Dry Mass ofMesocosm Roots
60 NIL+ was observed in outflow groundwater even before N0 3 -was added to the inflow groundwater. This suggests that NIL+ was internally generated within the mesocosms through mineralization of existing pools of organic N rather than produced by dissimilatory reduction ofN03-. The change in NIL+-N concentration over time noted in the mesocosms indicates that conditions responsible for NH 4 + flux within the mesocosms were not static. The NH4 + generation rates and patterns of outflow concentrations were substantially different between forested and suburban areas within sites. The brief elevated NIL+ concentrations in the early part of groundwater dosing suggest that the core extraction disturbance effects were short-lived in the mesocosms. NIL+ generation was not significantly different between similarly vegetated sites. The NIL+ generation trends are in striking contrast to the N0 3 -removal rates that did not reflect vegetation differences but rather exhibited site differences. Because of these different trends, NIL+ generation may be linked to a different set of processes than groundwater N0 3 -removal. However, because of the negligible net DIN removal rate at the North Kingstown site, the results from the mesocosm study do not support the concept that all PD soils are large sinks for N. I will focus my discussion on groundwater N0 3 -removal rather than DIN dynamics .
Groundwater N0 3 --N removal rates observed at both sites (Charlestown site mean : been shown in microcosm studies to decrease with depth into the soil profile . However, the majority of denitrification activity has been found to be isolated within the A horizon, generally within the top 10 cm of the soil profile. Since most of the Charlestown cores were taken entirely from the C horizon , the N0 3 -removal rates at 10 cm deeper in the profile may not have differed substantially from those observed in my study.
There has been much discussion in the literature over the relative role of plant uptake and microbial processes in riparian zone groundwater N0 3 -removal .
An advantage of thi s mesocosm approach is that by excluding plant uptake I could examine groundwater N0 3 -removal rates generated by microbially mediated processes.
However, plants may play a role in subsurface processes aside from plant uptake.
Anaerobic microsites can develop in the root vicinity, creating conditions favorable for denitrification (Woldendorp, 1963 ;. Plant roots are the primary source of organic matter in deeper soils . Living roots excrete organic substances into the soil which may become substrates for denitrification . This contribution of C by plant roots has been found to correspond with higher rates of denitrification in rhizosphere soil . However, the age and type of plant may determine the effects of root excretions on denitrification (Woldendorp, 1962;Haider et al. , 1985, Qian and. In general, metabolically active roots contribute more organic matter and stimulate denitrification more than senesced roots (W oldendorp, 1963 ;Volz et al. , 1976). Artificially killed roots may contain more reserve carbohydrates available for decomposition than naturally senesced roots which may have transferred these carbohydrates to other plant parts .
All mesocosrns in my study contained plant roots, some of which were probably killed artificially during the course of the experiment. Dead roots do not actively excrete organic substances, thus the only C contribution from these roots would have been via decomposition. I was not able to quantify the extent of decomposition and mineralization of the roots found in the mesocosms. Thus, some of the N dynamics observed in this study may be the result of roots artificially killed rather than from patch dynamics.
Consequently, the groundwater N0 3 -removal rates presented by my mesocosm study do not represent in-situ mi crobiall y mediated groundwater N0 3 -removal rates. 63 The dry mass of plant roots was not significantly correlated with N0 3 -removal rates. This suggests that plant roots may not be the sole indicator of N0 3 -removal potential. Since these roots were found to be significantly correlated with patches of elevated organic matter within mesocosms, plant roots may play a role in the formation of such patches. Groundwater N0 3 -removal rates were significantly correlated with the total mass of patch material C suggesting that these patches of elevated organic matter may be "hotspots" of microbial activity. The percentage of patch material and the mass of patch material C in the subsoil may be useful indicators ofN0 3 -removal potential in the subsurface.
Although no significant differences in N0 3 -removal between forested and suburban areas were observed at the Charlestown site, two of the three Charlestown forested vegetated mesocosms contained portions of the Bh horizon, an illuvial zone enriched in organic matter. These mesocosms had the highest mass of C, but did not consistently have markedly higher N removal rates than other mesocosms at the same site.
This suggests that the Bh horizon did not markedly influence groundwater N0 3 -removal to a great extent.
Land use legacy may have also played a role in the difference in N0 3 -removal rates between sites. The Charlestown site was cultivated and received horse manure applications for approximately 20 years until the early l 980 ' s. Manure applications have been shown to increase the amount of organic matter in surface soils Nat. Res . Council, 1993 ;. In addition, cultivation processes have been found to translocate C substrates deeper into the soil profile . This previous cultivation and manuring may have contributed a greater mass of patch material C in the subsurface of the Charlestown site. Also, vegetation differences between sites may have influenced the intersite differences in N0 3 -removal. Unlike the North Kingstown site, the Charlestown site contained speckled alder trees. Alders have the capability of fixing atmospheric N. In addition, alders have been shown to produce higher amounts of nutrient-rich litter biomass annually and increase soil organic matter more than other tree species .
The differences in growing season hydrology between the two sites may be responsible for some of the differences in mass of patch material C and/ or observed groundwater N0 3 -removal rates. The mean 1996 growing season water table depth observed at the Charlestown site was approximately 30 cm higher than at the North Kingstown site. In 1997, the maximum depth of the water table varied over 50 cm between sites. The greater mass of patch material C in the subsurface at the Charlestown site may be due to lower rates of decomposition that occur under saturated conditions . In addition, ifthe upper portion of the C horizon at the Charlestown site remains saturated, as in the growing season of 1996, it is more likely to remain anaerobic. These conditions are favorable for denitrification .
However, data from the 1997 growing season indicates that the difference in hydrology between sites may not be significant in all years. Also, the mesocosms remained saturated throughout the groundwater dosing experiment so water levels did not directly influence nitrate removal rates in my study. Christensen et al. (l 990b) found that hotspots of microbial activity had a limited longevity, persisting from a few days to a few weeks. Outflow groundwater N0 3 -removal gradually declined over the course of my study; this suggests that any hotspots of microbial activity within my mesocosms also had a limited longevity or a gradual decline in their strength. My results may reflect artificially induced microbial activity . The existing roots within the mesocosms may have died and generated a pulse of C available for microbial processes. Bijay-Singh et al. (1988) suggested that denitrification rates are likely to be controlled by the mineralization of organic C within the soil.
The variability in N0 3 -removal rates between replicates within the four treatments suggests that the mesocosms in this study were not large enough or more replicate cores should have been taken to encompass the natural soil heterogeneity. The coefficients of variation ranged from 4% to 191 % for the four treatments . In all treatments, there was at least one mesocosm with relatively high groundwater N0 3 --N removal rates, i.e. greater

Further Studies
In order to gai n a complete understanding of the processes responsible for N03dynamics in my mesocosms, one cannot simply rely on a mass balance approach of waterborn N0 3 -. Complementary studies to my thesis will be undertaken to examine the mechanisms controlling N0 3 -removal within the mesocosms . In order to examine the role of denitrification and immobilization, the N added to these mesocosms was labeled with 15 1-N (5 atom% )N). To compute a comp lete N mass balance, gas and soil samples need to be analyzed for 15 N. Also, fresh samples of patches of elevated organic matter and matrix material from each site in this study will be examined for denitrification potential to determine the extent of variation in the microbial activity of patch and matrix material between and within sites.
There is also the need for research beyond the extent of my project. Additional studies need to be completed to explore further the role of mowed vegetated riparian areas in removing groundwater N03-. In particular, work is needed in riparian areas that are isolated from woody vegetation. In addition to vegetation management, it would be interesting to study the effect of other alterations, such as changes in soil structure and soil drainage, in riparian zones on water quality functions .

LITERATURE REVIEW
Pollution is recognized to be deleterious to human health and the environment.
Point source pollution originates at a discrete, identifiable location, such as a smokestack or pipe. This characteristic makes point source pollution easier to contain and regulate than non-point source pollution (NPSP) . Consequently, research interest has shifted to studying the mechanisms affecting the fate and transport of NPSP . Nitrate (N0 3 -) is an example ofNPSP that is discharged over large areas from diffuse sources. N0 3 -is the most commonly detected pollutant in groundwater wells in the US (US EPA, l 990) and is a federally registered drinking water contaminant. At high concentrations, N0 3 -can cause methemoglobinema (blue-baby syndrome) in infants. N0 3 -also has been linked to eutrophication of coastal waters and watershed management of nitrogen (N) is a major coastal issue Howarth et al ., 1996;.
N0 3 -enters groundwater from a variety of sources. Undisturbed and unmanaged forests are expected to contribute very little N0 3 -to g roundwater. However, forests may contribute high levels of 0 3 -to groundwater when a large disturbance, such as fire or clear cutting, occurs (Keeney,l 986). Agricultural fertilizer and manure applications have the potential to leach high level s of N0 3 -into groundwater (Keeney, l 986). Groundwater 0 3-can also originate at se ptic tanks (Keeney, l 986; Weiskel and Howes, l 992) . A study conducted in southern New England found that silage corn fields and septic systems contributed substantially more N0 3 -to groundwater than home lawns treated with fertilizer or forests . Other potential sources of N03-includ e atmospheric 75 deposition, industrial discharges, and animal feedlots .
Our understanding of watershed N is beset by uncertainties surrounding the role of N "sinks" -areas within a watershed capable of removing N from the groundwater.
Researchers studying coastal watersheds ranging from local scales (1000 ha) to regional scales (Mississippi Basin) have found that less than one-third of the estimated net anthropogenic N input is discharged to coastal waters Howarth et al , 1996). These studies suggest that we need additional research to improve our understanding of watershed sinks ofN.

Riparian Zones
Riparian zones are functionally defined as areas where energy and matter are exchanged between terrestrial and aquatic ecosystems    (Wright and Sautter, 1988).
Redoximorphic features are spots in the soil where iron has been reduced and reoxidized causing a color change in the soil (Singer and Munns, 1991 ). Redoximorphic features typically occur where the water table fluctuates between the growing and dormant seasons (Wright and Sautter, 1988) 76 Catenas are sequences of soils derived from similar parent material but categorized into different soil drainage classes (Wright and Sautter, 1988). There are seven possible soil drainage classes in complete soil catenas that range from upland locations toward waterbodies: excessively drained, somewhat excessively drained, well drained (WD), moderately well drained (MWD), somewhat poorly drained (SWPD), poorly drained (PD) and very poorly drained (VPD) (Wright and Sautter, 1988). Soil drainage classes roughly estimate the frequency and duration of soil saturation (Wright and Sautter, 1988). The excessively well drained class is dry for most of the year, and the VPD class has the water table at or near the surface for a substantial portion of the year. Redoximorphic features form closer to the ground's surface in PD and VPD soils. A riparian zone is generally composed of soils ranging from MWD to VPD soils.

Water-Saturation and Oxygen
As previously mentioned, the water table is near or at the ground's surface for a significant portion of the year in PD and VPD areas . In the summer when evapotranspiration is active, the water table in these areas may drop further below the surface. In PD red maple swamps in Rhode Island, the water table fluctuates on average 71.2 cm throughout the year .
Diffusion of oxygen into soils cannot occur when interconnected soil pores are absent or blocked with water (Singer and Munns,l 991 ). In saturated soils, the oxygen cannot be replaced, and anoxic conditions develop once the oxygen is consumed by soil organisms and plant roots. Therefore, PD soils are frequently anoxic since they experience long-term saturation. 77 Organic Matter Organic matter consi sts of living or dead plant and animal residues . In general, cultivated soils only contain 1-5% organic matter, most of which is isolated in surface soils . Decomposition is influence by a number of factors, including soil moisture. Under anoxic soil conditions, both microbial decomposition and plant growth slow down . As mentioned previously, the period of water saturation increases and DO concentrations decrease toward the PD and VPD areas of a soil catena. As a result, decomposition proceeds at reduced rates in these areas, and organic matter accumulates in these soils.

Vegetation
Vegetation type in an area is influenced by the site' s hydrology (Mitsch and Gosselink, 1993). From MWD to VPD soils in a riparian zone the hydrology changes dramatically which may account for the great diversity in vegetation cover contained within a single riparian zone. Vegetation in the PD and VPD sections of riparian zones needs to be tolerant of or have adaptions to withstand frequently flooded and anoxic conditions (Mitsch and Gosselink, 1993).

Riparian Zones as Sinks for Nitrate
Riparian buffer zones have long been accepted as a Best Management Practice (BMP) for soil and water conservation in agricultural areas by removing sediment from surface water runoff . Riparian zone functions are not limited to surface water remediation. There is a general consensus in the literature that riparian zones are a significant sink for groundwater N0 3 -in the subsurface, but many of the details surrounding the processes and mechanics of this N0 3 -removal remain elusive .

Mechanisms of Groundwater Nitrate Removal
Recent studies in riparian zones have tried to elucidate the mechanisms responsible for N03-attenuation . N03-is often considered conservative in groundwater , but several studies suggest that N0 3 -may be removed from groundwater by selected mechanisms in some aquifers. Some possible removal mechanisms include: 1) immobilization in microbial biomass, 2) dissirnilatory N0 3 -reduction to ammonium (DNRA), 3) plant uptake, and 4) denitrification . Plant uptake and denitrification are generally considered to be the dominant processes responsible for N0 3 -removal in riparian zones .

Immobilization and Dissimilatory Nitrate Reduction to Ammonium
During immobilization, microbes consume N0 3 -for cell growth and metabolism.

Plant Uptake
An important mechanism ofN0 3 -removal is plant uptake. O' Neill and Gordon (1994) examined N03-removal due to plant uptake by setting up a laboratory study with large boxes filled with either soil only or soil with trees planted. When pumping N0 3amended water through the boxes, more N0 3 -was removed from the boxes with trees than without any vegetation. Therefore, they attributed the N0 3 -removal to plant uptake.
Carrier proteins help transport N0 3 -across the root cell membrane (Hopkins, 1995). Once inside the plant, the N03-must be reduced to NRi +for assimilation and utilization (Hopkins, 1995). Uptake ofN0 3 -by plant roots is affected by I) low temperatures, 2) inhibitors of respiration and protein synthesis, and 3) anaerobic conditions (Hopkins, 1995). Plant uptake of N0 3 -may experience seasonal patterns with the least amount of plant uptake occurring in winter months (Hopkins, 1995). Plant uptake also declines as plants age (Hopkins, 1995). Nadelhoffer et al. ( 1995) demonstrated that uptake rates differ between species.
Since the primary rooting zone of many plants extends less than 60 cm into the soil (Harris et al. , 1977), plant uptake is not thought to be an important mechanism of N03removal from deep subsurface soils. However, Huang et al. (1996) found that switchgrass had the capacity to remove a substantial quantity of N0 3 -from soil at depths greater than 120 cm below the soil ' s surface. Even when plant uptake ofN03-occurs, the N may eventually be recycled back into the soil when plant parts senesce and decompose.
Plant uptake of N0 3 -from groundwater may not be constant. Forests may experience "N saturation" after extended peri ods of N uptake (Aber et al. ,l 989) . saturation occurs when availability ofN in the soil-water environment exceeds plant and microbial needs, not including denitrification (Aber et al., 1989). As a result ofN saturation, there may b~ an excess of N leached from mature forests. To avoid N saturation in forests,  recommend selectively cutting mature trees thereby allowing younger trees, which can utilize more N, to establish. However, N saturation may also be avoided in mature forests if excess N0 3 -stimulates microbiological activity in the soil (Nadelhoffer et al. , 1995). This greater microbial activity may increase the need for N03-in the microbial community, and N03-uptake may consequently resume.

Denitrification
Denitrification occurs under the same anaerobic conditions as DNRA. A study by Tiedje et al. ( 1981) found that the role of DNRA was inferior to denitrification. DNRA is thought to occur in areas dominated by long-term anaerobic conditions while denitrification may occur more frequently in areas where the DO status is transient (Tiedje et al. , 1981).
Of all the previous mechanisms of N0 3 -removal mentioned thus far, denitrification is the only mechanism that completely removes N0 3 -from the soil-water environment.
N03-is converted to N20 or N2 gas as follows : In order for denitrification to proceed, the following are needed : 1) N03-, 2) denitrifying bacteria, 3) an energy source (organic C or an inorganic electron donor such as sulfur), and 4) anaerobic conditions .

Nitrate
Perhaps the most basic requirement for denitrification is that N0 3 -be present in the soil to be reduced . A study comparing an enriched N site and a control site found that the enriched site had a greater potential for denitrification .

Bacteria
Denitrifying bacteria are essential for denitrification to proceed. In the absence of N0 3 -or anaerobic conditions, a portion of the denitrifying population is able to survive and then utilize N0 3 -once it becomes available again (Schipper et al., 1994).

Anaerobic Conditions
Denitrification only occurs under anaerobic conditions. Oxygen inhibits the formation of N oxide reductases required for denitrification reactions . Oxygen even inhibits the activity of preformed reductase (Knowles, 1981 ). In order for denitrification to proceed, the DO concentration in the soil needed to be less than 2 mg L· 1 in a study conducted by Gillham and Cherry (1978) . Parkin and Tiedje (1984) found an average two-to four-fold increase in denitrification rates in unsaturated soil as the 0 2 concentration was decreased from 20% to 5%.
The water content of soil exhibits great control over the amount of 0 2 that is able to diffuse into soil; for example, in wetter soils more pore spaces are filled with water, and less 0 2 can diffuse into these pore spaces (Singer and Munns, 1991 ). Pennock et al. (1992) found that topography influenced denitrification rates by controlling the volumetric water content of soil. Positions lower on the landscape tended to be wetter and thus had lower DO concentrations. Myro ld ( 1988) also found that field denitrification rates were 82 closely correlated with soil water content. Many studies have concluded that denitrification rates are higher in wetter soil drainage classes. For example,  detected the highest denitrification rates in PD surface soils in a North Carolinian study; Groffman and Tiedje (1989a) found that PD surface soils were consistently wetter and had higher denitrification rates than other soil drainage classes in Michigan; Parsons et al. (1991) observed higher denitrification rates in SWPD surface soils than WD surface soils in Kentucky; and Groffman et al. ( 1992) and  detected higher denitrifying enzyme activity (DEA) in wetland (PD and VPD) surface soils than in transition zone (MWD and SWPD) surface soils in Rhode Island . Groffman and Tiedje (1989b) concluded that soil drainage is a strong predictor of annual denitrification activity on a landscape scale.

Organic Carbon
Soil moisture also affects the amount of carbon (C) available for denitrification.
Under low 0 2 conditions, decomposition of organic matter is reduced resulting in an accumulation of organic matter . Denitrification has been found to be limited by the amount of available C in soil (Myrold and Tiedje, 1985;Hiscock et al. , 1991). When soils were incubated under non-nitrate limiting and anoxic conditions in a study by , they suggested that denitrification rates are likely to be controlled by the mineralization of organic C within the soil C limitation of denitrificati on was also demonstrated in a laboratory study of aquifer sediments from a golf course in Florida (Bradley et al. , 1992); denitrification rates in this study increased once glucose was added to the soil. The quality of the C source has been shown to be an important factor in 83 denitrification (Schipper et al. , 1994;McKenney et al. , 1995). A study by Schipper et al. ( 1994) found thaNienitrification increased more when fresh plant material was added to laboratory soil slurries than when senescent plant material, with less labile C, was added.
Also, different plant species contribute qualitatively and quantitatively different organic matter to soil. For example, soils treated with rye grass residues had greater potential to remove N03-via denitrification than soils treated with legume residues in a study by McKenney et al . ( 1995).
Denitrifying activity may be unevenly distributed throughout soil with the greatest activity occurring in uppermost horizons . In one study, DEA was isolated within the top l 0 cm of the soil profile, 80% of this DEA was contained within the top 2 cm of soil (Lowarance, l 992). Spatial variability of denitrification rates has often been explained by the distribution of C in soil. Beauchamp et al. ( 1980) examined A, B, and C horizons of three different soils and found the highest correlation between denitrification rates and total organic C.  reported that denitrification did not occur in deep aquifer sediments due to the declining availability of organic C with depth below the ground ' s surface McCarty and Bremner ( l 992) also observed that the number of denitrifiers decreased with depth; however, substantial populations still existed at a depth of 150-200 cm below the surface.
They hypothesized that little organic C was leached into the subsoil and available to denitrifiers since decomposition of plant derived organic matter proceeds so rapidly in surface soils. When glucose was added to subsurface sediments,  84 noted a substantial increase in denitrifying activity.

Seasonality
Denitrification is affected by soil temperature. Lowering the temperature decreases denitrification rates dramatically. This may be due to a lowering in microbe metabolic activity (Hiscock et al., 1991;Knowles, 1981). As a result, seasonality of denitrification may be expected. In a study of two different cropping systems in Oregon, field denitrification rates were correlated with soil temperature, with the highest rates in spring (Myrold, 1988).  also observed the highest denitrification rates in the spring and fall in Rhode Island surface soils. Lower rates recorded during warmer summer months in this study may not be attributed to soil temperature but to drier/aerobic conditions. However, a study conducted by Parsons et al . (1991) did not find a correlation between denitrification and soil temperature at two sites. Groffman and Hanson (1997) found inconsistent seasonal patterns over a number of years in another Rhode Island study and suggested that denitrification was not sensitive to changing soil temperatures.

Soil Texture
Soil texture may also influence denitrification rates (Beauchamp et al., 1980; Groffman and Tiedje, l 989a and l 989b ). Beauchamp et al. ( 1980) found that finer textured soils had higher denitrification rates since these soils generally had higher levels of available organic C. Groffman and Tiedje ( l 98 9a) observed the same trend . They attributed this trend to smaller pores in finer textured soils which hold water more tightly.
The soils therefore become anaerobic more readily. 85 pH Denitrification can be influenced by pH. The optimum range for denitrification to proceed is between 7 and 8 (Knowles 198 l ;Hiscock, 1991 ). At low pH, denitrification rates may be limited. pH also influences the end products of denitrification (Knowles, 1982). N oxide reductases are inhibited at low pH (around pH 4.0) so that the primary end products of denitrification under such conditions would be NO and N 2 0 , rather than N 2 (Knowles, 1982). However, Ottow et al. (1985) did not find that pH was correlated with the end products of denitrification. Bradley et al. (1992) also found no significant relationship between potential denitrification and pH in subsurface soils; they suggested this may be due to the narrow range of pH (5.4-6.4) observed at their site. They did find , however, that actual denitrification rates were affected by short-term changes in pH.

Influence of Plants on Denitrification
The role of plants in N0 3 -removal is not limited to plant uptake. Plants can also influence denitrification by a few processes: 1) since plants uptake less N03-is consequently available to denitrifiers, 2) roots excrete organic substances into the soil which may become substrates for denitrification, and 3) roots may contribute 0 2 to rhizosphere soil that may stimulate nitrification, the formation ofN03- ). In addition, elevated 0 2 consumption by roots and rhizosphere organisms may create anaerobic microsites in the so il that are favorable for denitrification (Woldendorp, 1963 ; .  found increased N0 3 -removal in the presence of roots. He suggested this was because roots generate C from sloughed-off root cap cells and root 86 exudates that may be available for microbial activity. The roots of healthy plants naturally shed as part a( their development processes, and roots are the primary source of organic matter in deeper soils . In a Liriodendron tulipifura forest, roots accounted for 70% of the total organic matter from above-and below-ground sources in the soil (Cox et al., 1978).
Roots also contribute organic C to the soil when they senesce and decompose.
Roots do not have a predetermined life-span, but root hairs have a relatively short lifespan, ranging from a few hours to a few months, before they are subjected to decomposition and become a source of available C to microorganisms . In the northern hardwood Hubbard Brook Experimental Forest, New Hampshire, Fahey and Hughes ( 1994) found the median longevity of fine roots to be 6 months . These roots also underwent rapid decomposition; 3 0% of these fine roots disappeared during the period from October to August. Root systems seem to produce localized zones in the soil with anaerobic conditions and ample C supplies thus creating a favorable environment for microbial denitrification activity.
Rhizosphere soil was found to have higher denitrification rates than the matrix soil, with the highest rates closest to the roots . Another study suggested that there was 90 times more denitrifying bacteria in the rhizosphere than in the soil matrix (Stefanson, 1972). However, the age and type of plant may determine the effects of root excretions on denitrification (Woldendorp, 1962). For example, a study by Haider et al. ( 1985) found that root-deposited C from wheat and corn plants was not readily decompo sable and therefore not readily used by denitrifying bacteria.  found that organic compounds derived from com roots benefited only 4-11 % of the soil rr'ncrobial biomass. Furthermore, a study by Sharnoot et al. (1968) found that only between 30 and 48% of rhizosphere-deposited organic matter was decomposed over 14 weeks of incubation.
There are also distinctions between senescent roots and metabolically active roots.
Senescent roots may be resistant to decomposition while active roots may provide less complex and more readily available forms of organic matter to the soil for microbial activity (Volz et al. ,197 6). For example, the most probable number (MPN) of denitrifiers was higher in flooded soil planted with barley than in a flooded fallow soil that contained senescent plant roots (Volz et al. , 1976). Woldendorp (1963) found denitrification was more stimulated by living roots than senescent roots. Whether roots died by natural or artificial means may also influence root effects on denitrification. Roots that have died naturally may have already recycled reserve carbohydrates to the plant. However, roots that are artificially killed may maintain reserve carbohydrates that can be readily excreted to and utilized by microbes in the soil .
Plants in soil may also influence the major products of denitrification . In soil without plants the main product of denitrification was N 2 0 while in soils with plants the main product was N 2 gas in studies by Stefanson ( 1972) and .
More work is needed to elucidate the importance of plant roots in inducing microbial processes of groundwater N03-removal.

Denitrification versus Plant Uptake
There is a great deal of uncertainty as to the relative role of plant uptake and denitrification in removing groundwater N0 3 -in riparian zones . In an ecosystem scale 15 N study in a mixed forest, Nadelhoffer et al. (1995) suggested that plant and microbial processes in surface soils have approximately the same potential ofremoving N0 3 -. In field experiments, it is extremely difficult to determine if N03-removal can be attributed to plant uptake, denitrification, or a combination of the two . In studies where N0 3 -removal was measurable in subsurface soils even though denitrification potential had been demonstrated to be low in these soils, plant uptake is usually assumed to be the dominant mechanism of N0 3 -removal Groffman et al. , 1992). However, these studies did not directly measure in-situ denitrification. When N0 3 -removal was observed in winter months when plant activity was expected to be low, denitrification was generally suggested as the mechanism of removal Groffman et al. , 1996).
Also, if groundwater was sampled below the major rooting zone, denitrification was generally assumed to be responsible for N0 3 -removal as in a study by . Verchot et al. (1997) conducted a field study of subsurface N cycling in grassed and forested vegetated filter zones in which they attempted to eliminate the plant uptake variable. To do so, they dug trenches along the edge of their plots to cut tree roots extending into the area of interest, weeded out saplings and herbaceous plants, and applied herbicide to prevent plant growth. Under these conditions, substantial N03-removal was observed and attributed to denitrification. More research needs to be conducted to determine the relative role plant uptake and denitrification play in removing N03-from groundwater.

Riparian Zones as Favorable Nitrate Removal Sites
Riparian zones possess many of the properties that favor N0 3 -removal from groundwater. In particular, the PD and VPD sections of riparian zones seem to be particularly favorable sites for denitrification . Since these sites are wetter, they tend to be anaerobic for a longer period of time and to have an accumulation of organic matter -conditions favorable for denitrification.
Riparian zones may be "pulsating ecosystems" similar to tidally flushed estuaries (Odum et al., 1995 ;Gold et al., in review) . In such pulsating ecosystems, final steady states do not develop in nature. Rather, spatial or temporal oscillations bring the system to a pulsing steady state. For example, during summer months in riparian zones, the water table drops and roots may extend deeper into the subsurface. These roots shed organic matter and may senesce generating a C source for anaerobic microbial activity to occur when the water table again rises in the dormant season. Therefore, a site favorable for denitrification has been created in this pulsating ecosystem.

Microsites of Microbial Activity
An interesting aspect of riparian zone studies is the discrepancy between companion field and laboratory microcosm studies of groundwater N0 3 -removal. As opposed to counterpart field studies, laboratory studies of shallow aquifer media have found lower denitrification rates and low if any correlation between denitrification rates and soils of different drainage classes (Pennock at al , l 992;Groffman et al , 1996) 90 Substantial laboratory-detected denitrification potential seems isolated within the top 1 O cm of soil ~Ambus  even though field studies with groundwater well networks have reported groundwater N0 3 -removal deeper in the subsurface Haycock and Pinay, 1993). Laboratory measured microbial activity may not adequately represent fieldmeasured activity because field studies encompass a greater extent of the natural soil heterogeneity . Differences in sample volume between field and laboratory studies may explain such discrepancies (Groffrnan et al. , 1996).
Organic matter is often distributed unevenly through soil profiles.  suggested that processes, such as denitrification, nitrification, and plant growth, that experience rapid transformations dependent on local conditions may be particularly influenced by sample volume. Parkin et al. (1987) suggest that the bulk of denitrification may occur in microsites or "hotspots" of organic matter. If these micro sites occur in irregular patterns, microsites of high activity may be missed in the small samples, i.e. less than 50 grams, typically used in laboratory analysis. Up to a point, the larger the sample volume, the greater chance of capturing these hotspots and the denitrification variability may tend to decrease  .  found that 25-85% of denitrification potential came from 0 .40 -0. 08% of the core ' s mass while a study by Christensen et al. (I 990a) traced the majority of denitrification activity to 8% of the sample' s volume. Measurements of microbial activity need to be conducted with methods that take into consideration the heterogeneity in soil and microbial populations (Murray et al. , 1995;. Appropriate sample sizes to reflect realistic field conditions 9 1 have been suggested, ranging from 5 cm diameter cores  to 10-15 kg soil samples . Recently, a laboratory mesocosm study (Gold et al. , in review; Jacinthe et al. , in review) of 12 kg of soil produced N03-removal rates more comparable to a field study than microcosm studies at the same site Groffman et al., 1996). In addition, PD mesocosms exhibited higher N0 3 -removal rates than MWD, the same trend observed in the field studies. When the mesocosms were disassembled, MWD mesocosms did not contain any dark-stained patches of elevated organic matter. However, such patches were found in the PD mesocosms that exhibited high N03-removal rates. These patches appeared to be root-derived biological hotspots of microbial activity.

Nitrate Removal in Forested and Non-Forested Riparian Zones
Most recent riparian zone research has focused on forested areas Groffman et al. , 1996;. Throughout this work, it is generally accepted that forested riparian zones have the ability to serve as N0 3 -sinks. Fewer studies have examined the role of non-forested riparian zones in removing N0 3 -from groundwater (Haycock and Pinay, 1993 ;Osbourne and Kovacic, l 993 ;. In general, trees are thou ght to uptake more groundwater N03-than grasses because they have a potentially deeper rooting zone . Haycock and Pinay (1993) compared o,removal in poplar and grass riparian zone subsoils in winter.
The poplar site removed more groundwater N0 3 -. They suggested that C limitation in the grass site was responsible for the lower N03-removal . A study by Osbourne and Kovacic (1993) also found that a forested riparian subsoil removed more groundwater N0 3 -than a grassed riparian subsoil on an annual basis. They suggested that denitrification was the mechanism ofN0 3 -removal which was limited by the form of the C source.
However, other studies have shown greater N0 3removal in grassed riparian zones than forested riparian zones. It has been suggested that C from grassed areas may be more readily available to microbes as compared to the C from forested biomass . In general, prairie soils contain twice as much organic matter as similar forested soils (Plaster, 1997). Also, organic matter extends deeper into prairie soil where grass roots may decay as opposed to forest soils where most of the soil organic matter originates from surface litter (Plaster, 1997). Even different grass species may contribute different amounts of available C from their roots (Groffman et al. , 1991 ).  found greater denitrification rates in subsoils of a grassed riparian area than a forested riparian area in a paired comparison study. Lower denitrification rates in the wooded area were attributed to C limitation.  conducted a study in a restored riparian wetland . In this study, the area closest to the stream was replanted with trees while an outer fringe of herbaceous vegetation remained . They concluded that shallow groundwater denitrification rates in grassed riparian buffer zones could equal or exceed denitrifi cation rates in restored buffer zones recently replanted with trees. More studies are needed to determine whether there are any differences in groundwater N03removal rates between forested and non-forested riparian subsoils.
93 Nutrient Enrichment of Soil Hendricks et al. ( 1993) concluded in their review of fine root research that fine root turnover rates are likely to increase as site quality, i.e. N availability, improves.
These roots decompose readily releasing additional N and C into the soil. Therefore, the high site quality is further perpetuated. More research is necessary to explore fine root dynamics .

N-fixing Plants: Alders
N cycling under high-N environments, as in soils planted with N-fixing plants, is of particular interest to research on water quality functions of riparian zones. Alder (A/nus sp .) is an example of a genus that forms root nodules and is capable of fixing N from the atmosphere. Beaupied et al. (1990) found that soil mineral N only contributed to 6% of black alder (A/nus glutinosa) annual N requirements, the remainder corning from atmospheric N . Flooding can limit N-fixation in alders since this process requires aerobic conditions (Winship and Tjepkema, 1985). However, Winship and Tjepkema ( 1985) have suggested that red alder (A . rubra) root nodules have "compartments of low permeability" that enable some parts of the nodule to remain aerobic for a longer periods of time and thus continue to fix N despite saturation.
Since ancient times, alder has been recognized as improving soil fertility due to their ability to fix N (Yamaya, 1968). Yamaya (1968) studied the role of A. inokumae leaves in soil N cycling. In this study, the N content of alder leaves was found to be higher than that of many other broad-leaved and needle-leaved deciduous trees . These leaves were also found to decompose more rapidly than Japanese cedar leaves. This rapid 94 decomposition may be due to soil fauna being attracted by the high N content. Therefore, alder leaves may recycle more N back into the soil. However, alders also grow very rapidly (Y amaya, 1968). This rapid growth may correlate with a higher rate of N cycling in alder forest soils than other forest soils. As a result, the N content of alder forest soils may not be substantially higher than other forest soils. When alder forests are cut, however, there may be a substantial release of N into the soil (Y amaya, 1968).
The role of alder roots in N dynamics has also been studied. Roots of N-fixers may decompose at different rates than roots of plants that do not fix N. Black alder roots with higher N concentrations decomposed at a slower rate than black alder roots with lower N concentrations in a study by Camire et al. (1991). They suggested that N-lignin derivative compounds, resistant to decomposition, may have formed in these N-rich roots .
Denitrification in soils planted with alders has been explored in a few studies. Binkley et al. (1992) did not find denitrification to be a substantial sink of N0 3 -in red alder forests. As in other forested soil studies , the highest denitrification activity was found within the top few cm of the surface soil in an alder stand (Struwe and Kjoller, 1990). Denitrification rates were found to be low overall, but denitrification activity was seasonal with highest rates in the spring and summer months (Struwe and Kjoller, 1990). Duff and Triska ( 1990) observed substantial denitrification in shallow sediments in the hyporheic zone under a red alder riparian zone. Mander et al. (1997) compared the ability of two grey alder (A . incana)/wet meadow riparian zones to remove N from groundwater. They found that even though one site had greater N inputs (adjacent to a pig farm) , the output N concentrations were comparable.
This high N removal capacity, between 67 and 96% N removed, was attributed to high rates of denitrification as well as high rates of plant N uptake in both alder stands.

Agricultural Practices
Cultivation  compared denitrifi.er populations between a tallgrass prairie grassland and a cultivated field . They found that denitrifier populations in the cultivated field did not decrease with depth as in the grassland . They suggested that this trend was due to the cultivation process translocating C deeper into the soil profile.
Therefore, cultivation alters the surface and subsurface sediments by creating a more favorable denitrifying environment.

Manure Applications
Manure applications to farm land serves two purposes: l) it is a nutrient source that can replace fertilizer applications and 2) it is way for farmers to dispose of this waste product . Manure applications supply nutrients and organic matter to soil (Nat. Res. Council, 1993 ;. In a study of surface soils that received manure applications for 15 years, the manure treated soil had a greater organic matter content than control soils-3% more organic matter in fine sandy loam soils and over 25% for silt loam and clay loam  . This increase in organic matter may promote denitrification, as evidenced in a study by  where surface soils were treated with fermentation waste. 96 Alternatively, manure may not contribute to the long-term organic matter content of soil because a large portion of the C may be converted to C0 2 gas (Simpson, 1986). Lovell and Jarvis ( 1996) found that the application of cattle dung pats on pasture soils did not substantially impact the size or activity of the soil microbial biomass under field conditions. The extent to which denitrification is affected by manure applications depends on the manure composition which is influenced by many factors, such as the type of animal, type of feed the animal received, manner of manure storage, and extent of decomposition in this manure (Paul and Beauchamp, 1989;Nat. Res . Council, 1993). For example, sheep and poultry manure contains high N contents (23-28 lbs ton-1 ) while manure from cattle, horses, and pigs contains low N contents (I 0-13 lbs ton-1 ) (Plaster, 1997

Core Extraction
The core extracting apparatus consisted of multiple parts (Fig. 10). It began with a 2-section extendible metal pipe mounted on a metal plate that rested against one wall of the pit. At the other end of the extendible pipe, there was another metal plate on which we attached an 8 ton hydraulic bottle jack. The jack had a metal plate attached to its extension arm that supported the core ( 15 cm diam., 40 cm long Schedule 3 5 PVC pipe with one beveled end) . We rested this entire apparatus on two scissors jacks.
We randomly selected the side of the pit from which to extract the core. During the extraction procedure, we set the center of the core at least 3 0 cm below the dormant season water table level of the area. At the North Kingstown site, we extracted co res at a depth of 67 .5 cm (midpoint of the core) from the surface. We extracted the Charlestown cores at 57 .5 cm (midpoint of the core) from the su1face because an elevated water table   99 prevented deeper extraction at this site. Once we adjusted the scissors jacks to the appropriate height and leveled the apparatus, we began to press the beveled end of the PVC core horizontally into the wall of the pit by pumping the bottle jack. Once the jack' s arm was fully extended, we stabilized the PVC pipe by placing blocks of wood below it.
Then, we brought the extendible metal pipe out to another extension and reestablished the apparatus. As the core was pushed into the wall of the pit, we excavated the soil around the top and sides of the core with spades in order to relieve pressure on the core and prevent it from bending. We continued this procedure until the back end of the core was flush with the side of the pit.
When pressing of the core was complete, we placed blocks of wood below the core to ensure its stability while we removed the extracting apparatus from the pit. We secured the back face of the core with a sheet of plastic and duct tape. We excavated the core out of the side of the pit by digging around its top and sides and beyond its front end .
A piece of wire was used to slice the core cleanly away from the in-situ soil. We carefully lifted the core out of the pit, seated it vertically upright, and secured the other end with a sheet of plastic and duct tape. We brought the cores back to URl and stored them at 4 oc.
As mentioned previously, the water table rose in midsummer which made it necessary for us to extract the cores at that Charlestown site at 57 .5 cm rather than at 67.5 cm (midpoints of the core) as at the North Kingstown site. Since the water table at the Charlestown site only dropped to around 45 cm below the surface, we needed to use a pump to remove excess water out of the pits as we dug the pit and extracted the cores.

Mesocosm Set-up
My mesocosm study was not intended to replicate natural, undisturbed field conditions. While I attempted to minimize the disturbance of soil sediments, the state and dynamics of the roots and other organic matter contained within the cores cannot represent natural field conditions over l l 6 days . Therefore, my results are not equivalent to in-situ rates of groundwater N0 3 -removal from the respective sites. This method should, however, allow us to learn more about N0 3 -removal rates in the absence of plant uptake. In addition, mesocosm results should be more comparable to field conditions than microcosm studies upon which much of our insights into nitrogen dynamics of riparian zones are based.
lOl Figure 13 . Schematic of horizontal core extraction. The cores were extracted from depths of 57 .5 cm and 67.5 cm (midpoint of l 5 cm diam. core) at Charlestown and North Kingstown riparian study sites, respectively .