Interactions between seawater and benthic systems play an important role in global biogeochemical cycling. Benthic fluxes of some chemical elements (e.g., C, N, P, O, Si, Fe, Mn, S) alter the redox state and marine carbonate system (i.e., pH and carbonate saturation state), which in turn modulate the functioning of benthic and pelagic ecosystems. The redox state of the near-bottom layer in many regions can change with time, responding to the supply of organic matter, physical regime, and coastal discharge. We developed a model (BROM) to represent key biogeochemical processes in the water and sediments and to simulate changes occurring in the bottom boundary layer. BROM consists of a transport module (BROM-transport) and several biogeochemical modules that are fully compatible with the Framework for the Aquatic Biogeochemical Models, allowing independent coupling to hydrophysical models in 1-D, 2-D, or 3-D. We demonstrate that BROM is capable of simulating the seasonality in production and mineralization of organic matter as well as the mixing that leads to variations in redox conditions. BROM can be used for analyzing and interpreting data on sediment–water exchange, and for simulating the consequences of forcings such as climate change, external nutrient loading, ocean acidification, carbon storage leakage, and point-source metal pollution.
Oxygen depletion and anoxia are increasingly common phenomena observed in the World Ocean, inland seas, and coastal areas. Observations show a decline in dissolved oxygen concentrations at continental margins in many regions, and this has been linked to both an increase in anthropogenic nutrient loadings and a decrease in vertical mixing (e.g., Diaz and Rosenberg, 2008; Rabalais et al., 2002; Richardson and Jørgensen, 1996). Although bottom waters may be permanently oxic or anoxic, they oscillate seasonally between these extremes in many water bodies (Morse and Eldridge, 2007). Such oscillations typically result from variations in the supply of organic matter (OM) to the sediment–water interface (SWI), from the hydrophysical regime (mixing/ventilation), and from nutrient supply (river runoff). Frequently, oxic conditions during periods of intense mixing are followed by near-bottom suboxia or anoxia after the seasonal pycnocline forms, restricting aeration of the deeper layers. This occurs, for instance, on the Louisiana shelf (Morse and Eldridge, 2007; Yu et al., 2015) and in Corpus Christi Bay (McCarthy et al., 2008), the Sea of Azov (Debolskaya et al., 2008), and Eleusis Bay (Pavlidou et al., 2013). The redox state and oxygenation of near-bottom water varies due to the transport of oxidized and reduced species across the SWI and biogeochemical processes occurring in the sediments (Cooper and Morse, 1996; Jorgensen et al., 1990; Roden and Tuttle, 1992; Sell and Morse, 2006). The sediments generally consume oxygen due to the deposition of labile OM and the presence of reduced forms of chemical elements. Their capacity to exchange oxygen with the pelagic layer is limited, as near-bottom water is usually characterized by low water velocity and reduced mixing in the vicinity of the SWI (Glud, 2008). In some cases, a high benthic oxygen demand (BOD) associated with local OM mineralization and low mixing rates can cause anoxia in the bottom water. This may lead to death, migration, or changed behavior of the benthic macro- and meiofaunal organisms responsible for bioturbation and bioirrigation (Blackwelder et al., 1996; Sen Gupta et al., 1996; Morse and Eldridge, 2007), which in turn can greatly slow down the transport of solid and dissolved species inside the sediments and therefore the rates of oxidative reactions. Under such conditions, sedimentary sulfides can build up, and dissolution of carbonate minerals may come to a halt (Morse and Eldridge, 2007). When oxic conditions return, there can be an “oxygen debt” of reduced species in the water column (Yakushev et al., 2011) which may buffer and delay reoxygenation of the sediments (Morse and Eldridge, 2007).
In areas experiencing seasonal hypoxia/anoxia, the processes taking place in the water column and in the sediments are tightly coupled to each other, as well as to the fluxes and exchanges of organic matter over a range of timescales. An accurate understanding of physical, chemical, and biological processes driving changes in redox conditions is needed to predict the distribution of hypoxia/anoxia in a given environment. This “benthic–pelagic coupling” broadly encompasses the fluxes of OM to the sediments and the return fluxes of inorganic nutrients to the water column. Variations in supply, dynamics, and reactivity of OM affect benthic communities (Pearson and Rosenberg, 1978), sediment and porewater geochemistry (Berner, 1980), and nutrient and oxygen fluxes at the SWI (Boudreau, 1997).
Many previous studies have demonstrated the capability of sophisticated reactive transport codes for integrated modeling of biogeochemical cycles in sediments (Boudreau, 1996; Van Cappellen and Wang, 1996; Couture et al., 2010; Jourabchi et al., 2005; Katsev et al., 2006, 2007; Paraska et al., 2014; Soetaert et al., 1996). The water column redox interface was also specifically targeted in the models of Konovalov et al. (2006) and Yakushev et al. (2006, 2007, 2011). However, the process of integrating such models with pelagic biogeochemical models to produce benthic–pelagic coupled models has only begun in recent years.
As of the year 2000, benthic–pelagic coupling was either neglected or crudely approximated in many pelagic biogeochemical and early diagenetic models (Soetaert et al., 2000). One of the first fully coupled physical–pelagic–benthic biogeochemical modes was developed for the Goban Spur shelf break area to examine the impact of in situ atmospheric conditions on ecosystem dynamics, to understand biogeochemical distributions in the water column and the sediments, and to derive a nitrogen budget for the area. This model was most suited for testing the impact of short-term physical forcing on the ecosystem (Soetaert et al., 2001).
Later, several coupled benthic–pelagic models were produced with an emphasis on studying eutrophication (Cerco et al., 2006; Fennel et al., 2011; Soetaert and Middelburg, 2009) or hypoxia in various locations including Tokyo Bay (Sohma et al., 2008), the Baltic Sea (Reed et al., 2011), the North Sea oyster grounds (Meire et al., 2013), and the Southern Bight (Lancelot et al., 2005). Another model was created to investigate early diagenesis of silica in the Scheldt estuary, with benthic–pelagic coupling only of silica (Arndt and Regnier, 2007).
By coupling two quite sophisticated models ECOHAM1 and C.CANDI, a 3-D model for the North Sea was created where pelagic model output was used to force a benthic biogeochemical module (Luff and Moll, 2004). Another physical–biological model for the North Sea, PROWQM, is more complex than ECOHAM1 and has been coupled to a benthic module to simulate seasonal changes of chlorophyll, nutrients, and oxygen at the PROVESS north site, southeast of the Shetland Islands (Lee et al., 2002). Brigolin et al. (2011) developed a spatially explicit model for the northwestern Adriatic coastal zone by coupling a 1-D transient early diagenesis model with a 2-D reaction-transport pelagic biogeochemical model. Currently, the most known and established coupled model is ERSEM – the European Regional Seas Ecosystem Model, which was initially developed as a coastal ecosystem model for the North Sea and which has evolved into a generic tool for ecosystem simulations from shelf seas to the global ocean (Butenschön et al., 2016).
The BROM model described herein is a fully coupled benthic–pelagic model with a special focus on deoxygenation and redox biogeochemistry in the sediments and benthic boundary layer (BBL). The BBL is “the part of the marine environment that is directly influenced by the presence of the interface between the bed and its overlying water” (Dade et al., 2001). Physical scientists tend to prefer the term “bottom boundary layer”, but this is largely synonymous with the BBL (Thorpe, 2005). Within BROM, the term BBL refers to the lower parts of the fluid bottom boundary layer where bottom friction strongly inhibits current speed and vertical mixing, hence including the viscous and logarithmic sublayers up to at most a few meters above the sediment. This calm-water layer plays a critical role in mediating the interaction of the water column and sediment biogeochemistry and in determining, e.g., near-bottom oxygen levels, yet it remains poorly resolved in most physical circulation models. For BROM, we have developed an accompanying offline transport module (BROM-transport) that uses output from hydrodynamic water column models but solves the transport-reaction equations for a “full” grid including both water column and sediments. BROM-transport uses greatly increased spatial resolution near the SWI, and thereby provides explicit spatial resolution of the BBL and sediments.
The goal of this work was to develop a model that captures key biogeochemical
processes in the water and sediment and to analyze the changes occurring in
the BBL and SWI. As a result, BROM differs from existing biogeochemical
models in several key respects. BROM features explicit, detailed descriptions
of many chemical transformations under different redox conditions, and tracks
the fate of several chemical elements (Mn, Fe, and S) and compounds
(MnCO
BROM consists of two modules, BROM-biogeochemistry and BROM-transport. BROM-biogeochemistry builds on ROLM (RedOx Layer Model), a model constructed to simulate basic biogeochemical structure of the water column oxic/anoxic interface in the Black Sea, Baltic Sea, and Norwegian fjords (He et al., 2012; Stanev et al., 2014; Yakushev et al., 2009, 2006, 2007, 2011). In BROM-biogeochemistry, we extended the list of modeled compounds and processes (Fig. 1). BROM considers interconnected transformations of species of N, P, Si, C, O, S, Mn, and Fe, and resolves OM in nitrogen currency. OM dynamics include parameterizations of OM production (via photosynthesis and chemosynthesis) and OM decay via oxic mineralization, denitrification, metal reduction, sulfate reduction, and methanogenesis. To provide a detailed representation of changing redox conditions, OM in BROM is mineralized by several different electron acceptors and dissolved oxygen is consumed during both mineralization of OM and oxidation of various reduced compounds. Process inhibition in accordance with redox potential is parameterized by various redox-dependent switches. BROM also includes a module describing the carbonate equilibria; this allows BROM to be used to investigate acidification and impacts of changing pH and saturation states on water and sediment biogeochemistry.
Flow chart of biogeochemical processes represented in the
Benthic RedOx Model (BROM), showing the transformation of sulfur species
The physical domain of BROM-transport spans the water column, BBL, and upper layers of the sediments in a continuous fashion. This allows for an explicit, high-resolution representation of the BBL and upper sediments, while also allowing the boundary conditions to be moved as far as possible from these foci of interest, i.e., to the air–sea interface and to deep in the sediment.
BROM is integrated into an existing modular platform (FABM) and is therefore coded as a set of reusable “LEGO brick” components, including the offline transport driver BROM-transport and modules for ecology, redox chemistry, and carbonate chemistry. This means that BROM-transport can be used with all biogeochemical modules available in FABM, including, e.g., the modules comprising ERSEM, and that BROM biogeochemical modules can be used in all other 1-D and 3-D hydrodynamic models supported by FABM (e.g., GOTM, GETM, MOM5, NEMO, FVCOM). Individual BROM modules can also be coupled to existing ecological models to expand their scope, e.g., by providing descriptions of redox and carbonate chemistry. Using the FABM framework thus facilitates the transparent and consistent setup of complex biogeochemical reaction networks for the prediction of hypoxia/anoxia while harnessing the capabilities of various hydrophysical drivers.
BROM-biogeochemistry consists of three biogeochemical submodules: BROM_bio (ecological model), BROM_redox (redox processes), and BROM_carb (carbonate system). Interactions between modeled variables are either kinetic (e.g., OM degradation) or equilibrium processes (e.g., carbonate system equilibration) (Boudreau, 1996; Jourabchi et al., 2008; Luff et al., 2001). In general, the redox reactions are fast in comparison with the other processes and a typical model time step. Species involved in such reactions are therefore set to equilibrium concentrations using mass action laws and equilibrium constants for seawater (Millero, 1995). Total scale pH is also diagnosed at every time step, mainly as a function of dissolved inorganic carbon (DIC) and total alkalinity (Alk) which are both prognostic (state) variables.
The model has 33 state variables (Table 1), including frequently measured
components such as hydrogen sulfide (H
State variables of BROM.
The overall goal of the ecosystem representation is to parameterize the key features of OM production and decomposition, following Redfield and Richards stoichiometry (Richards, 1965). We divide all the living OM (biota) into Phy (photosynthetic biota), Het (non-microbial heterotrophic biota), and four groups of “bacteria” which may be considered to include microbial fungi. These latter are Baae (aerobic chemoautotrophic bacteria), Baan (anaerobic chemoautotrophic bacteria), Bhae (aerobic heterotrophic bacteria), and Bhan (anaerobic heterotrophic bacteria). OM is produced photosynthetically by Phy and chemosynthetically by bacteria, specifically by Baae in oxic conditions and by Baan in anoxic conditions. Growth of heterotrophic bacteria is tied to mineralization of OM, favoring Bhae in oxic conditions and Bhan in anoxic conditions. Secondary production is represented by Het, which consumes phytoplankton as well as all types of bacteria and dead particulate organic matter (detritus, which is also explicitly modeled). The effect of suboxia and anoxia is parameterized by letting the mortality of aerobic organisms depend on the oxygen availability.
A detailed account of processes representing the inorganic cycling of N, S, Mn, Fe, and P is given in the description of ROLM (Yakushev et al., 2007, 2013a), while the process parameterization, chemical reactions, rates, and stoichiometric constants values are summarized in Tables 2–4. Table 2 also describes the redox-dependent switches, nutrient limitation, and substrate consumption rates for heterotrophs. The redox-dependent switches are mostly based on hyperbolic tangent functions which improve system stability compared with discrete switches. The nutrient limitation and heterotrophic transfer functions are based on squared Monod laws for nutrient–biomass ratio, which also stabilizes the system compared with Michaelis–Menten and Ivlev formulations. Here, we describe the parameterization of carbon that was not considered in ROLM and was not described in Yakushev (2013).
Parameterization of the biogeochemical
processes:
Continued.
Continued.
Continued.
Total alkalinity,
Biogeochemical processes can lead to either increase or decrease of
alkalinity, and alkalinity can be used as an indicator of specific
biogeochemical processes
(Soetaert et al., 2007). Organic matter production can affect alkalinity via
the “nutrient-H
BROM also considers the effect on alkalinity of the following redox reactions
occurring in suboxic and anoxic conditions via production or consumption of
[OH
Equilibration of the carbonate system was considered as a fast process occurring within seconds (Zeebe and Wolf-Gladrow, 2001). Accordingly, the equilibrium solution was calculated at every time step using an iterative procedure. The carbonate system was described using standard approaches (Lewis and Wallace, 1998; Munhoven, 2013; Roy et al., 1993; Wanninkhof, 2014; Wolf-Gladrow et al., 2007; Zeebe and Wolf-Gladrow, 2001). The set of constants of Roy et al. (1993) was used for carbonic acid. Constants for boric, hydrofluoric, and hydrogen sulfate alkalinity were calculated according to Dickson (1992), for silicic alkalinity according to Millero (1995), for ammonia alkalinity according to Luff et al. (2001), and for hydrogen sulfide alkalinity according to Luff et al. (2001) and Volkov (1984). The ion product of water was calculated according to Millero (1995). Total scale pH was calculated using the Newton–Raphson method with the modifications proposed in Munhoven (2013). Precipitation and dissolution of calcium carbonate were modeled following the approach of Luff et al. (2001) (Table 2).
BROM-biogeochemistry can be coupled online with various hydrodynamic models using FABM, but this may require extensive adaptation of the hydrodynamic model to resolve the BBL and upper sediments. We have therefore developed a simple 1-D offline transport-reaction model, BROM-transport, whose model domain spans the water column, BBL, and upper layers of the sediments, with enhanced spatial resolution in the BBL and sediments. All options and parameter values for BROM-transport are specified in a runtime input file brom.yaml. A step-by-step guide to running BROM-transport is provided in Appendix A.
The time–space evolution of state variables in BROM-transport is described
by a system of 1-D transport-reaction equations in Cartesian coordinates. In
the water column, the dynamics are
Parameter names, notations, values, and units of the coefficients used in the model:
Continued.
Continued.
Continued.
Rates of biogeochemical production/consumption of the model compartments:
Continued.
In the sediments, dissolved substances or solutes obey the dynamics
The porosity
Simulated seasonal variability of the selected modeled
chemical parameters (
Diffusion within the sediments is assumed to be strictly “intraphase”
(Boudreau, 1997), hence the Fickian gradients in Eqs. (2)
and (3) are formed using the concentration per unit volume porewater for
solutes and per unit volume total solids for particulates. The total solute
diffusivity
Diffusion between the sediments and water column, i.e., across the SWI,
raises a subtle issue in regard to particulates. Here, any diffusive flux
cannot be strictly intraphase, because particulates are modeled as
[mmol m
The burial velocities
Finally, the process of bioirrigation, whereby benthic organisms flush out
their burrows with water from the sediment surface, is modeled as a
non-local solute exchange (following Aller, 2001; Meile et al., 2001; Rutgers Van Der Loeff
and Boudreau, 1997; Schlüter et al., 2000):
Vertical distributions of the modeled chemical parameters
(
Vertical distributions of the modeled chemical parameters
(
Equations (1)–(3) are integrated numerically over a
single combined grid (water column plus sediments) and using the same model
time step in both water column and sediments. All concentrations are stored
internally and input/output in units [mmol m
Diffusive updates are calculated either by a simple forward-time
central-space (FTCS) algorithm or by a semi-implicit central-space algorithm
adapted from a routine in the General Ocean Turbulence Model, GOTM (Umlauf et
al., 2005). Bioirrigation and reaction updates are calculated from forward
Euler time steps, using FABM to compute
BROM-transport also provides an option to divide the diffusion and
sedimentation updates into smaller time steps related to the
sources-minus-sinks time step by fixed factors, since the physical transport
processes are often numerically limiting (Butenschön et al., 2012).
The default time step is 0.0025 days or 216 s, which is much longer
than the characteristic equilibration timescale of the CO
The vertical grid in BROM-transport is divided into the pelagic water column, the BBL, and the sediments. The pelagic water column grid is either set as uniform with height/spacing set by the brom.yaml file (see Sect. S1 in the Supplement), or it is read from the NetCDF forcing input file (see below), with an option to decrease resolution by subsampling. In principle, the NetCDF input from the hydrodynamic model may already include a fully resolved BBL, but in practice we find this is rarely the case. BROM-transport therefore allows the user to “insert” a high-resolution BBL into the bottom of the input water column. This BBL has non-uniform grid spacing with layer thickness decreasing geometrically towards the SWI, reaching O (cm) thickness for the fluff layer, based on parameters from the brom.yaml file. For the upper sediments, the layer thickness is increased geometrically moving down from the SWI, from O (mm) thickness in the surface layer to O (cm) thickness deeper in the sediments, again based on brom.yaml parameters. The result is a full grid with non-uniform spacing and maximum resolution near the SWI. As in many ocean models (e.g., ROMS, GOTM) the vertical grid in BROM-transport is staggered: temperature, salinity, and biogeochemical concentrations are defined at layer midpoints, while diffusivities, sinking/burial velocities, and resulting transport fluxes are all defined on layer interfaces.
Initial conditions for all concentrations in Eqs. (1)–(3) can be provided
by either using the initialization values defined in the fabm.yaml file (see
Sect. S2 in the Supplement) as uniform initial conditions for each variable, or by providing
the initial conditions for all variables at every depth in a text file with
a specific format. Typically, these initial-condition text files are
generated by running the model to a steady state annual cycle and saving the
final values as the desired start date. Alternatively, they could be
generated by interpolating/smoothing data, in which case the user should
note that the input concentrations must be in units [mmol m
BROM-transport presently allows the user to choose between four different
types of boundary conditions for each variable and for upper and lower
boundaries: (1) no gradient at the bottom boundary (no diffusive flux) or
no flux at the surface boundary, except where parameterized by the FABM
biogeochemical model (i.e., for O
Under option 1, and using BROM-biogeochemistry, a surface O
BROM-transport includes two simple Beer–Lambert
attenuation models to calculate in situ 24 h average photosynthetically
active radiation (PAR) as needed by BROM-biogeochemistry and many other
biogeochemical models. The first is derived from the current ERSEM default
model (Blackford et al., 2004;
Butenschön et al., 2016) and models the total attenuation as
Simulated seasonal variability of biogeochemical
transformation rates just above the sediment water interface, showing the
rates of DON mineralization with oxygen, nitrate, nitrite, Mn(IV), Fe(III),
SO
Simulated seasonal variability of vertical diffusive
fluxes from the benthic boundary layer to the sediments of oxygen, hydrogen
sulfide, nitrate, silicate, ammonia, Mn(II), and Fe(II). Positive fluxes are
downward and negative fluxes are upward. Units are in mmol m
BROM-transport requires forcing inputs at least for temperature, salinity, and vertical diffusivity at all depths in the pelagic water column and for each day of the simulation. These may be provided from an input subroutine that creates simple, hypothetical profiles, or from text/NetCDF files containing data from interpolations of measurements or hydrodynamic model output. Forcing time series of surface irradiance and ice thickness may also be read as NetCDF input. BROM-transport then uses these inputs in combination with parameters set in the runtime input file brom.yaml (see Sect. S1) to solve the transport-reaction equations on a “full” vertical grid including pelagic water column, BBL, and sediment subgrids.
In order to run, BROM-transport must extend the input pelagic (temperature,
salinity, diffusivity) forcings over the full grid. Temperature and salinity
in the BBL and sediments are set as uniform and equal to the values at the
bottom of the input pelagic water column for each day. The vertical
diffusivity needs a more careful treatment, as it is the main defining
characteristic of the pelagic vs. BBL vs. sediment environments. Within the
water column, the total vertical diffusivity
Optional forcings for BROM-transport include 24 h average surface
irradiance Eair(
A North Sea hydrodynamic scenario was used to demonstrate the ability of BROM to reproduce the biogeochemical mechanisms of oxic/anoxic transformations. Complete lists of the model options and parameter values used are given in Sect. S1 (brom.yaml input file for BROM-transport) and Sect. S2 (fabm.yaml input file for BROM-biogeochemistry).
The BROM-transport water column extended from 0 to 110 m, with a pelagic spatial resolution of 1 m inherited from the GOTM hydrodynamic model used to provide forcings. A high-resolution BBL was inserted from 109.5 to 110 m, with layer thickness decreasing from approximately 25 to 3 cm in the fluff layer. Sediment grid points were added to cover the upper 10 cm of sediments with layer thickness increasing from 0.5 mm in the surface layer to 1 cm at depth. This choice of grid does not explicitly resolve the DBL (default thickness 0.5 mm) but the main DBL function of limiting solute exchange between the BBL and sediments is largely fulfilled by the fluff layer (thickness 3 cm) and upper sediment layer (thickness 0.5 mm). The model time step for BROM-transport was set to 0.0025 days (216 s).
Upper boundary conditions included sinusoidal, time-varying Dirichlet
boundary conditions for nitrate, phosphate, and silicate, implying net
influxes and outfluxes of surface nutrients, as well as the default
parameterized air–sea fluxes of O
The pelagic water column was forced by output from a GOTM hydrodynamical simulation for temperature, salinity, and vertical diffusivity (taken from the salinity diffusivity) and surface irradiance calculated using the sinusoidal option. We aimed for a solution representative of “present day” and therefore treated the GOTM forcing as representative of a “normal year”. BROM-transport was spun up from vertically homogeneous initial conditions for 100 model years with repeated-year forcings and boundary conditions. After this time, a quasi-stationary solution with seasonally forced oscillations of the biogeochemical variables had been reached.
The results of these calculations were written to an output file in NetCDF format, including the daily vertical distributions of model state variables, diagnostic rates of biogeochemical transformations, and fluxes associated with diffusion and sedimentation. This output can be visualized by any NetCDF-compatible software.
The model simulated the periodic replacement of oxic with anoxic conditions in the BBL following seasonal mixing and OM production. The simulation demonstrates the characteristic features of biogeochemical profiles in the water column, BBL, and upper sediments, as well as their variability under changing redox conditions (Figs. 2–4).
During intensive mixing conditions in winter, the water column is well
oxygenated and the oxic/anoxic interface is located at a depth of several centimeters
in the sediments (Figs. 2, 3). In summer, just after the spring bloom,
an enrichment of the sediment surface with fresh OM and a restricted oxygen
supply leads to the consumption of O
Figure 5 shows the rate of OM mineralization with a variety of electron acceptors. Oxygen is consumed during OM mineralization in summer and autumn and, after its complete depletion, denitrification dominates, with both nitrate reduction and nitrite reduction playing significant roles. The rate of mineralization of OM with Mn and Fe oxides is small, but as these processes prevent mineralization with sulfate, they cause a lag of a few days between the depletion of oxygen and the appearance of hydrogen sulfide in the water column (Figs. 2, 5). The amount of labile degradable OM is relatively small and mineralization with sulfate completely removes the remaining OM, thus preventing methanogenesis (Fig. 5).
The seasonal variability of the sediment–water fluxes clearly demonstrates the appearance in the bottom water of reduced forms of N, Mn, Fe, and phosphate (Fig. 6).
Generally, the concentrations, vertical distributions, and benthic–pelagic fluxes of the parameters considered in the model are reasonable and are within observed ranges for the North Sea (Queirós et al., 2014) and some other regions with temporary bottom anoxia (Almroth et al., 2009; McCarthy et al., 2008; Morse and Eldridge, 2007; Pakhomova et al., 2007; Queirós et al., 2014; Yu et al., 2015).
This paper presents a description of BROM, a fully coupled pelagic–benthic
model that provides an integrated framework to study the biogeochemistry of
a water column and upper sediments. BROM simulates changes in redox
conditions and their impact on the distributions of a wide range of
biogeochemical variables. In particular, BROM provides a detailed
description of the fate and availability of dissolved oxygen and hydrogen
sulfide: the former essential for macroscopic marine life, the latter highly
toxic to it. BROM can therefore provide valuable information to ecological
studies, particularly in the context of multistressor impacts. The model
suggests that the timing of hydrogen sulfide release into the pelagic is
linked to the dynamics of several electron acceptors that are themselves of
limited interest for biogeochemical and ecological purposes, and that are
therefore rarely included in models. The ability of BROM to simulate and
forecast H
This paper was not devoted to a detailed validation of BROM with in situ data; we plan to explore this in future work. A qualitative analysis of the model results (Sect. 3) suggests that the model can produce realistic distributions and fluxes of key biogeochemical variables during periodic changes in redox conditions.
In summary, we present a new benthic–pelagic biogeochemical model (BROM) that combines a relatively simple pelagic ecosystem model with a detailed biogeochemical model of the coupled cycles of N, P, Si, C, O, S, Mn, and Fe in the water column, benthic boundary layer, and sediments, with a focus on oxygen and redox state. BROM should be of interest for the study a range of environmental applications in addition to hypoxia, such as benthic nutrient recycling, redox biogeochemistry, eutrophication, industrial pollution from trace elements, organic loading, and ocean acidification.
The model as presented consists of two components. The first is a set of
biogeochemical modules (brom/redox, brom/bio, brom/carb, brom/eqconst),
available as part of the official FABM distribution (
Also, you can run BROM without any Fortran compiler using a Win32 executable
file (which can be downloaded from
As BROM's biogeochemical modules are built on FABM, they can be used from a wide range of 1-D and 3-D hydrodynamic models, including GOTM, GETM, ROMS, MOM, NEMO, and FVCOM (a ROMS-FABM coupler has been developed by P. Wallhead; NEMO-FABM and FVCOM-FABM couplers have been developed by the Plymouth Marine Laboratory; contact J. Bruggeman for information).
Results shown in this paper were produced with BROM-transport tag v1.1 and
the BROM-biogeochemistry code in FABM tag v0.95.3, available from the above
repositories. The simulation was run using the netCDF/.yaml input files
found in the data folder of the BROM-transport repository. However, we
envisage BROM to be further developed in a backward compatible manner, and
encourage users to adopt the latest version of the code. Step-by-step
instructions for running BROM are found in Appendix A. Both FABM and
BROM-transport are distributed under the GNU General Public License
(
Installation requires a Fortran 2003-capable compiler, e.g.,
gfortran 4.7 or higher, or the Intel Fortran compiler version 12.1 or higher.
In our demonstration, we used the Intel Fortran Compiler
version 15.0.4.221. Additionally, a NetCDF library compatible with the
chosen Fortran compiler is required. CMake software should be installed.
After ensuring these prerequisites are in place, create a directory to hold
the BROM model code and associated input and output files. Detailed
instructions for installation are provided at the BROM repository
( Preparation of input files consists of the model reading two .yaml files
with the model parameters (fabm.yaml and brom.yaml), as well as a NetCDF or
text file with the hydrophysical forcing data. Optionally, the biogeochemical
initial conditions can be read from a text file The brom.yaml (see Sect. S1) file specifies the values
of transport model parameters as well as various option switches and
input/output file and variable names. Text comments provide guidance and
references for setting parameter values. If using NetCDF input, the user
should pay careful attention to the NetCDF input parameters and names,
ensuring that this information is consistent with the input NetCDF file. The
selected-year parameter year must refer to a year that is covered by the
input forcing data. The fabm.yaml (see Sect. S2) file specifies the values
of biogeochemical model parameters, default initial values for state
variables, and the coupling of FABM modules. Text comments provide
annotation and references. The nns_annual.nc (in the example) file contains input
forcing data that may be derived from observations or hydrodynamical model
output (GOTM in our demonstration). It can be replaced by a text (.dat) file
if this is the format of the hydrodynamical model output. The start.dat is the text file with initial values for model state
variables at every depth. This file may be created by renaming the output of
a previous simulation (finish.dat is the state on 1 January of the
last modeled year). Output files are NetCDF and headed text files
generated automatically by the model during the simulation. Output files can
be readily imported into various software packages for visualization and
further analysis. Certain output files (Vertical_grid.dat and
Hydrophysics.dat) are generated early in the simulation and
should be checked by the user to ensure that the model grid and hydrophysical
forcings are set up as intended. Vertical_grid.dat is the text file with model layer indices,
midpoint depths, increments between midpoint depths, and thicknesses. Hydrophysics.dat is the text file with daily profiles of
hydrophysical variables (temperature, salinity, diffusivity, porosity,
tortuosity, burial velocities). The finish.dat is the text file with the state variables for the
1 January of the last modeled year. It can be used for visualization
or as initial conditions for further calculations. The output_NNday.dat is the optional text file with the state
variables and diagnostic variables for day NN to make plots of vertical
distributions (e.g., Fig. 3) BROM_out.nc is the NetCDF file with daily profiles of state
variables, rates of biogeochemical transformations, and vertical fluxes. For visualization of NetCDF output files, any software with
NetCDF input can be used. In the example, we used PyNcView for 2-D and
BROM_pictures for 1-D (available at
The conservation equations for liquid and total solid volume fractions in the sediments can be written as
Development of the model code was made by E. V. Yakushev, E. A. Protsenko, J. Bruggeman, P. Wallhead, and S. Yakubov; analyses of the model results and discussions were conducted by R. G. J. Bellerby, R.-M. Couture, and S. V. Pakhomova; and all authors contributed to the writing of the manuscript.
We wish to thank J. Middelburg, O. P. Savchuk, and G. Munhoven for detailed and constructive reviews that led to a greatly improved manuscript. We acknowledge funding from the EC 7th Framework Program (FP7/2007-2013) under grant agreement no. 265847 (“Sub-seabed CO2 Storage: Impact on Marine Ecosystems”, ECO2) and 240837 (“Research into Impacts and Safety in CO2 Storage”, RISCS), EC Horizon 2020 under grant agreement no. 654462 (“STrategies for Environmental Monitoring of Marine Carbon Capture and Storage”, STEMM-CCS), with additional development funds from FME SUCCESS; CO2Base; EEA CO2MARINE; Norwegian Research Council projects no. 236658 (“New knowledge on sea deposits”, NYKOS), no. 535640 (“Combined effects of multiple organic stressors from jellyfish blooms and aquaculture operations on seafloor ecosystems”, JELLYFARM), and no. 254777 (Environmental impacts of leakage from sub-seabed CO2 storage, Trykk CO2); the Research Council of Norway through its Centers of Excellence funding scheme, project number 223268/F50 (CERAD), contract no. 208279 (NIVA Strategic Institute Initiative “Climate effects from Mountains to Fjords”); NIVA OASIS; and VISTA – a basic research program and collaborative partnership between the Norwegian Academy of Science and Letters and Statoil, project no. 6164. The work of J. Bruggeman was funded by NERC National Capability in Marine Modeling. R.-M. Couture acknowledges funding by RCN project no. 244558. Edited by: D. Roche Reviewed by: O. P. Savchuk, G. Munhoven, and J. Middelburg