Volume 3, Issue 1 p. 210-224
Original Article
Free Access

The role of hydrodynamics in structuring in situ ammonium uptake within a submerged macrophyte community

Edward P. Morris

Corresponding Author

Edward P. Morris

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Present address Department of Ecology and Coastal Management, Instituto de Ciencias Marinas de Andalucía, Consejo Superior de Investigaciones Científicas, 11519 Puerto Real, Spain

Correspondence to Edward P. Morris [email protected]Search for more papers by this author
Gloria Peralta

Gloria Peralta

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Search for more papers by this author
Tom Van Engeland

Tom Van Engeland

Centre for Estuarine and Marine Ecology, Netherlands Institute of Ecology, 4400 AC Yerseke, the Netherlands

Search for more papers by this author
Tjeerd J. Bouma

Tjeerd J. Bouma

Centre for Estuarine and Marine Ecology, Netherlands Institute of Ecology, 4400 AC Yerseke, the Netherlands

Search for more papers by this author
Fernando G. Brun

Fernando G. Brun

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Search for more papers by this author
Miguel Lara

Miguel Lara

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Search for more papers by this author
Iris E. Hendriks

Iris E. Hendriks

Instituto Mediterraneo de Estudios Avanzados, Universidad de las Islas Baleares y el Consejo Superior de Investigaciones Científicas, 07190 Esporles, Spain

Search for more papers by this author
Javier Benavente

Javier Benavente

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Search for more papers by this author
Karline Soetaert

Karline Soetaert

Centre for Estuarine and Marine Ecology, Netherlands Institute of Ecology, 4400 AC Yerseke, the Netherlands

Search for more papers by this author
Jack J. Middelburg

Jack J. Middelburg

Centre for Estuarine and Marine Ecology, Netherlands Institute of Ecology, 4400 AC Yerseke, the Netherlands

Faculty of Geosciences, Utrecht University, 3508 TA Utrecht, the Netherlands

Search for more papers by this author
J. Lucas Perez-Llorens

J. Lucas Perez-Llorens

Faculty of Marine and Environmental Sciences, University of Cádiz, 11510 Puerto Real, Spain

Search for more papers by this author
First published: 04 December 2013
Citations: 19

Abstract

In low-nutrient, macrophyte-dominated coastal zones, benthic ammonium (NH4+) uptake may be influenced by the structural properties of plant canopies via their effect on near-bed hydrodynamics. Using a dual-tracer (uranine and 15NH4+) method that does not require enclosures, we examined how this process affects nutrient uptake rates within a tidally dominated, patchy Caulerpa proliferaCymodocea nodosa landscape. NH4+ uptake was determined by calculating tissue 15N excesses and correcting for 15N enrichment as derived from uranine concentration. Vertical hydrodynamic profiles were measured in the downstream flow direction from outside to inside of the C. nodosa bed by using an array of acoustic Doppler velocimeters. The transition from a C. prolifera to a C. nodosa bed included a change in both benthic canopy properties (short and dense to tall and sparse) and sediment topography (0.2-m increase in water column depth) that resulted in an increase in longitudinal advection and turbulent diffusivity within the C. nodosa canopy between 0.5 and 1.5 m from the leading edge. Vertical differences in canopy water exchange appeared to explain variations in uptake between biotic functional groups; however, no clear differences in longitudinal uptake were found. Using in situ labeling, this study demonstrated for the first time the role of hydrodynamics in structuring NH4+ uptake within an undisturbed, patchy macrophyte landscape.

Introduction

Ammonium (NH4+) uptake by submerged macrophyte communities is a major pathway by which nitrogen is removed from the water column and potentially retained within shallow, marine, benthic ecosystems (Wallentinus 1984; Touchette and Burkholder 2000; Fernandes et al. 2009). The high foliar NH4+ uptake capacity of many marine angiosperms (Touchette and Burkholder 2000; Rubio et al. 2007; Alexandre et al. 2011) and macroalgae (O'Brien and Wheeler 1987; Vergara et al. 1997; Malta et al. 2005) indicates that, in systems with low-to-moderate nutrient concentrations, uptake is often controlled by rates of transfer across concentration boundary layers, rather than by enzymatic processes (Bilger and Atkinson 1992). Hence, an increase in NH4+ concentration or water velocity strongly enhances macrophyte uptake (Hurd et al. 1996; Thomas et al. 2000; Koch et al. 2006).

Macrophyte communities often form distinctly patchy distributions consisting of bare and vegetated areas as well as patches of different species. Water flow interacts with these patchy landscapes, resulting in horizontal (Fonseca and Koehl 2006) and vertical (Gambi et al. 1990; Ackerman and Okubo 1993; Koch and Gust 1999) differences in water velocity and turbulence that influence NH4+ uptake rates of individual canopy elements (Cornelisen and Thomas 2002; Morris et al. 2008). That is, the physical interaction of macrophyte structures with water movement results in spatially structured variation in benthic NH4+ transport rates. Furthermore, intra- and interspecific differences in macrophyte canopy properties, such as canopy morphology (Verduin and Backhaus 2000; Stewart and Carpenter 2003; Hendriks et al. 2010), the density of elements (Fonseca et al. 1982; Peterson et al. 2004; Lacy and Wyllie-Echeverria 2011), stiffness (Gaylord and Denny 1997; Bouma et al. 2005; Peralta et al. 2008), and patchiness (Folkard 2005; Maltese et al. 2007), strongly influence near-bed flow properties. Consequently, macrophyte landscapes potentially have a wide range of nutrient transport niches determined solely by physical interactions with water flow.

Simple canopy structure metrics, such as the diameter (d) and density of structures (n), projected frontal area per unit volume (a = nd), canopy height (hc), and the canopy drag coefficient (CD~1), can help to anticipate these interactions (Nepf and Ghisalberti 2008; Ghisalberti and Nepf 2009). In sparse canopies (ah < 0.04, where h = height), the vertical water velocity profile is that of a rough boundary layer with notable within-canopy longitudinal advection over a transition length scale (LT), which may be larger than the patch length (Ghisalberti and Nepf 2009). In dense canopies (ah>0.3), the extent of vortex penetration into the canopy (δe) and LT are proportional to the drag length scale of the canopy, (CDa)− 1, and at longitudinal distances exceeding LT, exchange is dominated by Kelvin–Helmholtz vortices (Nepf and Ghisalberti 2008). For very dense canopies, δe is very small (~2d), and the majority of canopy exchange is controlled by stem-scale turbulence (White and Nepf 2008). Hence, the relevant hydrodynamic regime that controls canopy water exchange (and potential nutrient supply) can be very different, depending on a species' structural characteristics.

Differences in the relevant hydrodynamic regime controlling canopy nutrient exchange may be expected within subtidal Caulerpa proliferaCymodocea nodosa landscapes, which are made up of two very different canopy types. C. prolifera, a rhizophytic, soft-sediment dwelling alga, forms extensive, short (hc~0.05 – 0.13 m) and dense canopies (a>100 m− 1) that often surround patches of tall (hc~0.2 – 1 m), relatively sparse (a~1 – 5 m− 1) C. nodosa, a marine angiosperm. We estimate that LT is in the order of 0.02 m for C. prolifera and 5 m for C. nodosa; that is, longitudinal flow penetrates ~250 times farther into the C. nodosa than the C. prolifera canopy. Furthermore, as C. prolifera (ah)− 1 is generally < < 2d, δe/h is expected to be small (White and Nepf 2008), suggesting low rates of vertical water exchange. In contrast, for C. nodosa, canopy water exchange is expected to be dominated by a mix of longitudinal advection and vertical diffusivity associated with Kelvin–Helmholtz vortices. In other words, both longitudinal water flow and mixing are predicted to be much larger in the C. nodosa than in the C. prolifera canopy. As water flows over the upstream–downstream transition between the dense C. prolifera and sparse C. nodosa canopies, because of the difference in canopy density, a transition in the shape of the vertical velocity profile is predicted that results in a gradient of increasing near-bed water velocities across the transition. Furthermore, flow adjustment should be observable upstream of the leading edge of the C. nodosa canopy at a length scale comparable to the canopy height (Rominger and Nepf 2011), suggesting changes in flow properties at the transition that extend into the C. prolifera bed. We hypothesized that if NH4+ uptake is physically limited, these changes in water flow properties will manifest as spatially explicit differences in the uptake rates of biota within the canopies.

Few studies have examined how differences in hydrodynamics act to structure the ammonium uptake of natural benthic macrophyte communities. Traditionally, NH4+ uptake rates are measured using a chamber (Thursby and Harlin 1982; Hemminga et al. 1994; Naldi and Viaroli 2002), in which hydrodynamic conditions and the structure of macrophyte communities deviate strongly from natural conditions. Flow chambers that generate both unidirectional and oscillatory currents allow the preservation of hydrodynamic effects (Thomas and Atkinson 1997; Thomas et al. 2000; Thomas and Cornelisen 2003) and, when combined with stable-isotope labeling, allow the contributions of individual macrophyte community components to be assessed with minimal disturbance to canopy structure (Cornelisen and Thomas 2002; Cornelisen and Thomas 2004; Lepoint et al. 2007). Nevertheless, although crucial to the development of modeling approaches (Falter et al. 2004; Cornelisen and Thomas 2009), these studies are a simplification of natural systems; they usually require the transplantation of macrophyte communities or the use of a field flume with spatial dimension inappropriate for examining heterogeneous landscapes.

An alternative is stable-isotope labeling without enclosures, which has been applied successfully in a number of aquatic systems (Hughes et al. 2000; Gribsholt et al. 2005; Pace et al. 2007). Based on these principles, we devised a system that allowed us to spike the water column over a shallow, submerged Caulerpa proliferaCymodocea nodosa landscape with a mix of 15NH4+ and uranine (an Na salt of fluorescein). By measuring the spatial distribution of uranine and the incorporation of 15NH4+ within plant tissues, we examined in situ NH4+ uptake rates of the natural phototrophic community under undisturbed hydrodynamic conditions. Using this technique, we examined the hypothesis that spatial differences in hydrodynamics influence in situ NH4+ uptake within an undisturbed natural, patchy macrophyte community.

Methods

Study Site

Labeling experiments were carried out in Cádiz Bay located in the west of the Gulf of Cádiz, southwestern Spain, between 36° 23′ and 36° 37′ N latitude and between 6° 8′ and 6° 15′ W longitude (Fig. 1A). The bay is divided into two basins, a shallower basin (inner bay) with a mean depth of 3 m MLW (mean low water) (Freitas et al. 2008) and a deeper basin (outer bay) with a mean depth of 12 m MLW (Rueda and Salas 2003). The inner bay is protected from the action of large waves but is strongly influenced by semidiurnal co-oscillating tides with mean amplitude of 1.5 m (Alvarez et al. 2003).

Details are in the caption following the image

Map of study location: Bay of Cádiz, southwestern Spain with insert of larger region (A), and airborne photograph of the Caulerpa proliferaCymodocea nodosa boundary (B), with the patch boundary highlighted in blue and the position of the nutrient delivery system in yellow. Visualization of aerial photographs (kindly supplied by L. Del Río Rodríguez, University of Cádiz) was carried out by using QGIS (www.qgis.org).

Experiments were carried out in the shallow (0.5 m lowest astronomical tide), southwestern region of the bay (Fig. 1A, Santibañez) where maximum free-stream current velocities range between 0.03 and 0.1 m s− 1. A 10-m-long, relatively straight section of a large macrophyte community dominated by a Cymodocea nodosa Ucria (Ascherson) patch bordered by a dense bed of Caulerpa prolifera (Forsskål) J.V. Lamouroux was chosen to examine the transition between canopy types (Fig. 1B). Previous hydrodynamic measurements showed that ~30 min after low tide, current direction was essentially unidirectional, parallel to the shore (x-direction), and perpendicular to the edge of the C. nodosa patch (y-direction). Sediment height (zs) was not constant across the transition between the two species; the C. prolifera bed (x = − 1 m) was 0.2 m higher than the C. nodosa bed (x = 2 m) (Fig. 2). We defined water column height (z) relative to the sediment surface at x = 2 m; however, we also refer to height (h) above the sediment surface. The experiment was carried out on 2 August 2007 at ~13:00 (UTC+1) when z = 1.35 m; by the end of the release, z = 1.53 m. Photosynthetic photon flux measured using a custom data logger (Apogee SQ-100 sensor; Apogee Instruments, Inc., USA) connected to an Onset HOBO data logger (Onset Computer Corporation, USA) at z = 0.5 (within the C. prolifera bed) during the labeling period was 801 ± 95 μmol-quanta m− 2 s− 1 (mean ±  standard error); water temperature was 27.7 ± 0.2 °C, and salinity, 34.

Details are in the caption following the image

Schematic xz (A) and xy (B) diagrams of the experimental design. Filled circles represent the release points of the nutrient delivery system (NDS); open circles, water collection points; and squares, biomass samples. Horizontal distance (x-axis, m) is relative to the edge of the C. nodosa patch and parallel to the main flow direction. The increase in the sediment surface height within the C. prolifera canopy (dashed line) is represented by a brown polygon. Water column height (z-axis, m) is relative to the sediment surface at x = 2 m within the C. nodosa canopy (represented by a solid line). The green polygon represents the extent of the C. prolifera bed.

Canopy Properties

Canopy structural properties for each species are shown in Table 1. The density and biomass of above-sediment structures were estimated by collecting all plant material within randomly selected 0.2 × 0.2–m quadrants (n = 4). Canopy height (hc) was assessed by measuring the distance from the top of the canopy (assessed by eye in situ) to the sediment surface at a number of positions haphazardly situated throughout the experimental area and calculating the arithmetic mean (i.e., equivalent to the average roughness height). hc was relatively constant for C. prolifera, whereas for C. nodosa it varied between 0.5 m at the leading edge and 0.7 m at x = 4 m (Fig. 2). C. nodosa morphology (number of leaves, leaf length, width, and thickness) represents the mean values of samples collected in summer between 2005 and 2011 as part of a C. nodosa monitoring program (www.famar.wordpress.com). C. prolifera morphology was derived from Vergara et al. (2012). Vertical profiles of the frontal area of structures per water volume (a, m− 1) were calculated using a cumulative probability function of the distribution of structure lengths scaled to the in situ hc. Unfortunately, C. nodosa epiphyte biomass was not quantified. Rather, summer biomass estimates of 0.18 ± 0.08 gDMepiphytes gDMC. nodosa (n = 18; P. García-Marin, University of Cádiz, pers. comm.) were used. Unattached macroalgae (Gracilaria sp.) were sparsely distributed within the lower canopy of the macrophyte beds.

Table 1. Summary of macrophyte canopy properties [range or mean ± SD (n)]
Property C. prolifera C. nodosa
Structure density (structures m− 2) 8700 ± 146 (4) 356 ± 60 (4)
Leaves per shoot 3.5
Structure width (d, m) 0.03 0.006
Structure thickness (e, m) 1 × 10− 3 1 × 10− 3
Biomass (gDM m− 2) 200 ± 13 (4) 258 ± 25 (4)
Canopy height (h, m) 0.1 ± 0.02 (5) 0.5–0.7 (8)
Frontal area of structures per water volume (a, m− 1) 183 3.8
ah (dimensionless) 18.3 1.9–2.7
Canopy porosity (dimensionless) 0.87 0.996

Hydrodynamic Measurements

Vertical profiles of hydrodynamic properties were measured in the downstream flow direction (x) from outside to inside of the C. nodosa patch (~5 m offshore from the nutrient delivery system [NDS]) by using four acoustic Doppler velocimeters (ADV; two Vector, one NDV, and one Vectrino, Nortek AS, Norway) fixed on a single frame at x = − 0.8, 0, 0.7, and 1.5 m, perpendicular to the leading edge of the C. nodosa bed. Vertical profiles were collected by moving the frame by hand, up and down, cycling through heights of z = 0.48, 0.68, 0.88, 0.98, 1.08, and 1.28 m. ADV measurements were carried out over a 1-h time interval coinciding with the nutrient release. The ADV frame was left at each z-position for ~300 s before being moved to the next z-position, resulting in three measurement periods at z = 0.48 m, one at z = 1.28 m, and two at the remainder (z = 0.68–1.08 m). A single ADV also recorded water velocity and depth at a height of 0.5 m above the sediment surface in the C. prolifera bed.

Velocity components (u, v, and w in the x-, y-, and z-directions, respectively) were measured at 25 or 64 Hz. Coordinates were rotated in the xy plane so as to align u with the x-axis, correcting for slight variations in the orientation of the individual ADVs and the frame. All three velocity components with any of the individual beam correlations < 70% or signal-to-noise ratio < 5 were discarded. Velocity spikes were removed using the phase-space despiking algorithim in the free software package WinADV (Goring and Nikora 2002; Wahl 2003). Time-averaged velocity components (equation image, equation image, and equation image) and their respective fluctuations (u′, v′, and w′) were calculated by averaging the temporal measurements (n = 2.5–15 × 103) at each sampling point as described in Morris et al. 2008. equation image was fitted to a second-order polynomial surface to extract values corresponding to the positions of uptake measurements (see below). Because measurements were missing from the lower regions of both canopies, to aid the fitting procedure we inserted near zero (0.005 m s− 1) equation image values at midway within the C. prolifera canopy and just above the sediment surface in the C. nodosa canopy.

Turbulent Diffusivity Coefficients

Turbulent diffusivity in the x-, y-, and z-directions, at each sampling point of the vertical profiles was derived from Taylor's theory according to Holtappels and Lorke (2011). Normalized autocorrelation coefficients of the root-mean-squared velocity fluctuations (u′RMS, v′RMS, and w′RMS) were calculated for all time intervals (t) within each burst period (using the R function acf; http://www.r-project.org/). t values were translated into spatial distances (r) according to equation image, where equation image is the modulus of velocity (i.e., equation image), transforming the autocorrelation coefficients into Eulerian spatial correlation coefficients (EE). EE was integrated over the spatial distance (r) (using the R function cumsum). The integral length scale (LE) was defined as the first maximum of integrated EE, which corresponds to the first zero crossing of the autocorrelation coefficient. Turbulent diffusivity coefficients derived from Taylor's theory (K, m2 s− 1) were calculated as the product of the root-mean-squared velocity deviation and LE for each direction, respectively (e.g., Ku = uRMSLE). The time-ensemble median, 25%, and 75% quantiles were chosen to summarize K for each burst.

Tracer Addition

An NDS, consisting of six lengths of irrigation tubing designed to release water at a constant rate (2 L h− 1) through each exit port (spaced~0.2 m apart) when under pressure by a small submersible pump placed within a 50-L holding tank, was placed perpendicular to the main flow direction upstream of the C. nodosa leading edge at x = − 2 m (Fig. 2). Lines of irrigation tubing were positioned at h = 0.15, 0.3, 0.45, 0.6, 0.75, and 0.9 m, across a distance of 5 m (representing a yz labeling plane of 3.75 m2), resulting in an estimated tracer releases rate of 0.05 m h− 1. The first line of tubing was 5 cm above the top of the C. prolifera canopy.

Using a priori information about the water velocity, the delivery rate of the tracer, and water column NH4+ concentration, we added 25 g of 15NH4Cl (x(15N) = 98%) to the holding tank (for recent recommendations on stable-isotope terminology, see Coplen 2011) to increase in situ water column 15N concentration by ~5%. However, actual dilution of the 15NH4+ label in the water column was quantified using the fluorescent dye uranine (an Na salt of fluorescein; C20H10Na2O5, M = 376.28 g mol− 1, Flury and Wai 2003). The uranine concentration recovered in the water column varied between 1.2 and 28.5 μg L− 1.

Labeling of the water column was initiated by filling two 50-L containers with seawater from the site and, shortly before pumping the mixture through the NDS, spiking them each with 1 L concentrated tracer and thoroughly mixing. Pumping of the mixture continued for 30 min and did not completely empty the containers, resulting in a total volume discharged of ~90 L.

A water collection system, consisting of 28 tubes connected to two peristaltic pumps, was positioned to collect water samples at x = − 1, − 0.5, 0, 0.5, 1, 2, and 4 m and heights of h = 0.15, 0.3, 0.5, and 1 m (Fig. 2). Thus, samples were collected 5 cm above the C. prolifera canopy, as well as within and above the C. nodosa canopy. Water samples (20 ml) were collected at 5, 10, 20, 30, and 40 min from the start of tracer release and immediately stored on ice within a cool box. Sampling times were corrected for the time required for water to travel to the boat via the tubing (~2 min).

Chemical and Isotopic Analysis

Before the nutrient release, 9, 17, and 3 samples (0.5 gDM) of C. prolifera, C. nodosa, and the unattached macroalgae Gracilaria sp., respectively, were collected haphazardly from the experimental area for natural abundance stable-isotope analysis. At the end of the labeling period (40 min), divers carefully collected biomass samples by hand of C. prolifera, C. nodosa, and Gracilaria sp. (depending on their occurrence) from four 0.2 × 0.2–m plots situated next to each of the water collection points (Fig. 2).

Samples were stored in a cool box until they were transported to the laboratory (within 4 h). No obvious epiphytes could be observed on C. prolifera. Epiphytes on the leaves of C. nodosa were removed by carefully scraping with a razor blade, stored on pre-ashed GF/C glass fiber filters (Whatman, UK), and freeze-dried. Macrophyte tissues were separated from the sediment, cleaned in seawater, washed briefly with distilled water, and separated into above-sediment (assimilators, leaves, and sheath) and below-sediment (rhizoids, rhizome, and roots) parts before being freeze-dried and ground to a fine powder.

Dried samples of macrophyte tissues and epiphytes were analyzed using an elemental analyzer (Thermo EA 1112, Thermo Fisher Scientific, USA) coupled to an isotope ratio mass spectrometer (Thermo Delta V Advantage, Thermo Fisher Scientific) with a Conflo II interface (elemental analyzer-isotopic ratio mass spectrometer), allowing determination of nitrogen and carbon content (μmol (gDM)− 1) and 15N atomic fractions of dry matter (x(15N)DM).

The uranine concentration (μg L− 1) within unfiltered water samples was measured using a fluorometer (TD-700, Turner Designs, USA) fitted with fluorescein excitation (490 nm) and emission (580 nm) filters and a blue mercury lamp (kit 10-086R). Ammonium concentrations of filtered (GF/F filter, Whatman) water samples collected upstream of the NDS were analyzed colorimetrically.

Calculations and Statistics

Downstream uranine and upstream NH4+ concentrations were used to estimate the 15N atomic fraction of the water column (x(15N)aq.N) during the labeling period, that is, the actual dilution of the 15NH4+ label. Weighted-mean uranine concentration (weighted by the time interval 0–40 min; see below) was fitted to a second-order polynomial surface (xz; see below), and uranine values were extracted at each x-position corresponding to the respective heights of the macrophytes. These regions were defined as the top of the C. prolifera canopy, within the C. nodosa canopy, 0.05 m above the sediment surface for Gracilaria sp., and the top of the C. nodosa canopy for epiphytes. Concentrations of added NH4+(μmol NH4+L− 1) were calculated as
urn:x-wiley:21573689:media:lof32397024:lof32397024-math-0009
where Mtracer is the mass ratio of 15NH4Cl:uranine of the initial tracer mix (1.67 g g− 1) and equation image is the atomic weight of ammonium chloride (54.5).
The 15N atomic fraction of the dissolved N source pool in the water column (x(15N)aq.N) was then calculated as
urn:x-wiley:21573689:media:lof32397024:lof32397024-math-0017
where x(15N)tracer is the atomic fraction of the added tracer (0.98), [NH4+]water is 10.6 ± 3.5 μmol-NH4+L− 1 (mean ±  SE), and x(15N)initial is the atomic fraction of the initial water column (estimated as the mean atomic fraction of all macrophyte above-sediment natural abundance tissue samples, 3.68 × 10− 3).
Excess atom fractions (xE(15N)DM_sample) were calculated as the difference between the atomic fraction labeled (x(15N)DM_sample) and natural abundance x(15N)DM_nat.ab) of plant dry matter samples (Van Engeland et al. 2011). Dry-matter-specific uptake rates of NH4+ (Vsample(NH4+), μmol NH4+ gDM− 1 h− 1) were calculated as
urn:x-wiley:21573689:media:lof32397024:lof32397024-math-0010
where [N]DM_sample is the tissue nitrogen content (μmol-N gDM− 1) and t is the length of exposure to the tracer (0.67 h).

These calculations assume that variations in water column NH4+ concentration, depletion by biological activity, and isotope dilution due to regeneration were minimal within the labeling area. Furthermore, it is assumed that x(15N)initial was similar to the natural abundance of the community of macrophytes, which is supported by previous studies (Morris et al. 2009). To help assess assumptions about N depletion and dilution of the water column NH4+ source pool by uptake and regeneration, respectively, areal uptake rates (μmol NH4+m-benthos− 2 h− 1) were calculated by multiplying Vsample(NH4+) at each x-position by the areal biomass (gDM m− 2) of each tissue component and integrating over the study area to provide an estimate of the potential areal uptake rate of the community. Examples of the magnitude of benthic NH4+ regeneration rates were derived from Fulweiler et al. (2010) and Mortazavi et al. (2012).

Statistics were performed with the statistical program R 2.15 (R Core Team 2013). Significant isotope excesses were assessed using nonparametric Kruskal–Wallis rank sum tests (kruskal.test function) and nonparametric multiple test procedures (Behrens–Fisher, Satterthwaite t-approximation; npmc function) to examine differences between natural abundance and labeled sample atomic fractions. The same tests were used to examine differences between uptake rates at different positions for each biological component and to examine differences between components. Significance levels were set at p < 0.05. Uranine and water velocity data were fitted to a second-order-polynomial surface by using the function surf.ls (Venables and Ripley 2002). Pearson's product-moment correlation was used to test for significant correlation between water velocity and specific uptake (cor.test function). Error propagation (Monte Carlo simulations) calculations for areal uptake values were carried out using the function propagate (Ritz and Spiess 2008).

Results

The spatially (xz) weighted-mean uranine concentration reached a steady value of ~15 μg L− 1 after 15 min of labeling (Fig. 3A). Time- and z-weighted mean uranine concentration was relatively homogeneous in the x-direction, although slightly lower values were recorded at x = 4 m (Fig. 3B). This could be mainly attributed to differences in the vertical distribution of the tracer over time (Fig. 4); time-weighted mean uranine concentration during the labeling period (0–45 min) ranged from a maximum of 17.6 μg L− 1 observed at the top of the Cymodocea nodosa canopy at x = 2 m to a minimum of 6.2 μg L− 1 in the upper water column at x = 4 m. However, the lowest uranine concentration corresponding to a depth region with uptake measurements was 8.7 μg L− 1 at the sediment surface, x = − 1 m. Thus, effective 15NH4+ tracer concentrations were between 0.27 and 0.47 μmol NH4+L− 1 (an increase in total water column [NH4+] of between 2.5% and 4.5%), which represented an increase of 15N (x(15N)aq.N/x(15N)initial × 100) of between 650% and 1130%.

Details are in the caption following the image

Depth-averaged uranine concentrations (weighted mean ± 95% confidence interval) grouped by time (A) and x-position (B). Pumping of label ceased after 30 min.

Details are in the caption following the image

Weighted mean uranine concentration during the labeling period (0–45 min). Other details are as in Fig. 2A.

Free-stream water velocity (equation image at z = 1.08 m) was 0.040 ± 0.005 m s− 1 (mean ±  SD; Fig. 5A). Vertically averaged equation image within the C. nodosa canopy was 0.017 ± 0.004, 0.019 ± 0.012, and 0.025 ± 0.009 ms− 1 at x = 0, 0.69, and 1.46 m, respectively, suggesting that velocities increased up to a distance of at least 1.5 m from the leading edge. Root-mean-squared velocity deviations (u′RMS) ranged between 0.007 and 0.028 m s− 1 and appeared relatively similar between profiles (Fig. 5A, error bars). As expected, w′RMS values were much smaller than in the x-direction, apart from at x = 1.46 m (Fig. 5B, error bars), where values as high as 0.014 m s− 1 were observed near the C. nodosa canopy interface.

Details are in the caption following the image

Profiles of horizontal (A) and vertical velocity (B) and horizontal (C) and vertical turbulent diffusivity (D) measured during 1–3 bursts of 300 s depending on height at x = − 0.77, 0, 0.69, and 1.46 m. Schematics of relative sediment, canopy, and water level positions are shown as in Fig. 2 at x = − 0.77, 0, 0.69, and 1.46 m. Error bars represent root-mean-squared velocity deviations and quantiles (25% and 75%) for velocities and turbulent diffusivity, respectively. Dashed blue lines represent the air–water interface at the start (1.35 m) and end (1.53 m) of the experiment. Green polygons represent the relative vertical distribution of frontal surface area per water volume (a/amax) of the C. prolifera (bright green) and C. nodosa (transparent dark green) canopies. Vertically integrated a for each canopy is shown as text. Brown polygons represent the sediment height. Water column height (z-position, m) is relative to the sediment surface at x = 2 m.

Median turbulent diffusivity coefficients (K) in the x- and y-directions were similar (Kx and Ky were 1.08 × 10− 2 and 8.91 × 10− 3m2s− 1, respectively) and about three times larger than vertical diffusivity (Kz = 2.99 × 10− 3 m2 s− 1). Vertical profiles of diffusivity showed similar tendencies in the x-, y-, and z-directions (Fig. 5C–D); values decreased toward the canopy at x = − 0.77 m, whereas they were higher and relatively constant throughout the water column at x = 1.46 m. Profiles at the patch edge (x = 0 and 0.69 m) appeared to be intermediate between these two situations. This resulted in significantly different water column Kz values among positions (Kruskal–Wallis rank sum test, χ2(3,48) = 21.6, p < 0.001), with values at x = 1.46 m (0.01 m2s− 1) much higher than the other positions (0.002 m2s− 1; Behrens–Fisher test, p < 0.05). This effect could be attributed to an increase in both w′RMS (see above) and LE_z; LE_z was 1.78 m at x = 1.46 m and 0.47 m at the other positions (p < 0.05).

Significant increases in x(15N)DM_sample compared to x(15N)DM_nat.ab (data not shown) were observed for C. nodosa leaves (Kruskal–Wallis rank sum test, χ2(6,33) = 31.1, p < 0.001), C. nodosa epiphytes (χ2(6,31) = 30.0, p < 0.001), and C. prolifera above-sediment tissues (χ2(4,20) = 18.1, p < 0.01). Nonparametric multiple test procedures revealed that all labeling positions showed significant uptake of 15N for C. nodosa leaves and epiphytes (Behrens–Fisher test, p < 0.05), and only a single position of C. prolifera above-sediment tissues did not show significant uptake (x = − 1 m).

Mean specific NH4+ uptake rates of the macrophyte community components ranged from a minimum of 0 (i.e., no significant uptake, see above) to a maximum of 25.0 μmol NH4+ gDM− 1 h− 1 (Fig. 6; only significant uptake of the major components is shown). No differences in the uptake rates at different x-positions were found for C. nodosa epiphytes or leaves, whereas distance significantly affected the uptake of C. prolifera above-sediment tissues (Kruskal–Wallis rank sum test, χ2(2,9) = 31.1, p < 0.05), with uptake at x = 0.5 m higher than at x = 0 and − 0.5 m (Behrens–Fisher test, p < 0.05).

Details are in the caption following the image

Specific nitrogen uptake rates at distances relative to the leading edge of a C. nodosa patch for C. nodosa epiphytes (A), C. nodosa leaf (B), and C. prolifera above sediment (C). Gracilaria sp. and C. nodosa roots are not shown. C. prolifera was present only at positions − 1 to 0.5 m, and C. nodosa, at 0–4 m. Significant differences (Behrens–Fisher test, p < 0.05) in C. prolifera uptake are indicated by different letters. Note that the y-axis for epiphytes has a different scale.

When specific uptake rates were pooled per component, C. nodosa epiphytes had the highest uptake; no significant differences where found between C. nodosa leaves and C. prolifera above-sediment, and Gracilaria sp. and C. nodosa roots had the lowest rates of uptake (Table 2; Behrens–Fisher test, p < 0.05). Upscaling suggested that even though C. nodosa leaves had uptake rates four times lower than those of epiphytes, because of their high biomass they were the dominant component in terms of areal uptake (Table 2). When considered as a community, the seagrass habitat had an areal uptake rate 2.4 times higher than the Caulerpa habitat. Integrating areal uptake for both habitats over the whole study area (6 m × 1 m) suggested that at the in situ NH4+ concentration of 10.6 μmol NH4+L− 1, vertically averaged longitudinal velocity (Ub) of 0.03 m s− 1, and depth of 1.45 m, depletion of total NH4+ within a packet of water passing through the study area would be < 1% of the ambient concentration. Likewise, based on reported estimates of sediment NH4+ regeneration fluxes to the water column (~100–300 μmol NH4+m− 2 h− 1; Fulweiler et al. 2010; Mortazavi et al. 2012), dilution of the total-NH4+ source pool within a packet of water passing through the study area is not likely to be significant. Ammonium mass transfer coefficients (equation image), that is, areal uptake divided by water column NH4+ concentration, were 4.76 × 10− 5 and 2.02 × 10− 5 ms− 1 for the C. nodosa and C. prolifera habitats, respectively.

Table 2. Summary of specific uptake rates, biomass, and areal uptake for each of the community components under natural hydrodynamic conditions (uz = 1 m = 0.04 m s− 1) [mean ±  SD (n)]. Incubation time was 0.67 h, and in situ [NH4+] was 11 μmol L− 1. Superscript letters highlight significantly different specific uptake rates (Behrens–Fisher test, p < 0.05). Biomass estimates for Gracilaria sp. and C. nodosa roots were unavailable
Component Uptake (μmol NH4+ gDM− 1 h− 1) Biomass (gDM m− 2) Areal uptake (μmol NH4+m− 2 h− 1)
Cymodocea nodosa epiphytes 16.58 ± 11.47a (23) 46 ± 23 (4) 762 ± 699
Cymodocea nodosa leaf 4.10 ± 1.99b (23) 258 ± 50 (4) 1055 ± 559
Caulerpa prolifera above sediment 3.85 ± 2.88b (16) 200 ± 25 (4) 772 ± 590
Gracilaria sp. 1.01 ± 1.72c (13)
Cymodocea nodosa root 0.29 ± 0.30c (23)

Discussion

This study is one of the first attempts to measure in situ, specific NH4+ uptake rates of benthic macrophytes and their associated phototrophic community under natural, undisturbed flow conditions. Uptake rates of Cymodocea nodosa leaves (collected from the same area) matched well with measurements made at a free stream velocity of 0.05 m s− 1 in a flow chamber (3.57 ± 0.8 μmol NH4+ gDMleaf− 1 h− 1; Morris et al. 2008). Ammonium mass transfer coefficients (equation image) for the C. nodosa (4.76 × 10− 5 ms− 1) and Caulerpa prolifera (2.02 × 10− 5 ms− 1) habitats were similar to values derived by Cornelisen and Thomas (2009) (equation image = 0.00038 Ub0.69, equation image (for Ub = 0.03) = 3.38 × 10− 5m s− 1). Overall, uptake rates measured in this study fall within the range of reported values for seagrasses (Touchette and Burkholder 2000; Alexandre et al. 2011; Van Engeland et al. 2013), epiphytes (Cornelisen and Thomas 2002; Van Engeland et al. 2013), and macroalgae (Hein et al. 1995; Malta et al. 2005; Van Engeland et al. 2013).

Clear differences in flow properties were observed at the transition between canopy types. Hydrodynamic measurements suggested that longitudinal advection, as well as turbulent diffusivity, increased within the C. nodosa bed (Fig. 5). Vertically averaged equation image within the C. nodosa canopy was 0.017 ± 0.004, 0.019 ± 0.012, and 0.025 ± 0.009 ms− 1 at x = 0, 0.69, and 1.46 m, respectively, and Kz at x = 1.46 m (0.01 m2s− 1) was much higher than the other positions (0.002 m2s− 1). Hence, the prediction that flow over the transition between the dense C. prolifera canopy and the sparse C. nodosa canopy will lead to an acceleration of near-bed water velocities appears to be reasonable. The vertical profile at x = − 0.77 m and a similar pattern in the profile at x = 0 m also suggest that flow adjustment began at length scale comparable to the C. nodosa canopy height (0.5–0.7 m) upstream of the leading edge. However, despite these longitudinal differences in hydrodynamic properties, we were unable to detect significant longitudinal variation in specific uptake for C. nodosa leaves or epiphytes (Fig. 6). This was probably a consequence of the relatively small differences in flow velocity, resulting in small observable effects on uptake compared with the high within-replicate variability of measurements. Higher uptake rates of C. prolifera at the boundary (x = 0.5 m) compared with upstream of the C. nodosa patch were observed, suggesting a possible mechanism that may favor the expansion of C. prolifera; however, this effect was difficult to assess because of incomplete mixing of the tracer close to the release point (i.e., no significant uptake at x = − 1 m).

In contrast, the relatively larger variations in vertical velocity experienced by biotic components of the community at different heights above the bed indicate that water velocity is a relatively good predictor of uptake (Fig. 7). By modeling equation image at the corresponding x–z coordinates for each community component where uptake measurements were collected, we can see that specific uptake was significantly correlated with equation image (Pearson's product-moment correlation, r = 0.78, t15 = 4.9, p < 0.001). Hence, the higher uptake rates of epiphytes can be explained by their elevated position on the surface of leaves extending into the faster-flowing region at the top of the canopy. Likewise, the lack of significant uptake by Gracilaria sp. was probably related to its vertical position within the lower canopy of the C. prolifera bed where flow velocities (Hendriks et al. 2010) and vertical diffusivity are low (Nepf and Ghisalberti 2008).

Details are in the caption following the image

Specific nitrogen uptake rates (mean ± SD) plotted against time-ensemble mean water velocity (equation image, m s− 1) at the depth region corresponding to each community component between x = − 1 and 2 m (see text for details). Water velocities corresponding to the position of each component were extracted from an interpolation of vertical profiles collected at x = − 0.77, 0, 0.69, and 1.46 m.

The relative contributions of longitudinal advection and turbulent diffusion in determining mass transport can be summarized by comparing their respective time scales, Lx/Ub and L/K, using the dimensionless turbulent Péclet number, Pe = UbL2/KLx, where Ub = 0.03 m s− 1, Lx is the characteristic length scale of the C. nodosa patch (4 m), and L is a characteristic length scale for diffusion. Considering the sampling area and the mean canopy height of the C. nodosa patch as the characteristic length scales (x = 4 m, y = 2 m, and z = 0.6 m) gives Pe values in the x-, y-, and z-directions of 11, 3.5, and 0.9, suggesting that advection dominates longitudinal and lateral patch-scale transport. This observation is consistent with that of Lara et al. (2012), who estimated similar lateral K (between 1 and 5 × 10− 3 m2 s− 1) and Pe>1 in adjacent seagrass communities by using independent techniques. At the scale of the C. nodosa canopy height vertical Pe is close to 1, suggesting that neither process is dominant. Taking a length scale of 0.1 m (i.e., C. prolifera canopy height), vertical Pe is < < 1 suggesting that the contribution of turbulent diffusion to mass transport is much larger for the C. prolifera canopy.

Significance to Aquatic Environments

The labeling of a submerged Caulerpa proliferaCymodocea nodosa landscape in the field with a mix of 15NH4+ and uranine enabled the measurement of in situ specific NH4+ uptake rates of benthic macrophytes and their associated phototrophic community under undisturbed flow. The transition from a C. prolifera to a C. nodosa bed included both a change in benthic canopy properties (short and dense to tall and sparse) and sediment topography (0.2 m increase in water column depth) that resulted in an increase in longitudinal advection and turbulent diffusivity within the C. nodosa canopy between 0.5 and 1.5 m from the leading edge. Nevertheless, we were unable to clearly identify significant longitudinal differences in N uptake rates. However, vertical differences in flow velocities related to the position of organisms (i.e., near the top vs. near the bottom of the canopy) appeared to be a good indicator of uptake. These results highlight the important role that hydrodynamics plays in controlling the transport of nutrients to the benthos and how the interplay between the physical structure of organisms and flow within the benthic boundary layer may determine the nutrient transport niches of individual species within macrophyte communities. Interpretations and models of macrophyte ecosystem functioning should incorporate the spatial heterogeneity created through the interaction of canopy morphology, bathymetry, and hydrodynamics.

Acknowledgments

This research was supported by regional government of Andalusia project FUNDIV (“Functional diversity of submerged macrophytes: Influence on carbon and nitrogen processing”) (P07-RNM-2516), the Spanish project IMACHYDRO (“Interactions between marine macrophyte beds and hydrodynamics: From organisms to ecosystems” CTM2008-00012/MAR), a travel grant from the Schure-Beijerinck-Popping Fund (SBP/JK/2007-32), and the Netherlands Organization for Scientific Research. Thanks to P. García-Marin and I. Hernandez (Project EcoLagunes (Environmental Management of Aquaculture Achievement Wetlands), INTERREG-SUDOE-SOE1/P2/F153) for supplying epiphyte biomass values, Fidel Echevarría Navas (director of Centro Andaluz de Ciencia y Tecnología Marinas) for granting us access to facilities, to Bas Koutstaal for helping with sample processing, and to L. Del Río Rodríguez for supplying aerial photographs. E.P.M. is presently supported by a Junta para la Ampliación de Estudios (JAE-Doc-2010) contract, funded in part by the European Union (European Social Fund, ESF2007-2013) and the Spanish Ministry for Economy and Competitiveness. This is CEIMAR journal publication no. 34.