Episodic events of flood deposit in coastal environments
are characterized by deposition of large quantities of sediment containing
reactive organic matter within short periods of time. While steady-state
modelling is common in sediment biogeochemical modelling, the inclusion of
these events in current early diagenesis models has yet to be demonstrated.
We adapted an existing model of early diagenetic processes to include the
ability to mimic an immediate organic carbon deposition. The new model
version (FESDIA) written in Fortran and R programming language was able to
reproduce the basic trends from field sediment porewater data affected by
the November 2008 flood event in the Rhône River prodelta. Simulation
experiments on two end-member scenarios of sediment characteristics dictated
by field observation (1–high thickness deposit, with low TOC (total organic carbon) and 2–low
thickness, with high TOC), reveal contrasting evolutions of
post-depositional profiles. A first-order approximation of the differences
between subsequent profiles was used to characterize the timing of recovery
(i.e. relaxation time) from this alteration. Our results indicate a longer
relaxation time of approximately 4 months for
Coastal margins play a crucial role in the global marine systems in terms of carbon and nutrient cycling (Wollast, 1993; Rabouille et al., 2001b; Cai, 2011; Regnier et al., 2013; Bauer et al., 2013; Gruber, 2015). Due to their relatively shallow depth, sedimentary early diagenetic processes are critical for the recycling of a variety of biogeochemical elements, which are influenced by organic matter (OM) inputs, particularly carbon (Middelburg et al., 1993; Arndt et al., 2013). Furthermore, these processes have the potential to contribute to the nutrient source that fuels the primary productivity of the marine system. In river-dominated ocean margins (RiOmar; McKee et al., 2004), organic matter input can also be enhanced by flood events which provide a significant fraction of the particulate carbon (POC) delivered to depocenters (Antonelli et al., 2008). Organic matter derived from riverine input to sediment has biogeochemical significance in coastal marine systems (Cai, 2011). As a result, the coastal environment serves as both a sink for particulate organic carbon and nutrients, and an active site of carbon and nutrient remineralization (Burdige, 2005; McKee et al., 2004; Sundby, 2006).
In the context of early diagenetic modelling, numerical models with time-dependent capability are well-established (Lasaga and Holland, 1976; Rabouille and Gaillard, 1991; Boudreau, 1996; Soetaert et al., 1996a, b; Rabouille et al., 2001a; Archer et al., 2002; Couture et al., 2010; Yakushev et al., 2017; Munhoven et al., 2021; Sulpis et al., 2022), and they are used in many coastal and deep-sea studies. However, because of the scarcity of observations and their unpredictability, the role of massive deposit of sediment in these early diagenesis models has frequently been overlooked (Tesi et al., 2012). As these rare extreme events are being currently documented in various locations, there is a growing appreciation for their impact on the coastal margin (Deflandre et al., 2002; Cathalot et al., 2010; Tesi et al., 2012).
Attempts to use mathematical models to understand perturbation-induced events such as sudden erosion/resuspension event, bottom trawling, and turbidity-driven sediment deposition on early diagenetic processes have resulted in a variety of approaches that incorporate this type of phenomenon. As an example, Katsev et al. (2006) demonstrated that the position of the redox boundary (depth zone beneath the sediment–water interface that separates the stability fields of the oxidized and reduced species of a given redox couple) in organic-poor marine sediment can undergo massive shifts due to the flux of new organic matter on a seasonal basis, whereas on a longer timescale (e.g. decadal), redox fluctuation linked to organic matter deposition can induce the redistribution of solid-phase manganese with multiple peaks (due to depth-wise oxidation reduction of Mn). Another study in a coastal system revealed that coastal sediments change as a result of an anthropogenic perturbation in the context of bottom dredging and trawling (van de Velde et al., 2018). More recently, using a similar model, De Borger et al. (2021) highlighted that perturbation events such as trawling can possibly decrease total OM mineralization.
In river-dominated ocean margins, episodic flood events can deliver sediment
with varying characteristics depending on its source origin, frequency, and
intensity (Cathalot et al., 2013). Therefore, the flood characteristics have
direct impact on the deposited sediment's characteristics such as
scale/thickness of the deposited layer, composition (mineralogy and
grain size), OM content, and so on. For example, In the Rhône prodelta, a
single flash flood can deliver up to 30 cm of new sediment material in a
matter of days (Cathalot et al., 2010; Pastor et al., 2018). Despite the
large amount of sediment introduced by this episodic loading, vertical
distribution of porewater species like oxygen (O
The goal of this study is to better understand episodic events in the
context of flood-driven sediment deposition and their impact on benthic
biogeochemistry, post-flood evolution dynamics, and relaxation timescale. As
the relaxation dynamics represent a gap in our understanding of how coastal
systems respond to external drivers, we characterize the timescale of the
recovery of sediment porewater profiles using a first-order approximation.
To accomplish this, we developed an early diagenetic model called FESDIA.
The ability to explicitly simulate non-steady early diagenesis processes in
systems subject to perturbation events such as massive flood or storm
deposition is a novel contribution of FESDIA to early diagenetic models. In
the following ways, FESDIA differs therefore from the OMEXDIA model
(Soetaert et al., 1996a) by implementing:
an explicit description of the anoxic diagenesis including (i) iron and
sulfur dynamics, and (ii) methane production and consumption. In comparison,
OMEXDIA has a single state variable (ODU: oxygen demand unit) to describe
reduced species; the possibility to include sediment perturbation events such abrupt deposition
of sediment.
In this paper, we only discuss part of the FESDIA model concerned with
implementation of a perturbation event as it relates to some biogeochemical
indicators. The model is implemented in Fortran (for speed) and linked to R
(for flexibility). We demonstrate the model's utility in describing data
collected from a flood event in November/December 2008 (Pastor et al., 2018)
and numerically investigate the impact of varying degrees of flood
type characteristics on the system's relaxation dynamics. This work is a
foundation for a more in-depth investigation of the model–data
biogeochemistry of the porewater and solid phase components of core samples
from Pastor et al. (2018), and it provides a useful baseline for
understanding the spatiotemporal dynamics of coastal marine systems subject
to event-driven organic matter pulses.
The Rhône prodelta serves as a case study for the development of the
model used to evaluate sediment perturbation dynamics. This particular
coastal area acts as the transitory zone between the inland river channel
and the continental shelf (Gulf of Lion) of the Mediterranean Sea. The
Rhône River with a drainage basin of 97 800 km
Map showing the locations of sampling sites off the Rhône River mouth.
Relating to the episodic pulse of organic matter, numerous studies have documented instances of flood-driven deposition from the Rhône River from a hydrographic perspective (Boudet et al., 2017; Hensel et al., 1998; Pont et al., 2017). Pastor et al. (2018) go beyond sedimentology and hydrographic characteristics to provide a concise description of the various flood types, their diagenetic signatures, and biogeochemical implications. Furthermore, published porewater chemistry and solid-phase data have highlighted sediment characteristics following such an event (Cathalot et al., 2010; Toussaint et al., 2013; Cathalot et al., 2013; Pastor et al., 2018).
Following the description of the Rhône River flood types and the composition of the flood deposit (mainly in terms of organic carbon) at the proximal station A (Pastor et al., 2018), we proceed to describe the model developed to explore the observed data and their diagenetic implications in terms of relaxation times and their evolution following this transient perturbation.
Description of notations, phrases, acronyms, and abbreviations, as used in this paper.
Our model combines the development in the OMEXDIA model (Soetaert et al., 1996a), applied in the Rhône prodelta area (Ait Ballagh et al., 2021; Pastor et al., 2011) and which has recently been equipped with event-driven processes (De Borger et al., 2021). In De Borger et al. (2021), the authors specifically addressed the issue of bottom trawling as a mixing and an erosional process that removes an upper layer of sediment and mixes a certain layer below. In addition, the model considers a bulk categorization of reduced substance in a single state variable, ODU (oxidative oxygen unit). For our approach, the event is defined by an addition of a new layer on top of the former sediment–water interface (Table 1). Furthermore, we explicitly modelled pathways involving sulfur and iron. Following this preamble, the following sections go over aspects of the model description and parameterization. Table 1 provide some key glossary of mathematical notations used in the model.
The complete model describes the concentration of labile
(
State variables described in the model.
In some coastal settings, oxidation via sulfate reduction has been
highlighted as the primary pathway for organic carbon (OC) mineralization,
with minor contributions from manganese and iron oxidation (Burdige and
Komada, 2011). In addition, the flux of integrated remineralization products
such as DIC has previously been estimated to contribute up to 8 times that
of diffusive oxygen uptake (Rassmann et al., 2020) – thus highlighting its
importance in describing the amplitude of benthic recycling in coastal
water. As such in this paper, we focus our analysis on these proxy variables
(
Early diagenesis processes on the seafloor are driven by organic matter
deposition. For areas such as the Rhône prodelta, continental organic
carbon input is dominant, and it is difficult to identify the fraction of
labile fraction responsible for fast OM pool consumption (Pastor et al.,
2011). Moreover, observations show that some organic compounds are
preferentially degraded and become selectively oxidized (Middelburg et al.,
1997; Pozzato et al., 2018). As a result, the model assumed solid phase
organic carbon with two reactive modelled fractions with different
reactivities and
This rate of carbon mineralization of organic matter (mmol m
Secondary redox reaction includes reoxidation of reduced substances
(nitrification, Fe oxidation,
Transport processes in the model are described by molecular diffusion and
bio-irrigation for dissolved species whereas bioturbation is the main
process for mixing the solid phase. In addition, advection occurs in both
the solid and dissolved species. The model dynamics described as a partial
differential equation (PDE) is the general reaction–transport equation
(Berner, 1980). We use a similar paradigm and formulations to that of
Soetaert et al. (1996a). For substances that are dissolved:
For the solid phase:
Diffusive fluxes of solutes across the sediment–water interface are driven
by the concentration gradients between the overlying seawater and the
sediment column. Fick's first law is used to describe the solute flux due to
molecular diffusion,
As a simplifying assumption, material accumulation has no effect on
porosity. We further assumed the porosity profile decreased with depth but
invariant with time. Although, this assumption is a restrictive as the site of
flood deposition can undergo variation in grain size which might affect
their porosity (Cathalot et al., 2010), we proceed noting that the fixed
parameters which define the porosity curve can be changed when necessary.
Thus, using optimized parameters fitted with data in the proximal sites of
Rhône prodelta (Ait Ballagh et al., 2021), porosity (
Bioturbation in the model is characterized by the movement and mixing of
particles by benthic organisms. This is parameterized as a diffusivity
function in space (
Bio-irrigation is modelled in an identical manner to that of biodiffusion,
and acts as a non-local exchange process between the porewater parcels and
the overlying bottom water.
The model is vertically resolved with grid divided into 100 layers (
Schematic of model implementation for the deposition event scenario. Profile from previous time step (left) and after addition of new layer over a predefined depth layer (right). For the solid
The inclusion of the deposition event as a separate external routine to
modify the sediment properties (i.e. porewater species,
The event calculation was carried out dynamically within the same simulation
time. For the solid species, following the flux of organic carbon via the
boundary condition (see Sect. 2.2.7), the portion of organic carbon is
split between the fast and slow decaying component using a proportionality
constant (
It is important to note that this parameter differs from
For the solutes (
The boundary conditions for the model are of three types:
At the sediment–water interface, a Dirichlet concentration condition for
most solutes equalling the bottom water concentration was used,
The model also includes the ability to include time-varying organic carbon
flux with user-specific time series or a functional representation such as
sinusoidal pattern. In the case study presented here, this carbon flux
varies over the annual carbon flux (
The model parameters in Table 2 (for full model parameter, see Table S1 in
Supplement) were derived from previously published model in the Rhône
Delta (Pastor et al., 2011; Ait Ballagh et al., 2021). The organic matter
stoichiometry for both fractions is represented here by the NC ratio
(
The bottom water temperature was fixed at 20
Core parameters used in the model.
For the verification of the model output, data from Pastor et al. (2018) corresponding to the diagenetic situation 26 d after an organic-rich flood were used. We restricted our benchmark to data from the proximal station (station A) near the river mouth, where the impact of this flood discharge is more visible (Fig. 1).
Because the procedure is based on OMEXDIA, complete details of the
derivation can be found in that paper and is referenced therein (Soetaert et
al., 1996a). Here we recap the mathematical formulation of the
method of lines (MOLs) algorithm used by FESDIA. Direct differencing of Eqs. (8)–(10) results to
Equations (8)–(10) implemented as Eq. (22) are integrated in time using an
implicit solver, called lsodes, that is part of the ODEPACK solvers
(Hindmarsh, 1983). This solver uses a backward differentiation method (BDF);
it has an adaptive time step, and is designed for solving systems of
ordinary differential equations where the Jacobian matrix has an arbitrary
sparse structure. The model output time and its time step (
The model application starts by estimating the steady-state condition of the
model using the high level command
This relaxation timescale calculation based on the disappearance of the perturbed signal (via successive profile similarity) may differ from an approach in which the profile returns to a pre-defined “old profile”. Because the exactness of pre-flood and post-flood profiles is difficult to quantify numerically (Wheatcroft, 1990), and the return to the former is frequently driven by slow dynamics, the approach used here can provide a window of estimate for which a particular signal fades toward the background of a theoretically pre-perturbed signal.
The bootstrapping technique used to calculate the uncertainty in the relaxation timescale. The resampled median about a reference provides a replicate over which the standard error estimate is defined. The solid red represents the expected value of the quantity estimated while the vertical red line is the deviation from this expected value.
In this case, we employ a modified bootstrapping technique to estimate the
uncertainty in the relaxation timescale by resampling on the cutoff point
introduced in Eq. (23) (i.e. median,
The 95 % confidence intervals (
The model is initialized as explained in Sect. 2.2.9. Thereafter, for the dynamic
simulation, the model is spin-up for a sufficiently long time to attain
dynamic equilibrium (
For the numerical model experiment, we investigate the sediment's response
to two end-member types of deposition that can represent actual field
observations in the Rhône prodelta (Pastor et al., 2018).
Except for the
Lastly, we conducted a sensitivity analysis of the relaxation timescale for oxygen, sulfate, and DIC concentrations in terms of their variation to the thickness of the new sediment layer and the quantity of organic carbon introduced by the deposition.
We assumed a 15 cm average deposit thickness and conducted simulations with
a thickness variation ranging from 1 to 30 cm. A 5 cm thickness increment
was used for the sensitivity analysis. The
In order to compare the model evolution to field data, we made a comparison
between the simulated profiles 26 d after a flood layer deposition and
data collected in the Rhône prodelta in December 2008 (observed data
collected 26 d after a Cevenol flood). During this flooding period,
riverine discharge delivered
The general pattern of the simulated profile agrees well with the observed
data (Fig. 4). The newly introduced organic carbon-rich sediment resulted in
rapid oxygen consumption. The data for total organic carbon (TOC) shown in
Fig. 4 suggest a good agreement with the model, with high TOC (2.5 wt %–2.0 wt %) deposited at the upper 10 cm. Twenty-six days after the flood, the oxygen
concentration dropped from 250
Overall, the model–data trend was satisfactory with observed depth
distribution of sulfate (
The DIC profile shows a similar trend to the data collected after the flash
flood. Within the depth interval of data, the model tends to follow the
data. It drifts at lower depths, on the other hand, by overestimating the
concentration of DIC observed at deeper layers. Similarly, the modelled
Model and observation depth profile of TOC (%), SO
With a test case of 30 cm of new material deposited during the event (EM1) in the spring, the sediment changes as thus: prior to the event, the oxygen penetration depth (OPD) was about 0.17 cm. The OPD increases to 1.17 cm after the deposition of these low OC materials. The model showed a gradual return to its previous profile within days, with the OPD shoaling linearly with time (Table 4). By day 5, oxygen has returned to the pre-flood profile with similar gradient to the pre-flood state.
Against a background OM flux following the introduction of the flood layer,
the sediment responded quickly. As a result, the perturbation has a
significant effect on sulfate penetration depth, with concentration
remaining nearly constant within the perturbed depth (
Model vs. data comparison for oxygen penetration depth (OPD), flux of oxygen, sulfate, and DIC (26 d after deposition).
Scenario 1 (EM1): model evolution for sulfate following deposition. Relative deviation of successive profile with time shown below. Dashed vertical line signifies cutoff point by the median (dashed horizontal line). Inset: histogram of bootstrap estimate of sulfate relaxation timescale for EM1 with 95 % confidence interval.
Scenario 1 (EM1): model evolution for DIC following deposition. Relative deviation of successive profile with time shown below. Dashed vertical line signifies cutoff point by the median (dashed horizontal line). Inset: histogram of bootstrap estimate of DIC relaxation timescale for EM1 with 95 % confidence interval.
Below the interface with the newly deposited layer, the sediment is enriched
in OM whose mineralization results in a higher sulfate reduction rate (SRR)
at the boundary that delineates the newly deposited layer and the former
sediment–water interface. The simulated SRR falls from 437 mmol C m
Correspondingly, OC mineralization products (such as DIC) were significantly
lower in the upper newly deposited layer, as a consequence of the reduced
quantity of OC brought by the flood. This concentration increased with depth
to about 80 mM. Starting from the deposition, higher production of DIC
below the former SWI led to a distinct boundary in the sediment: a DIC-depleted layer above an increasing DIC with concentrations up to 75 mM trapped in the region below the new–old sediment horizon 20 d after deposition (Fig. 6). This increased DIC production continued
despite complete exhaustion of buried labile fraction with mineralization
driven by the slow decaying component. The depth gradient caused by the
increased DIC production enhances diffusive DIC flux. Following that initial
period, DIC began to revert to its previous state. This slow
re-organization, mostly driven by diffusion continues, with an estimated
recovery time of 5 months (with a 95 % bootstrap confidence interval of
137–147 d respectively), as it temporarily lags behind
A flood deposition scenario of 10 cm thick material with enhanced OC content
was used for the other end-member case experiment (EM2) in autumn. In this
scenario, the modelled sediment exhibits a variety of response
characteristics. The newly introduced sediment resulted in rapid oxygen
consumption. The OPD decreased to 0.74 cm shortly after the event, according
to the model, and stabilized there for days. There was no visible
deformation in the shape of oxygen during its recovery trajectory, and total
oxygen consumption for organic matter mineralization decreased by 8 %
during the first 2 d after the event, from 12 to 11 mmol O
The
Scenario 2 (EM2): model evolution for sulfate following deposition. Relative deviation of successive profile with time shown below. Dashed vertical line signifies cutoff point by the median (dashed horizontal line). Inset: histogram of bootstrap estimate of sulfate relaxation timescale for EM2 with 95 % confidence interval.
Scenario 2 (EM2): model evolution for DIC following deposition. Relative deviation of successive profile with time shown below. Dashed vertical line signify cutoff point by the median (dashed horizontal line). Inset: histogram of bootstrap estimate of DIC relaxation timescale for EM2 with 95 % confidence interval.
Relaxation timescale (
We then examine the sensitivity analysis of the relaxation timescale (
Over all runs varying the enrichment factor (
Similar variation of relaxation time for DIC was simulated for different
In highly dynamic coastal ecosystems, such as RiOmar (river-dominated ocean margins) systems, driven by seasonal variability and meteorologically extreme events, the response of early diagenetic processes to time varying deposition of organic matter is generally non-stationary (Tesi et al., 2012). While dynamic equilibrium as a steady-state condition may be reasonable in the case of seasonal variability, such an assumption may fail in cases of instantaneously event-driven deposition. An intermittent supply of sediment and OC, like those presented here, can cause a change in the system's properties on a short- or long-term basis. Previous works have highlighted excursions in sediment redox boundary (Katsev et al., 2006), flux of solutes at the sediment–water interface (Rabouille and Gaillard, 1990), and modification of other system properties due to depositional flux of organic matter. Thus, the premise of steady-state conditions in early diagenetic processes which often depends on the temporal resolution of the observation might need revisiting especially in areas of episodic sedimentation (Wheatcroft, 1990; Tesi et al., 2012). Here, we discuss the evolution and dynamics of a non-stationary sedimentary system following a singular perturbation.
Non-steady-state models are increasingly being applied in dynamic coastal
environments, but they are still primarily based on forcing from smooth
varying boundaries that mimic seasonal forcing or long-term variability
(Soetaert et al., 1996b; Rabouille et al., 2001a; Zindorf et al., 2021).
Explicit consideration of abrupt changes in the upper boundary of the model
caused by events such as landslides, flash flooding, turbiditic transfer of
materials on a continental slope, and trawling is still relatively
uncaptured by these models (but see De Borger et al., 2021, for inclusion of
erosion events). In this paper, we adapt OMEXDIA (Soetaert et al., 1996a), a
well-known reaction transport model, to investigate the changes in the solid
and liquid phases during massive deposition event. Our efforts highlight the
algorithm's utility in incorporating this process with minimal numerical
issues. The model represented the basic characteristics of the data derived
from the November/December 2008 flood event at station A in the Rhône
Delta's depocenter (Fig. 4). The simulated flux was also in agreement with
the estimate from field data, as diffusive oxygen uptake (DOU) rate sampled
26 d after the event (8 December 2008) was
Flooding events can transport large amounts of material through the river to transitional coastal environments such as deltas and estuaries. River floods can account for up to 80 % of terrigenous particle inputs (Antonelli et al., 2008; Zebracki et al., 2015), and they can have a significant impact on geomorphology (Meybeck et al., 2007), ecosystem response, and biogeochemical cycles (Mermex Group et al., 2011). If the source materials have a different organic matter composition (Dezzeo et al., 2000; Cathalot et al., 2013), the rapid deposition of these flood materials can alter diagenetic reactions and resulting fluxes.
Furthermore, the relaxation timescale associated with the sediment recovery
following this external perturbation can be important in terms of the process
affecting the biogeochemistry of solid and solutes species. With a series of
numerical experiments ranging in between two end-members of the input
spectrum for flood events such as those in the Rhône prodelta (Pastor et
al., 2018), our study revealed contrasting sedimentary responses and
associated typical timescales at which porewater profiles relax back to
undisturbed state. Using a simple metric for estimating relaxation timescale
of the perturbation, our calculations for the first end-member scenario (EM1)
show that the upper bound of the timescale of relaxation for oxygen is
Other terminal electron acceptors (TEAs), such as
Such a long time lapse for the recovery of an element with a complex
pathway, such as
In addition, our calculations show that the timescale of return to the
previous “pre-flood” profile is bracketed by the range of recovery due to
purely molecular diffusion, putting an upper bound on our estimate. For
example, using the Einstein's approximation, a species such as oxygen with a
sediment diffusion coefficient (
The relaxation time may also vary depending on the diagenetic interaction,
and the characteristics of the organic matter available for degradation.
This difference in characteristics was partially imposed in our study by
assuming variations of
With the sensitivity analysis, we further explore the variation of
relaxation timescale under variation of the thickness of layer and
enrichment factor of input material given by
In terms of the recovery time as a function of the availability of labile
OC, our results revealed a contrasting pattern for oxygen and sulfate.
Several factors related to how different oxidants react with sediment matrix
disturbances can explain these differences:
With oxygen that has a high molecular diffusion coefficient, variations in
relaxation time depend on the levels of labile OC, with thin sediments
containing a high level of labile OC showing a shorter recovery time than
thicker sediments with a low OC content. This pattern can be attributed to
the higher relative importance of oxygen consumption in OM-poor sediment
relative to the OM-rich sediment. For low thickness deposits, sulfate and DIC relaxation times were more or
less constant. However, a longer relaxation time was simulated for larger
deposits and higher labile OC. This can be attributed to the increased
distance required for solutes to migrate back after the event. This is
clearly the case for sediment thicknesses greater than 14 cm. Such two-way
dynamics could be explained by the fact that biological reworking and
physical mixing within the surface mixing layer (SML) can improve OC
degradation by promoting the replenishment of electron acceptors (i.e.
oxygen, sulfate, nitrate, and metal oxyhydroxides) (Aller and Aller, 1992;
Aller, 2004), resulting in a shorter recovery time for the porewater profile
to reorganize within the SML. This critical depth could also be the distance horizon at which the slow
diffusion of the profile when retracting back to its pre-flood profile
becomes an important factor in controlling the relaxation timescale. This is
especially true for DIC, where the connection is more obvious. It has been
proposed that when flood deposits extend beyond the sediment bio-mixing
depth, the relaxation time for the constituent species is determined by the
concentration gradient between the historical and newly deposited layers
(Wheatcroft, 1990). In our sensitivity analysis, higher
Because it is based on the well-established OMEXDIA model, FESDIA has
several capabilities that make it suitable for a wide range of application
domains for non-steady-state early diagenetic simulation. However, due to
assumptions made during model development, some limitations in model usage
must be considered:
First, we assumed that porosity is time independent. This may not be the
case in some coastal systems that receive sediment materials from regions
with distributary channels, which contribute particles of varying origin and
grain size (Grenz et al., 2003; Cathalot et al., 2010). The composite
sediment that is eventually transported to the depocenter by a flood event
may differ in porosity, and thus vary temporarily depending on when and where
the source materials are derived during the flood event. In this case, model
estimates of fluxes in dissolved species may be over/underestimated. The
resulting porosity in the new layer is barely predictable and could range
between 0.65 and 0.85 in the proximal zone of the prodelta (Grenz et al.,
2003; Cathalot et al., 2010), allowing us to justify our assumption. Second, in our examples, we assumed that the burial rate and bioturbation
were constant. With the introduction of these flood events, such assumptions
may be called into question (Tesi et al., 2012). In addition, benthic
animals respond to other perturbation event such as trawling in ways that
may warrant explicit description of their recovery, which is linked to
bioturbation (De Borger et al., 2021; Sciberras et al., 2018). While some
coastal sediment burial rates have been shown to vary seasonally (Soetaert
et al., 1996b; Boudreau, 1994), in the proximal zone of the Rhône
prodelta, approximately 75 % of sediment deposition can occur during the
flood (e.g. 30 cm d The current FESDIA version does not include a diffusive boundary layer,
which can be important for material exchange between the overlying bottom
water and the sediment. This is critical for calculating fluxes of species
such as
In terms of future development, we hope to improve the model's diagenetic
pathways, particularly for the iron and sulfur cycles. Furthermore,
processes such as calcite formation have been shown to affect DIC profile by
10 %–15 % in the proximal sites of Rhône prodelta (Rassmann et al.,
2020), thus might necessitate inclusion in future versions of the model. This
will enable FESDIA to account for carbonate system dynamics in marine
sediment which can play an important role in the coast carbon cycle (Krumins et
al., 2013).
While one main focus of this study is on providing a quantitative estimate
of relaxation time, the difficulty of objectively defining what
The need to comprehend extreme events and their relationship to marine biogeochemistry prompted the development of novel methods for diagnosing flood-driven organic matter pulses in coastal environments. In this paper, we propose a new model (FESDIA) for characterizing flood deposition events and the biogeochemical changes that result from them. This type of event can have an impact on the benthic communities and the response of the whole ecosystem (Smith et al., 2018; Bissett et al., 2007; Gooday, 2002). Our modelling study shows that the post-depositional sediment response varies depending on the input characteristics of the layer deposit. For instance, we tested the combined effect of enrichment of labile organic carbon and deposit thickness on space–time distribution and relaxation time of key dissolved species (oxygen, sulfate, DIC). This integral timescale of relaxation is constrained by the intrinsic properties of the solutes (diffusion) and the characteristics of the flood input (thickness and concentration of labile organic carbon). In essence, the findings from this study highlight the importance of the quantity and quality of organic carbon in modulating the sediment response following such a singular perturbation, and the role of flood events with heterogeneous quantitative contributions in the coastal ocean.
The full model equation explained in Sect. 2.3.2 is described fully below.
Organic matter is composed of three fractions: fast degradable organic
matter, slow degradable organic matter, and refractory organic matter. Given
the long timescale for the degradability of the refractory OM, it is
parameterized using
Degradation of organic matter:
Limitation terms: the limitation of a mineralization pathway by the availability of the
oxidant is modelled by a MOND-type hyperbolic limitation function with
inhibition of a pathway represented by a reciprocal hyperbolic function.
Depth-dependent kinetic reaction: this limitation is used to reconstruct the vertical distribution of the
successive mineralization pathways with a rescaling term “lim” to ensure
that the sum of the individual pathway equal the total degradation rate.
Secondary reaction: the re-oxidation of reduced substance and other secondary reactions are
modelled with a first-order reaction term.
As a whole, the model is bundled as an R package for easy accessibility and
can be downloaded from R-forge (
The data and paper used to evaluate the model (Pastor et al., 2018) can be
found at
The supplement related to this article is available online at:
All authors contributed to the paper in several capacities. The project was supervised by CR and EV. SIN, CR, and EV conceptualized the method design, result interpretation, and assist in the initial draft of the paper. Model development was jointly design by KS and SIN. LP and BL provided insight on the data used in the model.
The contact author has declared that none of the authors has any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research has been supported by the Université de Versailles Saint-Quentin-en-Yvelines (Ecolé doctorale des science de l'environment, Ile de France (grant no. ED129)) and the Centre National de la Recherche Scientifique (INSU EC2CO DELTARHONE).
This paper was edited by Sandra Arndt and reviewed by two anonymous referees.