Build-up and decline of organic matter during PeECE III

Increasing atmospheric carbon dioxide (CO2) concentrations due to anthropogenic fossil fuel combustion are currently changing the ocean's chemistry. Increasing oceanic [CO2] and consequently decreasing seawater pH have the potential to significantly impact marine life. Here we describe and analyze the build-up and decline of a natural phytoplankton bloom initiated during the 2005 mesocosm Pelagic Ecosystem CO2 Enrichment study (PeECE III). The draw-down of inorganic nutrients in the upper surface layer of the mesocosms was reflected by a concomitant increase of organic matter until day t11, the peak of the bloom. From then on, biomass standing stocks steadily decreased as more and more particulate organic matter was lost into the deeper layer of the mesocosms. We show that organic carbon export to the deeper layer was significantly enhanced at elevated CO2. This phenomenon might have impacted organic matter remineralization leading to decreased oxygen concentrations in the deeper layer of the high CO2 mesocosms as indicated by deep water ammonium concentrations. This would have important implications for our understanding of pelagic ecosystem functioning and future carbon cycling.


Introduction
Human activities such as the massive combustion of fossil fuels perturb the natural carbon cycle by increasing atmospheric carbon dioxide [CO 2 ] concentrations. Since the beginning of the industrial revolution atmospheric CO 2 levels have increased al- 15 ready from about 280 to 380 µatm, in the year 2006 and are projected to reach values as high as 700 µatm, by the end of the 21. century (IPCC, 2001). Together with changes in other climate relevant gases this substantial increase in atmospheric CO 2 drives global climate change: increasing global mean temperatures, rising sea level and accelerating extreme weather incidences, to name just those events most likely Introduction nine polyethylene bags (see Fig. 1 for a graphical representation), 2 m in diameter and extending about 9.5 m into depth, were deployed at a free floating raft, operated by the University of Bergen. The bags were filled simultaneously with unfiltered nutrient-poor fjord water, about 27 m 3 each, pumped from 13.5 m depth. Three different CO 2 levels Introduction EGU uously (15 min averages) with a Licor light sensor  connected to a datalogger (LI-1400) mounted on the raft (see Fig. 4). On some days this photon flux density (PFD) was also measured within the headspace and at 2.5 m depth of mesocosm M8. The headspace within the tents was continuously flushed with air at target CO 2 values throughout the experiment, simulating a 1x, 2x and 3x CO 2 atmosphere. 5 The phytoplankton bloom was initiated on day t −1 by additions of nitrate and phosphate to upper surface waters aiming for final surface water concentrations of 16 and 0.8 µmol kg −1 , respectively (see Fig. 5a, b). However, in mesocosms 1,2,3 and 4 these concentrations were lower than expected as some of the nutrients added must have been lost to deep waters during mixing. Therefore, additional nitrate and phosphate 10 was added to these mesocosms on day t 1 , establishing uniform nutrient concentrations within all mesocosms on day 2 (see Fig. 5). As pCO 2 in mesocosms 1-6 were below intended values, NaHCO 3 and equimolar amounts of HCl were added, reaching final concentrations of ∼25 and ∼18 µmol kg −1 in mesocosms 1-3 and 4-6, respectively. Regarding the changes in the CO 2 system this procedure is equivalent to 15 CO 2 aeration and increased pCO 2 in mesocosms 1-6 to desired values of ∼700 and ∼1050 µatm on day 2 (see Fig. 6).
Daily sampling was at 10:00 a.m. starting on day t 0 . Dissolved and particulate components (with the exception of DIC, alkalinity, pCO 2 , and other gases) were sampled with a 5 m long, 6 cm diameter tube lowered into the mesocosms, providing a depth-20 integrated sample. The samples representative for upper surface waters (∼20 L per mesocosm per day) were brought back to shore where sub-samples for various measurements and analyzes were taken immediately upon arrival. Every other day water from sediment traps, 60×10 cm Plexiglas cylinders deployed in each mesocosm at 7.5 m depth, was also collected. 25

Measurements and analyzes
In principle all measurements and analyzes of parameters presented in this article followed standard procedures. A complete list of all parameters measured together with 4543 Introduction EGU their individual methods is given in Table 1. For details on measurement methods of parameters not presented in this article the reader is referred to the respective reference given in Table 1. The nutrients nitrate, nitrite and phosphate were determined from GF/F filtered, silicate from 3 µm cellulose acetate filtered sample water and measured with an autoan-5 alyzer (AA II) according to Hansen and Koroleff (1999). Ammonium was determined following the approach described in Holmes et al. (1999).
For phytoplankton pigment analysis 250-500 ml of sample water were filtered onto GF/F filters which were frozen at −20 • C until extraction and analysis. For extraction filters were homogenized together with 2 and 4 mm glass beads in 2 ml 100% acetone by 10 a cooled Vibrogen cell mill for 5 min. Then the samples were centrifuged at ∼3000 x g and −10 • C for 10 min. The supernatant was 0.2 µm filtered and analyzed by rp-HPLC (reverse-phase high performance liquid chromatography) where pigments are identified by comparing retention times and absorption spectra, obtained with a diode array spectrophotometer (WATERS), with those of pigment standards. Commercially avail- 15 able pigments were also used as standards for calibration. All sample handling during the extraction and analysis procedures were carried out under dimmed light conditions to prevent pigment photo-oxidation. Calculation of the phytoplankton community composition and individual phytoplankton Chla contributions (see Fig. 7) was carried out with the program CHEMTAX (Mackey et al., 1996). 20 For determination of the particulate compounds TPC (total particulate carbon), POC (particulate organic carbon), PIC (particulate inorganic carbon), PON (particulate organic nitrogen), POP (particulate organic phosphorus) 250-500 ml of sample water were filtered onto precombusted (450 • C for 5 h) GF/F filters and stored at −20 • C. TPC, POC, and PON were determined on an EuroVector elemental analyzer (EuroEA 3000). 25 Before analysis all filters were dried for 6 h at 60 • C, while POC filters had been fumed overnight with concentrated HCl to remove PIC. Particulate inorganic carbon was then calculated as the difference between TPC and POC (see Fig. 8 EGU described by Hansen and Koroleff (1999). Every other day these measurements were also performed on 200-250 ml sub-samples collected from all nine sediment traps. Determination of the dissolved compounds DOC (dissolved organic carbon), DON (dissolved organic nitrogen) and DOP (dissolved organic phosphorus) was on GF/F filtered fjord water samples. DOC and DON were analyzed by means of high tem-5 perature catalytic oxidation (HTCO) as described in Qian and Mopper (1996), DOP by potassium persulphate oxidation and subsequent dissolved inorganic phosphorus determination described in Hansen and Koroleff (1999) (see Fig. 8).

Mesocosm performance
10 Unlike in a previous mesocosm CO 2 perturbation experiment (Engel et al., 2005), deep water was separated from the upper mixed surface by establishing a halocline. This allowed particulate organic matter produced in the nutrient-enriched surface waters to settle into depth, simulating export production. Detailed monitoring of the location and stability of the halocline showed that until day t 12 deep water loss (mixing of deep 15 waters into the surface waters and deepening of the halocline) was only moderate (∼8-21%) in most mesocosms, with the exception of mesocosm M6 (∼36%). Unusually stormy weather conditions around day t 12 , however, enhanced deep water mixing, and until the end of the experiment a substantial portion of the deep water was lost in all mesocosms (∼29-42%) with up to 58% and 50% in mesocosms M3 and M6, 20 respectively (Fig. 2).
Temperatures within the mesocosms, initially evenly distributed throughout depth at about 9 • C, increased to ∼11.5 • C in upper surface and ∼11 • C in deep waters until day t 19 (Fig. 3). Towards the end of the experiment on day t 25  EGU 90 and 560 µmol m −2 s −1 . These fluctuations corresponded well with those measured in solar radiation at the University of Bergen (Fig. 4). About 80% of incident PAR passed through the gas tight tents into the mesocosms' headspace of which up to 15% penetrated into ∼2.5 m depth, the center of the mixed surface layer (data not shown).
Having established desired pCO 2 values within the mesocosms' surface waters of 5 ∼350, 700 and 1050 µatm at day t 2 , CO 2 partial pressure continuously decreased until day t 12 to ∼200, 350 and 460 µatm in the 1x, 2x and 3x mesocosms, respectively. After a slight increase on day t 13 and t 14 pCO 2 remained rather constant until the end of the experiment (Fig. 6). All sediment traps, sampled every other day, showed significant over-trapping. In other words more particulate material was calculated to be exported into the deep than inorganic nutrients removed from surface waters. This was probably caused by continuous flapping of the mesocosm bags mainly forced by wind stress which extended into depth and resuspension of sedimented particulate matter. Hence, rather than absolute concentrations only element ratios (POC/PON, POC/POP and PON/POP) are reported 15 in this study (Fig. 11).

Inorganic nutrient uptake and organic material build-up
Following the addition of nitrate and phosphate on day t −1 and their adjustment on day t 1 to 15 and 0.7 µmol kg −1 in all mesocosms, respectively, inorganic nutrients (nitrate, phosphate and silicate) continuously decreased (Fig. 5). In all mesocosms silicate was 20 the first to get depleted about day t 7 (Fig. 5c), followed by phosphate about day t 10 (Fig. 5b) and nitrate about day t 13 (Fig. 5a). From their day of depletion onwards all inorganic nutrient concentrations remained comparatively low. An exception were ammonium concentrations which were close to zero until day t 12 and increased up to ∼0.5, 0.75 and 0.9 µmol kg −1 until day t 20 in the 1x, 2x and 3x CO 2 mesocosms, respectively ( Fig. 5d). Corresponding deep water values also increased to up to 3 µmol kg −1 towards the end of the experiment, with apparently higher values in 3x than in 2x and 2x EGU than 1x CO 2 mesocosms (Fig. 5e). The draw-down in inorganic nutrients was reflected by a concomitant build-up of organic matter. Total Chla concentrations increased from about 1.5µg l −1 on day t 0 to ∼11.0, 13.5 and 13.0 µg l −1 on day t 10 in the 1x, 2x and 3x CO 2 mesocosms, respectively (Fig. 7a). From then on Chla continuously decreased in all mesocosms, reaching 5 again starting concentrations around day t 16 . Separating total Chla concentrations into individual contributions by selected phytoplankton taxa showed that diatoms (mostly Skeletonema sp. and Nitzschia sp.) and prymnesiophytes (including Emiliania huxleyi) were of major importance, both reaching up to ∼5 µg l −1 in all mesocosms. However, diatom associated Chla peaked about day t 9 while that of prymnesiophytes coincided 10 with the total Chla maximum on day t 10 (Fig. 7b, c). During and after the decline of the phytoplankton bloom, from day t 10 onwards, prasinophyte, cyanobacteria and dinoflagellate Chla concentrations were slightly increasing in all mesocosms ( Fig. 7d, e, f). The build-up of particulate organic carbon, nitrogen and phosphorus was very simi-  (Fig. 8a, b, c). Due to CaCO 3 production of coccolithophores during the phytoplankton bloom (compare Fig. 8c), also about 15 µmol l −1 of PIC was built-up until the peak of the bloom, however almost all CaCO 3 produced was lost from the upper mixed layer towards the end of the experiment (Fig. 8d). On the other hand, concentrations of all 25 dissolved organic compounds increased steadily from day t 0 to day t 24 in all mesocosms, DOC from about 75 to 100 µmol l −1 , DON from about 2.5 to 9.5 µmol l −1 and DOP from about 0.2 to 0.25 µmol l −1 (Fig. 8e, f, g).
Finally, in all mesocosms metazooplankton biomass increased from an average EGU of about 0.7 µmol carbon per liter at the beginning of the experiment to about 3.1-3.5 µmol l −1 towards the end (Fig. 9).

Element ratios of particulate and dissolved organic matter
Individual build-up and removal of both, particulate and dissolved organic matter in the mixed surface waters lead to specific temporal variations in the respective ele-5 ment ratios. The ratio of carbon to nitrogen to phosphorus of particulate matter accumulating in the upper 5.7 m of all mesocosms was close to Redfield with 106/16/1 (POC/PON/POP) until about day t 9 . From then on only POC/PON remained close to 6.625 while POC/POP and PON/POP increased about 2-fold to values of ∼200 and 30, respectively ( Fig. 10a, b, c). This can be the result of preferential POP removal from 10 the upper surface waters. The individual element ratios of the dissolved organic matter build-up during the experiment, however, did not seem to change with time but significantly deviated from Redfield in all mesocosms. DOC/DON was relatively low with about 3, while DOC/DOP and DON/DOP were relatively high with values around 300 and 100, respectively (Fig. 10d, e, f). About 2x lower DOC/DON, 3x higher DOC/DOP 15 and 5x higher DON/DOP in comparison to the corresponding Redfield ratios can be the result of either relatively higher DOP than DOC than DON removal or relatively lower DOP than DOC than DON build-up within the upper surface waters. The ratios of carbon to nitrogen to phosphorus of the particulate organic material collected in the sediment traps reflect processes of accumulation and remineralization. 20 In all mesocosms sedimentary POC/PON was close to Redfield ratios throughout the experiment (Fig. 11) as were the corresponding values in the upper surface waters (compare Fig. 10a, b, c). Sedimentary POC/POP and PON/POP increased after about 14 days to higher ratios, however, apparently higher in the 3x than 2x than 1x CO 2 mesocosms. Introduction

Mesocosm performance
The 2005 PeECE III mesocosm experiment differs in several aspects but one substantial from many previous studies, the establishment of a halocline. This was thought to create two distinct water masses in which pelagic key processes such as primary pro-5 duction (surface waters) and remineralization (deep waters) are separated. Any water exchange between these two realms can make data interpretation difficult as most efforts were directed towards sampling in the upper surface waters. Hence, the stability and location of the halocline was monitored on a daily basis. However, even without the storm event on day t 12 , substantial deep water mixing into the upper surface would have occured. This could have been caused by wind force acting on the mesocosm bags being transmitted into depth. This would also explain the observed over-trapping of the sediment traps caused by significant resuspension of settled particulate organic matter. Complete element budgets and process assessment of primary production and remineralization therefore require sampling in both water masses. Sedimentation pro- 15 cesses on the other hand appear to be difficult to evaluate with this kind of mesocosm setup.

Inorganic nutrient uptake and organic material build-up
The draw-down of the major nutrients nitrate, phosphate and silicate during the bloom phase of the experiment followed the same temporal pattern in all mesocosms, irre-20 spective of pCO 2 (for details see Bellerby et al., 2007). The concomitant build-up of organic material, reflected by increasing Chla concentrations showed also no treatment effect, although at the peak of the bloom total and diatom associated Chla concentrations were seemingly higher in the 2x and 3x compared to the 1x CO 2 mesocosms. The build-up of particulate organic matter (POC, PON, POP), however, appeared slightly EGU in dissolved organic material (DOC, DON, DOP) on the other hand, was again not significantly different between CO 2 treatments. Also the build-up of metazooplankton (>90 µm) biomass was not significantly different between treatments. In summary, the temporal development of nutrient uptake, build-up and decline of particulate organic matter and increase of dissolved organic matter during the experiment followed the 5 same patterns in all mesocosms, irrespective of CO 2 condition (compare Engel et al., 2005). The similarity of nutrient uptake and corresponding organic matter build-up between treatments could be caused by phytoplankton nutrient utilization (nitrate and phosphate) and division rates which are insensitive to changes in seawater pCO 2 . Indeed, although there is a critical CO 2 concentration for all phytoplankton species below 10 which their growth decreases, in the CO 2 range of the present study (350-1050 µatm) division rates of diatoms, coccolithophores and cyanobacteria have been found to be rather constant (Burkhardt et al., 1999a;Rost et al., 2002;Langer et al., 2006;Barcelos e Ramos et al., 2007). Changes in phytoplankton C/N and C/P stoichiometry due to variations of seawater pCO 2 would then have to explained by changes in inorganic 15 carbon utilization (see below). Another interesting finding is that deep water ammonium concentrations towards the end of the experiment increased to lower values in the 3x than in the 2x and than in the 1x CO 2 mesocosms (corresponding surface layer values show the same trend and increased from day t 12 onwards due to deep water mixing). This can be explained by dif-20 ferences in remineralization rates of organic nitrogen compounds at depth. Ammonium regeneration in sediments appears to be oxygen dependent with steadily decreasing rates at decreasing oxygen concentrations Chapelle (1995); Serpa et al. (2007). But what could be the mechanisms responsible for reduced deep water oxygen concentrations at elevated CO 2 ? Although standing stocks of dissolved and particulate organic 25 carbon in the upper surface waters were quite similar in all mesocosms, significantly more dissolved inorganic carbon was utilized in the 3x than in the 2x and than in the 1x CO 2 mesocosms (for details see Riebesell et al. (2007) and Bellerby et al., 2007). This implies that considerably more organic carbon must have been lost into deep wa- EGU ters, about twice as much in the 3x compared to the 1x CO 2 mesocosms. The process operating could be enhanced TEP formation which has been shown to be higher at elevated CO 2 (Engel , 2002), leading to intensified particle coagulation and hence organic carbon export from the surface to depth. Intensified organic carbon sedimentation at higher CO 2 levels could have fueled organic carbon remineralization by heterotrophic 5 bacteria, leading to stronger oxygen reduction at depth in the 3x than in the 2x and than in the 1x CO 2 mesocosms. If only half of the additional organic carbon exported in the 3x CO 2 mesocosms would have been remineralized in the lower 1.5 m of of the mesocosms, oxygen would have been considerably stronger depleted at depth in comparison to the 2x and 1x CO 2 mesocosms, by about 100 µmol kg −1 at the end of the 10 experiment. This could explain reduced ammonia regeneration at elevated CO 2 .

Element ratios of particulate and dissolved organic matter
While in the surface layer element ratios of particulate and dissolved organic matter reflect processes of production and removal, corresponding ratios in the deep water of the mesocosms are indicative for processes of accumulation and remineralization. 15 Until day t 9 element ratios of particulate organic matter were close to corresponding Redfield ratios. As from then on inorganic phosphorus was depleted while nitrate and dissolved inorganic carbon could still be assimilated, POC/POP and PON/POP steadily increased. The same trend was found for dissolved organic matter, although due to relatively high standard deviations statistically not significant. Interestingly, however, 20 DOC/DON was considerably lower than the Redfield ratio , indicating preferential loss of dissolved organic carbon compared to dissolved organic nitrogen in all mesocosms. This can be the result of TEP formation, favoring preferential DOC export from the surface waters to depth (compare Sect. 5.2). In this context it is also important to note that the loss of organic matter from the surface, defined as the difference between 25 organic matter build-up and the inorganic counterparts drawn-down, seems to be a process in operation right from the start of the experiment (data not shown but compare Riebesell et al., 2007

EGU
Balancing the processes of accumulation and remineralization at depth is difficult because of the problems associated with sediment trap sampling and the lack of dissolved organic matter data in deep waters. However, the scenario described above of 1) increased loss of organic carbon from surface waters to depth at elevated CO 2 , resulting in 2) increased respiration of this organic carbon and hence increased oxygen 5 consumption at depth, and hence in 3) reduced organic nitrogen remineralization (ammonium regeneration) is not at odds with the observed patterns in sedimentary organic matter ratios. Constant sedimentary POC/PON, for instance, could indicate increased organic carbon remineralization at elevated CO 2 . Furthermore, increased sedimentary PON/POP in the 3x and 2x in comparison to the 1x CO 2 mesocosms could be the 10 result of decreased PON remineralization.

The pelagic ocean in a high CO 2 world: indications from mesocosm studies
Bulk parameters of natural phytoplankton blooms, especially those representative for biomass such as POC, PON, POP, Chla, but also DOC, DON and DOP appear to a great extent insensitive to changing CO 2 conditions (compare Engel et al., 2005). Also 15 corresponding elemental ratios of these standing stocks appear largely unaffected by seawater pCO 2 . However, in this study there were pronounced treatment differences in the calculated amount of organic carbon lost from upper surface waters, with higher values in the 3x compared to the 2x and 1x CO 2 mesocosms (compare Riebesell et al., 2007). This was fueled by significantly increased dissolved inorganic carbon (DIC) to 20 inorganic nutrient (nitrate, phosphate) uptake ratios at elevated CO 2 (compare Bellerby et al., 2007). Hence, elemental ratios of organic matter exported from surface to depth must have been enriched in organic carbon compared to nitrogen and phosphorus under elevated CO 2 conditions. Furthermore, this "extra" organic carbon could have resulted in higher bacterial respiration and oxygen consumption at depth, reducing 25 ammonium regeneration in the 3x compared to the 2x and to the 1x CO 2 mesocosms.
In summary, mesocosm studies have identified various processes in pelagic ecosystems potentially sensitive to increasing atmospheric CO 2  EGU et al., 2007 1 ): 1) higher inorganic carbon consumption in the euphotic zone, 2) higher export production, 3) reduced ammonium regeneration at depth, 4) increases in climate relevant trace gas production (Wingenter et al., 2007), 5) reduced CaCO 3 production by coccolithophores such as Emiliania huxleyi (Delille et al., 2005) and 6) diminished growth and survival rates of calcifying molluscs (Antia et al., 2007). All these processes 5 are part of feedback mechanisms impacting CO 2 partitioning between atmosphere and ocean. However, whether changes in these processes will dampen or enhance global climate change is still impossible to forecast.   BGD 4,2007 Build-up and decline of organic matter during PeECE III