The copyright line for this article was changed on 13 APR 2015 after original online publication.
 A closed eddy core in the Subantarctic Atlantic Ocean was fertilized twice with two tons of iron (as FeSO4), and the 300 km2 fertilized patch was studied for 39 days to test whether fertilization enhances downward particle flux into the deep ocean. Chlorophyll a and primary productivity doubled after fertilization, and photosynthetic quantum yield (FV/FM) increased from 0.33 to ≥0.40. Silicic acid (<2 µmol L−1) limited diatoms, which contributed <10% of phytoplankton biomass. Copepods exerted high grazing pressure. This is the first study of particle flux out of an artificially fertilized bloom with very low diatom biomass. Net community production (NCP) inside the patch, estimated from O2:Ar ratios, averaged 21 mmol POC m−2 d−1, probably ±20%. 234Th profiles implied constant export of ~6.3 mmol POC m−2 d−1 in the patch, similar to unfertilized waters. The difference between NCP and 234Th-derived export partly accumulated in the mixed layer and was partly remineralized between the mixed layer and 100 m. Neutrally buoyant sediment traps at 200 and 450 m inside and outside the patch caught mostly <1.1 mmol POC m−2 d−1, predominantly of fecal origin; flux did not increase upon fertilization. Our data thus indicate intense flux attenuation between 100 and 200 m, and probably between the mixed layer and 100 m. We attribute the lack of fertilization-induced export to silicon limitation of diatoms and reprocessing of sinking particles by detritus feeders. Our data are consistent with the view that nitrate-rich but silicate-deficient waters are not poised for enhanced particle export upon iron addition.
 Iron limits primary productivity across large areas of the oceans, which hence contain perennially high NO3− and PO43− stocks, but low chlorophyll a [Boyd et al., 2007]. Artificial Fe fertilization experiments (FeAXs) in these regions have induced blooms of large-celled diatoms, drawdown of macronutrients and fCO2 in the surface mixed layer, and enhanced downward particle flux [Boyd et al., 2007; Coale et al., 2004; de Baar et al., 2005; Smetacek et al., 2012]. Analogous results are found in naturally iron-rich waters downstream of Southern Ocean islands [Blain et al., 2007; Pollard et al., 2009]. Ocean Fe fertilization has hence been proposed as a means to sequester CO2 by increasing the downward flux of particulate organic carbon (POC), although the viability and side effects remain unclear [Aumont and Bopp, 2006; Lampitt et al., 2008a; Lenton and Vaughan, 2009; Smetacek and Naqvi, 2008; Zeebe and Archer, 2005]. Downward POC flux clearly influences atmospheric CO2 [Kwon et al., 2009; Parekh et al., 2006], but we do not fully understand the links between iron supply and POC flux and how they are influenced by planktonic community structure.
 Only two FeAXs were conducted in low-silicon (Si) waters (Southern Ocean iron enrichment experiment (SOFeX) North and SOLAS air-sea gas exchange experiment (SAGE)). However, diatoms were abundant and, apparently, not Si limited during SOFeX North [Coale et al., 2004], while downward particle export was not measured during SAGE [Harvey et al., 2010]. Recent work has suggested that Fe supply to low-Si Southern Ocean regions may not enhance POC flux, although appreciable POC flux occurs there naturally [Bowie et al., 2011; Henson et al., 2012; Trull et al., 2001a].
 It is important to note that the shallow export flux of POC, often measured at 100 m, generally does not sequester carbon from the atmosphere for climatically relevant time scales. Long-term sequestration requires POC to sink below the permanent thermocline, and it is this deeper flux that would need enhancing for geoengineering to work [Lampitt et al., 2008a]. POC flux can decrease sharply between these two depths, and the magnitude of this decrease depends on the community structure in the surface and mesopelagic [Buesseler and Boyd, 2009; Boyd and Trull, 2007; Jacquet et al., 2011; Lam and Bishop, 2007]. Enhancing POC export does not necessarily enhance POC sequestration, as heterotrophic activity in the mesopelagic might be stimulated [Lomas et al., 2010]. POC flux past the permanent thermocline must hence be measured during FeAXs, which so far has only been done in Si-rich waters [Smetacek et al., 2012].
 We measured production, export, and deep POC flux using multiple independent methods during LOHAFEX (loha is the Hindi word for iron), a 39 day FeAX in low-Si waters. We first present results from each method individually, examine the reasons for the lack of export, and, finally, discuss the depth horizons where and the processes by which flux was attenuated.
2.1 Site Selection, Fertilization, and Patch Tracking
 LOHAFEX was conducted aboard R/V Polarstern from 26 January to 06 March 2009 in the Atlantic sector of the Southern Ocean. The closed core of a stable cyclonic eddy in the Antarctic Polar Frontal Zone (48°S, 15°W; Figure S1 in the supporting information) was selected for the experiment based in part on real-time Eulerian and Lagrangian altimetry analyses, which were continued throughout the experiment (Okubo-Weiss and Lyapunov exponent techniques [d'Ovidio et al., 2009; Smetacek et al., 2012, supplementary information]).
 Starting on 27 January 2009, 2 t of Fe (10 t of FeSO4 × 7 H2O) were dissolved in SF6-labeled seawater with HCl and spread across 300 km2 in the putative center of the eddy in a spiral pattern around two drifting buoys used to mark the patch center (theoretically yielding 2 nM Fe). Another 2 t of Fe were applied after 18 days, but an instrument fault prevented more SF6 injection. The fertilized patch was studied for 39 days, comparing in-patch measurements to control out-patch observations in unfertilized waters of the eddy.
 The patch drifted within the eddy and was tracked via the drifting buoys, SF6 concentration, the photosynthetic quantum efficiency FV/FM of phytoplankton, and the concentration of chlorophyll a. Because the SF6 outgassed within about 2 weeks, we had to rely mostly on the buoys, chlorophyll a, and FV/FM, which were elevated within the patch until the end of the experiment. A total of five buoys had to be deployed in succession because the first two became detached from the main part of the patch.
2.2 Macronutrients and FV/FM
 NO3− + NO2−, NH4+, PO43−, and Si(OH)4 were measured at sea on a Skalar autoanalyzer using standard procedures.
 Phytoplankton photosynthetic quantum efficiency (FV/FM) was measured continuously from the underway seawater supply using a Chelsea Technology Group fast repetition rate fluorometer and averaged over 2 min intervals. Due to strong daytime fluorescence quenching, only FV/FM measurements from 19:00 to 06:00 local time were used.
2.3 Net Community Production (NCP) and Patch Model
 The seawater O2 concentration is governed by biological and physical factors, but that of the inert gas argon (Ar) is governed only by physical factors. The seawater O2:Ar ratio thus reflects biological O2 supersaturation, ΔO2/Ar [Craig and Hayward, 1987].
 O2:Ar ratio was measured continuously in surface seawater by mass spectrometry, calibrated against outside air every 2–4 h, and averaged every 2 min [Cassar et al., 2009]. ∆O2/Ar was calculated following Craig and Hayward  and the biological O2 concentration, [O2]Bio, according to Cassar et al. . In-patch [O2]Bio was corrected for dilution with unfertilized waters using the dilution rate from a Lagrangian model of the patch and a weighting function based on the ventilation history of the mixed layer (see supporting information).
 NCP was estimated using both a steady state [Reuer et al., 2007] and a non–steady state calculation [Hamme et al., 2012]. The latter accounts for changes in [O2]Bio over time.
 The piston velocity k was calculated from wind speed and water temperature measured in the entire eddy [Wanninkhof, 1992] (see supporting information).
 Routine calibration and instrument problems caused gaps of <1 h to several days, and there are no data for most of the final week. Thus, NCP was only analyzed up until Day 30 (∆O2/Ar at the very end of LOHAFEX was roughly equal to Day 30 values). Because of the gaps, the mean of the observations might not reflect the true mean NCP over the period; hence, loess models were fit to the in-patch data (see supporting information). NCP was then imputed from the models at 2 min frequency across all gaps. Mean NCP was calculated as the overall mean of measured and imputed NCP data up until Day 30 and converted to carbon as C = O2/1.4 [Laws, 1991].
 Postcruise, the extent, trajectory, and dilution of the patch were modeled hourly using a filament-resolving Lagrangian model based on satellite altimetry data, FV/FM measurements, and surface buoy positions [d'Ovidio et al., 2010] (Figures S1a–S1f). Each O2:Ar measurement was classed as in-patch, out-patch, or out-of-eddy based on this model, and the classification adjusted manually by comparison to FV/FM, chlorophyll, salinity, and ship location. The model agreed well with a satellite chlorophyll a image (Figure S2).
2.4 234Th Measurements
 Total 234Th was measured in 4-L samples after manganese co-precipitation with a 230Th yield monitor [Cai et al., 2006; Pike et al., 2005]. Since 234Th deficits did not change over time, steady state downward 234Th flux was calculated [Coale and Bruland, 1985, 1987]. In situ pumps (ISPs) with sequentially mounted 53 and 10 µm Nitex mesh collected particles at 10 stations from 100 m to measure POC:234Th ratio. Particles were washed off the mesh with filtered seawater and sonication and filtered onto precombusted Whatman QMA filters, dried, and β counted. Following background β counting after 234Th decay, filters were acidified with 0.1 M HCl, oven dried, and C and N measured on a Eurovector C/N element analyzer. POC:234Th ratios were also measured in six sediment traps.
 Particulate (≥1 µm, QMA filter) and dissolved 234Th were automatically sampled at 4 h resolution from the underway supply [Rutgers van der Loeff et al., 2004, 2006]. Any automated measurements taken outside of the eddy were omitted, and the rest were designated as in- or out-patch measurements based on the patch model up to Day 31. Later measurements were designated “In” if they were within 10 nautical miles of the buoy. 234Th data are presented as the activity ratio to 238U (238U = 0.0713 × salinity ± 3%), which is 1 at secular equilibrium. The activity ratio for particulate 234Th is to total 238U in the water sample.
 Since In- and Out-patch 234Th measurements did not differ, they were not affected by dilution.
2.5 Sediment Traps
 Neutrally buoyant PELAGRA traps [Lampitt et al., 2008b] were deployed inside and outside of the patch at 200 and 450 m for 5–6 days each (Figure S3). Argo float profiles from the region suggested that 450 m would be close to, but still below, the winter mixed layer. Trap cups contained 2% borate-buffered formaldehyde in 0.2 µm filtered seawater with 0.5% wt/vol NaCl. Additional cups contained polyacrylamide gels [Ebersbach and Trull, 2008].
2.6 Sediment Trap Sample Analyses
 Samples from each trap were pooled, divided with a rotary splitter, and swimmers removed at sea (60–120× magnification). Samples were filtered on precombusted, preweighed Whatman GF/F filters (mass + POC + particulate organic nitrogen (PON)), polycarbonate filters (0.4 µm, particulate inorganic carbon (PIC) and biogenic silica (BSi)), or QMA filters (POC:234Th ratios) and rinsed once with MilliQ. Blanks were prepared by filtering preservative through the different filters. POC:234Th ratio was measured as for ISP samples; the other filters were stored at −20°C. Splits for phytoplankton cell counts and fecal pellet analysis were stored at +4°C.
 Dry weight, POC, PON, PIC, and BSi were measured as in Martin et al. . PIC samples were size fractionated [Bairbakhish et al., 1999], but since the small fraction contained foram fragments, not coccoliths, we present the sum of both fractions.
 Fecal pellets were removed manually from one split onto a precombusted Whatman GF/F filter, acid fumed, oven dried, and POC analyzed at the University of California Davis Stable Isotope Facility.
 Too little material was available to replicate analyses. However, sample processing and analytical errors of 10–15% are likely for POC, PON, PIC, and BSi [Martin, 2011].
 Polyacrylamide gels were photographed on board following Ebersbach and Trull . Aliquots of each sample were settled in sedimentation chambers for 48 h and unicellular organisms counted under inverted light and epifluorescence microscopy. Mean biovolume was measured from 10 to 20 specimens per taxon [Hillebrand et al., 1999] and converted to organic carbon [Menden-Deuer and Lessard, 2000] to calculate unicellular plankton POC flux.
2.7 Underwater Video Profiler (UVP)
 The UVP is a rosette-mounted camera that photographs particles at ~0.2 m vertical resolution on the conductivity-temperature-depth downcast [Picheral et al., 2010]. Custom software calculates equivalent spherical diameter (ESD) and volume of all particles ≥100 µm and classes particles ≥ 630 µm ESD as either aggregates, fecal sticks/pellets, or live zooplankton [Gorsky et al., 2010] (Figure S10). Data were averaged over 5–10 min intervals for each individual profile. We present the median values of all in- and out-patch profiles here.
3.1 Surface Biological and Biogeochemical Response to Fertilization
 Upon fertilization, FV/FM increased from ~0.33 and remained elevated at 0.4–0.5. In-patch chlorophyll a approximately doubled to 1–1.5 mg m−3. Primary productivity from 14C incubations was <80 mmol C m−2 d−1 outside, but rose up to a peak of 130 mmol C m−2 d−1 in the patch (M. Gauns, personal communication, 2010). In-patch NO−3 declined from 20 to 17.5 µmol L−1. Si(OH)4 in the patch was 0.6–1.6 µmol L−1 and did not decrease over time.
 Diatoms were present but small for their species, and flagellates <10 µm contributed >90% of phytoplankton biomass (I. Schulz et al., in preparation, 2013). The coccolithophore Emiliania huxleyi declined after fertilization. Copepod grazing pressure was very high: fecal pellet production rates of Calanus simillimus implied grazing of >30% of net primary productivity (range: 0.7%–240%) (H. González et al., in preparation, 2013). Oithona spp. were particularly abundant: on average, 100,000 m−2 between 0 and 200 m in the patch (range: 35,000–235,000) (M. G. Mazzocchi, personal communication, 2010). Bacterial leucine and thymidine uptake increased somewhat upon fertilization, but cell abundance and species composition did not change [Thiele et al., 2012].
3.2 Movement of the Patch and Trap Trajectories
 The patch rotated inside the eddy core until Day 32 (27 February 2009, Figures S1a–S1f) and was then filamented when the fertilized eddy was entrained by a nearby anticyclone centered around 48°S, 13°W (Figure 1). The patch model (see section 2.3) indicated that the hot spot of the fertilized patch maintained its integrity until the end of the experiment, albeit shrinking due to erosion of its borders by stretching along the frontal jet. The patch model estimated an upper bound of dilution of the hot spot to 50% by Day 20 and to 20% by Day 39, mostly due to diffusion (Figure S4).
 The trap trajectories mirrored the surface circulation indicated by the buoys and shipboard acoustic Doppler current profiler (Figures 2 and S5), implying homogeneous circulation down to 450 m. Although D7#430 and D8#230 surfaced within the patch model, they were in fact recovered from waters outside of the patch. During this time, the patch was squeezed up against the eastern side of the eddy, with a very sharp boundary to unfertilized waters. Traps were designated as in or out (Table 1), yet none is truly unambiguous. Tracking the drift and boundaries of the patch was very challenging, and we cannot be sure that the In traps only collected below the patch. Conversely, time constraints limited how far away the out traps could be deployed, so they might have been influenced by the patch.
Traps are referred to as Deployment Number # Depth. All fluxes are given in millimoles per square meter per day, rounded to two significant figures, except for 234Th, which is in disintegrations per minute per square meter per day; the POC:PON ratios are in mol mol−1. “NM” = not measured.
 Steady state in-patch NCP rose from about 0 to 50 mmol O2 m−2 d−1 by Day 10 and returned to zero by Day 30 (Figures 3 and S6). Out-patch NCP was consistently lower than in-patch NCP, and the out-patch data least likely to have been influenced by the patch (orange points in Figure 3) remained close to zero throughout the experiment. The nominally out-patch data on Days 10–14 were mostly very close to the patch, which were collected while steaming back and forth across the patch boundaries as we tried to map its extent, so their elevated NCP is likely due to mixing with fertilized waters. The autocorrelation function of [O2]Bio indicated a strong diurnal cycle (Figure S7). The non–steady state estimate yielded higher NCP for the first half and lower NCP for the second half of the experiment (Figure S8). These differences cancelled each other out such that the overall mean NCP up to Day 30 was very similar to the steady state NCP (Table 2). We only discuss the steady state estimate below.
Table 2. Mean In-Patch NCP From Days 0–30, Corrected for Dilution With Unfertilized Watersa
Mean NCP Steady State
Mean NCP Non–Steady State
Span of Loess Model
(mmol C m−2 d−1)
(mmol C m−2 d−1)
Loess models spanning different percentages of the data were used to impute data across gaps; NCP values quoted are the mean of the measured and imputed data.
10% of data
15% of data
20% of data
 A loess model spanning 15% of points was judged to fit the data best. Using fewer points yielded unlikely interpolations across gaps (green line in Figure 3), while using a higher percentage of points did not change the estimate significantly (Table 2).
 Our best estimate of mean in-patch NCP is thus 29 mmol O2 m−2 d−1, or 21 mmol C m−2 d−1 (Table 2), while out-patch NCP was −6.2 mmol O2 m−2 d−1, or −4.4 mmol C m−2 d−1. The dilution correction only had a modest effect: uncorrected in-patch NCP was 25 mmol O2 m−2 d−1, or 18 mmol C m−2 d−1. This is because mixed layer ventilation was ~sixfold faster than horizontal dilution, so most of the O2 produced in the patch was lost to the atmosphere, not by dilution.
 Quantifying the uncertainty in NCP is unfortunately rather difficult. Estimating the piston velocity at high wind speeds is a major source of uncertainty [Ho et al., 2006]. However, O2:Ar measurements seem to reflect NCP quite accurately in the Southern Ocean, though they may underestimate NCP by around 20% when productivity is high and Zmix >50 m [Jonsson et al., 2013]. We therefore assume an uncertainty of at least ±20%, since the dilution correction and loess interpolation introduce additional uncertainties.
3.4 Export Based on 234Th
 We calculated downward 234Th flux assuming steady state, since the deficit did not change over time (Figure S9). 234Th-derived export at 100 m was 5.2–7.8 mmol POC m−2 d−1 inside and 4.7–6.4 mmol m−2 d−1 outside of the patch (Figure 4). Most of the deficit was above 75 m, and 234Th excesses relative to 238U rarely exceeded the analytical error (Figure S9). The average 234Th-derived in-patch export at 100 m was 6.3 mmol POC m−2 d−1.
 There was no evidence of a fertilization-induced export event: export flux during Days 0–6 varied as much as during the entire experiment. Since it is bloom collapse that would trigger enhanced particle export [Buesseler et al., 1992, 2001; Cochran et al., 2000; Martin et al., 2011; Smetacek et al., 2012], the range in 234Th-based export estimates over the first 6 days probably reflects spatial variability in the patch, not an increase upon fertilization.
 POC:234Th ratios were 2.9–6.9 µmol dpm−1 in the >53 µm ISP samples and 2.1–8.2 µmol dpm−1 in the three 200 m traps; the overall mean was 4.6 ± 2.0 µmol dpm−1. In- and out-patch ratios fell within the same range, so one ratio was used for all stations. However, >53 µm Nitex-filtered particle samples are not necessarily representative of sinking particles (e.g., fragile particles may disintegrate and pass through the mesh). Hence, we also calculated the combined POC:234Th ratio of all particles >10 µm from the ISP samples, which was 3.1 ± 0.7 µmol dpm−1. Our 234Th-derived POC export may thus be overestimated by about 30%, in which case export would actually have been in the range of 3.5–5.3 mmol POC m−2 d−1.
 The automated surface measurements did not indicate a large export event either, and in- and out-patch surface 234Th depletions were equal (Figure 5). While the total activity ratio of 234Th:238U declined from 0.8 initially to 0.75 by Day 39, ranging ±0.1 at any time, this does not indicate increased 234Th depletion to 100 m depth. However, the particulate 234Th fraction nearly doubled by Day 20. This evident increase in the surface area available for 234Th scavenging could reflect either buildup of new or fragmentation of existing particles.
3.5 Trap Samples
 The traps recorded very low particle flux, and in-patch versus out-patch differences were not evident (Figure 6 and Table 1). POC flux at 450 m was 0.70–1.9 mmol m−2 d−1 inside and 0.24–0.91 mmol m−2 d−1 outside of the patch. However, the lowest out-patch value (trap D4#470) was probably due to a sample processing error, while the highest in-patch value was from a trap that surfaced during adverse weather and could be recovered only 48 h later (trap D6#440); both values are hence suspect. At 200 m, POC flux was 0.46 mmol m−2 d−1 inside the patch (Days 0–2), but 2.4 mmol m−2 d−1 in one sample outside of the patch. POC:PON ratios were high: 8.4–9.6. Intact fecal pellets contributed around 45% of total POC flux (probably underestimated, as trap recovery and sample splitting might disintegrate pellets). The polyacrylamide gels were also dominated by fecal pellets. Unicellular plankton contributed only 0.3%–9% of total POC flux, mostly as dinoflagellates and other flagellates (Table 1). Broken and empty diatom frustules far outnumbered intact diatom cells.
 CaCO3 flux exceeded opal flux by a factor of 2–7. Si:POC ratios were hence low (0.04–0.25), while moderate PIC:POC ratios were found (0.20–0.59).
 Strangely, 234Th flux into the first and third traps was only 60 dpm m−2 d−1, far lower than the >1000 dpm m−2 d−1 predicted at 100 m from 234Th profiles. The other traps collected 510–780 dpm m−2 d−1.
3.6 UVP Particle Profiles
 Particles <250 µm ESD were most abundant in the mixed layer, decreasing between 70 and 120 m. Particles >250 µm ESD peaked at 75 m, decreasing down to 150 m (Figure 7). Total particle volume peaked at 75 m and decreased to about 150 m; while mean particle size and the slope of the particle size spectrum both indicate a higher proportion of large particles below the mixed layer. Moreover, while fecal abundance peaked at 50 m and then decreased sharply to 150 m, unrecognizable detritus (that would include fecal pellets disintegrated by coprorhexy) [Lampitt et al., 1990] increased sharply from 50 to 80 m (Figures 7 and S10). Total copepod abundance peaked at 75–100 m.
 Since particles in the 250–630 µm and >630 µm size classes had very similar depth profiles, the two classes are combined in Figure 7. However, particles <630 µm ESD were more abundant inside than outside the patch (Mann-Whitney U test, W = 196, n = 26, and 9, p = 0.02), and the mean abundance and volume of particles >630 µm ESD decreased with time in 100 m below the mixed layer inside the patch (Spearman's rho = −0.55, n = 26, p = 0.004). No other significant trends with time or in-patch versus out-patch differences were found (for time series of abundance and volume, see Figure S11).
4.1 Effect of Fertilization on Downward Particle Flux
 Neither the 234Th nor the sediment trap data indicate major fertilization-induced export, despite the clear increase in NCP. Moreover, the UVP showed no increase in particles >100 µm upon fertilization. In contrast, evidence is mounting that iron fertilization of Si-replete waters, leading to diatom blooms, can induce severalfold higher export than during LOHAFEX and enhance flux to deep waters (EIFEX [Smetacek et al., 2012], CROZEX [Salter et al., 2007; Morris and Sanders, 2012], SEEDS II [Aramaki et al., 2009], SERIES [Boyd et al., 2005], SOFeX [Buesseler et al., 2004], KEOPS [Blain et al., 2007], and IronEx-II [Bidigare et al., 1999]).
 The LOHAFEX data thus suggest that iron fertilization of Si-limited Southern Ocean waters, which does not stimulate diatom blooms, enhances neither shallow export nor deep POC flux. This is consistent with the view that diatoms are major contributors to new production [Dugdale and Wilkerson, 1998, 2001], given the importance that sinking may have in diatom ecology [Smetacek, 1985; Salter et al., 2012]. It has hence been questioned whether Southern Ocean iron fertilization would work at all to enhance carbon sequestration if it does not do so under Si limitation, because Si is already fully utilized in the Southern Ocean [Trull et al., 2001b]. However, iron fertilization can lower the Si:C ratio of exported material and, thus, can sequester more carbon for the same amount of Si [Smetacek et al., 2012; see also Salter et al., 2012]. Thus, we do not believe that the LOHAFEX results imply that iron fertilization cannot enhance Southern Ocean carbon sequestration.
 However, we cannot readily disentangle the effects on downward POC flux of the lack of diatoms on the one hand and the very high grazing pressure and particle reprocessing by zooplankton on the other. Thus, LOHAFEX provides no conclusive proof that downward POC flux in low-Si sub-Antarctic waters will never be enhanced by iron fertilization, especially since significant export and deep POC flux do occur in low-Si regions [Cardinal et al., 2005; Henson et al., 2012; Honjo et al., 2008; Planchon et al., 2013; Trull et al., 2001a]. Organic carbon did accumulate in the mixed layer (section 4.2), leaving open the possibility that enhanced export occurred after the end of the experiment, although the heavy grazing and particle reprocessing by zooplankton would probably have strongly attenuated any future export event.
 Nevertheless, our results agree with those of SAZ-SENSE, which reported lower export and greater mesopelagic remineralization in naturally iron-replete than in iron-limited low-Si sub-Antarctic waters [Bowie et al., 2011; Ebersbach et al., 2011; Jacquet et al., 2011]. Only a modest response, mostly by nondiatom phytoplankton <20 µm, was found upon iron fertilization of sub-Antarctic low-Si waters during SAGE, suggesting that export was probably not greatly enhanced [Harvey et al., 2010; Peloquin et al., 2010]. In contrast, POC export at the low-Si sub-Antarctic SOFeX North site was enhanced by iron fertilization, with NO3− depletion similar to LOHAFEX [Bishop et al., 2004; Coale et al., 2004]. However, Si(OH)4 in SOFeX North was above limiting concentrations and, apparently, replenished in the elongated patch by admixture of surrounding water. Weakly silicified diatoms contributed 44% to total phytoplankton POC and aggregated eventually [Coale et al., 2004]. Thus, while SOFeX North is nominally considered a “low-Si” experiment, diatoms were not initially Si limited and did bloom, in strong contrast to LOHAFEX.
4.2 Comparison Between NCP, 234Th, and Sediment Traps
 Comparing these three methods is fraught with complications, since export may lag production, the methods integrate over different time scales and depths, and each suffers from biases and uncertainties [Lampitt et al., 2008b; Le Moigne et al., 2013; Morris et al., 2007; Savoye et al., 2008]. However, the long duration and Lagrangian nature of LOHAFEX mitigate some of these problems, and while significant uncertainties are associated with each of our estimates, we do not believe that any of the methods is grossly biased. Figure 8 summarizes our main conclusions.
 NCP was 21 mmol POC m−2 d−1, exceeding the 100 m export flux by ~15 mmol m−2 d−1, implying organic carbon accumulation in the mixed layer and/or flux attenuation between the mixed layer depth (MLD) and 100 m. Direct measurements do suggest accumulation in the mixed layer of ≤6 µmol L−1 of total organic carbon in the patch (S. W. A. Naqvi et al., in preparation, 2013), accounting for ≤13 mmol m−2 d−1 of the NCP. This would allow for export out of the mixed layer of at least 8 mmol POC m−2 d−1, of which around 6 mmol m−2 d−1 sank below 100 m (as diagnosed from 234Th). This implies that POC flux was attenuated by around 2 mmol m−2 d−1 between the mixed layer and 100 m. Thus, a little more than half of the in-patch NCP appears to have accumulated in the mixed layer, while the remainder was exported below the mixed layer as sinking POC flux.
 The POC flux diagnosed from 234Th exceeded trap fluxes threefold to sixfold. Since the flux of 234Th itself was just 2–3 times lower in the traps than that diagnosed from the profiles, the discrepancy cannot be attributed purely to biased trap collection. The 234Th and trap data thus indicate a strong reduction in particle flux from 100 to 200–450 m.
 Between the base of the mixed layer and the sediment traps at 200–450 m, POC flux was probably attenuated about eightfold, or about sixfold between 100 and 200–450 m. These estimates must be treated with caution, since the export estimates at each depth carry significant uncertainty. However, such intense attenuation contrasts with the higher transfer efficiencies of flux to depth that have been reported upon collapse of diatom blooms [Buesseler and Boyd, 2009; Martin et al., 2011; Smetacek et al., 2012]. Interestingly, subsurface 234Th excesses indicative of remineralization [Maiti et al., 2010; Savoye et al., 2004] were not consistently found, although excesses are often confined to narrow depth horizons. They might hence have been missed by our 50 m vertical resolution in the mesopelagic.
 The UVP data are also consistent with strong flux attenuation: particle stocks declined with depth below the MLD, and there was a shift from intact fecal pellets to unrecognizable detritus. This shift was most pronounced at the depth of highest copepod abundance, implying coprorhexy [Lampitt et al., 1990] and, generally, particle reprocessing by zooplankton. The high abundance of Oithona spp. during LOHAFEX also suggests substantial flux reprocessing: Oithona spp. are reported to be coprophagous and, hence, likely to attenuate POC flux [González and Smetacek, 1994]. However, intact fecal material contributed ~45% to the sediment trap catches, underscoring the importance of unreprocessed fecal pellets in downward POC flux.
 This contrasts with the enhanced mesopelagic particle stocks seen during the Kerguelen Ocean and Plateau Compared Study (KEOPS) [Jouandet et al., 2011]. Overall, the UVP revealed that the most intense particle transformations took place between the base of the mixed layer and around 150 m (Figures 7 and S8), and flux attenuation probably took place throughout this depth range.
 Mesopelagic communities of high- and low-Si regions may actually respond differently to iron fertilization: mesopelagic remineralization as estimated from excess barium was a relatively low proportion of export flux in the high-Si iron fertilized areas of EIFEX and KEOPS [Jacquet et al., 2008a, 2008b]. In contrast, at the iron-replete low-Si sub-Antarctic site in SAZ-SENSE a greater proportion of export flux was remineralized than at either of the iron-limited sites [Jacquet et al., 2011]. Moreover, export from SOFeX North was initially reduced owing to a response by mesopelagic grazers, though an export event did occur later [Bishop et al., 2004; Lam and Bishop, 2007]. We observed no drastic changes over time, but the upper mesopelagic community appeared to attenuate particle flux heavily.
 Downward particle flux out of the fertilized patch and through the mesopelagic was tracked successfully for 39 days. Net community production, but not 100 m export flux, increased relative to unfertilized waters; mixed layer organic carbon accumulation and flux attenuation above 100 m can account for this difference. Particle flux appeared to decrease strongly between 100 and 200–450 m. Our results add further evidence to support the idea that Fe fertilization does not necessarily stimulate POC export and sequestration under Si limitation in the Southern Ocean. Zooplankton community composition and activity under the mixed layer may strongly regulate the export by reprocessing sinking particles and altering the particle size distribution.
 We thank the captain and crew of R/V Polarstern. Kevin Saw ensured the success of the PELAGRA deployments, Christine Klaas gave advice on the dilution correction, and two anonymous reviewers provided constructive criticism that significantly improved the manuscript. The altimeter products were produced by Ssalto/Duacs and distributed by AVISO with support from CNES. N.C. was partly supported by an Alfred P. Sloan Fellowship. This work formed part of the PhD research of P.M.