Beyond the Fe-P-redox connection: preferential regeneration of phosphorus from organic matter as a key control on Baltic Sea nutrient cycles

. Patterns of regeneration and burial of phospho­ rus (P) in the Baltic Sea are strongly dependent on redox conditions. Redox varies spatially along water depth gra­ dients and temporally in response to the seasonal cycle and multidecadal hydrographic variability. Alongside the well-documented link between iron oxyhydroxide dissolution and release of P from Baltic Sea sediments, we show that pref­ erential remineralization of P with respect to carbon (C) and nitrogen (N) during degradation of organic matter plays a key role in determining the surplus of bioavailable P in the water column. Preferential remineralization of P takes place both in the water column and upper sediments and its rate is shown to be redox-dependent, increasing as reducing conditions be­ come more severe at greater water-depth in the deep basins. Existing Redfield-based biogeochemical models of the Baltic may therefore underestimate the imbalance between N and P availability for primary production, and hence the vulnerabil­ ity of the Baltic to sustained eutrophication via the fixation of


Introduction
Hypoxia has been intermittently present in the Baltic Sea throughout the Holocene (Zillen et al., 2008), but its severity and spatial extent have increased greatly over the past 1 0 0 years (Conley et al., 2009a: Fonselius and Valderrama, 2003: Jonsson et al., 1990).This has led to multiple ecological im pacts, such as the disruption of benthic foodwebs (Karlson et al., 2005) and fish habitats (Cardinale and Modin, 1999).Rising anthropogenic nutrient loadings in the early 20th cen tury are believed to have provided the initial trigger to the modern eutrophication of the Baltic, with the resulting in creased biological oxygen demand forcing the exacerbation of hypoxia (Savchuk et al., 2008).However, both eutrophic conditions and hypoxia have persisted to the present day, de spite a decline in fertilizer use in the catchment since the 1990s (Emeis et al., 2000).Natural feedbacks within the cy cles of phosphorus (P) and nitrogen (N) in low-oxygen sys tems (Middelburg and Levin, 2009) likely play a key role in allowing the Baltic to self-sustain hypoxia.Regeneration of P as phosphate (HPO^-) is known to be enhanced from sed iments underlying hypoxic or anoxic bottom waters.Phos phate is released during the remineralization of organic mat ter, and the rate of P remineralization may accelerate rela tive to that of carbon and nitrogen during anaerobic degra dation (Ingall et al., 1993).Additionally, the dissolution of P-bearing Fe-oxyhydroxides in surface sediments increases the rate of P regeneration during the transition into hypoxia or anoxia (Einsele, 1936: Mortimer, 1941, 1942) .On the other hand, nitrification and denitrification rates may also vary as oxygen is depleted (Kemp et al., 1990).Basin-scale biogeochemical modeling of the Baltic Proper through the last century shows the net result of altered anthropogenic nutrient loading and oxygen-dependent feedbacks to be an excess of phosphate in the modern water column (Savchuk et al., 2008).This situation stimulates the fixation of atmo spheric N by cyanobacteria (Vahtera et al., 2007), maintain ing high primary productivity in the basin.In turn, decay of sinking biomass maintains hypoxia in the bottom waters, providing the conditions for ongoing phosphate regeneration from the sediments.
Within the context of a generally eutrophied, hypoxic modern Baltic, the spatial distribution of hypoxia is con trolled by the physical characteristics of the water column (Fonselius, 1981).In the Baltic Proper, a persistent strat ification exists between surface waters of salinity 7-8 and bottom waters of salinity 11-13, the latter generated by ma jor inflows across the Danish Straits (Matthaus and Franck, 1992: Meier et al., 2006: Schinke and Matthaus, 1998).Vertical profiles of salinity and dissolved oxygen through out the basin are strongly anti-correlated, with a mid-depth halocline marked by a steep decline in oxygen (Fonselius, 1981), across which turbulent diffusive exchange is limited (Gustafsson and Stigebrandt, 2007).The frequency of ma jor saline inflows affects the spatial extent of hypoxia by de termining the depth and gradient of the halocline.During stagnations (i.e. the multiannual periods between major in flows) , the halocline may be eroded vertically, reducing the total area of bottom-water hypoxia.New inflows then re establish a strong halocline, allowing bottom water hypoxia to spread (Gustafsson, 2000).Major inflows always occur during winter westerly storms across the Danish Straits, but their frequency has reduced dramatically from nearly annual in the early to mid-2 0 th century to roughly decadal since 1983 (Schinke and Matthaus, 1998), with the most recent oc curring in 2003.This change in frequency of major inflows is believed to result from the relative weakness of easterly winds in the early winter "preconditioning" phase during re cent decades (Lass and Matthaus, 1996), and has led to highamplitude variability in the area of hypoxic bottom waters during this period (Conley et al., 2002).The size of the hy poxic area also displays a large seasonal fluctuation due to oxygen consumption after the spring-summer phytoplankton bloom (Conley et al., 2002).
Given the coupling between hypoxia and nutrient cycles, the strong physically-driven variability in bottom water hy poxia during the inflow-stagnation cycle may be expected to influence regeneration and burial of nutrients on the basinscale, and hence to influence nutrient budgets of the Baltic (Gustafsson and Stigebrandt, 2007: Savchuk andWulff, 2009).In the case of P, it has been proposed that sediments in the Baltic Proper act as a net source of phosphate dur ing expansion of the hypoxic area, as Fe-oxyhydroxides at the sediment surface are reductively dissolved in the absence of oxygen (Conley et al., 2002).Conversely, the opposite should be true as the hypoxic area contracts, namely that the sediments become a net sink for P. In accordance with this theory, the basin-scale modeling approach of Savchuk and Wulff (2009) considers a linear enhancement of P regenera tion with expansion of the hypoxic area.Similarly, Gustafs son and Stigebrandt (2007) assume the uptake or release of P by sedimentary iron oxyhydroxides to account for any devi ation in water column phosphate concentrations away from the value predicted by remineralization of organic matter in Redfield proportions.However, the behavior of phosphorus in Baltic Sea sediments has only recently begun to be studied in detail (Hille et al., 2005: Lukkari et al., 2009: Mort et al., 2 0 1 0 ), and a large amount remains to be learned about the response of phosphorus regeneration, and burial, to spatial and temporal hydrographic variability.In particular, the re spective roles of the iron oxyhydroxide-P (or "Fe-P") interac tion, and preferential remineralization of P during anaerobic degradation of organic matter, are poorly constrained.Al though some recent models, such as BALTSEM of the Baltic Nest Institute in Stockholm, include preferential remineral ization of P from organic matter as a process (Oleg Savchuk, personal communication, 2 0 1 1 ), a lack of comprehensive geochemical data hinders validation of its magnitude.The present study aims to address this issue.Firstly, we investi gate the characteristics of P regeneration during the current stagnation using a broad network of sediment and porewater profiles throughout the Baltic Proper and the Gulf of Fin land.We adopt a water-depth based approach, to investigate the contrasting behavior of sites along the redox gradient into the deep basins.Secondly, we investigate the response of P burial to the inflow-stagnation cycle by studying the accu mulation rate of P in 2 1 0 Pb-dated sediment cores.The results allow a re-examination of existing theories about the rela tionship between hypoxia and the phosphorus cycle in the Baltic.

Onboard sampling and analyses
A total of 17 sites in the Baltic Proper and Gulf of Finland, covering a range of water depths, were visited during the HYPER/COMBINE cruise of the R/V Aranda in May/June 2009 (Fig. 1).Water column salinity and dissolved oxygen profiles were measured using a CTD system, and additional bottom-water oxygen concentrations were measured by Win kler titration.A multicorer retrieving sub-cores of length <50 cm was deployed at each site.At 9 sites, a bottom wa ter sample was taken from one sub-core using a 2 0 ml plas tic syringe, and transferred immediately to an anoxic glove box.The same sub-core was inserted into the glove box through an air-tight hole in the base and the remaining bot tom water was removed.The sub-core was then sliced at a resolution of 0.5 cm (0-2 cm depth), 1cm (2-10 cm depth) and 2 cm (>10 cm depth).An aliquot of each wet sediment sample was transferred to a 50 ml plastic centrifuge tube, which was capped, removed from the glove box and cen trifuged at 2500 g for 10-30 min.The centrifuge tubes were then returned to the glove box.The supernatant water from each sample, including the bottom water sample, was dis pensed via a 20 ml plastic syringe through a 0.45 jtm filter and collected in a 15 ml plastic centrifuge tube.Subsam ples were taken for analysis of dissolved phosphate and iron (acidified with lOgl of 37% HC1 per ml of subsample and stored at 5 °C) and ammonium, nitrate and sulfate (stored at -20 °C).The remainder of each wet sediment sample was transferred to a pre-weighed 15 ml glass vial for solid-phase analyses.The vial was capped and stored in a nitrogen-filled airtight jar.At the remaining 8 sites, a bottom water sample was taken as above, and porewater samples were taken us ing rhizon samplers (Seeberg-Elverfeldt et al., 2005) from a sub-core pre-drilled with 3 mm diameter holes at 1 cm ver tical spacing.The uppermost rhizon sample was taken from the first hole below the sediment water interface and was as sumed to collect porewater from the upper 1 cm of sediment, centered on 0.5 cm depth.The rhizons were attached to 10 ml plastic syringes, which were allowed to fill with porewater (for 10-30 min) and transferred to the glove box for sub sampling as above.Site BY2 was also visited during a cruise with the R/V Skagerak in September 2007 and sampled for porewater and solid-phase parameters as per the above pro tocols, with porewater extraction by centrifugation (Mort et al., 2 0 1 0 ).

Laboratory analyses
Phosphate, ammonium and nitrate in porewater samples were determined spectrophotometrically using an Autoanalyser.Sulfate was determined using a Dionex Ion Chromatograph.Iron and other elements in porewaters were analyzed by Perkin Elmer Optima 3000 Inductively Coupled Plasma-Optimal Emission Spectroscopy (ICP-OES; all iron was pre sumed to be present as Fe2+).Wet sediment samples from BY2, GOF3, GOF5, LF1 , GOF6 , LF1.5, LF3, LL19, F80 and BY15 were frozen within the airtight jars, then removed from the jars and freeze-dried to estimate water content.The dried sediment was transferred to an anoxic glove box and ground in an agate mortar.Aliquots of 0.5 g dried sediment were de calcified by shaking in excess 1 M HCl, initially for 12 h and for a further 4 h after addition of new acid.Tests have shown that the amount of organic carbon hydrolyzed by this proto col is negligible (van Santvoort et al., 2 0 0 2 ).The decalcified sediment was dried, ground in an agate mortar and analyzed by combustion for organic carbon (Corg) and total nitrogen (Ntot) using a Fisons NA 1500 NCS (precision and accu racy < 2 % based on an atropine/acetanilide standard calibra tion and checked against internal laboratory standard sedi ments).The non-organic contribution to Ntot is assumed neg ligible in these sediments, hence Ntot ~ Norg.A second 0 . 1 g aliquot of dried sediment was dissolved in 2.5 ml HF (40 %) and 2.5 ml of a HCIO4/HNO3 mixture, in a closed Teflon bomb at 90 °C for 1 2 h.The acids were then evaporated at 190 °C and the resulting gel was dissolved in IM HNO3 , which was analyzed for selected elemental concentrations by ICP-OES (precision and accuracy <5 %, based on calibration www.biogeosciences.net/8/1699/2011/Biogeosciences, 8, 1699-1720, 2011 to standard solutions and checked against internal laboratory standard sediments).A third 0.1 g aliquot was subjected to the SEDEX sequential extraction procedure to determine the solid-phase partitioning of phosphorus (Ruttenberg, 1992).All H2 O rinses were omitted from the original protocol (see Slomp et al., 1996), but three MgCH rinses were included.The full spéciation was as follows: exchangeable-P (ex tracted by 1M MgCl2 , pH 8 , 0.5 h), iron oxyhydroxide-bound P ("Fe-P", extracted by citrate-bicarbonate-dithionite (CDB), pH 7.5, 8 h, followed by 1M MgCD, pH 8 , 0.5h), authigenic Ca-P (including carbonate fluorapatite, biogenic hydroxyapatite and CaCC^-bound P, extracted by Na-acetate buffer, pH 4, 6 h, followed by 1M MgCH, pH 8 , 0.5 h), detrital Ca-P (extracted by 1 M HCl, 24 h) and organic P (Porg, after ashing at 550 °C for 2 h, extracted by IM HC1, 24 h).
The initial 1M MgCH and CDB rinses were performed in side the anoxic glove box, to eliminate the potential conver sion of authigenic Ca-P to Fe-P due to pyrite oxidation upon oxygen exposure (Kraal et al., 2009).Dissolved phosphate in all rinses was analyzed colorimetrically (Strickland and Parsons, 1972) on a Shimadzu Spectrophotometer (precision and accuracy < 2 %, based on calibration to standard solu tions and checked against internal laboratory standard sedi ments), with the exception of the CDB rinse, in which P and Fe were analyzed by ICP-OES.Only the Porg data (from all analyzed sites) and Fe-P data from BY2 (June and Septem ber) are presented in this paper.A fourth 0.1-0.5 g aliquot of dried sediment from sites LF1, LFE5 and LF3 was re served for 210Pb analysis.This was spiked with 209Po and digested in 10 ml concentrated HC1 in a microwave oven for 3h. 2 ml 3.5% HCIO4 was then added and the acids were removed by evaporation.The resulting precipitate was re dissolved in 5 ml concentrated HC1 for 30 min.Thereafter, 40 ml 0.5 M HCl (with 12 g I-1 boric acid), 4 ml NH4 OH and 5 ml 40 g I-1 ascorbic acid (in 0.5 M HCl) were added.Poisotopes were deposited by suspending silver disks in the so lution, which was heated to 80 °C for 4 h then left overnight at room temperature.The activity of 210Po was measured by a-spectrometry with Canberra Passivated Implanted Planar Silicon (PIPS) detectors, allowing 210Pb activity to be backcalculated.1er counting uncertainty was on average 5 %.

Porewater profiles and regeneration fluxes
A complete overview of the water column and porewater data from the 17 sites sampled in May/June 2009, and site BY2 from September 2007 (Mort et al., 2010), is presented in Fig. 2 (raw porewater data is presented in Table AÍ).The sites range in water-depth from 48 m (BY2) to 238 m (BY15).For simplicity and synergy with Mort et al. (2010), we categorize all sites of <90 m depth as "Group 1 (oxicseasonally hypoxic)", whereas sites of > 9 0 m water depth are categorized as "Group 2 (semi permanently hypoxicanoxic) ".At the time of sampling in May/June 2009, all sites except BY2, LF1 and LF1.5 experienced bottom water oxy gen concentrations below 2 m lL_1, hence were "hypoxic" as per the commonly adopted definition (Diaz and Rosen berg, 1995).The five Group 2 sites were anoxic, with hydro gen sulfide present in the water column (plotted in Fig. 2 as "negative oxygen" concentrations, as per Fonselius (1981)).
The stratification control on water column oxygen concen tration is clear from the inverse relationship between oxygen and salinity at all sites, although the depth and gradient of the halocline vary between and within sub-basins.Most sites display evidence for sulfate reduction in the porewaters, with complete removal of sulfate within the depth of the multicore at a number of sites.Porewater ammonium (NH(j~) profiles at all sites show an asymptotic increase with depth, typical of NHj" release during organic matter remineralization in sedi ments undisturbed by bioturbation (Burdige, 2006).The flux of NHj" across the sediment-water interface, calculated from the concentration gradient between the bottom water and up permost porewater sample, therefore provides a minimum es timate of the rate of organic matter remineralization at each site.Although the true rate of remineralization at sites with oxygen penetration into the sediments may be greater, due to the oxidation of ammonium to nitrate in the uppermost sam ple, very low porewater nitrate concentrations at these sites (not shown) and the strongly asymptotic NHj" profiles sug gest this effect to be negligible.Hence, the relationship be tween the porewater profiles and fluxes of NH)¡", phosphate (HPO^-) and iron (II) (Fe2+) may justifiably be used to de convolve regeneration of HPO^-due to iron oxyhydroxide dissolution from regeneration due to remineralization of or ganic matter at the sites in this study.
Throughout both groups of sites, no clear trend is observed with water depth for absolute fluxes of NHj" or HPO^-(Fig.2).This statement holds whether sites sampled with rhizons (marked "R" in Fig. 2, for which a greater degree of er ror is expected due to the difficulty in capturing the true con centration gradient across the sediment-water interface) are included or excluded.However, a marked depth-dependence is observed in the NH^/HPO^-flux ratio (Fig. 3a).Most striking is the cluster of Group 1 sites recording values of 0-3, while all Group 2 sites record values of 10-20.This distribution is explained by closer examination of the upper 10 cm of the porewater NHj" and HPO^-profiles at each site.Group 1 sites with NH^/HPO^-flux ratios of 0-3 display a peak of porewater HPO^-in the shallow subsurface which is not replicated in NHj" (Figs. 2,3b).This peak causes the dif fusive HPO^-flux to exceed the value predicted by organic matter remineralization alone, suppressing the NH^/HPO^flux ratio.Coincident peaks in Fe2+, which drive non zero diffusive fluxes of Fe2+ (Figs. 2,3c), provide evidence that dissolution of P-bearing iron oxyhydroxides close to the   (Sept), for w hich profiles w ere m easured during a cruise w ith the R /V Skagerak in Septem ber 2007 (M ort et al., 2010).Sub-zero oxygen concentrations are estim ated from hydrogen sulfide concentrations using the ratio 1 m ol E ^S = 2 m ol O 2 (see Fonselius, 1981).M iddle panel: D iffusive fluxes o f NH j" and H P O ^ across the sedim ent-w ater interface, calculated from the concentration gradient betw een bottom w ater and the upperm ost porew ater sam ple (0-0.5 cm depth for cores sliced for porew ater extraction by centrifugation, 0 -1 cm depth for cores sam ple w ith rhizons, indicated "R ") using F ic k 's first law: w here F = flux in p m o lm -^ d a y -1 , and m inus sign indicates an efflux from the sedim ents, D = ion-specific diffusion coefficient, corrected for tem perature, taken from B oudreau (1997) and adjusted for salinity according to Van C appellen and W ang (1995), </> = porosity and 9 = tortuosity, defined as per B oudreau (1997): L ow er panel: Porew ater profiles o f HPO^", NH^", F e2+ and SO ^-.See text for details o f sam pling resolution and processing.SO ^_ profiles for six sites are presented as logarithm ic best-fit lines through raw data due to high scatter (data point in brackets for BY2 (June) not included in fit).A ll other profiles represent raw data only.o_ sediment-water interface was actively elevating the H P 0 4 presently involved in the P cycle at these water depths, and flux (relative to the expected flux from organic matter rem-hence that organic matter remineralization alone dictates the ineralization alone) at these sites at the time of sampling.
rate of H P 0 4_ regeneration.Two Group 1 sites (BY2 and Conversely, the lack of shallow-subsurface HPO^" peaks GOF5) record anomalously high NHpHPO^-flux ratios of at Group 2 sites suggests that iron oxyhydroxides are not profiles at these sites display low values in the uppermost ~5 cm, and an increase below this depth (Fig. 2).Such pro files are characteristic of sediments with an intact layer of iron oxyhydroxides at the sediment-water interface (Slomp et al., 1996;Slomp et al., 1998) which adsorb H P04_ , thereby suppressing the diffusive flux of H P04_ and elevating the NH+/HPO;;-flux ratio.
The total hypoxic area in the Baltic varies annually by some 10 000km2 (in a total of 20 000-70000 km2, depend ing on the position in the inflow-stagnation cycle (Conley et al., 2 0 0 2 ), due to the annual cycle of organic matter loading to the sediment-water interface (Stigebrandt, 1991).A sub stantial area of sediments above and within the halocline thus experiences major variability in redox conditions throughout the year, which may be expected to influence the interac tion between iron oxyhydroxides and H P04_ regeneration (Reed et al., 2011).Closer inspection of the data from site BY2, sampled both in June 2009 and September 2007, con firms the strong influence of seasonal redox variability on the P cycle.In the season of greatest oxygen stress (rep resented by September 2007), porewater peaks of H P04_ and Fe2+ are present close to the sediment-water interface (Fig. 4), driving high diffusive fluxes of both species and suppressing the NH^/HPO^-flux ratio (Figs. 2, 3).At this time, the concentration of Fe-P in the surface sediments is ~20pm olg-1 (Fig. 4).In contrast, during more oxic conditions (represented by June 2009) the subsurface pore water peaks of H P04_ and Fe2+are smaller and displaced downwards, enhancing the NH^/HPO^-flux ratio, while the surface sediments contain up to 40-60 pm olg-1 Fe-P (note: raw Fe-P data for BY2 in June 2009 are presented in Ta ble BÍ).Furthermore, the mass of Fe-P in the upper 20 cm of a 1 cm2 sediment column in September 2007 was 21.4 pmol, compared with 49.1 pmol June 2009 (calculated from the profiles in Fig. 4 and dry bulk density data, not shown).These results confirm that a substantial fraction of the pool of P associated with iron oxyhydroxides in surface sediments at BY2 (~27.7pmolcm -2 ) may be quickly regenerated to the water column during seasonal outbreaks of hypoxia, and conversely that P may be efficiently sequestered back to the sediments as oxygen conditions improve.It is noteworthy that seasonal outbreaks of even mildly "hypoxic" bottom wa ter conditions, such as those observed at BY2 in September (Fig. 4), are able to trigger such a strong response in sed imentary Fe and P dynamics.Hence, although geochemi cal processes in bottom waters may not be expected to vary greatly across the biologically-defined "hypoxia" threshold of 2 m lL_ 1 dissolved oxygen, the vertical migration of re dox zones within the underlying sediments leads to strong responses in sediment-and porewater-chemistry.
Due to the effects of seasonality, the data in Fig. 3 should be observed as a snapshot of conditions at the time of sam pling.All sites currently experiencing seasonality in bottom water oxygen concentrations -hence all Group 1 sites -may reasonably be expected to show a variable flux of HPO^across the sediment-water interface throughout the year.The H P 04-flux at such sites is strongly coupled to bottom water oxygen stress and the rate of iron oxyhydroxide dissolution in the sediments (Reed et al., 2011).The precise timing of the maximum HPO^-flux may vary from site to site, due to the changing shape of the organic carbon loading curve in differ ent regions of the Baltic (Stigebrandt, 1991), the ventilating action of small-scale inflows, and the water-depth dependent severity of oxygen depletion.A combination of these factors may account for the apparent offsets between the character istics of BY2 and GOF5, versus the other Group 1 sites sam pled in May/June 2009 (Fig. 3).In contrast, the Group 2 sites have experienced continuous hypoxia/anoxia throughout the year since early in the current stagnation, and the absence of iron oxyhydroxides in the surface sediments may be consid ered permanent within that time frame.
Closer inspection of the NH(j7HP04_ flux ratios at the five Group 2 sites (Fig. 3a) shows a trend of decreasing val ues from ~20 at the shallowest of the sites (LF3, 95 m) to ~10 at the deepest (BY15, 238 m).This observation sug gests that the remineralization rate of P relative to N from organic matter in the surface sediments increases with wa ter depth, supporting the concept of accelerated regenera tion of P from sediments with increasing severity of anoxic conditions (eg., Algeo and Ingall, 2007;Ingall et al., 1993;Kraal et al., 2010;Slomp et al., 2002).The mechanism of this process remains uncertain, but may be related to redoxdependent variability in the storage or release of P by sed imentary micro-organisms during the breakdown of organic matter (see discussion in Ingall and Jahnke (1997)).

Surface-sediment organic matter composition and relationship to regeneration fluxes
Surface sediment organic C:N:P ratios allow the hypothesis of accelerated remineralization of P with increasing water depth to be tested.Hypothetically, if organic matter arriving at the sediment-water interface at all sites had a consistent N:P ratio (Redfield or otherwise), surface sediment NtouPorg should display the inverse trend to that of the N H (jV H P 0 4_ flux ratio; namely an increase with water depth, reflecting the increase in preferential remineralization of P. With some scatter, mean Ntot : Porg in the top 2 cm of sediment at the ten sites selected for solid-phase analysis indeed shows an in crease with water depth, from 32 at GOF3 to 57 at LL19 (Fig. 5a).The trend includes both Group 1 and Group 2 sites, suggesting that NH^/HPO^-flux ratios at Group 1 sites would likely show the inverse relationship if the Fe-P interaction was not active.A similar pattern is also ob served in C0rg:P0rg, ranging from 328 at LF1 to 626 at LL19 (Fig. 5b), confirming that P is also preferentially remineral ized relative to C with increasing oxygen stress (note that all raw Corg, P org and N tot data are presented in Interestingly, the convergence point of the exponential curves describing the NH(j7HP04_ flux ratio and surface sediment Nt0t:Porg is close to N:P = 27 (Fig. 5a).One in terpretation of this value is that it approximates the Ntot : Porg ratio of organic matter arriving at the sediment-water inter face, and that during subsequent organic matter degradation, the NH(j7HP04_ flux ratio and Ntot:P0rg in the upper 2 cm of sediment then diverge to lower and higher values, respec tively.The extent of the divergence is controlled by redox conditions in the sediments, with strongly reducing condi tions causing the greatest offset.The initial value of 27 rep resents a two fold enrichment relative to Redfield stoichiom etry and appears high compared to the mean N:P values of material caught in sediment traps in the Gotland Deep (~9-11, Emeis et al., 2000;Struck et al., 2004).However, these studies report Ntot:Ptot> which may be suppressed by sinking or suspended non-organic P phases such as Fe-P and detrital-P.Dellwig et al. (2010) show maxima in suspended MnOx-Fe0 0 H-P0 4 in the vicinity of the redoxcline in the Got land Deep, and in suspended Fe0 0 H-P0 4 over a broad (at least 40 m) depth range immediately below the redoxcline.Thus, both the 140 m and 230 m traps reported in Emeis et al. (2000) could potentially have collected non-organic P from these suspended phases, influencing the Nt0t:Ptot ra tios.The convergence point in our sediment and diffusive flux data therefore provides a useful alternative estimate of the change in the N :P ratio of organic matter from its initial Redfield proportions during degradation in the water column, i.e. 16 -^ 27 (Fig. 5a).At Group 1 sites, this may surpass the magnitude of the change which subsequently occurs dur ing degradation in the sediments (27 -^3 2 -48), whereas for Group 2 sites the change after deposition (27 -> 48 -57) is greater than that in the water column.Lateral transport of organic material across the seafloor (Emeis et al., 2002;Jonsson et al., 1990) may allow older, partially degraded organic material to accumulate at deeper sites, potentially influenc ing the measured surface sediment NtoflPorg-Specifically, organic material experiencing degradation during downslope transport -thus in relatively oxic conditions compared to those in the deep basin accumulation zone -would be ex pected to acquire depleted NtoflPorg and C 0rg :P 0ig ratios, and therefore to dilute deep-basin surface sediment NtopPorg to wards lower values.This phenomenon could account for part of the scatter in the sediment data of Fig. 5a, most notably the relatively depleted N toflPoig and C 0rg :P 0ig ratios at F80 and BY15.However, the lack of such scatter in the N H ^/ H P O ^flux ratio (R 2 = 0.94 for all Group 2 sites) suggests that the laterally transported material is refractory once accumulated in the deep basins and does not contribute significantly to remineralization at these depths.

Implications of preferential P remineralization for basin-scale DIP during inflow-stagnation cycles
Due to preferential remineralization of P, Redfield-based modeling of organic matter breakdown in the Baltic may underestimate the magnitude of the direct remineralization flux to the total rate of P regeneration during stagnations, and hence the total pool of P in the water column.Simple budget calculations based on the modeled deep-basin car bon and phosphorus flows of Gustafsson and Stigebrandt (2007) Hille, 2005 andHille et al., 2005), it can be shown that pref erential remineralization of P from the buried fraction (48 %) of the organic matter adds 0.13-0.29 g m2 yr-1 P to the 0.32-0.39gm 2 yr_ 1 P released during complete remineralization of the other 52 %.This is of a similar order of magnitude to the mean value of Rpe -the "additional" amount of P re quired to balance observed water-column concentrationsduring five completely anoxic stagnations since 1965 in the Gotland Deep (Gustafsson and Stigebrandt, 2007).Rpe is as sumed to represent the dissolution of sedimentary Fe-P, thus the non-zero values during completely anoxic stagnations are unexpected, since no iron oxyhydroxides should be present in the surface sediments in such intervals.Lateral transport of sedimentary material downslope into the deep basins of the Baltic (e.g., Emeis et al., 2002) may potentially contribute P via dissolution of Fe-P transported across the redoxcline.While this flux remains poorly quantified, our calculated magnitude of the preferential remineralization flux is suf ficient to explain 0.13-0.29gm2yr_1 of the 0.30gm 2yr_1 "additional" P (Rpe) observed during anoxic stagnations (Ta ble 1).Considering the uncertainties involved, we may con clude that preferential remineralization accounts for a major proportion of Rpe during anoxic stagnations, with an addi tional contribution most likely derived from dissolution of laterally-transported Fe-P.These factors can also explain the scatter in the "Rpe vs. anoxic bottom increase" trend during variable redox stagnations (Fig. 10 in Gustafsson and Stige brandt 2007).
In light of the proposed mechanisms for accelerated ac cumulation of HPO^-in the deep waters, the relationship between total hypoxic area (A) and the total water column pool of HPO^-(DIP) proposed by Conley et al. (2002) may be re-examined.Although the period 1970-2000 investi gated by these authors encompasses several shorter multi annual inflow-stagnation cycles, a clear secular trend is ob served of contraction in the total hypoxic area from 1970 1993, followed by an expansion from 1994-2000, with the major inflow of 1993 acting as the switching point (Fig. 6 a and Conley et al., 2002).The extended dataset of 1962-2006 presented in Conley et al. (2009a) confirms that this pattern forms part of a multidecadal oscillation in hypoxic area in the Baltic since the middle of the 20th century.Con ley et al. (2002) showed that the net annual change in A (AA) and ADIP are positively correlated in the 1970-2000 dataset, due to net uptake or release of P by sedimentary iron oxy hydroxides from one year to the next.However, the rela tionship between the raw values of A and DIP for the full 1970-2000 dataset is less conclusive (Fig. 6 b), implying that the Fe-P interaction cannot fully explain the secular trends in DIP.When the data is divided into the intervals 1970-1993 and 1994-2000, it becomes clear that while the re lationship between A and DIP during the expansion phase (1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000) is indeed strongly positive, during the contrac tion phase   confirm the generally positive relationship between AA and ADIP identified by Conley et al. (2002), with 15 of the 22 annual steps displaying a positive gradient.However, the re sults clearly indicate that during contraction of the hypoxic area on the multidecadal timescale from 1970-1993, a net trend towards higher DIP is observed, despite the ongoing uptake of P by iron oxyhydroxides as a larger area of sur face sediments becomes oxic.This observation implies that during the contraction phase, release of HPOjj-into the wa ter column below the halocline, via direct remineralization from organic matter in the surface sediments and in the wa ter column itself, on balance outweighs the uptake of P by iron oxyhydroxides at shallower depths.Simply stated, al though the volume of hypoxic water shrinks, the total mass of HPOjj-in the water column increases.Both preferential remineralization of P from organic matter and dissolution of Fe-P transported laterally into the deep basins may con tribute to accelerated accumulation of HPOjj-.In contrast, the strong positive correlation between A and DIP in the ex pansion phase of 1994-2000 can be explained by the fact that both organic matter remineralization and the dissolution of surface-sediment iron oxyhydroxides along the basin mar gins act in unison to release HPOjj-into the water column as the hypoxic area expands.

Temporal variability in P burial with expansion and contraction of hypoxic area
The strong multidecadal oscillations in total hypoxic area may be expected to influence spatial patterns not only of P regeneration, but also of P burial, especially around the mar gins of the deep basins where redox conditions are most vari able.Mort et al. (2010) showed that organic P (P org) is the dominant burial phase of reactive (non-detrital) P throughout most of the modern Baltic, due to the poor preservation of Fe-P below the sediment surface and the lack of authigenic Ca-P formation within the upper 30 to 50 cm of the sediment column.To investigate the recent evolution of P org burial in response to changes in hypoxia, we focus on dated sediment records from three sites (LF1, LF1.5 and LF3, Fig. 1) along a depth transect into the Färö Deep in the central Baltic (de tails of the 210Pb age models are presented in Table Cl).A long archive of water column monitoring data is available for comparison from site F80 in the central Färö Deep, allow ing reconstructions of water column oxygen conditions from 1910 onwards.Conditions in the central basin do not trans late directly to those at equivalent depth along the basin mar gins, as shoaling of the halo/redoxcline in shallower waters dictates an offset towards lower oxygen concentrations at the shallower sites (see Fig. 16 in Al-Hamdani and Reker, 2007).However, the multidecadal trends in oxygen concentration in the 70-90 m depth range are expected to be qualitatively sim ilar throughout the basin, making such a comparison valid.
Water column oxygen concentrations at 70-90 m in the Färö Deep declined from 1910-1970, crossing the threshold into hypoxia (2 ml L-1) in the lower part of the depth range.From 1970-1993, during contraction of the hypoxic area (and thus deepening of the water column redoxcline), oxygen concentrations increased, before declining again during ex pansion of the hypoxic area from 1994-2000.The accumula tion rates of P org at LF1, LF1.5 and LF3 show a clear inverse co-variation with oxygen concentrations throughout the 2 0 th century, with maxima close to 1970 and minima close to 1992, confirming the observations of Mort et al. (2010) that Porg accumulation in Baltic Sea sediments increases with hy poxia.Although organic matter degradation within the upper sediments is expected to remove a portion of accumulated Porg prior to permanent burial, reactive transport modeling of organic matter degradation at these sites (unpublished re sults) suggest the influence of such breakdown at the current depth of the 1970-2000 interval is negligible, hence accumu lation may be considered equivalent to burial.Furthermore, the accumulation rate of C org, and sedimentary C org :P 0rg ra tios, show a roughly parallel evolution to that of P org accu mulation.This indicates that although enhanced preferential P remineralization occurred at these sites during intervals of lower oxygen concentrations (thus maximum hypoxic area), net P org burial nevertheless increased due to enhanced burial of its host phase, organic matter.
The effects of the multidecadal expansion and contraction of the hypoxic area on P regeneration and burial are summa rized in Fig. 8 , in which the Färö Deep is shown as represen tative for the deeper sections of the Baltic Proper.During in tervals of maximum hypoxic area such as the early 1970s, the halocline and redoxcline were comparatively shallow, with hypoxia occurring below ~7 0 m water depth in the central part of the basin.The zone of sediments along the basin mar gins containing Fe-P was therefore comparatively small, and LF1, LF1.5 and LF3 were all situated below the redoxcline.Hence, LF1, LF1.5 and LF3 in the early 1970s would qualify as Group 2 sites, characterized by the absence of Fe-P at the sediment surface, a high net rate of Porg burial, and enhanced preferential remineralization of P relative to C. Conversely by 1992, the halocline and redoxcline were displaced much deeper in the water column, qualifying LF1, LF1.5 and LF3 as Group 1 sites with Fe-P seasonally present in the surface sediments, a comparatively low net rate of Porg burial, and limited preferential remineralization of P relative to C. How ever, despite the deeper redoxcline, the long stagnation of the bottom waters in the years leading up to 1992 generated a comparatively large total pool of water column DIP in the deepest part of the basin.

Wider implications for P-driven eutrophication/ hypoxia feedbacks and basin-scale biogeochemical modeling
Preferential remineralization of P from organic matter in the water column and sediments, and the consequent accumu lation of HPO^-below the halocline, clearly exert a strong control on the basinwide pool of DIP in the Baltic.However, since the total DIP pool largely consists of DIP stored in the deep waters, there remains much debate over its influence on P availability in the photic zone, and hence the efficiency of the positive feedback between P regeneration, eutrophication and sustained hypoxia.Upwards transport of HPO^-across the halocline is predicted by the principles of turbulent dif fusion (Gustafsson and Stigebrandt, 2007) and has been cal culated to account for 60 % of the annual replenishment of HPO^-to surface waters (Vahtera et al., 2007).However, recent evidence suggests that up to 40-60 % of the flux esti mated from the [HPO^-] gradient may in fact be trapped and recycled by particulate iron and manganese oxyhydroxides in the water column (Turnewitsch and Pohl, 2010).This pro cess may contribute to the low correlation observed between winter HPO^-concentrations above and below the halocline at sites in the open Baltic (Liiover and Stips, 2008).Further more, the physical mechanisms determining P transport to the surface, ie.wind-driven mixing in winter (Janssen et al., 2004) and summer upwelling along coastal margins (Nausch et al., 2009) are sensitive to climatic parameters and thus vary on multiannual to multidecadal timescales.
Changes in the spatial distribution of the deep-water DIP pool (Fig. 8 ) may influence its accessibility to vertical trans port processes such as mixing and upwelling, and thus may be equally as important as the size of the pool itself in de termining the total rate of P supply to the surface waters.A major challenge for future biogeochemical modeling of the Baltic is thus to unravel how the two contrasting -and nat urally forced -states presented in Fig. 8 differ in their ca pability to sustain eutrophication and hypoxia.Especially with regards to the planning of hypoxia remediation strate gies (Conley et al., 2009b), knowledge of the expected re sponse of the system to changes in total hypoxic area is cru cial.The enhanced C org burial observed along the deep basin margins during intervals of maximum hypoxic area (Figs. 7,   8 ) is likely a combined signal, both of deteriorating redox conditions and of higher export productivity, analogous to the mechanisms by which Mediterranean sapropel sediments are formed (e.g., de Lange et al., 2008).If so, intervals of maxi mum hypoxic area such as the early 1970s may well be char acterized by a more efficient P regeneration-eutrophicationhypoxia feedback loop than intervals of minimum hypoxic area, with all fluxes in the P cycle accelerating despite the smaller total DIP pool in the water column.This would im ply that the accessibility of the DIP pool (which is greater due to its shoaling during intervals of maximum hypoxic area) www.biogeosciences.net/8/1699/2011/Biogeosciences, 8, 1699-1720, 2011 indeed outweighs the size of the pool in determining the net upwards flux of P to the surface waters.

Conclusions
A comprehensive coring survey has facilitated an unprece dented investigation of the controls on regeneration and burial of P throughout the Baltic Sea.The network of pore water data provide a snapshot of conditions during the cur rent stagnation interval (2003-present), and reveal a strong depth-dependence of the processes controlling regeneration of P from the sediments.Sites of <90 m water depth cur rently display evidence for the interaction between P and iron oxhydroxides in the surface sediments, confirming the earlier work of Mort et al. (2010).Seasonal variability in bottom-water redox conditions at such sites results in a strong annual cycle in the efflux of P, with maximum release during the most reducing months when iron oxhydroxides reductively dissolve.In contrast, sites below 90 m display a continuous efflux of P, reflecting degradation of sedimen tary organic matter.Moreover, the rate of P regeneration (relative to C and N) from degrading organic matter is con trolled by the ambient redox conditions, and hence is itself depth-dependent, with maximum preferential P remineralization occurring in the deep basins.Such preferential P remineralization leads to depletion of the diffusive flux ra tio of NH^"/HPO^-, and enrichment of sedimentary Ntot:P0rg and Corg :Porg, with increasing water depth.The data sug gest a roughly two fold enrichment of the Ntot:Porg ratio of organic matter from its original Redfield proportions during descent through the water column, followed by further en richment after accumulation at the sediment-water interface.These processes lead to a substantial build-up of excess P in the water column, which is underestimated by existing Redfield-based modeling of biogeochemical cycles during previous stagnation intervals.Furthermore, the magnitude of preferential remineralization of P in the deep basins was great enough to generate a net increase in the total water col umn pool of DIP during contraction of the hypoxic area from 1970-1993, despite the large scale sequestration of P into iron oxyhydroxides at shallower depths during this time.Burial of reactive P in the Baltic takes place principally as organic P, and the rate of organic P burial along the margins of the deep basins is redox dependent, with higher burial dur ing intervals of more reducing conditions.Expansion of the total hypoxic area on the multidecadal timescale thus leads to expansion of the area of enhanced burial of organic P, despite a simultaneous increase in the rate of preferential remineralization of P from organic matter and the large-scale dissolu tion of P associated with iron oxyhydroxides.This may im ply an accelerated P cycle during intervals of maximum hy poxic area, in which the rates of P incorporation into organic matter, regeneration from and burial within the sediments are all enhanced.Such a scenario may be supported physically by efficient upwards transport of deep water HPO^-across the shallow halocline present during these intervals.C oncentrations o f porew ater species in p m o lL -1 (R = sam pled w ith rhizons) and bottom w ater oxygen concen trations in m lL -1 .

Redoxcline
S ub-zero oxygen concentrations are e s tim ated from hydrogen sulfide concentrations using the ratio 1 m o lF l2S = 2 m o l 0 2 (see Fonselius, 1981)

Fig. 1 .
Fig. 1.B athym etric m ap o f the Baltic Proper, show ing the locations o f the coring sites used in this study.Circles represent G roup 1 sites ( < 9 0 m w ater depth), squares represent G roup 2 sites ( > 9 0 m w ater depth).B athym etric and coastline data are presented in equidistant cylindrical projection, taken from the G eneral B athym etric C hart o f the O ceans (GEBCO) D igital A tlas (Intergovernm ental O ceanographic C om m ission et al., 2003).
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I

Fig. 2 .
Fig. 2. D ata array o f sites used in this study, arranged by increasing w ater depth from left to right.U pper panel: W ater colum n salinity and oxygen profiles.A ll profiles w ere m easured during the H Y P E R /C O M B IN E cruise o f the R/V Aranda in M ay/June 2009, except site 1.BY2

Fig. 3 .
Fig. 3. (a) R atio o f diffusive fluxes o f NH^"/H P O ^ across the sedim ent-w ater interface vs. w ater depth for all sites, (b) C oefficient o f determ ination (R 2) for linear regression betw een porew ater profiles o f NH^" and H P O ^ in the upperm ost 10 cm o f each core.L ow R 2 indicates deviation o f H P O ^ profile from asym ptotic shape, caused by dissolution o f Fe-P in surface sedim ents.V ertical dashed line represents perfect correlation (R 2 = 1).(c) D iffusive flux o f Fe2+ vs. w ater depth for all sites.N egative = efflux.
Fig. 5. (a) N H + /H P O J diffusive flux ratio at ali G roup 2 sites (left trendline and data), and sedim entary Ntot/Porg ratio in the upper 2 cm o f sedim ent at sites selected for solid phase analyses (from both G roup 1 and 2, right trendline and data), plotted against w ater depth.Vertical dashed lines indicate Redfield ratio (N:P = 16) and the intersection o f the tw o trendlines (N:P = 27).A rrow s represent inferred progression o f N:P ratio o f organic m atter during descent through the w ater colum n ("O M (w /c)"), from 16 -2 7 , and in the surface sedim ents (OM (sed)), from 27 -a higher value, w hose m agnitude is determ ined by w ater depth-dependent preferential rem ineralization o f P. (b) S edim entary Corg/Porg ratio in the upper 2 cm o f sedim ent at the sam e sites plotted for Ntot/Porg in (a).
Sum m ary o f carbon and phosphorus fluxes in deep basins o f the Baltic.lc u la te d fro m flu x es in b u d g e t o f G u sta fsso n a n d S tig e b ra n d t et al. (2 0 0 7 ).** E stim a te d fro m to ta l P a c c u m u la tio n ra te s o f H ille et al. (2005) a n d m ea n Porg: to ta l P ratio o f to p 2 c m o f F80, B Y 1 5 a n d L L 1 9 (0 .3 6 , T ab le B Í ) .

Fig. 6 .
Fig. 6.(a) T im e series o f total hypoxic area o f the Baltic ("A ", square sym bols), i.e. the area o f sedim ents in contact w ith bottom w aters o f < 2 m lL -1 O 2 , and the total pool o f dissolved inorganic phosphate (DIP, circles).B oth series averaged for January-M arch for each y e ar from 1 9 70-2000 (data from Table 1 in C onley et al., 2002).O pen sym bols indicate 1 9 70-1993 "co n tractio n " o f hypoxic area, closed sym bols represent 1994-2000 "exp an sio n " phase, (b) C ross-plot o f the sam e data.T rend for com plete dataset show n as a fine solid line.D ata points for 1970-1993 (open circles) are jo in e d chronologically w ith a dashed line.N ote the frequently positive relationship betw een annual changes in A and DIP but the overall negative trend for this interval (dashed trendline).D ata points for 1 9 94-2000 (filled circles) are jo in e d chronologically w ith a solid line (heavy solid trendline represents 1994-2000 trend).

Fig. 7 .
Fig. 7. T im e series o f Corg and Porg accum ulation rates, and Corg/Porg, at sites LF1, LF1.5 and LF3 on the m argins o f the Färö Deep in the central B altic, and relevant w ater colum n oxygen data.Corg and Porg accum ulation rates w ere calculated from 2 1 0 Pb-derived m ass accum ulation rates (estim ated using a C onstant Rate o f 210Pb S upply (CRS) m odel (Boer et al., 2006) and Corg and Porg contents).Total hypoxic area data from C onley et al. (2009) and w ater colum n oxygen concentrations at site F80 from the B altic Sea E nvironm ental D atabase at S tockholm U niversity are plotted as 5 year binned m ean values.
oxic degradation releases P from organic matter, to be trapped as Fe-P in surface sediments High: inefficient anoxic degradation of organic matter Low: despite efficient degradation of organic matter, oxic degradation pathways favor retention of P relative to C org orgHigh: anoxic degradation pathways favor release of P from sediments

Fig. 8 .
Fig. 8. (a) W ater colum n oxygen concentrations m easured during the R/V Aranda cruise o f M ay/June 2009 at sites LF1, LF3 and F80 along a depth transect into the Färö D eep betw een the islands o f G otland and Saarem aa, show ing the vertical offset o f the redoxcline betw een central (F80) and m arginal (LF1) sites, (b ) Scaled cross-sectional sketches through Färö D eep in 1971 and 1992 (sim plified bathym etry from D ata A ssim ilation System o f B altic E nvironm ental D atabase, Stockholm U niversity, depth scale as for (a)), show ing w ater colum n salinity and oxygen profiles from F80 (also know n as BY20, data from Baltic E nvironm ental D atabase) in the m onth o f February (chosen for com parable data availability).W ith the exception o f F80, sites are positioned accurately by depth.The sim plified bathym etry aliases the true depth o f F80 (191 m, Fig. 2) and generates m inor lateral offsets from true position o f all sites.D ashed lines show approxim ate position o f hypoxia threshold (O2 = 2 m l F -1 ) based on profiles at F80 and assum ption o f 15 m vertical offset o f redoxcline betw een F80 and FF1 o_ (see (a)).Shading in w ater colum n below halo/redoxcline qualitatively represents deep w ater [H P04 ]. S edim ent zones are divided based on the assum ption that iron oxyhydroxides are absent from surface sedim ents below a threshold o f oxygen stress w hich m ay be spatially variable but generally deeper than that o f bottom w ater hypoxia, (c)Table sum m arizing characteristics o f sedim ent zones w ith respect to the ............... .........................
Table BÍ).The shape of the N t0t:Porg and C org:P 0rg vs. water depth relation ships most likely reflects the change in water column oxygen with depth, with a higher rate of change per vertical meter in the Group 1 sites where oxygen decreases rapidly close to the halocline.Although reported as Corg: total P (Ptot) and not explicitly stated by the authors, a depth-dependence in surface sediment C :P ratios is also seen in the data of Emeis et al. (2 0 0 0 ), with highest values at three stations of > 2 0 0 m depth in the Gotland Deep.Our maximum C org:Porg values of 500-650 indicate a ~5 -6 fold sedimentary enrichment of C relative to P from the Redfield ratio of 106:1, which is comparable to the values in Mediterranean sapropels(300- 1085(300-  , Kraal et al., 2010;; Slomp et al., 2002), modem anoxie Black Sea sediments (mean value 558, Van Cappellen andIngall, 1997)and Arabian Sea oxygen minimum zone sedi ments (400-800,Schenau and de Lange, 2001).
support this claim (Table1).The model ofGustafsson  and Stigebrandt (2007)calculates a mean burial efficiency of 48 % of all Corg passing through the 150 m water depth hori zon in the Gotland Deep.Using this value, and measured deep-basin Corg and Porg accumulation rates (from this study,

Table A l.
.

Table A l
. C ontinued.

Table A l
. C ontinued.

Table A l
. C ontinued.

Table C l
. S edim ent solid phase param eters used in the calculation o f 2 1 0 Pb-based age m odels.