Introduction
Intertidal estuarine mudflats are transitional areas between land and sea.
This intermediate position explains the important horizontal, vertical (in
the sediment column) and temporal heterogeneities in physical and chemical
sediment properties. It also causes heterogeneous ecological niches with
scales ranging from micro- to hectometres. When studying such heterogeneous
environments, the observational scale has to be chosen as a function of the
scale of the studied ecological niche variability (Wu et al., 2000; Morse et
al., 2003; Martiny et al., 2006; Wu and Li, 2006). This is a fundamental
prerequisite to further identify potential parameters controlling the
heterogeneity of the niches.
Ecological studies of benthic foraminifera attempt to describe the main
factors controlling foraminiferal communities, as well as their variability on
different spatial and temporal scales (Buzas et al., 2015). The best
described pattern concerns the spatial variability of their vertical
distribution in open marine environments, on a 100 km scale. The
conceptual model proposed by Jorissen et al. (1995) considers a regional
variability of the spatial organization of foraminiferal taxa in the sediment
column, where they occur in a succession of so-called microhabitats. The
stratified succession of inhabited sediment layers is supposed to be a
response to oxygen and organic matter availability, which changes not only vertically
in the uppermost sediment but also geographically, when changing from
oligotrophic (deep water, offshore) to eutrophic (shallow water, nearshore)
conditions. In estuarine areas, on smaller scales, other major controls are
invoked (e.g. emersion time, grain size, salinity), but they are less well
documented. At the kilometre scale, the salinity, salinity variations and more
generally the frequency of chemical exchanges with the ocean are often
invoked as controls of foraminiferal assemblages (Debenay and Guillou, 2002;
Debenay et al., 2006). Within the estuary, especially in cross-shore
transects, emersion time seems to be a major controlling factor of species
distribution at the decametre scale (Berkeley et al., 2007). But other
parameters, such as grain size, pH or organic carbon lability, could also have
a significant impact. Estuarine foraminiferal faunas seem to show substantial
patchiness at the metre scale at the sediment surface (Buzas, 1970; Hohenegger et
al., 1989; Buzas et al., 2002, 2015). At the decimetre scale, the rare studies
performed on intertidal mudflats highlight that grain size and topography
could be important controls (Lynts, 1966; Morvan et al., 2006).
Finally, according to our knowledge, only three publications have analysed
the spatial surface organization at the centimetre scale, using an adequate
sampling grid (Buzas, 1968, in Rehoboth Bay, Delaware; Olsson and Eriksson,
1974, on the Swedish coast; and de Nooijer, 2007, in the Wadden Sea). These
three studies show that foraminiferal densities present a patchy
distribution. Buzas (1968) hypothesized that this could be due to individual
reproduction, leading to very localized and intermittent density maxima, so-called “pulsating patches” (Buzas et al., 2015). Another field approach, at
the centimetre scale, is to sample around inhabited burrows, using a
non-regular sampling scale, by defining position, size and shape of each
sample according to the burrow geometry. In this way Aller and Aller (1986)
and Thomsen and Altenbach (1993) studied the foraminiferal distribution
around macrofaunal burrows at subtidal stations and observed a 3-fold
enrichment of foraminiferal density in the burrow walls. With a similar
sampling strategy, Koller et al. (2006) showed a 300-fold
enrichment of foraminiferal densities in the burrow walls of an intertidal
station. These studies highlight the importance of macrofaunal activity at
the centimetre scale as a potential control of foraminiferal spatial
organization. They suggest the presence of oxic microenvironments around the
burrows generated by bioirrigation, attractive because of organic matter
enrichment (Aller and Aller, 1986). Foraminifera could specifically colonize
these environments favourable for aerobic respiration and therefore be found
at depths below oxygen penetration.
However, another possible explanation for the presence of rich foraminiferal
faunas in deeper anoxic layers could be the ability of some species to switch
to alternative (e.g. anaerobic) metabolisms (Leutenegger and Hansen, 1979;
Bernhard and Alve, 1996; Risgaard-Petersen et al., 2006; Heinz and Geslin,
2012). These two possible mechanisms lead to contrasting conclusions
concerning ecological strategies. For example, a high density of living
foraminifera along burrow walls compared to anoxic surrounding sediments may
be explained by a positive response of the foraminiferal community to the
availability of oxygen and labile organic matter (Aller and Aller, 1986;
Loubere et al., 2011) or as the involuntary consequence of passive downward
transport due to macrofaunal bioturbation followed by the development of a
short-term survival strategy based on a metabolism modification (Douglas,
1981; Alve and Bernhard, 1995; Moodley et al., 1998). In situ distribution
can answer this question by determining whether subsurface high density is
only concomitant with burrows or whether living A. tepida are able
to modify their metabolism in order to survive in suboxic environments
(without both oxygen and sulfide) independently of burrows. Unfortunately,
the sampling strategies used in the above-mentioned references did not allow
for establishing the importance of burrows compared to other environmental
physico-chemical parameters because the increased density observed in burrow
walls was not compared to a “background heterogeneity” at the same scale.
This precaution is necessary, especially when the increase in foraminiferal
density is not at least of 1 order of magnitude. Consequently, a large
uncertainty remains about the ubiquity and the nature of
macrofauna-independent mechanisms that could cause foraminiferal
heterogeneity.
The recent development of pore water sampling techniques with high resolution
in two dimensions offers the advantage of providing simultaneously
geochemical information on vertical and horizontal submillimetre scales
(Stockdale et al., 2009; Santner et al., 2015). Several studies have
evidenced important spatial variability of dissolved iron release into pore
water (Jézéquel et al., 2007; Robertson et al., 2008; Zhu and Aller,
2012; Cesbron et al., 2014). This can be due to iron oxide consumption caused
by local labile organic matter patches that favour anaerobic respiration (by
dissimilatory bacteria; Lovley, 1991) or by enhancement of sulfide transport
from the deeper layers through burrows and subsequent abiotic dissolution
(Berner, 1970). Conversely, macrofaunal water renewal is also likely to bring
oxic water into the burrows, which consumes reduced dissolved iron and
replenishes the stock of iron oxide. Direct burial of iron oxide by
macrofauna may also contribute to the replenishment (Burdige, 2011). The
overall role of macrofaunal activity on the sedimentary iron cycle is still
unclear (Thibault de Chanvalon et al., 2015; Robertson et al., 2009).
Phosphorus is also likely to have a heterogeneous geochemical pattern. Very
marked centimetre-scale patches have been reported (Cesbron et al., 2014),
apparently due to nutrient recycling from organic matter. However, iron oxide
dissolution can also release adsorbed phosphorus according to a P / Fe ratio of up to ∼ 0.2 (based on ascorbate extractions; Anschutz et al.,
1998), which can be compared to the theoretical anaerobic respiration ratio of
P / Fe ∼ 0.002 (Froelich et al., 1979). Using geochemical
fingerprints, the combination of submillimetre resolution analyses of
dissolved iron and phosphorus is thus likely to (1) confirm the burrow
activity (iron oxidation) and (2) identify potential hotspots of organic
matter consumption (phosphorus production independent of iron).
In the present paper, we present a new two-dimensional sampling technique
allowing (1) the investigation of the relation between benthic
foraminifera and dissolved iron, (2) analysis of the heterogeneity
of foraminiferal distribution and (3) identification of the scale of
potential controls such as active burrows or labile organic matter patches.
Material and methods
Site description
The Loire estuary (NW coast of France) is hyper-synchronous: it shows an
increasing tidal range upstream (Le Floch, 1961) reaching a maximum spring
tidal range of about 7 m at 40 km from the mouth. At Donges (in the high
tidal range area, right shore) the daily surface salinity range is about 20.
Seasonally, surface salinity fluctuates from 0 during floods to 30 during
low-water periods (SYVEL network, GIP Loire Estuaire). On the opposite shore,
the largest mudflat of the estuary (“Les Brillantes”, ∼ 1350 ha)
extends downstream from the city of Paimbœuf. During high tide,
hydrodynamics (tide, wind-induced waves, flow) constrain the sediment
deposition/resuspension cycle, whereas during low tide, biological factors
(bioturbation, biofilm stabilization, benthic primary production; Round,
1964; Vader, 1964; Paterson, 1989) become more important and generate
sediment burial and chemical transformations. Microphytobenthic biofilms vary
annually between 20 in January and 60 mg m-2 in July
(Benyoucef et al., 2014). Our sampling site
(47∘16′56.00′′ N, 2∘3′47.00′′ W) is located on Les Brillantes mudflat, below the mean high water neap tide
(MHWNT) level, about 20 m offshore from an active 1 m high eroded
cliff. Sediment is mainly composed of silt (92 %), with some clay
(6 %) and sand (2 %) (Benyoucef, 2014).
We sampled in May 2013, 2 weeks after a major flood (discharge volume at
Paimbœuf > 2500 m3 s-1, hydro.eaufrance.fr). During sampling,
the river discharge was 835 m3 s-1 on average. Air temperature
was 12.7 ∘C, the weather was cloudy, and salinity in the surface
waters of the main channel ranged from 0.6 to 20 (data from SYVEL network).
Sediment samples were collected at the beginning of low tide. Porosity
decreased from 0.917 to 0.825 in the first 5 cm (Thibault de Chanvalon et
al., in preparation). The calcite saturation state, calculated from alkalinity,
sodium and calcium concentrations and pH (Millero, 1979, 1995; Mucci, 1983;
Boudreau, 1996; Mucci et al., 2000; Hofmann et al., 2010) was above 1.0 until
9 cm depth (data not shown). The macrofauna was mainly composed of
Hediste diversicolor (Annelida: Polychaeta, 630 individuals m-2) and
Scrobicularia plana (Mollusca: Bivalvia, 70 individuals m-2)
(I. Métais, personal communication, 2015).
1-D sampling and processing
Four cylindrical cores (diameter 8.2 cm) were sampled using Plexiglas tubes.
The first two cores were dedicated to foraminiferal analysis and were sliced
immediately after sampling: every 2 mm from 0 to 2 cm and every
half centimetre between 2 and 5 cm. Surface microtopography induces high
uncertainty in the volume of the upper slice. Within 1 h after
retrieval, in order to distinguish living foraminifera, sediments were
incubated with the staining molecule CellTracker™ Green (CTG) in a
final concentration of 1 µmol L-1 in 50 mL of estuarine
water for 10–19 h (Bernhard et al., 2006). CTG is a
non-fluorescent molecule which is hydrolysed by nonspecific esterases,
producing a fluorescent compound. After incubation, samples were fixed in
3.8 % borax-buffered formalin and stored until analysis. In the
laboratory, samples were sieved over 315, 150, 125 and 63 µm
meshes, and the 150–315 µm fraction was examined using an
epifluorescence stereomicroscope (i.e. 485 nm excitation, 520 nm emission;
Olympus ZX12 with a fluorescent light source (Olympus U-RFL-T) or Nikon SMZ 1500
with a PRIOR Lumen 200). All foraminifera that fluoresced continuously and
brightly were wet-picked, air-dried, identified and counted.
The two other cores were used to constrain geochemistry. The first core was
dedicated to microelectrode profiling and solid-phase geochemistry. The solid
phase was characterized by total organic carbon and reactive iron, manganese
and phosphorus, extracted by an ascorbate reagent (buffered at pH 8) over
24 h (Kostka and Luther III, 1995; Anschutz et al., 1998, 2005; Hyacinthe et
al., 2001; Hyacinthe and Van Cappellen, 2004). For further details, see the Supplement
(S1). Oxygen was analysed with Clark-type electrodes (50 µm tip
diameter, Unisense©, Denmark) within the first 5 mm at a 100 µm vertical resolution. In the second core, one-dimensional DET (diffusive equilibrium in thin film; adapted from Davison and Zhang, 1994; Krom et al., 1994) probes were incubated over one night to sample dissolved sodium, iron, manganese and phosphorus. Gel samples were eluted
in 0.01 M HNO3 and analysed by inductively coupled plasma atomic emission spectroscopy (ICP-AES). Salinity was estimated from
sodium concentration. For further details, see Supplement S2.
2-D sampling and processing
For the two-dimensional sampling, we used a “jaw device”, composed of two
main parts (jaws; Fig. 1). The first jaw is a DET gel probe which samples
the dissolved chemical species from the pore water at high resolution,
whereas the second jaw samples a 2 cm thick slice of the adjacent sediment,
from which we subsampled 1 cm3 aliquots for foraminiferal analysis.
The first jaw is a 250 mm × 200 mm × 2 mm polycarbonate plate with a central depression of 1 mm that
holds a 2-D gel probe. The probe is made of two layers: (1) a
180 mm × 97 mm × 0.92 mm polyacrylamide thin film
prepared and rinsed with Milli-Q water (Krom et al., 1994) which reaches
equilibrium in a few hours once incubated (called “2-D DET gel”) and (2) a
PVDF porous (0.2 µm) membrane to protect the gel and prevent it from falling
out of the depression and control diffusion. The 2-D DET gel was prepared and
mounted less than 1 week before sampling, conserved in a wet clean
plastic bag, and then de-aerated by N2 bubbling for about 6 h before
deployment. The sampler was deployed into the sediment at low tide. On both
lateral sides of the central depression (Fig. 1), plastic rails (2 cm high)
were fixed in order to guide the second jaw to slide along the plate. The
second jaw is a stainless steel plate (1.5 mm thick) bent on both sides.
After equilibration (5 h) of the 2-D gel, the second jaw was inserted along
the guides of the first jaw and the whole device was gently pulled out of the
sediment. Once onshore, the 2-D gel was separated from the sediment, covered
with a plastic-coated aluminium plate and stored in an icebox with dry ice
pellets (Cesbron et al., 2014) until final storage in a freezer
(-18 ∘C).
Schematic view of the “jaw device” for simultaneous sampling of
sediment and porewater.
The sediment plate was manually cut (with stainless steel trowels) within
30 min in 1 cm3 cubes for a surface of 8 cm × 8 cm. The
resulting sampling map is presented in Fig. 2 together with the 1-D sampling
scheme of foraminifera. Next, these sediment cubes were labelled with CTG so
that living foraminifera could be recognized (as for the core slices; see Sect. 2.2).
Considering an error of 1 mm for each cut, the volume uncertainty was
∼ 14 %, except for surface samples where the microtopography of the
sediment surface considerably increases volume uncertainty.
Sediment sampling methodology for living foraminiferal analyses. (a)
Usual 1-D hand coring and layer slicing. (b) Sediment plate sampling with the
second jaw of the “jaw device” (Fig. 1) and representation of the sediment
cubic slicing.
The 2-D DET probe was analysed in order to obtain the concentrations of
dissolved iron and dissolved reactive phosphate (DRP) (Cesbron et al., 2014).
After thawing at ambient temperature, the sample gel was quickly recovered
by a reactive gel equilibrated in specific colorimetric reagents. Twenty-five
minutes after contact, a photograph (reflectance analysis) of superposed gels
was taken with a hyperspectral camera (HySpex VNIR 1600) and analysed. The resolution (surface area of pixels) was
211 µm × 216 µm. The estimated incertitude is
10 % for iron and 11 % for DRP. For further details, see Supplement S3. To
compare the geochemical species distribution (at submillimetre resolution)
and foraminiferal density (at centimetre resolution), an R code was
written that allowed for the downscaling of chemical resolution from 0.2 mm to
1 cm.
Statistical analyses
The patchiness effect or autocorrelation, interpreted as the fact that the
density of one square depends on its neighbours, was explored using spatial
correlograms built using Moran's index (I), computed with R (package
“spdep”, following Fortin and Dale, 2005; Bivand et al., 2008; Legendre and
Fortin, 2010; Borcard et al., 2011; Eq. 1). This index was applied to
benthic meiofauna by Blanchard (1990) and Eckman and Thistle (1988) and to
foraminifera by. Hohenegger et al. (1993). This index calculates the
similarity of pair values for one neighbourhood – a neighbourhood being
defined by a weight (wi,j) function of the distance (d) between pairs.
Id=∑i,jnwi,jdxi-x¯(xj-x¯)∑inxi-x¯2×n∑i,jnwi,j(d)
Here, the n=40 cubes used for Moran's index have neighbourhoods defined
as cubes in direct contact (four neighbours per sample with a weight of 1 and the others have 0, also known as “rook connectivity”; Fortin and Dale, 2005).
With this configuration, Moran's index is -1 for a contrasted organization
(perfect negative correlation between neighbours) and +1 in the case of grouped
organization (perfect positive correlation between neighbours). A value close
to zero (I0=(n-1)-1) corresponds to no organization or random
distribution. The correlogram plots Moran's index versus the order of the
neighbours (o.n.). A decrease in the Moran's index from positive to negative
values characterizes a patchy distribution. The characteristic length of the
patchiness is defined as the order of neighbours when Io.n.=0
(Legendre and Fortin, 1989). Two-dimensional non-random organization was
tested with the alternative hypothesis: Io.n. > I0. The
second test examines if there is a preferential direction in the organization
(isotropy). Again, the alternative hypothesis
for Moran's index, Io.n. > I0, is used, restricting the distance to the tested dimension
(vertical or horizontal). Thus, in our case, each sample was compared only
with its lateral or vertical neighbours (i.e. two neighbours per test).
Results
1-D geochemical features
Figure 3 shows both solid and dissolved chemical species obtained from the
dedicated cores. Total organic carbon (Corg, black circles,
Fig. 3a) decreased from 2700 to 1900 µmol g(dry sediment)-1 in the first centimetre, then
increased sharply until 1.5 cm depth, and finally decreased progressively
from 2700 to 2400 µmol g(dry sediment)-1 at 5 cm depth. Salinity (Fig. 3a) ranges from 7.5 to 1.7 with an
offset of ∼ 2 between replicates and a decrease of ∼ 3 in the
first 13 cm. Figure 3b shows the vertical distribution of dissolved
oxygen. The 3 profiles shown (out of 18) are considered representative of
the lateral variability in the sediment. Most of the oxygen concentration
profiles show the exponential trend typical for undisturbed marine sediments
(two profiles in Fig. 3b, with light-grey and white diamonds; Revsbech et al.,
1980; Berg et al., 1998). However, one-third of the O2 profiles diverged
from the exponential model, showing an interruption of the decreasing trend,
or even a local increase, at depth (e.g. the profile with dark-grey diamonds
represented in Fig. 3b). The oxygen penetration depth (OPD) remained
relatively constant around 2.0 mm (SD = 0.2 mm, n=18) despite this
heterogeneity.
1-D geochemical features. (a) Vertical profile of total solid organic
carbon (filled circles, uncertainty smaller than symbol size) and profiles of
salinity (white and grey diamonds). (b) Typical profiles of dissolved
oxygen; the profile with dark-grey diamonds is considered bioturbated. (c, d, e)
Vertical profiles of manganese (c), iron (d) and phosphorus (e) in dissolved
(white and grey diamonds for DET replicates) and reactive solid phase
(ascorbate-leached) from the core (black circles).
Figure 3c, d and e show the distribution of manganese, iron and
phosphorus, respectively, both in the dissolved phase (grey and open
diamonds) and in the easily reducible solid phases (black circles, extracted
by ascorbate leaching; Anschutz et al., 2005; Hyacinthe et al., 2006).
Extracted manganese (mainly (hydr)oxide, black circles in Fig. 3c) showed a
strong enrichment of the easily reducible solid phase (until
13 µmol g(dry sediment)-1) in the first 2 mm, where an
important upward diminution was visible in both replicates of the dissolved
phase (grey and open diamonds in Fig. 3c). Below, the solid phase showed a
slightly decrease from 7.9 to 5.6 µmol g(dry sediment)-1 until 5 cm
depth. The dissolved manganese concentration decreased between 4 and 9 cm
depth in both replicates (from 70 to 30 µmol L-1). In the
solid phase, iron, phosphorus and manganese are strongly correlated when the
surface sample is not considered (r2=0.70 between iron and
manganese and r2=0.55 between and iron and phosphorus). Profiles of dissolved iron
and phosphorus are also strongly correlated (r2=0.90, slope = 1.87
and r2=0.47, slope = 1.31 for replicates A and B). Iron and
phosphorus were remobilized, and therefore appeared in the dissolved phase,
between 1 and 9 cm. Both replicates of dissolved iron showed the same four
well-described maxima (at least six samples for each maximum) at 2.3, 3.3,
5.9 and 7.3 cm depth but with different concentrations. In replicate A (open
diamonds) these maxima have 5 times higher iron concentrations (up to
700 µmol L-1) than in replicate B.
Visual features on the sediment plate
Figure 4a shows the sediment slice obtained from the “jaw device” facing
the 2-D DET gel. In order to facilitate the description, the figures were
subdivided into centimetre squares labelled with letters for the horizontal
position and numbers for the vertical position. The black rectangle
corresponds to the 2-D DET gel position, the blue rectangle to the gel signal
exploited and the red rectangle to the 2-D foraminiferal sampling. Burrows
parallel to the cutting plan are visible over their entire length. When
perpendicular to the cutting plan, they appear as a dark hole (B14 in
Fig. 4a). Figure 4b summarizes burrow distributions superimposed on a picture
of the gel after equilibration with the colorimetric reagents (pink
coloration corresponds to iron and blue to dissolved reactive phosphorus
(DRP)). Five burrows were visibly connected to the sediment surface; their
traces mostly extended vertically down to 10 cm depth, where their track is
lost. Between 10 and 15 cm depth, visible burrow density decreased. Below
15 cm depth, burrows were rarely observed and the sediment was dark
(Fig. 4a). During slicing of the sediment plate, living polychaetes
(Hediste diversicolor) were observed in some burrows.
(a) Picture of the sediment plate before cube slicing for
foraminiferal analysis (sediment–water interface at the top). (b) Picture of
the analysed gel after colorimetric reactions: dissolved iron shown in dark
pink and dissolved phosphorus in turquoise (burrows superimposed). The black
rectangle corresponds to the gel limit, the blue rectangle to the limit of
available data set of dissolved iron and phosphorus and the red rectangle to
the limit of the available data set of foraminiferal distribution.
2-D DET gel
Figure 5 shows the two-dimensional data sets, with the distribution of dissolved
phosphorus (Fig. 5a) and iron (Fig. 5b) obtained from the 2-D DET gel. For
comparison, burrow distribution is shown in Fig. 5a. Dissolved iron and
phosphorus both appeared a few millimetres below the sediment–water
interface. They are positively correlated for the whole plate (r2=0.59,
slope = 2.7). Despite their patchy distribution, both species can be
observed along the entire length of the gel probe (i.e. 17 cm depth). A main
feature was the occurrence of two prominent vertical structures enriched in
dissolved iron and phosphorus (A-B/6–9 and F-G/5–14; Fig. 5). The highest
concentrations, of about 170 and 50 µmol L-1 for iron and
phosphorus, respectively, were found within the structure at the right
(squares F/8–9). In the structure on the left (A/6–8), iron and phosphorus
maxima were around 120 and 25 µmol L-1, respectively.
(a, b) Two-dimensional concentrations after numerical analysis of
dissolved reactive phosphorus (DRP) and dissolved iron. The distribution of
burrows is shown on the DRP plot. Red lines represent the boundary of
foraminiferal analysis. (c) 2-D distribution of A. tepida densities
from the sediment plate with burrow distribution.
Most burrows seem to impact the iron concentration. For example, burrows 1, 3
and 5 clearly correspond (in the first 4 cm) to a drastic decrease in
or even disappearance of dissolved iron, whereas other burrows seem to
correspond to a dissolved iron enrichment (F-G/5–9). However, some
centimetre-sized patches (e.g. A-B/6–9, H-G/8–9 and F-G/17) seem to be
unrelated to burrow structures. Below 15 cm depth, the sediment was dark and
dissolved iron generally decreased, whereas DRP increased.
Living foraminiferal distribution
Figure 5c shows the distribution of CTG-labelled Ammonia tepida
determined for 1 cm3 samples in the sediment facing the 2-D DET gel.
The analysis of living foraminifera in the 64 cubes (8 cm
width × 8 cm depth) takes roughly the same time as the analysis of
one core of 8.2 cm diameter (until 5 cm depth). Ammonia tepida
was by far the dominant species, accounting for 92 % of the total
assemblage. The second most frequent species, Haynesina germanica,
represented 5 %, but its low density (mostly 0, 1 or 2 individuals per
cubic centimetre) was not sufficient to support a serious discussion. For
this reason the data relative to this species are omitted from the present
paper. A. tepida density ranged from 0 to 38 individuals cm-3 with
important lateral and vertical variability. The relative standard deviation
(RSD) calculated for each row is, on average, 45 %, whereas for each
column the RSD is 60 %, suggesting a slightly more pronounced vertical
organization. This is confirmed by the stratification of the richest samples
(≥27 individuals cm-3), which were found in the topmost centimetre and below
6 cm depth, whereas the poorest samples (≤5 individuals cm-3) were found
between 1 and 3 cm depth. Each row from the 2-D distribution can be
represented by a box-and-whisker plot (Fig. 6). The results confirm a three-step
pattern with high densities at the surface (13 to 38 individuals cm-3), lower
density between 1 and 3 cm depth (0–12 individuals cm-3 and one outlier at
24 individuals cm-3) and increasing values below 3 cm (7 to
31 individuals cm-3).
This vertical pattern is also visible in the two studied sediment cores
(Fig. 6): high densities of A. tepida (26 ± 0 individuals cm-3)
are observed in the first 2 mm, with a rapid decrease to minimal densities in the
1.0–1.2 cm layer (3 ± 0 individuals cm-3), followed by a progressive,
somewhat irregular increase until 9 ± 0 individuals cm-3 below 2 cm
depth to 8 cm depth. Despite the different vertical sampling resolution, the
densities observed in the cores are in agreement with the average densities
observed in the sediment slice cubic samples.
Vertical comparison of A. tepida densities from the two cores (filled and open
triangles) and the “jaw device” sampling (each box plot represents the
distribution of one layer; bars are first and third quartiles for the boxes
length and whiskers are below 1.5 interquartiles; open circles are
outliers).
Discussion
A methodological improvement to characterize heterogeneity
Here, we present for the first time a methodology allowing the simultaneous
study of the vertical and horizontal heterogeneity of dissolved chemical
species and living foraminifera (determined by CTG labelling) in the first 8 cm of the sediment. Figure 6 compares the vertical density
distribution of A. tepida between the cores (triangles) and the jaw
device (box-and-whisker plots), sampled a few decimetres apart. Despite the different
vertical sampling resolution, the densities observed in the cores (sampling
surface of 53 cm2) are in agreement with the average densities observed
in the sediment slice samples (sampled with the “jaw device”, sampling
surface of 8 cm2). This similarity suggests a limited horizontal
heterogeneity of A. tepida at the decimetre scale, although it is
impossible to draw firm conclusions on the basis of only three samples (the
two cores and the jaw device).
The jaw device (box plot whiskers, Fig. 6) reveals a heterogeneous horizontal
distribution at the centimetre scale. The centimetre-scale heterogeneity is
quantified by calculating the Moran's index that estimates the characteristic
length of foraminiferal niches. Figure 7 shows Moran's index correlograms
applied between 3 and 8 cm depth (suboxic sediment), where high densities of
living foraminifera were observed. Figure 7a shows that the spatial
organization of A. tepida is patchy at the centimetre scale
(I1=0.24, p value = 0.013). For farther neighbours the Moran's
index values drop to zero, meaning random organization. Concerning
vertical and horizontal heterogeneities, Moran's index values for direct
neighbours are 0.02 and 0.47, with p values of 0.38 and 0.001,
respectively. For second-order neighbours, values do not significantly differ
from 0 in either direction (data not shown). This means that A. tepida specimens tend to be grouped in horizontal spots with a
characteristic length of 1 to 2 cm.
Moran's index correlograms for 3 to 8 cm depth. (a) Moran's index
correlogram for A. tepida with a 1 cm resolution. (b) Moran's index
correlogram for [Fe]dissolved with a 1 cm resolution. An asterisk indicates significant
differences from zero; error bars are twice the standard deviation. The
numbers are the number of pairs for each order of neighbours.
Figure 7b shows the Moran's index correlogram for iron at 1 cm scale
resolution (phosphorus is similar and not shown). It shows strong patchiness
(I1=0.7) for direct neighbours in either directions, with a
characteristic length of 3–4 cm. The fact that the characteristic lengths
of A. tepida (Fig. 7a) and dissolved iron (Fig. 7b) patches are
longer than 1 cm suggests that the impact of different sampling thicknesses
(roughly zero for dissolved iron compared to 1 cm for foraminifera) would not
result in major bias. Moreover, this characteristic length is important as it
likely corresponds to the characteristic length of the controlling mechanisms
(Clark, 1985; Wu and Li, 2006). In fact, the difference in Moran's index
between chemical species and the A. tepida density distribution
suggests that not exactly the same mechanisms control these parameters. This
is an unexpected result, since most conceptual models explain benthic
foraminiferal distribution in the sediment as a direct response to
geochemical gradients, especially oxygen and sulfide (Jorissen et al., 1998;
Van der Zwaan et al., 1999; Fontanier et al., 2002; Langezaal et al., 2006;
Langlet et al., 2013), which intimately control iron remobilization.
Factors generating chemical heterogeneity
The heterogeneity of geochemical patterns is mainly explained by the
availability of oxidants mineralizing organic carbon. In the generally
applied conceptual model of Froelich et al. (1979), organic matter
remineralization is characterized by a succession of horizontal layers where
specific oxidants are used. Figure 3 confirms this theoretical vertical
stratification: oxygen is rapidly consumed by respiration (about 2 mm depth,
Fig. 3b), and thereafter reduced dissolved manganese appears (Fig. 3c). Dissolved iron
appears still deeper, with a first maximum at 2 cm depth. The slopes of the
concentration profiles are steeper and the reactive solid phase (Fig. 3d and
c) is more concentrated for iron than for manganese, suggesting a higher
reactivity. However, the strictly vertical succession of redox layers is no
longer respected in the deeper suboxic layers, as suggested by the presence
of multiple maxima of iron (Fig. 3d) and by the high lateral heterogeneity
observed in Fig. 5a and c. This high lateral heterogeneity cannot be
explained by vertical diffusion of oxygen. It appears therefore that a
strictly vertical stratification of redox zones, defining a similar
foraminiferal microhabitat succession, is not a reasonable assumption for our
study area.
Macrofaunal impact on heterogeneity
Macrofauna is assumed to be the most important cause of chemical
heterogeneity at a scale of 0.01 cm (roughly the foraminiferal scale) to
100 cm (station scale), because of its ability to reorganize the sediment.
In this way, macrofauna determines whether other factors can impact the
heterogeneity of dissolved iron and/or A. tepida. Macrofauna
modifies (i) the sediment texture/composition (burrow walls or faecal
pellets); (ii) the redox conditions, by ventilation of their burrows with
oxygenated water (bioirrigation); and (iii) particle arrangement, by crawling
or burrowing (biomixing) (Meysman et al., 2006). The efficiency of biomixing
to homogenize the sediment mainly depends on two aspects (see Wheatcroft et
al., 1990, or Meysman et al., 2010a, for a more detailed discussion):
(1) The biomixing species assemblage. At the Les Brillantes
mudflat, the main macrofaunal species are Hediste diversicolor (630 individuals m-2) and Scrobicularia plana (70 individuals m-2; I. Métais, personal communication, 2015). H. diversicolor is a
gallery diffusor (particle mixing due to burrowing activity), whereas
S. plana is an epifaunal biodiffusor (particles are mixed in a
random way over short distances along the surface; e.g.
François et al., 2002;
Kristensen et al., 2012). These two species generate homogeneity or
heterogeneity according to the second criterion. See below.
(2) The relation between the average time of existence of the studied objects
(here foraminifera and dissolved iron) in the bioturbated area and the
average time between two bioturbation events. Frequent bioturbation events
generate efficient mixing (homogeneity), whereas rare bioturbation events
generate heterogeneity. The average time between two bioturbation events is
estimated to be days to months by tracer modelling (Wheatcroft et al., 1990;
Meysman et al., 2003, 2008), while the longevity of foraminifera in suboxic
environments is estimated to be roughly 1 year (Langlet et al., 2013; Nardelli
et al., 2014). The mean residence time of iron in the dissolved phase is
estimated between 2 and 3 days (Thibault de Chanvalon et al., in preparation).
Therefore, biomixing should generate a homogeneous distribution of
foraminiferal density distribution, contrasting with a heterogeneous
distribution of dissolved iron (and DRP). The different time spans also
suggest that most of the living foraminifera were already present in the
suboxic sediment before the visible (most recent) burrows were created.
Conversely, the heterogeneity of the dissolved chemical species should be
directly related to biomixing and to others factors that have not been
homogenized by biomixing, i.e. with a short time of existence in suboxic
environments.
Geochemical impact of biogenic factors
The factors likely to generate chemical heterogeneity are (1)
bioirrigation, which mainly causes an increase in oxidant availability (Aller
and Aller, 1986; Aller, 2004; Arndt et al., 2013), and (2) biogenic particles
(e.g. decaying macrofauna, faecal pellets), which cause an increase in labile
carbon availability. Dissolved iron shows two opposite types of behaviour
(Aller, 1982). (1) Iron precipitates as a hydroxide when the
oxidative state of the pore water surrounding active burrows increases
(Meyers et al., 1987; Zorn et al., 2006;
Meysman et al., 2010b). This is confirmed by visible burrows in Fig. 5 in
which both dissolved iron and DRP are depleted (Fig. 4, numbers 1, 3, 5
(above 6 cm depth) and burrows in B-D/13, E/9–11, G-H/10–15 and A-B/9). These
structures are mainly vertical and have a length often exceeding 3 cm, in
agreement with the Moran's index correlogram. Conversely, within the long
burrow F-G/5–9, dissolved iron is enriched, indicating that this burrow is
abandoned and no oxygen renewal occurs. This feature was also observed for
some burrows by Zhu and Aller (2012) and Cesbron et al. (2014). (2) Dissolved
iron is produced by anaerobic respiration where biogenic particles increase
labile carbon availability, and thereby decrease the oxidative state of
surrounding pore waters (Robertson et
al., 2009; Stockdale et al., 2010). The geometry and isolation from visible
burrows of patches A/7–8, G-H/8–9 and F-G/17 in Fig. 5a and b suggest that
they could represent centimetre-wide labile organic matter patches. We
hypothesize that these patches correspond to intense remineralization of
biogenic particles that dissolves iron oxides.
Mechanisms controlling the A. tepida distribution
Figures. 5c and 6 clearly describe a three-step pattern in the distribution
of A. tepida, with high densities at the surface, low densities
between 1 and 3 cm depth, and a somewhat surprising increase below (in
suboxic sediments). A similar pattern has been reported, but not discussed, for
other intertidal environments (Alve and Murray,
2001; Bouchet et al., 2009). In our study, the consistency of the eight vertical
columns from the plate sampling confirms the robustness of this pattern, and
the two-dimensional approach reveals an organization of A. tepida in
centimetre-wide patches in the suboxic sediment. The next subsections discuss
possible mechanisms that could explain these features, especially in the
suboxic environment, where active burrows (supporting biomixing and
bioirrigation) and biogenic particles have been identified as factors likely
to generate such heterogeneity.
Foraminiferal metabolism
Generally, aerobic metabolism is considered the dominant mechanism in oxic
conditions since it is energetically most efficient. In fact, Figs. 5c and 6
clearly describe maximal densities of A. tepida at the sediment
surface (0–2 mm depth) and low densities below (6–18 mm depth). This strong
gradient of A. tepida density highlights the presence of a
continuously oxygenated microhabitat enriched in organic matter (see TOC and
O2 profiles, Fig. 3a–b) close to the sediment–water interface,
favourable for A. tepida. Energetic considerations and some
observations that report a strong seasonal variability in the oxic zone
(Moodley, 1990; Barmawidjaja et al., 1992) led to the assumption that foraminifera
reproduce preferentially in the oxic layer (de Stigter et al., 1999; Berkeley
et al., 2007). Together, these factors explain the maximum density in the
surface layer.
Since the work of Richter (1961), numerous publications have
reported living benthic foraminifera in suboxic sediment layers (Jorissen et
al., 1992; Moodley and Hess, 1992; Bernhard and Sen Gupta, 2003). For
intertidal environments, studies have reported living (rose bengal stain)
foraminifera in subsurface environments since the 1960s (e.g. Buzas,
1965; Steineck and Bergstein, 1979). Several in situ (Goldstein et
al., 1995; Bouchet et al., 2009) and laboratory (Moodley and Hess,
1992; Moodley et al., 1998; Pucci et al., 2009; Nardelli et al., 2014; Nomaki
et al., 2014) studies with A. tepida have also reported survival, activity and
even calcification in suboxic conditions. Anaerobic metabolism would be a
logical mechanism to explain the presence of large amounts of living
foraminifera in suboxic layers. Complete or partial (with endo- and/or
ectobionts; Bernhard and Alve, 1996) denitrification co-occurring with
nitrate storage has been demonstrated for some foraminiferal taxa
(Risgaard-Petersen et al., 2006). Nomaki et al. (2014) suggested
denitrification by endobionts for A. tepida. However,
denitrification has not been measured in A. tepida, and only very
low intracellular nitrate concentrations were found
(Pina-Ochoa
et al., 2010; Geslin et al., 2014). It appears therefore unlikely that the
abundance of living A. tepida in deeper suboxic layers can be explained by active
colonization.
Burying and burrow microenvironment
It is clear that biomixing is a likely mechanism to explain the introduction
of foraminifera in deeper sediment layers by passive transport
(Alve
and Bernhard, 1995; Goldstein et al., 1995; Moodley et al., 1998; Saffert
and Thomas, 1998; Alve and Murray, 2001; Jorissen, 2003). However, the
spatial distribution resulting from this passive transport has never been
well described, or modelled. According to the theory of biomixing, we
suggest that the vertical distribution of A. tepida can be approached with a diffusion
model, which should lead to an exponential downward decrease, with the slope
as a function of the mortality rate. A. tepida is possibly able to survive in suboxic
environments using an intermittent aerobic metabolism, using the oxygen that
can be occasionally available due to bioirrigation
(Fenchel, 1996; Wang et al.,
2001; Wenzhofer and Glud, 2004; Pischedda et al., 2012). Their activity
should progressively decrease once oxygen is depleted; Phipps (2012) suggested that they could finally be immobilized before dying
as a result of a prolonged absence of oxygen supply. We think that repeated
introduction by macrofaunal bioturbation, followed by reduced metabolic
activity, leading to immobilization, is the most likely mechanism to explain
the high abundances of living foraminifera in suboxic sediments.
Figures 4a and 5b show no relation between visible burrows and living
A. tepida. This result is in agreement with the different
timescales of the foraminiferal lifespan and the burrows, and with the idea
that biomixing homogenizes the A. tepida density. It also suggests
that the oxygen obviously brought by formation of new burrows is
consumed too fast to allow all infaunal A. tepida to migrate to
these active burrows. Thus, recent burrow walls are apparently not colonized
by specimens of A. tepida already present in the suboxic sediment.
Our observations contrast with earlier studies, showing increased
foraminiferal densities (up to 300 times higher than in the surrounding
sediment, rose bengal staining) in burrow walls. For example, data from
burrows of Amphicteis sp. at 4800 m depth (Aller and Aller, 1986), of
Echiurus echiurus at 42m depth (Thomsen and
Altenbach, 1993), and of Pestarella tyrrhena in intertidal sand flats
(Koller et al., 2006) all presented high foraminiferal densities. The
observed differences could be due to the fact that burrows of various
macrofaunal taxa may represent very different environmental conditions and
possibly due to a difference in sampling scale, since
Thomsen and Altenbach (1993) and Koller et al. (2006)
applied an irregular millimetre sampling around burrows. To summarize,
macrofaunal activity would explain transport to and survival in suboxic
layers. However, it does not explain the density minimum at 1–3 cm depth.
Sensitivity to geochemical gradients
We think that the most probable explanation for the 1–3 cm density minimum
of A. tepida is an active upward migration of the specimens, back to
the sediment surface, before they are completely immobilized by a lack of
oxygen and a strongly lowered metabolism. Numerous studies have already
reported that vertical migration of foraminifera allows them to move to more
hospitable environments
(Jorissen,
1988; Van der Zwaan and Jorissen, 1991; Alve and Bernhard, 1995; Moodley et
al., 1998; Gross, 2000; Langezaal et al., 2003; Geslin et al., 2004; Ernst et
al., 2005). In an experiment in which populations of Haynesina germanica were uniformly mixed in a 6 cm sediment column, Ernst et al. (2006) saw a clear migration back to the surface for the foraminifera living
between 1 and 3 cm depth, and suggested that foraminifera living at greater
depth were unable to do so. Similarly, Hess et al. (2013) showed that
benthic foraminifera are able to migrate through suboxic sediment to reach
oxic sediments over a maximal distance of a few centimetres. Active migration
towards directly detected oxygen or organic matter over distances beyond
1 cm seems improbable, since this distance is much higher than the typical
pseudopodial length (about 1 cm; see Travis and Rabalais, 1991).
However, as described above, the presence of oxygen could be indirectly
detected by other geochemical gradient (e.g. NO3-, Mn2+ or
Fe2+, dissolved organic carbon, pCO2). However, when gradients
generated by the oxygen front are imperceptible for A. tepida,
because they are living too deep in the sediment, or when such gradients are
hidden by other sources of geochemical gradients (as organic-rich patches),
this upward migration could no longer occur. This could explain why, below
3 cm depth, A. tepida remains in the deeper sediment layer after
being transported there accidentally.
However, the organization of the foraminiferal in 1–2 cm wide horizontal
patches identified by Moran's index suggests that A. tepida detects
not only vertical geochemical gradients but probably also lateral gradients
around degrading biogenic particles. The characteristic length of patches
corresponding to biogenic particles identified by dissolved iron maxima
(A/7–8, G-H/8–9 and F-G/17 in Fig. 7c and d; see Sect. 4.2.2) is in agreement with
the characteristic length of foraminiferal density maxima. For instance, in
the first 8 cm, the two identified biogenic particles patches
(A/7–8, G-H/8–9 in Fig. 5b) both correspond to a higher density of A. tepida (28/19 and 21/30 individuals cm-3 on average for A/7–8 and
G-H/8, respectively, Fig. 5c). In agreement with these results and despite a lowered
metabolism, we hypothesize that foraminifera could move towards patches of
labile organic matter even in deeper suboxic layers. Nevertheless, a better
identification of labile carbon patches, replicate sampling with the here-developed strategy, and experimental studies with artificial geochemical
gradients are necessary to confirm our hypotheses about the behaviour of
A. tepida in suboxic environments.
To summarize, we suggest that the distribution of A. tepida can be
interpreted as the result of not less than five interacting mechanisms
(Fig. 8). (1) High foraminiferal densities at the surface are the result of the
presence of abundant labile organic matter and reproduction in the oxygenated
layer (Sect. 4.3.1); (2) downward transport by macrofaunal biomixing
introduces living foraminifera into deeper sediment layers (Sect. 4.3.2); (3) in the first 3 cm foraminifera are capable to migrate
back to the oxygenated, organic-rich surface layers once they detect redox
gradients, whereas in deeper sediment layers, they are no longer capable to
find their way back to the superficial oxygenated layer (Sect. 4.3.3);
(4) after a prolonged presence in suboxic conditions, foraminifera lower their
metabolism and become inactive; and (5) foraminifera can be temporarily
remobilized during intermittent bioirrigation events, and can eventually
migrate towards organic-rich microenvironments in their vicinity
(Sect. 4.3.3). A better identification of labile carbon patches – for
example based on alkalinity (Bennett et al., 2015), pCO2 (Zhu et al.,
2006; Zhu and Aller, 2010) or dissolved organic carbon – should allow the interpretation to be taken further.
Putative mechanisms explaining the A. tepida density profile (OPD: oxygen
penetration depth).