Observations indicate that resuspension and associated fluxes of organic
material and porewater between the seabed and overlying water can alter
biogeochemical dynamics in some environments, but measuring the role of
sediment processes on oxygen and nutrient dynamics is challenging. A modeling
approach offers a means of quantifying these fluxes for a range of
conditions, but models have typically relied on simplifying assumptions
regarding seabed–water-column interactions. Thus, to evaluate the role of
resuspension on biogeochemical dynamics, we developed a coupled hydrodynamic,
sediment transport, and biogeochemical model (HydroBioSed) within the
Regional Ocean Modeling System (ROMS). This coupled model accounts for
processes including the storage of particulate organic matter (POM) and
dissolved nutrients within the seabed; fluxes of this material between the
seabed and the water column via erosion, deposition, and diffusion at the
sediment–water interface; and biogeochemical reactions within the seabed. A
one-dimensional version of HydroBioSed was then implemented for the Rhône
subaqueous delta in France. To isolate the role of resuspension on
biogeochemical dynamics, this model implementation was run for a 2-month
period that included three resuspension events; also, the supply of organic
matter, oxygen, and nutrients to the model was held constant in time.
Consistent with time series observations from the Rhône Delta, model
results showed that erosion increased the diffusive flux of oxygen into the
seabed by increasing the vertical gradient of oxygen at the seabed–water
interface. This enhanced supply of oxygen to the seabed, as well as
resuspension-induced increases in ammonium availability in surficial
sediments, allowed seabed oxygen consumption to increase via nitrification.
This increase in nitrification compensated for the decrease in seabed oxygen
consumption due to aerobic remineralization that occurred as organic matter
was entrained into the water column. Additionally, entrainment of POM into
the water column during resuspension events, and the associated increase in
remineralization there, also increased oxygen consumption in the region of
the water column below the pycnocline. During these resuspension events,
modeled rates of oxygen consumption increased by factors of up to
Understanding and quantifying the role that physical processes play on coastal water quality remains a scientific and management concern. Management solutions to hypoxia, the occurrence of low oxygen concentrations, as well as other water quality issues, have focused on reducing riverine delivery of nutrients and sediments (Bricker et al., 2007). Yet temporal lags between these reductions and water quality improvements (Kemp et al., 2009), and increased cycling of nutrients within coastal systems (e.g., Testa and Kemp, 2012), indicate that temporary storage of nutrients in the seabed and subsequent release to the water column via diffusion and/or resuspension can affect water quality in some coastal environments. Neglecting these processes impairs managers' ability to develop and evaluate strategies for improving coastal water quality (e.g., Artioli et al., 2008).
Resuspension-induced fluxes of sediment, particulate organic matter (POM), and dissolved chemical species between the seabed and water column can significantly affect biogeochemistry in coastal waters, including oxygen dynamics (Glud, 2008). Entrainment of seabed organic matter and reduced chemical species into the water column can increase remineralization and oxidation rates, thereby decreasing oxygen concentrations in bottom waters in some environments. For example, Abril et al. (1999) observed that oxygen concentrations were inversely correlated with tidal fluctuations of suspended particulate matter concentrations in the Gironde Estuary, France. Recently, Toussaint et al. (2014) collected high-resolution time series of microelectrode oxygen profiles on the Rhône River subaqueous delta that showed resuspension may also increase oxygen consumption in the seabed. This experiment revealed increases in diffusive fluxes of oxygen from the water column to the seabed during erosional events. Other observational studies have estimated resuspension-induced increases in oxygen consumption within the seabed and bottom-waters using measurements of turbulent oxygen fluxes (Berg and Huettel, 2008) and erodibility experiments (e.g., Sloth et al., 1996). Yet, it remains difficult to distinguish and quantify the relative influences of different biogeochemical (e.g., remineralization, oxidation) and physical (e.g., diffusion, resuspension) processes on oxygen dynamics in both the seabed and bottom-waters.
Hydrodynamic–biogeochemical models often complement observational studies of
water quality (e.g., Moll and Radach, 2003; Aikman et al., 2014), but these
simulations usually neglect or simplify seabed–water-column fluxes. Water
quality models often assume that organic matter and nutrients reaching the
seabed are permanently buried, instantaneously remineralized, resuspended
without remineralization, or a combination thereof (e.g., Cerco et al., 2013;
Fennel et al., 2013; Feng et al., 2015; Bruce et al., 2014; Liu et al.,
2015). Yet, numerical experiments showed that switching among relatively
simple parameterization methods for seabed–water-column fluxes can alter the
estimated area of low-oxygen regions by about
We therefore developed a modeling approach that accounts for physical and biogeochemical processes at the seabed–water interface, including resuspension of POM and porewater, and implemented it for the dynamic Rhône Delta. Previously, one-dimensional box models with a few vertical levels have been used to study the role of organic matter resuspension on oxygen (Wainright and Hopkinson, 1997) and contaminant levels (Chang and Sanford, 2005). Additionally, three-dimensional circulation models have been coupled to biogeochemical models with a single seabed layer and implemented to investigate the role of POM resuspension on Baltic Sea carbon budgets (Almroth-Rosell et al., 2011) and Black Sea biogeochemistry (Capet et al., 2016). To the best of our knowledge, however, no previously existing models have sufficient vertical resolution to resolve changes in the vertical biogeochemical profiles that drive diffusive seabed–water-column fluxes, or the ability to account for the entrainment of reduced chemical species into the water column.
This paper presents a model called
This section describes the Rhône Delta (Sect. 2.1), and HydroBioSed (Sect. 2.2), before explaining how the model was implemented to address the research questions (Sect. 2.3). Tables 1 and 2 list related symbols and vocabulary.
Description of symbols used in this paper. Note that concentrations
are porewater or bottom-water concentrations, not bulk concentrations, unless
otherwise noted, but units of length and area (i.e., meters, m, and meters squared, m
Continued.
Description of phrases, acronyms, and abbreviations, as used in this paper.
Located in the Gulf of Lion at the northwest end of the Mediterranean Sea,
the Rhône River subaqueous delta in France is an excellent study site for
these research questions, in part because of the available observations
(Fig. 1). Our study is co-located with the site from Toussaint et al. (2014)
at the “Mesurho” station (Pairaud et al., 2016) and is only a few kilometers away
from Site A in Pastor et al. (2011a); both locations are at
This site experiences frequent seabed disturbance due to centimeters of
erosion superimposed on rapid fluvial deposition. Over timescales of decades,
due to its proximity to the Rhône River (Fig. 1), accumulation rates at
this site are
The delivery of organic matter to the shelf drives oxygen consumption
directly via aerobic remineralization, and indirectly, as reduced chemical
species produced during remineralization are oxidized (Lansard et al., 2009).
Organic material comprises about 2–12 and < 1–5 % of
water-column and seabed particulate matter, respectively, and about
four-fifths of it originates from a terrestrial source, with little marine
influence at the study site (Bourgeois et al., 2011; Pastor et al., 2011a;
Lorthiois et al., 2012; Cathalot et al., 2013). Yet, the material settling to
the seabed at this site is relatively labile, and has been estimated to have
remineralization rate constants of 11–33 yr
The fully coupled HydroBioSed numerical model was developed within the Regional Ocean Modeling System (ROMS), a community-based and well-utilized ocean modeling framework (Haidvogel et al., 2000, 2008; Shchepetkin, 2003; Shchepetkin and McWilliams, 2009). In addition to its core hydrodynamic components, ROMS includes widely used modules for sediment transport (CSTMS: Community Sediment Transport Modeling System; Warner et al., 2008), and water-column biogeochemistry (e.g., Fennel et al., 2006, 2013). We built on those previous studies by coupling the sediment transport and water-column biogeochemistry components (Fig. 2a), enabling the model to account for storage of POM and nutrients in the seabed, and subsequent resuspension and redistribution of the organic matter and nutrients. As part of the coupling, we also incorporated aggregation of detritus, seabed–water-column diffusion, and a multi-layer seabed biogeochemical model based on Soetaert et al. (1996a, b). Below, we briefly describe the sediment transport and water-column biogeochemistry modules used, highlighting differences from standard ROMS implementations and the addition of the seabed biogeochemistry model.
Suspended sediment tracers in the ROMS–CSTMS module are transported by ocean
currents, experience downward settling, may be deposited and eroded from the
multi-layer seabed model, and are subject to source and sink terms such as
river discharge (Warner et al., 2008). As discussed in Warner et al. (2008),
the rates of deposition,
ROMS water-column biogeochemistry modules have typically included variables
for multiple nutrient, plankton, and detrital classes and accounted for
processes such as growth, grazing, and remineralization (e.g., Fennel et al.,
2006). Here, the ROMS biogeochemical model from Fennel et al. (2013) was
modified so that HydroBioSed converts some of the large detritus into
faster-sinking aggregates in the water column. In Fennel et al. (2013), small
detritus and phytoplankton in the water column may coagulate to form large
detritus. HydroBioSed builds on the Fennel et al. (2013) framework by
partitioning coagulated material into three types of particulate matter:
(1) large detritus, (2) labile aggregates, and (3) refractory aggregates
(Fig. 2b). Based on estimates that roughly half of the deposited particulate
organic matter is refractory in the Gulf of Lion (Tesi et al., 2007; Pastor
et al., 2011a), the model partitions coagulated material into 50 %
refractory aggregates and 50 % labile material (
Environmental conditions and parameters for the Standard Model implementation.
Continued.
A seabed biogeochemistry module (Soetaert et al., 1996a, b) was added to ROMS to account for changes in oxygen, dissolved nitrogen, and POM due to remineralization, oxidation of reduced chemical species, and diffusion across the seabed–water interface. This model has performed well in many environments including areas near river deltas (Wijsman et al., 2002; Pastor et al., 2011a), on the continental shelf and slope (Soetaert et al., 1998; Epping et al., 2002), and in the deep ocean (Middelburg et al., 1996). To incorporate the Soetaert et al. (1996a, b) model into HydroBioSed, we used the code developed by Wilson et al. (2013), and adapted it for the ROMS framework and the Rhône Delta. Calculations use the first-order accurate Euler method.
This seabed biogeochemistry model specifically tracks degradable particulate
organic carbon (POC), oxygen, nitrate, ammonium, and oxygen demand units
(ODUs), defined as the moles of reduced chemical species that react with 1 mole of O
Merging the Soetaert et al. (1996a, b) seabed biogeochemical model with the sediment transport and water-column biogeochemistry modules allows HydroBioSed to account for exchanges of biogeochemical tracers across the seabed–water interface due to deposition, erosion, and diffusion (Fig. 2b). Upon settling to the seabed, phytoplankton, detritus, and labile aggregates are incorporated into labile seabed organic matter in the surficial seabed layer. Refractory aggregates are added to the pool of refractory seabed organic matter in that layer. Porewater in newly deposited sediments is assumed to initially have concentrations of dissolved nutrients and oxygen equal to those in the overlying water column. This material may be reentrained into the water column when bed shear stress exceeds the critical shear stress of the seabed. Specifically, any POM or dissolved chemical species in the porewater within an eroded layer(s) of sediment is also entrained into the bottom water-column layer. The flux of sediment entrained into the water column is determined by the CSTMS module (see Sect. 2.2.1). In addition to erosion and deposition, dissolved oxygen and nutrients may be transported across the seabed–water interface by diffusion, as described in Appendix A.1.
During erosional periods, resuspended labile and refractory seabed organic matter is incorporated into the pools of labile or refractory aggregates suspended in the water column, respectively. Like other coagulated material in the water column, this material may be repartitioned based on Eqs. (3–5). Usually, the seabed organic matter is enriched in refractory material compared to the water column. Thus, this repartitioning reclassifies a fraction of the resuspended refractory organic matter, i.e., refractory aggregates, into the labile organic matter classes, i.e., large detritus and labile aggregates. This modeling approach is supported by laboratory experiments by Stahlberg et al. (2006) indicating that organic matter remineralization rates increased during and in the days following resuspension events, and that changes in remineralization rates were not only due to changes in oxygen availability. Due to the limited availability of pertinent research, we also considered literature related to the effect of redox oscillations on organic matter remineralization (e.g., Gilbert et al., 2016; Sun et al., 2003; Caradec et al., 2004; Aller, 1994; Wakeham and Canuel, 2006; Arzayus and Canuel, 2004). Yet, because guidance from this literature was inconclusive, we chose the simple approach described above for the partitioning of organic matter that mimics the changes in remineralization described in Stahlberg et al. (2006). We also tested an alternative, “no-repartitioning” approach that did not repartition resuspended organic matter, but this approach caused decreases in oxygen gradients across the seabed–water interface during depositional periods, inconsistent with observations from Toussaint et al. (2014) (Fig. 2c).
Overall, HydroBioSed represents POM in the seabed until it is resuspended, remineralized, or buried. Similarly, dissolved chemical species in the porewater may undergo biogeochemical transformations, diffuse into or out of the seabed, or be exchanged with the water column during periods of erosion and deposition. Thus, unlike Soetaert et al. (1996a, b) and other classical seabed biogeochemistry models (e.g., Berner, 1980; Boudreau, 1997; Soetaert et al., 2000; DiToro, 2001), HydroBioSed can quantify the effect of resuspension on biogeochemical dynamics (Fig. 2).
To evaluate the coupled model and explore the role of local resuspension on oxygen dynamics, we implemented a one-dimensional version of HydroBioSed for the Rhône Delta. This section describes the standard model run and sensitivity tests, and summarizes our methods for model evaluation and analysis. See Table 3 for a list of model input and parameters.
To isolate the effect of resuspension on seabed–water-column fluxes,
water-column concentrations of oxygen, nitrogen, and ODU, as well as the
supply of POM (excluding that from resuspension) were strongly nudged to
temporally constant values. Hourly to daily oxygen observations from the
bottom boundary layer (Toussaint et al., 2014), were used to constrain modeled
concentrations in the water column. These observations indicated that oxygen
concentrations 1 m above the bed varied between 216 and 269 mmol
O
Model forcing and parameters were chosen based on a combination of observed
values (wave height, bottom-water oxygen concentrations), climatology
(inorganic sedimentation rate, salinity, temperature), and values used in
previously implemented models (fraction of labile material, nitrification
rate, rates of diffusion within the seabed). See Table 3 for more details. A
few parameters, i.e., critical shear stress for erosion and erosion rate
parameter, were tuned to reproduce the 1–2 cm of observed erosion. For
initialization, the model was run without resuspension until it reached
steady state. As the biogeochemical profiles reached a state of
quasi-equilibrium within days following perturbations, using alternative
initialization techniques primarily affected estimates for the first
resuspension event and did not have a large effect on our results. The model
used a 30 s time step, the MPDATA advection scheme (Smolarkiewicz and
Margolin, 1998), the generic length scale turbulence
closure (Umlauf and Burchard, 2009), and a piecewise parabolic method
(Colella and Woodward, 1984) with a weighted essentially non-oscillatory
scheme (Liu et al., 1994) to estimate particle settling. It saved output in
3 h increments, and took
List of sensitivity tests. Additionally, for each simulation listed
here, an identical model run was completed that neglected resuspension (i.e.,
with
Additionally, “no-resuspension” model runs were completed to evaluate the
role of cycles of erosion and deposition on biogeochemical dynamics.
Specifically, for each sensitivity test and the standard model run, a
corresponding simulation was conducted that was identical to the original,
except that erosion was prevented by increasing the critical shear stress to
This section evaluates the skill of the standard model run by comparing it to observations (Sect. 3.1), analyzes the effect of resuspension on oxygen dynamics (Sect. 3.2), and evaluates the results' sensitivity to model parameters (Sect. 3.3).
Comparison of the standard version of HydroBioSed to Toussaint et
al.'s (2014) time series of oxygen profiles showed that model results were
consistent with measured concentrations, and changed during resuspension
events in a manner similar to the observations (Fig. 3). During quiescent
conditions when bed shear stress was low, modeled and observed oxygen
concentrations decreased with depth into the seabed, falling from about
250 mmol O
Time series of modeled (blue lines and
To quantify the changes in seabed oxygen profiles, the oxygen gradient near
the seabed–water interface was calculated from both the observed and modeled
profiles (Table 5). Specifically, the slope of the oxygen profile was
averaged over the oxygen penetration depth (OPD; variables are defined in
Table 1):
Statistics for model–observation comparison, including the root
mean square difference (RMSD) and the correlation coefficient (
Differences in the modeled and observed oxygen profiles derive at least
partially from differences in estimating seabed elevation (i.e., erosion and
deposition). As a one-dimensional vertical model, HydroBioSed assumed uniform
conditions in the horizontal, and so all resuspended material was
redeposited in the same location within a few days following an event. Yet,
at the actual study site, it is likely that some material was carried out of
the area and that deposition following the erosional periods was more gradual
than estimated in the model (e.g., see the late April and early May event in
Fig. 3c). Also, the model provided higher temporal resolution than possible
with the sampling gear, and may capture peaks in
dO
Overall, the combined seabed–bottom-water oxygen consumption increased from
Time series of bed stress and oxygen consumption in the seabed and bottom water (BW) for both the standard (blue solid line) and no-resuspension model runs (pink line). Shading indicates resuspension events, i.e., cycles of erosion and redeposition, as listed in Fig. 3. The red dashed line indicates the critical shear stress for erosion, and the black dashed lines indicate the times at which profiles in Fig. 5 were estimated.
O
The cycles of erosion and deposition that affected biogeochemical cycles are
illustrated by time series of seabed profiles (Fig. 5). Before resuspension
events, the porewater in surface sediments was typically equilibrated with
the overlying water column, with oxygen penetrating
Seabed profiles of oxygen (top row; mmol O
Physical (top) and biogeochemical (bottom) sources and sinks of
oxygen within the seabed for the standard model run. Sources and sinks of
oxygen to the seabed are positive and negative, respectively. Small
biogeochemical sinks < 1 mmol O
The next two sections provide a more detailed and quantitative analysis of how these exchanges of porewater and particulate matter between the seabed and the overlying water increased oxygen consumption and affected related biogeochemical processes within the seabed (Sect. 3.2.1) and bottom waters (Sect. 3.2.2).
Resuspension directly altered the supply of oxygen to the seabed. In this environment, where oxygen penetration was limited to the top few millimeters of the seabed, resuspension events typically removed the entire seabed oxic layer; the oxygen that had been in the porewater was entrained into the water column. Similarly, during deposition, incorporation of oxygen within the porewater of newly deposited sediment provided a source of oxygen to the seabed, accounting for up to a quarter of oxygen input to the seabed on a timescale of hours to days. Overall, this “pumping” of oxygen into and out of the seabed when sediments were deposited or eroded provided a small net source of oxygen to the seabed during a typical resuspension cycle; based on time-integrated fluxes of oxygen across the seabed–water interface for the 2-month period (Fig. 6a), these exchanges accounted for 4 % of the net oxygen supply to the seabed.
The remaining supply of oxygen (96 %) was delivered to the seabed via
diffusion across the seabed–water interface. Although these diffusive fluxes
of oxygen were always directed into the seabed, erosion and deposition caused
fluctuations in the rate of diffusion. During periods of resuspension,
erosion of the oxic layer sharpened the oxygen gradient at the seabed–water
interface, thus increasing diffusion of oxygen into the seabed by about
77 % (Fig. 6a). In contrast, during periods of deposition, incorporation
of oxygen-rich porewater into newly deposited surficial seabed layers reduced
the oxygen gradient at the seabed–water interface, decreasing diffusion of
oxygen into the seabed by about 71 %. However, “erosional oxygen
profiles” with thin oxygen penetration depths persisted longer and induced
larger changes in the rate of diffusion, compared to “depositional oxygen
profiles” with thick oxygen penetration depths. This imbalance occurred
because the additional oxygen available in the seabed during periods of
redeposition (i.e., oxygen available due to the incorporation of oxic water
into the porewater of newly deposited sediments) was rapidly consumed by
aerobic organic matter remineralization and nitrification, and so oxygen
profiles returned to their quasi-steady state condition within hours to
Rate of oxygen consumption in the
In addition to impacting the supply of oxygen to the seabed, resuspension
altered the magnitude of various biogeochemical oxygen sinks within the
seabed (Table 6, Fig. 6b). For example, erosion of organic matter, and labile
organic matter in particular, decreased rates of oxic remineralization in the
seabed from about 5 to < 1 mmol O
Resuspension primarily affected oxygen dynamics within the water column by
entraining POM into the layer of water below the pycnocline, i.e., bottom
waters, which increased remineralization rates there (Table 6). Turbulence
entrained this material as high as
In addition to entraining POM into the water column, resuspension increased
fluxes of reduced chemical species from the seabed into bottom waters,
further increasing oxygen consumption in the water column (Table 6). During
quiescent periods, oxidation of ammonium (nitrification) resulted in a
background level of oxygen consumption of
Like the standard model run, results from every sensitivity test showed that
resuspension increased bottom-water oxygen consumption during both individual
resuspension events and when estimates were averaged over 2 months
(Fig. 7d). All sensitivity tests except one showed that resuspension also
increased seabed oxygen consumption (Fig. 7b). In all model runs, oxygen
consumption in bottom waters was larger than that in the seabed for every
sensitivity test by at least a factor of
Over timescales ranging from hours to 2 months, seabed oxygen consumption
was more sensitive to changes in the rate of diffusion within the seabed
(
Within the standard model run and most sensitivity tests, resuspension
accounted for about 14 % of the cumulative seabed oxygen consumption when
integrated over 2 months. The role of resuspension, however, was especially
sensitive to the partitioning and delivery of organic matter because POM
entrained into the water column was subject to repartitioning (see
Sect. 2.2.3; Fig. 2b) and so resuspension increased the amount of labile
material available to redeposit on the seabed. This additional source of
seabed labile organic matter increased seabed oxygen consumption directly,
due to oxic remineralization, and indirectly, as ammonium produced during
this process was oxidized via nitrification. Overall, altering the
partitioning of organic matter between labile and refractory classes changed
the effect of resuspension on seabed oxygen consumption by up to 60 %
over 2 months (Cases L1 and L2; Fig. 7b). Specifically, decreasing
(increasing) the fraction of organic matter that is labile,
Oxygen consumption in bottom waters averaged over 2 months was more
sensitive to changes in the critical shear stress for erosion,
Within the standard model run and most sensitivity tests, resuspension accounted for about 57 % of bottom-water oxygen consumption when averaged over 2 months (Fig. 7d). Similar to the above analysis, the extent to which resuspension affected oxygen consumption was especially sensitive to the critical shear stress (Cases T1, T2). Over the 2-month model run, halving (doubling) the critical shear stress changed the fraction of bottom-water oxygen consumption that occurred due to resuspension to 34 % (71 %).
This discussion focuses on the importance of resuspension-induced changes in oxygen budgets in different environments (Sect. 4.1), compares our approach to other modeling techniques (Sect. 4.2), and suggests future research (Sect. 4.3).
Resuspension-induced oxygen consumption that occurred during short time periods (hours to days) increased model estimates of oxygen consumption integrated over longer timescales of weeks to months for all model runs (Figs. 7, 8). In other words, erosion and deposition did not just add variability to the time series of oxygen consumption; resuspension impacted the oxygen budget of the Rhône subaqueous delta. This section discusses the environmental conditions that caused this effect and the extent to which we expect resuspension to increase oxygen consumption in other coastal systems (Sect. 4.1.1); and the importance of these changes relative to seasonal variability (Sect. 4.1.2).
Box and whisker plot indicating the 0th, 25th, 50th, 75th, and 100th percentiles of combined seabed–bottom-water (BW) oxygen consumption averaged over different timescales for the standard model run. The pink lines indicate estimates from the no-resuspension model run.
Several characteristics of the Rhône subaqueous delta favor the increased
rates of oxygen consumption due to local resuspension. First, frequent
resuspension, e.g., three events in 2 months (Fig, 3c), ensures that the
entrainment of seabed organic matter into the water column and erosional
seabed profiles occur often, increasing resuspension-induced oxygen
consumption in both bottom waters and the seabed. Second, oxygen
concentrations in bottom waters and near the seabed–water interface are
relatively high, i.e., over 200 mmol O
We expect that the effect of local resuspension on oxygen dynamics in other systems that share characteristics of the Rhône subaqueous delta would be similar to our results. For seabed oxygen dynamics, this implies that the importance of local resuspension increases in energetic, oxic, and coastal areas with high organic matter input but relatively little bioturbation, including other river deltas (Aller, 1998; Aller et al., 1996, e.g., Amazon Delta, Brazil). For water-column oxygen dynamics, the above criteria suggest that local resuspension is most important in similar coastal areas with organic-rich, muddy seabeds, but relatively low background concentrations of organic matter in the water column. These characteristics may be found in regions with historically high nutrient loading and where organic matter has accumulated in the seabed (e.g., Gulf of Finland, Almroth et al., 2009). In sites that meet some, but not all of the above criteria, local resuspension may have a reduced effect on oxygen dynamics compared to the Rhône subaqueous delta.
The model estimated that resuspension increased seabed and bottom-water
oxygen consumption by about 16 and 140 %, respectively, when integrated
over April–May 2012 (Fig. 7); however, seasonal variations in environmental
conditions such as temperature may change the importance of resuspension for
oxygen dynamics. The 2-month model run presented here assumed a constant
bottom-water temperature of 15
Seasonal variations in resuspension frequency and magnitude may have a similarly large effect on oxygen consumption. During the winter when easterly storms are more frequent (Guillén et al., 2006; Palanques et al., 2006), resuspension-induced oxygen consumption could be more important than was estimated for the April–May period in this study. At the 32 m deep “Sète” site in the central coastal region of the Gulf of Lion, significant wave heights exceeding 2 m were observed an average of 3.5, 1, and 2 times per month in November–December 2003, January-February 2004, and March–April 2004, respectively (Ulses et al., 2008). Approximately doubling the resuspension frequency during the winter storm season could roughly double resuspension-induced oxygen consumption, counteracting reductions in wintertime oxygen consumption due to colder temperatures. Overall, accounting for the effect of erosional and depositional cycles on oxygen consumption may vary in importance throughout the year on the Rhône subaqueous delta, but it is likely more important during Fall compared to the Springtime period that was analyzed for this study.
Finally, oxygen dynamics may vary in response to seasonal or episodic variations in organic matter availability and lability. Following a flood in 2008, seabed oxygen consumption on the Rhône Delta decreased by one-third to one-half when riverine inputs of relatively refractory organic matter lowered remineralization rates in surficial seabed sediments, reducing seabed oxygen consumption (Cathalot et al., 2010). This result is consistent with results from our L1 sensitivity test, indicating that reducing the ratio of labile to refractory organic matter lowered seabed oxygen consumption (Fig. 7a). Thus, although variability in the amount and quality of organic matter delivered to the delta could be episodic, it may also substantially affect estimates of seabed oxygen consumption oxygen, similar to temperature and resuspension.
HydroBioSed differs from other models by accounting for resuspension-induced
changes in millimeter-scale biogeochemistry, a feature that was necessary to
reproduce Toussaint et al.'s (2014) observed temporal variations in seabed
oxygen consumption on the Rhône subaqueous delta. In contrast, other
models neglect resuspension-induced changes in biogeochemical dynamics or
assume that increases in water-column oxygen consumption due to
remineralization of resuspended organic matter during erosion are at least
partially offset by decreases in remineralization and associated oxygen
consumption in the seabed (e.g., Feng et al., 2015; Capet et al., 2016).
Results from these model parameterizations therefore conflict with our
HydroBioSed results that show that both water-column and seabed oxygen
consumption
First, resuspension increased the importance of bottom waters relative to the seabed for oxygen consumption. During quiescent conditions, bottom waters and the seabed each accounted for similar rates of oxygen consumption. However, when POM and porewater were entrained into the water column via resuspension, bottom-water oxygen consumption increased by a factor of 8, while seabed oxygen consumption only doubled. This disproportionate increase of oxygen consumption within bottom waters affirmed the importance of observing and modeling oxygen dynamics within bottom waters during resuspension events. Also, only accounting for quiescent time periods would underestimate the role of bottom waters, which accounted for 75 % of the total oxygen consumption over the 2-month model run for the Rhône Delta site, but only accounted for about 50 % when resuspension was neglected.
Second, diffusion of oxygen across the sediment–water interface dominated the supply of oxygen to the seabed in the model, regardless of the timescale or time period considered. The other transport mechanism, the “pumping” of oxygen into and out of the seabed as layers of sediment were deposited or eroded, provided at most a third of the instantaneous flux to the seabed (during depositional time periods; Fig. 5). Also, pumping contributed much less to seabed oxygen supply over time, primarily because the entrainment of porewater from the seabed into the water column during erosional periods partially offset the depositional flux of oxygen (Fig. 5). Over the 2-month simulation, diffusion across the seabed–water interface accounted for 96 % of the seabed oxygen supply, whereas pumping via erosion and deposition accounted for only 4 % of seabed oxygen fluxes. Thus, for environments like the Rhône Delta, future observational and modeling efforts should include resuspension-induced changes to diffusive fluxes across the seabed–water interface (Jørgensen and Revsbech, 1985).
Although resuspension can affect oxygen dynamics in coastal environments, the large spatial or temporal scale of some biogeochemistry models may make incorporating a full sediment model undesirable. For environments similar to the Rhône Delta, we suggest parameterizations for bottom-water and seabed oxygen consumption that focus on the role of resuspended organic matter and seabed–water-column diffusion. For example, various approaches have been used to parameterize the effect of resuspension on particulate organic matter fluxes (e.g., Cerco et al., 2013; Feng et al., 2015). Approaches accounting for temporal lags between deposition and reentrainment of organic matter into the water column seem especially promising for modeling oxygen dynamics in episodically energetic environments like the Rhône Delta (e.g., Almroth-Rosell, 2011; Capet et al., 2016). In addition, future parameterizations for seabed–water-column fluxes should focus on diffusion of oxygen across the seabed–water interface as well as the supply of organic matter and reduced chemical species (e.g., Findlay and Watling, 1997; De Gaetano et al., 2008; Hetland and DiMarco, 2008; Murrell and Lehrter, 2011; Testa et al., 2013; Laurent et al., 2016). Methods combining parameterizations for seabed–water-column fluxes and seabed resuspension may be particularly helpful for environments similar to the Rhône Delta where erosion and deposition may affect these processes.
This study focused on oxygen dynamics while holding the supply of organic matter and sediment, water-column concentrations of nutrients and oxygen, and temperature constant in time based on conditions observed on the Rhône subaqueous delta. Future work should therefore include analysis of the role of resuspension on oxygen dynamics for a variety of environmental conditions and investigation into how temporal variability in environmental conditions affects the relative importance of resuspension for oxygen dynamics. Additionally, applying HydroBioSed for a three-dimensional system would further facilitate its application to additional scientific and water quality concerns. For example, transport of organic matter from regions near the Mississippi and Atchafalaya river mouths, shallow autotrophic waters, and wetlands to “Dead Zones” has been speculated to encourage the depletion of oxygen in bottom waters there (Bianchi et al., 2010). However, the importance of organic matter transport within a single season of hypoxia, and on interannual timescales, is difficult to quantify with observations and has been debated on the northern shelf of the Gulf of Mexico (Rowe and Chapman, 2002; Boesch, 2003; Turner et al., 2008; Forrest et al., 2012; Eldridge and Morse, 2008) and other locations (Kemp et al., 2009 and references therein). Modeling efforts that account for resuspension of organic matter, as well as oxygen and nutrients, can help quantify the extent to which organic matter supply, resuspension, and transport affect biogeochemistry in these dynamic coastal environments (e.g., Almroth-Rosell et al., 2011; Capet et al., 2016).
Our analysis focused on oxygen, but resuspension also affected model estimates of nitrogen dynamics. For example, during quiescent periods, nitrification roughly balanced production of ammonium from remineralization of organic matter in the seabed, consistent with Pastor et al. (2011a). Yet, during erosional periods, the exposure of ammonium-rich porewater to oxygen increased seabed nitrification, enhancing fluxes of nitrate out of the seabed, consistent with observations from other systems (e.g., Fanning et al., 1982; Sloth et al., 1996; Tengberg et al., 2003). Overall, resuspension roughly doubled nitrate fluxes out of the seabed during resuspension, which led to about a 10 % increase overall for the 2-month model run.
HydroBioSed did not represent all processes that occur near the seabed–water-column interface. For example, future work could include accounting for turbulence-induced changes in diffusion, advective fluxes through the seabed, and variations in seabed porosity, as well as improving the model's representation of organic matter. Within HydroBioSed, for example, the steepening of the oxygen gradient at the seabed–water interface occurred because of changes in oxygen concentrations within the seabed and bottom waters (Fig. 3). HydroBioSed did not account for the thinning of the viscous layer at the seabed–water interface in response to wave-induced turbulence, which would act to further increase the oxygen gradient during erosional time periods (Gundersen and Jorgensen, 1990; Chatelain and Guizien, 2010; Wang et al., 2013). This implies that our current model estimates of oxygen diffusion into the seabed during resuspension events are conservative. Additionally, the model could be adapted for locations where waves and currents drive flows of water through non-cohesive seabeds, stimulating biogeochemical reactions (Huettel et al., 2014), or to account for vertical gradients in seabed porosity (Soetaert et al., 1996a, b). Finally, the uncertainty about both how to partition organic matter into classes for numerical modeling efforts and the effect of resuspension on remineralization rates, as noted in Sect. 2.2.3, has a large effect on model estimates (Fig. 7, Cases L1, L2, C1) and deserves attention from both the modeling and observational research communities.
Finally, this modeling effort incorporated time-dependent reactions into the ROMS sediment transport module and could be adapted for other research applications for which both resuspension and time-dependent tracers are important. For example, the model has been adapted to account for short-lived radioisotopes (Birchler, 2014) and could be adapted to include time-dependent particulate tracers such as the following: (1) particle-reactive nutrients and contaminants (Wiberg and Harris, 2002; Chang and Sanford, 2005); (2) other “particulates” such as cysts of harmful algal bloom species, (Beaulieu et al., 2005; Giannakourou et al., 2005; Butman et al., 2014; Kidwell, 2015), or fecal pellets (Gardner et al., 1985; Walsh et al., 1988); and (3) temporal variability in organic matter lability, oxygen exposure time, and carbon budgets (Aller, 1998; Hartnett et al., 1998; Burdige, 2007).
A model called HydroBioSed was developed that couples hydrodynamics, sediment transport, and both water-column and seabed biogeochemistry. A one-dimensional (vertical) version of the model was then implemented for the Rhône River subaqueous delta. This work expanded on the commonly used ROMS framework by accounting for non-conservative tracers, the resuspension of organic matter and entrainment of porewater into the water column, diffusion of dissolved tracers across the seabed–water interface, and feedbacks between resuspension and diffusion across the seabed–water interface. Including these processes created a new model capable of reproducing previously observed changes in seabed profiles that occurred during resuspension events on the Rhône River subaqueous delta.
Resuspension increased model estimates of oxygen consumption over the range
of timescales considered (hours to 2 months). In the seabed, resuspension
increased the exposure of anoxic, ammonium-rich sediment to oxic,
ammonium-poor bottom waters, thus stimulating seabed oxygen consumption via
nitrification during erosional periods. This oxygen consumption compensated
for or exceeded the decrease in oxic remineralization rates that occurred as
organic matter was resuspended into the water column. Additionally,
entrainment of seabed organic matter and reduced chemical species from the
porewater into the bottom portion of the water column, i.e., below the
pycnocline, increased oxygen consumption there. Overall, resuspension
increased peak oxygen consumption rates more in bottom waters (factor of 8)
than in the seabed (factor of 2). When averaged over a 2-month period that
included intermittent periods of erosion and deposition, accounting for
resuspension increased oxygen consumption by
These results imply that observations collected during quiescent periods, and models based on steady-state assumptions, may underestimate net oxygen consumption. This finding is consistent with results from laboratory erodibility experiments (e.g., .Sloth et al., 1996), observations using eddy correlation techniques (Berg and Huettel, 2008), and microelectrode profiles (Toussaint et al., 2014). While all of these studies showed increased oxygen consumption during resuspension events, they each had limitations; i.e., erodibility experiments are limited to low levels of erosion and timescales of hours, eddy-correlation methods can only be used for time periods without abrupt shifts in hydrodynamic and oxygen conditions (Lorrai et al., 2010), and microelectrodes can only be deployed in soft muddy seabeds. Thus, models like HydroBioSed, which resolve both biogeochemical processes and resuspension, may help observational studies quantify oxygen dynamics over longer time periods, during storms, and in a variety of environments.
Certain characteristics of the Rhône subaqueous delta study site, including its oxic water column, shallow oxygen penetration into the seabed compared to the thickness of eroded layers, fast rates of oxygen consumption, and the high concentrations of labile seabed organic matter, enhance the effect of resuspension on oxygen dynamics. Together, these characteristics ensure the following: (1) oxygen consumption in bottom waters is limited by the supply of organic matter and reduced chemical species, as opposed to oxygen availability; (2) resuspended material is rich in organic matter and reduced chemical species that increase oxygen demand in the water column – oxygen consumption in the seabed is dependent on the supply of oxygen, as opposed to the rate of consumption; (3) oxygen is available to be supplied to the seabed during resuspension; and (4) erosion exposes anoxic regions of the seabed to oxic regions of the water column. The dependence of oxygen dynamics on those environmental conditions caused modeled estimates of oxygen consumption to be particularly sensitive to the supply and lability of organic carbon, rates of diffusion within the seabed, nitrification rate, and the frequency of resuspension. Our results imply that local resuspension may affect oxygen dynamics in other environments with similar characteristics.
Model datasets (Moriarty et al., 2017) are publicly
available through the College of William & Mary's Digital Archive at
This study modified the seabed layering scheme from Warner et al. (2008) to include biogeochemical tracers and diffusion of dissolved tracers between the seabed and water column (Sect. A.1), and to resolve millimeter-scale processes in surficial sediments while maintaining centimeter-scale resolution deeper in the seabed (Sect. A.2).
Parameters for new seabed layering scheme, as implemented for the Rhône study site. Dashed lines indicate that no symbol was assigned to that parameter.
To couple the sediment transport and biogeochemical modules, we incorporated tracers representing particulate organic carbon and dissolved chemical species including oxygen and nutrients into the seabed module. To elaborate on the information presented in the Methods section (Sect. 2.2), this section details how the sediment transport module was adapted from Warner et al. (2008) to account for them. The inclusion of particulate organic carbon was relatively straightforward because the model treats it similarly to sediment classes, except that it decays in time. Inclusion of dissolved oxygen, nitrogen and ODU in the model, however, necessitated accounting for the formation of porewater within newly deposited layers and the entrainment of porewater into the water column during erosion, as described in Sect. 2.2.3, as well as diffusion of dissolved chemical constituents across the seabed–water interface, which is described below.
Our model parameterizes diffusion across the seabed–water interface by
assuming that concentrations of dissolved tracers in the bottom water column
and surficial seabed layer are equal. At each step, dissolved tracers move
into or out of the seabed so that concentrations in the surficial seabed
layer match those in the bottom water-column cell, while conserving tracer
concentrations (symbols defined in Table 1):
Our seabed layering scheme is based on Warner et al. (2008), whose model
includes a single, thin, active transport layer with thickness
Specifically, the layering scheme includes
Incorporating multiple types of layers within the seabed and maintaining high
resolution near the sediment–water interface affects how the layering scheme
handles erosion and deposition. During depositional periods, new sediment is
incorporated into surficial seabed layer(s), as described in Warner et
al. (2008). When deposition increases the thickness of the surficial layer so
that it exceeds
Additionally, the method of calculating the thickness of the surficial seabed
layer,
Katja Fennel is Co-Editor in Chief of the journal.
Observations from the Rhône River delta observatory (Mesurho) were provided by F. Toussaint (Laboratoire des Sciences du Climat et de l'Environnement). R. Wilson (formerly Dalhousie University) provided model code from Wilson et al. (2013). Feedback from E. Canuel, C. Friedrichs (Virginia Institute of Marine Science; VIMS), two anonymous reviewers, and Biogeosciences Associate Editor Jack Middelburg improved this paper. A. Miller, D. Weiss (VIMS), and E. Walters (the College of William & Mary; W&M) provided computational support and access to W&M's computing facilities, which are funded by the National Science Foundation, the Commonwealth of Virginia Equipment Trust Fund, and the Office of Naval Research. Funding was provided by the US National Oceanic and Atmospheric Administration Center for Sponsored Coastal Ocean Research (NA09NOS4780229, NA09NOS4780231, NGOMEX contribution 217) (Moriarty, Harris, Fennel, Xu), VIMS student fellowships (Moriarty), and MISTRALS/MERMEX-River and ANR-11-RSNR-0002/ AMORAD (Rabouille). This is contribution 3618 of the Virginia Institute of Marine Science. Edited by: J. Middelburg Reviewed by: two anonymous referees