Even though the effects of benthic fauna on aquatic biogeochemistry have been long recognized, few studies have addressed the combined effects of animal bioturbation and metabolism on ecosystem–level carbon and nutrient dynamics. Here we merge a model of benthic fauna (BMM) into a physical–biogeochemical ecosystem model (BALTSEM) to study the long-term and large-scale effects of benthic fauna on nutrient and carbon cycling in the Baltic Sea. We include both the direct effects of faunal growth and metabolism and the indirect effects of its bioturbating activities on biogeochemical fluxes of and transformations between organic and inorganic forms of carbon (C), nitrogen (N), phosphorus (P) and oxygen (O). Analyses of simulation results from the Baltic Proper and Gulf of Riga indicate that benthic fauna makes up a small portion of seafloor active organic stocks (on average 1 %–4 % in 2000–2020) but contributes considerably to benthic–pelagic fluxes of inorganic C (23 %–31 %), N (42 %–51 %) and P (25 %–34 %) through its metabolism. Results also suggest that the relative contribution of fauna to the mineralization of sediment organic matter increases with increasing nutrient loads. Further, through enhanced sediment oxygenation, bioturbation decreases benthic denitrification and increases P retention, the latter having far-reaching consequences throughout the ecosystem. Reduced benthic–pelagic P fluxes lead to a reduction in N fixation and primary production, lower organic matter sedimentation fluxes, and thereby generally lower benthic stocks and fluxes of C, N and P. This chain of effects through the ecosystem overrides the local effects of faunal respiration, excretion and bioturbation. Due to large uncertainties related to the parameterization of benthic processes, we consider this modelling study a first step towards disentangling the complex ecosystem-scale effects of benthic fauna on biogeochemical cycling.
Coastal ecosystems are highly productive, consist of diverse biological communities and carry out important functions including those supporting a growing world population (Costanza et al., 1997, 2014). However, they are facing multiple anthropogenic pressures such as nutrient loading and climate change (Cloern et al., 2016; Halpern et al., 2008). Elucidating the mechanisms of the coupled biogeochemical cycling of carbon (C), nitrogen (N) and phosphorous (P) in these systems is important to understand how they respond to current and future pressures but also because they contribute to the regulation of global climate and nutrient cycles by processing anthropogenic emissions from land before they reach the ocean (Ramesh et al., 2015; Regnier et al., 2013a, b; Seitzinger, 1988).
In contrast to the deep open ocean, benthic–pelagic coupling plays a large role in biogeochemical cycling in coastal and estuarine ecosystems (Soetaert and Middelburg, 2009). Coastal sediments act as hotspots for organic matter degradation and permanent removal of elements from biological cycling through burial and denitrification (Asmala et al., 2017; Regnier et al., 2013a; Seitzinger, 1988). The bioturbating activities of benthic fauna alter the physical and chemical properties of surface sediments, which in turn strongly influence organic matter degradation processes and benthic–pelagic biogeochemical fluxes (Aller, 1982; Rhoads, 1974; Stief, 2013). Here, we define bioturbation as all biological processes that affect the sediment matrix, including burrow ventilation (bio-irrigation) and reworking of particles (Kristensen et al., 2012). Additionally, benthic fauna retains carbon and nutrients in its biomass and transforms them between organic and inorganic forms through metabolic processes (Ehrnsten et al., 2020b, and references therein; Herman et al., 1999; Josefson and Rasmussen, 2000). Together, these direct and indirect effects of benthic fauna have far-reaching consequences for ecosystem functioning in the benthic and pelagic realms (Griffiths et al., 2017; Lohrer et al., 2004).
Even though the importance of benthic fauna for sediment biogeochemistry and benthic–pelagic fluxes has long been recognized (Rhoads, 1974), the combined effects of animal bioturbation and metabolism have seldom been studied together (Ehrnsten et al., 2020b; Middelburg, 2018; Snelgrove et al., 2018). A long-standing assumption in biogeochemical sediment research is that animals contribute considerably to transport of solids and solutes through bioturbation, but their consumption of organic matter is of minor importance (Middelburg, 2018). However, several studies show that this assumption does not hold in many shallow coastal systems, as recently reviewed by Middelburg (2018) and Ehrnsten et al. (2020b).
Further, empirical studies of faunal effects often focus on temporally and spatially limited parts of the system, omitting important interactions and variability occurring in natural ecosystems (Snelgrove et al., 2014). It is logistically challenging to study multiple drivers and interactions in the benthic and pelagic realms, such as the interactions between benthic and pelagic production, empirically. Mechanistic or process-based models are powerful tools to conduct such studies (Seidl, 2017). Here, we extend a physical–biogeochemical model of the Baltic Sea ecosystem (BALTSEM; Gustafsson et al., 2014; Savchuk et al., 2012) with benthic fauna components based on the Benthic Macrofauna Model (BMM; Ehrnsten et al., 2020a). We include both the direct feedbacks from animal growth and metabolism and the indirect effects of their bioturbating activities on biogeochemical cycling to evaluate their relative contributions.
We use the Baltic Sea as a model area for three reasons: (i) the shallow depth (mean depth 57 m) and enclosed geography with a long water residence time (about 33 years) contribute to strong benthic–pelagic coupling (Snoeijs-Leijonmalm et al., 2017; Stigebrandt and Gustafsson, 2003), (ii) the relatively simple, species-poor benthic communities facilitate model development, and (iii) the major features of biogeochemical cycling of C, N and P in the Baltic Sea are well known due to a wealth of oceanographic measurements and studies performed over the past century, making it an ideal system for process-based modelling (Eilola et al., 2011; Gustafsson et al., 2017; Savchuk and Wulff, 2009, 2001). However, the sediment pools and the role of sediment processes in benthic–pelagic exchange are not as well quantified as pelagic pools and fluxes. The higher uncertainty in benthic compared to pelagic processes and the traditional focus on pelagic eutrophication are probable reasons why physical–biogeochemical models of the Baltic Sea have omitted benthic fauna as state variables (e.g. Eilola et al., 2011; Lessin et al., 2018). Here, we aim to fill this knowledge gap and explore the role of benthic fauna in biogeochemical cycling of C, N and P on a long-term ecosystem-level scale.
The Baltic Sea is a semi-enclosed coastal sea in northern Europe with strong latitudinal and depth gradients in salinity, temperature and productivity shaping the distribution of species and ecosystem functioning (Bonsdorff, 2006; Elmgren, 1984; Snoeijs-Leijonmalm et al., 2017). The diversity of benthic fauna is low due to the low salinity, and large, deep-burrowing species are only found near the entrance to the Baltic Sea (Bonsdorff, 2006; Remane, 1934). Thus, the sediment layer mixed by bioturbating animals is very shallow compared to other coastal and shelf seas (Teal et al., 2008). Nonetheless, several studies have measured significant effects of benthic fauna on benthic nutrient processing in the Baltic Sea (e.g. Berezina et al., 2019; Lehtonen, 1995; Norkko et al., 2013, 2015).
Due to its large catchment area and limited water exchange with the North Sea, the Baltic Sea is heavily influenced by anthropogenic nutrient emissions (Andersen et al., 2017; Gustafsson et al., 2012). Although emissions have been significantly reduced since the peak in the 1980s, recovery from eutrophication is slow with limited reductions in nutrient pools and primary productivity seen to date (Gustafsson et al., 2012; Savchuk, 2018; Zdun et al., 2021). This is due to the long water residence time and the build-up of nutrient stores in soils and marine sediments during several decades (McCrackin et al., 2018; Savchuk, 2018, and references therein).
In this study, we focus on comparing results from the Baltic Proper and the Gulf of Riga (Fig. 1), two basins with a similar benthic community composition but differing in physical and biogeochemical properties such as depth, openness, productivity and bottom oxygen conditions. We expect these differences to be reflected in the strength of benthic–pelagic coupling processes and the role of benthic fauna therein.
The Baltic Sea hypsography and basin divisions in the BALTSEM model. This study focusses on the Baltic Proper (basin 9) and Gulf of Riga (basin 12), outlined in blue.
The Baltic Proper is the central, deepest basin of the Baltic Sea with a maximum depth of 459 m and a mean depth of ca. 75 m. A permanent halocline at ca. 60–80 m limits the vertical mixing between the low-salinity surface waters (5–8 psu) and the deeper waters with a salinity of 9–13 (Snoeijs-Leijonmalm et al., 2017). A majority of the waters below the halocline are hypoxic or anoxic because the mineralization of organic matter sinking through the water column and in the sediments consumes oxygen faster than it is replenished by infrequent salt water intrusions and vertical turbulent mixing. The expanding hypoxia has severely reduced the area habitable by benthic fauna in the Baltic Sea (Carstensen et al., 2014a, b). In the reducing environment, P bound to iron-humic complexes is released from sediments and contributes to the dissolved inorganic P (DIP) pool in the water column. The excess DIP promotes the fixation of atmospheric N by cyanobacteria, in turn promoting primary production by other phytoplankton, which leads to increased sinking and mineralization of organic matter, in turn expanding hypoxia. This feedback loop, termed the “vicious circle” (Vahtera et al., 2007), is further strengthened by climate change as increasing water temperatures promote cyanobacterial blooms (Kahru et al., 2020; Kahru and Elmgren, 2014).
The Gulf of Riga is a semi-enclosed coastal bay with mean and maximum depths of 23 and 51 m, respectively, and a salinity of 4–7 (Snoeijs-Leijonmalm et al., 2017). In contrast to the Baltic Proper, the Gulf of Riga is relatively well mixed, and hypoxia only occurs intermittently under the summer thermocline (Kotta et al., 2008), accompanied by increased release of phosphate from the sediments (Eglite et al., 2014). When occurring more often in the most recent decade, the intensity and extent of both sporadic hypoxia and phosphate release have somewhat increased (HELCOM, 2018; Stoicescu et al., 2022). Similarly, the summer cyanobacteria blooms sporadically occurring in the Gulf of Riga before the 2010s (Kahru and Elmgren, 2014) have regularly and extensively covered the gulf since 2015 (Mati Kahru, personal communication, 2021).
The biogeochemical cycling in the BALTSEM model was extended to include benthic fauna. BALTSEM simulates physical circulation and biogeochemical transformations of C, N, P, O and Si in the Baltic Sea in response to climatic conditions and nutrient inputs from rivers, point sources and atmospheric deposition. It describes the Baltic Sea as 13 horizontally homogenous boxes with a dynamic depth resolution of generally less than 1 m in the pelagial (Fig. 1). Sediments are represented as terraces at 1 m depth intervals with an area corresponding to the hypsography of each basin. The new benthic components were constructed from the carbon-based Benthic Macrofauna Model (BMM) described in Ehrnsten et al. (2019a, b, 2020a), extended to include nitrogen (N) and phosphorus (P) components.
Below we give a short description of the benthic dynamics in the new model version, referred to as BALTSEM–BMM, with a focus on the effects of benthic fauna on biogeochemical processes (Fig. 2). A full mathematical description of the benthic biogeochemical processes of the model is found in Appendix A. For a description of the pelagic biogeochemistry and physics, we refer the reader to Gustafsson et al. (2012, 2014) and Savchuk et al. (2012). Additionally, all benthic and pelagic state variables are listed in Table A1.
BALTSEM–BMM includes C, N and P contents in biomass of three functional
groups of benthic fauna. The facultative deposit- and suspension-feeding
bivalve
The biomass of all groups of fauna are modelled as a dynamic mass balance
between fluxes formed by food uptake, assimilation, respiration or
excretion, and mortality. The formulations for dynamics of the functional
groups and their food banks were kept as in BMM
(Ehrnsten et al., 2019a, 2020a) as far as
possible. The main change is the addition of N and P components to each
state variable. Consumers generally regulate their inner stoichiometry
within tight limits (Sterner and Elser,
2002); therefore the fauna was given a constant
Food uptake is modelled as a function of food availability in C units. The
uptake of N and P components of a food source is thereafter calculated
proportionally to the
Schematic overview of benthic model processes shared by benthic C,
N and P components
Respiration and excretion of inorganic C, N and P is divided into three
parts: (1) a basal maintenance part related to biomass; (2) a growth and
activity part related to food uptake as a proxy for activity; and (3) excess
excretion. As the stoichiometry of assimilated food varies, excretion of
excess elements is calculated dynamically to keep the fixed stoichiometry of
the benthos. Formulations are similar to those used for zooplankton in
BALTSEM and for benthos in other ecosystem models
(Ebenhöh
et al., 1995; Spillman et al., 2008). Respiration and excretion fluxes add
to the bottom water pools of dissolved inorganic carbon (DIC), nitrogen (NH,
representing total ammonia) and phosphorus (PO, representing total
phosphate). Respiration also consumes bottom water oxygen with a
respiratory quotient of 1 mol O
As in the standard BALTSEM, sediment bioavailable C, N, P and Si are represented as vertically integrated concentrations in the biogeochemically active surface layer of unspecified thickness. The concentrations are modelled as a dynamic mass balance between fluxes formed by sedimentation, mineralization and burial, extended by interactions with benthic fauna. Sediment C, N and P pools are further divided into three banks of different ages to resolve the food limitation of benthic fauna (Fig. 2), while benthic Si is represented as a single pool that does not interact with the fauna. Oxygen is not a state variable in sediments, but several benthic processes interact with simulated bottom water oxygen.
Bioturbation by benthic fauna, including sediment reworking and burrow
ventilation, generally increases the oxygenation of sediments
(Michaud et
al., 2005; Volkenborn et al., 2012), promoting the binding of phosphate to
iron oxides (P sequestration) and stimulating nitrogen oxidation
(Ekeroth
et al., 2016; Norkko et al., 2012; Renz and Forster, 2014). Similar to
Isaev et al. (2017), we use simple
formulations for the effects of faunal activities on the oxygen-dependent
processes of sediment nitrification and denitrification and P sequestration
through a bioturbation enhancement factor
Within each of the three sediment banks, C, N and P components share the
same source and sink processes. Sinking organic matter is integrated into a
bank of fresh organic matter available as food for deposit-feeders. This
bank ages into a slightly older bank available as food for
For each element, mineralization fluxes from the three sediment banks are
combined into a total flux (
Depending on oxygen concentrations in the bottom water layer and
bioturbation intensity, mineralized sediment N is released to the water
column as ammonia (
The model was run over 1970–2020 forced with observed nutrient loads and
actual weather conditions as described in Gustafsson et al. (2012, 2017) with forcing
time series extended to 2020. The physical circulation was forced by
3-hourly meteorological conditions and monthly time series of river run-off
and state variable concentrations and sea level at the North Sea boundary.
Monthly inputs of N, P, C and Si from land via rivers and from coastal point
sources, as well as atmospheric deposition of N, P and C, were used as
biogeochemical forcing. Initial conditions in 1970 were based on
observations for pelagic variables and hindcast simulations for benthic
variables as described in Gustafsson et al. (2012) and
Ehrnsten et al. (2020a). Briefly, the benthic fauna and their
food banks (SED1
Results of this default simulation were validated against observations of salinity, temperature, oxygen, ammonium, nitrate and phosphate (Appendix B), as well as biomasses of benthic fauna (Appendix C).
It is difficult to constrain the new parameters and to validate the
system-scale dynamics against observations from the field or laboratory,
which are usually made on much smaller temporal and spatial scales. Instead,
we made a sensitivity analysis testing the effects of changing the parameter
Additionally, we estimated the contribution of bioturbation to
benthic–pelagic nutrient fluxes by calculating the theoretical fluxes
without bioturbation enhancement (i.e. with
Finally, to study the relationship between nutrient loads and the role of
benthic fauna in biogeochemical cycling, we ran two future scenarios for
2021–2100 with either decreasing loads of N and P according to the Baltic
Sea Action Plan (BSAP scenario, total loads to the Baltic Sea of 739 kt N yr
As the purpose of this study was to evaluate large-scale dynamics, results were aggregated as means and standard deviations of the last two decades (2000–2020 or 2080–2100) to capture differences in long-term averages while accounting for inter-annual variations.
The BALTSEM–BMM behaves very similarly to the standard BALTSEM model, which has been extensively validated (Gustafsson et al., 2012, 2014; Savchuk et al., 2012) and shown to perform favourably in relation to similar Baltic Sea models (Eilola et al., 2011; Meier et al., 2018). A comparison of the main pelagic state variables (salinity, temperature and concentrations of oxygen, NH, NO and PO) to observations over time (1970–2015) and depth shows an overall relative bias of 1.40, while the relative bias of the standard BALTSEM is 1.41. The relative bias index compares model–data difference with variability in the data, giving an estimate of how well the model captures variability in nature on seasonal, annual and decadal scales (Savchuk et al., 2012). A detailed description and results of this analysis are found in Appendix B.
Ehrnsten et al. (2020) did a comprehensive validation of simulated biomasses
of benthic fauna against observations over depth intervals in the largest
basins of the Baltic Sea. We re-ran this analysis with the results of the
coupled model and extended it to include the entire Baltic Sea. The extended
analysis, based on 7774 observations, confirms previous results that the
model captures the main observed patterns of biomass over latitude and depth
with reasonable accuracy (Fig. 3, Appendix C), defined as cost function
values
Comparison of simulated total biomasses of benthic fauna to
observations at four depth intervals in six basins from south to north.
Observations are shown as both the sum of the three functional groups –
In addition, we compared the simulated biomasses of benthic fauna in the
Gulf of Riga to estimates from the literature (Table 1). The simulated
biomass of benthic fauna in the gulf varied substantially over depth and
time (29–284 g wwt m
Comparison of simulated benthic fauna biomass in the Gulf of Riga
and estimates based on field sampling (g wwt m
Carman and Cederwall (2001) have estimated the
amounts of C, N and P in Baltic Sea sediments based on core samples. It is
not straightforward to compare the total amounts to simulations as the
thickness of the simulated sediment is not defined. However, the estimated
Long-term (2000–2020) average benthic budgets of C, N and P are shown in
Fig. 4 for the Baltic Proper and Gulf of Riga. The results for the Baltic
Proper are restricted to the depth interval 0–90 m as benthic fauna is
practically absent in the oxygen-poor waters below this depth (Fig. 5a).
Figure 5 also shows the long-term average depth distribution of the
bioturbation factor
Average (2000–2020) benthic fluxes (g m
According to the default simulation, the benthic fauna made up a minor part of the benthic organic C, N and P stocks (1 %–4 %) but had a proportionally larger share in benthic–pelagic fluxes of DIC (23 % and 31 % in the Baltic Proper and Gulf of Riga, respectively, Fig. 4a–b), DIN (43 % and 51 %, Fig. 4c–d) and DIP (25 % and 34 %, Fig. 4e–f). The budgets also show that input of organic matter to the sediments was higher in the Gulf of Riga compared to the Baltic Proper, resulting in overall higher benthic stocks and benthic–pelagic fluxes (Figs. 4 and 5).
Depth distribution of benthic fauna biomass and the bioturbation
coefficient
When accounting only for the immediate local effects of bioturbation, it
increased NO outflux by 0.41 g N m
Immediate local effects of bioturbation on benthic fluxes. Benthic
fluxes directly affected by bioturbation in the default run with
bioturbation and when calculated for each time step without bioturbation.
Averages for 2000–2020
When also accounting for the effects of bioturbation throughout the
ecosystem in the sensitivity analysis, changes were more complex (Figs. 7–8). For example, comparing the default run (
Sensitivity analysis of effects of three levels of bioturbation on
benthic fluxes of carbon
Sensitivity analysis of effects of three levels of bioturbation on
total primary production
In general, increasing bioturbation led to a decrease in most benthic stocks
(Fig. 9) and fluxes (Fig. 7). This can be explained by the following chain of
effects. Increased P sequestration in the sediments (Fig. 7e–f) led to less
pelagic DIP available for phytoplankton, especially cyanobacteria, growth
and thereby lower N fixation and primary production (Fig. 8a–b) which in
turn led to lower organic matter sedimentation rates (Fig. 7a–f), lower
sediment stocks of organic matter (Fig. 9b–c), and consequently lower rates
of most sediment biogeochemical transformations and fluxes (Fig. 7).
Decreasing organic matter sedimentation also led to decreased biomass of
benthic fauna (primarily due to a reduction in
Sensitivity analysis of effects of three levels of bioturbation on
stocks of benthic fauna and sediment C, N and P. Averages for 2000–2020
Bioturbation also improved oxygen conditions (Fig. 8c–d). In the Baltic
Proper the effect on hypoxic areas was marginal, but the extent of anoxic
areas was reduced by ca. 8000 km
All results below are calculated from means of 2080–2100 for the BSAP and HIGH nutrient load scenarios and 2000–2020 for the default model run in the Baltic Proper (0–90 m depth) and Gulf of Riga.
With increasing nutrient loads, primary production and input of particulate organic matter (POM) to the sediments increased (Fig. 10a–e), resulting in an increase in most benthic stocks and fluxes. The biomass of benthic fauna responded more strongly to changing nutrient loads than the bioturbation enhancement coefficient linked to the feeding activities of fauna (Fig. 10f–g).
Primary production
With changing loads, the relative roles of faunal and microbial processes in the sediment changed (Fig. 11). With increasing loads, an increasing proportion of benthic–pelagic fluxes of inorganic nutrients originated from faunal metabolism. Expressed as percent of POM input to the sediments, the respiration and excretion of fauna were 12 %–13 % in the BSAP scenario and 26 %–27% in the HIGH scenario in the Baltic Proper (Fig. 11a, c, e). In the Gulf of Riga, respiration and excretion were 23 %–24 % of POM input in the BSAP scenario and 35 %–37 % in the HIGH scenario (Fig. 11b, d, e). Correspondingly, the proportions of POM input released as dissolved inorganic substances resulting from microbial processes in the sediment (DIC, NO and PO outflux in Fig. 11) were higher in the BSAP than in the HIGH scenario. An exception is the NH outflux in the Baltic Proper (Fig. 11c) that increased with increasing loads due to an expansion of anoxic bottoms.
Apportionment of benthic fluxes of carbon
The relative proportion of particulate organic phosphorus (POP) input sequestered shows a complex pattern in the Gulf of Riga (Fig. 11f): the proportion of POP input sequestered was lower in the BSAP scenario compared to the default model run due to less fauna and thereby less bioturbation. However, the proportion was also lower in the HIGH load scenario as the increased occurrence of hypoxia (Fig. 10h) counteracted the effects of increased bioturbation. In the Baltic Proper, relative P sequestration shows a decreasing pattern with increasing loads (Fig. 11e) driven by increasing occurrence of hypoxia (Fig. 10f).
Even though the total amount of POM input to the sediment increased with increasing nutrient loads, it constituted a decreasing proportion of primary production. In the BSAP scenario, almost half of the annual primary production reached the seafloor (48 % and 47 % in the Gulf of Riga and Baltic Proper, respectively) compared to 21 % and 27 % in the HIGH load scenario (Fig. 10a, c). Thus, the proportion of primary production mineralized by fauna varied only slightly with nutrient load scenario because of the opposite responses of sinking organic matter and fauna: 4.7 % (BSAP) to 4.1 % (HIGH) in the Baltic Proper and 10.7 % (BSAP) to 9.6 % (HIGH) in the Gulf of Riga.
We have created a new tool to simulate the long-term and large-scale effects of benthic fauna on biogeochemical cycling in the Baltic Sea by fully merging two existing process-based models. First simulations with the new model indicate that the benthic fauna makes up a small part of benthic organic stocks but contributes substantially to organic matter mineralization and benthic–pelagic fluxes of inorganic C, N and P through its metabolism. Further, the stimulation of P binding in sediments by bioturbation significantly reduced N fixation and primary production in the simulations, indicating that benthic fauna can alleviate the “vicious circle” of eutrophication.
In general, the BALTSEM-BMM model reproduces the observed Baltic-Sea-scale
patterns of decreasing biomass of benthic fauna with latitude and depth
reasonably well, as also shown for a previous one-way coupled model version
(Ehrnsten et al., 2020a). Compared to observations, the
model seems to underestimate the biomass of benthic fauna in the Bothnian
Sea and overestimate it in the Gulf of Finland (Figs. 3, C2). The former may
be due to an underestimation of primary productivity in the Bothnian Sea by
BALTSEM, while the omission of possible negative effects of low salinity on
The simulated mean biomass of benthic fauna in the Gulf of Riga was considerably
higher than estimated by Gogina et al. (2016) (Table 1, Fig. 3). Possible
reasons for overestimation may be that the model does not take into account
the limitations by mobile substrates and low salinity, especially in the
southern part of the basin
(Carman
et al., 1996; Kotta et al., 2008). In this region, a reduction in benthic
biomass (e.g. of
The modelled patterns in total benthic biomass are strongly driven by
changes in
The BALTSEM model was neither improved nor worsened by the addition of benthic fauna, according to the performance analyses comparing pelagic nutrient and oxygen concentrations to observations (Appendix B). This shows that increasing model complexity does not necessarily increase accuracy, especially when the functions and/or variables added are not well known (Ehrnsten et al., 2020b; Levins, 1966). In general, though, the previous assessments of model performance showing that the model is able to reproduce seasonal and decadal variations in biogeochemical variables and performs well in comparison to other Baltic Sea models remain valid (Eilola et al., 2011; Gustafsson et al., 2012, 2014; Meier et al., 2018; Savchuk et al., 2012).
Similar to estimates made with previous uncoupled versions of the model
(Ehrnsten et al., 2019a, 2020a), the
results of this study suggest that respiration by fauna constitutes a
significant part of organic matter mineralization in sediments (Fig. 4). The
fauna mineralized about 8–17 g C m
The sensitivity analysis showed a large effect of bioturbation on primary
production levels mainly due to increased P retention (Figs. 7–9). When
bioturbation increased P sequestration (Fig. 7e), this led to a weakening of
the “vicious circle” in the Baltic Proper where less DIP in the water column
led to less N fixation and organic matter production (Fig. 8a, b), which in
turn led to less organic matter input to sediments, less heterotrophic
oxygen consumption, and less hypoxia and anoxia (Fig. 8c, d) and thereby further
increased P sequestration in the oxygenated sediments. Also in the Gulf of
Riga, where hypoxia was rare (Fig. 8c, d), the bioturbation-induced
reduction in pelagic DIP had large effects on primary production and
especially N fixation (Fig. 8a, b). In the two runs with bioturbation there
was no or very little pelagic DIP surplus available for the N-fixing
phytoplankton group in contrast to the run without bioturbation, in which N
fixation added on average 0.97 g N m
The significant effect of bioturbation on P retention found here is in line
with the results of studies on the effects of the invasive polychaete
The effects of bioturbation on sediment N dynamics were less important for
eutrophication processes than the effects on P dynamics in this study. We
assumed a very simple process formulation, in which bioturbation increases
oxygen penetration depth in the sediments leading to a larger proportion of
organic N mineralization in oxic environments, thus promoting outflux of
nitrates over benthic denitrification. In reality, denitrification is a
complex process depending on, for example, the 3D structure of redoxclines in the
sediment. If the biogenic structures of tube-dwelling bio-irrigators
increase the area of the oxic–anoxic interface in the sediment, this can lead
to the opposite effect in which a larger proportion of nitrate is denitrified
at the enlarged redoxcline
(Aller, 1988; Gilbert et al.,
2003). However, we believe this to be a special case unlikely to dominate in
the Baltic Sea. Henriksen et al. (1983) measured an increased
proportion of nitrate denitrified in sediments with large burrows of animals
with low irrigation activity (e.g.
To better capture alterations in redoxclines, a depth-resolved sediment model with oxygen as a state variable would be needed. We also recognize that many other possible effects of bioturbation, e.g. on burial (Josefson et al., 2002) and resuspension (Cozzoli et al., 2021), were not included. However, there is always a trade-off between model complexity and generality, with few models to date combining a depth-resolved sediment module with a full pelagic model (Ehrnsten et al., 2020b; Lessin et al., 2018). One of the main advantages of the BALTSEM model is that its simplicity and fast running time promote the development of additional features and experimentation with a large number of simulations (e.g. Gustafsson et al., 2015; Soerensen et al., 2016; Undeman et al., 2015).
Experimental studies from the Baltic Sea report a range of positive,
negative or negligible effects of benthic fauna on benthic–pelagic DIN
fluxes and denitrification rates, showing that these processes are highly
context-dependent (Griffiths et al., 2017, and
references therein). Studies with
The Gulf of Riga had higher simulated benthic stocks and fluxes than the
Baltic Proper. This can partly be attributed to the slightly higher primary
production (171
Besides depth, the amount of organic matter export also depends on, for example, the
type of plankton and temperature (Tamelander et al., 2017).
Despite large inter-annual variations, there was a clear decreasing trend in
the proportion of primary production exported to the seafloor of about 5
percentage points per decade in the Baltic Proper during 1970–2020 (
The proportion of primary production arriving at the seafloor also varied with nutrient loads, constituting almost half of the annual primary production in the reduced-load BSAP scenario (47 %–48 %) compared to 21 %–27% in the HIGH load scenario (Fig. 10a, c). Simultaneously, the amount of benthic fauna decreased and mineralization of sediment organic matter became more dominated by microbial processes with reduced loads. Thus, we can conclude that both the absolute amount and the relative proportion of POM input mineralized by macrofauna increased with nutrient loads, but in relation to primary production the proportion mineralized by fauna was almost independent of changes in loads because of the opposite responses of sinking organic matter and fauna.
Using a newly developed modelling tool, significant effects of benthic macrofauna on C, N and P cycling were simulated in the semi-enclosed brackish-water Baltic Sea, with impacts on the ecosystem from the extent of hypoxic bottoms to the rates of pelagic nitrogen fixation and primary production. Our results suggest that in addition to bioturbation, relatively more studied in the modelling context, the metabolism of benthic fauna should be given more attention in future studies as it may play a significant role in benthic mineralization of organic C, N and P in coastal seas and estuaries.
The magnitude of effects of benthic fauna on biogeochemistry generally decreased with depth and increased with productivity, as shown by the comparison of two basins and different nutrient load scenarios. Thus, these simulations confirm the notion that benthic–pelagic coupling is strongest in shallow coastal areas (Griffiths et al., 2017; Nixon, 1981), but they also show that this relationship is modified by multiple interacting physical and biological drivers, which may change over time. For example, we found that the proportion of primary production reaching the seafloor decreased with increased nutrient loading and increased temperature as both led to an intensification of pelagic nutrient cycling. Further, simultaneous positive and negative feedback loops led to complex relationships, e.g. between productivity and P cycling (as seen in Fig. 11f). On the one hand, increased productivity can increase the amount of bioturbating fauna, stimulating P sequestration. On the other hand, increased productivity can increase benthic oxygen consumption for the mineralization of sinking organic matter, leading to deteriorating oxygen conditions and increasing P leakage from sediments. Unravelling the many interacting drivers and responses on a system scale is important to understand how coastal and global biogeochemical cycles are responding to changes in, for example, nutrient loads and climate.
Even though these large-scale simulations contain a large degree of uncertainty, they are an important complement to empirical studies, which for practical reasons can only consider temporally and spatially limited parts of the system (Boyd et al., 2018; Snelgrove et al., 2014). To improve the confidence in simulation results, we see two major ways forward. First, as all models contain different formulations, assumptions and uncertainties, implementing benthic fauna components in other physical–biogeochemical models and comparing the results would greatly increase the strength of evidence for those results where different models agree. This kind of ensemble modelling is increasingly used in climate change research and has also been applied in the context of Baltic Sea biogeochemistry (Meier et al., 2012, 2018; Murphy et al., 2004). We hope that the publication of the benthic model formulations stimulates the development of benthic fauna modules in other models of the Baltic Sea ecosystem and beyond. Even though the current model implementation is only applicable to the brackish parts of the Baltic Sea due to a lack of functional groups present in the marine parts, the inclusion of additional functional groups using the existing groups as a template would be straightforward technically. The main challenges are the parameterization of group-specific rates and managing the increased complexity.
Second, a comprehensive compilation of observational data on sediment stocks and fluxes would be needed for improved model validation. Such data are collected for monitoring and research purposes by a great number of institutions around the Baltic Sea, but a comprehensive, open-access, quality-controlled collection of these data is lacking. The Baltic Environment Database (BED) has been invaluable for both model development and validation of pelagic physics and chemistry. While these data can be used as indirect validation of benthic model processes in the strongly coupled system, we call for the development of a “Benthic BED” to facilitate future model development. A comprehensive collection of observational data would also facilitate the identification of knowledge gaps and future research priorities.
All state variables are listed in Table A1 and benthic parameters in Tables A2–A4. A graphical overview of benthic state variables and processes is
provided in the main paper (Fig. 2). Bioavailable surface sediments are
represented as terraces at 1 m depth intervals with an area corresponding to
the hypsography of each basin. All benthic state variables are calculated in milligrams per square metre (mg m
State variables in BALTSEM-BMM.
Sediment concentrations of elements
Mineralization (Eq. A5) and ageing (Eq. A6) of element
A proportion
Benthic parameters in the BALTSEM-BMM model taken from the standard BALTSEM (Gustafsson et al., 2014). Recalibrated parameter values are indicated in bold. Subscript indices C, N, P and Si refer to the respective elements. Where different elements share the same parameter value, the indices are listed in a row.
For each element
Depending on oxygen concentrations in the bottom water layer (OXY),
mineralized sediment N is released to the water column as ammonia
(
In anoxic or nearly anoxic conditions, the mineralized N is released to the
water column as NH, defined by the fraction
Subsequently, the oxidized fraction can be denitrified into N
A fraction
Benthic parameters in the BALTSEM-BMM model taken from the carbon-based BMM (Ehrnsten et al., 2020a). Parameters are applied equally to C, N and P components.
The fraction sequestered is positively related to oxygen concentrations
(enhanced by bioturbation) and negatively related to salinity according to
Eq. (A16), where
Sediment C mineralization and nitrification consume oxygen, while
denitrification causes reimbursement of O
New parameters in the BALTSEM-BMM model.
The model includes three functional groups of benthic fauna –
Since the model does not include recruitment or migration, biomass change is
set to 0 when biomass falls below 0.01 mg C m
The change in N and P components of a group BF
Food uptake of element
Predators feed on deposit-feeders and
Deposit-feeders are restricted to feeding on freshly deposited organic
matter in SED1
Additionally, feeding stops at anoxia for all groups.
The ingestion rate of food source component
Ingested food is divided into assimilated uptake (AU
Respiratory release of DIC (
Excretion of N and P (
To calculate excess respiration or excretion, first the limiting element for
growth (lim
For the limiting element, excess respiration or excretion is 0. The other two
elements are then released to restore the stoichiometry of the fauna. Thus,
total respiration and excretion of group
Respiration consumes bottom water oxygen according to Eq. (A32).
Mortality is divided into hypoxia-induced mortality and other mortality.
Other mortality rate
The hypoxia-induced mortality rate
Similar to Isaev et al. (2017), we
use simple formulations for the effects of bioturbation on sediment
denitrification (Eq. A13) and P sequestration (Eq. A16) through a
bioturbation enhancement factor
Temperature (
As a measure of model performance, we calculate the relative bias of simulated and observed long-term monthly means for some of the main pelagic state variables as described in Savchuk et al. (2012) but with data extended to 2015. The index compares model–data difference with variability in the data, giving an estimate of how well the model captures variability in nature on seasonal, annual and decadal scales.
Spatial distribution of the relative bias between simulated and
observed dynamics of salinity, temperature and concentrations of oxygen,
total ammonia (NH), nitrate
Observations of salinity, temperature and concentrations of oxygen,
ammonium, nitrate and phosphate were collected from the Baltic Sea
Environment Database (BED) and other major data sources around the Baltic
Sea such as the IOW (Germany), NERI (Denmark), SYKE-FMI (Finland) and SHARK
(SMHI, Sweden) databases. The full list of the data contributors can be
found at
Basin-wide monthly time series were prepared from available long-term observations in the following way. All the measurements found in monthly intervals over 1970–2015 for all frequently sampled water layers within every BALTSEM basin, i.e. usually at 5 m intervals for the top 20 m of the water column and 10–25 m intervals for the deeper parts of basins, were pooled together and averaged. Coastal measurements, defined as being sampled within 12 nmi from the shore, were excluded for all basins except the three Danish straits basins, where the 12 nmi coastal strip covers almost the entire basins. Measurements from several deep and isolated trenches in the northern Baltic Proper were excluded as they often display their own dynamics, asynchronous to that in the domain of the larger basin.
To emphasize both long-term changes and seasonality of variables, time
series of a model–data difference of pairwise monthly means were used.
Because the seasonal cycle is also reflected in monthly standard deviations,
especially in the upper part of the water column these differences were
scaled with month-specific standard deviation SD
The BALTSEM-BMM performs very similarly to the standard BALTSEM. The average relative bias of the analysed variables over all basins is 1.40, which can be compared to a relative bias of 1.41 for the standard BALTSEM model.
The model captures variations in physical parameters (salinity and temperature) and oxygen concentrations with a relative bias of mostly less than 2 (Fig. B1). Simulated ammonium (NH) concentrations are lower than measured in the well-oxygenated Bothnian Sea (basin 10) and upper parts of Bothnian Bay (basin 11) as all ammonium is oxidized to nitrate (NO) in the model under these conditions. Together with an underestimation of NO utilization in intermediate depth layers by phytoplankton, this results in an overestimation of NO concentration in the upper part of several basins, including the Gulf of Riga (basin 12). Variations in PO concentrations are well captured in the Baltic Proper (basin 9) but overestimated in the Bothnian Sea (basin 10) and Bothnian Bay (basin 11). A detailed discussion of performance and sources of errors can be found in Savchuk et al. (2012) and a comparison of BALTSEM to other similar models in Eilola et al. (2011).
The simulated biomasses of benthic fauna were validated against observations using the method of Ehrnsten et al. (2020a) but with data extended to include the southern and southwestern parts of the Baltic Sea (basins 1–8 in Fig. 1), as well as the Gulf of Riga (basin 12). The comparison includes a visual comparison of biomasses over depth intervals in the different basins, as well as a cost function assessment (cf. Eilola et al., 2011).
Number of observations of benthic fauna wet weight used for model
validation per basin and depth interval. Total
Cost functions comparing simulated and observed biomasses of
benthic fauna. ND: no data. NA: not applicable. In the deeper sections of
Kattegat, observed biomasses of
As described in detail in Ehrnsten et al. (2020a), data from
samples of benthic macrofauna biomass taken between 1990 and 2012 from the
national databases of Sweden (
The results of the default simulation were compared to observations using a
cost function (CF),
The model is not applicable to the high-salinity areas at the entrance to the Baltic Sea (Kattegat, Öresund and Arkona Basin) as it does not include the high diversity of functional groups present in these areas (Figs. C1–C2). In the Baltic Sea sensu stricto (basins 7–13), the CF values are mostly good to reasonable (Fig. C1), but a visual comparison shows that the standard deviations of observations are very large in most cases, allowing for large deviations between modelled and observed means (Fig. C2). However, the visual comparison also shows that the main observed trends in decreasing total biomass with increasing latitude and depth are reasonably captured by the model, as well as the order of magnitude of the individual functional groups.
The closest match between simulated and observed means is seen in the Baltic
Proper, which is the largest basin and which contains the major part of
benthic fauna stocks in the Baltic Sea. In the deepest segment of the
Bornholm Basin (70–100 m), observations are predominantly from hypoxic areas
with biomasses of
Comparison of simulated biomasses of benthic fauna to
observations at four depth intervals in BALTSEM basins from south to north.
Observations are shown as both the sum of the three functional groups –
The model code is available upon request from the corresponding author.
No original datasets were produced in this study. Observational data for validation of benthic and pelagic state variables were extracted from open-access databases as detailed in Sects. B1 and C1. Unpublished data on benthic biomass collected by
All authors contributed to the experimental design and interpretation of results. EE and BGG developed the model code. EE performed the experiments, analysed results and wrote the manuscript with significant contributions from BGG and OPS.
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.
We thank Bärbel Müller-Karulis for the development of and help with the implementation of a script for relative bias calculations and Alexander Sokolov for access to environmental validation data through the BED database. Erik Smedberg is acknowledged for contributions to artwork (Fig. 1).
This research has been supported by the Swedish Agency for Marine and Water Management (grant no. 1:11 – Measures for marine and water environment).The article processing charges for this open-access publication were covered by Stockholm University.
This paper was edited by Jack Middelburg and reviewed by Hans Cederwall and two anonymous referees.