Modelling the vertical distribution of bromoform in the upper water column of the tropical Atlantic Ocean

Modelling the vertical distribution of bromoform in the upper water column of the tropical Atlantic Ocean I. Hense and B. Quack Leibniz Institute for Baltic Sea Research, Rostock, Germany Leibniz Institute of Marine Sciences, IFM-GEOMAR, Kiel, Germany Received: 30 September 2008 – Accepted: 17 October 2008 – Published: 16 December 2008 Correspondence to: I. Hense (inga.hense@io-warnemuende.de) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
A considerable part of the ozone destruction in the troposphere and stratosphere is attributed to bromine species (see, e.g.Prather et al., 1984;Platt and H önninger, 2003;von Glasow et al., 2004;Sinnhuber and Folkins, 2006) formed from bromoform (CHBr 3 ) of oceanic origin.Hence, bromomform sources and sinks in the ocean are of interest for studying atmospheric trace gas cycles.Particularly high atmospheric concentrations of bromine species (see Read et al., 2008) and corresponding air-sea fluxes of CHBr 3 (Quack et al., 2004) have been found in the tropics and subtropics, both offshore and in coastal regions.Recent CHBr 3 observations in the tropical eastern Atlantic Ocean (Quack et al., 2004(Quack et al., , 2007) ) have revealed a pronounced subsurface maximum in the depth of the subsurface chlorophyll maximum, suggesting a phytoplanktonic source of bromoform.This is consistent with laboratory experiments and field data that suggest a coupling between bromoform production and phytoplankton growth for diatoms and cyanobacteria (Moore et al., 1996;Schall et al., 1996;Karlsson et al., 2008).Whether CHBr 3 production serves a specific purpose (e.g., protection against grazing and cell poisoning by hydrogen peroxide -see e.g.Manley, 2002) or has to be regarded as the byproduct of another process is currently not known.It can also not be ruled out that bromoform is produced during the decay of organic matter, i.e. phytoplankton (Quack and Wallace, 2003) when the enzyme bromoperoxidase (necessary to catalyse the reaction with bromide ions) is set free.Whichever it may turn out to be, a close link to growth and/or decay of phytoplankton seems highly plausible.Alternative ideas about open ocean bromoform sources (coastal production by macroalgae and/or advective transport, see e.g., Moore and Tokarczyk, 1993;Carpenter and Liss, 2000) are not considered here for this region.
Several processes have been found responsible for the reduction of CHBr 3 concentrations in the ocean: Outgassing to the atmosphere may be the major factor in the surface mixed layer (see Quack et al., 2004).Other known decay processes include halide substitution of CHBr 3 which leads to an estimated oceanic half life between 5 and 74 years (at 25 • C and 2 • C, respectively) as well as hydrolysis working ten to hundred times slower (Geen, 1992;Vogel et al., 1987).In addition, it is assumed that photolysis plays an important role in the decay of CHBr 3 in the surface ocean (Carpenter and Liss, 2000).Several biologically mediated processes including the decomposition of organic material by bacteria (Fetzner, 1998) accelerate bromoform degradation.
Recent observations indicate also a co-metabolism of bromoform during nitrification: The decay of bromoform has been shown to be more rapid in the presence of Nitrosomonas which oxidizes ammonium to nitrite (Wahman et al., 2005).The pronounced nitrite maximum below the subsurface chlorophyll maximum layer in the oligotrophic Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion open ocean (data from Meteor expedition M55, Wallace and Bange, 2004) might be an indication of not only nitrification1 , i.e. ammonium oxidation (Olson, 1981), but also enhanced bromoform decomposition.Observational evidence of a decrease of CHBr 3 in the nitracline, where nitrification is usually maximum, exists (Quack, unpublished data from the Meteor cruise 60/5).
In summary, the magnitude of water column losses, their vertical distribution and their effect on surface mixed layer bromoform concentrations and the associated outgassing are only poorly known (see e.g., Abrahamsson et al., 2004).In this study, we use numerical modelling to elucidate the role of different sources and sinks of open ocean CHBr 3 in controlling the oceanic bromoform levels and the resulting air-sea flux.

Physical model setup
The physical model is the one-dimensional water column model GOTM (General Ocean Turbulence Model, Umlauf et al., 2005, www.gotm.net).The model has been previously applied to several oceanic regimes, including the subtropical oligotrophic North Atlantic Ocean (BATS: Bermuda Atlantic Time-series Study, Weber et al., 2007) and the Baltic Sea (Burchard et al. 2006).
Here, GOTM is configured for the eastern tropical North Atlantic Ocean for the Cape Verde region.The model set-up is based on the BATS case, with only few modifications: a relatively high minimum value for turbulent kinetic energy (TKE) of 10 −5 m 2 s −2 is prescribed to parameterise the effects of double diffusion in this region (see e.g., Martínez-Marrero et al., 2008).In addition, we have applied the bulk air-sea parameterisation for heat and freshwater fluxes of Fairall et al. (1996).The model covers the upper 700 m of the ocean; the lower boundary has been placed at this depth because it Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion is close to the nutrient maximum where the diffusive fluxes vanish (see also Beckmann and Hense, 2007).The vertical resolution is 1 m and the time step 1 h.

Biogeochemical model setup
Our ecosystem model is based on the relatively simple nitrogen-phytoplanktonzooplankton-detritus (NPZD) model of Schartau and Oschlies (2003).Weber et al. (2007) have coupled it to GOTM for their study of the oligotrophic situation at BATS; we found no need to change any of their model parameters.For our purposes, we have added a bromoform compartment B (in nmol CHBr 3 m −3 ) that evolves due to internal sources (Q) and sinks (S) as well as vertical diffusion: where A v is the time-and depth-dependent turbulent vertical diffusivity coefficient.
The source terms are either coupled to primary production (Q 1 ) or to the total phytoplankton losses (Q 2 ) Here, P and Z are the phytoplankton and zooplankton concentrations, respectively (in mmol N m −3 ), µ is the actual (in situ) growth rate of phytoplankton, γ is the mortality rate of phytoplankton, r and g Z are phytoplankton respiration and phytoplankton-dependent grazing rates of the NPZD model.We call the bromoform production to phytoplankton production ratio β, and a conceptually similar bromoform production to phytoplankton loss ratio β * .These proportionality factors were computed based on a suite of laboratory studies which report bromoform increase during the exponential growth phase of phytoplankton (Moore et al., 1996).First, we extracted the maximum specific growth rates ω Introduction

Conclusions References
Tables Figures

Back Close
Full of the three species under investigation from the exponential phase2 ; then the corresponding temporal changes in bromoform concentrations (∆CHBr 3 ) over time interval ∆t were used to determine the bromoform production to phytoplankton production ratio according to The resulting values for β are 4.2, 6.71, and 6.73×10 −7 nmol CHBr 3 (mmol N) −1 , using typical cellular carbon contents of the species (Wasmund, pers comm.: 176 pg C cell for Nitzschia sp. and Nitzschia arctica, 2015 pg C cell −1 for Porosira glacialis) and the Redfield ratio for conversion into nitrogen.The first value stems from an experiment (with Nitzschia arctica) which has been terminated prematurely (i.e., during the exponential growth phase).We therefore choose the average of the other two β=6.72×10−7 nmol CHBr 3 (mmol N) −1 for our numerical experiments, assuming that this value (obtained for cold water diatoms) is generally appropriate for other species and phytoplankton groups.Not only do we assume that many more species do produce bromoform, but also that the rate is relatively constant across the phytoplankton assemblage.Supporting evidence for the first assumption is that bromoperoxidase enzymes which are necessary for bromoform production have been identified for diatoms, cyanobacteria, and Prymnesiophyceae (Moore et al., 1996;Jakopitsch, 2001;Hughes et al., 2006;Karlsson et al., 2008).We realise that the second assumption may be more problematic, but to our knowledge there are no studies that disprove it.This is an area that needs further research.An alternative interpretation of the laboratory experiments by Moore et al. (1996) is that bromoform production is correlated with the release of bromoperoxidase from the phytoplankton cell as a result of respiration and during the death of the cell.Hence, Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion a similar approach leads to the formulation Q 2 and β * can be determined from the decay phase in Moore et al. (1996)'s experiments.These data are more difficult to extract, but it seems that β * is close to β, so for simplicity we assume them to be equal.Bromoform is lost from the ocean through outgassing: using the gas transfer velocity k after Nightingale et al. (2000) and the atmospheric equivalent concentration B equi (based on the parameterisation by Moore et al., 1995).
Four different processes are assumed to play a role as internal sinks for bromoform: photolysis, halide substitution and hydrolysis, degradation during remineralisation of detritus and nitrification.
Photolysis is considered in all experiments.We use a depth-decreasing time scale for degradation by ultraviolet radiation according to where I 0 denotes the instantaneous surface irradiance, I ref is the annual average irradiance for the uppermost meter in this region (75 W m −2 ), a w the attenuation coefficient for UV-light (0.33 m −1 ) and z the depth (positive downward).The decay time scale (l uv ) −1 is 30 days as given by Carpenter and Liss (2000).Due to the short penetration depth of ultraviolet radiation, photolysis can only play a role in the surface mixed layer.
In addition, we have implemented the following loss terms: halide substitution and hydrolysis is assumed to be constant throughout the water column.A constant decay rate l con for CHBr 3 is used, where the exponential time scale (l con ) −1 =6.33 years (a half life of 4.37 years) is at the lower end of the range of the estimated half life of bromoform (halide substitution, hydrolysis) from observations (5-74 years, Quack and Wallace, 2003): Tables Figures

Back Close
Full This approach is the most simple choice, as it requires the specification of only one parameter.
bromoform degradation related to the remineralisation of detritus (D) (following Fetzner, 1998) using an exponential decay time scale of (l rem ) −1 =7.92 years (half life: 5.49 years), and taking into account the temperature dependence for remineralisation from the NPZD model: The reference concentration for detritus is as a (crude) parameterisation for the degradation of CHBr 3 by nitrifiers the decay rate has been inversely coupled to the light field I, assuming that nitrification (ammonium oxidation) is inhibited by high light levels (Olson, 1981).To our knowledge, no empirical information exists about the light dependence of nitrification limitation.We therefore adopt the functional dependence of the phytoplankton light limitation to describe this process based on 24 h averages of irradiance: The exponential decay time scale is assumed to be (l nit ) −1 =4 years (half life: 2.74 years); the additional parameters are a=3.0×10−5 m 2 W −1 s −1 and λ=3.1×10 −6 s −1 .
In this paper, we shall present results from four experiments, which use different combinations of source and sink terms (see Table 1).

Forcing and initialisation
The model is forced by climatological monthly means of 2-m atmospheric temperature, air pressure and dew point temperature, 10-m zonal and meridional wind velocities, cloud cover as well as precipitation based on ERA40 reanalysis (Uppala et al., 2005).The physical variables temperature and salinity are initialised with climatological profiles from the World Ocean Atlas (WOA01) (Conkright et al., 2002).Lateral effects in the one-dimensional model are taken into account by restoring salinity and temperature (except for the upper 20 m) towards climatological monthly means of WOCE with a five day-timescale.All biogeochemical variables are initialised homogeneously except nitrogen for which the vertical profile of dissolved inorganic nitrogen derived from measurements at Cape Verde (Bange, 2007) has been taken.At the bottom, the DIN concentration is restored to the observed value of 35.7 mmol m −3 with a relaxation time scale of one hour.The atmospheric concentration of CHBr 3 is set constant, using a value of 0.14 nmol m −3 (3.4 ppt) which is a mean (and typical) value for atmospheric bromoform concentrations in that particular region (Quack et al., 2004).Although initialization and forcing fields are from Cape Verde at 17.4 • N, 24.5 • W, we believe that our results are representative for the entire eastern tropical Atlantic Ocean.

Results and discussion
The model has been run for 50 years after which a perpetual seasonal cycle has developed; we have analysed the last year.The relatively long spin-up time is required for the slowly evolving bromoform distributions.The physical fields reach a periodically steady state after about 6 years, the biological variables after about 10 years.

BGD Introduction Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion

Hydrographic and biogeochemical characteristics
Our simulation of the physical conditions at Cape Verde reproduce the temperature and salinity fields reasonably well (Fig. 1).The annual mean vertical profile of salinity shows a subsurface salinity maximum in about 100 m depth (Fig. 1a) which is a feature of the tropical eastern North Atlantic Ocean resulting from the salt water intrusion from the South Atlantic, the SACW (South Atlantic Central Water).As expected for this region, the main thermocline is found between 50 and 100 m (Fig. 1b).
The annual mean vertical profile of nutrients shows a depletion in the upper 50 m (Fig. 1) due to the uptake by phytoplankton and subsequent sinking of dead organic matter.The maximum concentrations of phytoplankton biomass are found subsurface in the upper nutricline where light and nutrient supply are optimal.This is in agreement with observations in that region (Quack et al., 2004;Baham ón et al., 2003).Due to sinking, the detritus maximum is located below the phytoplankton maximum.

Vertical distribution of bromoform
As expected for a phytoplankton related source of CHBr 3 in the oligotrophic ocean, the simulated annual mean vertical distribution of bromoform shows a pronounced subsurface maximum in all experiments (Fig. 2, Table 1).The maxima are located in 66 m depth, a few meters below the phytoplankton biomass maximum and occur at the depth of the strongest temperature and nutrient gradient.
We note that differences between the experiments are smallest at the surface and largest below 200 m depth.The surface concentrations are controlled by the air-sea flux, photolysis and entrainment, which are similar in all experiments (Fig. 3).The subsurface concentration of CHBr 3 in the water column, however, depends strongly on the assumed internal decay process (Fig. 3d).
The most striking difference between our experiments are the comparatively high concentrations of CHBr 3 below 200 m depth in EXP2 (Fig. 2, green line).This means that degradation coupled to remineralisation leads to high concentrations at depth, be-Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion cause most of the remineralisation takes place above 200 m depth.Hence, all the bromoform that escapes into the deep ocean will accumulate there.This feature is related to the biogeochemical model and its values for detrital sinking and remineralisation rates.The few available observational data of CHBr 3 concentrations below 200 m show that the concentration at 630 m is close to 1.0 nmol CHBr 3 m −3 (Quack, unpublished data, 2004).While similar concentrations can be found for EXP1, EXP3 and EXP4 (0.8-1.3 nmol CHBr 3 m −3 ), they are significantly higher in EXP2 (6.5 nmol CHBr 3 m −3 ) indicating that the degradation is underestimated.One obvious way to decrease the deep CHBr 3 concentrations in EXP2 is to increase l rem , but this erodes the subsurface maximum to the point of being unrealistic.
The other three experiments show similar vertical profiles.In particular, we note that the type of bromoform source (Q 1 vs. Q 2 ) has hardly an effect on the annual mean profiles 3 .This is both encouraging and disappointing: on the one hand we do not need to know more about the details of the production process to obtain a realistic result; on the other hand we cannot use our experiments to distinguish the different hypotheses.
The highest maxima and the largest gradients of bromoform are produced in EXP3 and EXP4 (Fig. 2, red lines).A slightly reduced subsurface maximum of CHBr 3 and higher concentrations in 700 m depth are found in EXP1 (Fig. 2, blue curve).
A more quantitative evaluation of the model results can be performed for the month of November.Observations of bromoform in the eastern North Atlantic Ocean down to 200 m are available for a number of stations from a cruise in the beginning of November 2002 (Quack et al., 2004).Spatial (and to a lower degree temporal) variability yields profiles that differ in the depth, thickness and amplitude of bromoform maximum layers.Our one-dimensional water column model, forced by climatological fields, does not represent an individual observed profile.Therefore, we have constructed a "typical" (averaged) CHBr 3 distribution, applying the method of Hense and Beckmann (2008) Instead of averaging along depth horizons, all the maxima are aligned to obtain a "mean" profile (Fig. 4).The advantage of this approach is that two important characteristics of the distributions (amplitude and thickness) are reliably preserved while the drawback is that absolute depth information is lost (see Hense and Beckmann, 2008).Figure 4 should therefore be used only to compare amplitude and thickness between this composite of observations and the simulated profiles.
The model-data comparison for EXP1-EXP3 shows that the simulated vertical distribution patterns of CHBr 3 capture the most important features of the observations: the magnitude of the subsurface maximum, a strong decrease upward to the surface and a somewhat slower decrease downward (Fig. 4).Given the uncertainties of model parameters and forcing, the robustness of the results support the validity of our assumptions.
The systematic overestimation of the surface concentration in all models may be due to the use of a constant atmospheric bromoform concentration and the climatological wind fields which both affect the air-sea flux.Additional sensitivity experiments indicate that slightly changed atmospheric data can easily account for concentration changes of 2 nmol CHBr 3 m −3 in the surface mixed layer.
Differences between simulations and observations occur also with respect to the thickness of the bromoform maximum layer.The observed "typical" profile of bromoform has a thickness of about 60 m (based on a "full width at half maximum" criterion).
The simulated profiles show thicker maximum layers (EXP1: 105 m, EXP2: 126 m, EXP3 and EXP4: 100 m), most likely due to an overestimation of the thickness of the subsurface biomass maximum by the ecosystem model.Simultaneous measurements of bromoform and phytoplankton biomass at high vertical resolution would be required to make conclusive statements in this respect.
As mentioned before, the largest maxima and the steepest downward gradients of bromoform are obtained for S 3 , the "nitrification case" (Fig. 4, red curve).This is a consequence of the vertical separation of the production layer from the destruction layer below.It seems that the observed profiles can be best explained by adjacent, non-

Conclusions References
Tables Figures

Back Close
Full The quantitative evaluation of our model results shows that the concentration in the subsurface maximum is best represented by EXP3 or EXP4 (Table 1).Overall, however, the differences in the concentrations of bromoform in the upper 200 m between the experiments, particularly EXP1 and EXP3 are rather small.This is also true with respect to air-sea fluxes: The simulated flux from the ocean into the atmosphere ranges only between 163-188 pmol CHBr 3 m −2 h −1 for November for the individual experiments.They are close to the observed 175 pmol CHBr 3 m −2 h −1 in this region (Table 1).
A sensitivity experiment with photolysis as the only internal loss process resulted in an imbalance between sources and sinks below the subsurface maximum and lead to unrealistically high and continuously increasing subsurface bromoform concentrations (no steady state is reached within 100 years of model integration).Vice versa, omitting photolytic destruction leads to an increase in the surface concentrations and a higher air-sea flux (8-36%).
In summary, EXP3 shows the best agreement with observations and thus, for further analyses, we focus on the results of this case.

Simulated seasonal cycle of bromoform and air-sea fluxes
Having established that the model does represent the main features of the observed bromoform distribution reasonably well, we can now look at unobserved aspects of the problem, e.g., the seasonal cycle.The highest subsurface maximum of bromoform (>24 nmol CHBr 3 m −3 at around 66 m) occurs in late autumn while the lowest concentrations can be found in spring (<18 nmol CHBr 3 m −3 in about 80 m, Fig. 5).
We note a time delay of about one month between the occurrence of the subsurface maximum and the occurrence of the surface mixed layer maximum, related to the gradual deepening of the surface mixed layer and the entrainment of the bromoform signal (Figs. 5, 6).The lowest subsurface concentrations are found in April.During spring and summer, the surface layer is essentially decoupled from the subsurface bromo-

Conclusions References
Tables Figures

Back Close
Full form production layer and the surface concentrations decrease until early August.
The seasonal cycle of bromoform surface concentration and air-sea flux shows winter maxima and summer minima (Fig. 6).Concentrations vary by a factor less than two, fluxes by a factor of almost ten.It is important to note that the seasonal flux variations are only partly caused by the variations in production; the variability in mixed layer depth, which in turn is driven by changes in heat and freshwater fluxes as well as the wind speed, is the main factor.While the absolute maximum of the air-sea flux occurs in winter, secondary maxima are found in early summer and autumn (Fig. 6).From April to June relatively high wind velocities (Fig. 6c) promote the outgassing from the surface ocean.In September and October, the combined effect of the increasing wind speed and the deepening of the mixed layer (Fig. 6a,c) inducing the entrainment of bromoform into the surface layer leads to a rapid increase in the air-sea flux (Fig. 6b).We notice that the temporal variation of bromoform in the water column is similar in all our experiments irrespective of differences in the absolute subsurface concentrations.

Summary and conclusions
We have simulated the climatological seasonal cycle of the vertical distribution of bromoform in the upper 700 m of the tropical eastern Atlantic Ocean.Using the results from laboratory experiments (Moore et al., 1996) bromoform production has been coupled to the growth and losses of phytoplankton, respectively.The differences between these two approaches are marginal and leave open which physiological processes are ultimately responsible for the production of CHBr 3 .The generally good agreement of the model results with observations, however, suggests that the subsurface maximum can be entirely explained by in-situ production in which phytoplankton plays the most important role.This result is of more general relevance because observations indicate that the subsurface bromoform maximum is an ubiquitous feature of the eastern tropical Atlantic Ocean (Quack et al., 2004(Quack et al., , 2007)) that besides the near surface processes photolysis and air-sea flux, additional water column losses are necessary to explain the downward decrease in concentrations below the subsurface maximum.Among the three cases we compared, the experiment with a degradation proportional to nitrification (parameterised by an inverse proportionality to the light field) is in best agreement with observations, i.e., has the highest maximum and the lowest deep concentration of bromoform.This is achieved by the vertical separation of the bromoform production layer from the bromoform destruction layer below.In the assessment of our model results we have to keep in mind that although we have investigated the various bromoform degradation mechanisms individually, they are likely to be active simultaneously (at unknown relative strengths).In any case, nitrification-related degradation of bromoform seems essential in generating the observed vertical bromoform distribution.This is not an unreasonable result, given that observations point at a co-metabolism of CHBr 3 in the presence of nitrifiers (Wahman et al., 2005).In addition, recently obtained field data show a maximum of dibromomethane (a degradation product of CHBr 3 from reductive debromination processes) below the maximum of CHBr 3 (Quack, unpublished data, 2004, Butler et al., 2007).Further studies on subsurface bromoform degradation are clearly needed.
The model results further suggest that the seasonality of oceanic bromoform outgassing is larger than often assumed.While surface concentrations range from 5 to 8 nmol CHBr 3 m −3 (summer versus winter), the air-sea flux varies by almost an order  Full of magnitude.Strongly controlled by the seasonal changes in wind speed and mixed layer depth, outgassing is 40 pmol CHBr 3 m −2 h −1 ) in summer and 360 pmol CHBr 3 m −2 h −1 in winter.Note that the use of seasonally varying atmospheric concentrations of CHBr 3 will lead to (probably minor) quantitative changes in these values.The vertically integrated annual production rate of bromoform in the model is 4.6 µmol CHBr 3 m −2 a −1 (525 pmol CHBr 3 m −2 h −1 ) with a corresponding annual mean air-sea flux of 1.1-1.3µmol CHBr 3 m −2 a −1 (125-148 pmol CHBr 3 m −2 h −1 ) for the different experiments; i.e., about 25% of the production is transferred to the atmosphere.There, as a carrier of bromine, CHBr 3 of phytoplanktonic origin will contribute to the
. Our numerical experiments with different degradation mechanisms of CHBr 3 show Figures

Table 1 .
(Quack et al., 2004)nts and selected results: observed and simulated bromoform concentrations (nmol CHBr 3 m−3 ) at the surface and in the subsurface maximum as well as air-sea flux (pmol CHBr 3 m −2 h −1 ) for the month of November.Observed data are averages of offshore stations from the cruise M55(Quack et al., 2004)west of 23• W.