JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (6)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rutherford, J. C.
Right arrow Articles by Nguyen, M. L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rutherford, J. C.
Right arrow Articles by Nguyen, M. L.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Rutherford, J. C.
Right arrow Articles by Nguyen, M. L.
Related Collections
Right arrow Wetlands and Aquatic Processes
Right arrow Nutrients
Right arrow Solute Transport Models
Right arrow Nutrient Management
Right arrow Nitrogen
Published in J. Environ. Qual. 33:1133-1143 (2004).
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA

TECHNICAL REPORTS

Wetlands and Aquatic Processes

Nitrate Removal in Riparian Wetlands

Interactions between Surface Flow and Soils

J. C. Rutherford* and M. L. Nguyen

National Institute of Water and Atmospheric Research, P.O. Box 11-115, Hamilton, New Zealand

* Corresponding author (k.rutherford{at}niwa.co.nz).

Received for publication May 14, 2003.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Riparian wetlands containing springs are thought to be ineffective at removing nitrate because contact times between the upwelled ground water and the underlying microbially active soils are short. Tracer experiments using lithium bromide (LiBr) and nitrate (NO3–N) injected at the surface were used to quantify residence times and NO3–N removal in a riparian swale characteristic of New Zealand hill-country pasture. An experimental enclosure was used with collecting trays at the downstream end to measure flow and concentration, shallow wells to measure subsurface concentrations, and an array of logging conductivity probes to monitor tracer continuously. The majority of added tracer reached the outlet more slowly than could be explained by surface flow, but more quickly than could be explained by Darcy seepage flow. There was evidence from the wells of tracer diffusing vertically to a depth of at least 5 cm into the surface soil layer, which was permanently saturated and highly porous. During dry weather 24 ± 9% of added NO3–N was removed over a distance of 1.5 m largely by denitrification. The net uptake length coefficient for this wetland (K = 0.08 ± 0.03 m–1) is slightly higher than the range (K = 0.01–0.07 m–1) measured in a small stream channel infested with macrophytes. Nitrate removal is expected to decrease with increasing flow. Seepage flow is estimated to have removed only 7 ± 4% of the added NO3–N and we hypothesize that vertical diffusion substantially increases NO3–N removal in this type of wetland. Riparian wetlands with springs and surface flows should not be dismissed as having low NO3–N removal potential without checking whether there is significant vertical mixing.

Abbreviations: DEA, denitrification enzyme activity • SG, specific gravity


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
IN MANY COUNTRIES nitrate runoff from farmland contributes to problems of eutrophication and exceedance of standards in drinking water. Nitrate loading to streams from pastoral farming is of particular concern in New Zealand where livestock are grazed outdoors throughout the year and losses from urine patches, applied nitrogenous fertilizers, and clover–ryegrass pasture are high (Wilcock et al., 1999). In addition, nuisance growths of periphyton and phytoplankton in streams, lakes, and estuaries are limited by nitrogen in some parts of the country (White and Payne, 1977). Appropriate land-use management practices such as targeting optimum fertilizer application rates to critically generating source areas, and especially the provision of riparian buffer zones to intercept runoff from farmland, have the potential to reduce nitrate loads to streams.

It is well documented that riparian soils can significantly reduce ground water NO3–N concentrations, principally by microbial denitrification (e.g., Gilliam, 1997). High rates of NO3–N removal occur where ground water comes into contact with anaerobic, organic soils (Hill, 1996; Fennessy and Cronk, 1997). For example, in a small headwater stream draining pasture in Scotsman's Valley, New Zealand, the annual NO3–N efflux was only 5% of that leached below the root zone. The NO3–N efflux to the stream during base flow was only 2 to 68% of the ground water influx to the riparian zone, indicating 32 to 98% removal largely by denitrification in the riparian organic soils (Cooper, 1990).

There is, however, high variability in the reported ability of riparian buffer zones to reduce nitrate loads to streams. Even when soil conditions favor a high rate of denitrification, there must be a sufficiently long soil–water contact time before a significant reduction occurs in NO3–N concentration (Hill, 1991). Low removal can occur where high-nitrate water bypasses microbially active wetland soils either by flowing in deep ground water under the riparian zone (Burt et al., 1999); in mole, tile, or pipe drains through the riparian zone (Lowrance et al., 1984); or as surface water across the top of the wetland (Gold et al., 2001).

Along the banks of low-order streams draining hill-country pasture in New Zealand it is common to find permanently wet swales, hereafter termed "riparian wetlands." They range in size from 1 to 1000 m2, are well vegetated (with pasture grasses, sedges, and rushes), and are usually grazed by stock (especially in summer). Investigations in such wetlands have found that the near-surface soils are very fragile and organically enriched, redox potentials are low, and rates of denitrification are high (Cooper, 1990; Nguyen and Downes, 1997; Matheson et al., 2002). In one such wetland Burns and Nguyen (2002) injected tracer (LiBr plus NO3–N) at a depth of 10 to 20 cm and monitored its movement in an array of down-slope wells. The measured seepage velocity at this depth was low (7.5–30 cm d–1), the measured soil denitrification enzyme activity (DEA) was high (5.7 ± 1.8 mg kg–1 h–1), and the observed NO3–N removals were >90% and >99% over distances of 60 and 100 cm, respectively. Clearly, substantial reductions of nitrate concentration can occur in subsurface flow within such wetlands.

However, these riparian wetlands usually occur in flow convergence zones of a catchment and hence a disproportionately large fraction of the total ground water and surface flow from the catchment passes through them (Cooper, 1990). There are several springs at the head of the wetland studied by Burns and Nguyen (2002), water is visible at the surface most of the year, and there is appreciable surface flow, especially after rain. Hill (1996) found minimal reductions in nitrate concentration where water emerged as springs and flowed across the surface of a wetland. This led Rosenblatt et al. (2001) to classify riparian zones with springs as likely to have low nitrate removal potential. While surface flow clearly has the potential to convey high-nitrate water across the riparian zone, there is very little published information about the extent to which surface flow interacts with the underlying wetland soils and the effects this has on nitrate removal.

In this study we conducted three experiments in the same riparian wetland studied by Burns and Nguyen (2002) in which we injected tracer at the surface to mimic the behavior of upwelling spring water. The objectives were to (i) measure the residence time distributions of conservative tracer using an array of conductivity probes (Experiments 1–3), (ii) measure the removal of nitrate tracer injected at the surface (Experiment 3 only), and (iii) infer the principal flow pathways in nitrate removal.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Study Site
The streamside wetland studied was located in steep (10–30°) hill-country sown with ryegrass (Lolium perenne L.)–clover (Trifolium repens L.) pasture and grazed by sheep and cattle. The catchment (area = 1.3 ha) is part of a research farm situated at the Whatawhata Research Station west of Hamilton, New Zealand (37°48' S, 175°5' E) (Fig. 1) . The climate is humid–temperate with mean annual temperature of 13.7°C and rainfall of 1614 mm. The catchment is predominantly Waingaro steepland soil (a northern yellow brown earth, classified Umbric Dystrochrept; USDA, 2002) derived from sedimentary greywacke parent material. There is a shallow (50–75 cm depth) clay loam topsoil of fine and medium nut structure, underlain by a subsoil of firm clay with weakly developed nut structure (Bruce, 1978).



View larger version (13K):
[in this window]
[in a new window]
  		Fig. 1. Map of the North Island, New Zealand, showing the location of the study site.

 
The wetland was a permanently wet swale with a surface area of 350 m2 and a slope of 8 to 9° (Fig. 2) . It filled the small valley at the bottom of the hillslope, and appears to have been formed by the accumulation of sediment and organic matter washed in from the upland pastoral catchment that had been partially stabilized by vegetation. During the study the wetland was well vegetated with floating sweetgrass (Glyceria declinata Brébiss.), rush (Juncus spp.), sedge (Carex spp.), and lotus (Lotus pendunculatis Cav.). A series of 20 auger holes (maximum depth 1 m) indicated that the top 20 to 30 cm was dark brown-black, organically enriched fine clayey textured soil (much finer than the upland soils) containing plant roots, twigs, and occasional tree branches. The surface soils were very poorly consolidated and could not support the weight of a person, so boardwalks were built to provide access and minimize soil disturbance during sampling.



View larger version (27K):
[in this window]
[in a new window]
  		Fig. 2. Plan views of (a) the catchment, (b) the wetland, (c) elevations of four transects across the wetland, and (d) elevation of XS6 showing the approximate depth of the clay layer surveyed after the surface soils were washed out.

 
At a depth of 50 to 60 cm there was a transition to bluish-gray clay that became increasingly consolidated with depth (see Fig. 2). Shortly after this study the top 50 cm of wetland soil was scoured out during a storm down to the clay layer; behavior that has been observed in several adjacent wetlands. There was clear evidence of several springs near the head of the wetland and their location was confirmed after the surface soils were washed out. The wetland surface was moderately even (probably because of periodic stock grazing until 1999 when the wetland was fenced). Water from the springs rose to the surface, spread out across the wetland, and moved down-slope either over the surface or in shallow, subsurface seepage in the poorly consolidated topsoil. In dry weather a small amount of surface flow was visible underneath the grass. Typically the water depth was 1 to 3 mm and occurred in poorly defined preferred flow paths (microchannels) around root mounds that occupied 10 to 30% of the total surface area. When it rained, surface flow increased within a few minutes, rapidly spread across the entire wetland, increased to a depth of 5 to 10 mm, and persisted for about 12 h.

Experimental Design
A single rectangular patch of wetland soil was isolated from lateral flow within an enclosure. Collection trays were fitted at the downstream end of the enclosure to collect and measure the outflow at two depths. The upstream end was left open to allow inflow from higher up the wetland to enter the enclosure. Three surface injections of LiBr tracer were made at the upstream end of the enclosure during autumn (April 2001). There was a gap of 2 to 3 wk between injections, which was sufficient for outlet conductivity to return to background. Conductivity and flow were monitored continuously in the outlet and the percentage recovery of the added tracer was estimated by comparing conductivity "input" and "output" (as detailed below). The conductivity versus time profile at the outlet was used to determine the residence time distribution of tracer within the enclosure. An array of piezometers (slotted to admit water from depths of 5–15 cm) was installed within the enclosure to detect any tracer mixing into the wetland soils. Flow rates were steady during each of the three experiments, but were varied between experiments to give an indication of the effects of flow on tracer residence time. In one of the three experiments nitrate was added to the LiBr tracer and samples were collected at the outlet to estimate nitrate removal.

The enclosure (240 cm long by 106 cm wide, surface area 2.5 m2) was created by embedding plywood sheeting 60 cm into the ground (i.e., into the clay layer). The upstream end of the enclosure was left open. Steel collectors (boxes open on the upslope side) were driven horizontally into the wetland soil at the downstream end of the enclosure (150 cm from the tracer injection point), sealed, and clamped against the plywood sides (see Fig. 3) . The upper box collected all the surface flow plus the horizontal subsurface flow from the top 0- to 5-cm soil layer. The lower box collected subsurface flow from the 5- to 15-cm soil layer plus any leakage arising from imperfect seals between the upper box and the sides of the enclosure. The flow rate from each collector was measured separately using two tipping-bucket gauges fitted with Hobo event recorders (Onset Computer Corporation, Bourne, MA).



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 3. Sketch of the experimental enclosure. The diffuser was 150 cm upslope from the collector boxes.

 
Lithium bromide was used as the conservative tracer (Hill, 1996) because Br is not readily incorporated by soil microorganisms or utilized by plants (Simmons et al., 1992) and it has a very low background concentration in the wetland studied (Burns and Nguyen, 2002). Sufficient LiBr was added to increase conductivity above background (here by a factor of 10), but care was taken to minimize possible buoyancy effects associated with using a tracer that is denser than the surrounding surface or ground water. Stock LiBr solution (specific gravity [SG] = 1.39, conductivity = 413 mS cm–1) was diluted with tap water to give 5 to 10 L of tracer (SG = 1.03–1.06, conductivity = 36–67 mS cm–1). Approximately 20 mL of dye (Rhodamine WT) was added to enable visual observations of surface flows. Wetland vegetation was trimmed to a height of 1 to 2 cm to aid visual observations of surface flow. Tracer was discharged onto the wetland surface at the upstream end of the enclosure over 20 to 30 min using a Marriott bottle (Experiments 1 and 2) or peristaltic pump (Experiment 3). A diffuser (15-mm-diameter PVC pipe, 100 cm long drilled with 10 evenly spaced 2-mm holes) distributed tracer evenly across the width of the enclosure.

In all three experiments we used an array of conductivity probes to measure the residence-time distributions of tracer. Conductivity is a nonspecific measurement since it can change as a result of changes in the concentration of any ions in the wetland, and not necessarily just as a result of the added tracer. Nevertheless, with an array of logging conductivity probes, spatial and temporal patterns of tracer movement can be monitored more easily than using traditional sampling and laboratory analysis. In each collector box and below each flow gauge conductivity was logged every 5 min using conductivity probes (Model 247L; Campbell Scientific, Logan, UT) connected to data loggers (Campbell CR10). Conductivity was also logged in each of six piezometers (three each at equal distances across the channel 75 and 100 cm from the injection point). Piezometers were made from 25-mm-diameter PVC pipe slotted to admit water from depths of 5 to 15 cm. They were inserted into augered holes 25 cm deep, back-filled with sand, and the top 5 cm was plugged with bentonite to minimize the ingress of surface water. Because conductivity is only semiquantitative, we also analyzed water samples for Li and Br during one experiment (Experiment 3).

In Experiments 1 and 2 the tracer contained only LiBr, but in Experiment 3 nitrate was also added as KNO3 (final concentration = 1 g NO3–N L–1). In addition to logging conductivity, water samples were collected below the flow gauges every 30 to 120 min using ISCO (Lincoln, NE) automatic samplers. Samples were filtered to remove suspended material and subsequently analyzed for Li and Br both by inductively coupled plasma emission spectroscopy–mass spectrometry and NO3–N by a cadmium column reduction method after complexation with 1-napthyl-ethylenediamine followed by analysis on an autoanalyzer (American Public Health Association, 1998). The change in NO3–N to Br ratio and the difference between input and output NO3–N and Br mass were used to estimate the rate of nitrate removal (Schnabel et al., 1995).

There was no rain during each of the periods for which results are reported and flows were steady, although rain caused the early cessation of Experiment 3. Flows varied between experiments as a result of antecedent rainfall and for Experiment 2 because an upstream spring was diverted into the enclosure.

Four soil cores (30–40 cm long) collected just outside the enclosure were returned to the laboratory and cut into three equal lengths. Cores were held vertically, a constant head of water was maintained above the soil, the outflow was measured over 12 to 24 h, and the saturated hydraulic conductivity was determined from:

[1]
where K = saturated hydraulic conductivity (cm d–1) (saturated is omitted hereafter for brevity), Q = total volume of water (cm3) passing through soil column over time period t (d), A = cross-sectional area of the column (cm2), L = length of soil (cm), and H = depth of overlying water (cm). Soils were then extruded and oven-dried at 108°C for 48 h to determine porosity. The horizontal seepage flow in each soil layer within the wetland was estimated using Darcy's Law:

[2]
where Qi = seepage flow in layer i (cm3 s–1); s = slope; B = channel width (cm); Ki = average hydraulic conductivity in layer i (cm s–1); and Hi = thickness of layer i (cm). Pore water velocity was also estimated using Darcy's Law:

[3]
where vi = pore water velocity in layer i (cm s–1) and pi = porosity in layer i. Fresh soil samples from the same four locations from depths of 0 to 10 and 10 to 15 cm were bulked by depth, passed through a 5-mm sieve to remove twigs, and analyzed for denitrification enzyme activity (DEA) (Tiedje et al., 1981).

To determine how much inert tracer was recovered at the outlet and how much nitrate was removed, it was necessary to compare the quantities injected with the quantities leaving the enclosure taking into account any measurement uncertainties. Uncertainty in any measured variable (e.g., flow, concentration or conductivity) or derived quantity (e.g., increase above background or mass) is quoted as either the standard deviation (SD) or the coefficient of variation (CV; i.e., SD/mean). Where two variables were subtracted (e.g., concentration minus background), the variances (squared SD) were added. Where two variables were multiplied or divided (e.g., flow times concentration, or output/input) the squared CVs were added. The mass of NO3–N, Li, and Br applied (hereafter termed "input") was the product of the volume discharged and tracer concentration. The volume discharged was affected by spillage, air blockage in the pipes, and tracer remaining in the container and was estimated to have a SD of 15 mL. The SD of tracer concentration was determined by analyzing three replicates. The mass leaving the enclosure (hereafter termed "output") was the time integral (over the period when concentrations exceeded the background level) of the product of flow and outlet concentration less background. The volume delivered by the tipping buckets varied as a result of uneven friction in the pivot and splash and averaged 1650 ± 80 mL (CV = 4.8%). Timing errors were negligibly small and the CV of flow was taken as 4.8%. Conductivity "input" and "output" were estimated in the same way as for NO3–N, Li, and Br. They have the units L mS cm–1, are directly analogous to mass, and were used to assess tracer recovery. Conductivity probes were calibrated in the laboratory over the range observed in the field (0.1–4.2 mS cm–1) and the CV was found to average 3.1% (calibration error). Background conductivity in the field was the average measured for 30 min before injection and for 3 to 6 h after concentrations had returned to a steady level. It was found to have a SD averaging 0.005 mS cm–1 (background error). The uncertainty in conductivity "output" was estimated by adding the squared CV for flow and conductivity. The uncertainty in background affected not only the "corrected" conductivities but also the time period of integration. To account for this an extra term was added to the "output" uncertainty: the product of the time period of the integration, the average flow, and the SD in background conductivity.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The water level in the piezometers was noted during each inspection and found to lie at, or just below, the level of the surrounding soil. Thus there was no evidence that spring water upwelled within the enclosure.

Experiment 1 was conducted at low flow (Table 1) when surface water was 1 to 3 mm deep and confined to two or three microchannels that occupied about 30% of the enclosure width. Dye was observed to travel along these microchannels and first reached the outlet 5 to 10 min after injection. Over the next 30 min dye spread across most of the enclosure width and remained visible in both the enclosure and the top collector for several hours. Outlet conductivity started to increase 10 min after injection (Fig. 4) at much the same time as dye arrived. The tracer injection rate was uneven because of an air blockage in the tubing that was cleared as soon as it was noticed. The blockage gave rise to two minor peaks in outlet conductivity during the first 60 min of the experiment (Fig. 5) . These minor outlet peaks occurred 10 to 15 min after the injection pulses that caused them, which implies that the mean velocity of surface flow along the 1.5-m enclosure is 144 to 216 m d–1. The main peak and the centroid for conductivity occurred after 2.6 and 6.4 h, respectively (Table 2). Conductivity exceeded 25% of the peak for 7 to 8 h and did not return to background levels for >36 h. The outlet conductivity profile was markedly skewed with a very long "tail" (Fig. 4). Of particular importance is the fact that the outlet conductivity centroid occurred about 6 h after injection stopped. This indicates a residence time for the bulk of the added tracer significantly longer than expected given that the enclosure was 1.5 m long and that surface flow had a mean residence time of only 5 to 10 min. It can be inferred that much of the injected tracer was delayed within the enclosure.


View this table:
[in this window]
[in a new window]
  		Table 1. Summary of experiments.{dagger}

 


View larger version (19K):
[in this window]
[in a new window]
  		Fig. 4. Conductivity at the outlet (solid) following surface injections of LiBr. Also shown is mean and maximum (dashed) conductivity in piezometers 5 to 15 cm deep. Background conductivity has been subtracted.

 


View larger version (21K):
[in this window]
[in a new window]
  		Fig. 5. Tracer injection rate (thin) and outlet conductivity (thick) during the first 90 min of Experiments 1 and 2. Quick flow is separated by a straight line (dash) drawn from the time when tracer was first detected at the outlet to the point of inflexion after the peak. Flow was low during Experiment 1, but was increased for Experiment 2 by diverting an upstream spring.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Summary of tracer conductivity measurements.

 
Experiment 2 was conducted at steady, high flow (Table 1) when surface flow was 5 to 10 mm deep across the entire enclosure. The injection rate was higher during the second part of the injection because of a partial air blockage in the tubing, and this variation was mirrored by the outlet conductivity profile (Fig. 5). The plateau and peak in outlet conductivity occurred 6 to 8 min after the injection pulses that caused them. The peak outlet conductivity occurred sooner in Experiment 2 (20–25 min) than in Experiment 1 (2.6 h) and was significantly higher than subsequent conductivities (Fig. 4).

Experiment 3 was conducted at a similar flow to Experiment 1 (Table 1) and the outlet conductivity profile during the first 32 h (at which time the experiment was stopped because of rain) was similar to that observed in Experiment 1 (Fig. 4).

During all three experiments, conductivity in the piezometers increased soon after tracer injection and returned to background between experiments (Fig. 4). There was considerable variation between individual probes but conductivity typically peaked after about 6 h and took >36 h to return to background.

For Experiments 1 and 2 there was no significant difference between the conductivity input and output, within the experimental uncertainties of the measurements (Table 1). Assuming that the changes in conductivity above background are attributable to the added tracer, then it can be concluded that all of the added tracer left the enclosure via the two collectors, and that losses to deep ground water and/or retention of conservative tracer within the wetland soils were negligible. Since Experiment 3 was stopped 32 h after the tracer injection, at which time conductivity was still significantly higher than preinjection background levels, "conductivity output" was lower than the "conductivity input" (82 ± 7% recovery). The Li and Br concentrations were also higher than preinjection background concentrations when the experiment was stopped. The Br recovery (83 ± 9%) was similar to that for conductivity, but Li recovery (93 ± 9%) appeared to be somewhat higher for reasons that are not clear.

The top 10 cm of soil was very poorly consolidated and contained numerous grass roots, decaying plant fragments, and reddish flocks of precipitated iron. It had low bulk density, high porosity, and high hydraulic conductivity (Table 3). Soils became more consolidated with increasing depth. An exponential regression model was fitted to the hydraulic conductivity results (Fig. 6) , giving:

[4]
where K(y) = hydraulic conductivity (cm d–1) and y = depth (cm) below the surface. The root mean square error was 56%.


View this table:
[in this window]
[in a new window]
  		Table 3. Summary of measured soil physical properties.{dagger}

 


View larger version (12K):
[in this window]
[in a new window]
  		Fig. 6. The variation of saturated hydraulic conductivity with depth in wetland soils. Measurements were made in the laboratory using soil cores. Error bars are ±1 SD. The fitted line (see Eq. [4]) has a root mean square error of ±56%.

 
Seepage flow was estimated in each 1-cm-thick soil layer using Eq. [2] and [4]. Predicted seepage flow decreased with depth and in the top 30 cm totaled 0.34 ± 0.19 cm3 s–1 (Table 4). Estimated seepage flow was 6 and 8% of the measured outflow during Experiments 1 and 3 (conducted at low flow), but only 1% of outflow during Experiment 2 (conducted at high flow). Estimated pore water velocity was low, even in the 0- to 5-cm soil layer (52 ± 29 cm d–1), and decreased with depth. Neglecting vertical mixing, it is estimated that tracer in the 0- to 5- and 5- to 10-cm soil layers would take 3 and 15 d, respectively, to travel 1.5 m from the injection point to the enclosure outlet. This compares with an observed mean travel time of 2 to 7 h for the conductivity centroids (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 4. Summary of estimated seepage flows.{dagger}

 
For Experiments 1 and 2, "quick" and "slow" flow were separated graphically (Fig. 5). For Experiment 1, an estimated 5% of the applied tracer was transported to the outlet in quick flow during the first 50 min. For Experiment 2, which was conducted at high flow, an estimated 30% of tracer was transported in quick flow during the first 30 min (Table 5).


View this table:
[in this window]
[in a new window]
  		Table 5. Summary of results from the experiments.{dagger}

 
Some tracer was detected in the bottom collector (5–15 cm), but the measured flow was only 0.27 to 0.45 mL s–1 and the bottom collector accounted for about 2% of the total output (Table 5). Even this low percentage is a likely overestimate because slight leakage was observed underneath and around the top collector, as indicated by a buildup of iron. Darcy's Law predicts a subsurface flow of only 0.06 mL s–1 in the 5- to 15-cm soil layer (Table 4). Thus, almost all the added tracer was transported out of the channel via the top collector (namely, in surface flow and/or subsurface flow in the top 0–5 cm of soil) (Table 5). Flow in the top collector could not be separated experimentally into surface and subsurface flow. However, for the 0- to 5-cm soil layer, Darcy's Law predicts a seepage flow of 0.27 mL s–1. This is only 5 to 7% of the total flow in Experiments 1 and 3 (conducted at low flow) and 1% in Experiment 2 (conducted at high flow). This may be an underestimate because it was not possible to sample the topsoil without compressing it slightly, which would have reduced its measured hydraulic conductivity. Even so, there is strong evidence that only a small fraction of the measured flow out of the enclosure occurred as subsurface seepage flow. By inference the majority was surface flow across the top of the soil. This is consistent with our observation that surface water was always visible even in dry weather flowing in a thin layer underneath the wetland vegetation.

In Experiment 3, NO3–N returned to background levels after about 20 h whereas conductivity, Li, and Br were all significantly higher than background when the experiment was stopped after 32 h (Fig. 7) . For the first few hours after injection the NO3–N to Br ratio in outlet samples (0.028 ± 0.002) was similar to that of the tracer (0.025) indicating negligible nitrate removal (Fig. 7). After a lag of about 4 h the NO3–N to Br ratio decreased linearly with time (at a rate of 0.0018 h–1) indicating rapid nitrate removal. Of particular note is the fact that the NO3–N to Br ratio reached zero about 12 h before the rain started (Fig. 7). We can infer that, had the experiment not stopped because of rain, no more of the added nitrate would have appeared at the outlet. Consequently, the difference (1.1 ± 0.4 g) between NO3–N input (4.6 ± 0.2 g) and output (3.5 ± 0.2 g) is a reliable estimate of nitrogen removal (24 ± 9%). The NO3–N removal of 1.1 ± 0.4 g over 20 h in the enclosure whose surface area was 1.5 m2 equates to a removal rate of 0.9 ± 0.3 g m–2 d–1.



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 7. Observed conductivity, nitrate and bromide concentrations, flow, and NO3–N to Br ratio in the top collector following surface injection of tracer during Experiment 3. Rain occurred 32 h after injection.

 
The DEA of wetland soils measured in the laboratory averaged 4.1 ± 0.3 mg N kg–1 soil h–1 (mean ± SD, n = 4). This is comparable with the DEA of 5.7 ± 1.8 mg N kg–1 h–1 measured in the same wetland by Burns and Nguyen (2002). An estimate can be made of the mass of NO3–N likely to have been removed by denitrification during Experiment 3. Assuming that: (i) the added NO3–N mixed with the top 10 cm of soil, (ii) the bulk density of the top soil layer was 0.14 ± 0.01 g cm–3 (Table 3), (iii) NO3–N removal occurred at a constant rate during the period when the outlet NO3–N concentration was above the background (20 h), and (iv) NO3–N was removed at the measured DEA rate of 4.1 mg N kg–1 h–1, then estimated denitrification within the enclosure (surface area = 1.5 m2) during Experiment 3 is 1.7 ± 0.2 g. This is comparable with the measured difference of 1.1 ± 0.4 g between NO3–N output and input, and indicates that denitrification can explain the observed nitrate removal.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
The DEA measurements confirm the earlier findings of Burns and Nguyen (2002) that the soils in the study wetland are microbially active and have the potential to remove nitrate by denitrification from the shallow ground water upwelling within the wetland. The wetland studied also supports dense surface vegetation that may well remove nitrate from surface and subsurface flow. In the current study this vegetation was trimmed back to aid visual observation of surface flow pathways and as a result we may have inadvertently reduced plant productivity and hence nitrate uptake. However, the principle focus of this study was the role of vertical mixing on tracer retention and its implications for microbial denitrification in the wetland soils. Further studies are required to quantify the relative contributions to nitrate removal from denitrification and uptake by wetland plants.

Experiment 3 indicates that substantial nitrate removal occurs provided water remains in contact with microbially active soils for about 1 d. This is consistent with the >90% and >99% nitrate removal after travel times of 2 to 8 and 3 to 13 d, respectively, observed in subsurface (10–20 cm) soils in the same wetland (Burns and Nguyen, 2002). An important question is, What proportion of the high nitrate water delivered to the surface of the wetland (e.g., in upwelling spring water or surface inflow during rain) was in contact with microbially active soils for at least 1 d, and what proportion was transported quickly to the outlet, thereby bypassing the microbially active soils?

The undetectable nitrate removal from tracer reaching the enclosure outlet during the first 4 h of Experiment 3 can be explained by one of two reasons: (i) the tracer never came into contact with the microbially active soils or (ii) there was contact but a lag before nitrate removal. We have no evidence to either support or refute the later explanation and focus subsequent discussion on the prior explanation. There is clear evidence that some of the tracer added onto the surface of the wetland traveled directly to the outlet of the enclosure in surface flow across the top of the wetland soil (Pathway 1 in Fig. 8) . We estimate that at low flows 5% of injected tracer was transported by surface flow 1.5 m along the enclosure from the injection point to the outlet within 10 min. At higher flow this fraction increased to 30% and the travel time decreased to 3 to 5 min. Surface flow is unlikely to come into contact with the underlying microbially active soils, and hence nitrate removal from surface flow (Pathway 1) is likely to be negligible.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 8. Conceptual diagram showing the important flow pathways inferred from experimental results.

 
However, in the low flow experiments (Experiments 1 and 3) surface flow accounted for only 5% of the tracer transport. A striking feature of these experiments is that the conductivity centroid occurred 6 to 7 h after the injection of tracer had ceased. This indicates that 50% of the added tracer was retained in the enclosure for >6 h while some tracer was retained for >36 h, compared with the mean travel time for surface flow of 10 min. In the high flow experiment (Experiment 2) surface flow accounted for 30% of tracer transport, but 50% was retained in the enclosure for >2 h. One possible explanation is that tracer was temporarily held in "dead zones" on the wetland surface. During low flows (Experiments 1 and 3) surface flow was concentrated in several microchannels that occupied about 30% of the enclosure width, and it is conceivable that some tracer was temporarily held in surface dead-zones along the edges of these microchannels. However, surface dead-zones seem an unlikely explanation during high flow (Experiment 2) when surface flow occurred across the entire width of the enclosure.

The fact that conductivity increased in the piezometers soon after injection is evidence of vertical mixing of tracer from the surface water into the underlying soil. There are, however, three possible criticisms of our experimental methods that may have caused vertical mixing. First, the added tracer was denser (SG = 1.03–1.06) than the surrounding surface and ground water, and there is a possibility it may have sunk into the surface soils immediately after discharge. However, the 5 to 10 L of added tracer was discharged across the entire width of the enclosure over 20 to 30 min, during which time it would have mixed with another 10 L (Experiments 1 and 3) or 50 L (Experiment 2) of surface water. This is estimated to have reduced the SG to 1.01 to 1.02, which we believe is unlikely to have resulted in significant buoyancy effects. Nevertheless, during our experiments conductivity, Li, and Br concentrations significantly exceeded background concentrations giving high precision. In future experiments it would be feasible to use smaller quantities, thereby reducing the SG of the added tracer. However, background concentrations of NO3–N are moderately high in these wetlands, which means that fairly high concentrations must be used in the tracer. A preferable, but more expensive, alternative would be to inject small quantities of labeled nitrate (e.g., 15N-NO3). Second, tracer may have "leaked" into the piezometers directly from surface water as a result of imperfect seals around the plastic pipe. Every effort was made to seal the piezometers by backfilling the top 5 cm of the augered hole with bentonite, which (on contact with water) swelled and formed a mound around the top of each well tube that extended above the level of surface water. In addition every effort was made to avoid compressing the soil surface or knocking the well tubes when sampling. Our piezometers were only embedded to a depth of about 25 cm into the poorly consolidated wetland soils and any disturbance during sampling may have broken the bentonite seals. Thus, we cannot completely discount the possibility of leakage. Third, the discharge of 5 to 10 L of tracer caused local mounding of surface water near the injection point, which may have increased pressure gradients and accelerated seepage flow into the surface soil layers.

If we accept the appearance of tracer in the piezometers as evidence of vertical mixing into the soil, then a possible transport pathway for tracer is vertical mixing from surface flow into the soil followed by shallow horizontal subsurface (seepage) flow to the outlet (Pathway 3 in Fig. 8). However, Pathway 3 cannot completely explain our observations for two reasons. First, the estimated travel time of the tracer within the top (0–5 cm) soil layer (2.9 ± 1.6 d, see Table 4) is significantly longer than the measured travel time of the conductivity centroids during the low flow Experiments 1 and 3 (6–7 h) and the high flow Experiment 2 (2 h). Second, this estimated travel time along Pathway 3 is sufficient for almost complete nitrate removal (based on results in Fig. 7). However, seepage flow was estimated to be only 7 ± 4% of total flow during Experiment 3 and so we should have measured 7 ± 4% nitrate removal if there was negligible removal for Pathway 1. In fact we measured 24 ± 9% removal. Thus, seepage flow (Pathway 3) does not appear to be the main route by which tracer reached the outlet.

This conclusion is based on our laboratory measurements of hydraulic conductivity (see Fig. 6). However, despite care being taken to sever grass roots with a serrated knife while slowly rotating and inserting the cores, the topsoil (0–10 cm) was slightly compressed during collection (by about 10% in length), which may have reduced the hydraulic conductivity. Hydraulic conductivity in the 0- to 5-cm soil layer would need to be 3700 cm d–1 for the travel time of seepage flow to be 6 to 7 h: an order of magnitude higher than the value of 350 cm d–1 estimated from Eq. [4] in the 0- to 5-cm layer. Nevertheless, it is desirable to make in situ measurements of hydraulic conductivity, seepage flows, and porewater velocities in the top soil layers that avoid the problems of disturbance and compaction associated with coring. This requires the development of new experimental techniques for use in these highly porous and very "weak" soils.

If Pathways 1 and 3 cannot explain the observed outlet tracer profiles, and the very high tracer recoveries indicate there is negligible tracer movement via Pathway 4, then we must postulate another pathway. One possible explanation of our observations is vertical diffusion between the surface flow and porewater in the underlying soils (Pathway 2 in Fig. 8). Pathway 2 is consistent with classical diffusion theory. Soon after tracer addition concentrations would have been higher in the surface water than the underlying porewater and tracer would have diffused into the porewater. Surface water moved rapidly downslope whereas porewater seeped downslope only very slowly, so concentrations would have dropped more quickly in the surface water than in the porewater. Once the concentration in the surface water was lower than in the porewater water, tracer would have diffused back into the surface water and been transported to the outlet.

There do not appear to be any published studies of vertical tracer mixing in wetland soils. However, Thibodeaux and Boyle (1987) describe the process termed "pumping" that induces vertical flow into and out of bed sediments in rivers and estuaries. Bed forms (e.g., ripples and dunes) and/or heterogeneity of porosity and hydraulic conductivity (e.g., stones and debris) cause uneven pressure distributions on the bed surface, which in turn induce seepage flows. The physical properties of the surface soil layer in the study wetland were conducive to pumping flows. First, the top 10 to 20 cm of the wetland was very poorly consolidated with high hydraulic conductivity so that even small pressure gradients are likely to have induced seepage flow. Second, the soil surface was uneven as a result of nonuniform grass growth (notably grass root mounds), compression by stock (before 1999), and erosion. Since the soils were saturated at the surface, this unevenness is likely to have caused pressure gradients and induced seepage flows with a vertical component. More detailed investigations are desirable to quantify the rates of vertical diffusion in these wetlands and identify the mechanisms that drive it.

It seems reasonable to postulate that the very top soil layer is aerobic and supports little denitrification, although there may be nitrate uptake by plants in this layer. We did not measure profiles of dissolved oxygen or redox during this study. However, Matheson et al. (2002) studied wetland soils from the same location and found very low oxygen concentrations below depths of 2 cm even in the presence of sweetgrass. We postulate that tracer reaching the outlet during the first 4 h of Experiment 3 did so either in surface flow or after mixing into and back out of the aerobic surface layer. This could explain why there was negligible nitrate removal from this water.

We also postulate that water reaching the outlet from 4 to 20 h after injection mixed deeper into, and spent longer in contact with, the anaerobic and microbially active soils. It is not possible to determine the depth of mixing precisely because our piezometers were slotted at depths from 5 to 15 cm. This means that tracer measured in the wells could have seeped in from depths anywhere in the range 5 to 15 cm. Assuming there was no leakage down along the edge of the piezometers, the detected tracer indicates that the mixing depth is at or beyond 5 cm.

It is common in stream and wetland studies to quantify nutrient removal using the equation:

[5]
where Co and Cx = nutrient concentrations measured at sites separated by a distance x (m); and K = the net uptake length coefficient (m–1). During dry weather when surface flows were low, Experiment 3 indicates that 24 ± 9% of the NO3–N tracer applied on the wetland surface was removed over a distance of 1.5 m. This implies a net uptake length coefficient of K = 0.08 ± 0.03 m–1. Nguyen (unpublished data from an earlier study in the same wetland; Burns and Nguyen, 2002) found that NO3–N concentration in springs entering the wetland in autumn (April 1999) was 1.25 mg NO3–N L–1, but that concentration in the wetland outflow was only 0.028 mg NO3–N L–1. This 98% reduction over a distance of 30 to 40 m is equivalent to a net uptake length coefficient of K = 0.05 ± 0.01 m–1, which is broadly consistent with the estimate made from our enclosure study results. Net nitrate uptake length coefficients in our study wetland were comparable with values in the range K = 0.013 to 0.068 m–1 measured during summer low flows (300–950 mL s–1) in the Purukohukohu Stream that is heavily infested with sweetgrass [Glyceria fluitans (L.) R. Br.] (Hoare, 1979; Cooper and Cooke, 1984). In stream studies, the net uptake length coefficient has been found to vary inversely with flow (Rutherford et al., 1987) and the same is likely to occur in the study wetland. We did not measure nitrate removal rates during the high flows and this would be a valuable addition to the present study.

The enclosures used in our experiments offer several advantages. First, they allow a homogeneous area of wetland to be isolated and studied. One obvious weakness of our current study is that, by conducting three tracer experiments in the same enclosure, we have achieved only pseudo-replication. It is desirable to investigate the spatial variability in vertical mixing and nitrate reduction, within the study wetland and between different wetlands, and to relate these to soil properties. Second, the enclosures and collector trays allow flow to be measured in two layers (0–5 and 5–15 cm). Although this study showed that seepage flow in the lower layer was negligibly small (i.e., about 2% of the total), it would be desirable to further separate flow in the 0- to 5-cm layer into surface (0–1 cm) and shallow seepage (1–5 cm) flow. This might be feasible by adding a third collector above the top collector and resting the base on the soil after first cutting back the wetland vegetation to the level of the roots.

We found very high variability between piezometers in tracer dynamics. It is not clear whether this reflects spatial heterogeneity in soil properties or is an artifact of inconsistencies in the bentonite seals that allowed tracer in surface water to leak down the PVC tubes (e.g., if accidentally knocked during sampling). One possible improvement for future experiments would be to embed the piezometer tubes deeper into the underlying clay (e.g., to a depth of >50 cm), which would make them less likely to move and break the bentonite seals during sampling, but continue to sample water only in the 0- to 5- and 5- to 15-cm surface layers.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
Although conductivity is a nonspecific measurement, an array of logging conductivity probes provided valuable, semiquantitative information about the residence time and percentage recovery of an added conservative tracer more easily than traditional water sampling and laboratory Li and Br analyses.

The wetland swale studied, which is typical of hill-country pasture in many parts of New Zealand, was permanently saturated and had very poorly consolidated surface soils. Water upwelled from several springs near the head of the wetland and spread fairly evenly across the wetland surface, and there was visible surface flow even in dry weather. When tracer containing NO3–N was applied to the surface in an experimental enclosure within the wetland, 24 ± 9% (1.1 ± 0.4 g) was removed in a distance of 1.5 m during dry weather. This equates to an areal removal rate of 0.9 ± 0.3 g m–2 d–1 and a net uptake length coefficient of K = 0.08 ± 0.03 m–1. We did not measure nitrate removal during wet weather but, based on stream studies, we would anticipate that nitrate removal would vary inversely with flow.

Tracer leaving the experimental enclosure within 4 h after injection had the same NO3–N to Br ratio as the injected tracer, which is consistent with negligible NO3–N removal from surface flow and tracer that spends a short time in contact with surface soils. Tracer leaving the enclosure 20 h or more after injection had NO3–N concentrations comparable with preinjection background concentrations indicating that soil–water contact time of approximately 1 d is sufficient to achieve substantial reductions in NO3–N concentration. This latter finding is consistent with previous studies on subsurface flows in the same wetland (Burns and Nguyen, 2002).

Tracer was transported more slowly than can be explained by surface flow, but more quickly than can be explained by Darcian seepage flow. There was evidence from shallow wells that tracer mixed vertically into the surface soil soon after injection. We postulate that vertical diffusion into the top-10-cm porous soil horizon occurred soon after tracer release when surface concentrations were high, followed by diffusion out of the subsoil again when surface concentrations decreased. Such diffusion is unlikely to occur beyond a 50-cm depth since there is an aquaclude at this depth.

This study confirms that some of the spring water that upwells within a riparian wetland may bypass the microbially active soils and hence experience very low NO3–N removal as has been suggested previously (Hill, 1996; Gold et al., 2001; Rosenblatt et al., 2001). However, we postulate that uneven microtopography of the soil surface, combined with spatial heterogeneity of hydraulic conductivity, have the potential to promote vertical exchange between surface flows and porewater in the very poorly consolidated surface soils in wetlands with similar characteristics to that studied. At low flows there appears to be sufficient vertical diffusion for a significant proportion of the surface flow to interact with microbially active soils, especially in the poorly consolidated surface layer. We suggest that riparian wetlands with upwelling springs and surface flows should not be dismissed as having low NO3–N removal potential without checking whether there is significant vertical diffusion.


   ACKNOWLEDGMENTS
 
Sally Rutherford carried out much of the fieldwork. Four anonymous reviewers made very helpful suggestions about how to improve the paper. This work was funded by the New Zealand Foundation for Research Science and Technology through Contract C01X0010.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 


Related articles in JEQ:

This Issue in Journal of Environmental Quality

JEQ 2004 33: 799-804. [Full Text]  



This article has been cited by other articles:


Home page
 J. Environ. Qual.Home page
R. Collins
Fecal Contamination of Pastoral Wetlands
J. Environ. Qual., September 1, 2004; 33(5): 1912 - 1918.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (6)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rutherford, J. C.
Right arrow Articles by Nguyen, M. L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rutherford, J. C.
Right arrow Articles by Nguyen, M. L.
GeoRef
Right arrow GeoRef Citation
Agricola